Designing Artificial Metalloenzymes by Tuning of the Environment beyond the Primary Coordination Sphere

Abstract

Metalloenzymes catalyze a variety of reactions using a limited number of natural amino acids and metallocofactors. Therefore, the environment beyond the primary coordination sphere must play an important role in both conferring and tuning their phenomenal catalytic properties, enabling active sites with otherwise similar primary coordination environments to perform a diverse array of biological functions. However, since the interactions beyond the primary coordination sphere are numerous and weak, it has been difficult to pinpoint structural features responsible for the tuning of activities of native enzymes. Designing artificial metalloenzymes (ArMs) offers an excellent basis to elucidate the roles of these interactions and to further develop practical biological catalysts. In this review, we highlight how the secondary coordination spheres of ArMs influence metal binding and catalysis, with particular focus on the use of native protein scaffolds as templates for the design of ArMs by either rational design aided by computational modeling, directed evolution, or a combination of both approaches. In describing successes in designing heme, nonheme Fe, and Cu metalloenzymes, heteronuclear metalloenzymes containing heme, and those ArMs containing other metal centers (including those with non-native metal ions and metallocofactors), we have summarized insights gained on how careful controls of the interactions in the secondary coordination sphere, including hydrophobic and hydrogen bonding interactions, allow the generation and tuning of these respective systems to approach, rival, and, in a few cases, exceed those of native enzymes. We have also provided an outlook on the remaining challenges and future directions gain a deeper understanding of the secondary coordination sphere, and in turn guide the design of a broader and more efficient variety of ArMs.

Graphical Abstract

1. INTRODUCTION

Metalloproteins are ubiquitous in nature; a bioinformatic search of three-dimensional protein structures deposited in the Protein Data Bank in 2008 found ~40% to bind metal ions, with estimates ranging between 36 and 59% in different classes.¹ A metallomics study in 2010 indicated that microbial metalloproteomes are largely uncharacterized,² putting the estimate of metalloproteins among all proteins over 50%. Many of these metalloproteins function as metalloenzymes, performing the essential processes to support life as it is known, ranging from metabolism^3-14 and DNA synthesis^15-18 to the generation of atmospheric O₂^19,20 and nitrogen fixation.^21-24 Even more so, the stunning range of chemistry performed in these native systems is done so not only within the limited conditions conducive to life but also using a comparatively narrow range of biologically available ligands and metallocofactors. It is little wonder then that metalloenzymes have captivated chemists and biochemists alike for decades, and tremendous efforts have been undertaken across these fields to understand the intimate details of metalloenzyme functions.

The biologically available ligands for the formation of the primary coordination sphere (PCS) in metalloenzymes include the side chains of nucleophilic amino acid residues such as His, Asp, Glu, Arg, Met, Ser, Tyr, and Cys, the amine and carboxylate protein termini, the carbonyl and amide moieties of the protein backbone, small molecules such as CO, CN⁻, H₂O, O₂, S²⁻, and ⁻OH, and biosynthesized organic cofactors such as protoporphyrins, corrins, pterins, azadithiolate, and homocitrate.²⁵ These PCSs play critical roles in defining the available chemistry of an active site. A primary example is the heme b center, which consists of an iron center equatorially coordinated by the four nitrogen atoms of the protoporphyrin IX macrocycle and axially by one or two amino acids above and/or below the protoporphyrin ring. While undeniably important, abundant evidence from the study of metalloenzymes has shown that the same PCS environment, such as heme b, can display a variety of different properties from electron transfer to reversible O₂ binding, dioxygen reduction, and small molecule functionalization.²⁶ As a result, these heme proteins perform a wide range of biological processes, including O₂ transport (globins),^27-29 cell signaling,^30-32 electron transfer,^33-35 and enzymatic activities such as catalases,^36-40 oxidases,^41,42 oxygenases,^43-45 peroxidases,^46-48 nitric oxide reductases,^49-51 and sulfite reductases.^52-54 Therefore, features of the protein environment beyond the PCS are essential in not only fine-tuning the activities of metalloenzymes but also conferring new activities in otherwise similar systems.

Unlike the PCS itself, the environment surrounding the PCS consists of many weak and noncoordinating interactions such as hydrogen bonding (H-bonding), electrostatic, and hydrophobic interactions. Some may contribute little, if any, to protein function, while others play major roles that enable many metalloenzymes to catalyze reactions that are currently unparalleled by synthetic catalysts in terms of activity and/or reaction efficiency. Therefore, elucidating the structural features in these secondary coordinate spheres (SCSs) that are responsible for activity not only advances our knowledge about biology but can also result in novel synthetic catalysts or artificial metalloenzymes (ArMs) with new activity or reactivity for many industrial applications, such as producing energy from sustainable sources and synthesizing pharmaceutical drugs.

The study of native metalloenzymes is undeniably essential in providing a detailed understanding of their function. Despite major progress in many facets, however, it has been difficult to comprehensively elucidate how these enzymes catalyze so many reactions highly efficiently using structural features beyond the PCS. Such difficulties arise in large part because the PCSs of many native enzymes reside in large and complicated proteins, some of which are membrane proteins, that are not (yet) compatible with heterologous expression. As a result, it is very difficult to carry out site-directed mutagenesis (SDM) to probe the roles of SCS residues and to purify them to homogeneity in high enough yields for traditional studies. In addition, many native metalloenzymes contain multiple metallocofactors that can produce overlapping spectroscopic signals which make it challenging to focus on the PCS of interest.^12,55,56 Furthermore, many similar PCSs reside in very different protein scaffolds containing numerous different residues, making it difficult to identify the key SCS residues responsible for the different reactivity. Designing artificial metalloenzymes (ArMs) using small robust protein scaffolds that can be readily constructed, expressed, and purified in simple hosts such as E. coli can overcome the above limitations, because these ArMs contain only the PCS of interest and free of interferences from other cofactors. By placing the same PCS in the same protein scaffold and systematically introducing SCS residues from different enzymes with varying reactivity, we can provide a holistic view of the key structural features responsible for the differences in activity. Therefore, designing ArMs that reproduce the catalytic capacities of native enzymes not only provides the ultimate basis to test what we know about native catalytic mechanisms but also reveals new insights that are not readily obtainable from studying native enzymes, much in spirit with a quote from the late Prof. Richard Feynman, “What I cannot create, I do not understand.”⁵⁷ Given their small, robust scaffolds, ArMs provide a new, divergent opportunity to combine the benefits of both chemical catalysis and biocatalysis, allowing the power of SCS interactions to be harnessed to enhance catalytic selectivity in new-to-nature reactions.

ArMs can be generally designed using one of the following systems: (1) modifying an existing meta-binding site in a native enzymatic scaffold to promote non-native reactivity, (2) introducing a new metal-binding site to a native protein scaffold without such a site, and (3) designing a metal-binding site in a de novo designed protein scaffold. To achieve this goal, the most common strategies employed include rational design and directed evolution. Rational design involves the purposeful mutation of select residues or regions of a protein scaffold using SDM to incorporate a metal-binding site and/or modulate activity based on careful deliberation over available structural information and often from comparisons with related systems. Such design is often aided by comparing key residues between the native enzymes and ArMs and using the experience and knowledge gained from studying native enzymes, or by computations, to predict and then experimentally verify the roles of these residues. Although immensely successful, this approach is still limited by human intuition and/or the ability of computational programs to capture the structural and functional properties of metalloenzymes, and remains challenging because of highly variable PCSs and weak SCS interactions. Alternatively, directed evolution based on a random library to screen or select for desired activity followed by detailed structural and mechanistic studies of the “winners” can provide an understanding of the roles of residues in not only the SCS but also those much further away in conferring and fine-tuning the activities.^58-60 More recently, computational methods using programs that employ large structural databases and machine learning, such as Rosetta^61,62 and AlphaFold,^63,64 have emerged as powerful tools for the exploratory design of novel ArMs in silico, allowing a much broader scope of protein scaffolds and variants to be surveyed for desired properties, which can then be tested in vitro.

In the present review, we focus on how rational design and directed evolution have provided insight into the roles of the SCS in tuning the functionality of ArMs, with particular emphasis toward systems employing natural protein scaffolds and exhibiting catalytic activity. A more comprehensive perspective on catalysis and electron transfer by de novo metalloenzymes can be found in the review of Pecoraro and co-workers in this issue, and several recent reviews provide detailed coverage of the design and breadth of applications of these scaffolds.^65-68 For extensive reviews of the design and catalytic capabilities of ArMs, we refer the reader to many outstanding reviews in this area.^69-77 Fe- and Cu-based ArMs constitute a major portion of designed metalloenzymes, and therefore, this review has been broadly organized into sections addressing the fields of heme Fe, nonheme Fe, and Cu followed by multinuclear ArMs, and finally systems containing alternative metal cofactors including V, Cr, Mn, Co, Ni, Zn, Ru, Rh, Pd, and Ir.

2. CATALYSIS BEYOND THE PRIMARY COORDINATION SPHERE BY HEME PROTEINS WITH NON-NATIVE FUNCTIONS

Heme proteins are among the most, if not the most, studied metalloproteins, and their functions range from electron transfer to O₂ carrier, catalases, peroxidases, oxidase, and oxygenase. They perform these functions using different types of heme, such as heme a, b, c, d, and f. The different types of the heme have a similar core consisting of Fe and a porphyrin ring. Even the same type of heme (e.g., heme b) can catalyze very different and diverse reactions. Therefore, structural features beyond such PCSs must play important roles in defining and tuning the functions. Designing ArMs to mimic these native heme proteins using much smaller and robust heme proteins, such as myoglobin (Mb), cytochrome c (Cc), and cytochrome c peroxidase (CcP), allows not only testing of what we know about similarities and differences among these proteins but also revelation of new insights.

2.1. Rational Design Using Myoglobin as a Scaffold

Myoglobin (Mb) is a molecular oxygen carrier that natively serves to reversibly bind O₂ using a heme b center.^78,79 The ready heterologous expression of Mb variants in E. coli. has made this enzyme a popular scaffold for rational designs of heme enzymes with a range of activity far beyond its native function.^80-82 Furthermore, the small size, high solubility in water, and ready crystallizability of Mb have also made it accessible to an array of biochemical and biophysical methods to investigate the role of SCSs in catalysis.^83-88

Myoglobin belongs to the globin superfamily of proteins, consisting of eight α-helices connected by a short loop region (Figure 1a).⁸⁹ The endogenous active site of Mb consists of a heme b center (Figure 1b) in which Fe is proximally coordinated by histidine through the N_ε position. This PCS is ubiquitous among globins and is presented in some heme peroxidases as well.^90-92 Therefore, the differences in activity observed between Mb, other globins, and peroxidases must arise from differences in structure beyond the PCS.

Figure 1.

(a) Crystal structure of WT-sperm whale Mb (swMb) (PDB ID: 5YCE).⁸⁹ (b) Chemical structure of heme b. (c) Crystal structure of the active site of cytochrome c peroxidase (PDB ID: 1ZBY) with the distal His52 and Arg48 residues shown in yellow.⁹⁰ (d) Crystal structure of the active site of the F43H/H64L Mb mutant (PDB ID: 1OFK)¹⁰² with the two mutated residues shown in yellow. The distances between the N_ε atom of the distal His residue and the heme iron in each crystal structure are labeled.

2.1.1. From Myoglobin to Peroxidase.

In addition to its well-known function as an O₂ carrier, Mb is also capable of performing oxidation reactions in the presence of hydrogen peroxide (H₂O₂).^93-95 However, the peroxidase activity of wild-type (WT) Mb is much lower than that of native peroxidases (2.7 M⁻¹s⁻¹ vs 130 M⁻¹s⁻¹).⁹⁶ Native peroxidases are a large group of enzymes that catalyze several critical redox reactions, including alkene oxidation, sulfoxidation, and dehalogenation. All peroxidases share similar PCS environments to that of ferric Mb, with the fifth coordination site of Fe occupied by a proximal histidine side chain and the sixth coordination position occupied by a water in the resting state. However, peroxidases possess a highly conserved Arg residue in the distal pocket, which is absent in Mb (Figure 1c).^97,98 Interestingly, both Mb and peroxidases contain a His residue in the distal pockets (His64 in Mb and His52 in CcP). Comparing the X-ray structures of CcP and Mb reveals that the distal His of Mb is closer to the heme iron than in CcP, which may be responsible for the differences in the interaction of these enzymes with the H₂O₂ substrate.^89,97 By relocating the distal His residue in Mb through a F43H/H64L double mutant, the Watanabe group was able to produce a Mb with catalytic rates for sulfoxidation and epoxidation similar to those of native CcP, without introducing Arg.⁹⁹ The distal His64 in WT-Mb, which is 4.3 Å from the heme iron, was determined to be a critical residue for the lifetime of the active intermediate (compound I) and, therefore, activity with H₂O₂. The presence of His64 in WT-Mb presumably results in rapid electron transfer from compound I to His64. Replacement of His64 with a non-oxidizable Leu residue may prevent generation of the protein radical. As the His64 in native Mb is too close to the heme Fe to support the heterolytic cleavage of H₂O₂, the relocation of the distal His by F43H mutation led to a proper distancing between the N_ε atom of His and the heme Fe, which perfectly resembled the corresponding distances in native peroxidases. The appropriate positioning of the distal His in F43H/H64L allows the distal His to function as both a general base to facilitate the H₂O₂ binding through the deprotonation step and a general acid to aid the heterolytic cleavage of the O─O bond in H₂O₂ via H-bonding. As another corroboration of the importance of the proper positioning of the distal His in Mb for peroxidase functions, the L29H/H64V mutant exhibited a ~3–6 slower reaction rate with H₂O₂ relative to the native Mb as the imidazole of His29 is too far away to interact with the H₂O₂ substrate.¹⁰⁰ As another example of mimicking the structural features of the native peroxidase, Wu, et al. found that introduction of two His residues near the active site through the L29H/F43H double mutant produces a SCS resembling the His-Arg pair that is frequently observed in native peroxidases.¹⁰¹ An additional H64A mutation further opened a channel to the heme active center, positively influencing the peroxidase activity.

In addition to enhancing the sulfoxidation and epoxidation reactions in Mb, arene oxidation activity has also been introduced into Mb by mimicking the active site of cytochrome P450 (P450). The differences in reactivity between WT-Mb and P450 may arise from variations in the SCS; for example, d-camphor is capable of tightly binding to the hydrophobic heme pocket of P450 through a hydrogen bonding interaction with the hydroxyl group of Tyr96,¹⁰³ while the active site of WT-Mb does not afford specific interactions for accommodating organic substrates. Using crystallography, it was found that the distance between carbon atoms C6 and C7 of the electron-rich Trp43 residue and the heme iron in the F43W/H64L Mb mutant was similar to the distance between C5 of d-camphor and the iron in P450_cam (Figure 2). The development of the double mutant F43W/H65L Mb resulted in a six-electron oxidative modification of Trp43 in the presence of strong sacrificial oxidant, m-chloroperbenzoic acid (mCPBA), representing the first example of oxidation of an aromatic ring by Mb.¹⁰⁴

Figure 2.

(a) Crystal structure of P450_cam (left, PDB ID: 2CPP)¹⁰³ and structural model of F43W/H64L Mb (right). (b) Chemical structures of tryptophan and the six-electron modified tryptophan products.

As conserved residues near the active site of Mb, both His64 and Val68 have been shown to influence the affinity of ligand binding (e.g., O₂ and CO).^105,107 To investigate the role of Val68 in influencing the rate and enantioselectivity of oxidative reactions, Yang et al. engineered a series of H64D/V68X Mb mutants (X = Ala, Ser, Val, Ile, Leu, and Phe) (Figure 3).¹⁰⁶ Along this series, the V68L mutation was observed to activate H₂O₂ most efficiently, while decreased peroxidase activity was found for both V68I and V68A variants, highlighting that subtle changes in the interactions between residue 68 and the substrate can have profound effects on catalysis. In this case, the restricted rotation of the Ile side chain was found to hinder the association of H₂O₂ at the heme iron center. While removing this steric constraint via a V68A mutation should hypothetically allow association, structural data indicate that this variant allows for the introduction of a distal water molecule which then blocks substrate entrance, demonstrating the balance required for optimization of catalytic activity.

Figure 3.

Crystal structures of the (a) V68L (rose brown, PDB ID: 1MLS),¹⁰⁵ (b) H64D/V68A (orange, PDB ID: 1LUE),¹⁰⁶ and (c) V68I (magenta, PDB ID: 1MLO)¹⁰⁵ Mb variants. Different residues at position 68 are shown in cyan. Water molecules are denoted as red spheres. The hydrogen bonds are represented by black dashed lines.

As another route to engineering peroxidase activity, the introduction of stabilizing forces through disulfide bond formation offers a great opportunity to increase protein stability while tuning enzymatic function. For instance, disulfide bonds have been observed in neuroglobin (Ngb)¹⁰⁸ and cytoglobin (Cgb)¹⁰⁹ and have been shown to regulate ligand binding and migration. Similar disulfide bonds have been incorporated into Mb at different positions through variants containing F46C/M55C and V21C/V66C/F46S mutations (Figure 4).^110,111 The presence of the disulfide bond in F46C/M55C Mb was found to increase both protein stability and peroxidase function.¹¹⁰ Meanwhile, the V21C/V66C/F46S Mb variant demonstrated comparable activity to F46S Mb, but with a ~30% reduction in K_m, indicating that the presence of the disulfide bond stabilizes a favorable conformation for substrate binding.¹¹¹

Figure 4.

Crystal structures of the V21C/V66C/F46S Mb variant (PDB ID: 5ZEO).¹¹¹ The disulfide bond between the Cys21 and Cys66 residues is highlighted in the red dashed circle. The hydrogen bonds involving the distal His64 and the Ser46 residues are shown by black dashed lines.

2.1.2. From Myoglobin to Dehaloperoxidase.

Dehaloperoxidase (DHP) is a heme protein that catalyzes the dehalogenation of halophenols to nontoxic quinone products.^112,113 Halophenols are involved in the synthesis of pesticides and even used as pesticides themselves, which may lead to severe damage to the mammalian liver and immune system.⁸² They can also be readily oxidized to produce carcinogenic radicals capable of modifying DNA bases. Therefore, native peroxidases such as DHP may have important environmental and economic implications for degradation of industrial dyes and halophenols. Despite their potential applications, the use of bioremediation is often hindered by the limited catalytic efficiency of native enzymes.⁸² Rationally designed artificial enzymes offer an alternative route to generating functional biocatalysts. Mb-based ArM models for DHP have emerged as efficient catalysts for biodegradation. Despite the low sequence homology between DHP and Mb, the two proteins share a remarkably similar protein fold, with the RMSD between the positions of C atoms being 1.8 Å (Figure 5).⁹²

Figure 5.

(a) Overlay of the crystal structures of native DHP (red, PDB ID: 1EW6)⁹² and WT-Mb (blue, PDB ID: 5YCE).⁸⁹ (b) Active sites of native DHP and WT-Mb. The distances between the N_ε atoms of the distal His residues and the heme irons in DHP, Mb, and the G65T and G65I Mb variants are summarized in the table on the right.

The mechanism of dehalogenation by dehaloperoxidases has also been analyzed using both “peroxidase-like” Mb mutants and “Mb-like” DHP.¹¹⁴ To mimic the distal His-to-Fe distance in DHP, the authors successfully generated G65T and G65I Mb variants in an effort to position the distal His64 farther from the heme iron (Figure 5). As a result, a more than 5-fold increase in k_cat for the oxidative dichlorination reaction was observed relative to WT-Mb, highlighting the subtle changes of nonactive site residues can induce changes in the PCS through both the restrictions imposed by the position of the peptide backbone as well as through the hydrogen bond network.

Studies have shown that a novel tyrosine-heme covalent C─O bond cross-link can be formed in the F43Y Mb mutant (Figure 6a),¹¹⁵ which has provided a platform for engineering Mb variants as biosynthetic models of dehaloperoxidases. The F43Y Mb variant itself displays enhanced dehaloperoxidase activity 115-fold higher than WT-Mb and 9-fold higher than native DHP in the presence of 2,4,6-trichlorophenol (TCP) substrate.¹¹⁶ The presence of multiple Tyr and Trp residues in native dye-decolorizing peroxidases (DyPs) has also inspired the engineering of Mb variants with additional Tyr or Trp mutations, building on the F43Y Mb platform. Interestingly, the double mutant F43Y/F138W exhibits ~4.3-fold greater peroxidase activity than native DyP from Vibrio cholerae.¹¹⁷ This enhanced activity presumably arises from the more rapid generation of the oxidizing species and protein radicals.

Figure 6.

(a) Crystal structure of F43Y Mb (PDB ID: 4QAU)¹¹⁵ showing the Tyr-heme covalent C─O bond and two distal water molecules forming H-bonding interactions (dotted lines).¹¹⁵ (b) Overlay of the crystal structures of ferric F43Y/H64D Mb (blue, PDB ID: 5ZZF)¹¹⁸ and the heme active site of native chloroperoxidase (orange, PDB ID: 1CPO),¹¹⁹ showing the H-bonding network. (c) Crystal structure of ferric F43Y/H64D Mb in complex with TCP (PDB ID: 5ZZG),¹¹⁸ showing the conformation of Tyr43 and TCP and the H-bonding interactions in the heme center. The distance between the Cl4 atom and the heme Fe (3.91 Å) is indicated. Reproduced with permission from ref 118. Copyright 2018 American Chemical Society.

In 2018 Yin et al. combined structural features from both chloroperoxidase (distal Glu) and DHP (distal Tyr) by generating a F43Y/H64D Mb mutant (Figure 6b), which resulted in a 1000-fold increase in catalytic efficiency relative to native DHP from A. ornata.¹¹⁸ The crystal structure (Figure 6c) of this mutant in complex with TCP revealed that Asp64 adapts an open conformation when substrate is bound, highlighting the crucial role of this residue in modulating the substrate binding.

Beyond dehalogenation, the F43Y Mb platform has been employed in the biosynthesis of indigo from indole. While it has been shown that replacement of the distal His64 in Mb with Asp would favor the H₂O₂-dependent oxidation reaction,¹²⁰ the double mutant F43Y/H64D exhibits the highest efficiency toward the oxidation of indole to indigo reported to-date.¹²¹ More recent work employing a triple mutant based on the F43Y platform has achieved comparable catalytic efficiency to the most efficient native horseradish peroxidase (HRP), where the combination of T67R and F138W mutations is used to mimic the His-Arg pair and conserved Trp residues of native peroxidases.¹²² The scope of engineered Mb-based catalysts has also been expanded to biodegradation applications; for example, the F43H/H64A Mb mutant successfully catalyzed the biodegradation of malachite green (MG) with even higher efficiency than natural enzymes such as dye-decolorizing peroxidase and laccase.¹²³ Molecular modeling indicates that these active site mutations favor the binding of MG in the heme distal pocket.

2.1.3. From Myoglobin to Carbene Transferase.

Beyond peroxidase and dehaloperoxidase activity, Mb has also been redesigned to promote carbene transfer reactions.^124-128 Carbenes are formally neutral divalent carbon species whose activity can be attributed to an incomplete octet electronic configuration.¹²⁹ Recent protein engineering efforts have revealed that carbene transfer can be catalyzed in heme proteins, including Mb, via an iron porphyrin carbene intermediate, providing a whole new pathway for the biosynthesis of important molecules (Scheme 1).¹³⁰ In 2015, Bordeaux et al. discovered that WT-swMb exhibits cyclopropanation activity without asymmetric induction.¹²⁴ Inspired by these results, further Mb-based catalysts have since been developed with both higher activity and selectivity toward cyclopropanation, as well as expanded the substrate scope, including olefin cyclopropanation,¹²⁴ carbene NH insertion,¹³¹ and carbene SH insertion reactions.¹²⁵

Scheme 1. Carbene Transfer Reactions Catalyzed by Evolved Hemoproteins^a.

^a Reprinted with permission from ref 130. Copyright 2021 American Chemical Society.

One of the first efforts to improve the cyclopropanation activity of Mb involved varying the steric bulk properties at the active site residues Phe43, His64, and Val68 of Mb, resulting in the development of the H64V/V68AMb variant, which exhibits high activity and selectivity toward the cyclopropanation of styrene with ethyl diazoacetate (EDA) (Figure 7).¹²⁴ Based on the available crystal structures of related Mb- and Fe-(porphyrin)/carbene complexes, it has been proposed that the H64V mutation aids in promoting substrate access to heme, while the V68A mutation expands the size of the distal cavity and, thus, has a dramatic effect on enantioselectivity.

Figure 7.

(a) Top view of the crystal structure of WT-swMb (PDB ID: 5YCE) with the key residues in the distal heme pocket shown in magenta.⁸⁹ (b) Pocket cavity space (mesh) above the heme cofactor in H64V/V68A Mb (left) and WT-swMb (right). The pocket volumes were calculated to correspond to 243 and 125 ų, respectively. Reproduced with permission from ref 135. Copyright 2019 American Chemical Society. (c) Structural model of the carbene-bound transition state (TSO/ins) of H64G/V68A Mb. The benzofuran substrate is shown in cyan. Reproduced with permission from ref 133. Copyright 2019 John Wiley & Sons.

The reaction scope of the H64V/V68A Mb mutant has also been explored beyond cyclopropanation, demonstrating activity toward carbene N-H insertion with arylamine substrates.¹³¹ For the H64V/V68A Mb mutant, the yield for N-methylaniline was relatively low. An improved activity could be achieved through the L29A, H64V, and L29A/H64V variants. The beneficial effects of the L29A mutation may arise from a decrease in steric bulk directly above the heme iron site, allowing the amine substrate to be accommodated. In addition, H64V/V68A Mb has also exhibited promising reactivity toward C-H functionalization of indole at the C3 position.¹³² This enhanced activity (relative to WT-Mb and H64V Mb) is mainly attributed to the V68A mutation, which enlarges the active site near the heme group to facilitate attack of the heme-bound carbene by indole. The presence of an N1 or C2 substituent on indole resulted in significantly diminished substrate conversions, a limitation that was partially overcome in L29F/H64V Mb. This increase in activity was hypothesized to arise from Phe29 helping to orient the indole substrate into a favorable position via π-stacking interactions.

Along a similar vein, Tyagi et al. identified the L29A/H64V mutant as a promising catalyst for carbene S-H insertion by examining a panel of Mb variants with mutations at residues lining the active site, specifically 29, 43, 64, 68, and 107.¹²⁵ The greatest enhancement for a single mutation was observed in the L29A Mb variant, which may arise from its role in facilitating the nucleophilic attack of thiol or amine to the heme-bound carbene intermediate by expanding the distal cavity above the heme. Further examination of mutations at these positions has also revealed the F43V/V68F Mb variant, which catalyzes the olefination of aldehydes with excellent efficiency and E diastereoselectivity.¹²⁸

In 2019, Vargas et al. expanded the substrate scope of Mb to accommodate the larger benzofuran substrate and, in doing so, established the important role of residue 64 in H64X/V68A variants in controlling access of the olefin substrate.¹³³ By decreasing side chain steric bulk, progressively improved catalytic activity could be achieved along the series His≪Val < Ala < Gly. This “gating” effect has also been manifested in more recent work, where a decreasing size of the residue 64 side chain enhanced the cyclopropanation activity in the presence of a diazoketone carbene donor by removing the steric hindrance.¹³⁴

Taken together, these experimental results illustrate how the catalytic activity and selectivity toward different amine substrates can be modulated by modification of the Mb active site. Although the engineered Mb variants represent a superb platform for carbene transfer catalysis, the precise structural features that govern protein-mediated stereocontrols are yet to be fully elucidated. This problem was not resolved until a crystallographic and computational (DFT) analysis of the H64V/V68A mutant (Figure 7) revealed that stereocontrol of the cyclopropanation reaction by active site mutagenesis is achieved by favoring a specific conformation of the heme-bound carbene and control geometry of the attack of the olefin through steric obstruction, van der Waals forces, and π−π interactions.¹³⁵

2.1.4. From Mb to Nitrite Reductase (NiR).

To study the Tyr-Cys cross-linking phenomenon, which is ubiquitous in a variety of metalloenzymes, an unnatural amino acid 2-amino-3-(4-hydroxy-3-(methylthio)phenyl)propanoic acid (MtTyr) was incorporated into a functional model of T. nitratireducens cytochrome c nitrite reductase (TvNiR) engineered into swMb by Zhou, et al. The enzymatic activity was determined based on the reduction of hydroxylamine to ammonium by models both containing and lacking MtTyr (MtTyrMb and TyrMb, respectively) as mimics for the presence/absence of the Tyr-Cys cross-link. A 4-fold increase in k_cat was found for the MtTyrMb variant, supporting the hypothesis that thioester modification on Tyr helps to enhance NiR activity.¹³⁶

2.2. Rational Design Using Cytochrome c Peroxidase as a Scaffold

Peroxidases are a large family of enzymes that function to catalyze the reduction of H₂O₂ in the presence of various substrates. Notably, cytochrome c peroxidase (CcP) and ascorbate peroxidase (APX), two iconic members of this family, do not share any common reducing substrates despite having numerous structural similarities. Since the structures of CcP and APX mainly differ in the ascorbate binding site, Meharenna et al. engineered an ascorbate-binding site into CcP to repurpose the function of CcP toward the non-native substrate (Figure 8a).¹³⁷ This hybrid protein was generated by replacing native residues 30–42 in CcP with the ascorbate binding site residues 27–32 in APX. In addition, a conserved Arg residue in APX, which is thought to facilitate ascorbate binding via H-bonding, was anchored at position 184 in CcP. The resulting hybridized mutant was found to catalyze the peroxidation of ascorbate at a rate of about 12 min⁻¹.

Figure 8.

(a) Overlay of the crystal structures of WT-CcP (red) and APX (green). Reproduced with permission from ref 137. Copyright 2008 American Chemical Society. The ascorbate binding site and the corresponding loop in CcP are highlighted in the black dashed square. (b) Crystal structure of yeast CcP (PDB ID: 2CCP).¹³⁹ Three residues that are mutated to Ala are highlighted in magenta.

Peroxidases are natural platforms for engineering peroxygenation reactions since they react efficiently with hydrogen peroxide. The low peroxygenase activity of CcP may be attributed to the low affinity of the heme pocket for apolar substrates. In 2013, Erman et al. employed the CcP platform to build specific peroxygenation activity into CcP to mimic P450 monooxygenases.¹³⁸ In particular, a triple CcP variant, R48A/W51A/H52A (Figure 8b), was designed to replace the H-bonding residues in the distal heme pocket with residues containing apolar side chains in order to promote binding of the organic substrate 1-methoxynaphthalene. The naphthalene hydroxylation activity of this engineered CcP mutant was 34-fold greater than that ofWT-CcP and even comparable to that of WT-P450.

In many cases, the catalytic potentials of native enzymes originate from key transition metal centers in the active site. However, redox active amino acids, such as Trp and Tyr, can also play a pivotal role in catalysis, and the surrounding microenvironments of these residues are essential in tuning their redox potential.^140,141 This is true for lignin peroxidase (LiP), a fungal enzyme that plays a key role in the ligninolytic cycles for the degradation of the complex polymer lignin.¹⁴² The surface Trp171 residue of LiP is proposed to be the site of lignin oxidation, where the surrounding negatively charged residues (Asp264, Asp165, Glu168, and Glu250) provide an acidic microenvironment that enhances the oxidation potential of Trp^•+.^143,144 Taking inspiration from LiP, Smith, Ivancich, and co-workers redesigned Coprinus cinereus peroxidase (CiP) by incorporating a surface Trp residue (D179W) along with two negatively charged neighboring residues (R258E and R272D) to impart activity toward veratryl alcohol.¹⁴⁵ More recently, Field et al. created a similar series of CcP variants (A193W, Y229W, and A193W/Y229W) with Trp residues near the protein surface to mimic the function of LiPs.¹⁴⁶ The double mutant A193W/Y229W was found to exhibit the highest oxidizing activity, arising from an efficient electron hopping pathway that acted to accelerate oxidation of the aromatic amino acid.

2.3. Rational Design Using Cytochrome c as a Scaffold

Cytochrome c (Cc) is an electron transfer protein that contains a six-coordinated Fe heme center with axial ligation by one Met and one His (Figure 9a).¹⁴⁷ The electron provided by Cc can be utilized by other enzymes, such as cytochrome c oxidase (CcO), for their functions. Unlike typical peroxidases that harness pentacoordinate heme iron with a vacant distal coordination site for substrate binding, the occupation of the distal coordination site by Met80 in Cc seems to preclude peroxidase activity. However, Cc-catalyzed peroxidase reactions still occur despite an initial lag phase that corresponds to the in situ conversion of the six-coordinate precatalyst to an active form, which involves permanent displacement of the Met80 ligand from the heme iron.¹⁴⁸ Therefore, the intrinsic peroxidase activity of Cc is suppressed by the protein matrix in its native Met80-bound state when compared to other peroxidases such as CcP and HRP. While the proximal His18 ligation to heme iron in Cc is very robust, the 70–85 Ω loop (also known as Ω-loop D) is quite flexible,^149,150 contributing to a relatively weak Met80-Fe^III bond.¹⁵¹ Consequently, displacing the Met80 ligand by mutating it to a noncoordinating residue and/or perturbing the unstable Ω-loop D provides a route to converting Cc into peroxidase-like metalloenzymes. Moreover, the remarkable stability and compact globular protein structure of Cc, as well as the covalent bound heme group, make it an excellent structural framework for reengineering.

Figure 9.

(a) Crystal structure of yeast iso-1 Cc (top, PDB ID: 2YCC).¹⁴⁷ The flexible coordination loop, Ω-loop D, is colored in red. At the bottom is the multiple sequence alignment of the distal heme region of some classical peroxidases which is compared to the sequence of the distal heme region of yeast iso-1 Cc.¹⁵⁴ (b) Substrate recognition sites (SRSs) in P450 enzyme (CYP2C9, PDB ID: 1R9O)¹⁵⁵ shown by arrows: SRS1 (red), SRS2 (green), SRS3 (blue), SRS4 (yellow), SRS5 (orange), SRS6 (magenta).¹⁵⁶

In order to impart peroxidase-like activity in Cc, the M80V, T67H, and T67H/M80V mutants have been employed to examine the effects of removing the distal Met80 as well as of adding a His residue to the distal pocket.¹⁵² The M80V mutation was made in order to displace the native distal Met80 and generate an open coordination site at the heme iron, while the T67H mutation adds a distal His residue with an N_ε(His67)-to-Fe distance resembling that found in HRP and CcP (5.84 and 5.55 Å, respectively). Interestingly, the single mutant T67H Cc showed the highest peroxidase activity, even higher than that of T67H/M80V Cc, suggesting that the presence of a distal His residue is more important than eliminating the axial Met ligand for the enhancement of the peroxidase activity in Cc. Further modeling studies suggest that the additional M80V mutation leads to improper positioning of the His67 side chain, thus negatively impacting its function as an acid–base catalyst in the peroxidase cycle.

In addition to SDM, new routes have been explored to design novel metalloenzymes by replacing a full segment of the template protein, such as a loop, with another segment from a different protein. In Cc, residues of Ω-loop D (including Met80) constitute the distal coordination environment of the heme iron. Replacing these residues with key protein segments from other heme-containing enzymes that are responsible for substrate recognition and binding offers a great opportunity for protein engineering. In 2011, Ying et al. reported a P450-mimetic system in Cc by constructing hybrid proteins in which the segment of the coordination loop (residues 78–85) of yeast iso-1 Cc was substituted by the “substrate recognition sites” (SRSs) of P450 (Figure 9b).¹⁵³ Two constructs, HY1 and HY2, which possess SRS-1 and SRS-2 of P450, respectively, were successfully expressed and further exhibited P450-like function by catalyzing the oxidation of styrene while maintaining the high stability of WT-Cc. In this manner, these hybrids inherited the advantages of both parent proteins. Further enhancement of the peroxidation activity was accomplished by incorporation of the heme pocket of a peroxidase into Cc by structure-assisted mutagenesis in the distal-ligand-containing loop (Tyr67, Asn70, Pro71, and Met80) of Cc along with a highly conserved amino acid sequence.¹⁵⁴ The enhanced activities in these variants were primarily ascribed to the stabilization of compound I through the H-bonding interactions involving His and Arg residues. Taken together, these results indicate that loop residues 71–85 are critical for the function of Cc and that protein engineering targeting this region is a promising strategy for generating novel biocatalysts.

2.4. Rational Design Using c-Type Cytochrome Scaffolds

Although peroxidases are ubiquitous enzymes that catalyze the oxidation of many organic compounds, their poor thermal and environmental stability limits their range of practical applications.¹⁵⁷ Therefore, engineering peroxidases with improved thermal stability offers a promising route to broadening their industrial application. In 2008, Watanabe and co-workers presented the first design of a thermally stable artificial peroxidase in cytochrome c₅₅₂ (Cc₅₅₂) from Thermus thermophilus (Figure 10).¹⁵⁸ In these studies, mutations were made to remove the sixth Fe-coordinating ligand (M69A) and to introduce a general acid–base catalyst (V49D). Although V49D/M69A Cc₅₅₂ displayed an increase in peroxidase activity, significant heme degradation prevented this mutant from achieving sustained catalysis over long periods of time. Shortly after, an improved design of this artificial peroxidase was reported by the additional replacement of Tyr45 with phenylalanine (Y45F) or tryptophan (Y45W) in the basal mutant (V49D/M69A), which resulted in both higher peroxidase activity and thermal tolerance.¹⁵⁹ Many class I peroxidases, including CcP and APX, contain a Trp residue in the vicinity of the heme iron, acting as a functional receiver of the oxo-ferryl heme-π-cation radical. EPR studies of the V49D/M69A mutant of Cc₅₅₂ indicate weak magnetic coupling between the oxo-ferryl heme and Tyr45 radical. Replacing Tyr45 with Trp or Phe resulted in delocalization of the radical and increased magnetic coupling between the oxo-ferryl heme and the protein radical, which helped to suppress heme degradation while increasing enzymatic activity. These results suggest that engineering proteins from thermophiles could be a promising approach to generating artificial enzymes for practical applications. Moreover, this work demonstrates the importance of fine-tuning the microenvironment of redox active protein residues in improving catalytic performance.

Figure 10.

Crystal structure of Cc₅₅₂ (PDB ID: 1C52).¹⁶⁰ The key residues for mutations are shown in light yellow.

2.5. Rational Design Using Cytochrome P450 as a Scaffold

Cytochrome P450s (P450) have enormous biotechnological potential due to their unparalleled activity and selectivity toward oxidative chemistry. Members of the class VIII subgroup of this superfamily, and particularly its’ first discovered member, P450_BM3, are commonly employed as scaffolds for redesign as they are fused with their complimenting reductase, forgoing the need for purification and assaying of an additional protein component.¹⁶¹ Furthermore, the high turnover rate and bountiful structural data available for P450_BM3 have solidified its instrumental role in the advancement of protein engineering using P450 enzymes. Although numerous studies using both rational design and directed evolution approaches have unearthed common residues that can be altered to expand the scope of the reactivity, the amount of structural evidence capable of providing insight into how mutations guide the binding of new substrates or enhance activity is relatively limited. In 2013, Butler et al. reported the crystal structures of A82F and F87V/A82F mutants of P450_BM3 with and without substrate bound (Figure 11a), showing for the first time how the A82F mutation can serve as a “gatekeeper” to alter the substrate binding pocket toward the nonnative omeprazole substrate.¹⁶² The enhancement of ligand binding affinity in the A82F mutant was attributed to the large change in free energy between the ligand-free and ligand-bound conformations. Notably, the A82F mutation significantly destabilizes the ligand-free state, thus favoring access of the substrate to the binding site.

Figure 11.

(a) Mode of binding of omeprazole shown for the F87V/A82F P450_BM3 double mutant. Reproduced from ref 162 under the Creative Commons CC-BY license. (b) Active site of substrate-free P450_BM3 (PDB ID: 2IJ2).¹⁶⁸ Conserved residues Thr268, Phe393, and Cys400 are shown in green, and the heme cofactor is shown in purple. (c) From left to right are the crystal structures of WT-P450_BM3h (PDB ID: 1BU7),¹⁶⁹ the F393A mutant (PDB ID: 1P0V),¹⁷⁰ and the F393H mutant (PDB ID: 1JME)¹⁷¹ with the side chain of F393H orienting in two fashions. In the structure of F393A, a hydrogen bond is formed between the side chain amide nitrogen of Q403 and the thiolate of the heme ligating Cys400. For the two orientations of the His side chain in F393H, in one orientation the His side chain hydrogen bonds to an exterior water molecule. In a ring-flipped conformation, the side chain may form a weak hydrogen bond with the axial Cys.¹⁶⁷

Engineering efforts in repurposing P450_BM3 for cyclopropanation reactions have resulted in highly active, trans selective enzymes. However, engineering in P450_BM3 for cis-diastereomers has had limited success, requiring a large number of mutations.¹⁶³ To access stereoselective cyclopropanation catalysts using simpler variants, Gober et al. applied a strategic mutation, T268A, into a panel of 13 diverse P450 scaffolds and identified T286A P450_BM3 as having extraordinary trans-selectivity.¹⁶⁴ Although the detailed function of this mutation is not well understood, the crystal structure of T286A P450_BM3 suggests that this mutation has little impact on either secondary or tertiary protein structure but rather serves to alter protein dynamics and/or H-bonding in the active site to disfavor the O₂ binding and enable nitrene/carbene precursor binding¹⁵⁶

Mutation of the conserved heme ligating Cys residue (Figure 11b) to Ser or His in the PCS has provided a basis for P450_BM3 variants with substantially increased cyclopropanation activity.^165,166 However, these changes in the PCS can result in the stability of the protein itself being compromised. A complementary route toward engineering P450-based biocatalysts for highly active and selective cyclopropanation reactions is through modulation of the SCS.¹⁶⁷ It has been demonstrated that mutation of the conserved Phe393 residue to His or Ala increased the heme reduction potential and conversion of the cyclopropanation reaction to comparable levels of the PCS mutants without sacrificing protein stability. Structural analysis (Figure 11c) of the Phe393 variants suggests that the F393A and F393H mutations led to the formation of an electron-withdrawing H-bond to the Cys ligand via Gln403 or His393, thus increasing the heme reduction potential, which is the key to increasing catalytic reactivity.

2.6. Development of Heme Enzymes beyond the Primary Coordination Sphere by Directed Evolution

While the rational design of enzymes based on structure and calculations can offer insights into the role of residues beyond the PCS in promoting catalysis, this approach can still be limited by human intuition or computer programs that may not be able to capture or reproduce the functional properties of PCS and SCS features, especially subtle differences of metal ions and weak SCS interactions. Directed evolution of enzymes allows a route of overcoming these limitations to not only improve enzymatic activity but also unveil new structure–function relationships that are not readily intuited from a rational design approach, including residues far away from the PCS. The general approach of directed evolution uses various techniques to generate a random library of protein mutants, which is then screened based on desired activity.^59,130,172 The main requirements for this process to work are (1) starting from an enzyme with nonzero target activity, (2) sufficient stability of the target enzyme to undergo several rounds of mutagenesis, and (3) an effective screening process for assaying the target activity.¹⁷³ Several ensuing directed evolution studies focused on the P450_BM3 enzyme to enhance its substrate or product scope.^174-178 In one variant, the positions of several activity enhancing mutations were found to be surprisingly distant from the active site, with less than half of mutations appearing in the substrate binding region.¹⁷⁴ Notably, mutations that have been employed by rational design were not found in the selection process of these enzymes.¹⁷⁹ Based on activity and binding assays, it was concluded that the electronic environment around the heme binding site was similar to that of the native enzyme. However, despite the directed evolution engineered variant having a higher turnover with native substrate than its native enzyme counterpart, it did not exhibit the same regio- or enantioselectivity, demonstrating the importance of the screening method in acquiring the desired activity.¹⁷⁹ In such cases where the development of an appropriate assay is not possible, rational design can be used complementarily. For example, the addition of an A238V mutation to another engineered variant, 9-10A, results in the positioning of a residue close to the alkane methyl group at the active site, drastically improving the regioselectivity for alkane oxidation.

While directed evolution has successfully tailored a number of enzymes for specific activity with target substrates, the roles of many of the resulting mutations are not particularly well understood. However, based on previous rational design studies, a number of mutations in particular scaffolds can be rationalized. P450 has been particularly thoroughly studied, and the reader is referred to ref 180 for an in-depth discussion of the enzyme itself. For unnatural reaction landscapes explored by directed evolution, the reader is referred to refs 77 and 180-187, which cover the most recent advancements in the field, specifically for heme-enzymes, and also focus on the mechanistic aspect of these unnatural transformations.

2.6.1. Discovery of Noncanonical Reactions for Cytochrome P450.

The main driving force behind the P450 mediated oxidation reaction is the formation of an Fe^IV=O− (heme b^•+) intermediate termed as compound I. The core of this system and the subsequent reduced system termed compound II is isoelectronic with iron carbenes and nitrenes, the former of which is well studied (Figure 12).¹⁸⁸ Inspired by the similarities between synthetic Fe-porphyrin carbene complexes and compound I, Arnold and co-workers employed P450_BM3 for the study of carbene transfer reactions, starting with cyclopropanation of styrene.¹⁶³

Figure 12.

Structural comparison of the free heme cofactor bound to the cysteine thiolate compared with compound I, iron-carbene, and iron-nitrene intermediates. Reproduced with permission from ref 186. Copyright 2021 American Chemical Society.

To identify an optimal starting candidate for further reaction optimization via directed evolution, native P450_BM3 (Figure 13) and 92 other variants were included in an initial selection pool.¹⁶³ Native P450 is capable of catalyzing the cyclopropanation of styrene; albeit, very low yield numbers (TONs) above 100 have been found. Notably, a number of these active variants, including H2A10, H2-5-F10, and H2-4-D4, contain several Ala residues in the active site across positions Leu75, Met177, Leu181, Ile263, Thr268, and Leu437. To elucidate the effects of each, a series of SDM studies with respect to each active site residue was performed on different select variants. As a result, the T268A mutation was identified as the most important factor for modulating cyclopropanation activity, with a 60-fold increase in activity observed for the single mutant relative to WT-P450_BM3. This residue has also been identified as fundamental in native P450 activity, in which its elimination results in almost complete deactivation.¹⁸⁰ Based on these observations, it can be inferred that one of the main factors in improving selectivity for newly introduced non-native reactivity is the suppression of the native reaction pathway. These findings concluded on the development of a highly reactive and cis-selective variant termed P450_BM3-CIS containing 13 mutations. Next, site saturation mutagenesis (SSM) on the active site residues Leu181, Ile263, Ala328, Leu437, and Thr438 in this variant helped to elucidate the influence of each on reactivity. Out of these residues, only mutations at Ala328 and Thr438 improved the cis-selectivity. The A328G, T438A, T438S, and T438P mutations were found to be the most beneficial without significantly reducing TON.¹⁸⁰

Figure 13.

Structure of the active site of WT-P450_BM3 (PDB ID: 4ZFA) shown (a) as the full domain and (b) at the enzyme active site.¹⁸⁹ Commonly targeted residues for mutagenesis are highlighted in purple, Thr268 is displayed in green, and heme b is in orange. Some residues are not labeled for clarity.

P450 enzymes exhibit redox switching behavior in which the Fe^III/Fe^II potential is decreased upon binding of substrate. Importantly, this behavior only enables reduction of the heme cofactor (and subsequent O₂ binding) in the presence of substrate, helping to mitigate the generation of reactive oxygen species (ROS) which could potentially degrade the enzyme or waste reductant via direct formation of compound I or by peroxide-to-water reduction (the oxidase and peroxide shunts, which have been extensively studied).^190-192 In ArMs, however, this redox switch is no longer intrinsically present, requiring both stronger reductants to effectively reduce ferric heme and raising potential concerns regarding both inefficiency and protein degradation. Therefore, in the absence of such a mechanism, practical applications of designed P450 mimics as carbene and nitrene transfer catalysts would be limited.

To overcome this barrier, a series of varying proximal ligands in P450 were assayed to optimize activity under in vivo conditions in which C400H and C400S variants were found to be the most active. The latter resulted in a shift of the Soret band for the ferrous-CO bound state to 411 nm, and therefore, the C400S P450_BM3 variant is referred to as P411_BM3.¹⁶⁵ The increased activity of this variant was attributed to an increase in the Fe^III/Fe^II redox potential of the active site and subsequent stabilization of the carbene intermediate. Interestingly, this also drastically increases the cis-selectivity of the enzyme, which could allude to enhanced bond rigidity of the carbene intermediate. Based on these findings, other groups utilized the T268A and C400S mutations on homologous P450 enzymes to achieve similar reactivity with other substrates.¹⁹³

Combining chemical intuition with directed evolution has proven fruitful for advancing the reaction landscape of these developed enzymes. Wang et al. developed a new P450_BM3 cyclopropanation variant with only five mutations relative to the wild-type enzyme (C400H, T268A, L437W, V78M, and L181V) capable of reacting with N,N-diethyl-2-phenylacrylamide in very high yields, as well as high enantio- and diastereoselectivity.¹⁹⁴ This showcases the importance of the evolution pathway in reaction optimization.⁷⁷ In contrast to P450_BM3-CIS, this variant was able to achieve higher reactivity with only five mutations, albeit with different substrates. Furthermore, this variant has shown wide alkene selectivity.¹⁶⁶

Direct evolution has also been used to expand the substrate scope of cyclopropanation. Brandenberg et al. developed two sets of new variants, P411-VAC_cis and P411-VAC_trans, which were optimized for the catalysis of N-vinylphthalimide cyclopropanate to cis- and trans-cyclopropanes, respectively.¹⁹⁵ Previous studies have demonstrated the activity of P450 variants toward N-H insertion,¹⁹⁶ and this work further highlighted the challenges involved in designing enzymes with new reactivity while suppressing possible side reactions. Interestingly, the P411-VAC_cis and P411-VAC_trans constructs were also capable of catalyzing O- and S-vinyl compounds, albeit with one additional mutation. The substrate scope of this system was further diversified by the development of a new P411_BM3 variant capable of accepting nonaromatic substrates for cyclopropanation,¹⁹⁷ and a subsequent study found additional new variants capable of catalyzing C₃ or C₂ selective pyrrole alkylation of indoles and cyclopropanate cyclic hydrocarbons.¹⁹⁸ The key to enabling reactivity has been the utilization of two generations of directed evolution, initially for the development of indole reactivity and subsequently for diversification toward either hydrocarbons or pyrroles.

Directed evolution has also been used to develop P411_BM3 into a C-H insertion carbene transfer enzyme.¹⁹⁹ Specifically, P411-CHF was developed by directed evolution of P411_BM3 and shown capable of sp³ C-H activation with aromatic substrates with high enantio- and regioselectivity for α-carbon over aryl or allylic carbons in alkene substrates. It was also found to be highly selective for C-H insertion over cyclopropanation. Further development of P411-CHF also led to selectivity for highly fluorinated alkanes.²⁰⁰

The scope of the insertion activity of P450 has also been expanded further to include tandem cyclopropanation of bicyclobutanes.²⁰¹ Previously developed variants of P411 and other heme proteins with nitrene insertion activity were screened for carbene transfer activity with alkyne substrate to provide a stepping stone for subsequent directed evolution of the most active construct (P411-E10). Structural analysis suggested the Trp263 acted to obscure the active site, inhibiting its ability to perform tandem cyclopropanation. Additionally, a second P411_BM3 variant was separately developed for cycloprenation catalysis, starting from a less reactive alkyne substrate. The mutations incorporated to achieve this reactivity, however, have not been rationalized. In another study, cyclopropene formation was developed further to incorporate less reactive internal alkynes.²⁰²

In most studies of P450-based carbene transfer activity, the carbene precursor, with a few exceptions,^198-200 has been ethyl diazoacetate (EDA). Recently, a new P411-CIS variant has been developed which binds α-diazo lactone rather than EDA to perform enantio- and diastereoselective cyclopropanation reactions as well as S─H bond insertion with aromatic substrates.²⁰³ The activity of this precursor with the S400C variant (reinsertion of the native Cys ligand) has been studied as well, demonstrating that the proximal ligand affects reaction TON but not selectivity. This reaction has also been independently studied using another variant of P411_BM3, which was also capable of binding the same lactone and performing enantioselective C─C bond formation.²⁰⁴

Nitrene transfer reactivity has also been explored using the same strategy employed for carbene transfer, and the same mutants optimized for cyclopropanation also exhibit activity toward intramolecular amination.²⁰⁵ To this end, P411_BM3 variants have been commonly employed, including one based on the CIS variant¹⁶³ and another a double mutant of P411_BM3.²⁰⁵ A similar cyclization study performed by Singh et al. showcased the radical character of the aziridation reaction.²⁰⁶ In this study, different stereoisomers reacted with the enzyme to yield the same product, indicating a common intermediate and ruling out a concerted insertion mechanism. Furthermore, comparison of the kinetic isotope effect (KIE) for the cleaved C─H bonds to those of known, concerted, C─H insertion reactions and the correlation of the reaction rate with the C─H bond strength further suggested an intramolecular hydrogen abstraction by the nitrene as the rate-limiting step. Comparisons with previous research²⁰⁷ and analysis of undesired side products demonstrated that the main side reaction for the putative nitrene intermediate is an enzyme mediated reduction similar to native hydroxylation reactions.

In 2016, Arnold and co-workers utilized the sigmatropic rearrangement of allylic sulfimides to form an N─C bond following the initial enzyme-catalyzed nitrene transfer to the sulfur atom.²⁰⁸ The finalized variant, developed through five rounds of SSM, was capable of reacting with a variety of different phenyl thioethers. The developed mutant (P5) was developed to target allylic substrates and was also found to react with nonallylic substrates with high enantioselectivity and TON.

The first enzymatic olefin aziridation catalyst was constructed by Farwell et al. based on P411-CIS, which produced significant quantities of amido-alcohol products due to the styrene-based skeleton of the substrate.²⁰⁹ In 2017, one of the first cases of C-H amination by P411_BM3 variants was reported.²¹⁰ In this variant, SSM of the previously developed P411-CIS variant culminated in a variant termed P411-CHA This variant contains a “reversal” of Phe263 to Leu in addition to three other mutations which activate the enzyme toward enantioselective reactions with different aromatic substrates. Using this approach, Brandenberg and co-workers engineered a P411_BM3 variant capable of selective indole amination.²¹¹ Initial rounds of SSM revealed a variant capable of [3 + 2] cycloaddition reaction, showcasing the possibility of substrate binding without nitrene generation. Interestingly, this behavior was enabled by a W1046F mutation at the C-terminus of the enzyme. The motivation behind including this region for SSM was to reduce the rate of substrate reduction, and it was initially attempted by truncating the enzyme after its reduction domain to leave only the P450 region. After observing the adverse effects of truncation, SSM of the NADPH binding site was attempted. As Trp1046 stacks on top of the NADPH ring, the mutation to Phe was speculated to disrupt binding, which is supported by the resulting 10-fold decrease in electron transfer rate.

Another study in 2020 expanded the substrate scope to allylic and benzylic amines.²¹² The nitrene precursor has been changed from tosylazide to pivaloylhydroxyamine, which was based on a previous study²¹³ inspired by the findings of Tsutsumi et al.²¹⁴ and Morandi et al.²¹⁵ This serves to disable the direct reduction of the substrate and provides a final primary amine as the product instead of a secondary tosylamine. For benzylic amines, P411-BPA has been engineered, which acts with high selectivity toward them. Another variant, termed P411-APA, has been found to be successful at allylic aminations. The difference between these variants is the addition of two mutations at N395R and S438A.

In 2021, Athavale et al. reported a P411 variant developed through nine rounds of directed evolution which is capable of nitrene transfer of pivaloxyl-amides.²¹⁶ In contrast to the P411 enzymes designed for tosylazide as the substrate, the rate-determining step (RDS) in this variant is not C-H activation but nitrene formation as shown by KIE studies. These variants extend the substrate scope of the benzylic amines while also increasing the TON compared to that of P411-BPA.

As discussed vide supra, the redox switch mechanism is not intrinsically present in engineered enzymes, the O₂ binding capacity has been drastically altered via numerous mutations, and the active intermediate formation and substrate binding steps can be at least partially uncoupled. As a result, for the carbene intermediate, side reactions with nearby residues are one of the main factors limiting the enzyme turnover number (TON).²¹⁷ For nitrene transfer, the main barrier has been the reduction of the nitrene intermediate to its corresponding amine prior to reaction with the substrate. These side reactions could be potentially suppressed by either mutation of the residues responsible for the side reaction or optimization of the enzyme reactivity to increase the reaction rate of the generated compound one-like intermediate.²¹⁸

2.6.2. Myoglobin as a Scaffold.

One of the first reports of directed evolution in a heme protein was presented by Wan et al., who enhanced the peroxidase activity of Mb 25-fold through a quadruple mutant without sacrificing the O₂-binding affinity.²¹⁹ This same mutant was later used as a basis for the development of Mb variants active toward carbene and nitrene transfer. Several key amino acids that define the distal pocket in Mb significantly influence the activity and selectivity of this artificial biocatalyst in carbene- or nitrene transfer reactions.^124-127,131 In particular, mutations at the “gating” residue His64 lead to a general activity enhancement for many carbene-mediated reactions. SSM at the active residues Leu29, Phe43, His64, Val68, and Ile107 (Figure 14a) has proven to be a valuable strategy in generating biocatalysts with high proficiency and superior selectivity.

Figure 14.

(a) Crystal structure of WT-swMb (PDB ID: 5YCE)⁸⁹ with key residues in the distal heme pocket for directed evolution shown in magenta. (b) Stereochemical model for intramolecular cyclopropanation catalyzed by the stereodivergent Mb variants. Reproduced from ref 220. Copyright 2019 American Chemical Society.

By screening the active site mutation landscape library, in which active positions are subject to substitution by all other possible 19 amino acids, two triple mutants of Mb, L29A/H64V/V68A and F43A/H64W/V68F, were developed that displayed high but complementary stereoselectivity toward intramolecular cyclopropanation (Figure 14).²²⁰ Although detailed structural evidence for this stereodivergence is currently lacking, it is clear from mapping these mutations onto the Mb structure that the active site configurations of these two mutants are distinct; smaller residues in positions 29 and 68 in L29A/H64V/V68A render an expanded cavity in the upper side of ring A/D of the heme, while increased steric bulk in the F43A/H64W/V68F variant enlarged the cavity at the opposite side of the cofactor (Figure 14b). Therefore, by fine-tuning the distribution and expansion of the distal heme pocket by active site mutagenesis, one can modulate the stereoselectivity of Mb-based biocatalysts.

The active site of Mb has also been configured using directed evolution to enhance opposite enantioselectivities for intramolecular cyclopropanation via the quadruple mutants F43L/H64A/V68G/I107V Mb and F43Y (or F43H)/H64A/V68G/I107F Mb.²²¹ Notably, R-selectivity was achieved by increasing the steric bulk at position 107 (I107F) together with addition of an aromatic group with hydrogen bonding abilities (F43Y or F43H) at position 43. In contrast, S-selectivity could be made favorable by reduced the steric bulk at both positions 43 and 107 (F43L and I107V). To expand the substrate scope to more sterically demanding intramolecular cyclopropanation reactions, SSM was used to target active site residues to further expand the distal heme pocket, resulting in an optimized F43Y/H64V/V68A/I107V Mb mutant.²²² These results also supported that positions Phe43 and Ile107 are especially critical for the stereoinduction of substrate when large (i.e., aryl) substituents on the olefinic group are present.

Inspired by previous work investigating mutations at His64 and Val68 to tune the stereoselectivity of the cyclopropanation reactions,^124,135 Hernandez et al. mutated the analogous residues (T45 and Q49) in B. subtilis truncated globin to one of three nonpolar residues (Leu, Phe, and Ala) in an effort to enhance the performance toward cyclopropanation.²²³ The double mutant T45A/Q49A exhibited high selectivity, possibly due to the increased volume of the active site.

It is notable that the structure-guided rational design and directed evolution methods are not alone in guiding protein engineering for developing artificial metalloenzymes. Although these two approaches have provided fruitful results, both methods are time consuming and labor intensive. The efficiency of protein engineering can be significantly improved by employing calculations to identify promising target residues and mutations. Thanks to advances in computational tools, calculation-guided methods have emerged as effective comple-mentary tools for protein engineering. For the practical use of artificial enzymes, protein stabilization against thermal and chemical denaturation is a long-standing goal in enzyme engineering. In 2017, Moore et al. reported the development of a Rosetta-guided protein design to optimize protein stability for better performance in enzymatic catalysis.²²⁴ Starting from the H64V/V68A Mb construct, which has been found to perform carbene-mediated cyclopropanation reactions with enhanced activity and high diastereo- and enantioselectivities, the authors utilized the Rosetta enzyme design framework to identify the optimal positions for incorporating a thioether covalent staple to improve its robustness for synthetic applications, obtaining nine designs that were further examined for stability and catalytic activity in cyclopropanation reactions. This methodology allowed them to uncover stabilized variants with superior catalytic performance compared to the parent enzyme when exposed to organic solvents.

2.6.3. Rma Cytochrome c as a Scaffold.

Protein machinery promiscuity is observed not only in the P450 enzymes but also in bacterial cytochrome c from Rhodothermus marinus (Rma-Cc, Figure 15a), which is an electron transfer protein and has no catalytic functions in living systems. However, Arnold and co-workers have uncovered activity in Rma-Cc toward carbene-mediated C─Si bond formation and thereafter pursued the laboratory evolution of Rma-Cc variants to perform many new-to-nature functions with high activity and selectivity.²²⁵

Figure 15.

(a) Crystal structure of WT-Rma-Cc (PDB ID: 3CP5)²³⁰ with the key residues involved in directed evolution shown in purple. (b) Crystal structure of carbene-bound TDE-Rma-Cc (PDB ID: 6CUN) with the carbene species in cyan. (Inset) Interactions between the carbene and amino acid residues are shown with yellow lines: (i) Thr75, 3.7 Å; (ii) Met76, 4.3 Å; (iii) Pro79, 3.5 Å; (iv) Ile83, 3.6 Å; (v) Met89, 3.5 Å.²²⁶

In 2018, Kan et al. generated a triple mutant, V75T/M100D/M103E (TDE-Rma-Cc), by SSM at three positions near the distal heme pocket to perturb the distal ligand (M100) and the two residues that are close to the heme iron (Val75 and Met103).²²⁶ The evolved variant catalyzed the C─Si bond formation with more than 15-fold higher turnover than state-of-art synthetic catalysts and extremely high enantioselectivity (>99% ee). In 2018, Lewis and co-workers resolved the high-resolution crystal structure of the TDE-Rma-Cc mutant, which revealed a mutation-generated pocket in the distal side of the heme.²²⁶ Molecular dynamics (MD) simulations indicated that the front loop (Thr98-Glu103) of the active site is quite flexible, adopting an “open” conformation which favors the accessibility of substrate to the iron porphyrin carbene (IPC) intermediate with pro-R facial selectivity. The more recently solved high resolution crystal structure of carbene-bound TDE-Rma-Cc reveals how the laboratory-evolved enzyme achieved perfect carbene transfer by holding the iron porphyrin carbene in a single orientation (Figure 15b).²²⁶

In addition to C─Si bond formation, the Rma-Cc-based biocatalyst has been evolved for C─B bond formation by multiple rounds of SSM.²²⁷ By incorporating different substitutions at Val75 and Met103, the borylation reactivity was directed to diverse diazo ester substituents. The substrate scope of these biocatalysts was later expanded to perform C─B bond-forming chemistry with more rigid, cyclic, lactone-based carbenes by creating a V75R/M99/M100D/T101Y/M103V mutant (BOR^LAC).228

Encouraged by the beneficial mutations targeting the active site front loop, Cho et al. continued to explore the new-to-nature functions of the thermostable Rma-Cc platform by generating a biocatalyst for enantioselective aminohydroxylation of styrenyl olefins.²¹³ Apart from the SSM targeted residues (Thr98, Met99, Thr101, and Met103) within the same loop as the Met100 ligand, mutations at Met76 and Tyr44, which are relatively far from the active site, also improved the catalytic performance. Similarly, a Y44I mutation enhanced the borylation activity and enantioselectivity of an Rma-Cc-based biocatalyst for the synthesis of chiral α-CF₃ organoboranes.²²⁹ MD simulations revealed that the -CF₃ group pointed toward the active site, whereas the bulkier phenyl group lay between the side chains of the M103D and Y44I substituted residues, implying that the primary roles of these mutations were to facilitate the binding of the diazo substrate. Therefore, residue 44 could be a key to modulating substrate selectivity in this artificial enzyme.

The TDE-Rma-Cc mutant was revisited in 2021 by Garcia-Borràs et al. to study the origin of the chemoselectivity of this variant toward C─Si over C─N bond formation.²³¹ A combination of computational and experimental methods suggests that the protein front loop conformation and dynamics dictate silane binding, which is a major contributor to the rate-limiting step. The more favorable silane binding affinity in the TDE-Rma-Cc mutant may result from the increased flexibility of the front loop, induced by the hydrophilicity of the Asp100 and Glu103 side chains. Redesign of several mutations in the front loop led to the generation of a new protein (N80F/M99P/E103I TD-Rma-Cc) in which chemoselectivity was switched from silylation to amination.

2.7. DNA as a Scaffold for Heme-Based Biosynthetic Models

While proteins have long been known to serve enzymatic functions, it has also been discovered in recent decades that both RNA and DNA can function as enzymes.^232-234 Naturally occurring oligonucleotides are composed of a much narrower range of building blocks in comparison to proteins, and therefore, recruiting metallocofactors such as heme can help oligonucleotides to expand their enzymatic capabilities. Based on this hypothesis, Li, Geyer, and Sen used in vitro selection to obtain single-stranded DNA aptamers that specifically bound N-methylmesoporphyrin IX (NMM), a stable transition-state analogue for porphyrin-metalation reactions.²³⁵ A library of 1.8 × 10¹⁵ DNA molecules was screened for NMM binding affinity using NMM-bound acrylic beads through HPLC, subjected to both positive and negative selection in 12 rounds. In the DNA aptamers that bound strongly, sequence analysis revealed a common guanine-rich motif. Further characterization of the resulting DNA aptamers showed that hemin could form complexes with guanine-rich, short single-stranded DNA molecules with submicromolar affinities. The authors presented spectroscopic evidence that this complex more closely resembled a peroxidase than it did uncoordinated hemin, consistent with the observed increase in peroxidase activity from the hemin/DNA complex when compared to hemin alone.²³⁶

Shortly thereafter, the electron paramagnetic resonance (EPR) and UV/visible spectra of a folded guanine-rich oligonucleotide and hemin complex (wholly referred to as PS2.M-hemin) were shown to be similar to the spectra observed for high-spin ferric heme with axial symmetry, as observed with myoglobin.^237,238 The DNA component of PSM2.M-hemin was known previously to form a G-quadruplex structure, in which tetrads of guanosine stack atop each other to form a four-stranded supermolecule under physiological salt conditions.²³⁹ The initial work characterizing the peroxidase activity of a G-quadruplex/heme complex used ABTS as a substrate.²³⁶ A little more than a decade later, the G-quadruplex/heme complex was demonstrated to act as an NADH oxidase and NADH peroxidase mimic, oxidizing NADH via O₂ to generate H₂O₂.²⁴⁰ The complex of G-quadruplexed DNA and heme has since been characterized as a peroxidase-like DNAzyme, able to accommodate ligands in the distal sixth coordination site such as CO, O₂, and imidazole.^241,242 Heme-G-quadruplexed-DNA-zymes, however, have a major difference from protein peroxidases, as they lack a proximal imidazoyl ligand from a histidine amino acid residue.

Ten years after the spectroscopic investigation on PS2.M-hemin, the same complex was shown to catalyze two-electron oxidation reactions with substrates such as thioanisoles, indole, and styrene, and ¹⁸O labeling demonstrated that the transferred oxygen atom originates from hydrogen peroxide through what is presumably a ferryl moiety.²⁴³ Docking simulations of heme to a G-quadruplex structure suggested a lack of a structured environment distal to the heme, and it was suggested this is linked to the lack of enantioselectivity observed with oxidation reactions catalyzed by PS2.M-hemin, as the oxygen transfer reaction to thioanisole produces a racemic mixture and indole oxidation yields several products.²⁴³

Although no additional complexes of oligonucleotides with heme have been reported since, several other interactions do exist with other porphyrin molecules in complex with DNA.²⁴⁴ In G-quadruplex DNA formed from the repeating human telomere sequence d(TTAGGG), ferrous heme binds to the 3′-terminal G-quartet through π−π stacking interactions.^245-247 Chemical modifications of the heme in the d(TTAGGG)-heme complex to form different electron densities at the catalytic iron site altered the peroxidase activity of the complex, with decreasing peroxidase activity observed with decreasing ρ_Fe.²⁴⁸ Resonance Raman spectroscopy has revealed that the vibrational feature ν_CO for the [d(TTAGGG)]₄-heme complex with a bound CO adduct appears at a lower frequency than in myoglobin, suggesting a weaker donor strength of the proximal ligand and slightly longer Fe─C bond.²⁴⁸ This agrees well with earlier NMR work of chemically modified hemes in complex with G-quadruplex DNA, where the fifth (proximal) ligand appears to be a water molecule.²⁴⁷

In 2019, a different G-quadruplex/heme complex (called G4) was shown to work as a catalyst for carbene insertion into styrene, expanding the range of heme-DNAzymes beyond oxygen transfer reactions.²⁴⁹ As the heme molecule is situated above the 3′ G-quartet, the 3′ end of the oligonucleotide is free to interact with the cofactor—the authors report that the presence of 3′-terminal AAA or TTT results in indistinguishable effects upon carbene insertion, while AAA shows higher oxidation rates relative to TTT.²⁴⁹ The addition of multiple pendant nucleotides shows increasing stereoselectivity compared to porphyrin/DNAzymes that do not possess nucleotides on the distal side of the porphyrin. Figure 16 shows the “active site” of the heme-DNAzymes studied. Reactions were limited to oxene and carbene insertions, but they are mechanistically distinct from the Fe^IV-based, radical-dependent mechanism of heme monooxygenases.¹⁸³

Figure 16.

Active site of heme/DNAzymes (GQ-heme complexes), where the heme end-stacks on the 3′-most guanine-quartet (gray squares): (a) “G4-AAA” and “G4-TTT” oligonucleotides both form intramolecular, propeller looped GQs, with the 3′-terminal AAA or TTT overhanging (shown as blue rectangles). (b) An intermolecularly assembled GQ such as “(dA4G5A4)4” is expected to have a complex “distal” environment above the stacked heme, potentially imparting stereoselectivity to reactions catalyzed by the DNA-bound heme. Reproduced with permission from ref 249. Copyright 2019 American Chemical Society.

While there is a lack of crystal structure data for a G-quadruplex/hemin complex, efforts from the Mergny and Petitjean groups produced a planar Pt^II complex that binds in the same proposed position where hemin binds atop the G-tetrads at the 3′ end.²⁵⁰ This serves as a good structural model for the heme binding position on G-complexes in general, but the Pt/DNA complex does not display the proximal water ligand that has been demonstrated in peroxidase-type heme/DNAzymes.

3. CATALYSIS BEYOND THE PRIMARY COORDINATION SPHERE BY DESIGNED NON-HEME IRON ENZYMES

Non-heme Fe enzymes constitute a broad class of proteins that bind Fe to promote a variety of reactions, many of which parallel those of heme enzymes. Many non-heme Fe enzymes employ mono- or dinuclear Fe sites to promote O₂ activation and incorporation into organic substrates.^251,252 Additionally, non-heme Fe enzymes have been found to promote reactions including nitric oxide reduction, the reversible of oxidation of H₂, and even the reduction of atmospheric nitrogen to ammonia.^56,253-256 In many O₂-activating mononuclear non-heme Fe enzymes, the active site is composed of a (His)₂-carboxylate binding motif arranged to facially coordinate Fe^II, commonly referred to as the “facial triad”, with one or two labile ligands (such as H₂O) that allow facile binding of O₂.^257-260 Dinuclear non-heme Fe enzymes that serve to activate O₂ or reduce NO typically exhibit mixed His and carboxylate-coordinated active sites, often with one or more carboxylates (Asp or Glu) bridging the two irons that can appear in both the terminal and bridging positions and additional non-amino acid coordinating atoms such as H₂O, OH⁻, or O²⁻.^252,261 Further non-heme Fe enzymes may offer more exotic coordination environments, such as the 5,10-methlenetetrahydromethanopterin cofactor in Fe hydrogenase (H₂ase),²⁶² the bridging dithiolate/pendant amine cofactor and CO/CN⁻ ligands of FeFe H₂ase,²⁶³ and hetero iron-sulfur-based clusters of nitrogenases and carbon monoxide dehydrogenase (CODH).^264-267

The broad nature of non-heme iron systems naturally affords a high degree of modularity in the PCS and SCS. We have organized our discussion of these systems first in terms of mononuclear vs dinuclear systems and then further between approaches toward the active site, namely, (1) the use of amino acid residues to provide the PCSs of metal centers, (2) those which employ unnatural amino acids for metal coordination and those which incorporate a guest organometallic complex in a host protein scaffold (also referred to as the “Trojan horse” approach) using (3) covalent interactions and (4) supra-molecular interactions.

3.1. Design of Mononuclear Non-heme Iron Proteins

3.1.1. Design Using Natural Amino Acids.

One of the earliest studies to rationally design a catalytically active mononuclear non-heme Fe site into a non-native protein scaffold was performed by Pinto, Hellinga, and Caradonna.²⁶⁸ Utilizing the scaffold thioredoxin (Trx), a facial triad binding site was designed with the assistance of the program DEZYMER via Leu7His, Phe27Asp, Ile60His, and Asn63His mutations (Figure 17). Fe binding was observed via the appearance of a new absorption feature at 350 nm, assigned to a His → Fe^III ligand-to-metal charge transfer transition. When tested via colorimetric assay, the Fe-Trx ArM exhibited significant superoxide dismutase (SOD) activity, with a dismutation reaction rate of k ~ 10⁵ M⁻¹ s⁻¹.

Figure 17.

Crystal structure of Trx from E. coli (PDB ID: 2TRX).²⁶⁹ First coordination residues in designed Trx-SOD were made to residues highlighted in light gray, and SCS mutations are highlighted in blue.²⁶⁸ Atom coloring: S (yellow); O (red); N (blue).

Azurin (Az) has also served as an effective scaffold for the design of low-coordinate mononuclear non-heme Fe binding sites. Az is a type 1 copper metalloprotein but has garnered attention for protein engineering efforts due to its stability and ease of expression. The use of Az to bind Fe was first demonstrated by Holland and co-workers,²⁷⁰ who were able to incorporate Fe^II into both WT-Az and the M121A Az variant. Surprisingly, WT-Fe^II-Az appeared to resist oxidation and reduction by a variety of redox agents, suggesting that the accessible iron reduction potentials are outside the physiological potential range from approximately −1 to + 1 V vs SHE. Characterization by a combination of ¹H NMR and ⁵⁷Fe Mossbauer spectroscopies strongly indicates that there is little structural difference between WT-Fe^II-Az and M121A Fe^II-Az, implying the absence of Met121 does not significantly affect the coordination geometry. X-ray crystallography reveals clear coordination of Fe by His46, Cys112, and His117 and likely Gly45 as well. Further N₃⁻ binding experiments reveal that replacement of Met121 for Ala opens an additional binding pocket without perturbing metal binding. Additional binding experiments demonstrated the ability of CN⁻ to coordinate Fe but that alternative ligands such as N-methylformamide, hydrazine, and CO are incapable of binding.

Following the initial discovery of Fe binding in Az, the Liu group further engineered the binding pocket to impart superoxide reductase activity via a M121E substitution (Figure 18b).²⁷¹ Fe^II incorporation into M121E Az was supported by a combination of absorption spectroscopy, inductively coupled plasma (ICP) analysis, and electrospray ionization mass spectrometry (ESI-MS). Successful crystallization of M121E Fe^II-Az allowed structural determination at 2.0 Å resolution. Interestingly, M121E Fe^II-Az crystallizes as a tetramer of two heterodimers, with subunits A and C in conformation 1 and subunits B and D in conformation 2. No observable occupancy was found in conformation 1, while 100% occupancy was found in conformation 2. In addition to Gly45, His46, C112, and His117, Glu121 was found to coordinate Fe^II through the extended carboxylate side chain. Unlike WT-Fe^II-Az or M121A Fe^II-Az, the Fe^II → Fe^III redox couple of M121E Fe^II-Az was found to be readily accessible, with a reduction potential of +0.327 V. Further testing for superoxide reductase (SOR) activity using the Fridovich method²⁷³ revealed an SOR activity, with a reduction rate of k = 1.8 × 10⁴ M⁻¹ s⁻¹. Based on observations for native SOR, a second substitution was made exchanging Met44 for Lys. In the native system, it is believed this SCS residue aids in guiding superoxide to the metal center or even potentially stabilizes binding of the negatively charged superoxo via its positively charged side chain. This rational modification further only slightly lowered the Fe^II/Fe^III redox couple to +0.320 V but boosted the SOR activity of M121E/M44L Fe^II-Az by 2 orders of magnitude to 1.1 × 10⁶ M⁻¹ s⁻¹ relative to M121E Fe^II-Az, supporting the importance of this SCS interaction in promoting the SOR activity.

Figure 18.

(a) Crystal structure of M121E Az (PDB ID: 4QLW).²⁷¹ (b) Fe-bound metal binding site of M121E Az. (c) Cu-bound metal binding site of M121H/H46GE (PDB ID: 4WKX).²⁷² PCS atoms are highlighted in light gray, while SCS target residues are highlighted in blue. Atom coloring: S (yellow); O (red); N (blue); Fe (orange); Cu (copper).

More recently, the Lu group has further utilized Az to engineer the formation of a dinitrosyl iron complex (DNIC).²⁷⁴ Nitric oxide (NO) plays an important role in numerous processes ranging from cellular signaling to immune defense, and understanding how these processes are regulated by non-heme Fe proteins remains an intense subject of investigation. Note that in the following we will use Enemark–Feltham notation, {MNO}ⁿ, to describe the overall oxidation state of NO-bound metal (M) systems, where n denotes the number of valence electrons (metal d plus NO π* electrons).²⁷⁵ Using the M121H/H46E Az scaffold previously developed for Cu^II binding aimed at S-nitrosylation²⁷⁶ (Figure 18c), Fe^II binding was observed via a combination of absorption and ⁵⁷Fe Mössbauer spectroscopy. Further addition of 1 equiv of NO produced significant spectroscopic changes in the absorption spectrum, as well as the appearance of a new S = 3/2 signal as observed by EPR, indicative of {FeNO}⁷ formation. This new S = 3/2 species was actually a convolution of two signals with unique zero-field splitting (ZFS) properties, representing two different conformations of the {FeNO}⁷ species at low temperature, with one slightly more axial and the second more rhombic. By combining EPR and field-dependent ⁵⁷Fe Mössbauer spectroscopic measurements with density functional theory (DFT) calculations, it was proposed that these species represented two different PCSs. Specifically, the more axial species was proposed to arise from {FeNO}⁷(His)(Glu)(Cys) coordination, while the more rhombic from {Fe-NO}⁷(His)₂(Glu). Interestingly, room temperature resonance Raman (rR) measurements revealed a single set of Fe-NO and NO vibrations, indicating that the appearance of two conformers as observed by EPR and ⁵⁷Fe Mössbauer spectroscopy may only arise at low temperatures.

Exposure of M121H/H46E {FeNO}⁷-Az to an excess of NO produced a new novel species, which exhibited an axial S = 1/2 signal as observed by EPR. Using ¹⁴N/¹⁵N Q-band electron double nuclear resonance (ENDOR) spectroscopy, two isotope-dependent signals with relatively weak hyperfine interactions (3.6 and ~ 7 MHz in ¹⁵NO) were observed, supporting the formation of a {Fe(NO)₂}⁹ DNIC species. This hypothesis was further bolstered by the presence of two unique sets of Fe-NO/NO stretching frequencies as observed by Fourier transform infrared (FT-IR) and rR spectroscopies. Further ¹H Q-band ENDOR measurements supported Cys coordination, and by combining these results with ⁵⁷Fe Mössbauer measurements and DFT calculations, a {Fe(NO)₂}⁹(His)(Cys) coordination environment was proposed. Finally, investigation of the mechanism of {Fe(NO)₂}⁹ formation from {FeNO}⁷ revealed the presence of an additional S = 1 intermediate species assigned as {FeNO}⁸. Single-value decomposition of time-dependent absorption measurements shows this species is on-path to {Fe(NO)₂}⁹ formation, and researchers further propose this species to be {FeNO}⁸(His)(Glu)(Cys) coordinated based again on ⁵⁷Fe Mössbauer measurements and DFT calculations. This study not only provides insight into the stepwise nitrosylation of Fe in a protein scaffold and the resulting coordination environments but strongly illustrates the deep structural insight that can be obtained by applying a combination of multiple orthogonal spectroscopic and computational methods.

In addition to Az, reengineering of rubredoxin (Rd) has also allowed for successful {FeNO}⁷ formation. Rd is a low-molecular-weight electron transfer protein which natively binds Fe through the coordination of 4Cys (Figure 19). Substituting either Cys8 or Cys41 for Ala in Clostridium pasteurianum (Cp) Rd it has been shown to open a binding site, forming Fe(Cys)₃(OH) coordination.²⁷⁸ Using this approach, Lin and He were able to generate an open binding site for Fe^III in Rd from Pyrococcus furiosus (Pf) and subsequently bind NO following reduction with sodium dithionite (DT) to form an air-stable C9A {FeNO}⁷-Rb (and equivalently C41A {FeNO}⁷-Rb).²⁷² The dominant S = 3/2 species was characterized by EPR and absorption spectroscopies; interestingly, an additional S = 1/2 signal was also observed and assigned to a nonspecifically bound {FeNO}⁷ species. Irradiation with red light (625–650 nm) resulted in the stoichiometric release of NO from both C8A {FeNO}⁷-Rb and C41A {FeNO}⁷-Rb systems under either aerobic or anaerobic conditions. Under anaerobic conditions, photolysis of NO resulted in the Fe^II precursor, and further air exposure produced the starting Fe^III complex, which in turn was capable of binding NO to reform the {FeNO}⁷ complex. Meanwhile, photolysis of C8A {FeNO}⁷-Rb under aerobic conditions showed similar properties to the Fe^III starting material, but with a ~16 Da increased molecular weight. Based on this difference in molecular weight, it was hypothesized that oxygenation of either a binding Cys or Fe itself may occur. This species further decomposed to C8A Fe^III-Rd, which then could be used to also reform the {FeNO}⁷ complex. Further incubation with plasmid pBR322 DNA demonstrated photolysis of C8A {FeNO}⁷-Rb promoted DNA cleavage, demonstrating the practical biological applications of this system.

Figure 19.

Crystal structure of Pf Rd (PDB ID: 1VCX).²⁷⁷ Atom coloring: S (yellow); Fe (orange).

3.1.2. Design Using Unnatural Amino Acids.

One of the first examples utilizing an unnatural amino acid (UAA) to generate a biosynthetic non-heme iron binding pocket was accomplished by Schultz and co-workers, who genetically encoded E. coli to incorporate (2,2′-bipyridin-5-yl)alanine (Ala_Bpy) into the sequence of a DNA-binding protein).^279,280 Earlier studies have shown that incorporation of Fe^II-EDTA in DNA binding enzymes at the DNA interface induces oxidative cleavage at deoxyribose,^281,282 which has inspired efforts to directly bind Fe^II using bipyridyl as a high-affinity ligator. The homodimeric catabolite activator protein (CAP) was chosen as a scaffold due to its specificity for a 2-fold symmetric 22 base pair binding site. Inspired by the crystal structure of the CAP-DNA complex,^283,284 Lys26 was chosen for substitution to form the L26A_Bpy CAP variant (Figure 20a). Determination of the K_d of L26A_Bpy CAP vs WT-CAP for a 50 base pair DNA strand containing the 22 base pair binding site revealed similar affinities for both proteins, demonstrating that incorporation of Ala_Bpy did not significantly perturb the CAP-DNA interaction. Incubation of L26A_Bpy CAP with the same sequence in the presence of Fe^II, cyclic adenosine monophosphate (cAMP), and a reducing agent such as ascorbic acid or 3-mercaptopropionic acid resulted in DNA cleavage. Analysis of the products revealed cleavage primarily at either end of the recognition sequence, near the anticipated location of the Ala_Bpy. As cleavage was not observed in the absence of either Ala_Bpy or Fe^II (or equivalently Cu^II), it was inferred that the formation of an Fe^II-Bpy complex was necessary for activity. Continuing their work in DNA binding and regulation, Schultz and co-workers utilized the Zif268 transcription factor protein to design a novel Fe^II binding site, again using Ala_Bpy as a chelating ligand.²⁸⁵ Zif268 is a zinc finger protein, which uses Zn^II to stabilize the protein fold through Zn^II(Cys)₂(His)₂ coordination (Figure 20b). While Zif268 natively binds a specific nine base pair double-stranded DNA sequence, modification of the finger domain can lead to binding of both alternative sequences and other metals. Following successful incorporation of Ala_Bpy into Zif268 at three different positions (26, 28, and 29), it was found that all three mutants were capable of binding DNA with comparable affinities and with the same site specificity as WT-Zif268. Furthermore, substitution of Ala_Bpy for Ala_Bph (Bph = biphenyl) eliminated DNA binding affinity, as did the presence of EDTA, together demonstrating a requirement for both Ala_Bpy and a metal ion. Using a combination of ESI-MS along with absorption and EPR spectroscopies, it was elucidated that DNA binding by Zif268 Ala_Bpy is Fe^II-dependent. Comparison of the rR spectra from Fe^II-Zif268 Ala_Bpy and [Fe^II(bpy)(CN)₄] was indicative of Fe^II-bpy coordination. Further ⁵⁷Fe Mössbauer measurements revealed a single species with d = +0.41 mm/s and ∣ΔE_Q∣ = 0.34 mm/s, consistent with low-spin Fe^II. This result was further corroborated by temperature-dependent magnetic circular dichroism (VT-MCD) measurements, where the similarity of the low- and high-temperature spectra revealed a diamagnetic (S = 0) system. Together, these results support that incorporation of the UAA Ala_Bpy to Zif268 results in a similar fold to WT-Zif268, but with alternative metal selectivity.

Figure 20.

Crystal structures of target proteins for Ala_Bpy incorporation. (a) Catabolite activator protein (CAP) with the Lys26 target location of Ala_Bpy incorporation highlighted (PDB ID: 1J59).^280,284 (b) Transcription factor protein Zif268 with target residues for randomization highlighted. Native metal binding residues are highlighted in light gray, while additional target residues are highlighted in blue (PDB ID: 1ZAA).^288,285 (c) LmrR with target residues for Ala_Bpy incorporation highlighted (d) in light gray, with additional SCS target residues highlighted in light blue (PDB ID: 3F8B).^286,287 (e) Designed Ala_Bpy-incorporated protein scaffold based on indole-3-glycerolphosphate synthase containing bound Ni²⁺ (PDB ID: 4IXO) and (f) Co²⁺ (PDB ID: 4IWW). PCS residues are highlighted in light gray, while SCS target residues are highlighted in light blue.²⁸⁹ Atom coloring: N (blue); O (red); S (yellow); Fe (orange); Co (salmon); Ni (green).

Incorporation of UAAs has also been used to design metal binding sites for radical intermediate stabilization. Continuing from their previous work using a Ala_Bpy-incorporated lactococcal multidrug resistance regulator (M89X LmrR) protein (Figure 20c and d), Roelfes and co-workers probed the influence of varying the surrounding AA residues on the binding affinity for 3d transition metals ranging from Mn^II to Zn^II.^286,287 In order to bind Fe^II (as evidenced by the appearance of a 310 nm LMCT absorption band), additional substitutions of F93D and V15G were required, evidencing the need for available carboxyl ligands to induce Fe^II binding to bpy. Furthermore, Fe^II was found to bind and stabilize 3,5-di-tert-butylsemiquinone (DTB-SQ).

A significant challenge in designing competent metal catalysts in non-native scaffolds is understanding how each substitution will impact both structure and function. This problem can be even greater when employing unnatural amino acids, as most studies have focused around the use of natural AAs. Using the program Rosetta,^290,291 Baker and co-workers computationally designed a series of potential divalent metal binding sites utilizing Ala_Bpy, with the goal of forming a catalytic site mimicking catechol dioxygenase.²⁸⁹ Potential models were screened around a PCS involving Ala_Bpy and dopamine (a catechol). Genes for the best 13 candidates were synthesized and expressed in E. coli, of which five successfully produced soluble protein and two (CB_02 and CB_12) were capable of binding Fe^II. A successful crystallization of CB_02, a scaffold based on bacterial sialidase, revealed the Ala_Bpy side chain to be oriented outward toward solvent rather than aiding formation of a well-organized metal binding site. To overcome this, a second round of computational design was undertaken with a more constrained metal binding site geometry involving Ala_Bpy, Asp, Glu, and His, as well as two H₂O molecules replacing the catechol moiety. Of the 28 designs investigated, eight appeared to bind Fe^II in a buried fashion. Model MB_07 (based on indole-3-glycerolphosphate synthase) was successfully crystallized with both bound Co^II and Ni^II (Figure 20e and f) showing good homology with the respective computational designs. Although Fe^II-MB_07 could not be crystallized for direct comparisons, this pioneering work demonstrates the power of computational approaches in rationally designing PCS and SCS interactions to form novel metal binding sites.

3.1.3. Design via Covalent Interactions.

In addition to direct natural AA or UAA coordination, protein–metal binding sites can be generated via direct incorporation of an exogeneous organometallic complex into a host scaffold, commonly referred to as the “Trojan horse” method. Early applications of this approach were pioneered in the late 1990s by Distefano and co-workers by incorporation of a metal-binding 1,10-phenanthroline and pyridoxamine derivatives into adipocyte lipid binding protein.^292-295 This method is less commonly used to investigate the function of natural metalloenzymes and instead imparts altogether new catalytic activity. Here, the protein serves to impart SCS effects, such as controlling solvent access and hydrophilicity/hydrophobicity, charge screening, hydrogen bonding, enforcing conformers, and even imparting stability. Anchoring of these complexes can be accomplished either directly through covalent linking or via supramolecular interactions.

Maleimide reacts facilely with thiols via Michael addition to form thiosuccinimide under very mild basic conditions (pH ~ 7.5), providing an excellent route to forming a covalent linkage with Cys side chains (Scheme 2). Although use of this route requires careful consideration of cysteine accessibility, it offers a powerful tool for the incorporation of guest complexes into a wide range of protein scaffolds.

Scheme 2.

Banse and co-workers prepared one of the first examples of a mononuclear nonheme Fe site generated using a thiosuccini-mide anchor by incorporation of [Fe^II(L₅²_maleimide)Cl]⁺ into isoform B of bovine β-lactoglobulin (β-LG, Figure 21a) (L₅²_maleimide = 1-(2-{[2-(Bis-pyridin-2-ylmethylamino)ethyl]-pyridin-2-ylmethylamino}ethyl)pyrrole-2,5-dione)).²⁹⁶ The [Fe^II(L₅²_maleimide)Cl]⁺ complex was specifically chosen based on its reactivity with thioanisole and H₂O₂ to form phenyl-methylsulfoxide. Interestingly, incorporation into β-LG resulted in a 5-fold decrease in reactivity while simultaneously producing a 20% enantiomeric excess (ee). In the absence of thioanisole substrate, time-dependent absorption measurements revealed a transient signal at 660 nm maximizing around 4 min; further EPR measurements of the reaction mixture revealed this to be an S = 1/2 species, consistent with the formation of a low-spin Fe^III intermediate. Based on the energy of the 660 nm transition, this intermediate was further proposed to be Fe^III(H₂O₂) at room temperature, undergoing an acid/base equilibrium to form Fe^III(OOH) at low temperatures.

Figure 21.

Crystal structures of target proteins for Cys-maleimide linkages. (a) β-Lactoglobulin isoform B (β-LG) with the target Cys121 for thiosuccinimide formation highlighted in light gray (PDB ID: 4IB9).^296,297 (b) Steroid carrier protein SL (SCP-2L) containing an Ala100Cys mutation as the target for thiosuccinimide formation (highlighted in light gray). An additional target residue, Phe94, targeted for SCS modifications, is highlighted in light blue (PDB ID: 6Z1W).^298,299 (c) Xylanase A (Xln) with target residue for Ser212 (for the Ser212Cys incorporation highlighted in light gray, with additional SCS target residues highlighted in light blue) (PDB ID: 1E0W).^300,301 Atom coloring: O (red); S (yellow).

Further work by Jarvis and co-workers looked to utilize the oxidizing power of iron-peroxo to attack the β-O-4 linkage of lignin, a notoriously difficult to process material with important consequences as a potential renewable resource.²⁹⁹ A series of organic ligands were designed to enable both iron binding and covalent linkage via thiosuccinimide formation to the A100C variant of steroid carrier protein 2L (SCP-2L, Figure 21b). Of the explored organometallic complexes, only the maleimide-substituted tris(2-pyridylmethyl)amine exhibited activity to activate a lignin model complex via hydroxide to ketone conversion, which was improved by incorporation of the complex into SCP-2L. In an attempt to further stabilize the metal site (via either hydrogen bonding or direct dative bonding), F94E and F94H substitutions were introduced on the basis of MD simulations. While no effect was observed with histidine, glutamate resulted in improved conversion and yield. It is still uncertain whether direct coordination of Fe by Glu occurs, but it is possible the carboxylate side chain serves to stabilize the Fe center.

Counter to the pyridyl-based complexes used to impart oxygenase activity in studies with β-LG and SCP-2L, Mahy and co-workers combined the S212C variant of xylanase (Ser212Cys Xln, Figure 21c) with a maleimide Knölker’s Fe complex to generate a biohybrid with hydrogenation activity.³⁰¹ Using Rosetta,^290,291 the backbone perturbations arising from Michael addition of the maleimide-Knölker complex at S212C were modeled using a combination of MD and normal-mode analysis (NMA) which indicated some conformational rearrangement may be necessary for favorable binding to occur. However, successful incorporation was achieved, and hydrogenation of trifluoroacetophenone (TFAC) to its alcohol complement was boosted approximately 2.5-fold relative to Knölker’s complex in ethanol solution.

3.1.4. Design via Supramolecular Interactions.

While covalent attachment is one method of inserting a guest complex into a host scaffold, supramolecular interactions provide an equally powerful approach. Several systems have been employed to generate mononuclear non-heme Fe active sites via, for example, salt bridge, hydrogen bonding, and/or π-stacking interactions.

Ménage and co-workers have extensively employed the protein scaffold NikA (nickel-binding periplasmic protein, Figure 22a) to incorporate mononuclear non-heme Fe organometallic complexes to study the impact of the surrounding protein scaffold on oxygenation and hydroxylation activity. NikA natively serves as a Ni^II transporter protein which is essential for the maturation of a variety of microbial energy-converting enzymes.³⁰³ Interestingly, initial studies reinvestigating the structure of recombinantly expressed NikA resulted in the unexpected coordination of [Fe^IIIEDTA(H₂O)]⁻ (EDTA = ethylenediaminetetraacetic acid).³⁰⁴ Binding of this complex was proposed to be stabilized through several possible salt bridges, most importantly via Arg137 to the apical carboxylate of EDTA. Spurred on by these results, Ménage and co-workers generated an EDTA-inspired organometallic ligand L0 (L0 = N-benzyl-N′-(2-hydroxybenzyl)-N,N′-ethylenediaminediacetic acid, Scheme 3) as a mimic of Fe oxygenase and further incorporated the Fe^IIIL0 complex into the NikA scaffold to form Fe^IIIL0-NikA.³⁰² Crystallography revealed stabilization via the previously observed carboxylate-Arg137 salt bridge and further π-stacking between an equatorial phenol and Trp398 (Figure 22b). Exposure of Fe^IIIL-NikA crystals to dithiothreitol (DTT) induced a transformation of the equatorial ligand phenol → catechol and of benzyl → phenol, and further exposure to O₂ intramolecular resulted in dihydroxylation of the L0 scaffold. Further combination of crystallography and rR allowed the characterization of two O₂-bound intermediates, leading to the proposal that homolytic O₂ cleavage occurs to form an Fe^IV═O/Fe^III─O^• intermediate (Figure 22c-f).

Figure 22.

(a) Crystal structure of NikA with bound Fe^IIIL0. (b) Fe^IIIL0 binding site of NikA with interacting SCS residues Arg137 and Trp398 highlighted in blue. (c) Dihydroxylated product formed following exposure of NikA/FeL0 to 50 mM DTT and O₂. (d) O₂-bound intermediate formed following four successive back soaks of DTT-reduced NikA/FeL0 in DTT-free mother liquor, followed by O₂ exposure. (e) NikA/FeL0 soaked in 50 mM DTT. (f) Partially hydroxylated O₂-bound intermediate formed following three successive back soaks of DTT-reduced NikA/FeL0 in DTT-free mother liquor, followed by O₂ exposure. Respective PDB IDs for parts a—f: 3MVW, 3MVW, 3MW0, 3MVY, 3MVX, 3MVZ.³⁰² Atom coloring: N (blue); O (red); Fe (orange).

Scheme 3.

While the dioxygenase activity of Fe^IIIL0-NikA is promising, the dihydroxylated product results in coordination of Fe by the newly formed phenol, hindering any further turnover.³⁰² This inspired further efforts to deter the intramolecular reaction via two approaches. Under the first approach, substitutions on the benzyl and phenol groups of L0 inhibit the intramolecular reaction, while the second employed organic ligand was modified to provide N₂Py₂ coordination (N = substituted amine, Py = pyridyl).

To inhibit intramolecular hydroxylation, Cavazza, Ménage, and co-workers modified the origin L ligand by blocking the sites of aromatic hydroxylation observed previously.^302,305 To do so, a thiomethyl substituent was introduced at the para position of the benzyl group (complex L1, Scheme 3), and an additional methoxy substituent was placed at the meta position of the phenol (complex L2). Crystal structures were obtained for the Fe^IIILY-NikA and Fe^IILY-NikA systems (Y = 1, 2), as well as Fe^IILY-NikA following exposure to O₂ (FeLY-NikA-O₂). This structural characterization revealed that, in addition to the previously observed Arg137 salt bridge, H-bonding from Tyr402 and π-stacking from Trp398 provide important stabilizing interactions with Fe^IIILY-NikA. Upon reduction to form Fe^IILY-NikA, the conformational change resulting from rotation at the thiomethyl-substituted benzylic group results in substitution of H₂O for thioether coordination and the formation of a new salt bridge with Arg97. While O₂ exposure of Fe^IIL1-NikA resulted in intramolecular hydrogenation to form a catechol, Fe^IIL2-NikA appeared unimpacted. Through this process, products of both still preserve an open coordination site. Activity studies demonstrated some monooxygenase activity in FeLY-NikA, which was also greatly improved by cross-linking NikA to form enzymatic crystals (CLECs).³⁰⁵

To explore N₂Py₂ coordination, a series of Fe-EDTA derivatives FeLX (X = 3–7) were generated with pyridyl-substituted amines and a varying number of carboxylate groups, and their interactions with NikA were subsequently investigated (Scheme 3).³⁰⁸ The presence of a carboxylate moiety was found to be necessary for binding, and the affinity was significantly boosted by the presence of two carboxylates. Addition of steric hindrance by substitution of the 1,2-diaminoethane backbone for 1,2-diaminocyclohexane additionally decreased the binding affinity. Comparing complex charges, it was found that binding was also influenced by charge screening, where negatively charged complexes generally exhibited higher affinity. This effect was partially overcome in complexes that were capable of forming CH/π interactions. Further catalytic studies for these complexes were performed to probe their oxygenase activities toward sulfide complexes.³⁰⁹ Substrates were screened using molecular docking calculations and chosen on the basis of probing the presence/absence of H-bonding interactions with Glu247 and Gln385 residues. With a select substrate, the activities for sulfoxidation in the presence of NaOCl were monitored, demonstrating that the presence of the NikA host around FeLX-NikA is critical for catalysis. To further boost the catalytic capabilities of the NikA-hosted N₂Py₂ coordination complexes, CLECs of NikA were grown and substituted with FeLX-NikA (X = 8, 9) containing ortho methyl substituents on either pyridyl group.³¹⁰ The sulfoxidation activity was significantly improved in the FeLX-NikA/CLEC system relative to solution FeLX-NikA, promoting both higher yields and significantly greater turnover numbers (TONs). The harsh conditions used in the sulfoxidation reaction (including the presence of NaOCl as oxidant) were suggested to degrade the catalyst and protein during turnover, and the stability imparted by the CLEC likely mitigates this issue to some degree.

Ménage and co-workers have also explored sulfoxidation via incorporation of a Fe-N₂Py₂ complex in human serum albumin (HSA) by taking advantage of this protein’s ibuprofen recognition site.³⁰⁶ An ibuprofen-substituted N,N′-bis(2-pyridylmethyl)-N-methyl-1,2-ethanediamine complex (L_ibu) was used to coordinate Fe^IICl₂, forming the Fe^IIL_ibuCl₂ complex, which was further successfully incorporated into HSA as supported by ICP, absorption, and fluorescence measurements (Figure 23). While Fe^IIL_ibuCl₂ was found to completely oxidize thioanisole to sulfone in the presence of H₂O₂, product formation was limited predominately to sulfoxide in the Fe^IIL_ibu-HSA complex. Significant activity was also observed for HSA alone, although both selectivity and TON were improved by the presence of Fe^IIL_ibu.

Figure 23.

(a) Fe-chelating complex L_ibu.³⁰⁶ (b) Crystal structure of human serum albumin (HSA) with bound ibuprofen (abbreviated as “ibu”) highlighted in light gray (PDB ID: 2BXG).³⁰⁷ Atom coloring: N (blue); O (red).

In addition to NikA and HSA, the streptavidin (Sav)-biotin system has been effectively used to generate a wide range of artificial metalloenzymes, including several examples of mononuclear non-heme Fe sites. The first example incorporating a biotinylated Fe complex into Sav comes from Renaud, Ward, and co-workers, who utilized a series of modified Knölker’s complexes to promote the enantioselective hydrogenation activity (Figure 24).³¹¹ While most exhibited low catalytic activity, the use of an intermediate length linker was found to promote ketone hydrogenation with 26% conversion and 34% ee. Ward and co-workers have also incorporated a biotinylated Fe-TAML complex (Fe-TAML_biot, TAML = tetraamido macrocyclic ligand) into Sav to impart hydroxylation activity into the scaffold.³¹² While the Fe-TAML_biot exhibits similar activity independently of the Sav scaffold, incorporation of a S112R mutation resulted in >40% ee in the conversion of ethylbenzene to 1-phenylethanol, favoring the R enantiomer. Meanwhile, the L121R and S112R/L121E variants favored formation of the (S)-1-phenylethanol enantiomer. Structural characterization via crystallography of the S112R Sav and S112R/L121E Sav variants revealed salt bridge formation between Arg112 and the TAML ligand, forcing the Fe-TAML complex into a conformation allowing for Lys121 of the neighboring subunit (Lys121′) to bind Fe. In the S112R/L121E variant, the Arg112 side chain is turned away from TAML, with the side chain of Glu121′ coordinating Fe in an h² conformation. These changes in coordination are believed to impact the resulting product distribution and ee of these systems.

Figure 24.

(a) Fe-TAML_biot with a C₄ linker. (b) Fe-TAML_biot with a C₄ linker bound in Ser112Arg/Lys121Glu Sav. SCS mutations are highlighted in light blue (PDB ID: 6Y25).³¹¹ Atom coloring: N (blue); O (red); Fe (orange).

The Sav-biotin system has also been employed to simulate the classic facial triad present in many mononuclear non-heme Fe enzymes. To accomplish such a binding site, Borovik and co-workers designed a biotinylated bis(2-pyridylmethyl)amine (dpa) which was bound to a reengineered Sav scaffold with S112E, L121A, and E101Q mutations (2xm-S112E-Sav), each respectively intended to provide direct metal coordination, reduce sterics, and prevent exogeneous metal binding.³¹³ The crystal structure of Fe^IIdpa-(2xm-S112E-Sav) revealed an unusual binding mode in which only one of the two pyridines of dpa coordinates Fe^II, resulting in N₂O₃ coordination (Figure 25). Meanwhile, structural characterization of Fe^IIdpa-(2xm-S112E-Sav) showed facial coordination of all three N-donors of dpa, forming an N₃O₃ coordination set. In both Fe^II- and Fe^IIIdpa-(2xm-S112E-Sav), Glu121 serves to coordinate Fe in a k¹ fashion.

Figure 25.

(a) Fe^III-biot-et-dpa. (b) Crystal structure of Ser112Glu Sav-bound Fe^III-biot-et-dpa. Two configurations were observed (over-lapping, highlighted as light gray and blue) (PDB ID: 6UIY).³¹³ Atom coloring: N (blue,) O (red), Fe (orange).

3.2. Designed Dinuclear Non-heme Iron Enzymes

Catalytic nonheme diiron enzymes can be broadly separated into two categories based on the nature of their bridging moieties, namely, those involving μ-carboxylate and those involving μ-thiolate. The former represents a large family with a diverse array of functionality, including methane and toluene monooxygenases, ribonucleotide reductases, flavodiiron nitric oxide reductases, rubrerythrin, and stearyl-ACP desaturase, to name a few. Meanwhile, the primary example of a catalytic (μ-thiolato)diiron site is FeFe hydrogenase (FeFe H₂ase).

Despite the ubiquitous nature of natural (μ-carboxylato)-diiron enzymes, there are few examples of incorporating such an active site into an ArM due to the difficulty involved in tailoring a binding environment for not one but two metal centers. The de novo Due Ferri (DF) scaffold designed by Lombardi, DeGrado, and co-workers has proven to be extremely valuable for the generation and study of such active sites and is discussed in depth by Pecoraro and co-workers.^314-316 Recently, the Sav-biotin system was reengineered to form a unique (μ-carboxylato)diiron binding site, employing a combination of Trojan Horse and direct AA coordination approaches.³¹⁷ To do so, Borovik and co-workers took advantage of the quaternary structure of Sav, which positions two biotin-binding sites in close proximity and with an intermediary pocket. Utilizing variants of biot-et-dpa, a series of ligands (biot-x-dpa) with varying linker length (x = ethyl, propyl, and butyl) were cross-correlated with a series of Sav variants, namely, WT, S112Y, L121Y, and L121A/L124Y.³¹³ Tyr was chosen due to the intense CT absorption feature that arises from Fe^III-O_Tyr coordination,³¹⁸ which allowed for the design of a visual assay testing for Fe─O coordination. Successful binding was observed when employing a combination of the biot-bu-dpa ligand (bu = butyl) with the L121A/L124Y Sav variant. Using a combination of crystallography, ⁵⁷Fe Mössbauer spectroscopy, and DFT calculations, it was demonstrated that the newly formed site contained a [Fe^III-(μ-OH)-Fe^III] core with a labile proximal position at either Fe (Figure 26). Interestingly, an unusually long Fe─Fe distance of 3.96 Å and a small O_Tyr─Fe─O_μ-OH angle of 65° were observed, presumedly due to the strain imposed by the rigid Sav host, serving as an example of how deeply the SCS structure can impact the PCS.

Figure 26.

Fe^III-biot-bu-dpa bound in Lys121Ala/Leu124Tyr Sav forming a [Fe^III-(μ-OH)-Fe^III] core. PCS residues are highlighted as light gray, and SCS in light blue (PDB ID: 6VP1).³¹⁷ Atom coloring: N (blue); O (red); S (yellow); Fe (orange).

Although (μ-thiolato)diiron enzymes are found in a much smaller family of enzymes, the most prominent member, [FeFe] H₂ase, has been the subject of intense interest for several decades due to its ability to efficiently and reversibly catalyze the production of H₂. The catalytic H-cluster of [FeFe] H₂ase, [Fe₂(CO)₃(CN)₂(adt)] (adt = ⁻SCH₂NHCH₂S⁻), is unto itself an organometallic complex covalently tethered to a neighboring [Fe₄S₄] cluster via a cysteine-thiol bridge, making it an attractive target for incorporation in biosynthetic designs.

Given the H-cluster is a stable entity unto its own, one of the critical questions posed for the development of biologically inspired H₂ase catalysts is the extent of the SCS influence on H₂ production. As a result, many studies have utilized minimal oligopeptide maquettes to either form or bind H-cluster mimics. One of the first attempts was made by Jones et al., who incorporated a [Fe(CO)₃-(μ-S_Cys)₂-Fe(CO)₃] complex via a CXXC motif in a single-stranded 36-mer oligopeptide maquette.³¹⁹ Based on this work, Hayashi and co-workers provided one of the first examples attempting to incorporate an H-cluster mimic into a non-native protein scaffold.³²⁰ Inspired by the success of Jones et al., apo-cytochrome c (apo-Cc) was chosen due to its native CXXC motif (Figure 27a and b). Formation of a [Fe(CO)₃-(μ-S_Cys)₂-Fe(CO)₃] complex in apo-Cc to form H-apo-Cc was evidenced by a combination of ESI-TOF MS, absorption spectroscopy, and IR spectroscopy. Addition of the photosensitizer [Ru(bpy)₃]²⁺ and excess ascorbate to H-apo-Cc led to visible light-controlled H₂ formation, with a TON reaching ~80 after 3 h. The drop in efficiency observed with decreasing amounts of [Ru(bpy)₃]²⁺ inspired further investigation of an alternative scaffold that could prospectively also bind a photosensitizer close in proximity to the CXXC motif. Using a single-stranded α-helical octadecapeptide maquette (Pep18) based on the sequence of cytochrome c₅₅₆, a CXXCH motif was employed (Figure 27c and d).³²¹ Similar to the case of Cc, ESI-TOF MS, absorption spectroscopy, and IR spectroscopy were used to evidence both formation of [Fe(CO)₃-(μ-S_Cys)₂-Fe(CO)₃] and binding of [Ru(bpy)₃]²⁺ via His. Importantly, this maquette was able to evolve H₂ in the presence of light and excess ascorbate, which was otherwise not found without datively tethered [Ru(bpy)₃]²⁺.

Figure 27.

Summary of H₂ase-inspired (μ-thiolato)diiron enzymes. (a) Synthetic process for the formation of Cc-bound [Fe(CO)₃-(μ-S_Cys)₂-Fe(CO)₃] and (b) ensuing H₂ production system.³²⁰ Reproduced with permission from ref 320. Copyright 2011 Royal Society of Chemistry. (c) Synthetic process for the formation of Pep-18 [Fe(CO)₃-(μ-S_Cys)₂-Fe(CO)₃] with bound Ru(bpy)₃ and (d) ensuing H₂ production assay system.³²¹ Reproduced with permission from ref 321. Copyright 2012 Elsevier. (e) Model of the [Fe(CO)₃-(μ-S_Cys)₂-Fe(CO)₃] binding single-stranded α-helical maquette containing the dithiol UAA “Dt”.³²² Reproduced with permission from ref 322. Copyright 2012 Royal Society of Chemistry. (f) Proposed schematic for [Fe₄S₄]-[Fe₂(CO)₃(CN)₂(adt)] bridge formation in the FdM maquette.³²³ Reproduced with permission from ref 323. Copyright 2019 Royal Society of Chemistry. (g) Binding of a [Fe(CO)₃-(μ-S_Cys)₂-Fe(CO)₃]-maleimide complex in the β-barrel protein nitrobindin.³²⁴ Reproduced with permission from ref 324. Copyright 2014 American Chemical Society.

Alternative to the CXXC motif, Ghirlanda and co-workers explored the formation of a [Fe(CO)₃-(μ-S_Cys)₂-Fe(CO)₃] complex with a single-stranded α-helical maquette using a UAA containing a dithiol side chain (Figure 27e).³²² Interestingly, reaction of this system with [Ru(bpy)₃]²⁺, citrate buffer, and light produced a H₂ TON of 84 after 2.3 h, comparable to that observed by Sano et al. in their H-apo-cyt c system.³²⁰

Berggren and co-workers have recently expanded on the maquette-binding H-cluster model by design of a [Fe₄S₄]-binding oligopeptide (FdM) capable of binding [Fe₂(CO)₃(CN)₂(adt)] via a bridging Cys to generate a more comprehensive H-cluster structural model (Figure 27f).³²³ While [Fe₂(CO)₄(CN)₂(adt)] or [Fe₄S₄]-FdM was not observed to evolve H₂ as isolated in the presence of methyl viologen, the [Fe₂(CO)₃(CN)₂(adt)]-(μ-S_Cys-[Fe₄S₄]-FdM complex was capable of producing H₂ with a TON = 11 over the course of 1 h. Together with the maquette studies of Hayashi and Ghirlanda,^320-322 these results emphasize the importance of an electron transfer site in the proximity of the H-cluster.

Further studies have incorporated mimics of the H-cluster into larger scaffolds. Hayashi and co-workers have employed the thiosuccinimide-anchor method to generate a covalently bound H₂ase model using a modified nitrobindin (NB) β-barrel peptide scaffold with a Q96C substitution in combination with a maleimide-containing synthetic [Fe(CO)₃-(μ-S)₂-Fe(CO)₃] complex (Figure 27g).³²⁴ This combination yielded up to a TON of 130 over the course of 6 h in the presence of [Ru(bpy)₃]²⁺, ascorbate, and light. Interestingly, in the absence of NB this complex still is capable of reaching a TON ~ 130 over the course of just 2 h. In this case, it appears that the limitations of [Ru(bpy)₃]²⁺ access to the catalytic site imposed by NB hinder the reaction as opposed to promoting it.

Most recently, Ghirlanda and co-workers have employed the Sav-biotin system to generate a synthetic H₂ase using a biotinylated [Fe(CO)₃-(μ-S)₂-Fe(CO)₃] complex.³²⁵ Use of [Ru(bpy)₃]²⁺ to drive catalysis in the presence of visible light and ascorbate resulted in slow H₂ production, reaching ~48 TON over the course of 11 h, significantly greater than the TON = 6 found in the absence of the Sav scaffold, demonstrating that the presence of a surrounding protein scaffold significantly improves the catalyst integrity. This hypothesis was further supported by transient absorption measurements, which evidenced an increase in the lifetime of the catalytically competent intermediate in the presence of Sav.

4. ACTIVITY BEYOND THE FIRST COORDINATION SPHERE IN ENGINEERED COPPER ARMS

Like Fe, Cu serves as an important cofactor for a variety of protein functions, including electron transfer (ET) and catalysis.^326-332 These functions can be regulated by tuning both the PCS and SCS. Cu proteins can function as ET proteins using one of two types of copper centers, known as type 1 Cu (T1Cu) (Figure 28a) and copper A (Cu_A) (Figure 28b); structurally homologous proteins (or protein subunits) containing either type of the copper site are referred to as cupredoxins.^333-335 The T1Cu center is coordinated by one cysteine and two histidine residues in a trigonal plane; for many T1Cu cupredoxins, another interaction from the axial position above the trigonal plane can form a distorted tetrahedral geometry that is between the coordination geometries preferred by the Cu^II and Cu^I states, thus facilitating ET.³³⁶ Meanwhile, the Cu_A center is more complicated, involving a spatially close, electronically delocalized dicopper core bridged by two cysteines together with one histidine and one variable axial ligand on either side.^337,338 The geometry of Cu_A is more rigid than T1Cu with a comparably lower reorganization energy for the Cu^1.5+-Cu^1.5+/Cu²⁺-Cu²⁺ redox pair. Cu_A sites are often found in conjunction with metalloenzymes containing catalytically active sites, including heme-copper oxidase (HCO),³³⁹ nitrous oxide reductase (N₂OR),³⁴⁰ and non-heme nitric oxide reductase.³⁴¹ In HCO, the Cu_A center serves to transfer electrons from the periplasm to the buried heme-Cu_B site for the reduction of O₂. Meanwhile, the Cu_A site of N₂OR shuttles electrons to the copper Z (Cu_Z) site to catalyze the reduction of N₂O to N₂.³⁴² A Cu_A center with a unique PCS was recently identified in the protein PmoD.^343,344 Although this protein is related to two copper enzymes, ammonia monooxygenase (AMO) and particulate methane monooxygenase (pMMO), the function of this Cu_A center is not yet well understood.

Figure 28.

Active site structures of several representative natural copper binding proteins. (a) T1Cu from Pseudomonas aeruginosa (PDB ID: 4AZU).³⁴⁵ (b) Cu_A site of the cupredoxin domain from Thermus thermophilus cytochrome ba3 oxidase (PDB ID: 2CUA).³⁴⁶ (c) Substrate-bound (O₂) T3Cu site of laccase from Melanocarpus albomyces (PDB ID: 1GW0).³⁴⁷ (d) Substrate-bound (nitrite) T2Cu site of Alcaligenes xylosoxidans nitrite reductase (PDB ID: 2XWZ).^319,348 (e) Cu_z site of the Pseudomonas stutzeri N₂OR (PDB ID: 3SBR).³⁴⁹ Atom coloring: N (blue); C (cyan); O (red); S (yellow); Cu (copper).

While the T1Cu and Cu_A sites serve to transfer electrons, type 2 copper (T2Cu) is directly involved in biocatalysis as part of a catalytic active site.³⁵⁰ The ligand set of T2Cu is typically composed of nitrogen- and oxygen-based ligands and, in special cases, methionine as well. Unlike the trigonal or distorted tetrahedral geometry of the PCS of T1Cu sites, T2Cu is in a distorted square-planar geometry in the Cu^II resting state with one or more empty sites allowing for the binding of small molecular oxidants such as O₂ or superoxide and enzyme substrates (Figure 28d). T2Cu-containing enzymes are generally oxidoreductases with vastly different structures, including nitrite reductase (NiR),^351,352 SOD,³⁵³ and numerous O₂-utilizing enzymes known as oxygenases and oxidases.^354-356 Oxygenases function to incorporate an oxygen atom from molecular oxygen into the substrate. For example, lytic polysaccharide monooxygenase (LPMO) catalyzes the hydroxylation of the C─H bond in the glycosidic linkage using molecular O₂.^357,358 Meanwhile, oxidases use O₂ to reoxidize copper in the catalytic cycle without incorporation of O into the substrate. For example, laccase catalyzes the oxidative cross-linking of phenolic substrates while reducing O₂ to water.^359,360 For T2Cu enzymes, secondary coordination sphere (SCS) interactions function to not only tune the copper center but more importantly stabilize and/or activate substrates by hydrophobic interactions or hydrogen bonds or directly participate in the formation of reaction intermediates.

Beyond native functionality, the development of novel reactivity is an important perspective of artificial metalloenzymes. Copper complexes are important catalysts in organometallic chemistry, a topic covered by several in-depth reviews.^361-366 By combining a copper cofactor with a protein, one can take further advantage of the SCS environment imposed by the surrounding scaffold while maintaining the versatile chemistry of the organometallic copper compounds. There have been emerging examples using proteins or peptides as hosts for copper cofactors to achieve nonbiological reactivity, including Diels–Alder coupling,³⁶⁷ Friedel–Crafts alkylation,³⁶⁸ the Michael addition reaction,³⁶⁹ and atom transfer radical polymerization (ATRP).³⁷⁰ In many of these cases, the interactions between the SCS of the protein and the substrate molecules were shown to have a significant impact on the activity and selectivity.

The following section introduces a series of case studies on the SCS tuning effect of the two essential aspects of copper protein function—the electron transfer properties of T1Cu and Cu_A and the catalytic behaviors of T2Cu sites—in biosynthetic proteins or peptides. This section also covers nonbiological coppers catalysts including protein-fused synthetic copper complexes and the use of natural copper centers to catalyze nonbiological reactions. Due to their scarcity in nature and high complexity, other copper centers, such as type 3 copper (T3Cu)^371-373 and Cu_Z,³⁴⁹ are rarely studied using biosynthetic models and, therefore, are not discussed in this section. Despite their significance, the coverage of de novo designed copper proteins will be limited in this article, and we refer the reader to the review by Pecoraro and coauthors in this same issue for an in-depth discussion.

4.1. Designed Copper Proteins with Electron Transfer Functionality

The ET functions of T1Cu and Cu_A are essential for many biological processes requiring the transfer of electrons to active sites that are not close in contact with electron donors or acceptors. T1Cu can carry out efficient ET across long (>10 Å) distances, which is rationalized by the optimization of the driving force (ΔG) and reorganization energy (λ) according to Marcus theory.^333,374 Cu_A centers have been reported to perform even faster electron transfer than T1Cu in a comparative study.^375,376 Therefore, tuning the driving force—the reduction potential of copper (E°′)—is essential to understand electron transfer processes in nature and to build better systems with improved ET properties.

To avoid the complexity of many native enzymes, the electron transfer properties of T1Cu have been studied using simple cupredoxins such as Az.^378-381 The E°′ values of the native T1Cu proteins typically exhibit a wide (500 mV) range despite their similar structures.^382-384 Three main factors are considered responsible for such a large variarion in E°′, namely, the identity of the axial ligand, hydrophobicity, and H-bonding (Figure 29). Axial ligation is not conserved among T1Cu proteins, taking varying identities such as Met in Az, Gln in stellacyanin, and a variety of hydrophobic residues in fungal laccases, etc.³⁸⁵ To exclude other structural variables, the relationship between the axial ligand and E°′ has been demonstrated in Az, in which the native Met has been mutated into nearly every other natural amino acid, as well as a series of unnatural amino acids.^386-391 A positive correlation was found between the hydrophobicity of the axial ligand and the E°′ value of the T1Cu copper site, yet the range of observed potentials (150 mV) was not nearly as high as that observed between natural T1Cu cupredoxins. The influence of hydrophobicity in the T1Cu site SCS environment was examined in another study in which substitution of three more distant residues (Leu33, Met44, and Leu86) with Phe resulted in an increase of the reduction potential by 80 mV, which is additive to the tuning effect established by varying the axial residue.³⁹² Additionally, Lu and co-workers have demonstrated that the hydrogen bonding network may play a more subtle but equally important role.³⁷⁷ The F114P mutation disrupts H-bonding between a backbone amide nitrogen and the S(Cys), increasing S-Cu covalency and causing a 110 mV decrease in E°′. The N47S mutation further alters the H-bonding interaction with Cys112, leading to a 140 mV increase in E°′ without significantly changing the PCS. The effects of H-bonding are additive to those from axial ligand modifications, producing a cumulative tuning range of ~800 mV.³⁷⁷ SCS mutations also play a role in tuning the reorganization energy of the copper center. Interestingly, most of the mutants investigated were found to have lower reorganization energies and, hence, higher intra-molecular electron transfer rates than WT-Az, with the N47S/F114N Az mutant exhibiting the fastest electron transfer rate.³⁹³ A following study of electron transfer in these mutants suggested a Marcus-inverted trend in which a higher reduction potential may lead to slower electron transfer,³⁹⁴ therefore supporting the hypothesis that maximizing the rate of electron transfer requires optimizing the reduction potential.

Figure 29.

Three SCS targets of reduction potential tuning in Cu-Az: the axial residue hydrophobicity, hydrogen bonding network, and distal residue hydrophobicity. The plot shows the cumulative effect when different strategies are combined (PDB ID: 4AZU).³⁴⁵ Reprinted with permission from ref 377. Copyright 2011 Elsevier.

Besides homology-based biosynthetic modeling, de novo design emerged in the 1990s as a powerful alternative approach to building metal binding sites in simple protein scaffolds.^397,398 Hildebrandt and co-workers reported a series of designed four-helix bundle proteins with Cys and His residues in the core of the protein. The position of these residues was shuffled throughout helices, among which 30 candidates displayed copper binding affinity.³⁹⁹ The design was improved in a subsequent study by optimization of SCS interactions. By varying core residues with different hydrophobicity and bulkiness, the copper binding site was shown to be between T1Cu and T2Cu based on electronic absorption spectroscopy (color) and EPR (A_∥).⁴⁰⁰ A more recent report by Shiga et al. showcased an exact blue T1Cu center in a four-helix protein (Figure 30a).^395,401 The reduction potential of copper was determined to be 328 mV, close to the reduction potential of Az (~280 mV). The Ala residues in the SCS were proposed to be replaceable, allowing for enhanced packing of the core and, thereby, tuning of the configuration of the T1Cu, although this is yet to be experimentally verified. Quaranta, Pecoraro, and co-workers reported a three-helix T1Cu protein with spectral features resembling the red copper nitrosocyanin.^402,403 Recently, the electronic properties of red copper were found to depend on Glu41 in the SCS that lengthens both the Cu─S(Cys) and the Cu─N(His) bonds (Figure 30b).^396,404 This residue was also found to tune the T1Cu reduction potential, as a E41D substitution was shown to increase the reduction potential by 50 mV. Two additional designs were made from the initial red copper construct to mimic the T1Cu center of enzymes such as plastocyanin and nitrite reductase, which are blue and green copper, respectively. The primary difference between these two proteins is the absence or presence of a SCS axial methionine.⁴⁰⁴ Interestingly, the reduction potentials of the three mutants did not vary significantly, which opened up opportunities to further study the influence of the SCS in these three helical proteins to tune reduction potential.

Figure 30.

(a) De novo designed T1Cu protein by Shiga et al. with SCS Ala residues. (b) De novo designed T1Cu protein by Koebke et al. with an axial Glu residue. Reprinted with permission from refs 395 and 396. Copyright 2010 and 2018 American Chemical Society.

The necessity of redox potential tuning has been demonstrated in several T1Cu-containing enzymes, where a fine-tuning of the T1Cu potential can significantly affect the enzymatic activity. Multicopper oxidases (MCOs) are a class of enzymes that catalyze the oxidation of organic substrates while reducing O₂ to H₂O, such as ascorbic oxidase, bilirubin oxidase, copper efflux oxidase, and laccase (Figure 31a).^326,354 Despite the different structures and functionalities, all MCOs require efficient electron transfer from T1Cu to the active site, which is directly dependent on the redox potential of the T1Cu.^410-412 The strategies of redox potential tuning in single T1Cu protein have been applied in MCOs.⁴¹³ Mutation of the axial T1Cu ligand from Met to either Leu or Phe in Bacillus subtilis laccase resulted in a 100 mV increase in reduction potential, which, in turn, decreased the laccase activity by 2–4-fold in the Leu mutant, with a near complete loss of activity in the Phe mutant (Figure 31b).⁴¹⁴ This trend was corroborated by another study in which mutation of the native axial Phe of Thapsia villosa laccase to Met lowered the T1Cu potential by 100 mV.³⁸⁴ This mutation increased k_cat and K_M while shifting the optimal pH to a more basic regime. Kamitaka et al. reported a M467Q mutant of Myrothecium verrucaria bilirubin oxidase which lowered the formal potential of T1Cu by 230 mV while improving the kinetics of the electrocatalytic oxygen reduction reaction (Figure 31c).⁴¹⁵ Similar to the case of cupredoxins, introduction of an additional hydrogen bond donor can pose a significant impact on the T1Cu potential in MCO. Kakaota et al. studied the effect of the SCS hydrogen bonds of the T1Cu in E. coli copper efflux oxidase (CueO).^416,417 The conserved Pro444 was mutated to Ala, Leu, or Ile to introduce a second hydrogen bond to Cu-bound S(Cys) (Figure 31d). Similar to its Az counterpart, this modification decreased the covalency of the S(Cys)─Cu bond while increasing the T1Cu reduction potential by 40–70 mV. As a result, k_cat was observed to decrease by 40–60% among these mutants. Meanwhile, the P444G mutant showed higher activity than the native enzyme; however, the T1Cu reduction potential in this variant was not reported.

Figure 31.

(a) Structure of wild-type laccase from B. subtilis (PDB ID: 1GSK). The black arrow shows the electron transfer from the T1Cu to the T2/T3Cu site to reduce O₂. (b) Axial mutants of B. subtilis laccase modeled from the wild-type (PDB ID: 1GSK)⁴⁰⁵ using UCSF Chimera.⁴⁰⁶ (c) M467Q axial mutant of bilirubin oxidase from M. verrucaria (PDB ID: 6IQX).⁴⁰⁷ (d) Modifying Pro444, the H-bond donor to the Cys residue of T1Cu in E. coli CueO (PDB ID: 1KV7).⁴⁰⁸ (e) Modifying Glu506, a proton transfer mediator residue in E. coli CueO. Reprinted with permission from ref 409. Copyright 2012 Elsevier.

Another study by the same group focused on proton transfer in the SCS between the T1Cu and T2/T3Cu sites in E. coli CueO.⁴⁰⁹ Mutation of the proton donor Glu506 to Ile resulted in complete loss of activity, whereas a less bulky Ala mutation led to a compensated proton transfer pathway bridged by two water molecules. The E506N mutant also showed no activity, demonstrating the importance of a dissociable proton. The correlation between the T1Cu redox potential and the MCO activity has been previously reviewed⁴¹¹ and recently demonstrated in a comparative study by Solomon and co-workers.⁴¹⁸ The authors compared the high-potential fungal laccase of Trametes versicolor (axial Phe) with the low-potential plant laccase (axial Met) from Rhus vernicifera. The higher potential was found to result in an 8 kcal/mol decrease in the intramolecular electron transfer driving force. While electron transfer from T1Cu to the trinuclear Cu active site is the rate-limiting step of fungal laccase, this slow electron transfer is partially offset by the higher reduction potential of the trinuclear center (TNC, the T2Cu and T3Cu sites) due to a more positive protein environment and is still faster than the decay of the native intermediate during catalytic turnover.

In addition to laccases, the tuning of T1Cu and correlation with enzymatic activity have been studied in copper-dependent nitrite reductase (CuNiR), a more complex enzyme that catalyzes the reduction of nitrite (NO₂⁻) to nitric oxide (NO), a process essential for denitrification as part of the nitrogen cycle.⁴²² The reduction of NO₂⁻ at the T2Cu of CuNiR requires the injection of one electron and two protons, the former of which is mediated by T1Cu through a Cys-His bridge (Figure 32a).^348,423 A subtle balance exists between the two copper centers that their reduction potentials are only slightly different to prevent premature reduction of T2Cu in the rest state.⁴²⁴ Inspired by the studies on individual cupredoxins, the axial ligands of T1Cu in CuNiR were subjected to mutagenesis. The M144A mutation of the Alcaligenes xylosoxidans NiR (AxNir) profoundly increased the reduction potential of the T1Cu from 255 mV to 314 mV and, therefore, reversed electron transfer from the low-potential T2Cu (230 mV) (Figure 32b).⁴²¹ An even greater increase of the T1Cu potential was observed in the more hydrophobic M144L mutant (336 mV), whereas incorporation of the hydrophilic M144N mutation decreased the reduction potential to 173 mV.⁴²⁰ Interestingly, the M144L mutants showed higher NO₂⁻ reduction activity than the native enzyme while the Gln mutants exhibited no activity, suggesting that an artificially tuned, moderately uphill electron transfer is beneficial to the CuNiR activity (Figure 32c). In another study by Sato et al. the T1Cu binding loop of NiR was replaced with that from amicyanin, which maintains the PCS ligand sets but with a much shorter length (nine residues) (Figure 32d and e).⁴¹⁹ The swap caused a significant increase of 200 mV in the T1Cu reduction potential together with diminishing NiR activity due to the excessively negative driving force of electron transfer. This effect does not appear to corroborate with the trend observed between the decreased hydrophobicity and redox potential, as indicated by the coordinating water molecule, suggesting a profound but unknown influence from the SCS of the binding loop.

Figure 32.

(a) Structure of the T1Cu (right, cyan) and the T2Cu (left, tan) of AxNiR (PDB ID: 2XWZ).³⁴⁸ (b and c) Axial residue engineering of the T1Cu in AxNiR. (d and e) NiR with the T1Cu loop exchanged with the loop from amicyanin. Reprinted with permission from refs 419-421. Copyright 2008, 2005, and 2002 Wiley-VCH, Elsevier, and Elsevier.

Likewise, Cu_A has been studied in both native cupredoxins and biosynthetic models using similar tuning strategies as employed to study the electron transfer properties of T1Cu. For example, mutation of the axial Met227 to Ile in Paracoccus denitrificans cytochrome c oxidase (CcO), was shown to result in the loss of electron delocalization in the Cu-Cu core and impair electron transfer to heme a.⁴²⁹ A study of Rhodobacter sphaeroides CcO showed that the M263L mutation maintained the electron delocalization between the copper atoms but caused a 120 mV increase in the reduction potential and, correspondingly, a 2.5-fold decrease in the electron transfer rate and 90% diminishment of the CcO enzyme activity.^430,431 Other studies have investigated Cu_A by truncating native proteins, effectively separating the electron transfer and catalytic components.³⁴⁶ In a series of studies on the Cu_A domain of Thermus thermophilus cytochrome ba3 oxidase (TtCu_A), the weakly bound axial Met160 was mutated to Glu and Gln (Figure 33b).⁴³² In both cases, the mutated axial residues showed strong interactions with Cu possessing features similar to the PCS residues. Early EPR studies suggested an elongated Cu─Cu distance and change in geometry of the Cu₂S₂ core,⁴³³ while recent results from X-ray spectroscopy indicated that the Cu─Cu distance remains mostly unchanged in the M160Q mutant.⁴²⁸ The M160Q mutation resulted in a change of redox potential by −170 mV at neutral pH, with a Cu─Cu distance nearly identical to that of the wild-type (2.45 Å). An even greater difference of −190 mV was observed in another M160H mutant. Despite similar reduction potentials, the electronic structures of M160H and M160Q were found to differ significantly. The M160Q mutant has a significantly higher energy gap between the two accessible ground states π_u and σ_u* (900 cm⁻¹, pH = 6.0) than the wild-type (600 cm⁻¹) and the M160H mutant (200 cm⁻¹) (Figure 33f). A broader set of axial mutations were studied by Ledesma et al. with M160S, M160L, and M160Y mutations.⁴³⁴ The hydrophobic M160Y and M160L mutants exhibited similarly high reduction potentials (348 and 346 mV, +55 and +53 mV vs WT), while the hydrophilic M160S mutant exhibited lower reduction potentials (209 mV, −84 mV vs WT).⁴³⁴ In a study by Blackburn et al., the axial Met was replaced by Met_Se. The electrochemical properties were mostly identical between Met and Met_Se, and the EXAFS data indicated very similar geometries at the Cu_A core.⁴³⁵

Figure 33.

(a) Schematic of the O₂ reduction reaction catalyzed by heme-copper oxidase (cytochrome ba3 oxidase of T. thermophilus, PDB ID: 1XME).⁴²⁵ The electron transfer from Cu_A through heme b to heme a3 and the proton pumping pathway are shown with black arrows. (b) Mutation of the axial Met of the Cu_A from T. thermophilus cytochrome ba3 oxidase (PDB ID: 2CUA).^346,426 (c) Loop swapping strategy to tune the Cu_A center.⁴²⁷ (d) Hydrogen bonding network surrounding the artificial Cu_A site in engineered azurin. (e) Mutation of the axial Met of the artificial Cu_A in engineered azurin (PDB ID: 1CC3).⁴²⁶ (f) Tuning of the π_u and <_u* energy gap by axial ligand mutation of T. thermophilus Cu_A cupredoxin.⁴²⁸ Parts c and f were reprinted with permission from ref 428. Copyright 2019 American Chemical Society.

Cu_A centers have also been introduced into the homologous cupredoxins to explore the influence of the SCS. Biosynthetic Cu_A centers have been built in amicyanin^436,437 and Az^438,439 by replacing the original Cu binding loop with a Cu_A binding loop. Similar to studies of T1Cu, the axial Met has been mutated to hydrophilic Asp and Glu and hydrophobic Leu residues, altering the Cu_A redox potential by +16, −5, and −8 eV (Figure 33e).⁴⁴⁰ To modulate the H-bonding interaction with Cys112, N47S and E114P mutations were made in Cu_A azurin (Cu_AAz) (Figure 33d), leading to + 30 and −42 mV changes in the Cu_A reduction potential, which are modest in comparison to the + 120 and −46 mV differences seen in T1Cu.⁴⁴¹ Both strategies resulted in a smaller tuning effect compared to the T1Cu counterpart, which was attributed to either the higher robustness of the Cu_A diamond core or the fact that these alterations produced more subtle changes in the coordination environment, such as the possible presence of a water ligand in M123L Cu_AAz.⁴⁴²

The loop replacement strategy has also been applied in the native Cu_A cupredoxins, where the use of different loops allows for the PCS of Cu_A to be kept consistent while modifying the SCS. The three loops of TtCu_A, referred to as the His loop, entry loop, and ligand loop, were replaced by the loops from the native Cu_A subunits of human or plant CcO (Figure 33c). A moderate ±30 mV tuning of the reduction potential was observed in the loop-exchanged mutants compared to wild-type TtCu_A.⁴²⁷ Other electronic properties, including the ground state (π_u/σ_u*) energy and delocalized spin density on axial Met, were also tuned by the loop replacements.^443,444

In addition to cupredoxins, Cu_A has been constructed in both natural nonhomologous proteins and de novo designed proteins. Mirts et al. designed a Cu_A site in a cytochrome c peroxidase (CcP) scaffold (Figure 34).⁴⁴⁶ The new CuA center resides in the buried α-helical pocket where the heme is docked in the unmodified CcP. The reduction potential of the Cu_A center was found to be 251 mV as measured by redox titration, comparable to the case of Cu_A in native CcO (240 mV). Electron transfer between this Cu_ACcP protein and cytochrome c was observed, albeit at a significantly lower rate compared to the case of native Cu_A (10⁻³ s⁻¹ vs 10⁵ s⁻¹). Cu_A has also been incorporated into de novo designed helical bundles by Hildebrandt and co-workers, as evidenced by EPR, MCD, and UV–vis absorption spectroscopies.⁴⁰⁰ A later study by Tanaka and co-workers reported another example of Cu_A in a coiled-coil protein in a rationally designed Cu₂(Cys)₂(His)₂ core, exhibiting a Cu─Cu distance of 2.55 Å, longer than in native Cu_A.⁴⁴⁵ Unfortunately, no electrochemical characterization was carried out in either of these studies, and the ET activities of these constructs remain unknown. Studies using nonhomologous scaffolds have shown that the Cu_A structure can be well mimicked by transplanting the PCS ligand set. While full functionality is dependent on outer sphere interactions, it can be partially restored in these biosynthetic models.

Figure 34.

(a) Cu_A site of TtCu_A. (b) Original heme pocket of CcP. The three residues Trp51, His52, and Arg48 were engineered as the PCS core for Cu_A. (c) De novo designed Cu_A protein and (d) Cu_A structure. Reprinted with permission from references 445 and 446. Copyright 2012 and 2020 American Chemical Society.

4.2. Designed T2Cu Proteins with Catalytic Functions

Compared to the electron transfer copper proteins, modeling the catalytic sites of T2Cu proteins requires more detailed consideration. T2Cu proteins are, in general, more complex due to the numerous auxiliary units required to promote native catalytic activity. It is also difficult to achieve functional interactions with the substrates, as well as to stabilize redox-active intermediates during enzymatic turnover, roles which are fulfilled by well-positioned amino acid residues in the native protein. Three approaches, de novo design,^83,447-449 incorporation of T2Cu into natural proteins by coordination with amino acid side chains, and bioconjugation of organometallic complexes, have been applied to mimic the active site of catalytic Cu enzymes, and each of these approaches has provided insights into the role of SCS interactions.

The functionalities of several Cu enzymes, such as catechol oxidase (COx),⁴⁵¹ superoxide dismutase (SOD),⁴⁵² and nitrite reductase (NiR),⁴⁵³ have been applied in artificial protein models, among which models of NiR are the most well studied with respect to probing the influence of the SCS. The first functional model of the catalytic T2Cu was reported by Pecoraro and co-workers.^450,453 A tri-His copper binding site was introduced into the core of the three-helical peptide TRI-W. The NiR activity was observed by generation of N₂O without side reactions, albeit with lower activity than that of native NiR (Figure 35a). The activity was improved by tuning the SCS through electrostatic interactions, for which several solvent-exposed Lys and Glu residues were mutated to their counterparts with opposite charge.⁴⁵⁴ A 4-fold increase in the activity was observed between the most positively charged E27K and the most negatively charged K24E mutants. The reduction potential also decreased with increasing negative charge in the SCS by a maximum of 100 mV. Meanwhile, the influence of the steric hindrance was tuned by mutating the Leu residues below or above the active site into Ala or Asp.⁴⁵⁵ All variants showed elevated activity, with an increase in reaction rate up to 75-fold. These changes were attributed to an interplay between the coordination geometry of the Cu^I state and the electronic structure of the transition states. Recently, similar designs have been applied to construct Cu-bound α-helical peptides for other activities, including superoxide dismutation⁴⁵² and peroxide activation,⁴⁵⁶ with the effects beyond the PCS unexplored but promising for future studies.

Figure 35.

(a) De novo designed NiR peptide and the active site structure. Reprinted with permission from ref 450. Copyright 2018 American Chemical Society. (b and c) Structures of two AzNiR mutants modeled by UCSF Chimera using PDB ID 4AZU.^345,406

In addition to de novo design, strategies employing natural protein scaffolds have also been reported for NiR. Berry and co-workers have incorporated a His₃ or a His₂Asp site on the surface of Az (AzNiR), producing a system that includes both the T1Cu and T2Cu sites of CuNiR (Figure 35b and c).⁴⁵⁷ Both UV–vis and EPR spectra showed a new T2Cu species in addition to the native T1Cu. The incorporated T2Cu is solvent-exposed, leading to a moderately high potential (350 mV), which is between those of the previous model and native NiR, which are both hydrophobically buried. The NiR reaction rate (0.34–0.59 min⁻¹) in this Az model is lower than that of the native NiR yet comparable with those of the previously mentioned de novo designed peptides. The native T1Cu site serves as an optional electron donor which may participate in NiR catalysis directly, although the T2Cu was also found to independently catalyze nitrite reduction when the T1 site is blocked by Hg²⁺.

Similar cupredoxin scaffolds have been adopted to model catalytic Cu sites other than NiR. In Az, the substitution of the chromogenic Cys residue with a hard Asp ligand resulted in a typical (His)₂Asp T2Cu center, with large A_∥ as measured by EPR and a weak absorption feature at 600 nm.^460,461 Surprisingly, it has been found in a more recent study that further mutation of the axial Met residue to a hydrophobic, noncoordinating Leu, Phe, or Ile residue results in a new type of copper center that resembles neither T1Cu nor T2Cu.⁴⁵⁸ The new copper center, called “Type 0 Cu” (T0Cu, Figure 36), showed a weak d─d transition band at 800 nm but no absorption in the visible region. The A_∥ splitting falls into the low 300 MHz range, which is between the values from T1 (<200 MHz) and T2 Cu (>400 MHz). The reduction potential of T0Cu is higher than that of T2Cu and closer to that of the native T1Cu. The authors mutated the axial ligands to Glu in a following study, which reverted the spectroscopic features back to T2Cu with a comparably high potential (300 mV).⁴⁵⁹ The axial Glu exerts only an electrostatic effect and does not bind directly to copper at lower pH. At higher pH (9.0), the H-bonding network in the SCS caused a structural rearrangement of the T2Cu site as indicated by XANES and X-ray crystallography, leading to binding of the deprotonated Glu to Cu and a down-tuning of the reduction potential to 120 mV. Both modified T0Cu and T2Cu possess promising features, such as the low coordination number, high reduction potential, and flexible coordination pattern, for mimicking the catalytic T2Cu sites in nature.⁴⁶²

Figure 36.

(a) C112D Az (PDB ID: 3FQY). (b) C112D/M121L Az (PDB ID: 3FPY).⁴⁵⁸ (c) C112D/M121E Az at neutral pH. The SCS residues enforcing the binding configuration of the T0Cu site are circled in red. (d) C112D/M121E Az crystallized at basic pH. Reprinted with permission from ref 459. Copyright 2010 American Chemical Society.

Starting from the same Az model, the Lu group turned to investigate Cu-mediated post-translational modifications of S(Cys). A report in 2014 demonstrated that the M121G mutation decreased the hydrophobicity of the T1Cu pocket while opening an axial binding site (Figure 37a).⁴⁶³ Reduction to Cu^I enabled the interaction of the T1Cu site with hydrogen peroxide to yield a species with an absorption maximum at 400 nm that decays quickly in air to form the starting Cu^II-M121G Az. Further investigation of this intermediate by resonance Raman spectroscopy revealed an S-O vibration which could be attributed to the formation of sulfenic acid (Cys-SOH) formed at the Cys112 as further supported by proteomic analysis. The formation of sulfenic acid was attributed to the Cu-activated H₂O₂ based on ¹⁸O-isotope labeling studies. A computational study suggested the SCS served to stabilize the sulfenate by removing a negatively oriented dipole near Cu. This study was one of the first to demonstrate metal-sulfenate coordination in an artificial metalloenzyme system, whose natural presence was considered essential for the activity of nitrile hydratase and thiocyanate hydrolase. Although this process was accomplished using a different metal cofactor, this Az model demonstrated a plausible reaction mechanism for this post-translational modification in nature. Furthermore, the Lu group investigated the intramolecular nitrosylation of the Cys residue to form nitrosothiol (Cys-SNO), which serves as one of the most important biological reactions in NO regulation.²⁷⁶ To accomplish this, a H46E/M121H Az mutant was created to mimic the T1Cu of nitrosocyanin (Figure 37b). An additional F114P SCS mutation was introduced to eliminate H-bonding with S(Cys) and decrease the Cu potential. Under anoxic conditions, the T1Cu of the triple mutant catalyzed the formation of a Cu-Cys-SNO species on the ligating Cys residue, while a facile and complete release of NO could be achieved in vacuum or air. The Cu-mediated nitrosylation reaction displayed rapid kinetics (k = 6.7 × 10⁵ M⁻¹ s⁻¹), high yields (>90%), and efficient scavenging of NO at micromolar concentration.

Figure 37.

(a) M121G azurin and the oxidative modification (PDB ID: 4MFH).¹⁰³ The first step could only be achieved by H₂O₂ supported by the experiment with isotope labeled agents. The sulfenate could no longer be oxidized by H₂O₂, but further oxidation by O₂ from air was observed. (b) Active site structure of H46E/F114P Cu^II-Az. (c–e) EXAFS results for (c) M121H/H46E/F114P Cu^II-Az, (d) M121H/H46E/F114P Cu^I-Az, and (e) NO-treated M121H/H46E/F114P Cu^II-Az. Reprinted with permission from ref 276. Copyright 2016 Springer Nature.

Ward and co-workers have also designed a copper peroxidase using 6-phosphogluconolactonase, a natural hydrolase scaffold, that has a potential (His)₂Asp copper binding motif discovered by the STAMPS algorithm (Figure 38).⁴⁶⁴ The protein was expressed and cocrystallized with CuSO₄, showing that the two predicted His residues bound Cu as anticipated, although the third ligand was found to be H₂O stabilized by H-bonding with a side chain carbonyl moiety rather than the anticipated Asp131 residue. The peroxidase activity was found through the Cu-catalyzed oxidation of o-dianisidine using tert-butyl-hydroperoxide (t-BuOOH) as oxidant, with a maximum turnover number of ~25 achieved over the course of 1 h. Computational docking simulations revealed a possible hydrogen bond between the substrate and residues Asp131 and Tyr69, leading to the arrangement of the aromatic rings in-plane with the copper ion. A mutagenesis study of these two residues did not result in a significant change of catalytic efficiency when varied, although one mutant, Y69L, showed a 3-fold increase in turnover frequency.

Figure 38.

(a) Copper binding form of engineered 6-phosphogluconolactonase (PDB ID: 4TM8).⁴⁶⁴ (b) Peroxidation of o-dianisidine catalyzed by the Cu enzyme and the computationally docked substrate in the active site. The Asp131 and Tyr69 residues were believed to align the substrate. Reprinted with permission from ref 464. Copyright 2015 Royal Society of Chemistry.

In addition to natural proteins, Borovik and co-workers have used the Sav scaffold together with a biotinylated CuN₃ synthetic complex to study Cu-H₂O₂ interactions (Figure 39),⁴⁶⁵ a similar strategy previously adopted to create a T1Cu in the same scaffold.⁴⁶⁶ The protein/copper complex was used to model lytic polysaccharide monooxygenase (LPMO), an enzyme with a special Cu-His-brace active site to catalyze the oxidative degradation of long-chain polysaccharides.^358,467 The SCS of LPMO is known to stabilize the transient Cu-hydroperoxo (Cu-OOH) complexes but eventually lead to the cleavage of the O─O bond and the oxidation of the substrate.^468,469 Such interaction was mimicked by this Sav-Cu model but resulted in a room-temperature-stable Cu-OOH complex. Upon addition of hydrogen peroxide, an intense shoulder around 380 nm was observed and attributed as an ⁻OOH → Cu LMCT transition. The Sav-Cu protein crystal was soaked in H₂O₂-containing buffer, and the structure was determined with X-ray crystallography. The structure showed an electron density of two oxygen atoms on the axial site of copper, which was assigned as the axial hydroperoxo ligand. A hydrogen bond network was depicted in the crystal structure. The distal oxygen of the hydroperoxo anion formed a hydrogen bond with the amide nitrogen of the Asn49 residue. The proximal oxygen was involved in a more complicated hydrogen bond network consisting of Ala86, Ser112, Lys121, and two water molecules. The latter interaction was proposed to be important for the stability of the Cu-OOH complex as it offsets the push–pull effect from the distal hydrogen bond and avoids breaking the O─O bond, similar to the His-O₂ interaction in myoglobin.⁴⁷⁰ The study provided key insight into how a well-positioned hydrogen bond alters the stabilization or activation of a Cu-oxygen adduct.

Figure 39.

(a) Crystal structure of the H₂O₂-bound Sav-Cu protein (PDB ID: 6ANX). (b) Structure of the copper complex conjugated to Sav. (c) Proposed mechanism to control Cu-OOH stability by different positioning of hydrogen bonds. Reprinted with permission from ref 465. Copyright 2017 American Chemical Society.

Aside from the α-helical bundle proteins mentioned previously, another type of de novo designed peptides was reported by Rufo et al. as copper-based oxidases (Figure 40). These β-sheet peptides self-assemble to form an amyloid fibril, serving as robust heterogeneous biocatalysts. A tri-His motif was built on the outer face between the β-strands to bind metals such as Zn and Cu.⁴⁷¹ These solvent-exposed metal binding sites may be able to interact with substrates that are incompatible with hydrophobic environments in common enzymes. A report in 2016 showed that such amyloid assembly catalyzed the oxidative C─C coupling of 2,6-dimethoxyphenol using molecular oxygen.⁴⁷² In a following study, the Cu-amyloid assembly showed versatile activities, catalyzing the cascade hydrolysis and oxidation of 2′,7′-dichlorofluorescin diacetate (DCFH-DA).⁴⁷³ It also showed that pure hydrolase activity could be achieved by catalyzing the hydrolysis and detoxification of paraoxon, a toxic organophosphate insecticide.

Figure 40.

Schematics of (a) Cu-bound amyloid assembly protein catalyzing the oxidation of DMP and (b) Cu-bound amyloid assembly protein catalyzing the cascade hydrolysis/oxidation of a fluorescence derivative and the hydrolysis of paraoxon. Reprinted with permission from refs 472 and 473. Copyright 2016 and 2018 Wiley-VDH and American Chemical Society.

4.3. Designed Copper Enzymes with New-to-Nature Activities

The concept to combine artificial organometallic complexes with protein scaffolds was first raised in the 1970s by Whitesides, Kaiser, and co-workers.^474,475 Not limited by the natural amino acid ligand set, this bioconjugation-based strategy takes advantage of the non-natural reactivities of artificial cofactors while using the outer coordination spheres imposed by the surrounding protein matrix to impart unique properties on reactions, including enhanced rates, yields, stability, and selectivity. Various transformations have been reported in the past decades and reviewed extensively.^69,476 This section will focus on the protein-conjugated copper catalysts that specifically utilize the outer coordination spheres of the protein hosts.

Copper, known for its Lewis acidity, has been used as a catalyst for the Diels–Alder (D-A) reaction since the 1960s.⁴⁸⁶ To mimic this reactivity, Reetz et al. used commercially available Cu-phthalocyanine imbedded in serum albumin (SA).⁴⁸⁷ The Cu macrocycle was proposed to be noncovalently anchored in the IB subdomain, analogous to the natural hemin. Among a series of substrates, an endo/exo product ratio of up to 95:5 with an ee value up to 98% could be accomplished. The high enantioselectivity was not experimentally probed beyond product analysis, but it was proposed that protonation and/or hydrogen bonding with the substrate nitrogen atom was relevant. A later study by the same group created a Cu-(His)₂Asp facial triad site in a β-barrel core of the tHisF protein (Figure 41a), which was found to catalyze the D─A reaction on the same set of substrates but with lower enantioselectivity.⁴⁷⁷ The same Cu-(His)₂Asp core was later adopted and improved by Ghattas et al. in a 1-aminocyclopropane carboxylic acid oxidase (ACAO) scaffold by substituting the native Fe center with Cu (Figure 41b).⁴⁷⁸ Yields over 85% for all substrates were accomplished, with a maximum endo/exo ratio of 99:1 and ee of 95%. A computational docking study indicated that the substrate is flanked and orientated by Lys158 and Ile184, while Phe250 packs with the substrate from the noncoordinating side. These SCS restraints only enable one possible direction for the attack by cyclopentadiene, explaining why such high enantioselectivity was achievable.

Figure 41.

Examples for artificial Diels—Alderase. (a) tHisF protein with the copper binding site. (b) ACCO protein with the copper-bound cavity and docked substrate. (c) FhuA protein and the copper binding site. (d) Oligonucleotide and the conjugated copper complexes. (e) LmrR protein with the conjugation sites for copper complexes shown in red. (f) αRep A3 protein with the central pocket to conjugate copper complexes. (g) Neocarzinostatin protein with the copper complex noncovalently docked in the pocket via the testosterone tail. (h) FhuA protein with the copper complex conjugated. (i) apo-nitrobindin protein with the covalently conjugated copper complex. (j) Schematic representation that the copper complex with the adenosine tail may bind to the cell-surface receptor and catalyze the D—A reaction. (k) Bovine pancreatic polypeptide and the unnatural amino acid residues introduced to form a copper binding site. Reprinted from references 367, 369, 477-485. Copyright 1021, 2009, 2010, 2019, 2019, 2011, 2017, 2016, 2016, 2016, and 2018 Wiley-VCH and American Chemical Society.

Meanwhile, a series of artificial enzymes were reported using covalently fused Cu-phenanthroline or Cu-dipyridine compounds and scaffolds including polynucleotides,^480,488 LmrR,^367,489 αRep A3,⁴⁸¹ neocarzinostatin,⁴⁸² FhuA,⁴⁸³ aponitrobindin,⁴⁸⁴ adipocyte lipid binding protein, and human adenosine receptor (Figure 41c-j).⁴⁸⁵ Most of these examples adopt a hydrophobic SCS, either inside the protein or formed between dimer surfaces, to stabilize the aromatic copper complex, facilitate substrate binding, and accelerate the reaction. Another strategy using amino acid residues as ligands has also been adopted to design a Diels–Alderase in bovine-pancreatic polypeptide (BPP)³⁶⁹ (Figure 41k), LmrR⁴⁹⁰ and mTFP.⁴⁷⁹ Nonetheless, many of the enzymes in these early works have slow catalytic rates and relatively low conversion. This issue was addressed in a very recent study by Basler et al.⁴⁹¹ An α-helical protein binding Zn (a structural analogue of Cu with different acidity) was designed and developed through several rounds of directed evolution. The catalytic efficiency of this construct was improved by over 10⁵-fold between the starting and final designs. The final design, DA7, has a catalytic efficiency of 2.9 × 10¹⁰ M⁻¹, a TON of 10⁴, and >99% endo selectivity. Examination of the crystal structure and a docking model revealed the alignment of the diene substrate by the hydrogen bond donor Arg28 from the carboxylate side and by Gln31, Gln80, and Tyr84 from the pyridyl side. The dienophile substrate was packed underneath the diene by the aromatic ring of Trp68 with the two dienedienophile moieties closely facing each other. The precise orientation of the substrate explained how these high numbers were accomplished, the most efficient currently reported for a Diels–Alderase.

Another application of Cu-protein hybrid catalysts is the Friedel–Crafts (F-C) alkylation reaction. To this end, the Roelfes group used a LmrR protein that forms a homodimer with a large hydrophobic pocket.³⁶⁸ The two Trp residues at the interface exert strong π-π stacking to allow tight binding to an aromatic Cu^II-phenanthroline (Cu^II(phen)) or Cu^II-2,2′-bipyridine (Cu^II(bpy)) complex with micromolar affinity (Figure 42a). These artificial metalloenzymes catalyzed the site selective alkylation of an indole substrate with near full conversion and ee up to 93%. Interestingly, the reactivity of this artificial enzyme could be tuned by mutation of two SCS residues. The A92E mutation altered the side chain orientation toward an outer Asn14 residue, enhancing binding of the Cu^II(phen) cofactor between the two Trp residues according to MD simulations and eliminating the minor reactivity of the tandem F-C alkylation/enantioselective protonation reaction (FC-EP).⁴⁹² Removal of the packing Trp96 residue, however, switched the reactivity toward the FC-EP reaction with moderate enantioselectivity due to the binding dynamics of the cofactor. The hybrid catalyst was further applied in whole-cell biocatalysis, and directed evolution was applied to the outer sphere residues to enhance the yield and selectivity.⁴⁹⁴ A similar noncovalent cofactor docking method has also been applied in other proteins from the TetR family to catalyze the same Friedel–Crafts reaction (Figure 42b).⁴⁹³ An alternative strategy of genetic unnatural amino acid incorporation was adopted by the same group to introduce the Ala_Bpy unnatural amino acid residue as Cu ligand.⁴⁹⁰ Mutation of three SCS residues, Asn19, His86, and Phe93, had a moderate impact on the conversion rate, enantioselectivity, and substrate preference, which was attributed to π-stacking interactions and the steric freedom of the hydrophobic pocket. This Ala_Bpy LmrR mutant was further used to stabilize a semiquinone radical (Figure 42c), reminiscent of the chemistry of COx.²⁸⁶

Figure 42.

(a) LmrR protein with two tryptophans that docks the aromatic ligand of the copper complex. The active site structure at the bottom right shows the targets for SCS mutagenesis studies. (b) Different proteins from the TetR family that were engineered for F-C catalysis. The proteins are LmrR (top left), CgmR (top right), RamR (bottom left), and QacR (bottom right). (c) The Ala_Bpy LmrR mutant that stabilizes a semiquinone radical. Reprinted with permission from references 286, 368, 492, and 493. Copyright 2017, 2015, and 2020 American Chemical Society and Wiley-VCH.

The Michael addition reaction, another type of C─C bond formation reaction, can be catalyzed by artificial copper enzymes. In the late 2000s, Roelfes and co-workers reported the enantioselective catalysis of the Michael addition reaction by a Cu binding oligonucleotide⁴⁸⁸ and bovine pancreatic peptide scaffold,³⁶⁹ respectively. Recently, the group expanded the previously mentioned Cu^II(phen)-Trp docking strategy with an unnatural adjacent p-aminophenylalanine (pAF) residue.⁴⁹⁵ The two active centers work in a synergistic manner, in which Cu binds and activates the Michael donor while the pAF residue activates the Michael acceptor substrate by forming a Schiff base (Figure 43a). An improved yield (36%) and ee (86%) were observed in the dual-functional catalyst versus controls. The M8L mutation provided moderately improved activity and selectivity (99%), which were correlated to a hydrophobic interaction with Trp96, as described in a previous study vide supra. The Itoh group has reported a cupin-based catalyst using only natural amino acids (Figure 43b).⁴⁹⁶ Starting from a natural (His)₄ binding motif, a ligand set library was screened by decreasing numbers of His residues. Both the tris-His mutant H52A (S product) and the bis-His mutant H54A/H58A (R product) were found capable of catalyzing the Michael addition reaction with both high yield and selectivity, but interestingly, their enantioselectivities were reversed relative to one another. A docking study revealed that the hydrogen bond from Cys106 (sulfinic acid form) may swing to the Si-face of the azachalcone substrate in the H52A mutant, favoring the attack by nitromethane from this side. On the other hand, the H54A/H58A mutant appeared to have an open Re-face with the Phe104 residue blocking the Si-face, and as a result the Cys106 swings to the Re-face to enable attack from the Re side. This effect was supported by a subsequent mutagenesis study, in which the H52A/C106A or H52A/C106S mutation removes or disori-entates the H-bond and reduces the Si selectivity, whereas the H54A/H58A/F104W mutation enhances the steric hindrance from residue 104 to increase the Re selectivity. Another recent study by Rimoldi et al. used two simple peptides derived from the methionine-rich motif in hCtr-1 transporter to catalyze the Michael addition reaction.⁴⁹⁷ The yield was comparable to those of previous studies with a relatively lower enantioselectivity, possibly due to the absence of the SCS interactions of more complex proteins.

Figure 43.

(a) LmrR variant catalyzing the Michael addition reaction with an aniline moiety on an unnatural amino acid residue to activate the ketone Michael acceptor. The structure was modeled from PDB ID 6R1L,⁴⁹² and the scheme was adapted from ref 464. (b) Two cupin variants with different enantioselectivities of the Michael addition reaction. The insets at the bottom demonstrate how the hydrogen bond with Cys106 changes the orientation of the substrate and thereby alters the product enantioselectivity. Reprinted with permission from refs 495 and 496. Copyright 2020 Nature Publishing Group and Wiley-VCH.

Artificial Cu enzymes have also been reported for hydration reactions and small molecule polymerization. Both covalent conjugation and noncanonical amino acid incorporation strategies were used in two studies to introduce phenanthroline or Ala_Bpy Cu complexes into the LmrR scaffold (Figure 44).^498,499 Hydration of α,β-unsaturated ketone substrates was achieved in both cases with moderate enantioselectivity (60–80%). Two residues, Phe93 and Asp100, were found to serve as important SCS residues in both cases. The Phe residue provided a π−π stacking interaction to support binding of aromatic copper complexes, while the Asp residue acts as a general Lewis base, forming hydrogen bonds and activating the water substrate. In the latter case (Cu-Ala_Bpy), adding an alternative general base such as V15E significantly increased both the conversion rate and enantioselectivity due to better positioning of the carboxylate. As for polymerization, the Bruns group reported a protein cage nanoreactor system using thermosome (THS).³⁷⁰ This protein is a group II archaea chaperonin with a large (130 nm³) cavity, that was used to confine the Cu complex and substrate, as well as wide pores, to release polymeric products. The nanoreactor produced poly(N-isopropyl acrylamide) by ATRP with an average molecular weight of 1500 g/mol and a polydispersity index of 1.11. Other catalysts, such as horseradish peroxidase, have also been used in this nanoreactor system.⁵⁰⁰

Figure 44.

(a) LmrR variant as a hydratase with important SCS residues shown as sticks. (b) LmrR variant with an Ala_Bpy residue to bind copper and carry out hydration of alkene. Reprinted with permission from refs 498 and 499. Copyright 2013 and 2017 Royal Society of Chemistry.

5. CATALYSIS BEYOND THE PRIMARY COORDINATION SPHERE BY RATIONALLY DESIGNED MULTINUCLEAR METALLOENZYMES CONTAINING HEME

By incorporating another metallocofactor next to the heme, nature evolved a number of multinuclear metalloenzymes to fulfill different and more challenging functions. One example is manganese peroxidase (MnP), in which a Mn^II-binding site next to the heme macrocycle can oxidize Mn^II to Mn^III upon reaction with H₂O₂. The generated Mn^III is then used to oxidize other organic substrates, in some instances.⁵⁰¹ On the other hand, heme-copper oxidase (HCO) consists of a heme-Cu_B heteronuclear center. It functions as a terminal oxidase catalyzing the reduction of O₂ to water to generate the transmembrane proton gradient required for ATP synthesis.^41,502 Interestingly, bacterial nitric oxide reductase (NOR) has a non-heme iron center (Fe_B site) at the corresponding Cu_B position of HCO and catalyzes the one-step, two-electron reduction of NO into N₂O, which is an important step in denitrification.^422,503 Dissimilatory sulfite reductase (SiR) catalyzes a six-electron reduction of sulfite into sulfide using a combination of a heme macrocycle (siroheme) and a [4Fe-4S] cluster as cofactor. It makes up one of the important steps in sulfate respiration in sulfate-reducing bacteria.⁵⁰⁴ In order to elucidate how nature evolved these different multinuclear heme-containing metalloenzymes to catalyze different reactions, efforts have been made using synthetic⁵⁰⁵ and biosynthetic models^69,476,506 to provide more insights. Moreover, inspired by the water-bridged homodinuclear magnesium mechanism for some natural nucleases,⁵⁰⁷ an artificial nuclease was made by the design of a Mg²⁺-heme heteronuclear Mb system.⁵⁰⁸ In this section, we will focus on how SCS affects the performance of these multinuclear heme-containing ArMs.

5.1. Manganese Peroxidase Using Cytochrome c Peroxidase as a Scaffold

To engineer a functional mimic of MnP, the Lu group successfully designed a Mn binding site into CcP through the introduction of G41E, V45E, and H181D mutations (Figure 45a and b).⁵⁰⁹ The resulting triple mutant (MnCcP) could oxidize Fe^II at least five times more efficiently than the native CcP. A deeper structural analysis and comparison of the secondary coordination spheres of CcP and MnP uncovered two residues of interest, namely Trp51 and Trp191 of CcP, which sit in the same positions occupied by Phe in MnP (Figure 45c and d). Hence, the W51F and W191F mutations were made to further tune the reactivity of MnCcP. While no significant change in activity on Mn^II oxidation was observed for W191F MnCcP, W51F MnCcP was found to have a 4.3-fold increase in V_max compared to MnCcP, and the W51F/W191F MnCcP double mutant displayed a further 14.3-fold increase in V_max over that of W191F MnCcP. This compounding effect may be rationalized if one assumes there is an equilibrium between the Fe^IIITrp^+. state (the radical of compound I in CcP is believed to preside on Trp191 rather than the porphyrin ring, as Trp is readily oxidized) and the ferryl state of CcP compound II. Both Trp51 and Trp191 are capable of stabilizing compound II, with either Trp51 through H-bonding to the ferryl oxygen or Trp191 through the proposed equilibrium. Although W191F MnCcP cannot stabilize compound II through radical delocalization, it can still be stabilized by H-bonding interactions with Trp51; therefore, no increase in activity is observed. For W51F MnCcP, the delocalization from Trp191 can stabilize compound II, but the H-bonding interaction is missing, so an increase in activity is observed. By mutating both Trp to Phe, both stabilizing effects are removed, producing the largest increase in activity.^510,511 The resonance Raman spectra of these mutants also reveal that the W51F mutant lacks the H-bonding interactions between tryptophan and a weakly bound distal water. The loosening of this water may lead to quicker formation of compound I and an increase in activity.⁵¹² However, all the MnCcP and its mutant mentioned did not match the activity of native MnP. The k_cat/K_M for W51F/W191F MnCcP was 0.599 s⁻¹ mM⁻¹, while the one for native MnP was 3952 s⁻¹ mM⁻¹.⁵¹⁰

Figure 45.

(a) Mn-binding site in MnP (PDB ID: 1MNP).⁵¹³ (b) Putative Mn-binding site in CcP, forming MnCcP (PDB ID: 2CCP).¹³⁹ (c) Position of Phe45 and Phe190 around the active site of MnP. (d) Position of Trp51 and Trp191 around the active site of CcP. Atom coloring: N (blue); O (red); C_MnP (magenta); C_CcP (green); Fe (orange); Mn (lilac).

To enhance the activity of MnCcP and investigate the roles of other secondary sphere residues, MnCcP was further redesigned by replacing Asp181 with Glu, reverting the G41E mutation, and further mutating Asp37 to Glu. The resulting construct is referred to here as MnCcP.1. MnCcP.1 was found to have a 2.5-fold higher catalytic efficiency (k_cat/K_M) than MnCcP.⁵¹⁴ Based on MnCcP.1, several additional alterations to the secondary coordination sphere were made to further tune its activity.⁵¹⁵

First, comparison of the crystal structures of MnP and MnCcP revealed that Glu39 in MnP is very flexible and can adapt different rotamers. As a result, binding of Mn^II can occur with a relatively short distance of 2.10 Å. Meanwhile, the correspond-ing Glu45 of MnccP.1 is rather rigid, retaining its position in either Mn^II-free form or Mn^II-bound form with a distance of 3.94 Å in the latter. This results in a different metal center geometry and may be responsible for the different Mn^II binding affinities and MnP activities. Tyr36 was found to be within H-bonding distance of the carboxylate group of Glu45 in MnCcP.1 (Figure 46), an interaction that is otherwise missing in MnP. This H-bond may compete with Mn^II coordination by stabilizing rotamers that are unfavorable for coordination. To address this, a Y36F mutation was introduced into MnccP.1 and was found to have a 2.8-fold improvement of K_M, indicating the binding affinity of Mn^II was increased.

Figure 46.

Different SCS residues included to MnCcP.1 and their effects on the catalytic efficiency. Reprinted with permission from ref 515. Copyright 2016 American Chemical Society.

Second, residues Glu37 and Glu181 were found to closely neighbor Mn^II in MnCcP.1, with distances of 3.23 Å and 4.08 Å, respectively. However, these distances are considerably longer than the corresponding coordination bonds of Mn^II─Glu35 and Mn^II─Asp179 in MnP, which are 2.4 Å and 2.41 Å, respectively. A structure-based sequence alignment of MnPs from different organisms revealed that a conserved Arg was located below Glu35 in MnP. Bearing in mind that this arginine might play a key role in the positioning of Glu35 and Asp179 in MnP through the formation of a salt bridge, a K179R mutation was made in MnCcP.1 (Figure 46). MD simulations showed that Arg179 moved farther from Glu181, causing a reorientation of Glu181 and, consequently, reducing the interatomic distance of the carboxylate side chain to Mn^II by nearly 2 Å. Activity assays showed that, as a result, a 2.6-fold improvement in K_M was observed.

Additionally, it was discovered that Ile40 in MnCcP.1 may sterically clash with Glu37 in MnccP.1, causing the Glu side chain to reorient away from the Mn^II center. In most MnPs, the corresponding position is occupied by glycine, and therefore, an I40G mutation was introduced in MnCcP.1 (Figure 46). The corresponding MD simulations showed a decrease in bond length between the ligands and Mn^II. That for Glu37 decreased from 3.23 Å to 2.54 Å, and that for Glu45 decreased from 3.94 Å to 2.34 Å. This resulted in a 3.6-fold improvement for K_M. For this mutant, a 2.4-fold increase for k_cat is also observed and resulted in an overall 8.6-fold increase in catalytic efficiency.

5.2. Heme-Copper Oxidase Using Myoglobin as a Scaffold

5.2.1. Design Using Natural Amino Acids.

The practical application of enzymes such as HCO outside of the cellular environment is often challenging due to their large size and complexity. Therefore, designing ArMs based on small and robust protein scaffolds is an attractive solution to this problem.

While Mb contains a single heme center, HCOs contain a high-spin heme-Cu_B center and a low-spin heme dinuclear center (Figure 47). As a subclass of HCO, CcO also possesses a dinuclear Cu center called Cu_A.³³¹ To understand the role of the Cu_B site for the function of HCOs, Lu and co-workers reported their first design of the HCO biomimetic system using the Mb scaffold in 2000.⁵¹⁸ To accomplish this, a Cu-binding site (Cu_B site) was engineered in the vicinity of the heme Fe by introduction of two His residues, L29H and F43H, in the distal pocket of the heme site, closely resembling the heterobinuclear active site found in HCOs (Figure 47c). Herein, the L29H/F43H Mb mutant is referred to as Cu_BMb. Initial studies on Cu_BMb suggest that the copper ion in the Cu_B center is essential for the O₂ reduction activity.⁵¹⁹ However, Cu_BMb generates verdoheme, instead of ferryl-heme, during the O₂ reduction due to the lack of the H-bonding network that delivers protons to promote the heterolytic O─O cleavage. Since the Tyr residue in the active site of HCOs has been thought to be critical to the enzymatic functions,⁵²⁰ Miner et al. later found that introduction of Tyr to the active site of Cu_BMb (F33Y-Cu_BMb and G65Y-Cu_BMb) enabled this artificial enzyme to convert O₂ to water with minimal release of reactive oxygen species (ROS), which in turn prevented the degradation of heme to verdoheme (Figure 48).⁵¹⁷ Repositioning of Tyr in the G65Y Cu_BMb construct further improved the rate of H₂O production relative to that of F33Y Cu_BMb, illustrating for the first time that the positioning of the Tyr is critical for modulating the catalytic rate. Moreover, the crystal structure of apo-F33Y Cu_BMb revealed additional water molecules in the distal heme pocket, suggesting that Tyr may be involved in the H-bonding network that activates the ferric-superoxo intermediate for water generation. Two years later, Yu et al. reported the direct observation of the Tyr33 radical in F33Y Cu_BMb by EPR (Figure 48b), providing firm support for the presence of a Tyr radical and the role of Tyr as an electron and proton donor in the enzymatic mechanism of HCO.⁵²¹ Moreover, in 2016, the EPR spectra of cryoreduced oxy-F33Y Cu_BMb and the crystal structure (Figure 48c) of this mutant provided further direct evidence of the importance of H-bonding networks in activating the enzyme to reduce O₂ to water.⁵²²

Figure 47.

(a) Crystal structure of ba₃ CcO (PDB ID: 3S8F).⁵¹⁶ (b) Structure of the catalytic site. The low-spin heme b and heme a₃ are depicted in cyan; Cu_B is indicated by a yellow sphere. The cross-linked residues His233 and Tyr237, located in proximity of the catalytic site, are shown. (c) Crystal structure of L29H/F43H Mb (PDB ID: 4FWZ).⁵¹⁷

Figure 48.

(a) Structures and computer models of native and designed oxidases. Left panel: overlay of the crystal structures of bovine CcO(black) and E-F33Y-Cu_BMb (cyan). Right panel: overlay of the crystal structures of cbb₃ HCO from Pseudomonas stutzeri (tan) and the E-G65Y-Cu_BMb computer model (yellow). The Cu_B copper is represented as an orange sphere. N (blue); O (red); Fe (green). Reproduced with permission from ref 517. Copyright 2012 John Wiley & Sons. (b) EPR spectra of ferric F33Y-Cu_BMb reacted with 1 equiv of H₂O₂ (red) and ferrous F33Y-Cu_BMb with O₂ (black). Reprinted with permission from ref 521. Copyright 2014 American Chemical Society. (c) Crystal structure of oxy-F33Y-Cu_BMb (PDB ID: 5HAV, left panel) compared with that of oxy-WT-Mb (PDB ID: 1A6M).⁵²³ Reprinted with permission from ref 522. Copyright 2016 American Chemical Society.

Taking advantage of the enhanced O₂-reduction activity of G65Y Cu_BMb, Mukherjee et al. reported the Mb-based biosynthetic model CcO, capable of electrochemically catalyzing O₂ reduction with a reaction rate higher than that of either any known synthetic analogue or the native CcO.⁵²⁴ The electrons required for the O₂ reduction reaction were directly injected into the heme cofactor via tethering by click reaction of the functionalized hemin-yne to azide terminated thiols bound to a gold electrode. This approach to direct electron transfer circumvents the rate-limiting dissociation of ferric hydroxide and the presence of the protecting Tyr65, resulting in an O₂ reduction activity that is an order of magnitude faster than that of CcO fixed by immobilization on an electrode.

In 2015, Yu et al. further developed a Mb variant (D44K/D60K/E85K) based on the G65Y-Cu_BMb platform by incorporating residues to promote favorable surface electrostatic interactions with the native redox partner of HCO, cyt b₅ (Figure 49).⁵²⁵ Using this approach increased the activity of the Cu_BMb model to a level comparable to that of the native enzyme.

Figure 49.

(a) Structures of G65Y-Cu_BMb(+6), showing the engineered Lys in blue, and cyt b₅ (PDB ID: 1CYO).⁵²⁶ (b) Oxidase activity of G65Y-Cu_BMb(+6) in comparison with that of native cyt cbb₃ oxidase. The black arrows indicate the addition of reductant, and the double arrow shows the injection of native cyt cbb₃ oxidase. Reprinted with permission from ref 525. Copyright 2015 American Chemical Society.

In addition to serving as functional models of HCOs, engineered Mb oxidases have also proven to be excellent models for studying the relationship between the reduction potential and O₂-reducing activity. In 2014, Bhagi-Damodaran et al. introduced an S92A mutation to F33Y-Cu_BMb, aiming to remove the H-bonding interactions between Ser92, the proximal His ligand, and a heme propionate as well as increasing the hydrophobicity in the heme pocket (Figure 50).⁵²⁷ The S92A mutation led to an increase in the reduction potential by about 30 mV, resulting in a higher driving force and faster O₂ reduction.

Figure 50.

Overlay of the crystal structures of F33Y-Cu_BMb (PDB ID: 4FWX, cyan)⁵¹⁷ and S92A-F33Y-Cu_BMb (PDB ID: 4TYX, orange).⁵²⁷ Heme b and the side chains of His29, His43, His64, Tyr33, His93, and Ser/Ala92 are shown in licorice. Ser92 forms H-bonds with heme propionate and His93. Reprinted with permission from ref 527. Copyright 2014 American Chemical Society.

In 2018, Mukherjee et al. incorporated one or two proton transfer glutamate residues (V68E and V68E/I107E) in Cu_BMb to mimic the proton donor acidic residues in HCOs (Figure 51a).^524,528 Although V68E and V68E/I107E-Cu_BMb showed O₂ reduction rates comparable to that of G65Y-Cu_BMb, which does not contain a glutamate residue in the active site, the presence of Glu resulted in a significantly reduced solvent kinetic isotope effect (SKIE), with V68E/I107E-Cu_BMb bearing the lowest SKIE of ~2.4 (Figure 51b). The dramatic difference in SKIE indicates a change in the rate-limiting step in the reaction mechanisms, with Glu mutations presumably providing a preorganized proton transfer channel that speeds up proton translocation.

Figure 51.

(a) Crystal structure of Fe_B^II-V68E/I107ECu_BMb (PDB ID: 3M39).⁵²⁹ (b) Cartoon showing that incorporation of the Glu residue near the active site facilitates the proton transfer from the external solvent to the active site, which results in lower KSIE. (c) Structure of the active site of oxy-I107E-Cu_BMb (PDB ID: 7L3Y).⁵³⁰ (d) Rates of O₂ consumption yielding water (blue) or reactive oxygen species (ROS) (red). Parts c and d are reprinted with permission from ref 530. Copyright 2021 American Chemical Society.

The crucial role of the active site Glu residue in regulating oxidase activity was further reinforced by Petrik et al. in 2021.⁵³⁰ An I107E mutation not only promoted the O₂ reduction rate but almost completely eradicated ROS formation. A combination of crystallographic and EPR studies revealed an extensive H-bonding network involving glutamate-facilitated reaction intermediate protonation (Figure 51).

5.2.2. Design Using Unnatural Amino Acids.

Although the Tyr-His cross-link is a well-established feature in native HCOs (Figure 52a), the mechanistic role of this cross-link is still unclear.⁵³¹ To better understand the impact of this feature in HCOs, Liu et al. reported the genetic incorporation of an UAA, imiTyr, into Cu_BMb at position 33 to mimic the Tyr-His cross-link of HCOs (Figure 52b).⁵³² Compared to the F33Y Cu_BMb mutant, the Mb model bearing imiTyr exhibited an 8-fold increase in selectivity and 3-fold increase in catalytic turnover (Figure 52c). It was postulated that the decrease in the pK_a of the phenol in imiTyr is key to facilitating proton delivery and Tyr radical formation.

Figure 52.

(a) Cu_B site of CcO (bovine numbering). (b) Structural model overlay of imiTyrCu_BMb (cyan) and F33Y Cu_BMb (yellow). (Inset) Chemical structure of imiTyr. (c) O₂ reduction turnover number measured during the stepwise addition of O₂. The reduction was catalyzed by imiTyrCu_BMb or F33Y Cu_BMb. Reprinted with permission from ref 532. Copyright 2012 Wiley-VCH. (d) Crystal structure of F33Y/L29H/F43H swMb (F33Y Cu_BMb, PDB ID: 4FWX)⁵¹⁷ and pK_a values of the Tyr and Tyr analogs. From top to bottom: Tyr, 3-chlorotyrosine (ClTyr), 3,5-difluorotyrosine (F2Tyr), and 2,3,5-trifluorotyrosine (F3Tyr). The red arrow indicates increased activity with the increased pK_a values of Tyr and Tyr analogs. Reprinted with permission from ref 533. Copyright 2015 American Chemical Society. (e) Structural model of the OMeY Mb mutant, constructed based on the crystal structure of F33Y Cu_BMb (PDB ID: 4FWX).⁵¹⁷ Reprinted from ref 534 with permission from The Royal Society of Chemistry.

Since Tyr has been suggested to function as both a proton and electron donor in the enzymatic cycle of HCOs, both the pK_a and reduction potential of Tyr can potentially affect its role in modulating the enzymatic activity. To analyze the correlations between these factors and the oxidase activity, Yu and co-workers established a linear relationship between the pK_a of the phenol group at position 33 and the O₂ reduction activity using a series of unnatural analogues of Tyr with varying pK_a values of the phenol ring (Figure 52d).⁵³³ However, the dependence of the oxidase activity on the reduction potential of Tyr is convoluted, since the pK_a and the reduction potential of Tyr are closely correlated. To tackle this problem, Yu et al. incorporated a Tyr analog, 3-methoxy tyrosine (OMeY), that has a comparable pK_a to that of Tyr but a lower reduction potential (Figure 52e).⁵³⁴ The F33OMeY Cu_BMb exhibited a higher O₂ consumption rate with a lower generation of ROS compared to the case of F33Y Cu_BMb. In this manner, they demonstrated that the electron donating ability of Tyr in the active site is a critical factor for the function of HCOs.

5.3. Bacterial Nitric Oxide Reductase Using Myoglobin as a Scaffold

NOR is structurally homologous to the largest subunit of HCO; rather than a three-histidine-coordinated Cu_B site in the distal position of the heme, the Fe_B site is incorporated into this position instead. Aside from the three His residues, two additional Glu residues were found to be essential for NOR activity. To mimic the structure and activity of NOR, the Lu group engineered a three His/one Glu site in the distal pocket of swMb, successfully established NOR activity.⁵³⁵ To make this model more similar to the native NOR and to investigate the role of the second Glu, the second Glu was introduced into Fe_BMb by a I107E mutation. The resulting crystal structure showed a bridging water molecule between the OE2 atom of Glu107 and the non-heme iron center (Figure 53). The NO reduction activity of Fe^II-I107E Fe_BMb was found to be higher than that of Fe^II-Fe_BMb, with ~24% N₂O for the former and ~10% for the latter over the course of ~20 h under single-turnover conditions. This suggests that unlike the initial three His residues and one Glu, which directly coordinate to the Fe_B site, the second Glu may potentially play a role in the proton transfer pathway during the reduction of NO.⁵²⁹ This hypothesis was further supported by making a I107F mutation in Fe_BMb, which inhibited N₂O formation.⁵³⁶

Figure 53.

Overlay of Fe^II-Fe_BMb and Fe^II-I107E Fe_BMb: Fe^II-I107E Fe_BMb, PDB ID: 3M39 (cyan); Fe^II-I107E Fe_BMb, PDB ID: 3K9Z (orange).⁵²⁹ Reprinted with permission from ref 529. Copyright 2010 National Academy of Sciences.

By utilizing this model Fe_BMb NOR enzyme, an important question as to why Cu is preferred over Fe for O₂ activation and reduction in HCO was resolved. The role of different metal ions in the distal pocket was explored by titration of the non-heme Fe_B center with Cu^I (Cu^I-Fe_BMb, mimic of HCO), Fe^II (Fe^II-Fe_BMb, mimic of NOR), and Zn^II (Zn^II-Fe_BMb, control).⁵³⁷ While the redox potential values of the heme center were similar (−64 ± 4 mV, − 58 ± 3 mV, and − 45 ± 2 mV, respectively), the redox potential values of the non-heme metal center were different, with +259 ± 20 mV for non-heme Fe^III/Fe^II and +387 ± 25 mV for Cu^II/Cu^I. Resonance Raman (rR) spectroscopy showed that, in the O₂-bound form, Zn^II-Fe_BMb exhibited a stronger Fe─O₂ bond, which, in turn, implies greater O─O bond activation compared to that of Fe_BMb in the absence of a distal metal ion. DFT calculations indicated that the O─O bond in Fe-Fe_BMb was longer than that in Zn-Fe_BMb and even longer in Cu-Fe_BMb. One possible explanation for this behavior is that Fe^II and Cu^I are redox active and capable of electron donation to O₂. Additionally, Cu^II may provide a higher degree of d-electron density (d⁹) than Fe^III (d⁵) to O₂ through d_π-p_π backbonding, further weakening the O─O bond. In fact, copper is the only 3d transition metal that is redox active and have maximum number of d-electrons for O₂ activation.⁵³⁷

The redox potential of heme can play an important role in tuning NOR reactivity. Bhagi-Damodaran et al. tuned Fe_BMb by exchanging heme b for monoformyl (MF-) or diformyl (DF-) heme, as well as by mutating Leu89, a residue close to the proximal His, into Ser. The exchange of heme increased the redox potential of heme center, while the mutation increased the hydrophilicity around heme and lowered the redox potential by stabilizing the higher oxidation states. This work found that a lower redox potential facilitated fast NO binding and subsequent decay of dinitrosyl species, while higher redox potentials facilitate faster electron transfer. Accomplishing multiple turnovers of NOR reactivity requires balancing optimizing the redox potential to balance these two factors.⁵³⁸

5.4. Sulfite Reductase (SiR) Using Cytochrome c Peroxidase as a Scaffold

In 2018, the Lu group successfully engineered a [4Fe-4S] binding site into CcP using six mutations (H175C, T180C, W191C, L232C, M230A, and D235V), which will be referred to as SiRCcP.1, with the purpose of creating a biosynthetic model of SiR.⁵³⁹ Following incorporation of heme b and the [4Fe-4S] cluster into SiRCcP.1, the resulting heme-FeS-SiRCcP.1 exhibited notable sulfite reducing activity, forming a basis for further work tuning the SCS to enhance its activity. A series of mutations were incorporated to form a positively charged cavity reminiscent of native SiR, including W51K, H52R, and P145K, mimicking Lys217, Arg153, and Lys215 in native SiR. The position of native Arg48 in CcP is similar to that of Arg83 in native SiR (Figure 54a). This W51K/H52R/P145K-SiRCcP.1 variant exhibited a 5.3-fold increase in activity over that of SiRCcP.1. Residue Asn481 in native SiR was oriented towards the iron sulfur cluster and may increase activity. Therefore, a D235N mutation at a similar position as SiR Asn481 was made to replace the original D235V mutation, with the resulting construct named SiRCcP.2 (Figure 54b). This new SiRCcP.2 construct was found to be 17-fold more active than SiRCcP.1. Further exchange of residue 235 for Cys in the D235C variant (SiRCcP.3) (Figure 54b) resulted in even higher activity. In combination with the W51K/H52R/P145K substitutions, W51K/H52R/P145K-SiRCcP.3 displayed a 63-fold increase in activity relative to the original SiRCcP.1 construct, achieving ~18% of the activity observed for native SiR from Mycobacterium tuberculosis.⁵³⁹

Figure 54.

(a) Computational model of the SiRCcP substrate-binding site with mutations W51K, H52R, P145K (cyan), and Arg48 (orange). (b) Computational model of the mutations D235V (SiRCcP.1, cyan), D235N (SiRCcP.2, green), and D235C (SiRCcP.3, magenta) in SiRCcP. Reproduced with permission from ref 539. Copyright 2018 American Association for the Advancement of Science.

5.5. An Artificial Nuclease Using Myoglobin as a Scaffold

By incorporating a Mg²⁺ into L29E Mb, an efficient artificial nuclease Mg²⁺-L29E Mb was made. Through molecular docking simulations of double-stranded B-DNA with the enzyme (Figure 55), it was determined that the distal Arg45 can interact with the sugar/phosphate group of DNA through H-bonding interactions. This basic residue was found to be conserved in the native nuclease, possibly enhancing the DNA binding affinity and further promoting cleavage. Introduction of a R45S mutation was introduced to eliminate the interaction and the DNA cleavage, and it was found to successfully inhibit nuclease activity by ~55–60%.⁵⁰⁸

Figure 55.

Docking structure of the Mg²⁺-L29E Mb-DNA complex (left) and enlarged view of the Mg²⁺-heme heterodinuclear center (right), highlighting the coordination of the Mg2+ ion and the H-bonding interactions with the sugar phosphate group of DNA. Reprinted with permission from ref 508. Copyright 2020 American Chemical Society.

6. BEYOND FE AND CU: INSIGHTS INTO THE SECONDARY COORDINATION SPHERE OF ENGINEERED METALLOENZYMES INCORPORATING NON-NATIVE METAL IONS

6.1. Replacement of Native Metal Ions and Metallocofactors

Besides mimicking the diverse array of metalloenzymes found in nature, the protein reengineering approach can be used to generate novel active sites to impart new reactivity with non-native metal ions and cofactors. However, doing so presents its own unique challenges, particularly in finding a scaffold capable of tightly binding the desired metal center or organometallic complex. To solve this problem, noncovalent, single-point covalent attachments, and dual covalent anchoring strategies can be employed. The Watanabe group has demonstrated that enantioselective enzymatic sulfoxidation can be regulated by unnatural metal complexes in a redesigned protein active site.⁵⁴⁰ Oxidative Cr^III and Mn^III Schiff base catalysts were inserted to Mb using His93 as an anchor, forming a metal–His93 dative bond. In addition to the noncovalent interactions, including π-stacking between the benzene ring of the salophen ligand and the Phe43 side chain, and C-H/π-interactions between the side chains of Leu89 and Ile99 and the ligand (Figure 56a), the two methyl groups of the salophen ligand were found to interact with Ile107, allowing the Mn complex to be inserted deep in the pocket (Figure 56b). The rate of sulfoxidation catalyzed by these complexes was accelerated by inserting into apo-Mb, and Cr^III-Mb and Mn^III-Mb exhibited 15- and 6-fold higher reactivities than that of the Cr^III Schiff base complex in buffer, respectively. Further modification of the ligand sterics regulated the position of the inserted synthetic metal complex, helping to tune the enantioselectivity, and additional mutation of H64D increased the substrate accessibility, enabling a larger substrate scope.

Figure 56.

Active site structure of Mn-1-apo-A71G Mb: (a) side and (b) top views. The heme of met-Mb¹²⁸ (blue) is superimposed on Mn-1-apo-A71GMb. (c) Metal-salophen complex (1) Reprinted with permission from ref 540. Copyright 2005 American Chemistry Society.

To further enhance both the rate and enantioselectivity of the sulfoxidation reaction, the Lu group constructed an artificial metalloenzyme by incorporating manganese salen complex into Mb using a dual covalent anchoring approach, generating MnSalen-Mb(T39C/L72C) (Figure 57).^541,542 Mb(T39C/L72C) was found to exclusively catalyze the formation of sulfoxide. To account for such high chemoselectivity, it was hypothesized that the polarity of the residue near the entrance of the protein pocket may play an important role in regulating substrate access. A single mutation, A71S, near the entrance of the protein pocket was made to increase hydrophilicity and, as a result, led to the further oxidized product sulfone. This demonstrates the delicate balance of the secondary coordination sphere in tuning selectivity.

Figure 57.

(a) Computer model of Mb(L72C/Y103C) with 1 covalently attached overlaid with heme. (b) Complex 1. Reprinted with permission from ref 541. Copyright 2004 American Chemistry Society.

To further enhance reactivity, the distal His64 residue was modulated by substitution with Phe to remove the H-bonding interaction with the active Mn-hydroperoxo species.⁵⁴³ Such mutation did not increase in the turnover numbers; however, substitution of His64 with Arg resulted in increased oxidation activity. This observation strongly suggested that the distal ligand His64 plays a key role in tuning the reactivity of the Mn-salen complex through H-bonding interactions.

Superoxide dismutases (SODs) serve to detoxify superoxides produced by aerobic metabolism and by the mammalian immune system in response to infection.⁵⁴⁴ Different SOD families are defined by the cofactor they utilize, including Ni SODs, Cu and Zn SODs, and Mn or Fe SODs. All members of the most common Mn/Fe-dependent SODs are related in sequence. They exhibit identical protein folds and have identical ligands coordinating the metal ion. Therefore, it is unclear why some enzymes require exclusively Mn for catalysis (MnSOD), while others require Fe (FeSOD) and yet others, such as camSOD, show metal cofactor flexibility.

The metal ion reduction midpoint potential (E_m) plays a critical role in which reactions an enzyme can participate and, therefore, must be carefully controlled. Miller and co-workers successfully tuned Fe- and Mn-containing SODs, generating FeSOD variants with E_m values spanning 900 mV (Figure 58a).⁵⁴⁵ The protonation state of the coordinated solvent accounted for the significant differences observed in both the redox potential and metal ion specificity. It was postulated that protonation of the coordinated solvent was suppressed in MnSOD by a strong H-bonding interaction with Gln69, leading to drastic differences in the E_m values relative to FeSOD (Figure 58b). Mutations Q69H and Q69E in FeSOD were designed to decrease the strength of this H-bond interaction with the coordinated solvent (Q69H) or to reverse the polarity of the H-bond (Q69E). The E_m was observed to increase moving from WT-FeSOD to Q69H to Q69E, supporting this hypothesis (Figure 58).

Figure 58.

(a) Redox tuning over almost 1 V of Fe-containing superoxide dismutase. (b) Hydrogen bonds present in the active sites of some FeSOD variants. Reprinted with permission from ref 545. Copyright 2008 American Chemistry Society.

To further explain the metal specificity of MnSOD and camSOD, Waldron and co-workers identified three residues (19, 159, and 160) that are spatially close to the metal binding site and vary between MnSOD and camSOD (Figure 59).⁵⁴⁶ Modification of position 159 in the G159L MnSOD increased its Fe-dependent activity while decreasing its Mn-dependent activity by ~2-fold. On the contrary, camSOD L159G diminished its Fe-dependent activity by >10-fold while increasing its Mn-dependent activity by >3-fold. Interconversion of activity and metal specificity could also be achieved by varying position 160, although alterations at position 19 seemed to have little impact on these properties. Surprisingly, although mutations at these positions change the metal specificity, there are essentially no changes observable in the protein backbone structure, at least within the resolution of the crystal structures. The primary coordination spheres, substrate access channels, and H-bonding networks all appear otherwise identical. However, such mutations had direct effects on the electronic properties of the Mn center, as demonstrated by high-field EPR. Different reactivities toward auto-oxidation and dithionite reduction suggested that each of the two WT-SODs have different redox potentials. These differences were inverted in the double mutants (MnSOD G159L/L160F and camSOD L159G/F160L), correlating also to an inversion of the Mn-dependent activity in these variants. Together, these results provide important insight into the role of secondary coordination sphere residues in modulating SODs and the factors that contribute to both altered metal specificity and reduction potential.

Figure 59.

Structure overlay of MnSOD (PDB ID: 5N56, yellow, F19/G159/L160) and CamSOD (PDB ID: 5N57, blue, I19/L159/F160).⁵⁴⁷

Cobalamin-dependent methionine synthase (MS) catalyzes the conversion of homocysteine to methionine in most mammals and bacteria. This large, modular protein (ca. 146 Da) consists of four distinct functional domains. Two kinds of methyl group transfers occur during the catalytic cycle. Namely, N-methyl tetrahydrofolate first acts as a methyl group donor to methylate cobalamin. This methyl-cobalamin is then used for methionine synthesis from homocysteine. Mechanistic studies of the native enzyme are challenging due to the structural complexities and the dynamic conformational changes that occur in MS during the reaction.⁵⁴⁸ Despite the importance of this enzyme, the number of protein model systems developed to investigate how MS initiates the S_n2-like transmethylation reaction and stabilizes reaction intermediates is limited.⁵⁴⁸ Using Mb reconstituted with a Co tetrahydrocorrin derivate (Co(TDHC)Mb), Hayashi and co-workers investigated the reaction catalyzed by methionine synthase.⁵⁴⁹ It was found that Co^I(TDHC)Mb could react with methyl iodide to form a methylated cobalt complex, a process that did not efficiently proceed for Co^I(TDHC) in organic solvents. These initial results highlight the importance of the heme pocket of Mb, which may serve to enhance the nucleophilicity of the Co^I species while simultaneously stabilizing the methylated Co^III species. Interestingly, transmethylation from the methylated cobalt complex to one of the nitrogen atoms of the His64 imidazole ring was observed. DFT calculations exploring this process supported a reduced barrier for His64 methylation together with significant stabilization of the methylated His product through the dielectric effect of the surrounding protein environment. In the crystal structure, the N_δ atom of His64 was found to be hydrogen bonded to a water molecule (Figure 60), suggesting barriers can be further reduced by the partial or full deprotonation of the methylated histidine.

Figure 60.

Crystal structure of horse heart myoglobin reconstituted with Co^II-tetradehydrocorrin (PDB ID: 3WFT).⁵⁵⁰ The N_δ1 atom of His64 hydrogen bonded to a water molecule is shown.

Ghirlanda and co-workers have also investigated the influence of the protein second coordination sphere on Mb substituted with cobalt-protoporphyrin XI (CoMb) (Figure 61).⁵⁵¹ Compared to free cobalt-protoporphyrin IX (CoPP(IX)), a 4-fold increase in the photoinduced hydrogen production activity was observed for CoMb. Substitution of the distal His residues in the engineered H64A/H97A CoMb mutant displayed a 2.5-fold higher reactivity relative to that of CoMb. It is hypothesized that the H64A substitution removes a slightly positive charge from His at physiological pH as well as a potential competitor for proton binding in the proximity of the CoPP(IX) active site. A hydrogen bond formed between His97 and one of the propionic acid groups of CoPP(IX) was also observed based on analysis of the crystal structure. Meanwhile, the H97A mutation was found to destabilize binding of CoPP(XI), alter the steric constraints on the porphyrin, and increase the degree of freedom between the scaffold and the CoPP(XI), effectively lowering the catalytic ability. However, the H97A substitution also increased the solvent accessibility of the active site, which may favor interaction between the photosensitizer and CoPP(XI), as the TON of H64A/H97A was found to be higher than that of the H64A mutant.

Figure 61.

Crystal structure of Co-Mb highlighting residues His64/93/97 (PDB ID: 1YOI).⁵⁵²

In 2016, Hartwig and co-workers constructed an artificial metalloenzyme (ArM) by incorporation of an Ir(Me)-heme cofactor into myoglobin (IrMb), capable of carbene insertion activity with up to 50% yields for a model reaction,⁵⁵³ Unlike many other anchoring strategies, the Ir(Me)-heme cofactor is embedded within the protein cavity, allowing SCS interactions to be readily employed to tune reactivity.⁵⁵⁴ Directed evolution of this IrMb was conducted by mutating a variety of residues, including the axial ligand and those in the substrate-binding region, to achieve substrate specificity. Aside from the axial and proximal ligands (His64 and His93, respectively), Phe43 and Val68 were mutated into hydrophobic or uncharged residues to maintain the hydrophobicity of the porphyrin/substrate binding site. Four additional residues around the active site (Leu32, Phe33, His97, and Ile99) were also investigated (Figure 62a). Different selectivities and yields were achieved for varying carbene insertion reactions across the range of designed mutants, with selectivities reaching up to a 92:8 enantiomeric ratio, yields up to 97%, and TONs up to 7260.⁵⁵³ The Ir(Me)-heme catalyst was also incorporated into the CYP119 scaffold (a variant of P450 enzymes), producing an ArM with better thermal stability and reactivity toward carbene insertion,⁵⁵⁵ cyclopropanation,⁵⁵⁶ and (sp³)C-H site selective functionalization reactions.⁵⁵⁷ Upon directed evolution of the residues near the active site, the authors achieved ArMs with improved yields, enantiomeric excess, chemoselectivity, total turnovers and turnover frequencies relative to those of IrMb. These mutations mainly focused around varying the heme axial residue (Cys317), residues lining the substrate binding pocket (Leu69, Ala209, Thr213, and Val254), and residues toward the distal side of the substrate binding site (Ala152, Leu155, Phe310, and Leu318) (Figure 62b).

Figure 62.

(a) Inner sphere (red), middle sphere (cyan), and outer sphere (orange) residues of the active site studied in Mb-based ArM for carbene insertion reactions (PDB ID: 1MBN).⁵⁵⁸ (b) Residues studied in CYP119-based ArMs: axial ligand for heme (green); residues lining the substrate binding pocket (cyan); residues to the distal sites of the substrate binding pocket (orange) (PDB ID: 1IO7).⁵⁵⁹

6.2. Design of New Metal Binding Sites

Nickel is intimately involved at the active sites of several important energy-converting enzymes, including NiFe H₂ases, carbon monoxide dehydrogenase (CODH), and acetyl coenzyme-A synthetase (ACS), all of which coordinate Ni in a S-rich environment. To better understand the nature of these enzymes and to mimic their unique activities, a number of protein reengineering efforts have been put forth.

Chakraborty and co-workers reported the rational redesign of a copper storage protein (Csp1) into a nickel binding protein as a novel mimic of [NiFe] H₂ase (Figure 63).⁵⁶⁰ The native active site of NiFe employs two Fe-bridging and two terminal Cys residues to coordinate Ni, making the already thiol-rich Csp1 an intuitive starting point. The 13 Cys residues along the central helical core of native Csp1 are capable of binding a total of 13 equiv of Cu^I. The Cu4 binding site (Cys26/62/87/113), which already enables tetrathiolate ligation, was chosen as the target location for the formation of a Ni(Cys)₄ site. During the redesign process, seven Cys residues away from the Cu4 site were mutated to Ala due to the similar side chain size of these residues in an effort to maintain a compact structure. Additionally, two bulky hydrophobic residues, Val (Cys94) and Leu (Cys103), were introduced to provide hydrophobic interactions at the exposed ends of the protein, aiding in folding. Spectroscopic and computational data supported Ni binding to occur in this redesigned construct, adapting a distorted square planar geometry. More strikingly, the Ni^II-incorporated protein was found to catalyze proton reduction to H₂ under photochemical and electrochemical conditions, with a TOF of ~210 under photochemical conditions, comparable to those of reported biomolecular catalysts.⁵⁶¹

Figure 63.

Redesign approach for the conversion of Csp1 (a and b) to a Ni-binding protein (NBP). Copyright NAMD model of apo NBP. (d) Overlay of the X-ray structures of apo Csp1 (purple, PDB ID: 5FJD) and Cu^I-Csp1 (cyan, PDB ID: 5FJE) and the minimized structure of apo NBP (green). Ball and stick model of the tetrathiolate site of minimized NBP viewed perpendicular to the helical axis (e) and from the top of the helical axis (f) with N-terminus at the top. Modeling performed using CHARMM22 force fields. Reprinted with permission from refs 560 and 562. Copyright 2019 and 2015 American Chemistry Society and Nature Publishing Group.

Also taking advantage of a naturally Cys-rich scaffold, the Shafaat group has developed a nickel-substituted rubredoxin (NiRd) and further probed the secondary coordination sphere of this construct to shed light on how the secondary sphere can modulate the hydrogenase activity.⁵⁶³ A rationally designed library of 30 mutants were generated in the metal-binding loops of Rd targeting residues V08, V34, and V37 (Figure 64), which exhibit H-bonding interactions with the Ni-coordinated cysteines through the amide backbone. By varying the identity of these residues, it was found that the strength of these H-bonding interactions is reflected in changes of the Ni^III/II reduction potential, in which stronger hydrogen bonds between the amide proton and sulfur atom of Cys were correlated with higher reduction potentials, likely due to the increased electronic delocalization and stabilization of the reduced state. Further-more, it was found that the steric constraints imposed by certain side chain residues are capable of influencing both H-bonding with the amide backbone and solvent accessibility. Solvent accessibility is a key factor in proton-dependent H₂ evolution reactions. Specifically, the greatest increases in activity correlated with mutations of the V08 residue. The TOF and solvent accessibility of V08X mutants increased relative to WT-NiRd due to the rotation of side chains away from Cys35, improving the solvent accessibility of the active site. Meanwhile, the solvent accessibility and catalytic performance of V34X mutants were comparable to those of WT-NiRd, with the exceptions of V34G and V34H, which reduced the degree of H-bonding between the amide backbone and nearby Cys32 due to substantial movement of the backbone. The dependence of the TOF on pH was found to correlate with secondary sphere perturbations as well. The V34H mutant showed a logarithmic dependence of reactivity on pH. MD simulations demonstrated that protonation of His34 at low pH resulted in a rotation of the His side chain away from the metal binding site, enabling direct access of water to the active site. This extensive work from the Shafaat group with a simple NiRd scaffold demonstrates the impact of the secondary coordination sphere in tuning both catalysis and reaction mechanism.⁵⁶³

Figure 64.

Structures of the (a) Dv MF NiFe H₂ase and (b) Dd NiRd active sites including key secondary sphere interactions. Reprinted with permission from ref 563. Copyright 2019 American Chemistry Society.

Besides H₂ase mimics, the Shafaat group recently reconstructed ferrodoxin using a [NiFe₃S₄] cluster to form a protein model of CODH.⁵⁶⁴ While native ferrodoxin is capable of binding CN⁻, only the [NiFe₃S₄] reconstituted enzyme was found capable of binding CO, as supported by an extensive spectroscopy study. Further systematic studies of the PCS and SCS residues of surrounding this ferredoxin-based model system will provide even greater insights into the mechanism of native CO₂ reduction by CODH.

The Baker lab has also successfully redesigned a mononuclear Zn metalloenzyme for organophosphate hydrolysis with the assistance of applied computational tools.⁵⁶⁵ Starting from an adenosine deaminase scaffold (PT3), eight mutations were introduced to help establish a Zn coordination site as well as stabilize a postulated transition state geometry for hydrolysis based on an extracted set of mononuclear Zn enzyme scaffolds. Based on these calculations, residue Glu217 was suggested to be crucial for proton shuttling in the deamination reaction. These calculations were supported by the low activity observed for the E217Q variant, highlighting the importance of Glu217 for organophosphate hydrolysis, which was suggested to aid in fine-tuning substrate alignment at the active site. Directed evolution was further performed to enhance the hydrolysis reactivity toward organophosphate substrates, resulting in the develop-ment of variant PT3.1, with approximately 2500-fold higher reactivity than the initial PT3 variant. Based on these results, it was postulated that substitutions in the vicinity of Glu217, such as V218F and Q58V, may increase the hydrophobicity around the putative catalytic base Glu217 (Figure 65). A change in pH of 1.6 accounts for the catalytic increase observed in PT3.1, containing V218F mutation relative to PT3, which suggests that the basicity of Glu217 is related to the activity enhancement obtained through the directed evolution pathway.

Figure 65.

Structure of engineered organophospate hydrolase PT3.1. Residues 58 and 218 influencing the hydrophobicity of Glu217 are shown in sticks (PDB ID: 3T1G).⁵⁶⁵

6.3. Design Using Strong Noncovalent Interactions

The biotin–avidin interaction is known to be one of the strongest noncovalent interactions in nature,⁵⁶⁶ and the use of this strong interaction to construct ArMs by connecting avidin-based protein scaffolds with biotin-substituted artificial metal-locofactors can be dated back to the last century.^475,567 In 2003, the Ward group reported an artificial metalloenzyme (ArM) using Sav and a Rh complex with a biotinylated amino-diphosphine ligand Biot-1 ([Rh(COD)(Biot-1)]⁺⊂Sav, Scheme 4, bottom) capable of achieving 92% ee (R) when catalyzing the enantioselective hydrogenation of α-acetamidoacrylic acid (Scheme 4, top). When mutating the Ser112 of the L7,8 loop of Sav (Figure 66), which is in the proximity of the metal active center, into glycine, an increase in enantioselectivity for the ArM [Rh(COD)(Biot-1)]⁺⊂Sav S112G to 96% (R) was observed.⁵⁶⁸ In addition to S112G, the group also studied the V47G, K80G, and P64G mutations. Val47 forms a close contact with Ser45, which interacts with the H-N group of biotin through H-bonding. Meanwhile, Lys80 is a neighboring residue of Trp79, which is a SCS residue of biotin, forming a hydrogen bond with Thr90, which further interacts with the thioether of biotin (Figure 66). Although Pro64 is located on a loop that is relatively remote from the biotin, mutations at this residue may have an impact on the loop geometry. It was found that mutations closer to the active site (V47G and S112G) had a more significant effect on the enantioselectivity compared to those that are relatively remote (K80G and P64G).⁵⁶⁹ Site saturation mutagenesis (SSM) on Ser112 in the same ArM system catalyzing α-acetamidoacrylic acid or α-acetamidophe-nylacrylic acid provided more insights into this site. Generally, the ArMs having S112A and S112G mutations more heavily favor the R products, while those with S112H, S112K, and S112R mutations favor producing S products. By modifying the artificial cofactor itself, a further increase in the ee of both R and S products was achieved. The authors speculated that the cationic side chains of residue 112 can interact with the carboxylate moiety of the substrate through hydrogen bonds, making one of the prochiral faces of the substrate more favorable for coordination.⁵⁷⁰ A directed evolution approach using the same Biot-1 ArM system for catalysis of the α-acetamidoacrylic acid ester substrate found that several residues ~4–6 Å away from the metal center can change the ee to favor either R or S isomers, possibly by directly affecting the conformation of the catalyst or catalyst/substrate complex. Among these, N49V increased the ee from 23% (R) to 65% (R), and N49H/L124F decreased the ee from 23% (R) to −7% (S).⁵⁷¹

Scheme 4. Enantioselective Hydrogenation of α-Acetamidoacrylic Acid and Structure of Biot-1⁵⁶⁸.

Figure 66.

Crystal structure of a wild-type Sav with bound biotin with the relevant residues show (PDB ID: 1STP).⁵⁷² The protein backbone is shown in green, and biotin C atoms are shown in white, relevant residues with direct interaction with biotin in cyan, and relevant residues distal to biotin in magenta.

Incorporation of a racemic biotinylated three-legged d⁵/d⁶-piano stool complex (Scheme 5) into Sav enabled the formation of a series of artificial transfer hydrogenases (ATHases). Starting from the Ru d⁶-piano stool complex, several mutations were made to study the influence of the secondary coordination sphere on ketone hydrogenation. S112G, a mutation close to the catalytic site, resulted in the highest conversion (90%) but lowest enantioselectivity (28% ee (R)). P64G, a mutation at a relatively remote site, has the greatest enhancement on enantioselectivity (72% ee (R)) but moderate conversion (54%). The double mutant P64G/S112G combined the positive effects of both single mutants, resulting in excellent conversion (95%) and reasonable enantioselectivity (58% ee (R)).⁵⁷³ Later, a series of alternative d⁵/d⁶-piano stool complexes coordinating Rh, Ru, and Ir, respectively, were screened for activity toward the reduction of ketones or cyclic imines. Select complexes were further developed by SSM to generate more advanced ArMs. Variants of Sav containing cationic residues at position 112 (S112R and S112K) generally have higher S-enantioselectivity. It was also found that the introduction of aromatic residues (i.e., S112F and S112Y) or using small amino acids at the same site (i.e., S112G and S112A) leads to higher R-enantioselectivity.^574,575 Based on the S112A and S112K mutations, a follow-up study was performed focusing on the SSM of the Lys121 and Leu124 sites. Sav is a homotetrameric protein, in which the Lys121 of one monomer interacts with the η⁶-arene ring of the complex, while the Lys121 from the adjacent monomer interacts with the incoming substrate (Figure 67a). Therefore, this residue may play a role in favoring certain prochiral substrate interactions. Leu124 may influence the biotinylated catalyst slightly because its methyl group is in close proximity (3.53 Å) with the sulfone group of the catalyst. Overall, SSM of Lys121 proved to be more effective for the optimization of enantioselectivity than when applied to Leu124.⁵⁷⁶ The crystal structure of S112K Sav containing Ru-cofactor clearly showed interaction between Lys112 and Leu124. Also, several other residues close to the active site, including Thr114 and Pro64, were investigated, where mutation of either to Gly was found beneficial in enhancing the enantioselectivity. Thr114 is close to the η⁶-arene ring, while Pro64 is located in the L4,5 loop, which neighbors the L7,8 loop from another monomer, the latter of which is close to the catalytic metal center (Figure 67a). Mutation of the Pro64 to Gly hypothetically perturbs the location of the L4,5 loop, which in turn should influence the positioning of the L7,8 loop and further influence the Ru complex.⁵⁷⁷

Scheme 5.

Formation of a Series of Artificial Transfer Hydrogenases^574,575

Figure 67.

(a) Crystal structure of a Ru complex-based Sav ArM tetramer showing the positions of relevant protein residues and loops (PDB ID: 2QCB):⁵⁷⁶ monomer A (orange); monomer B (cyan); monomer C (green); monomer D (magenta); Ru complex (white). (b) Crystal structure of Ir complex-based Sav-FPD S112V-K121A (PDB ID: 6GMI) overlaid with Ir complex-based Sav S112A (PDB ID: 3PK2): monomer A of Sav-FPD S112V-K121A (blue); monomer B (green); monomer C (lilac); monomer D (cyan); Sav S112A (gray); mutated residues S112V and K121A (magenta). Reprinted with permission from ref 579. Copyright 2018 American Chemical Society.

There are also some ATHases that have been constructed based on similar Sav-bound biotinylated Ir complexes. Several cationic and anionic residues around the metal complex were investigated using imine reduction as a model reaction. Both S112A/K121A and S112A/K121L displayed excellent catalytic efficiency (k_cat/K_M). An additional D67V mutation to the S112A/K121L construct was made to remove a negative charge near the catalytic site, resulting in a lowering of the catalytic efficiency due to a lower k_cat. A possible explanation for this effect is that the reduction of imine may proceed through an iminium ion intermediate, and removal of the cationic residue (K121A or K121L) next to the catalytic site would help with the catalytic process. The presence of an anionic residue (Asp67) would stabilize the cationic transition state and increase the k_cat.⁵⁷⁸

To further modulate the ATHase activity in the Ir-complex-based Sav, an additional 24-residue helix-turn-helix motif (FPD) was incorporated into the L7,8 loop and combined with other previously investigated point mutations to tune the catalytic performance of the ArM. The two best-performing mutants found for the reduction of a quinolium substrate were S112A/K121P Ir-Sav-FPD and S112V/K121A Ir-Sav-FPD. The crystal structure of S112V/K121A Ir-Sav-FPD revealed that the L4,5 loop was shifted 5.6 Å compared with that of Sav S112A, and the directionality of the L7,8 loop had rearranged to point toward the Ir cofactor, although the FPD-loop was not resolvable (Figure 67b). The cofactor occupancy for S112V/K121A Sav-FPD was also found to increase relative to that for S112V/K121A Sav (from 50% to 80%, indicating that the FPD-loop shielded the biotin-binding vestibule and helped the binding of the complex.⁵⁷⁹ Since Sav is a tetramer hypothetically capable of binding up to four biotinylated cofactors, the roles for different monomeric units were also investigated. To this end, two Sav subunits were fused together and mutated separately at the respective Ser112 and Lys121 positions. It was found that mutations made in the first subunit (Sav_A) had a greater impact on both substrate conversion and product enantioselectivity. Meanwhile, those made in the second subunit (Sav_B) do not show a clear trend for the influence on the reactivity; however, these residues do still appear to play a role in fine-tuning substrate interactions, as well as conversion and enantioselectivity. This likely arises from different conformations of the Ir complex in each subunit, with Sav_A serving as the predominately active form.⁵⁸⁰

The [Cp*RhCl₂]₂ (Cp* = pentamethylcyclopentadienyl) complex is a versatile catalyst for electrophilic aromatic C-H activation.⁵⁸¹ With the aim of introducing asymmetric activity and improving conversion, Hyster et al. synthesized a biotinylated [Cp*RhCl₂]₂ complex ([Cp*RhCl₂]₂-biot) and successfully incorporated the complex into Sav.⁵⁸² The low activity observed in the resulting ([Cp*RhCl₂]₂-biot)-Sav construct inspired further modification of the secondary coordination sphere based on molecular docking simulations. Initial efforts introduced a basic residue in the vicinity of the metal center through S112E and K121E mutations. Similar to previous reports,^570,574,575 both of these variants exhibited low conversion for the reaction (10% and 7%, respectively). However, mutation of K121 to Asp resulted in a significantly improved reaction conversion of up to 89%, demonstrating that both the presence of a carboxylate and its precise positioning can significantly impact aromatic C-H activation. Inspired by these results, the authors further added a N118K mutation to K121E ([Cp*RhCl₂]₂-biot)-Sav (Asn118 is adjacent to Lys121), and they observed a further increase in conversion to 99%, as well as an excellent regioisomeric ratio (rr) (15:1) and enantiomeric ratio (er) (82:18). By screening different mutations at the Ser112 site, S112Y was found to have the best enantioselectivity, with an er of 88:12. The synergistic effect of combining S112Y and K121E mutations also led to a total conversion of up to 95%, rr of 19:1, and er of 91:9. Through kinetic isotope effect (KIE) and competition studies, the authors found that the introduction of the carboxylate had the most dominant effect on improving the activity and selectivity.⁵⁸²

Further expanding the scope of the biotin-Sav guest–host scaffold, Ward and co-workers incorporated a biotinylated second-generation Hoveyda–Grubbs catalyst into Sav to generate a novel artificial metathase. Importantly, this ArM system was developed to function in whole cells as a catalyst for in cellulo abiotic transformations. Directed evolution was applied to enhance the activity of this ArM, and a quintuple mutant (V47A/N49K/T114Q/A119G/K121R) was found to increase the cell-specific activity by 5.4-fold. According to the B factors associated with the crystal structures of the wild-type and quintuple mutant, the biotin-vestibule loop of the quintuple mutant exhibited greater flexibility, with mutations T114Q, A119G, and V47A serving as the key mutations for this enhanced flexibility.⁵⁸³

Ward and co-workers further explored the use of ArMs as in vivo new-to-nature catalysts by developing a method for displaying Sav on the surface of E. coli. It was reasoned that a surface display would more readily bind designed biotinylated metallocofactors, as diffusion of the complex through the periplasm is not required. A biotinylated CpRu (2-quinoline-carboxylate) complex was successfully incorporated into surface-displayed Sav with the intention of catalyzing the uncaging of allylcarbamate-protected substrates, generating a functional allylic deallylase. SSM at positions in close proximity to the Ru complex (S112 and K121, vide supra) revealed two double mutants, S112Y/K121S and S112M/K121A, with activities 25- and 24-fold higher than that of the wild-type, respectively. The crystal structure of the S112M/K121A mutant showed that the aromatic quinoline ring was stabilized by a methionine CH₃/π-interaction in S112M. Also, the carbonyl groups from K121A and Ser122 helped the binding of the quinoline ligand.⁵⁸⁴

Beyond Ru, Rh, and Ir, the biotin–avidin method has also been used to create ArMs with many other different metal complexes, including Pd for asymmetric allylic alkylation⁵⁸⁵ and Suzuki-Miyaura cross-coupling,⁵⁸⁶ Au for hydroamination and hydroarylation,⁵⁸⁷ and OsO₄ for cis-dihydroxylation of olefins.⁵⁸⁸ Mutagenesis investigations have been conducted for all of the examples illustrated above, focusing predominately on varying residues Ser112 and Lys121. In the Suzuki-ase ArM generated by incorporation of a Pd complex (Scheme 6) into Sav, the crystal structure of the S112Y/K121E mutant was acquired, revealing that S112Y provided a significant steric barrier on the lower face of Pd, likely accounting for the modest improvement observed in the enantioselectivity. Additionally, an H-bonding interaction was found between Tyr112 and Glu121, further locking the conformation of Tyr.⁵⁸⁶

Scheme 6.

Suzuki-ase ArM Generated by Incorporation of a Pd Complex into Sav⁵⁸⁶

Aside from the strong biotin–avidin interaction for constructing ArMs, vanadium pentahydrate was found to interact with the Sav host in acidic media through H-bonding with H-donor residues that are also critical for biotin binding (e.g., Asn23, Ser27, Ser45, and Asp128). An ArM catalyzing enantioselective sulfoxidation reaction was, hence, constructed (Figure 68), and this artificially generated vanadyl enzyme successfully catalyzed the oxidation of prochiral sulfides with good enantioselectivities of up to 93% ee.⁵⁸⁹

Figure 68.

Superimposition of the docked structure [VO(H₂O)₅]²⁺ (ball-and-stick representation) with WT-Sav (monomers A and D, light blue schematic secondary structure) with the structure of biotin with WT-Sav (PDB ID: 2IZG): biotin (yellow stick and yellow transparent surface); monomers A and D (yellow schematic secondary structure). Reprinted with permission from ref 589. Copyright 2008 American Chemistry Society.

6.4. Engineered Metalloenzymes by Covalent Attachment of Metallocofactors

In addition to the noncovalent anchoring approaches of scaffolds, such as the biotin–avidin method, ArMs can also be generated through covalent anchoring of synthetic catalysts. For example, azide-alkyne cycloaddition (the classic “click” reaction) has been employed to anchor an alkyne-bearing dirhodium complex in propyl oligopeptidase containing the UAA 4-azidophenylalanine (POP-Z). Although initial reactions of otherwise unmodified POP-Z failed to incorporate the cofactor, mutations aimed at reducing the sterics at the end of the β-barrel were found to enable bioconjugation in POP-ZA₄. Further mutations of residues 99 and 594 to Phe, which neighbor the distal Rh based on homology modeling, were found to increase the yields from 61% to 74% and the ee from 85% to 92%,⁵⁹⁰ despite later crystal structures that showed the Phe99 was not positioned as suggested by homology modeling.⁵⁹¹ Random mutagenesis was also applied to POP-ZA₄ to generate several mutations which improved the enzymatic activity. Among these, only three mutations, G99S/S301G/Y326H (1-SGH-POP-ZA₄), were required to reach optimal selectivity. Based on crystallographic analysis of 1-SGH-POP-ZA₄, Ser99 and His326 were found in the active site while Gly301 was farther away.⁵⁹¹ 1-SGH-POP-ZA₄ was further subjected to directed evolution targeted at improving the catalytic performance of diazo coupling cascade reactions. Site saturating mutations were introduced in the β-strand across from the putative Rh-binding His to examine mutations that could significantly impact the product yield and selectivity. From this effort, mutations Q98P and S99H were identified, and a further combination of combinatorial codon mutagenesis across 25 active site residues followed by another round of SSM identified residues V71G and E283G to generate a final mutant with product yields as high as 76%.⁵⁹²

Thiol-maleimide bioconjugation has also been used to generate 4d/5d metal containing ArMs. C96 nitrobindin was employed to generate an ArM capable of C(sp²)─H bond functionalization via concerted metalation deprotonation (CMD) with 15% yield and 92:8 regioselectivity through the covalent binding of a maleimide-tagged Rh complex. Carboxylates are known to facilitate the CMD-type C(sp²)─H bond functionalization, and therefore, glutamate residues were introduced near the active site in a L100E/A125E mutant, which increased the yields to 40% with similar regioselectivity (90/10).⁵⁹³ Directed evolution was later applied to this system, and the T98H/L100K/K127E mutant was identified to increase the activity 2.2-fold. Interestingly, MD simulations suggested that all three of these residues are located opposite to the Rh metal center.⁵⁹⁴

Human carbonic anhydrase II (hCA II) exhibits a high binding affinity toward para-substituted aryl-sulfonamides at the Zn binding site, providing another tool for the development of ArMs. An {IrCp*}²⁺ moiety was introduced into hCA II to generate a novel ATHase. Structural analysis based on the resulting X-ray structure showed that Ile91 and Lys170 occupy a potential substrate-binding site. Therefore, I91A and K170A mutations were incorporated, with the former showing some increased activity for specific substrates; however, the apparent low occupancy of the catalytic complex precluded significant improvements.⁵⁹⁵ Computational redesign using the program Rosetta was applied to tackle the problem of low cofactor occupancy.⁵⁹⁶ It was speculated that the active site of WT-hCA II was not well defined toward the Ir complex. To solve this problem, a variety of hydrophobic residues were introduced to tune the shape of the binding site and optimize the packing interactions between the protein and cofactor. N67W, E69Y, and Q92F mutations were made to β-strands 2 and 3 to allow interactions between the biaryl and phenylsulfonamide groups. An L140 M substitution on β-strand 5 and L197 M on loop 7,8 were also made to pack against the coplanar biaryl moiety of the ligand, and several additional mutations were made to stabilize the protein backbone. L60V on β-strand 1 was introduced for stabilization, A65T was made on β-strand 2 to enable H-bonding with the carbonyl oxygen of Phe95, and C205S was introduced at the end of loop 7,8 to add rigidity to the loop via H-bonding with the carbonyl oxygen of Val134, maintaining the orientation of the cofactor (Figure 69). Based on these mutations, one of the variants d3 (containing L60V, A65T, N67W, E69Y, Q92F, L197M, and C205S) displayed a 64-fold increase in binding affinity for the Ir cofactor. Reverting selected residues such as E69Y (d2) or L197 M (d1) resulted in some reduction in the binding affinity, although still higher than that in the WT. All three mutants (d1, d2, and d3) showed improved ee and higher TONs than with the wild-type protein.⁵⁹⁷ This ArM system was later applied to another similar IrCp cofactor using a dual-anchoring system to further increase the occupancy and stabilize the geometric configuration.

Figure 69.

Model of Rosetta design on the crystal structure of the Ir complex in WT-hCA II (PDB ID: 3ZP9). Designed backbone-stabilizing H-bonds are represented as green dashed lines. Reprinted with permission from ref 597. Copyright 2015 American Chemical Society.

Directed evolution based on the dual-anchoring mutant I91C was conducted at the Leu60, Asn62, Asn67, Glu69, and Gln92 sites, all of which neighbor the terminal chloride of the Ir complex. All mutations at Gln92 led to decreases in activity, possibly due to the elimination of an optimal H-bonding interaction between Gln92 and the amide oxygen of the cofactor, which serves a crucial role in fixing the Ir cofactor in the active site and maintaining the pyridine–sulfonamide linkage. For purified protein, N67G/E69R/I91C displayed the highest activity toward the R-enantioselectivity of the substrate with a TON of 451 and ee of 96%. For S-enantioselectivity, N67L/E69Y/I91C and L60W/N67L/E69Y/I91C reached moderate ee values of −62% and −49%, respectively (TONs 458 and 221, respectively). In comparison, the starting I91C construct exhibited a TON of 265 and ee of −39%. The L60W/N67L/E69Y variant alone enhanced the activity by 4.3-fold as well as improved the S-enantioselectivity to 59% ee. From the crystal structures of N67G/E69R/I91C and I91C, it was surmised that N67G provided more space for the substrate, allowing a more facile approach to Ir-H, while E69R may serve to introduce a cation–π interaction between the residue and the substrate. As a result, the above two mutations together may aid in substrate binding to the active site, leading to the increase of the TON and enantioselectivity.⁵⁹⁸

The HaloTag (HT) protein utilizes a reactive aspartic acid in its binding pocket to react with haloalkanes, forming a covalent linkage. Taking advantage of this, Fischer and co-workers developed a metathase by linking a Ru complex to HT, which was shown to exhibit enhanced TONs when compared to the catalyst alone. Directed evolution was applied to residues around the binding pocket; however, no significant improve-ments in the activity were observed. This is likely due to the long distance between the cofactor and the protein scaffold caused by the length of the linker or possibly due to the minimal interactions between the SCS and the catalytic cofactor transition state.⁵⁹⁹

7. SUMMARY AND OUTLOOK

In this review, we have showcased how catalysis can be modulated in a wide range of designed ArMs by focusing on the roles of the environment beyond the PCS. While we have covered many examples involving the generation of ArMs through the redesign of existing PCSs or the introduction of a new PCS into a native protein scaffold, most of these ArMs display much lower activities in comparison to their native enzyme counterparts. The most impressive examples with high catalytic activities, including several that approach or exceed those of native enzymes, have arisen from those that have further employed SCS modifications to optimize properties such as the reactivity and substrate selectivity.

Throughout this review, we have seen both rational design and directed evolution used as effective tools to impart and enhance new reactivities in ArMs. At face value, the ethoses of rational design and directed evolution appear at odds, in-so-far as the first insists on the question “How?”, while the second avoids it. However, these two strategies can be powerfully used to complement one another and overcome their own respective limitations. In particular, the further optimization of rationally designed ArMs by directed evolution can uncover subtle SCS roles that may otherwise be easily overlooked. Likewise, the use of rational design to analyze ArMs directed by evolution can provide new insight to overcome barriers associated with limitations in screening methods and to design smaller, higher quality libraries.⁶⁰⁰ When synergized with computational methods, particularly those available for initial scaffold design, such as Rosetta^61,62 and AlphaFold,^63,64 and those for analysis, such as QM/MM methods, both our understanding of the roles of SCS residues and the practical development of ArMs can be greatly advanced. Doing so may not only enable further development of the systems discussed in the present review but also open doors to the design of ArMs modeling metalloenzymatic systems with more complex cofactors and more challenging reactions, such as photosystem II, nitrogenases, and methane monooxygenase.

Due to the prevalence, long history of study, and biological significance of heme enzymes, heme-based ArMs represent one of the most thoroughly studied classes in this review, particularly regarding the roles of SCS residues. Engineered heme ArMs generally fall into one of two groups: (1) nonenzymatic native scaffolds, such as Mb and Cc, that are engineered to generate new activities and (2) enzymatic scaffolds modified to exhibit activity toward non-native substrates or non-native functions. Additionally, three strategies have been employed to study and enhance catalysis in these systems. The first approach is to mimic the structural features of a desired native enzyme, such as the His-Arg pair or the positioning of the distal His in peroxidases.^101,122 The second is to employ SDM at SCS residues surrounding the active site to vary structural features such as sterics, polarization, and H-bonding to tune the activity. The third entails using directed evolution to repurpose a native heme protein for new, non-native functions. This is often accomplished through SDM of residues identified as “hot spots” near the active site, although it is worthwhile to note that directed evolution is also effective at identifying beneficial mutations that are surprisingly remote.^569,573 These strategies have resulted in the generation of several highly competent catalytic heme ArMs, with some of them approaching,⁵³⁹ rivaling,⁵²⁵ or even exceeding⁵²⁴ those of native enzymes. Through this process, we have learned not only how the presence of SCS features, such as H-bonding, influences catalysis but also how the positioning of the same SCS residue in different scaffolds (e.g., distal His in Mb and CcP) can play a key role in both conferring and fine-tuning activity. However, the practical development of these enzymes would benefit from further developments in terms of scaffold stability, and further optimization of these systems for varying activity stands to benefit from the more extensive application of computationally aided design and screening.

It is interesting that DNA has been used as an alternative scaffold for designing artificial heme enzymes. However, current DNA models do not support proximal ligation to the heme by the scaffold molecule, which limits the reactivity of heme/DNAzymes compared to heme enzymes. Extending the oligonucleotide beyond the G-quartet/heme interface provides a structure analogous to the active site of heme proteins;²⁴⁹ however, the much more disordered distal environment of a G-quadruplex scaffold is very different from typical protein scaffolds. This provides an opportunity for developing an active site de novo—importantly, as the DNA scaffold can be developed by in vitro selection, active sites made from pendant nucleotides can be selected against one another without the issues of expression or purification that can otherwise arise when employing directed evolution for proteins.

The PCSs of native mononuclear non-heme Fe enzymes predominately utilize the facial (His)₂-carboxylate triad, which has served as the major inspiration for initial designs of non-heme Fe ArMs. However, despite their prevalence in nature, relatively few non-heme Fe ArMs have been designed which employ natural amino acid side chains/protein backbone for metal ion coordination,^{268,246-248,274} at least when compared to those for other metals such as Cu, Ni, and Zn. As a result, the study of SCS interactions in non-heme Fe ArMs has not advanced to the extent discussed in some other sections, underlining the need for continued developments in this field, particularly toward mono- and dioxygenase activity. Similarly, the employment of the UAA Ala_Bpy has been effective in forming Fe^II binding sites (especially when combined with computational design), but UAA-based systems have yet to be turned into efficient catalysts.^{255-258,286,287} Meanwhile, non-heme Fe ArMs formed using a protein scaffold to host an organometallic cofactor have proven to be the most well-characterized in terms of catalytic competence, particularly through the efforts of Ménage and co-workers.^{302,305,306,308} In these systems, SCS residues have primarily been investigated to support fixation and positioning of the organometallic scaffold through π-stacking and salt bridge interactions to enhance the specificity and selectivity. Lastly, the development of dinuclear non-heme Fe sites in native scaffold-based ArMs has largely been inspired by FeFe H₂ase,^291-295 although little further development has been made in these systems to understand the role of the peptide scaffold beyond the PCS. Borovik and co-workers have developed a notable exception, forming a non-heme diiron site via bridging of Sav active sites that is reminiscent of native oxygenases, although further catalytic studies are yet to be presented.³¹⁷ Taken together, non-heme Fe ArMs present a promising field for continued study, particularly through more expansive investigations of the roles of SCS residues in modulating the properties of Fe.

Cu proteins serve diverse roles in nature, including ET and various catalytic roles. Two types of Cu sites natively serve in ET (T1Cu and Cu_A), both of which have been studied by the development of cupredoxin units that are isolated from native ET systems, as well as by incorporation into nonhomologous scaffolds. By modulating the SCS to vary factors such as hydrophobicity and hydrogen bonding, the ET properties of these centers can be effectively optimized by balancing E° with reorganization energy.^{349-355,392,393} The next development of these centers requires consideration of how to effectively couple these ET active proteins with catalytic ArMs to enhance enzymatic activity, including the realization of specific surface interactions. Additionally, SCS development optimizing pathways for long-range electron transfer offer a tantalizing challenge, requiring not only a well-balanced electrochemical gradient but also possible involvement of proton donor/acceptors.

Catalytic Cu centers mimicking natural reactivities present a greater challenge. Several reactions with small inorganic molecules have been demonstrated in Cu-based ArMs, including nitrite reduction,⁴⁵⁷ superoxide dismutation,⁴⁵² S(Cys) peroxidation,^456,464 and S-nitrosylation.²⁷⁶ In these examples, studies of the SCS have allowed the electrostatic environment and E° of the Cu active site to be tuned. The oxidation of organic substrates has also been reported, where SCS interactions have been employed to control protein–substrate interactions. Furthermore, artificial Cu proteins have been repurposed to catalyze a number of unnatural reactions including the D-A reaction,^478,487 F-C alkylation,³⁶⁸ Michael addition,⁴⁹⁷ and alkene hydration.⁴⁶⁸ In these cases, the SCS has been employed to enforce product selectivity via substrate–SCS interactions. It is worth noting that Cu centers have often been treated as Lewis acids, and their natural roles as oxidoreductases are surprisingly less studied. In particular, oxidative reactivity has been more difficult to achieve than reductive chemistry (in large part due to the damage that can occur to the protein scaffold from ROSs), despite numerous examples of Cu-based oxidases and oxygenases in nature. Therefore, much remains to be explored in the area of Cu ArMs by further consideration of SCS effects.

In addition to Fe and Cu metalloenzymes, there are many ArMs that have employed other biologically relevant metal cofactors, such as V,⁵⁸⁹ Cr,⁵⁴⁰ Mn,^540,541,547 Co,^549,551 Ni,^560,563 Zn,⁵⁶⁵ and Mg,⁵⁰⁸ as well as those with abiological metal centers, including Ru,⁵⁷⁷ Rh,^{568,581,590,594} Pd,⁵⁸⁶ Os,⁵⁸⁸ Ir,^{558,559,579,597} and Au.⁵⁸⁷ These studies can be divided into two categories, namely those that target understanding native metalloenzyme functionality and those that aim to utilize the ArM scaffold to enhance the catalytic properties of organometallic complexes. The role of the SCS is key to both of these aims, functioning both to tune the metal site and control substrate–metal interactions through varying factors including H-bonding, steric hindrance, hydrophobicity, and pK_a. Future studies using directed evolution to investigate the SCSs of novel ArMs could benefit from more thorough structural and mechanistic studies of the evolved enzymes to rationalize the specific influence of varying residues on ArM performance for each sequential round of selection. By tracing the evolutionary path, greater insight could be gained on how to further engineer ArMs with enhanced performance through tuning the environment beyond the PCS.

SCS interactions are often weak and difficult to pinpoint, with many effects remaining nonobvious even for experienced ArM designers and too subtle to be reliably predicted by current computational methods. To overcome these limitations, the study and development of ArMs stand to benefit from machine learning approaches. To achieve such goals, a large and well-defined set of structural and activity data for each target metalloenzyme is required. Since it is much easier to produce these data sets using ArMs, designed ArMs can play a key role in advancing this field. Given the rapid progress made in data collection and improvement of machine learning algorithms, it would be possible to routinely incorporate the best combinations of PCS and SCS features to design any metalloenzyme with desired activity.

ABBREVIATIONS

AA

natural amino acid

ABTS

2,2′-azino-bis(3-ethylbenzothiazoline-6-sulfonic acid diammonium salt

ACCO

1-aminocyclopropane-1-carboxylate oxidase

ACS

coenzyme-A synthetase

adt

-SCH₂NHCH₂S-

Ala_Bpy

(2,2′-bipyridin-5-yl)alanine

AMO

ammonia monooxygenase

apo-Cc

apo-cytochrome c

APX

ascorbate peroxidase

ArM

artificial metalloenzyme

ATHase

artificial transfer hydrogenase

ATRP

atom transfer radical polymerization

AxNiR

Alcaligenes xylosoxidans nitrite reductase

Az

azurin

β-LG

β-lactoglobulin

biot

biotin

Bph

biphenyl

BPP

bovine-pancreatic polypeptide

bpy

bipyridine

CA

carbonic anhydrase

cAMP

cyclic adenosine monophosphate

camSOD

cofactor-flexible superoxide dismutase

CAP

catabolite activator protein

Cc

cytochrome c

Cc₅₅₂

cytochrome c₅₅₂

CcP

cytochrome c peroxidase

CcO

cytochrome c oxidase

CiP

Coprinus cinereus peroxidase

CLEC

cross-linked enzymatic crystal

ClTyr

3-chlorotyrosine

CMD

concerted metalation deprotonation

CODH

carbon monoxide dehydrogenase

CoPP(IX)

cobalt-protoporphyrin IX

COx

catechol oxidase

Cp

Clostridium pasteurianum

Csp1

copper storage protein

Cu_AAz

Cu_A azurin

Cu_BMb

myoglobin containing a designed Cu_B binding site

CueO

copper efflux oxidase

CuNiR

copper-dependent nitrite reductase

Cgb

cytoglobin

D-A

Diels–Alder reaction

DF

diformyl

DFT

density functional theory

DCFH-DA

2′,7′-dichlorofluorescin diacetate

DHP

dehaloperoxidase

DNA

DNA

DNIC

dinitrosyl iron complex

dpa

bis(2-pyridylmethyl)amine

DTB-SQ

3,5-di-tert-butylsemiquinone

DTT

dithiothreitol

DyP

decolorizing peroxidases

E°′

reduction potential

EDA

ethyl diazoacetate

EDTA

ethylenediaminetetraacetic acid

ee

enantiomeric excess

E _m

midpoint potential

ENDOR

electron double nuclear resonance spectroscopy

ESI-MS

electrospray ionization mass spectrometry

ET

electron transfer

EXAFS

extended X-ray absorption fine structure

EPR

electron paramagnetic resonance spectroscopy

F2Tyr

3,5-difluorotyrosine

F3Tyr

2,3,5-trifluorotyrosine

F-C

Friedel–Crafts alkylation

FdM

[Fe₄S₄]-binding oligopeptide

FeSOD

iron superoxide dismutase

FT-IR

Fourier transform infrared spectroscopy

GQ

G-quadruplex

H-bonding

hydrogen bonding

H-cluster

[Fe₂(CO)₃(CN)₂(adt)]

H₂ase

hydrogenase

hCA II

Human carbonic anhydrase II

HCO

heme-copper oxidase

HRP

horseradish peroxidase

HSA

human serum albumin

HT

HaloTag

IR

infrared

ICP

inductively coupled plasma

imiTyr

Tyr-His cross-link

IrMb

myoglobin containing the Ir(Me)-heme cofactor

KIE

kinetic isotope effect

L0

N-benzyl-N′- (2-hydroxybenzyl)-N,N′-ethylenediaminediacetic acid

L₅²_maleimide

1-(2-{[2-(bispyridin-2-ylmethylamino)ethyl]-pyridin-2-ylmethylamino}ethyl)pyrrole-2,5-dione

L _ibu

N,N′-bis(2-pyridylmethyl)-N-methyl-1,2-ethanediamine complex

LiP

lignin peroxidase

LMCT

ligand-to-metal charge transfer

LmrR

Lactococcal multidrug resistance regulator

LPMO

lytic polysaccharide monooxygenase

Mb

myoglobin

MCO

multicopper oxidase

MD

molecular dynamics

MF

monoformyl

MG

malachite green

MnCcP

Mn-bound cytochrome c peroxidase

MnP

manganese peroxidase

MnSOD

manganese superoxide dismutase

MtTyr

2-amino-3-(4-hydroxy-3-(methylthio)phenyl)-propanoic acid

N₂OR

nitrous oxide reductase

NB

nitrobindin

Ngb

neuroglobin

NiR

nitrite reductase

NiRd

nickel rubredoxin

NMM

N-methylmesoporphyrin IX

NMR

nuclear magnetic resonance

NO

nitric oxide

NOR

nitric oxide reductase

OMeY

3-methoxytyrosine

P411

cytochrome P411

P450

cytochrome P450

Pf

Pyrococcus furiosus

pMMO

particulate methane monooxygenase

pAF

p-aminophenylalanine

phen

phenylalanine

PT3

adenosine deaminase protein scaffold

POP

propyl oligopeptidase

POP-Z

propyl oligopeptidase containing the UAA 4-azidophenylalanine

PCS

primary coordination sphere

RDS

rate-determining step

Rma-Cc

bacterial cytochrome c from Rhodothermus marinus

Rd

rubredoxin

ROS

reactive oxygen species

RMSD

root mean squared deviation

RNA

ribonucleic acid

rR

resonance Raman spectroscopy

SA

serum albumin

Sav

streptavidin

SCP-2L

steroid carrier protein 2L

SCS

secondary coordination sphere

SDM

side-directed mutagenesis

SHE

standard hydrogen electrode

SKIE

solvent kinetic isotope effect

SNO

nitrosothiol

SOD

superoxide dismutase

SOH

sulfenic acid

SOR

superoxide reductase

SRS

substrate recognition site

SSM

site-saturation mutagenesis

STAMPS

synchronous, transactional, and asynchronous multiprocessing

swMb

sperm whale myoglobin

t-BuOOH

t-butyl-hydroperoxide

TAML

tetraamido macrocyclic ligand

TCP

2,4,6-trichlorophenol

TDHC

tetrahydrocorrin

TFAC

trifluoroacetophenone

THS

thermosome

TOF

turnover frequency

TOF MS

time-of-flight mass spectrometry

TON

turnover number

Trx

thioredoxin

TvNiR

T. nitratireducens cytochrome c nitrite reductase

TtCu_A

Thermus thermophilus cytochrome ba3 oxidase

UAA

unnatural amino acid

UV

ultraviolet

VT-MCD

variable temperature magnetic circular dichroism

WT

wild-type

XANES

X-ray absorption near-edge spectroscopy

Xln

xylanase

ZFS

zero-field splitting

Biographies

Casey Van Stappen completed his B.Sc. (2011) and M.Sc. (2013) degrees in Chemistry at the University of Minnesota, Duluth, under the supervision of Prof. Viktor Nemykin. Following several years of postgraduate study at the University ofMichigan, Ann Arbor, under the advisement of Profs. Nicolai Lehnert and Vincent Pecoraro, he transferred his studies to the Max Planck Institute for Chemical Energy Conversion (MPI-CEC, 2015), as a member of the IMPRS-RECHARGE, and completed his D.Sc. (2019) in Chemistry via enrollment at Ruhr-Universität Bochum under the advisement of Prof. Dr. Serena DeBeer, Dr. Eckhard Bill, and Prof. Dr. Frank Neese. Following a short period as a postdoctoral research associate, also at the MPI-CEC, he joined the group of Prof. Yi Lu at the University of Texas at Austin as a postdoctoral research fellow. His research interests lie in combining spectroscopic, biochemical, and computational methods to investigate small molecule activation and catalysis, particularly in biological and biologically inspired systems.

Yunling Deng grew up in Changchun, China, and received her B.Sc. (2014) in Chemistry from Nankai University, China (2014). She completed her Ph.D. (2020) in Bioinorganic Chemistry at Dartmouth College investigating the effects of heme propionates on the structural and redox properties of c-type cytochromes under the guidance of Prof. Ekaterina Pletneva. She next moved to the University of Illinois at Urbana-Champaign (and recently to the University of Texas at Austin) as a postdoctoral research fellow to work with Prof. Yi Lu, working on developing artificial metalloenzymes for CO₂ reduction.

Yiwei Liu was born in Zhengzhou, China, and received his B.Sc. (2018) from Peking University, China, under the supervision of Prof. Jun-Long Zhang. He joined Prof. Yi Lu’s lab at the University of Illinois at Urbana-Champaign (now at the University of Texas at Austin) in the same year. As a Ph.D. candidate in Inorganic Chemistry, his research focuses on engineering simple, noncatalytic proteins such as cupredoxins into artificial metalloenzymes to mimic the coordination chemistry and reactivity of natural enzymes.

Hirbod Heidari was born in Hamedan, Iran, and completed his Chemistry B.Sc. (2018) under the supervision of Prof. Sirus Jamali at the Sharif University of Technology, Tehran, Iran, along with a minor degree in Physics. In 2017, he conducted a summer-long internship under the supervision of Prof. Truhlar and the late Dr. Steven Mielke on Molecular Partition Function Computations. In 2019, he joined the lab of Prof. Yi Lu at the University of Illinois at Urbana-Champaign (now at the University of Texas at Austin) as a Ph.D. candidate. His research interests lie at the interface of computational chemistry and artificial protein engineering.

Jing-Xiang Wang obtained his B.Sc. (2019) from Peking University, Beijing, China, under the supervision of Prof. Jun-Long Zhang. After he joined Dr. Yi Lu’s lab at the University of Illinois at Urbana-Champaign in the same year, he moved to the University of Texas at Austin in 2021 with Dr. Yi Lu, continuing to pursuit his doctorate degree. His research interests lie in metalloprotein engineering and constructing artificial metalloenzymes. His current research is focusing on enhancing activities for metalloenzymes for further applications.

Yu Zhou was born in Jiangxi, China. He received his B.Sc. degree (2019) from Zhiyuan College, Shanghai Jiao Tong University, under the supervision of Prof. Xiaomin Xie and Prof. Zhaoguo Zhang and conducted a half-year internship with Prof. David Sarlah in 2018 (University of Illinois at Urbana-Champaign, UIUC) as a visiting undergraduate. After a three-year undergraduate research experience in copper- and palladium-catalyzed organic reaction development, he decided to join Prof. Yi Lu’s lab at UIUC (now at the University of Texas at Austin) and study biocatalysis. His current research aims at designing novel, efficient, and highly selective biocatalytic systems and metabolic networks.

Aaron P. Ledray received a B.S. (2015) in Biotechnology and Biomedical Sciences from the University of Central Florida under the supervision of Prof. Stephen J. King, where he studied neuro-degenerative disease. He completed his Ph.D. (2021) in Biological Sciences from the University of California, Irvine, investigating reactive heme intermediates under the advisement of Prof. Michael T. Green. He joined Prof. Yi Lu’s group at the University of Texas at Austin as a postdoctoral research fellow in 2021, shortly after completing his doctorate. Aaron’s research interests include bioinorganic chemistry, enzyme design, and mechanisms of human disease.

Yi Lu received his B.S. degree from Peking University in 1986 and his Ph.D. degree from the University of California at Los Angeles in 1992. After two years of postdoctoral research in Professor Harry B. Gray’s group at Caltech, Dr. Lu started his own independent career in the Department of Chemistry at the University of Illinois at Urbana-Champaign in 1994. He was promoted to Associate Professor with tenure in 2000, full Professor in 2004, and endowed Jay and Ann Schenck Professor in 2010. In August of 2021, Dr. Lu moved to the University of Texas at Austin, becoming the Robert J. V. Johnson-Welch Regents Chair in Chemistry. Dr. Lu’s research interests lie at the interface between chemistry and biology. Specific areas of current interest include (a) design and engineering of functional metalloproteins as environmentally benign catalysts in renewable energy generation and pharmaceuticals; (b) fundamental understanding of DNAzymes and their applications in environmental monitoring, medical diagnostics, and targeted drug delivery; (c) employing principles from biology for directed assembly of nanomaterials with controlled morphologies and its applications in imaging and medicine and (d) engineer biocatalysts to address challenges in synthetic organic chemistry and applications of novel biocatalysts in synthetic biology for biomass conversion, valuable products generation in high yield and good selectivity.