Repurposing Metalloproteins as Mimics of Natural Metalloenzymes for Small-Molecule Activation

Abstract

Artificial metalloenzymes (ArMs) consist of an unnatural metal or cofactor embedded in a protein scaffold, and are an excellent platform for applying the concepts of protein engineering to catalysis. In this Focused Review, we describe the application of ArMs as simple, tunable artificial models of the active sites of complex natural metalloenzymes for small-molecule activation. In this sense, ArMs expand the strategies of synthetic model chemistry to protein-based supporting ligands with potential for participation from the second coordination sphere. We focus specifically on ArMs that are structural, spectroscopic, and functional models of enzymes for activation of small molecules like CO, CO₂, O₂, N₂, and NO, as well as production/consumption of H₂. These ArMs give insight into the identities and roles of metalloenzyme structural features within and near the cofactor. We give examples of ArM work relevant to hydrogenases, acetyl-coenzyme A synthase, superoxide dismutase, heme oxygenases, nitric oxide reductase, methyl-coenzyme M reductase, copper-O₂ enzymes, and nitrogenases.

Graphical Abstract

1. Introduction

1.1. Motivations for creating metalloenzyme models.

Many of the most important small-molecule transformations in nature, such as the reduction of protons, O₂, CO, CO₂, and N₂, are performed by complex multicomponent metalloenzymes. Unravelling the mechanisms of these redox transformations is challenging for several reasons. For example, they often contain multiple cofactors with overlapping spectroscopic signatures, hindering unambiguous monitoring of their reaction intermediates. In addition, their large size, and ability to bind other proteins through hydrophobic surfaces, can make them resistant to structural characterization with X-ray crystallography. Third, the most exciting intermediates (those that perform bond breaking or formation in the small molecule) are often short-lived: this short lifetime is advantageous for avoiding side reactions of highly reactive intermediates, but makes it difficult to characterize these intermediates and understand the details of the key reactions.

These challenges have led chemists to study simpler mimics of the enzymes, most often through synthetic chemistry. Decades of dedicated synthesis efforts have been devoted to the design of supporting ligands that bind to the relevant metals to give complexes that are “synthetic analogues,” in order to provide spectroscopic, structural, and mechanistic insight.[1] However, the synthetic analogue approach has limitations: specifically, the abiological surroundings of the metal site in the analogue typically lack the variety of dipoles and charges of the multifaceted protein environment, and specific second-sphere interactions that may be crucial for reactivity (see section 1.3 below). A recent review outlines some of the additional ways that a protein tunes its local environment, including solvent accessibility and conformational flexibility.[2]

In this Focused Review, we describe metalloenzyme modeling using artificial metalloenzymes (ArMs), which consist of a catalytically active protein scaffold surrounding a foreign metal atom or coordination complex.[3–9] The use of proteins (rather than synthetic ligands) to create model systems can combine many of the advantages of studying the natural protein (or mutated variants) with the advantages of studying synthetic models. Though it is outside the scope of this Focused Review, another advantage of modeling with ArMs is the potential for use of directed evolution to optimize or tune the activity of enzyme constructs, a strategy that has gained prominence with a broader set of biocatalysts.[10–14]

1.2. Types of ArMs and scope.

ArMs come from implanting a metal or inorganic complex into an enzyme scaffold or by protein engineering of an existing metalloenzyme so that it fulfills a different purpose than naturally found. Invariably, the metal or complex is the active site of the ArM. There are four general methods of attaching the metal or complex to the protein scaffold (Figure 1)[15]: (1) covalent attachment, in which the ligands of an inorganic complex are covalently linked to amino acid residues of the protein; (2) dative attachment, in which one or more amino acid residues of the protein coordinate directly to a new, introduced metal center; (3) metal substitution, in which the metal of a natural metalloenzyme is replaced with a different metal that binds in the same site; and (4) supramolecular attachment, in which the ligands of an inorganic complex are covalently linked to a natural product which binds strongly to the protein. The strategies and advantages of each of these methods have been reviewed.[7, 16]

Figure 1.

Examples of four approaches to ArM assembly: (a) dative attachment, (b) metal substitution, (c) supramolecular attachment, and (d) covalent attachment. Reproduced from ref [15] with permission.

This review highlights the use of ArMs as enzyme mimics, in order to demonstrate their utility as a tool for understanding metalloenzyme mechanisms in the style of traditional synthetic modeling. We focus on a limited number of small-molecule reduction reactions to give a reasonable scope, and discuss only systems using the same metal ion as the natural metalloenzyme. For reasons of space, we do not attempt to cover the many other valuable applications and insights from ArMs, such as catalysts for organometallic reactions,[7, 10, 12–14, 17–21] design of de novo proteins,[22–28] or ArM approaches to abiotic catalysts.[29–31] The reader may also want to consult recent reviews with practical perspectives on the criteria for choosing a protein host for ArM chemistry.[7, 15, 32, 33]

1.3. ArMs versus traditional synthetic models.

A longstanding approach to gain insight into metalloenzymes is to prepare and study synthetic analogues.[1] These molecules are intended to give insight into the metalloenzyme properties or mechanism, but it is necessary to consider the nature of the analogy between the metalloenzyme and the model complex. To facilitate this analysis, model complexes may be categorized (Figure 2) as structural models in which the geometry of the metal sites, the same donor atoms, or the disposition of multiple metals is similar to the metalloenzyme structure (or hypothesized structure), as spectroscopic models where one or more spectroscopic characteristics have an analogy to the observed ones in the enzyme, and as functional models if they reproduce the catalytic activity or reactivity trends of the natural system. The most useful model compounds have more than one of the analogies, which enables the chemist to learn the relationships between structure and function, between electronic structure and function, or spectroscopic signatures for characterizing the complex protein environment that gives natural activity.

Figure 2.

Types of model complexes (synthetic or ArM).

Typical synthetic model complexes require multistep organic synthesis and are prone to intermolecular reactions due to the greater exposure of the active site relative to the protected environment in a metalloprotein. In addition, synthetic models typically lack the diverse surroundings of natural protein environments. One advantage of ArMs as models of enzyme active sites is that they offer a protein environment, which may have hydrogen bonds, charged and polar groups, spatial constraints, or hydrophobic pockets that would be extremely challenging to incorporate into a synthetic complex. Further, the advent of robust protein engineering tools allows chemists to modify ArMs using point mutations, insertions, or deletions to give an environment for the cofactor that is more accurate or more tunable than synthetic complexes. Moreover, most ArM-based models employ proteins that are small, soluble, easily isolated and crystallized, with few or no extra metal sites to obscure spectroscopic signals of interest. Within this conceptual framework, the ArMs can be used as structural, spectroscopic, and functional models like the small-molecule models.[34] It is important to note that ArM models have one of the same disadvantages as other artificial models: the accuracy of mechanistic conclusions from functional modeling is contingent upon the assumption that the model has sufficient similarity of structure and reaction pathways to give the same mechanism. In this way, all models are limited to showing feasibility or analogies: they do not show how the enzyme itself works! However, models (synthetic or ArM) often can elucidate broader trends precisely because of their differences: for example, if a designed ArM can capture the function of a metalloenzyme with local structural similarity despite a very different scaffold, it teaches us that the local structure is the predominant factor controlling function. On the other hand, if a seemingly identical site does not react similarly in a different scaffold, it implicates factors outside the active site as important for function.

2. Iron ArMs for understanding hydrogenase

The catalytic reduction of protons to hydrogen has been a topic of interest in the chemistry community for decades, particularly as part of systems for water splitting. Two classes of enzymes are known that combine two protons and two electrons to form H₂ (or the reverse reaction, H₂ oxidation): the [FeFe] hydrogenases and the [NiFe] hydrogenases.[35–38] Both types of hydrogenases have cofactors with non-protein components (Figure 3). They are particularly notable because they contain CO and CN^‒, which are essential for the catalytic reactions and enable infrared spectroscopy as a convenient spectroscopic signature for different intermediates in the catalytic cycles. In the [FeFe] hydrogenase, an additional cofactor is a dithiolate that doubly bridges the iron atoms. For years, the identity of the bridging dithiolate was controversial, but ArMs played an important role in resolving it.

Figure 3.

Active sites of [FeFe] hydrogenase (left) and [NiFe] hydrogenase (right). In the left picture, the circles represent Fe (black) and S (white). Adapted from ref [38] with permission.

2.1. Insertion of synthetic cofactors into the hydrogenase apo-protein.

Based on crystallographic studies, it was not clear whether the central atom of the dithiolate bridge in [FeFe] hydrogenase was C, N, or O, because these three atoms have similar electron densities.[39, 40] Synthetic models had suggested that the azadithiolate was the most likely bridging ligand based on its ability to serve as an internal base in structural and functional models,[41, 42] but as noted above, synthetic models can only show feasibility, not the enzymatic mechanism itself. Therefore, in addition to advanced EPR studies of the enzyme,[43] a semisynthetic ArM strategy was pursued by Lubitz, Artero, Fontecave and coworkers. They used synthetic cofactors that were specifically prepared with either propanedithiolate (pdt, C bridge), azadithiolate (adt, N bridge), or bis(thiomethyl)ether (bte, O bridge) as bridges, as bona fide examples of each possibility. Each synthetic diiron complex was bound to the biosynthetic precursor HydF, and thence inserted into the hydrogenase apo-protein. Importantly, only the adt-derived species had high activity, indicating that this one had the correct components of the wild-type enzyme. Though the semisynthetic enzyme with adt is not technically an ArM (it is the full reconstituted enzyme), it is its comparison with two artificial analogues that illuminated the atomic nature of the cofactor.[44]

In subsequent work, this semisynthetic strategy led to a number of other ArMs with varying bridgehead groups (Figure 4).[45] It was possible to vary not only the identity of the central atom, but also add bulk or change electronics. The accumulated results showed that the nitrogen bridge is not simply a proton shuttle, rather it also participates in electrostatic or hydrogen bonding interactions with the protein scaffold. In addition, even larger substituents could be incorporated, even though none of the altered [FeFe] hydrogenases were more active than the native version. All in all, this work shows the power of the semisynthetic approach for testing the roles of active-site features.[46]

Figure 4.

Substitution of the Fe₂ cofactor in [FeFe] hydrogenases. Reproduced from ref [45] with permission.

Recently, Berggren modified a known [4Fe4S]-containing metallopeptide to accommodate the Fe₂ synthetic cofactor with a covalent connection to the iron-sulfur cluster.[47] The result is a small mimic of the entire cofactor assembly in [FeFe]-hydrogenases but with most of the protein scaffold stripped away. However, frequencies of the CO stretching vibrations corresponding to CO ligands in this ArM differ substantially from the frequencies observed in the hydrogenases, and the activity of the ArM is much lower. This is a case where a structural analogy does not correlate with spectroscopic or functional ability, and these results demonstrate the importance of second-sphere donors for activity in hydrogenases. This system’s simplicity is promising, because it will enable future studies to selectively incorporate second-sphere effects one at a time for insight into the factors that nature uses to maximize activity for H₂ production.

2.2. ArMs with diiron cofactors for hydrogenase modeling.

A number of chemists have used various de novo proteins to hold Fe₂ cofactors, in an effort to query the behavior of the [FeFe] hydrogenase catalytic cofactor in an environment that is simpler than that in the natural enzymes. Jones has used protein maquettes that hold carbonyl-Fe sites through Cys residues, phosphine-containing amino acids, or lysine amides.[48, 49] These systems were amenable to detailed characterization of the mechanism of electrocatalytic H⁺ reduction, in which the active site is reduced, protonated, reduced, and protonated again (ECEC mechanism) to yield H₂. Notable aspects include the ability to use organic/water mixtures to test solvent effects, and the ability to vary the nature of the bridging dithiol in [FeFe] hydrogenase models, which are opportunities afforded by the synthetic ArM approach. Ghirlanda has prepared other diiron models in a protein scaffold using an unnatural amino acid anchor, and showed that the models can reduce protons using electrons photogenerated from a [Ru(bpy)₃]²⁺ photosensitizer.[50]

In other cases, a natural protein supports the [FeFe] hydrogenase mimetic active site. Hayashi has used apo-cytochrome c, which can coordinate the diiron cofactor through the Cys residues that would normally be the covalent attachment for the c-type heme.[51] This system was active for photocatalytic H⁺ reduction. They learned that the full apo-cytochrome c gave higher activity than a shorter peptide with the CXXCH binding sequence,[52] demonstrating that the larger protein scaffold is beneficial for activity, though the specific interactions that facilitate catalysis are not yet clear. In another study, Hayashi attached a β-barrel protein to a diiron cofactor through a maleimide linker, and this ArM construct was active as well.[53]

Using a different strategy, Ghirlanda leveraged the binding of a biotinylated Fe₂ cofactor to streptavidin, which created an ArM with the biomimetic active site in the “vestibule” area of streptavidin.[54] This example is described here in detail to demonstrate some of the advantages of the ArM modeling approach. The CO stretching bands in the infrared spectrum shifted to slightly higher frequency, and a broad feature at 1990 cm⁻¹ split into more features, indicating a change to a less polar environment with more variability. This variability in polar environment is promising in catalysis, where transition state energies for different steps may be influenced by the environment differently to change the selectivity or rate. The numerous observed CO stretching bands indicated that there were multiple conformations of the diiron site, which agreed with the disorder that was evident in the crystallographic structure of the ArM. In terms of activity, the ArM was active in protein film voltammetric reduction of protons (proton reduction using an electrode as the electron source), and photocatalytic reduction using [Ru(bpy)₃]²⁺. Though the catalytic rates were slower for the cofactor in the protein than outside the protein, the longevity of the catalyst improved by encapsulating the cofactor in the protein. This was attributed to a longer lifetime of catalytic intermediates.[55]

2.3. ArMs with nickel cofactors for hydrogenase modeling.

A simpler ArM construct is based on rubredoxin (Rd), a small iron-containing electron-transfer protein with all-Cys coordination. In this case, an ArM preceded the enzyme structure: in 1988, Moura substituted the Fe site in Rd with Ni, showing that this ArM could reduce protons to H₂ using a chemical reductant.[56] The crystal structure of [NiFe] hydrogenases later demonstrated that the nickel site in the enzyme has a similar Cys₄ coordination environment, though the hydrogenase differs by having two of the cysteine thiolates bridging to a redox-inactive iron ion (cf. Figure 3 above).[57] Recently, Shafaat extended the study of nickel-substituted rubredoxin to photocatalytic and electrocatalytic hydrogen production, and sequential mutation of residues indicated the importance of proton-coupled electron transfer (PCET). In this example of PCET, there is proton transfer from second-sphere residues to coordinated Cys residues simultaneously with electron transfer that reduces the active site from nickel(II) to nickel(I).[58, 59] They also screened numerous variants of the Rd protein to elucidate the influence of specific second-sphere groups on the rate and redox potential of the nickel ArM catalyst. This demonstrated that a carboxylate facilitates proton transfer into the active site, and that the rate of catalysis can be increased independent of the driving force (Figure 5).[60] In the long run, this tunability of the environment of the catalytic site is a significant advantage of the protein scaffold of an ArM over synthetic catalysts. In addition, these Ni-rubredoxins are air-stable, which provides a practical advantage over the air-sensitive [NiFe] hydrogenases and many of the synthetic catalysts (although some synthetic Ni catalysts for H₂ production are air-stable).[61]

Figure 5.

The rates and overpotentials of [NiFe] hydrogenase modeling nickel-rubredoxin ArMs vary widely, and are not correlated because the changes in the protein environment affect different steps in the catalytic cycle. Reproduced from ref [60] with permission.

Another recent system by Chakraborty is similar in that it uses a Cys-rich site in a copper storage protein Csp1 to hold nickel.[62] In this case, the natural protein has many Cys and His residues that could potentially bind Ni, and most of which were mutated to Ala residues to better define the active site. Spectroscopic studies suggest that the nickel site in the resting state in the ArM is square-planar, in contrast with the tetrahedral site in Ni-rubredoxin. Despite this difference, the photocatalytic and electrocatalytic abilities are reminiscent of the rubredoxin systems, and it should also be amenable to second-sphere modification.

3. Enzyme mimics from unnatural metals in azurin

Azurin is another small, soluble metalloprotein with a relatively rigid metal-binding site, which naturally accommodates a copper ion for electron transfer.[63] Especially since its interaction with one of the donors (M121) is weak, it is a natural choice for ArM design where the sidechain is substituted.[64]

3.1. Iron in azurin mimics superoxide reductase.

The Holland group substituted the copper site of azurin with iron(II), and characterized the wild-type and M121A ArMs with spectroscopy and crystallography. These studies showed that the iron(II) species could not be oxidized to iron(III) with common oxidants, and could not be reduced to iron(I) except using radiolytic reduction.[65, 66] In contrast, the Lu group created the Fe-M121E ArM and found that the anionic glutamate donor gave an iron site that allowed oxidation to iron(III).[67] The Fe-M121E azurin active site has a similar geometry as the active site of superoxide reductase (Figure 6), and reproduces the enzymatic activity.[68, 69] In the ArM, mutation of a Met to a Lys residue near the iron site greatly increased activity, showing the importance of this second-sphere cationic group for binding the negatively-charged O₂⁻ substrate.

Figure 6.

Comparison of structures of (a) superoxide reductase, (b) the copper analogue of the iron(II)-substituted M121E/M44K azurin used as a mimic. Adapted from ref [67] with permission.

3.2. Nickel in azurin mimics acetyl-coenzyme A synthase.

More recently, an ArM from nickel substitution into M121A azurin has been influential because it can bind small-molecule substrates to the vacated axial site. In particular, binding of CO is relevant to the modeling of acetyl-coenzyme A synthase (ACS), which couples CO and methylcobalamin to coenzyme A as part of the Wood-Ljungdahl cycle for anaerobic energy transduction. Shafaat has demonstrated that the nickel-azurin ArM can bind both CO and CH₃ to the nickel site, using a combination of spectroscopy and reactivity studies (Figure 7).[70, 71] Since ACS is a terribly complicated enzyme system with quaternary structure dynamics, a “tunnel” that delivers CO from another site, and substantial lasting controversy about mechanism,[72] the relative simplicity of the Niazurin ArM is a major benefit. Much of the stoichiometric behavior of the ArM is similar to synthetic nickel complexes,[73, 74] but importantly the protein scaffold can stabilize transient nickel(III)-methyl intermediates that were elusive in synthetic systems.[75] These exciting results amply demonstrate the new opportunities available from the ArM modeling approach.

Figure 7.

Comparison of the active sites of acetyl coenzyme A synthase and nickel-substituted M121A azurin. Reproduced from ref [75] with permission.

4. ArM approaches to modeling heme enzymes for small molecule activation

4.1. Enzyme mimics for O₂ and NO reducing enzymes.

The heme protein myoglobin (Mb) has been used as a scaffold for ArM designs, mostly by the Lu[76] and Hayashi[17] groups. Many studies have modeled dinuclear enzymes in which a heme has a nearby metal with which it cooperatively reduces small-molecule substrates such as NO and O₂. In these designs, it is advantageous to leave the natural heme site while engineering the proximal cavity to accommodate a second metal that models the Fe (in NO reductase) or Cu (in cytochrome c oxidase, CCO) (Figure 8). The resulting models are small, soluble analogues of the multicofactor, membrane-bound natural proteins, highlighting the advantage of the ArM’s simplicity.

Figure 8.

The design of an engineered myoglobin active site for ArMs with cooperative bimetallic small-molecule activation.

In research from Lu, a copper-binding site of three histidines was installed just above the heme site in Mb.[77] In order to explore the postulated role of an unusual crosslinked tyrosine residue in the mechanism of CCO,[78] the authors first introduced a F33Y mutation into Mb, and the catalytic rate and selectivity improved dramatically.[79] Surprisingly, the choice and concentration of added metal did not influence the O₂-reducing ability, nor did the addition of EDTA, which suggested that the Cu site is not loaded in this ArM (though the authors did not consider the possibility that EDTA might be incapable of removing a metal ion that is introduced during expression or purification, a recurrent problem in azurin studies). In addition, it was possible to detect a Tyr radical using electron paramagnetic resonance (EPR) spectroscopy, which detects the unpaired electrons characteristic of radicals, during O₂ reduction catalysis.[80] In separate work, the authors explored the role of the tyrosine crosslink by introducing an unnatural amino acid composed of an imidazole-bearing tyrosine.[81] This enabled them to compare the tyrosine to the imidazole-tyrosine, and in the latter the catalytic rate and selectivity improved dramatically. In another case, the Tyr residue was substituted with halogenated Tyr analogues that lower the pK_a of the Tyr hydroxyl group, and this modification improved the rate and selectivity.[82] This is an example of the power of the ArM modeling approach to bring insight to enzyme mechanism, in this case the details of Tyr participation in selective reduction of O₂ to water without the production of H₂O₂.

In complementary work, Lu has shown that inserting an Fe atom into the engineered proximal site can yield an ArM that models nitric oxide reductase (NOR). In this case, the authors introduced two key Glu residues near the bimetallic site, resulting in a functional model of NO reduction to N₂O.[83] Transient NO intermediates were identified using rapid freeze-quench resonance Raman spectroscopy, which showed that NO binds first to the nonheme iron, followed by a second NO binding to the heme iron.[84] Raman spectroscopy indicated that the two NO ligands faced one another, providing key evidence toward the trans mechanism of N₂O production.[84, 85] Though this (like all modeling studies) does not show the mechanism of NOR itself, it indicates that the trans mechanism is viable for NOR enzymes, since the active site of the ArM is so similar to that of NOR.

4.2. ArMs with other porphyrinoids in the myoglobin pocket.

Another enzyme where ArM-based models have been advantageous is methyl-coenzyme M reductase (MCR), which catalyzes the reaction in Figure 9.[86] MCR has a complex heterohexameric structure, and the MCR reaction involves an organic coenzyme B as a part of the complicated mechanism in the enzyme.[87] The metal site of MCR is the nickel-containing F430 cofactor, in which the nickel is coordinated by a monoanionic tetrapyrrole (hydrocorphine) that can support nickel in formal oxidation states of +1, +2, and +3. A nickel(III)-methyl species has been detected in the enzyme from addition of the alternative substrate methyl iodide, and was previously proposed to be an intermediate from an analogous nucleophilic attack on the methyl thioether group of coenzyme M, but a 2016 study from Ragsdale indicates that the nickel(III)-methyl species is not kinetically competent to be an intermediate.[88] Therefore, these studies implicated an alternative mechanism with radical attack of nickel(I) on the methyl group of methyl-coenzyme M to produce a methyl radical and nickel(II).

Figure 9.

Methyl-coenzyme M reductase reaction, and its F430 cofactor. Adapted from ref [86] with permission.

Hayashi has placed an artificial nickel-tetradehydrocorrin cofactor into the Mb protein,[89, 90] to form an ArM (named rMb-NiTDHC or reconstituted myoglobin with nickel tetradehydrocorrin) that was used for insight into potential MCR mechanisms. Reduced rMb-NiTDHC from addition of dithionite generates methane from methyl donors like methyl tosylate and methyl iodide, but not methyl-coenzyme M itself. In addition, the turnover numbers (TONs) of alkane formation from rMb-NiTDHC, or numbers of alkanes formed per ArM, are in the range of 0.1–2, which is better than synthetic models but still low. Thus, in the ArM (as in the enzyme) nucleophilic attack from the nickel is possible with alternative substrates but not methyl-coenzyme M. It seems likely that additional unmodeled components of the natural MCR enzyme must be incorporated to capture its ability to follow the radical mechanism that can interconvert the thioether with alkane.

5. Biotin-streptavidin for metalloenzyme mimics

Section 2.2 above mentioned the use of biotin/streptavidin binding (pioneered by Whitesides [91] and popularized by Ward [20] for binding organometallic hosts within a protein) for incorporating the [FeFe]-hydrogenase cofactor into the streptavidin protein environment. Borovik has leveraged this strategy for stabilizing a copper-hydroperoxide complex within the protein host using the hydrogen bonding environment (Figure 10).[92] In this work, the copper is supported by a dipyridylamine ligand of the type well-known in the synthetic modeling literature,[93] but the incorporation of a biotin tether is used to suspend it inside the modifiable protein host, which permits precise modification of the environment that would not be possible with a traditional synthetic model complex. This hydrogen-bonding network was later probed in more detail in a copper-azide complex[94] and in an iron-azide complex.[95] Other work by Borovik has incorporated a Co₄O₄ cubane core into streptavidin,[96] which may lead to an enzyme-mimicking ArM that incorporates a manganese-oxo model of the oxygen-evolving cluster of photosystem II.

Figure 10.

(A) Copper-hydroperoxide complex stabilized within the pocket of streptavidin. (B) Hydrogen bonding network around the Cu-OOH site. Reproduced from ref [92] with permission.

6. Other examples of ArMs for O₂ and N₂ activation

6.1. Hydrocarbon-oxidizing enzymes.

Rieske dioxygenases have the impressive ability to add O₂ to aromatic rings in benzene, phthalates, toluene, and naphthalene to form cis-dihydrodiol products.[97] This is typically part of a sequence that leads to net dihydroxylation of the strong C–H bonds of these aromatics, and is involved in degradation of xenobiotic compounds. The O₂ is activated at a nonheme iron(II) center held by a “catalytic triad” of two His and one Asp/Glu residue. The crystal structure of naphthalene dioxygenase has been determined, including an O₂ adduct that precedes attack on the aromatic ring.[98] In an ArM approach, Ménage and coworkers made a model of this enzyme using the apo form of the nickel-binding protein NikA, which can hold small complexes in the vacated nickel site. [99, 100] The authors bound an iron(III) complex of a pentadentate N₂O₃ ligand by soaking it into a crystal of NikA, and found that after reduction it can react with O₂ to generate an O₂-bound intermediate that could be crystallographically characterized. With excess reductant and O₂, a pendant phenyl group is hydroxylated or dihydroxylated, and these were each crystallographically characterized as well. This work highlights another advantage of ArM constructs over synthetic complexes, the ability to observe intermediate species in crystallo, because protein crystals are much more flexible and solvent-rich than crystals of synthetic complexes. In later work, they blocked the position on the N₂O₃ ligand that had been oxidized, and the resulting ArM construct could perform intermolecular styrene oxidations, but the oxidation was at the exocyclic double bond rather than the aromatic ring.[101]

6.2. Potential for insight into nitrogenases.

Nitrogenases are a prime candidate for simplification, as these enzymes are very large with multiple cofactors, and have a complex dependence on conformational changes and electron transfer during catalysis. Unfortunately, it has not yet been possible to induce the cofactors to reduce N₂ in any environment except the nitrogenase proteins. On the other hand, it is possible to extract nitrogenase cofactors from the enzyme, and the extracted cofactors are able to reduce acetylene or CO without the natural protein.[102, 103] A number of studies have compared the reactivity of various nitrogenases with cofactors from different nitrogenases inserted.[104] This should be a fertile ground for future work with ArMs, which could modulate an artificial protein environment around the cofactors to elucidate which specific second-sphere features from nitrogenase are the ones that enable these metalloenzymes to perform the prodigious reduction of N₂ to ammonia.

A related strategy has been the insertion of artificial iron-sulfur clusters into the nitrogenase apoprotein. A collaborative team from Ohki, Tatsumi, Ribbe, and Hu have inserted a synthetic [6Fe9S] cluster into apo-nitrogenase and observed acetylene reduction with low activity.[105] Since the authors reported no spectroscopic experiments to verify that the synthetic cluster is still intact in the protein, the nature of the cofactor in this ArM is not known. It will be enlightening to learn further structural and spectroscopic characterization of these artificial versions of nitrogenase, because the opportunity to vary the metal environment within nitrogenases can be used to dissect the different roles of structural features of nitrogenase cofactors, and provide analogues to transient enzymatic intermediates.

7. Conclusions

Artificial metalloenzymes (ArMs) provide some of the best aspects of synthetic models and enzyme mutation, and they are already starting to have impacts on our understanding of metalloenzymes. Here, we highlight recent studies on metalloenzymes that activate small molecules as an example of how this particularly challenging set of metalloenzyme mechanisms has been impacted. This description also highlights some of the advantages and disadvantages of the ArM modeling approach, which should be of use to future chemists eager to take advantage of the field.

Figure 11.

ArM construct of an iron complex in apo-NikA, as a model of Rieske dioxygenase enzymes. (A) Crystal structure of an O₂ complex; (B) Crystal structure of dihydroxylated aromatic ring. Reproduced from ref [100] with permission.

Synopsis:

We review artificial metalloenzymes (ArMs) that aim to give insight into the mechanisms of metalloenzymes that transform small molecules like CO, H₂, and N₂.

Highlights.

• Artificial metalloenzymes (ArMs) can be models of natural metalloenzymes.

• ArMs are small, soluble, and tunable, yet have a protein environment around the metal.

• ArMs can help to distinguish the key determinants of enzyme activity.

• We review models for complex metalloenzymes that transform CO, CO₂, O₂, N₂, H₂, and NO.

Abbreviations

ACS

acetyl-coenzyme A synthase

adt

azadithiolate

ArM

artificial metalloenzyme

bte

bis(thiomethyl)ether

Mb

myoglobin

MCR

methyl-coenzyme M reductase

PCET

proton-coupled electron transfer

pdt

propanedithiolate

Rd

rubredoxin

rMb-NiTDHC

reconstituted myoglobin with nickel tetradehydrocorrin