Artificial metalloenzymes (ArMs) result from the incorporation of a catalytically competent metallocofactor into a protein scaffold. Initial attempts to construct ArMs were reported by Wilson and Whitesides as well as Kaiser and coworkers in the late 1970s. [1][2] Despite several fascinating aspects of these hybrid catalysts, Whitesides concluded his article by stating "the catalytic system (...) is not a practical asymmetric catalyst." 2 Unfortunately, following this pronouncement, relatively little progress was made toward ArM development until the new millennium. Setting aside the question of whether practicality should dictate the course of scientific inquiry, it is certainly the case that ArM construction in those early days was limited by the tools available for both organometallic synthesis and protein engineering. Significant progress in these areas contributed to the revival in ArM research that continues to this day.
The resurgence of interest in ArMs was also driven by increased appreciation of the potential benefits of combining attractive features of both homogeneous catalysis and enzymatic catalysis, Table 1. For example, the ArM secondary coordination sphere could interact with metal catalysts, substrates, or intermediates to facilitate reactions or to discriminate similarly reactive sites on substrates. ArMs could also be generated from protein scaffolds with inherent functionality (e.g. catalytic activity, substrate binding, redox properties, etc.) that could be used to augment ArM function. Perhaps more intriguingly, ArMs could endow organometallic catalysts with a genetic memory. Incorporating a organometallic catalyst within a genetically-encoded scaffold offers the opportunity to improve the ArM performance by mutagenesis. Ultimately, this could enable Darwinian evolution schemes for ArM optimization. Introduced in the early nineties by Frances Arnold and Pim Stemmer, [3][4] directed evolution has had a revolutionary impact on biotechnology, 5- 7 leading to catalysts that have supplanted well-established large-scale processes based on homogeneous catalysts. 8 Similar optimization of ArMs could ultimately led to systems suitable for practical applications, and provide greater insight into the role of second coordination sphere interactions in organometallic catalysis. Four complementary strategies have been implemented to localize metallocofactors within a well defined second coordination sphere environment, provided by the host protein: i) covalent, ii) supramolecular, iii) dative and iv) metal substitution, Figure 1.
i) Covalent anchoring, reminiscent of the well-established bioconjugation techniques, 9 involves a high-yielding and irreversible reaction between cofactors bearing a reactive functional group and an amino acid side-chain on the protein scaffold. Common reactions used for ArM formation include: a) nucleophilic attack by cysteine or another uniquely activated residue on an electrophilic moiety (maleimide, a-halocarbonyl, etc.) on the cofactor, b) disulfide bond formation between cysteine and a cofactor substituted with an electrophilic sulfur moiety, and c) Huisgen [3+2]-cycloaddition between an unnatural amino acid bearing a terminal alkyne or azide with an azide-or alkyne-substituted cofactor. 10 ii) Supramolecular anchoring exploits the high affinity that proteins may display for a limited set of non-covalent inhibitors, natural cofactors or substrates. Covalent modification of these with the cofactor may, in some cases, maintain a high affinity, thus ensuring quantitative localization of the cofactor within the host protein.
iii) Dative anchoring relies on the coordination of a nucleophilic amino acid residue (His, Cys, Glu, Asp, Ser etc.) to a coordinately unsaturated metal center. This type of anchoring and activation of the metal often complements either covalent or supramolecular strategies.
iv) Metal substitution builds upon the unique reactivity of non-native metals combined with the exquisitely tailored active site of natural metalloenzymes. Upon substituting the metal, new-to-nature reactivities can be introduced in the ArM's repertoire. 11 This strategy builds upon the very elegant enzyme repurposing approach introduced by Frances Arnold in 2013. [12][13][14] A complete coverage of the enzyme repurposing strategy is presented by Fasan and coworkers in this issue of Chemical Reviews.
Figure 1. Four anchoring strategies allow to firmly localize an abiotic cofactor within a protein scaffold: a) covalent, b) supramolecular, c) dative, and d) metal substitution. The following colour codes apply: protein and natural cofactor (green), supramolecular anchor (red), variable spacer and ligand (blue) and abiotic metal (black).
Since its renaissance in the early 2000s, the field of artificial metalloenzymes has been extensively reviewed and highlighted. With catalytic applications in mind, the present review summarizes the progress in ArM according to the reaction they catalyze.
The authors' initial intention was to classify the reactions according to the enzyme class system (EC) from the International Union of Biochemistry and Molecular Biology:
oxidoreductase, transferase, hydrolase, lyase, isomerase and ligase. It rapidly became evident that the EC system is not best suited to ArM as the reactions these catalyze are often non-natural: in what EC would one assign to an artificial metathase or Suzukiase?
We thus followed a more classical functional-group transformation classification.
In view of the focus on catalysis, artificial metalloproteins with no catalytic function are not covered comprehensively. [34][35] The same applies to the fascinating fields of DNA-and RNAzymes as well as metallopeptides with less than one hundred amino acids (an arbitrary length set for inclusion in this review). These DNA/RNAzymes and metallopeptides have been reviewed recently by key players in the field. 70,[96][97][98] This review begins with a summary of the proteins and the anchoring strategies used to date for the creation of ArMs (chapter 2), followed by a historical perspective (chapter 3).
Then follows a summary of the reactions catalyzed by ArMs (Reduction chemistry, C-C bond-formation, oxygen insertion and Hydration: chapters 4-7). This review ends with a critical outlook (chapter 8).
The authors have invested significant effort to cover all artificial metalloenzymes published to date, albeit with the restrictions outlined above. Despite the care for the detail and the search tools available, there is no doubt that we have missed some articles. We strive to apply a fair ethics of citation and thus wish to apologize for any unintentional omission. 99 In addition, this review served as a basis to set up a fully-searchable website on ArM that will be updated regularly by the authors. 100
The table presented below summarizes all the protein scaffolds that have been used to date to create artificial metalloenzymes as well as relevant data including i) the anchoring strategy, ii) whether mutants have been tested, iii) the reaction that the resulting ArM catalyzes etc. Protein scaffolds that have been subjected to mutagenesis to optimize the performance of the ArMs are displayed throughout the article where most relevant. m Commercially available.
n POP from Pyrococcus furiosus (PDB : 5T88), submitted.
The aim of this section is to cover work on artificial metalloenzymes dating from the earliest studies on hybrid metal-protein catalysts to those completed around the year 2000. As noted in the introduction, around this date, modern tools of molecular biology began making a significant impact on ArM research. Presumably as a result of this, the five-year average number of publications on ArM research each year substantially increased substantially in the early 2000's (Chart 1). These early studies are presented in chronological order to highlight how different advances were made using different scaffolds and ArM formation strategies at the time of the original reports.
Chart 1. Number of publications and the five-year moving average of this number plotted versus publication year for publications cited in this review.
The first example of transition metal-catalyzed asymmetric synthesis, reported in 1956, also constitutes the first realization of a protein-modified transition metal catalyst. [101][102] S.
Akabori et al. adsorbed palladium chloride on silk fibroin fibres and reduced the resulting material with hydrogen, likely resulting in the formation of Pd-nanoparticles embeded within an enantiopure protein environment. 103 The protein-immobilized palladium catalyzed asymmetric hydrogenation of dehydroaminoacid derivatives 1-3 to provide products 4 and 5 with significant enantiomeric excess (Scheme 1). The reduction of abenzildioxime, however, led to variable results, suggesting that the catalyst was poorly defined. Despite reproducibility issues, [101][102] this initial study has sparked research in biopolymer-immobilized Pd, sometimes termed Bio-palladium. 104 Pd-nanoparticles have recently also attracted attention for in vivo applications. [105][106] Scheme 1. Asymmetric hydrogenation of dehydroaminoacid derivatives with silk-fiber modified palladium reported in 1956.
Metal substitution has been used for the study of natural metalloenzymes since the late 1960s. [107][108] Metal-exchange studies often aimed at introducing spectroscopically observable metals as a means to elucidate structural details but did not focus on catalysis. [109][110] J. E. Coleman, however, observed that the CO2 hydration activity of carbonic anhydrase remains roughly half when Zn(II) is exchanged for Co(II), whereas the esterase activity increased after metal exchange for carbonic anhydrase B. 108 In the same year (1967), P. Cuatrecasas et al. reported the replacement of Ca(II) for Sr(II) in a staphylococcal nuclease. Whereas the Ca(II)-dependent enzyme catalyzed both DNA and RNA cleavage, the Sr(II) exchanged nuclease cleaved DNA exclusively. 107 Although not catalytic, the potential of metal-ion exchange to impact enzyme function was convincingly demonstrated in the activation of the zymogen trypsinogen by trypsin in 1970. 111 Trypsinogen contains two latent metal binding sites, the lower affinity of which is located at the cleaved N-terminus. 112 Ca(II)-ions bound to this site led to a lowering of the Km for the trypsin-trypsinogen interaction. Bound Nd(III) leads to a much more pronounced lowering of the Km compared to Ca(II), even at a 100 fold lower concentration of metal ions.
Presumably the first example where alternative catalytic function was introduced through metal-exchange in a metalloprotein was reported by K. Yamamura and E. T. Kaiser in 1976. 1 Carboxypeptidase A was converted to an active oxidase by exchange of Zn(II) for Cu(II) (Scheme 2). Limiting kcat and Km values of kcat = 6 min -1 and Km = 0.24 mM where determined for the oxidation of ascorbic acid 6 to dehydroascorbic acid 7. Scheme 2. Oxidation of ascorbic acid by carboxypeptidase A (CPA) after exchange of the active site metal ion Zn(II) for Cu(II).
M. E. Wilson and G. M. Whitesides were the first to realize the anchoring of a completely abiotic cofactor inside a protein cavity in a defined fashion. 2 Exploiting the remarkable affinity of the small molecule ligand biotin and some of its derivatives for avidin, a biotinylated Rh(I)-diphosphine precatalyst 9 was employed for the hydrogenation of aacetamidoacrylic acid 8 in the presence of the host protein (Scheme 3). The system showed significant, albeit moderate, stereoinduction in aqueous phosphate buffer and, importantly, a definite increase in activity when compared to the protein-free cofactor.
Avidin is a homotetrameric protein that can bind up to 4 equivalents of biotin. One equivalent of the biotinylated catalyst precursor vs. biotin binding sites led to a higher enantiomeric excess than 0.5 equivalents indicating that the cofactors interact with each other when localized in adjacent binding sites. This study was reported in 1978. Scheme 3. Asymmetric hydrogenation of a dehydroamino acid derivative by a biotinylated Rh(I)-complex upon binding to avidin.
Catalytic enantioselective oxidation of sulfides by oxidants such as NaIO4, t-BuOOH, or H2O2 in the presence of bovine serum albumine (BSA) was investigated by T. Sugimoto and T. Kokubo et al. in the late 70's and early 80's. [113][114][115] In one publication, the effect of stoichiometric metal additives, namely Cu(II), MoO4 2-, or WO4 2-, in combination with t-BuOOH was reported, though the results were less promising than other systems tested in terms of yield and selectivity. 113 Similarly, in 1983, T. Kokubo et al. investigated the asymmetric OsO4-catalyzed dihydroxylation of alkenes with BSA as a chiral scaffold with remarkable success (Scheme 4). 116 An ee of 68 at a TON of 40 in the dihydroxylation of a-methylstyrene 10 with t-BuOOH as a stoichiometric oxidant was reported, clearly demonstrating the potential to exploit non-metalloproteins as ArM scaffolds. Scheme 4. Asymmetric dihydroxylation of a-methylstyrene with OsO4/t-BuOOH in the presence of bovine serum albumin (BSA).
Cytochromes P450 have long attracted the attention of chemists. The ability of these enzymes to catalyze regio-and stereo-selective oxidative processes using molecular oxygen and reducing equivalents from NAD(P)H has so far been impossible to mimic reliably with purely synthetic methods. The sophisticated redox machinery required to transfer electrons from reduced nicotinamide cofactors to the heme center and generate an active iron-oxo species has contributed to the interest in these systems, 117 but it also hampers the application of P450 variants in synthetic processes. 118 Hemoglobin, another heme-containg metalloprotein, has been shown to catalyze P450-type oxidations when supported by an NADPH-cytochrome P450 reductase containing FAD and FMN cofactors. 119 T. Kokubo, S. Sassa, and E. T. Kaiser reported in 1987 on a catalytically competent 7-cyanoisoalloxazine-hemoglobin conjugate. 120 The flavin-substituted hemoglobin hybrid, named flavohemoglobin, catalyzed the para-hydroxylation of aniline 11 in the absence of NADPH-cytochrome P450 reductase with a higher rate than the reconstituted system and with a Km value similar to that of the hemoglobin scaffold (Table 2, Scheme 5). The flavin derivative was attached to cysteine b-93 through a glycineaminoethanthiol-linker. The X-ray structure (1988) of flavohemoglobin revealed that i) the flavin derivative is readily accessible to NADPH ii) the center of the isoalloxazine and the heme of the same subunit are 14 Å apart iii) the flavins of neighbouring subunits are in close contact, possibly enabling electron transfer between subunits. 121 Natural flavohemoglobins exist, for example in E. coli. 123 In a similar manner, various oligonucleotide binding proteins or protein domains were converted to oligonucleotide cleaving proteins via covalent modification with EDTA or phenanthroline complexes of Fe(II) or Cu(II), respectively. [124][125][126][127] The combination of a reducing agent such as a thiol or ascorbic acid with molecular oxygen resulted in oxidative cleavage of the oligonucleotide backbone. 128 The systems were typically employed in superstoichiometric amounts relative to the oligonucleotide.
The ability to raise monoclonal antibodies against seemingly any hapten inspired ambitious research into the generation of antibody-based artificial metalloenzymes. It was envisioned that these could possess the ability to discriminate substrates and hence provide specificity toward selectable cleavage motifs. 130 Monoclonal antibodies were raised against a peptide (trien)Co(III) complex 12 (trien = triethylentetramine). Substitution of the hapten for other metal(trien) complexes led to active peptidases that could distinguish between structurally related substrates 13
-18 and cleaved a specific peptide bond in the small set of peptidic substrates investigated 17, 18; Scheme 6. The antibody ArM design was based on the prospective exchange of a coordinatively inert octahedral Co(III) complex used as the hapten against a kinetically more labile complex for catalysis. A simple TLC assay for the liberated terminal amino group enabled the fast evaluation of a range of constructs. One construct, antibody 28F11 in combination with Zn(II)trien, was investigated in more detail, and the authors concluded that all components, namely metal(trien), specific substrate, and antibody are required for the reaction. Various metal(trien) complexes could be employed, and a turnover number of 400 with a turnover frequency of 6 ´ 10 -4 s -1 was reported. Scheme 6. Development of metal-dependent antibodies for specific cleavage of peptidic substrates. a) Kinetically inert cobalt(III) complex employed as hapten; b) peptidic structures not accepted as substrates; c) cleavage of peptidic substrates with antibody 28F11 in the presence of Zn(II)trien.
Following Kaiser's earlier work demonstrating that hemoglobin could be converted into a reductase-independent hydroxylase, T. Sasaki and E. T. Kaiser decorated Fe(III) coproporphyrin with four identical fifteen-residue peptides. 131 In the presence of acetylflavin and NADPH, the resulting construct 'helichrome' was shown to hydroxylate aniline with kcat and Km values of 0.02 min -1 and Km = 5.0 mM, respectively. The unmodified Fe(III) coproporphyrin showed only negligible hydroxylase activity under the same conditions. The study was published in 1989 and constitutes, to the best of our knowledge, the first report of a de novo 'heme'-enzyme.
A. G. Cochran and P. G. Schultz raised in 1990 monoclonal antibodies against Nmethylmesoporphyrin IX 19 as a hapten (Figure 2). 132 N-methylporphyrins show out-ofplane distortion and can be considered as a transition state model for the metalation of the porphyrin, a reaction typically catalyzed by ferrochelatase in the conversion of protoporphyrin to heme. Three purified antibodies were specific for the hapten, and two of these catalyzed Zn(II)-and Cu(II)-porphyrin complex formation. The faster of these two showed rates for Zn(II) insertion comparable with those of ferrochelatase (80 h -1 vs 800 h - 1 , respectively). Further studies revealed that the same antibody was effectively inhibited by iron(III) mesoporphyrin 20. Intriguingly, the resulting construct catalyzed the oxidation of a range of typical chromogenic peroxidase substrates (using H2O2 as a stoichiometric oxidant) with significantly higher rates than iron(III) mesoporphyrin itself. 133 Notably, ABTS and pyrogallol red were converted by the construct, but not by iron(III) mesoporphyrin. In a closely related study, A. Harada and coworkers reported several years later that antigen binding fragments in combination with meso-tetrakis(4-carboxyphenyl)porphyrin (TCPP) complexes of Mn(III) and Fe(III) catalysed the oxidation of pyrogallol with H2O2 exclusively while other substrates tested by Cochran and Schulz such as hydroquinone, resorcinol, catechol and ABTS with a higher redox potential were not converted. 134 In a conceptual paper published in 1990, E. Keinan et al. raised antibodies against the water soluble Sn(IV) porphyrin complex meso-tetrakis(4-carboxyvinylphenyl)porphinato tin(IV)dihydroxide (Sn(TCP)(OH)2) 21 (Figure 2). 135 Antibody metallation with Mn(III)(TCP)
cofactor 22 led to an ArM that catalyzed styrene epoxidation in the presence of iodosobenzene. The highest activity was observed under heterogeneous conditions in CH2Cl2, but no enantioinduction could be detected. Despite the prevalence of non-covalent cofactor incorporation in early ArM efforts, work on covalent incorporation also continued. For example, J. P. Germanas and E. T. Kaiser (posthumously) communicated an artificial oxidase based on a covalently bound Cu(II)bipyridine complex linked to the cysteine located in the active site of papain. The construct catalyzed the air-oxidation of ascorbic acid (compare Scheme 2) and more lipophilic derivatives at 15 -26 fold higher rates compared to the free cofactor. 136 Although no catalysis was probed, in 1993, H. B. Gray, B. Imperiali, and coworkers demonstrated photoinduced electron transfer between a covalently linked Ru-center and heme in a cytochrome c mutant. The heme-bearing mutant beared a Ru(bipy)2-fragment coordinated to a bipyridyl-alanine bearing peptide constructed by semisynthesis. 137 In 1993, R. A. Lerner, K. D. Janda, and coworkers disclosed a strategy to generate an antibody-based ArM hydrolase using a hapten that contained a methylpyridinium unit 23, rather than a metal moiety (Scheme 7). 138 Based on the activation of picolinic acid esters by Lewis acids coordinated to the pyridine nitrogen atom, it was envisioned that the methylpyridinium-substituted hapten would select for antibodies that address to scaffold design criteria. First, the methyl group in the hapten would act as a placeholder for a metal ion. Second, the positive charge would elicit the expression of carboxylate residues to serve as additional ligands near the prospective metal binding site and thereby increase metal affinity. The ester moiety envisaged as a cleavage site, or more specifically, the expected tetrahedal intermediate, was simulated with a hydroxyl functional group. Upon metallation with Zn(II), the resulting antibody-based ArM showed a >10,000-fold rate acceleration against background and a > 1,000-fold rate acceleration against Zn(II) in equivalent concentration. The affinity of the antibody for Zn(II) and the substrate complexed to Zn(II) was weak, whereas the affinity for the substrate itself was substantially higher.
Monoclonal antibodies catalyze the hydrolysis of a specific picolinic acid ester in the presence of Zn(II). a) N-methyl pyridinium employed as the hapten; b) substrates were not accepted by the antibody; c) hydrolysis of substrate by the antibody in the presence of Zn(II).
B. Imperiali and R. S. Roy reported in 1994 on an artificial transaminase generated by reassembly of the subtilisin-cleaved bovine ribonuclease A (RNase A) with a synthetic peptide fragment carrying a pyridoxal derivative (Figure 3). One reassembled construct displayed considerable rate acceleration (18 fold) for the conversion of L-alanine to pyruvate under single turnover conditions compared to the pyridoxal-derivative carrying the peptide fragment alone. Accelerating effects of metal ions observed with simple pyridoxal models had been studied previously and consequently their addition was probed in this complex system. 139 Cu(II)-ions had either beneficial or detrimental effects on the rate depending on the peptide fragment employed. 140 In a subsequent study, multiple turnovers could be realized in the conversion of pyruvate to alanine through the addition of L-phenylalanine accompanied by a moderate enantioselectivity. 141 Figure 4). Surprisingly, wild type trypsin was also able to cleave this substrate in the presence of Zn(II) at 10-fold lower rate, although cleavage after tyrosine is highly disfavored in the absence of zinc. This unexpected result was rationalized by complexation of the substrate with the participation of glutamate 151. Indeed, exchange of glutamate for glutamine led to a mutant unable to cleave the peptide AGPYAHSS. The metal binding site was modeled based on the X-ray structure of a rat-trypsin mutant. ) of adipocyte lipid binding protein (ALBP). 142 The X-ray structure confirmed the positioning of the phenanthroline unit inside the protein cavity (PDB ID 1A18). 143 Complexation with Cu(II) resulted in a construct (ALBP-Phen-Cu(II)) that accelerated the hydrolysis of the activated picolinic acid ester 24, and, more importantly, a small set of racemic alanine, serine, and tyrosine esters. The L-enantiomer of the isopropyl ester of alanine 25 was converted significantly faster than the D-enantiomer resulting in an ee of 86% after 24 hours and around 1 turnover, whereas around 8 turnovers were realized with the methyl ester of tyrosine 26 (39% ee, Scheme 8). In a subsequent study, the position of the phenanthroline inside the protein cavity was varied by site directed mutagenesis to place the linking cysteine residue in different locations.
The enantiomeric excess for the hydrolysis of the isopropyl ester of alanine could be improved to 94%, although only relatively minor improvements resulted from altering the cofactor linkage site (Figure 5). 144 These remarkable studies mark the first examples of enantioselective catalysis involving small molecule substrates using ArMs generated via covalent scaffold modification. formation. 145 One study scrutinized the effect of Cu(II), Ni(II), and Zn(II) on the catalytic performance of a range of constructs, including some where the location of the cysteine in the protein cavity required for immobilization of the artificial cofactor had been altered by site directed mutagenesis. 146 Added metal ions were observed to result in rate acceleration but also erosion of enantioselectivity for the transamination of keto-carboxylic acids with phenylalanine as a sacrificial donor. Both a Cu(II)-Phen dependent hydrolase and a pyridoxal-derivative dependent transaminase were characterized by X-ray crystallography. 143 Inspired by reported phytase inhibition by transition metal oxoanions and the structures of vanadium chloroperoxidases, R. Sheldon and coworkers doped phytase from Aspergillus ficuum with vanadate. 147 The resulting ArM oxidized thioanisole 27 in the presence of H2O2 at considerably increased rates relative to free vanadate and provided an ee of up to 66% (S)-28 (Scheme 9). Unexpectedly, phytase even in the absence of vanadate also catalyzed the reaction, albeit at reduced rates and enantioselectivity. The original study was published in 1998 and was followed by further investigations communicated in 2000. [148][149] These later studies examined a range of protein hosts from various sources, e.g. acid phosphatase, apo-ferritin, aminoacylase, sulfatase etc. with vanadate, and phytase doped with alternative transition metal oxoanions, namely MoO4 150 The antibodies (33F12 and 38C2) were originally raised against a hapten able to form a covalent intermediate with a reactive lysine residue in the antibody active site via reactive immunization. [151][152] A wide range of metal ions was examined, but substantial effects were only observed for Pd(II) salts: rate enhancements of 2.3-2.6 fold with variable enantioselectivity. The authors concluded that a direct binding of Pd in the active site seemed unlikely given the moderate effect and favoured instead an allosteric interaction as explanation for their observations. Following Whiteside's -at the time two-decade old -lead, A. S. C. Chan and coworkers published in 1999 a study in which more rigid and, importantly, enantiopure pyrphos ligand was biotinylated and the corresponding Rh(I)-complex 29 incorporated into avidin. 153 The authors studied the effect of hydrogen pressure, temperature and stereochemistry of the pyrphos moiety on the catalysis outcome in the hydrogenation of itaconic acid 30, Scheme 10. Although stereoselectivity was improved in the presence of avidin and could even be inverted compared to the protein free cofactor, only moderate enantioselectivities were reported.
Scheme 10. Asymmetric hydrogenation of itaconic acid catalyzed by an enantiopure biotinylated Rh-diphosphine complex in the presence of avidin.
S. Nimri and E. Keinan reported in the same year on antibodies raised against a modified hapten compared to Keinan's previous design from 1990 (Figure 2 c). 154 Apart from a variation of the porphyrin scaffold (TCPP instead of TCP), the axial hydoxo-ligands of the tin center were exchanged for bulky a-naphthoxy-groups with the intention of creating a substrate binding site in the antibody scaffold. Antibody metallation with TCPP-Ru(II)-CO led to an active and enantioselective ArM sulfoxidase in the presence of iodosobenzene.
A kcat of 24 min -1 , a KM of 10 mM, and an ee of 43% were obtained for the oxidation of thioanisole 27, although only 75% of the cofactor was bound to the antibody even using ~13:1 antibody:cofactor ratio. A significantly higher kcat-value (174 min -1 ) was determined for the electron-rich substrate 4-methoxythioanisole, albeit at the expense of a diminished enantioselectivity (27% ee).
M. Marchetti and coworkers studied the biphasic rhodium-catalyzed hydroformylation of alkenes in the presence of a range of protein scaffolds. The initial study was published in 2000 and followed by a more detailed investigation in 2002. [155][156] Most successful system reported was generated by combining human serum albumin with [Rh(acac)(CO)2]. A TON of 741,000 was observed at a TOF of 30,000 h -1 . At higher catalyst loadings (S : C = 10,400 : 1) good recyclability was demonstrated. 1,1-Diarylethenes were not viable ArM substrates despite their reactivity toward the common TPPTS-Rh(I) system (TPPTS = triphenylphosphine-3,3',3''-trisulfonic acid trisodium salt). A branched : linear-ratio of 90 :
10 was observed for styrene, whereas only a qualitative comment on stereoinduction (i.e. very low, but observable) was made. Stereoselectivity in biphasic hydroformylation reactions is generally considered to be problematic. 157 In 2002, K. M. Nicholas, K. D. Jandas, and coworkers published an unexpected application of a monoclonal antibody for the generation of an artificial metalloenzyme. 158 Whereas previously antibodies were raised against metal complexes or ligands for the generation of ArMs, the authors employed an antibody selected for a completely unrelated activity, namely aldolase activity, as a protein host for a copper(II)-bis-imidazole complex.
The antibody 38C2 discussed above contained a reactive lysine residue of low pKa inside its binding pocket. [151][152] This lysine residue was modified with an electrophilic derivative of the copper(II)-bis-imidazole complex and the resulting construct successfully employed for the hydrolysis of picolinic ester 24 (kcat = 2.3 min -1 , Km = 2.2 mM), while it was inactive for benzylic hydroxylation with ascorbate/O2 or t-BuOOH.
The advent of increasingly convenient site directed mutagenesis in the late 70's (Nobel prize in chemistry for M. Smith and K. B. Mullis in 1993) opened new opportunities to optimize natural enzymes enzymes for biocatalysis. [159][160] The possibilities for biocatalyst development were even further expanded by the development of directed evolution methods involving iterative round of random mutagenesis or gene shuffling followed by screening or selection. 161 These procedures involve large theoretical library sizes compared to site directed 'design'-protocols, but they enable identification of mutations that are not obvious from structural data. Moreover, the ability to simply iterate the mutagenesis and screening procedure eliminates the need to exhaustively (or even remotely) sample the full library diversity. A wide range of site-directed and random mutagenesis strategies are now commonly used to rapidly optimize biocatalyst activity. 162- 163 Despite these advances, the considerable effort required to express, purify, modify, and evaluate protein scaffolds for ArM catalysis posed and continues to pose a considerable obstacle to optimizaing ArMs via mutagenesis. In 2002, M. T. Reetz proposed in conceptual articles the application of directed evolution protocols to hybrid catalyst development for non-biological reactions such as hydroformylation, olefin hydrogenation, metathesis, and allylic substitution. [164][165] In contrast, T. R. Ward and coworkers initially selected a chemo-genetic strategy to reduce the number of ArM components (protein scaffolds and cofactors) and simplify ArM optimization. This strategy involves the combination of a small set of protein mutants with a small set of cofactor variants to generate ArM diversity. 31 Many exciting studies on systems related to ArMs but outside the scope of this review as outlined in the introduction were reported during the timeframe discussed in this section.
These include active site redesign of natural metalloenzymes to elucidate structurefunction relationships, 110 and early efforts on de-novo designed coiled-coil and helical bundle metallopeptides that would later be used for ArM catalysis. [166][167][168] Considering the widened opportunities for rational design guided by structural information and molecular modelling, standardized and new molecular biology protocols and kits, the renewed interest in water compatible transition metal catalysts and, a few years later, affordable gene synthesis, the ground was set for further research endeavors in the field of artificial metalloenzymes.
The discovery of homogeneous, phosphine-transition metal complexes capable of promoting the addition of molecular hydrogen to carbon-carbon double bonds ultimately precipitated the entire field of asymmetric transition metal catalysis. 101,169 Factors contributing to the importance of this reaction for organic synthesis include its i) perfect atom economy, ii) high reaction rates, iii) high tunability via ligand design, and iv) ability to generate valuable, chiral products. Unsurprisingly, asymmetric hydrogenation has also been extensively explored as a benchmark reaction for artificial metalloenzymes.
As previously noted, Whitesides' seminal example of asymmetric hydrogenation using a biotinylated Rh(I)-bisphosphine complex embedded within avidin constituted not only the first example of ArM-catalyzed hydrogenation, but also the first example of asymmetric catalysis using an ArM and the first example of ArM formation via non-covalent cofactor anchoring. This work was first revisited by A. S. C. Chan and coworkers in 1999 (Scheme 10). 2, 153 M. T. Reetz later demonstrated that Cu-, Pd-, and Rh-complexes of maleimidesubstituted dipyridine compounds (Figure 6) could readily alkylate the active site cysteine of papain, a cysteine protease. Only preliminary catalytic activity was reported for these ArMs, which were said to catalyze hydrogenation with low enantioselectivity. [164][165] Figure 6: Dipyridine complexes by Reetz. 164 Initial efforts in the Ward group on building an artificial hydrogenase aimed at improving the original Whitesides system. Instead of avidin, the close homologue streptavidin (pI = 6.2) was used, based on the idea that the high pI of avidin (pI = 10.4) might significantly reduce the affinity of cationic complexes such as [Rh(COD)(34)] + for the host protein.
Artificial hydrogenase [Rh(COD)(34)] + × WT Sav was used for the reduction of Namidoacrylic acid to yield (R)-N-acetamidoalanine 31 with an ee of 92 % and quantitative conversion with 1 mol % catalyst loading. To obtain a structural model of the hydrogenation catalyst, docking studies were performed. Aminoacid S112 was found to be located in close proximity to the position of the rhodium moiety. 170 Thus, Sav variants bearing the remaining nineteen proteinogenic amino acids in position S112 were expressed and purified. In addition, a library of 22 biotinylated ligands was produced consisting of two different bisphosphine ligands (34, 36) and eleven spacer units. 44,[171][172][173][174][175][176] All 20 Sav isoforms were screened in the presence of the 22 biotinylated Rh-complexes and two N-protected dehydroamino acid substrates to yield 31 and 32. The observed results (Scheme 11 and Table 3) and conclusions can be summarized as follows: i) Generally, the "chemical" Rh-complex library contributed more to diversity than the "genetic" Sav-mutant library.
ii)
The more flexible bisphosphine 34 scaffold afforded more active and selective ArMs than the rigid six-membered chelate formed with 36.
iii)
The flexible 34 is likely to adopt an enantioenriched conformation inside the scaffold protein reminiscent of the "induced lock-and-key" hypothesis.
iv)
The two substrates displayed similar reactivity and selectivity profiles. This suggested a broad substrate scope of the artificial transfer hydrogenase, reminiscent of homogeneous catalysts.
v) Mutation at position S112X enabled an inversion of enantioselectivity (Table 3, entries 13 and 14) vi) Good (S)-selectivity was obtained with complex 40 inside 'cationic' S112X Sav mutants (X = H, K, R; Table 3, entries 2-4).
vii)
The (S)-selectivity could be further improved by introduction of (R)-proline as a spacer between the biotin anchor and the flexible 34 using either Avi or Sav as host protein. In early studies, van Koten and Klein-Gebbink demonstrated that lipase inhibitors, such as nitrophenol phosphonate esters, can be used to selectively anchor metallopincers within Cutinase (44). [177][178][179] Cutinase from Fusarium solani pisi is a 21 kDa lipase containing a Ser120-Asp175-His188 catalytic triad that enables hydrolysis of fatty acids.
Phosphonate esters function as transition-state analogues for this reaction and react irreversibly with Ser120. A pincer-metal complex linked to a phosphonate moiety (45 or 46) may thus be suitable for covalent anchoring to cutinase. (Scheme 12). M. T. Reetz and coworkers discussed related studies with an esterase in 2002, but the resulting construct suffered from low hydrolytic stability. They suggested to improve hydrolytic stabilty by exchanging a nitrophenoxy-group at phosphorus for an ethoxygroup, a strategy later realized by van Koten and Klein-Gebink. 164 Scheme 12. Covalent anchoring of pincer-metal complexes via nucleophilic substitution of an activated phosphonate ester Ser 120 in the active site of cutinase 44. The anchoring is conveniently monitored by the release of para-nitrophenolate.
van Koten and Klein-Gebbink subsequently extended their proof-of-principle approach with Pd-and Pt-pincer complexes to Rh-based systems for the creation of artificial hydrogenases. 180 Building on the work of Kühn et al. concerning the hydrogenation of acetophenone in aqueous media with rhodium N-heterocyclic carbenes (NHCs), 181 Klein-Gebbink covalently linked the lipase inhibitor 47 within Cutinase 44 and CalB (33.5 kDa, lipase from Candida antarctica), Scheme 13. 182 Scheme 13. Covalent anchoring of a Rh(NHC) in serine-hydrolases for the creation of artificial hydrogenases.
The resulting hybrids 47•Cutinase and 47•CalB (obtained using Cutinase and CalB respectively), were tested in the hydrogenation of acetophenone 48, and methyl 2acetamidoacrylate 49 (Scheme 14). Acetophenone was reduced in 90% yield (18 TON) with the free Rh-NHC catalyst (Table 4, entry 1), while the Rh-ArMs showed a reduced catalytic activity, yielding 1-phenylethanol 50 in 27% and 0% yield for Rh-cutinase 47•Cutinase and 47•CalB, respectively (Table 4, entries 2 and 3). For the hydrogenation of the olefinic substrate, both artificial hydrogenases afforded complete conversion to methyl-acetylalaninate 51, (Table 4, entries 5 and 6), highlighting the preferred reduction of olefins over ketones. 183 Accordingly, 47•CalB was fully chemoselective for the hydrogenation of olefin 49 (Table 4, entries 3 and 6), highlighting the benefit of the deep binding pocket provided by the active site in CalB to discriminate between both substrates. Similar results were observed for 47•Cutinase at shorter reaction times (Table 4, entries 7-10). Unfortunately, no enantioselectivity was observed for any of the systems reported.
These examples highlight the potential of second coordination sphere interactions provided by a host protein to discriminate between similarly reactive substrates. Kamer demonstrated that a non-catalytic cysteine residue in apo-photoactive yellow protein (PYP) could be selectively acylated with a carboxylic imidazolide-substituted bisphosphine. This phosphine could also be metallated with [Rh(cod)(MeCN)2]BF4 (52) and subsequently reacted with PYP to generate an artificial hydrogenase, which catalyzed the reduction of dimethyl itaconate (53) with full conversion but low ee (Scheme 16). 185
Scheme 16. Hydrogenation of itaconic acid methyl ester catalyzed by 52.
A. Harada and coworkers elicited a monoclonal antibody 1G8 against an achiral rhodium complex 54, Scheme 17. 186 In the hydrogenation of the 2-acetamidoacrylic acid 8, the antibody 1G8 in combination with the hapten 54 afforded the reduction product with > 98% ee (S) and a TON of 854. In contrast, racemic product was obtained with either 54 or 54•BSA (BSA = bovine serum albumin) accompanied by lower TONs, Table 5. [Rh(cod)2]BF4. 187 In addition to Rh incorporation in the CA active site, Rh also bound to the surface of the scaffold (Rh : hCAII = 6.5 by ICP-MS), leading to cis/trans isomerization (hydrogenation : isomerization = 6.3), Scheme 18. Site-directed mutagenesis was used to remove 9 histidine residues on the surface of hCAII that were believed to be involved in non-specific Rh-binding. The resulting ArM, 9*His-hCAII-[Rh], was found to have a reduced Rh : hCAII ratio of 1.8, provided an increased hydrogenation : isomerization ratio of 20.6, and displayed improved specificity for cis-stilbene over trans-stilbene, Table 6. Eppinger designed cofactors comprised of not only a catalytically active metal complex and a reactive group (epoxide) to covalently link the cofactor to protein scaffolds, but also a non-covalent amino acid recognition element to control cofactor orientation within the scaffold. 188 Half-sandwich ruthenium and rhodium cofactors were constructed in this manner and linked to cysteine proteases in the papain family to generate ArM hydrogenases. Ketone hydrogenation could be achieved with up to 64% ee for pchloroacetophenone as substrate (Table 7 andScheme 19). 189 Renaud, Ward and coworkers investigated the potential of biotinylated Knölker-type complexes (Figure 7) to catalyze the enantioselective hydrogenation of aromatic ketones and imines inside Sav. 190 A library of seven catalysts was synthesized with a biotin anchor tethered to the cyclopentadienone moiety via various linker groups. Hydrogenation of ketone a,a,a-trifluoro-acetophenone afforded the corresponding secondary alcohol with 20 TON but only 9 % ee. The best enantioselectivity (34 % ee (R)-product) was obtained in the conversion of p-methoxyacetophenone albeit with only a single turnover.
Today, scientists around the world face the challenge of finding a sustainable source of energy to support tomorrow's societal needs. In this context, valorization of solar energy is regarded by many as the "Holy Grail". However, efficient ways to convert solar energy directly to electricity or to a fuel, for storage and transportation, remains challenging. The development of artificial photosynthesis would allow conversion of sunlight into a fuel.
Hydrogen (H2) is one of the most explored potential fuel candidates. 60 Interestingly, its production from light and water has been observed in a living organism, and biocatalytic hydrogen production is catalyzed by hydrogenases. These are classified according to the nature of their active site, namely [FeFe]-, [NiFe]-, and [Fe]-hydrogenases. They catalyze the generation of H2 from protons and electrons provided by the electron transport chain of photosynthesis under near thermodynamic equilibrium conditions (i.e. no overpotential required) with high catalytic rates (TOF up to 20 000 s -1 ). 68 Interestingly, hydrogenases operate in both in the forward and reverse reaction and approach the efficiency of platinum.
The technological applications of natural hydrogenases are impaired by (i) their biosynthesis, which requires complex maturation machinery, (ii) their air sensitivity, and
(iii) their high molecular weight, which limits the density of active sites that can be immobilized on an electrode. 58 The former limitation has been recently addressed by Fontecave and coworkers with the
[FeFe]-hydrogenase HydA. The natural formation of its di-iron catalytic site (H-cluster) requires a complete maturation machinery, involving at least HydE, HydG and HydF.
HydE and HydG provides the iron-ligands (CO, CN -, aza-propanedithiol), and HydF is the scaffold protein in which the di-iron complex is ultimately assembled. In the last step, the nearly-mature di-iron complex is transferred to apo-HydA. Fontecave and coworkers demonstrated that a hydrogenase could be maturated in vitro, without the need of the entire maturation machinery. By loading an inactive artificial di-iron cofactor into apo-HydF, they could transfer the synthetic complex into the apo form of the hydrogenase HydA, reconstituting a fully active HydA (Scheme 20, path a)). 191 This work unambiguously settled the controversy concerning the nature of the central atom of the bridging dithiolate ligand: catalytic activity was detected only in the presence of a nitrogen atom at this position (X in Scheme 20). Building on these results, Happe and Fontecave demonstrated that the use of HydF could even be avoided, that is, a direct loading of a synthetic complex into apo-HydA is possible (Scheme 20, path b). 192 Building upon this strategy, Hu and coworkers reported the direct reconstitution of an [Fe]-hydrogenase by loading its apoform with a synthetic complex (Scheme 20, path c)). 193 Scheme 20. Reconstitution of apo-hydrogenases with synthetic cofactors affords functional hydrogenases.
Besides shedding light on a long controversy concerning the key features of the cofactor, this methodology demonstrated that it is indeed possible to generate an active hydrogenase without the need of its complex maturation machinery. These results are of particular interest as: i) they allow the expression of the hydrogenase in a heterologous organism and ii) they open the possibility of introducing a fully synthetic cofactor within an apo-hydrogenase. The former feature should allow large screening campaigns of hydrogenases (e.g. homologous, directed evolution), to identify variants with higher activity and/or oxygen-resistance. As demonstrated by Hu and Lubitz, subtle variations of the cofactor lead, in most cases, to inactive hydrogenases, thus allowing to establish structure-activity relationships for the cofactor. 194 This strategy relies on a highly evolved scaffold, optimized for hydrogen production from its constituents (i.e. presence of proton-and electron-channels leading to and from the active site). The drawbacks of this strategy lie in the complexity and the size of the scaffold.
For technological applications, one would favor a scaffold reduced to its minimum size and with a structure that is stable over a wide range of conditions. With large scale applications in mind, some groups explored the possibility of assembling synthetic complexes within various protein scaffolds to produce hydrogen. Only a limited number of catalysts and protein scaffolds have been used in this strategy. Alternatively, the metalsubstitution strategy to repurpose natural metalloenzymes for hydrogenase activity has also been used. In this section, we summarize the reported studies fitting into either category, with a particular emphasis on the catalytic activity of the resulting artificial hydrogenases. Several recent papers have also reviewed other aspects of artificial hydrogenases. 57,59,61,66,68,195 Hayashi et al. inserted synthetic complexes that mimic the active site of [FeFe]hydrogenases into various host proteins. They datively anchored a di-iron complex into the active site of apo-cytochrome c. 196 This protein naturally contains a Feprotoporphyrin(IX) cofactor that is covalently attached to the protein via thioether linkage at a Cys-X2-Cys motif. After removal of the heme, the Cys-X2-Cys motif of the apo- TON in 2h), with a maximum observed TOF of 0.035 s -1 (Table 8, entry 1). Notably, the electron transport from the PS to the active site is proposed to be the limiting factor in the reaction. Although the structure of this artificial hydrogenase is poorly defined, the protein environment provides robustness to the catalytic activity. When compared to 62 bound to a synthetic heptapeptide containing the Cys-X2-Cys motif and a His residue to mimic the active site of cyt c, the entire protein scaffold performed better (i.e. ~4 fold higher TOF and ~8 fold higher TON), demonstrating a protective role of the protein environment.
Next, Hayashi et al. selected a more rigid scaffold to engineer an artificial hydrogenase.
The di-iron cofactor [(µ-S)2Fe2(CO)6]-moiety was anchored via a maleimide linker to the cysteine of apo-nitrobindin Q96C (Figure 8). 197 As suggested by the docked structure, the synthetic complex 62, (Scheme 21) is embedded in the cavity formed by the β-barrel of nitrobindin, surrounded by a well-defined protein environment. Upon irradiation, the 63•nitrobin generated hydrogen with a maximum TOF of 0.038 s -1 and up to 130 TON after 6 h of irradiation (120 after 2 h) (Table 8, entry 2). Under similar conditions, the free cofactor 63 performed c.a. 3 fold faster. This phenomenon is most likely due to a decreased accessibility of the complex inside the protein, decreasing the rate of electron transfer from the PS. Despite this decrease in rate, it is clear that the protein provides a stabilizing environment to 63 since its deactivation is slower than in solution. Compared to 62•cytochrome c, the activity of 63•nitrobin was less sensitive to the pH: raising the pH to 7.8 decreased the activity of the former by 93 % versus 72 % for the latter (Table 8, entry 3). In addition, due to the well-defined environment around the active site, improving catalytic activity by rationally mutating residues around at the β-barrel cavity can be envisioned. Another example of an artificial hydrogenase containing a biomimetic-center was recently reported by the group of Shafaat. This work is reminiscent of the pioneering work from the groups of Moura and LeGall. [201][202] In this example, the aim was not to mimic the active site of the [FeFe]-hydrogenases, but those of [FeNi]-hydrogenases. For these systems, the nickel, coordinated by four cysteinate residues, is the redox-active metal operating during the proton reduction. The first coordination sphere is reminiscent of rubredoxin. An iron to nickel substitution in the active site of rubredoxin yields an enzyme, termed NiRd, able to produce hydrogen, Scheme 22, a). When irradiated in presence of [Ru(bpy)3] 2+
and ascorbate, in phosphate buffer at pH 6.5, NiRd generated hydrogen at a rate of 0.0083 s -1 and performedapproximately 70 TON over 8 h (Table 8, entry 4). Using Ti(III)-citrate as a light-independent electron source, NiRd performed up to 300 TON in 8 h, at 22 °C at pH 9.4 (Table 8, entry 5), suggesting that, with a strong electron source, NiRd can generate hydrogen at pH > 7. Importantly, most of the catalyst can be recycled, indicating that the inactivation is not due to protein degradation. However, this system requires a high overpotential (540 mV at pH 3-5). It is proposed that mutagenesis of residues close to the Ni-center may help to improve the catalytic properties of this artificial hydrogenase.
Metal substitution in natural metalloenzymes yield artificial hydrogenases. The iron of rubredoxin can be replaced with nickel a). The iron in a heme protein can be substituted by cobalt either by demetallation followed by incubation with excess of [Co(Ac)2] b), or by removing the heme from the enzyme and inserting a [CoPP(IX)] complex c).
Following in Hayashi's footsteps, other groups have designed artificial hydrogenases based on heme-proteins. This family of proteins has been used extensively for their ability to bind heme moiety substituted by other metals as well as other planar complexes containing non-porphyrin ligands. 203 In addition, the potential of Co-porphyrins and cobaloximes for hydrogen production in aqueous solutions has been known since the 80's. [204][205] The latter usually require a smaller overpotential that the former, and both exhibit promising oxygen tolerance properties.
Bren and coworker demonstrated that substituting iron by cobalt in enzymes containing iron-protoporphyrin(IX) affords efficient artificial hydrogenases. The iron of the hemeprotein cytochrome c552 from Hydrogenobacter thermophilus (Ht c-552) was substituted with cobalt, taking advantage of the thioethers linking the protoporphryrin(IX) ligand and the cysteins of the protein that retains the porphyrin even under forcing conditions Scheme 22 b). 61 In this scaffold, the histidine residue acting as axial ligand was conserved.
However, the distal methionine was mutated to alanine (M61A) to keep a free coordination site for proton reduction. Though no data was presented for photocatalysis, the electrocatalysis experiments showed remarkable stability for proton reduction: 11,000 TON after 6 h and up to 27,000 TON in 24 h at pH 7.0 (Table 8, entry 6). The major drawback of this system lies in the high overpotential that is required (i.e. 830 mV). On the other hand, comparing the performance of Co•Ht c-552 and CoMP11-Ac, an eleven aminoacid proteolytic fragment of the horse cytochrome c containing a cobalt-substituted heme, highlights the beneficial effect of a full protein environment (for a related structure of Fe-microperoxidase-8, see 217). 206 CoMP11-Ac exhibits interesting proton reduction activity coupled with a remarkable oxygen tolerance. Although the cobalt ion possesses an identical first coordination sphere, it is exposed to the solvent in CoMP11-Ac but buried in the protein in Co•Ht c-552. The surrounding protein environment results in a longer catalyst lifetime (~24 h against ~4h), allowing to reach a ~10 fold higher TONs for the former. This clearly illustrates the protection provided by the second coordination sphere of the host protein. The presence of a defined protein environment however did not lead to a decrease of the overpotential required (830 vs 850 mV).
Ghirlanda and coworkers used a cobalt-substituted myoglobin termed 64•Mb (i.e. dative anchoring via a His and no cysteine thioether linkages, Scheme 22 c). 207 Upon irradiation at pH 7.0 in the presence of ascorbate and a photosensitizer, 64•Mb catalyzes the production of hydrogen in up to 518 TON in 12 h (Table 8,entry 7). This corresponds to a This increase may arise from a flexibilization of the protein structure in the proximity of the catalyst, facilitating interactions of the protein-bound cofactor with the photosensitizer (i.e.
facilitating the electron transfer). (Figure 15 for Mb active site).
More recently, Ghirlanda and coworkers applied a similar strategy to anchor 64 within cytochrome b562. 208 In this system, the metal of the heme is anchored via H102 and M7 (Figure 9). Mutation of the methionine-7 to an alanine (M7A) results in a 2.5 fold increase in photocatalytic activity (305 vs 125 TON), as it liberates one coordination site for catalysis. Continuous removal of hydrogen increased 64•b562 M7A TON to 1450 (Table 8, entry 8). Even in the presence of air, 64•b562 M7A afforded up to 400 TON. Upon catalyst stalling, addition of fresh photosensitizer ([Ru(bpy)3] 2+ ) partially restored catalytic activity, while addition of fresh 64•b562 M7A did not. This suggests that PS degradation is the primary limiting factor in this experiment rather than catalyst inactivation.
Next, 64•b562 M7D and 64•b562 M7E mutations were investigated. The authors speculated that the presence of a carboxylic acid in the vicinity of the catalytic center could enhance hydrogen production, similar to protonated amines. [209][210] Although better than the WT host, the activity of these carboxylate-bearing mutants however did not surpass that of the M7A mutant, suggesting that these residues do not behave as proton relays. Photochemical hydrogen production by cobaloxime•Mb ArMs was also performed using deazaflavin (DAF hereafter) and tris buffer as electron donor. At pH 7.0, 66•Mb performed up to 5 TON in 15 min (Table 8, entry 9). In this case however, the free cofactor 66 outperformed the artificial hydrogenase, TON 8.3. Other heme-binding proteins were also investigated as scaffolds to generate ArM hydrogenases using either 65 or 66. These included: rat heme oxygenase 1 (HO1 hereafter) and Corynebacterium diphteriae heme oxygenase (HmuO hereafter). 211 The resulting artificial hydrogenases 66•HO1 and 66•HmuO were more active than 66•Mb both under thermal-and photochemical conditions. 66•HO1 converted up to 82 % of the electron donor [Eu(EGTA)] 2+ into H2 within 5 min, corresponding to 6.2 TON (Table 8, entry 10). Both TON and TOF were higher than that of 66 free in solution. However, these did not outperform the free catalyst 66 under photocatalytic conditions: at 10 µM concentration, 66 afforded 20 TON in 15 min, versus 6.3 and 15.3 for 66•HO1 and 66•HmuO respectively (Table 8, entries 11 and 12). Despite the low overpotential required and the shielding effect provided by the protein, these artificial hydrogenases suffered from a poor affinity of the host for the cofactor. This may be traced back to the absence of interaction with the propionate side-chains of the ligand, as well as significantly reduced hydrophobic interaction between the cobaloximes and the apo-heme proteins, compared to the natural cofactor-protein interaction.
Scheme 23. Cobaloximes and nickel phosphines are synthetic complexes with high hydrogen production catalytic activity. Cobaloximes can be anchored via a histidine or a methionine residue of the apo-heme protein.
The artificial hydrogenases of Artero and coworkers proved to be better catalysts than the free cobaloximes at pH 7.0 in the case of thermal catalysis (using [Eu(EGTA)] 2+ as electron donor), while in photocatalysis, the free catalysts performed better. This could arise from the shielding effect of the protein, obstructing interaction between the PS and the anchored catalyst. In addition, the deactivation of the catalyst occurred rapidly. The This issue has been directly addressed by Utschig and coworkers. In a first attempt, Utschig, Tiede and coworkers made use of the direct interaction between the synthetic complexes 67 and 68 with photosystem I (PSI hereafter). [212][213] This natural component of the photosynthetic electron transport chain possesses both a light-harvesting center and a series of electron relays, scaffolded within a large protein assembly. They demonstrated that 67 and 68 interact with PSI in a non-covalent manner. In contrast to other artificial hydrogenases described herein, the interaction between PSI and the two complexes were not unambiguously defined: the authors hypothesize that interaction may occur at the hydrophobic patches of PSI (and possibly at a histidine residue in the case of 67).
Following incubation of 67 with PSI, ascorbate (as sacrificial electron source) and cytochrome c6 as electron transport (to reduce the oxidized PSI), afforded 67•PSI. Upon irradiation, dihydrogen evolution could be detected. Relative to the cobalt content, this system performed 2080 TON in 1.5 h, with a maximum TOF of 1.13 s -1 , at pH 6.3 (Table 8, entry 13). Under similar conditions, 68•PSI performed 1870 TON in 3 h (Table 8, entry 14). For both 67 and 68, analysis of the catalytic mixture upon completion of the reaction revealed thatapproximately 90 % of the metal was lost. The poorly defined and weak interactions between PSI and those catalysts combined with possible catalyst degradation pathways may lead to leaching during catalysis. To overcome this limitation, the authors combined 68 with apo-flavodoxin (Fld), a protein known to interact via electrostatic interactions with the FB cluster of PSI. The authors hypothesized that 68 could be inserted is the FMN pocket of apo-flavodoxin. The metal content of the assembly was quantified, confirming this hypothesis. Irradiation of PSI in the presence of 30 equivalents of 68•flavodoxin led to the rapid production of dihydrogen relative to PSI, improving the photon to hydrogen conversion. However, when converted to catalyst concentration (ie.
30 eq. vs. PSI), only 94 TON were obtained after 4 h (Table 8, entry 15). Despite its efficiency and elegant design, this biomimetic artificial hydrogenase is challenging to handle because of the membrane-bound nature of the proteins.
To reduce the complexity of the above system, Utschig and coworkers recently reported a simpler design, in which both the synthetic PS and the catalyst are anchored on a ferrodoxin protein, acting as scaffold naturally containing an electron relay. [214][215] This latter feature, as well as the defined and stable structure of the scaffold, differentiates their work with previous report from Hayashi and coworkers, in which a PS and a catalyst were attached on a peptide. 216 Ferredoxin is a small and soluble electron carrier protein In summary, the artificial hydrogenases reported to date show that anchoring a catalytically competent proton-reduction catalyst within a protein scaffold can impart compatibility with aqueous media at neutral pH and increase both TOF and TON. In some cases however, the artificial hydrogenases are less efficient than the cofactor alone. This is especially valid for photocatalytic systems whereby the PS catalyst interactions are hampered, eventually leading to slow catalytic rates or catalyst degradation. The recent studies from Utschig and coworkers offer promising avenues to overcome this issue.
Some advantageous aspects of artificial hydrogenases remain to be systematically explored. For example, very few reports on random mutagenesis or directed evolution schemes have been published, although such avenues are often suggested in the outlook of these publications. Only a handful of modifications in the direct environment of the metal have been tested either by mutating residues or by testing different scaffolds from a same family (e.g. heme-binding proteins). None of these examples led to a dramatic improvement of catalytic activity however. In contrast, directed evolution could be used to tune the protein environment, rather than the PS or the catalyst. 217 Finally, in a biomimetic spirit, artificial hydrogenases could be improved by engineering facilitated pathway for proton/electron delivery/removal at the metal center, and gas release pathways, reminiscent of the channels and electron relays present in natural hydrogenases.
Background. Catalytic asymmetric transfer hydrogenation (ATH) provides a powerful means for reducing unsaturated substrates including ketones, imines, nitro compounds, nitriles, oximes, a,b-unsaturated carbonyl compounds, heterocycles, alkenes, and alkynes. [218][219][220][221][222][223][224][225][226] It requires neither hazardous hydrogen gas nor pressure vessels. Instead, a number of cheap, non-toxic, and easy to handle reductants, including formic acid, formate salts or isopropanol, have been used as hydrogen donors. Recent advances and trends in ATH using homogeneous, heterogeneous, organo-and transition-metal catalysts were recently reviewed. 227 A major breakthrough in ATH was reported by Noyori and coworkers in 1995: enantiopure
diphenylethylenediamine) were found to be outstanding homogeneous catalysts for the asymmetric reduction of a variety of substrates. [228][229][230] Building upon this finding, some of the most effective catalysts reported to date are d 6 -pianostool complexes of rhodium, iridium and ruthenium. As amply demonstrated by Xiao and coworkers, closely related homogenous catalysts are efficient for the asymmetric reduction of prochiral ketones and cyclic imines in water. [231][232][233] This is obviously an important feature in the context of ArMs.
In nature, NAD(P)H-dependent enzymes including ketoreductases, imine reductases (IRED), and ene reductases can be used for the asymmetric reduction of C=O, C=N and activated C=C bonds respectively. While both keto-234 and ene-reductases 235 are well known and widely used on industrial scale, 162 IREDs were only recently discovered by Nagasawa in 2011. 236 These have been further exploited by, among others, the Turner and Hauer groups. 237,238 In this section, we summarize the progress in artificial transfer hydrogenases (ATHases)
for the asymmetric reduction of prochiral ketones, enones, imines as well as NAD(P) + and its analogs. For several groups, ATHase has provided a fertile playground to benchmark various challenging concepts including: catalyst immobilization, directed evolution, enzyme cascades, allosteric regulation etc. These are presented at the end of this section.
Transfer Hydrogenation of Ketones. To assemble an artificial metalloenzyme for the asymmetric transfer hydrogenation of prochiral ketones, the Ward group set out to adapt Noyori's homogeneous d 6 -pianostool complexes bearing a Ts-DPEN-ligand. 229 69)Cl] + were combined with an Sav sitesaturation mutagenesis library Sav S112X as well as selected single-and double point mutants (Scheme 24). 18,239 The aminoacid positions subjected to mutagenesis were selected based on their estimated proximity to the catalytic metal. Instead of screening the whole chemo-genetic diversity matrix, Ward and coworkers first screened all cofactors [(h n -arene)M( 69)Cl] + with a subset of Sav isoforms. Only the best biotinylated catalysts were subsequently screened with the entire set of proteins. The best catalyst×protein combinations were evaluated toward various substrates. Selected results are summarized in Table 9.
Noteworthy features for the ATHases included (Scheme 24 and Table 9):
Complexes bearing a para-substituted spacer (70) outperform those with orthoand meta-spacers.
ii)
Exchange of the capping arene ligand from benzene to p-cymene led to stereoinversion; this is most pronounced with mutants Sav S112K ((S)selective) and Sav S112A ((R)-selective), respectively (Table 9, entries 6 vs. 3).
Aryl alcohols 71-73 yielded good to very good (i.e. ee > 90%) whereas reduction of dialkyl ketones resulted in only modest ees. This is in line with findings for homogeneous pianostool catalysts: the key enantiodiscriminating interaction is between the h n -bound arene moiety and the aryl moiety of the prochiral substrate.
Sav S112X mutants bearing a potentially coordinating aminoacid side-chain have an inhibitory effect on catalysis, resulting in modest conversions.
An X-crystal structure of the most (S)-selective artificial transfer hydrogenase (ATHase hereafter) [h 6 -(benzene)Ru( 70)Cl]×Sav S112K was determined (PDB 2QCB, Figure 10). 240 The following features are apparent:
i)
The Ru-center is located in the biotin-binding vestibule in close proximity to residues S112KA, L124A, S112KB, K121B (subscripts A and B refer to Sav monomers). This observation highlights the importance of scrutinizing mutagenesis libraries Sav S112X, K121X and L124X for the genetic optimization of the ATHase.
ii) A (S)-Ru-Cl absolute configuration was determined for [h 6 -(benzene)Ru( 70)Cl]×Sav S112K. This configuration is reminiscent of the homogeneous catalytic system as both ArM and homogeneous catalyst, (S)reduction products are formed preferentially in the presence of an (S)-Ru-Cl moiety. This suggests that both catalysts rely on a similar enantioselection mechanism.
iii)
The short RuA v)
The overall Sav structure in the complex-bound Sav vs. apo-Sav is virtually identical (RMSD = 0.276 Å).
From the crystal structure, it was concluded that site-saturation mutagenesis libraries in position K121 and L124 may allow to further fine-tune the catalytic performance of the ATHase. Thus, the following libraries K121X, L124X, S112A-K121X, S112K-K121X, S112A-L124X and S112K-L124X were produced to afford eighty double mutants and forty single mutants. This Sav library was combined with both (S)-and (R)-selective cofactors [h 6 -(benzene)Ru( 70)Cl] and [h 6 -(p-cymene)Ru( 70)Cl] respectively. The resulting ATHase were screened for the reduction of aryl-alkyl and dialkylketones 71-75, Scheme 24. To speed-up the screening effort, the Sav mutants were tested in catalysis in semi-purified form. First, mutants were expressed and the cells were lysed.
Next, the lysates were centrifuged and biotin-sepharose beads were added to the supernatant. The sepharose-bound mutants were screened and the best hits were validated using purified and solubilized Sav isoforms. With this procedure, ATHases could be obtained that produced up to 97 % ee (R) and 92 % (S) for aryl-alkylketones 71, 72 and 73 (Table 9, entries 15-18) and up to ee 90 % (R) for dialkylketone 74 and 75 (Table 9, entries 19-21).
Scheme 24: An artificial transfer hydrogenase based on biotinylated Noyori-type d 6pianostool complexes anchored within Sav catalyzes the asymmetric reduction of ketones. Adapted with permission from ref. 239 Copyright 2006 American Chemical Society. respectively. Based on Sadler's report, 248 it seems reasonable to speculate that the Ru-pianostool complex is coordinated to H15 of HEWL. These results suggest that the formation of catalytic compartments upon incorporation of Ru(II) complexes favour the ATH and affords different enantiomers depending on the crystal form used (tetragonal vs. orthorhombic). Transfer Hydrogenation of Imines. Besides the ATH of ketones, the Ward group developed ATHases for the reduction of cyclic imines. In an initial screening, the complex ii)
The opposite enantiomer (S)-87 was obtained with ee 78 % at 5 °C and pH 7.5
in the presence of the cationic mutant [h 5 -(Cp*)Ir( 70)Cl]×Sav S112K.
iii)
The double mutant [h 5 -(Cp*)Ir( 70)Cl]×Sav S112A-K121A displayed a 7.6-fold improved catalytic efficiency compared to WT Sav for substrate 89. 250 iv)
Reducing the ArM concentration to 17 µM for the reduction of 88 resulted in TON > 4'000.
Scheme 28: Artificial transfer hydrogenases for the reduction of cyclic imines result from incorporation of an iridium d 6 -pianostool complex within Sav isoforms.
To identify the transition state for the reduction of isoquinoline 88 in the presence of either [h 5 -(Cp*)Ir( 70)Cl]×Sav S112A/K, Ward, Maréchal and coworkers relied on a custom tailored QM/MM strategy. 251 The computed results nicely corroborated the catalysis results:
i) In the transition state for [h 5 -(Cp*)Ir( 70)Cl] × Sav S112A, the imine cofactor is located in the biotin-binding site opposite to the position of the biotinylated cofactor (see below).
ii) Residue K121 B forms a cation-π with the electron-rich arene moiety of the substrate 88.
iii) The computed difference in transition-state free energy (1.21 kcal • mol -1 ) is in line with the enantioselectivity observed at room temperature. As mentioned above, enantiopure 1,2-aminosulfonamides, such as TsDPEN combined with d 6 -pianostool moieties are priviledged catalysts for asymmetric transfer hydrogenation. 230 Building upon this, Gandolfi-Rimoldi and coworkers investigated the properties of Ru complexes chelated with chiral 1,3-aminosulfonamides for the ATH of prochiral ketone in water. 253 Inspired by the work of Ward and coworkers based on 1,2aminosulfonamides, 38 Pellioni and Rimoldi extended this work to biotinylated substituted 1,3-aminosulfonamides for the reduction of prochiral imines. Both enantiopure and achiral bidentate ligands were evaluated, Figure 11. As for previous studies by Maréchal and coworkers, 252 the host protein dictates the absolute configuration at iridium, which in turn determines the preferred approach of one prochiral face of the substrate 88. Ward and Maréchal coined this chiral relay mechanism "induced lock-and-key" as the host protein favours one metal absolute configuration upon incorporation within Sav. With the diastereopure cofactor in place, the approach of the prochiral substrate is dictated by the Sav host.
Technology. Although incorporation of ArMs in Sav is best achieved via a biotin anchor, additional dative interactions between the metal cofactor and engineered amino acid side chains can be envisaged. This strategy may contribute to firmly localize the metal within the shallow biotin-binding vestibule. To evaluate the potential of this strategy, Ward designed a biotinylated Cp* ligand, thus ensuring localization of a d 6 -pianostool moiety but leaving the remaining three coordination sites available for coordination to amino acid residues and for catalysis. Accordingly, Ward and coworkers designed cofactor [(98)MCl2(H2O)] (M = Rh(III), Ir(III)) inspired by structure-based modeling. 255 The presence of the ethyl spacer projects the metal in proximity to both Sav S112 and Sav K121 positions. The authors hypothesized that the imidazole side chains of mutations S112H and K121H may coordinate thereby influencing the ArMs activity and selectivity. As a model reaction, the asymmetric transfer hydrogenation of salsolidine precursor 88 was selected. The reaction's stereoselectivity was very similar for complexes [( 98 Upon lowering the catalyst loading to 0.0013 mol% 16990 TON were achieved. The selectivity of complex [h 5 -(Cp*)Ir( 70)Cl]×Sav S112A for (R)-87 increased from 79 % ee to 90 % ee upon immobilization. The opposite effect was observed with (S)-selective mutant S112K where the ee decreased from 66 % ee to 56 % ee upon immobilization. The most active mutant tested was S112A-K121A which afforded up to 46'747 TON (in 24 hours) at a catalyst loading of 0.0075 mol%.
Neutralizing Cellular Glutathione for ATHases. Natural evolution of function relies on mutagenesis and selection. In contrast to homogenous catalysts, artificial metalloenzymes bear the potential to be genetically optimized by mutagenesis. Directed evolution relies on the efficient high-throughput screening of genetic diversity to identify and characterize functionally evolved mutants. Parallel in vivo expression of Sav in E. coli (e.g. in 96-well plates) followed by cofactor addition could in principle allow directed evolution of the corresponding ArMs. A challenge in the development of HTS methods for ArMs optimization is the presence of glutathione (GSH) in millimolar concentration in aerobic E. coli cells. 258 The soft acid character of the precious metals in ArMs favors their coordination to GSH. One way to decrease the cellular GSH content is to treat cell lysates with GSH scavengers such as electrophiles or oxidants. Ward and coworkers evaluated the potency of various GSH scavengers to recover artificial transfer hydrogenase activity in cellular media. 259 Experiments were carried out in the presence of (R)-and (S)-selective as GSH neutralizing agent. 260 To parallelize Sav mutant expression, a 24-well deep well plate format was introduced. In six milliliter medium, sufficient Sav (~0.5 mg) can be produced to run 1-2 catalytic reactions. Accordingly, twenty eight amino acids within a 15 Å distance to the metal were selected for mutagenesis. Instead of introducing all 20 amino acids at each position, the effort was reduced to 12 amino acids: A, V, L, D, E, Q, K, H, M, Y, S, P as representatives of the different types of amino acids (acidic, basic, polar, hydrophobic). The autoinduction medium Zyp-5052 was used to express all Sav isoforms, Scheme 32. After 24 h expression, cell free extracts of the mutants were prepared and biotin-4-fluorescein titration was applied to determine the number of free biotin-binding sites per well. To the cell-free extracts, 99 was added to neutralize GSH. After incubation, the catalyst, the substrate and the reductant were added. The reaction was followed by UPLC. The procedure allows the screening of 335 Sav mutants within 20 days (including mutagenesis, expression and activity screening). Scheme 33. Enzyme cascades combining an ATHase and an amine oxidase afford enantiopure amines. This can be applied for the synthesis of a/b) isoquinolines and pyrrolidines, c) nicotine, and d) pipecolic acid. Addition of either horseradish peroxidase or catalase prevents the oxidation of the iridium cofactor by hydrogen peroxide resulting from MAO activity. Adapted with permission from ref. 261 Copyright MacMillan 2013.
ATHases can be used to recycle NADH using formate as reductant. The NADH-recycling ATHase can be applied in a cascade with HbpA monooxygenase to produce dihydroxy biphenyl. Adapted with permission form ref. 261 Copyright MacMillan 2013.
The NADH regeneration employing ArMs and sodium formate as hydride source (Scheme 35) was also investigated by Salmain and coworkers. The Papain scaffold was used to host the precious metal d 6 -pianostool complexes using a covalent anchoring strategy via a maleimide moiety. [241][242] They evaluated the catalytic transfer hydrogenation performance for the reduction of NAD + 111 into NADH 110. The bispyridine-rhodium complex 77 proved most active upon anchoring in PAP. 262 The authors showed that The four-enzyme cascade yielded amine (R)-102 with >99 % ee and >99 TON, Scheme 37.
A biotinylated [( 98)Ir( 117)Cl] + cofactor anchored within Sav can rely on NADPH as reducing agent for the asymmetric transfer hydrogenation of imines. 265 This ATHase was integrated in a four-enzyme cascade to produce enantioenriched 102, relying on glucose as hydride source (blue frame) and MAO-N to upgrade the enantioselectivity of the ATHase (red frame).
Transfer Hydrogenation of Imines, Allosteric regulation. In living organisms, enzymatic activity is typically (cross)-regulated via weak molecular interactions. Molecular or physical (e.g. optical, magnetic, temperature etc.) tools with the ability to control artificial metalloenzyme's activity would be highly desirable to regulate enzyme cascade in vivo.
Ward and coworkers developed a latent artificial transfer hydrogenase that is activated by a protease, Scheme 38. 266 As cofactor, a biotinylated Ir-Cp* complex [( 98 269 The crystal structure of many of the hCAII-ligand complexes are deposited in the protein data bank (PDB) and provide insight for rational ArM design. Moreover, hCAII can be expressed in large amounts in E. coli. In an initial effort to build an ArM from hCAII, Ward and coworkers tethered pianostool complex [h 6 -(C6Me6)Ru(bispyridine)Cl] + to a benzene sulfonamide anchor. 270 The Ru-cofactor did not afford any transfer hydrogenation activity but bound to hCAII with nanomolar affinity. A second arylsulfonamide cofactor was based on iridium complexes [h 5 -(Cp*)Ir(pico)Cl] (119-120). 271 In the asymmetric transfer hydrogenation of cyclic imine 88 inside WT hCAII, this ArM afforded 9 TON and up to 70 ee for (S)-salsolidine 87, Scheme 39. A crystal structure of complex [h 5 -(Cp*)Ir(pico)Cl] 119×WT hCAII was solved which served as a model for a computational enzyme redesign. In a collaboration with the Baker group, Ward and coworkers relied on the Rosetta design algorithm to identify mutants that may stabilize the Ir-cofactor within hCAII. 272 The hypothesis was that cofactor stabilization and localization might increase the relative concentration of the active catalytic species vs.
unproductive conformations of the ATHase. Rosetta predicted four hCAII constructs containing up to 8 mutations (d1 -d4) that were predicted to display increased affinity.
Binding assays of the mutated hCAII revealed that indeed the designs had on average a ~50-fold increased affinity vs. WT hCAII (Kd = ~0.33-0.46 nM vs. 21 nM). Gratifyingly, the imine reduction activity was also positively affected: up to 98
Hydroformylation of olefins is widely used to synthesize aldehydes in both commodity and fine chemicals due to its perfect atom economy 276 and the availability of olefins and syngas.
Both linear and branched aldehydes can be accessed, in the latter case leading to a stereogenic center α to the aldehyde, presenting the opportunity for catalyst control over both regio-and enantiooselectivity. 277 As noted in section 3, early studies by Marchetti established that ArMs generated via metallation of HSA with Rh(acac)(CO)2 catalyze styrene hydroformylation with a branched-to-linear ratio of ~90:10. 156 Building on extensive work showing that the active site Zn of carbonic anhydrase could be substituted with different metals, 278 Kazlauskas replaced the active site Zn of hCAII with [Rh(acac)(CO)2] to generate a hydroformylase. 279 The branched aldehyde product of styrene hydroformylation was favored by the free cofactor and ArMs generated using wt hCAII, Scheme 41. Analysis of this ArM using ICP-MS indicated that it possessed a Rh/scaffold ratio of 5.4, which was believed to result from non-specific Rh-binding sites on the scaffold surface. Targeted mutagenesis and chemical modification of surface histidine residues was therefore carried out, and ArMs generated from the resulting hCAII variants had Rh : scaffold ratios of 1.2 and 1.4. The hydroformylation of styrene catalyzed by the modified ArMs reversed its regioselectivity to favor the linear product (linear/branched ratio up to 8.4, Scheme 41).
Scheme 41. Hydroformylation of styrene catalyzed by a Rh-substituted hCAII.
More recently, Hartwig reported detailed characterization of rhodium-and iridiumsubstituted hCAII ArMs. 280 Metallation of apo hCAII was monitored using a colorimetric assay, which indicated that metallation was significantly impacted by the ligands on the metal precursors used but suggested that surface histidine residues had only a minor impact. [Rh(nbd)2]BF4 and [Rh(acac)(CO)2] (nbd = norbornadiene and acac = acetylacetonate) afforded 71% and > 90% metallation, and the Rh/scaffold ratio was approximately 1:1 even with surface histidine present. A elegant series of HSQC NMR experiments, which had been previously used to study the catalytic mechanism of hCAII, 281 indicated that Rh was only coordinated by two of the three His residues in the active site. Multiple ligand substitution reactions of hCAII-[Rh(nbd)2] were also examined.
Natural enzymes performing reductions of small molecules are located in the super-group of the oxidoreductases (Enzyme Commission number: EC 1) and act on nitrite, nitric oxide, nitrous oxide, elemental nitrogen, sulfite, elemental sulfur as well as dioxygen, chlorate and perchlorate. Enzymes involved in these transformations are amongst others flavodiiron proteins 282 , heme proteins [283][284] and enzymes depending on copper 285 , molybdenum [286][287][288] and vanadium. 289 Beside the large variety of natural small molecule reductases, researchers have engineered artificial metalloenzymes for the reduction of nitrite, nitric oxide, dioxygen and carbon dioxide.
Inspired by the structural homology of heme copper oxidases (HCOs) and nitric oxide reductases (NORs), Lu designed and engineered a copper binding site in sperm whale myoglobin (Mb) by introducing two histidine residues (L29H, F43H) into the distal heme pocket. Metal binding in the resulting scaffold (CuBMb) was supported by UV-Vis and EPR, and increased O2 affinity was observed when Ag(I) (as a Cu(I) mimic) bound. 290 NO reduction(Equation 1) was catalyzed by CuBMb-Cu(I), with a turnover number close to a native enzyme (~ 2 mol NO • mol CuBMb -1 • min -1 ). 291 Lu later designed a heme/non-heme FeB binding site in Mb by incorporating L29H, F43H, and an additional glutamate residue (V68E) into this scaffold. 292 The binding of Fe(II) in the new artificial enzyme (FeBMb) was supported by the crystal structure of Fe(II)-FeBMb, in which the non-heme iron was coordinated with three histidines, one O atom of glutamate and one water molecule. Thirty percent conversion was observed for NO reduction catalyzed by Fe(II)-FeBMb. The catalytic activity was further enhanced (~100% increase) by introducing a second glutamate to the second coordination sphere of FeB binding site (I107E), which was believed to facilitate proton delivery via a hydrogen binding network. The same group reported an alternate binding pocket (FeBMb(-His)) for Cu, Fe and Zn ions, consisted of two histidines and one glutamate (L29E, F43H, H64). Both FeBMb(-His)-Cu and FeBMb(-His)-Fe catalyzed the reduction of NO, while the former ArM had better activities (32% vs 6% conversion after 20 h). 293 In 2006, Watanabe and coworkers studied the electron transfer between an artificial metalloenzyme and a natural enzyme. For this purpose, they incorporated a Fe-salophen (122 or 123) into heme oxygenase (HO). This enzyme converts heme to biliverdin using electrons provided by NADPH/cytochrome P450 reductase (CPR, Scheme 42). 294 The crystal structure of 123•HO highlighted the presence of a hydrogen bond between the propionic acid carboxyl group and R177 of HO. Accordingly, the electron transfer rate from CPRred to 123•HO is 3.5-fold faster than that of 122•HO, although the redox potential of 123•HO is lower than that of 122•HO (Table 12). 13). previously designed α-helical coiled-coil motif (section 7.1). 298 The construct (TRIL23H)3
is able to bind Cu(I) and Cu(II) with picomolar and nano/micromolar affinities respectively.
Binding occurs via coordination to three histidine residues (mutation L23H). The nature of the active site, a structural analogue of copper nitrite reductase, was characterized by spectroscopy. The designed artificial metalloenzyme Cu(I/II)(TRIL23H)3 +/2+ catalyzes the reduction of nitrite to nitric oxide with ascorbate as electron donor, Equation 2. Kinetic investigations revealed similar rates for the oxidation of ascorbate by Cu(II)(TRIL23H)3
and for the oxidation of Cu(I)(TRIL23H)3 + by nitrite (kascorbate = 4.6 • 10 -4 s -1 and kCu = 5.2 • 10 -4 s -1 at pH 5.8). This suggests that the electron transfer from Cu(I)(TRIL23H)3 + to nitrite with simultaneous formation of Cu(II)(TRIL23H)3 2+ is the rate-limiting step of the catalytic cycle. Under optimized conditions, the artificial nitrite reductase exceeded 5 turnovers and formed predominantly NO as product (i.e. almost no formation of N2O).
Allylic alkylation is a C-C bond forming reaction catalyzed by a range of transition metal complexes, though Pd-phosphine complexes are most commonly used for synthetic applications. 299 These catalysts activate heteroatom-substituted allyl compounds to attack by a soft nucleophiles via a metal-allyl intermediates. Though no natural enzyme has been reported to catalyze this reaction, terpene synthases rely on allyl cation intermediates that react in a similar fashion with an intramolecular carbon nucleophile to create a C-C bond. 300 Researchers have demonstrated that Ru-catalyzed allylic substitution proceeds even inside of cells, making this reaction particularly attractive for in vivo applications. 301- 303 Encapsulating the transition metal within a protein scaffold could help to protect the metal in a biological environment, induce substrate specificity, and enable selective cell or even organelle targeting.
Toward this end, Ward and coworkers generated a panel of ArMs by incorporating biotinylated Pd-bisphosphine complexes into different Sav variants and investigated their activity toward asymmetric allylic alkylation of 1,3-diphenylallylacetate using malonate as a nucleophile. 304 In total, fourteen Pd complexes were screened against 20 Sav isoforms bearing mutations at S112X and two additional positions (Scheme 44 and Table 14).
Didodecyldimethlyammonium bromide (DMB) was added to suppress 1,3diphenylallylacetate hydrolysis (Scheme 44). With this modification, ArMs containing ligands 37 (which yielded rac-product), 41 (mostly (R)-selective), and 42 (mostly (S)selective) were identified. The sterically constrained spacers in these ligands were believed to project bound Pd centers into a protein environment more suitable for alkylation rather than hydrolysis in bulk solvent. Replacement of DMB by DMSO was possible in case of ligand 41 and led to increased enantioselectivity and conversion (i.e.
95 % ee, TON 45, Table 14, entry 3). The (S)-selectivity was increased to 82 % ee upon combining mutants S112G and V47G mutations.
Scheme 44. A biotinylated bisphosphine palladium complex anchored within streptavidin affords an artificial allylic alkylase. Adapted with permission from ref. 304 Copyright Wiley VCH 2008.
Carbon-carbon cross-coupling reactions offer a highly versatile means to assemble functional molecules. The palladium-catalyzed Suzuki-Miyaura cross-coupling reaction of aromatic halides and aryl boronic acids to afford biaryls is particularly notable in this regard. 305 The importance of biaryls to the agrochemical, pharmaceutical, and polymer industries, coupled with the bioorthogonality of the Suzuki-Miyaura cross-coupling in general have led to extensive development of catalysts for this reaction. Researchers have also started to apply this and closely-related reactions for chemical biology applications, [306][307][308][309] although these efforts are limited by the high catalyst loadings required, presumably due to catalyst inactivation under biologically relvant conditions 306 . Anchoring Pd-catalysts within a protein scaffold could potentially i) protect the Pd-catalyst from inactivation in a biological reaction medium and ii) control the stereochemistry of the crosscoupling reaction.
To date, three studies have been reported on ArM Suzukiases. The first of these, reported by Ueno, Watanabe, and coworkers, was generated by loading [Pd(allyl)Cl]2 127 into apoferritin (Scheme 45). 310 Ferritin is an iron storage protein consisting of 24 self-assembled four helix-bundle subunits. The crystal structure of 127•apo-ferrtin reveals two dinuclear Pd complexes located at the so-called accumulation center and threefold channel, respectively (Figure 13). Localization of the Pd centers is ensured via coordination to cysteine, histidine, and glutamic acid residues. The resulting ArMs catalyzed the Suzuki-Miyaura cross-coupling between 4-iodoaniline 128 and phenylboronic acid 129 with up to TOF = 3500/hour (Scheme 45 and Table 15, entry 1).
Figure 13. Ferritin (PDB code 1DAT) was engineered to carry out Suzuki-Miyaura crosscoupling 310 and phenylacetylene polymerization reactions. 311 Crystal structure analysis of Pd-and Rh-bound ferritin indicated metal binding to amino acid side chains in the "threefold channel" (magenta) and the "accumulation center" (red). Only a fraction of ferritin is displayed for clarity.
Several mutations were introduced around the Pd binding site in 127 •apo-ferrtin to decipher the role of the coordination geometry on the catalytic activity (Figure 13). 312 Among the mutants, the ferritin E45C-R52H variant gave the highest TOF = 4300/hour (Table 15, entry 7).
The Heck reaction is an extremely powerful means of forging C-C bonds between widely available aryl hailides and olefins. Moreover, substituted olefins can give rise to regio-and enatiomeric products, which necessitates the development of catalysts that can impart selectivity to these transformations. Filice and Palomo therefore took on the challenge of developing an ArM Heckase. 314 Covalent modification of a variety of lipases, including CAL-B, was achieved using phosphonate-substituted Pd-pincer complex 136 (Figure 14 a) in analogy to the previous work of Gebbink and van Koten (Scheme 12). 180,182 Optimization of reaction conditions and immmobilization methods led to the finding that Even more remarkably, this catalyst provided high enantioselectivity (96.6 % ee) in the Heck coupling of 2,3-dihydrofuran 137 with phenyl iodide at 120 °C in 75% DMF, though no conversion was observed at 70 °C or with phenyl bromide (Figure 14 c). Beyond highlighting the potential for ArMs to catalyze selective Heck reactions, this study provides compelling support for further exploration of immobilized ArMs.
Carbon-carbon bond formation via carbon-hydrogen (C-H) bond functionalization remains a highly attractive yet equally challenging transformation in organic synthesis. 315 By eliminating the need for prefunctionalized starting materials, these reactions can enable new disconnection strategies, reduce synthetic manipulations, and decrease waste. All of these benefits, however, require a catalyst to activate a single C-H bond in the presence of many other, often similarly reactive, C-H bonds, and functionalize the resulting metallated position with the desired regio-and enatioselectivity. The impressive levels of selectivity exhibited by a number of natural enzymes that catalyze C-H bond functionalization clearly demonstrates the potential for protein scaffolds to control these transformations. 316 A number of researchers have therefore explored the possibility of using ArMs to control the selectivity of synthetic C-H functionalization catalysts.
Rovis and Ward, for example, hypothesized that incorporating a biotinylated Rh-complex [(98)RhCl2(H2O)] into Sav could provide a chiral environment for enantioselective hydroarylation reactions [317][318] cofactor ratio was varied and the enantioselectivity was determined using substrate 138.
Strikingly, the artificial benzannulase [(98)RhCl2(H2O)]×Sav S112Y-K121E retained a high ee even in the presence of a large excess of cofactor; a rate acceleration of ~100-fold was determined.
Scheme 47. An artificial benzannulase for the synthesis of enantioenriched dihydroisoquinolones based on the biotin-streptavidin technology. Reaction conditions and substrate scope and postulated transition state of the CMD step are displayed in a) and b), respectively. Adapted with permission from ref. 38 Copyright ACS Publications 2016.
Metal porphyrin complexes catalyze a range of C-H insertion reactions via generation of highly reactive metal-oxo, -peroxy, -carbene, and -nitrene intermediates upon reaction with suitable precursors for these species. 320 While oxo insertion reactions catalyzed by heme-dependent oxygenases and peroxygenases have been extensively studied, 321 reactions of these enzymes with carbene and nitrene precursors has only recently been explored in detail. 322 As discussed in the review by Fasan and coworkers elsewhere in this issue, this has been exploited to enable carbene insertion into olefins (cyclopropanation) 12 and nitrene insertion into C-H bonds using cytochromes P450, myoglobin, and other heme enzymes. [323][324] To obtain systems with further expanded function, a number of researchers have incorporated unnatural porphyrins and other heme-like cofactors into heme proteins and enzymes to generate ArMs. 55 These efforts take advantage of well-formed active sites that evolved to accommodate a planar (heme)
cofactors and small molecule substrates. While this approach has also been extensively explored for oxo insertion reactions (section 6.1), significantly less has been done for carbene and nitrene insertion reactions.
Despite the ability of heme enzymes to catalyze carbene insertions into olefins, analogous insertions into C-H bonds using these enzymes have not been reported. Hartwig and Clark therefore investigated the activity of ArMs generated by metallating eight different apomyoglobin variants containing a different mutation at the axial ligand position (H93X) with protoporphyrin IX (PPIX) cofactors containing different noble metals. Of these, an ArM generated from (PPIX)Ir(Me) catalyzed both cyclopropanation (vide infra) and intramolecular carbene insertion using diazo carbene precursors. 325 In the later case, reaction of diazoester 141 to form the corresponding chiral dihydrobenzofuran, which has not been reported for natural heme enzymes, was examined. Stepwise targeted mutagenesis of the myoglobin active site led to ArMs with good enantioselectivity reactions of multiple substrates (86% ee for C-H insertion, Figure_15). Variants were identified that catalyzed intramolecular carbene insertion in one particular substrate to provide either product enantiomer with up to 84% and 50% ee (Scheme 48 a)).
Figure_15. Cartoon representation of Myoglobin's active site with its natural cofactor (PDB: 1MBI). Highlighted in magenta and yellow are the axial (H93) and distal (H64) histidines, respectively. Highlighted in orange are additional amino acids modified to evolve the ArMs as C-H activase, cyclopropanase 325 and dihydrogenase. 207 Highlighted in beige are amino acids mutated to modify the affinity or the orientation of the artificial cofactors for the evolution of sulfoxidase 326 327-328 and peroxidase. [329][330][331] Heme displayed in green (C), red (O) and blue (N).
The same group also reported that a thermostable cytochrome P450 from Sulfolobus solfataricus (CYP119) could be reconstituted with (PPIX)Ir(Me) to generate a thermostable ArM (Tm = 69 °C). 332 Intramolecular carbene insertion using diazoester 141
was again selected as a model reaction (Scheme 48 b)). A catalytic efficiency (kcat/KM) of 0.071 min -1 •mM -1 was observed for the single mutant C317G, which was improved by introducing hydrophobic and uncharged residues into the ArM active site (Figure 16). The quadruple mutant (L69V-T213G-V254L-C317G) showed a more than 4000-fold higher catalytic efficiency (kcat/KM = 269 min -1 • mM -1 ) than the initial wild type construct, accompanied by an initial turnover frequency of 43 min -1 and a high enantioselectivity (94% ee). Diazoester 141 was transformed into the corresponding dihydrobenzofuran on a preparative scale (1.0 g substrate, 55% isolated yield, 93% ee), whereby the ArM exceeded 3200 turnovers. On a smaller scale (10 mg substrate) a TON > 35,000 was achieved. Supported on sepharose beads, the ArMs could be recycled at least 4 times without erosion of enantioselectivity and retaining 64% of its initial activity.
Transition metal catalyzed olefin metathesis has become a highly versatile tool for C-C bond formation. 333 The development of Z-and E-selective cross metathesis catalysts has enabled construction of a wide range of molecule containing a double bond. Moreover, the bioorthogonality of many olefin metathesis reactions has led to their use in chemical biology, [334][335][336] particularly for selective functionalization of biopolymers. [337][338] More recently, olefin metathesis has been considered as a means to expand the chemical toolkit of biochemistry given that it is thus far unknown to living organisms. A number of ArM metathases have been developed toward this end.
The first ArM metathases were introduced in 2011 by Ward et al. and Hilvert et al. [339][340] The former employed (strept)avidin as a host for biotinylated cofactors based on a Grubbs-Hoveyda 2 nd generation catalyst (GH-type catalyst). In fact, all the artificial metathase published to date use this type of cofactor. This is most probably due to its high stability toward air and water. Different reaction conditions were explored, and the best yields were obtained at low pH with high concentrations of MgCl2 (Table 16). Similarly, Matsuo et al. investigated the effect of added chloride ions in aqueous metathesis and found that catalyst stability is significantly enhanced by the addition of KCl. 341 They proposed that the replacement of the chloride ligands by hydroxy ions may lead to decomposition of the catalyst. This could explain why the addition of chloride salts and lowering of the pH has a beneficial effect. In a subsequent article, Ward and coworkers tested a large number of biotinylated catalysts where the biotin anchor was placed not at the backbone but on one of the mesityl units of the NHC ligand. The construct of Hilvert et al., on the other hand, was generated via covalent modification of a small heat-shock protein from Methanocaldococcus jannaschii (MjHSP). 340 This protein assembles into a spherical substructure with large pores, which allow for the diffusion of small molecules. Similar to streptavidin, the protein displays a remarkable stability towards high temperatures and low pH. GH-type catalyst 150 substituted with a a-bromoacetyl group was reacted with the thiol of a surface cysteine in the G41C mutant of MjHSP to yield the artificial metathase 151 (Scheme 50). Scheme 50. Covalent anchoring strategy to produce a metathase by Hilvert et al. 340 The same trend observed for the metathase of Ward et al. was also apparent for this system: the lower the pH, the better the performance of the metathase. A total turnover number of 25 could be achieved for ArM 151 at pH 2, but the free catalyst 150 gave slightly better results (Table 17, compare entries 1-3 with entries 4-6). An covalent anchoring strategy was also used by Matsuo et al. to generate an ArM metathase. 343 GH cofactor 152, which contains an L-phenylalanine moiety bearing an achloroketone, was used to both covalently anchor the cofactor to a-chymotrypsin and selectively interact with the S1 site of a-chymotrypsin via hydrophobic contacts.
Nucleophilic attack of the imidazole of H57 on the chloroketone resulted in a covalent link between the cofactor and a-chymotrypsin to provide 153. Interestingly, the authors showed that the enantiomer of 152 with inverted stereochemistry did not react with the protein. This provided convincing evidence that the functionalization indeed occured inside the active site of a-chymotrypsin which is known to only recognizes L-phenylalanine.
The authors tested three different substrates for metathesis and reported good turnover numbers when using the glycosylated diene 154. In contrast, they observed only low TONs for the cationic substrate 155 and the lipophylic diallyl tosylate 147 (Table 18, entries 1-6 and Scheme 51). Acetate pH 5 20 a conditions: Catalyst (5 mol%), MgCl2(10 eq.), Buffer, CH2Cl2 5 %(v/v), 25 °C, 20h.
An artificial metathase based on on human carbonic anhydrase II (hCA II) was reported by Ward et al. 345 The hCA II protein has a strong affinity (typically nM -pM) toward arylsulfonamides due to sulfonamide coordination to an active site zinc ion. Three different sulfonamide-substituted catalysts were prepared (159-161), and their activity toward ring closing metathesis of substrate 147 was evaluated. Catalyst 160, bearing orthoisopropoxy groups at the NHC aryl groups, performed best. The ArM had similar activity to the free cofactor, but moderate turnover was observed even at pH 7.0, which may be relevant for in vivo applications (Table 20 and Scheme 52). Scheme 52. Human carbonic anhydrase as host for an artificial metathase. Okuda and Schwanenberg reported the first artificial metathase for ring opening metathesis polymerization (ROMP). In their original publication, the authors described a strategy to engineer the b-barrel protein FhuA via modification of an internal cysteine mutant (K545C) of the protein (Figure 17). 346 Two additional positions (N548V and E501F) were also mutated to ensure cysteine accessibility and improve catalyst performance.
Both maleimide-and a-bromoacetyl-functionalized GH-type catalysts reacted with the internal cysteine residue in the FhuA triple mutant to generate ArM metathases. Of the metathases evaluated, 166•FhuA, generated from maleimide functionalized GH-catalyst 166, provided the highest activity for polymerization of 7-oxanorbornene 162, but higher activity was observed for the free catalyst. In a subsequent publication, the authors performed chemical optimization of the linker length. 347 Gratifyingly, shortening the linker increased the efficiency of the metathase, and modest selectivity towards cis-double bonds was observed in the polymer product (Table 21, entries 1-8 and Scheme 53). A breakthrough in the field of artificial metalloenzymes and metathases in particular was achieved in the groups of Ward and Panke. 349 Relying on the biotin-streptavidin couple, researchers in these groups selectively assembled an ArM metathase within the periplasm of Escherichia coli. The secretion of streptavidin into the periplasm was achieved by fusing Sav with the signal peptide OmpA. The E. coli strain combined a biotinylated GH-type catalyst 171 and periplasmic streptavidin as the protein host. This approach offers several important benefits. First, the metathase is not exposed to metabolites that could potentially interfere with the reaction or poison the catalyst. Glutathione, for example, is known to inhibit many precious metal catalysts, 259 but in the periplasm, this metabolite is only present in low concentrations and primarily in its oxidized form. The latter was shown to In conclusion, it has been shown that metathesis is a versatile tool for the construction of artificial metalloenzymes, performing non-natural reactions. Within the past five years, the efficiency of metathases rose from a TON of 20 to hundreds of turnovers. Furthermore, all of the common metathesis reactions including cross-metathesis, ring closing metathesis, as well as ring opening metathesis polymerization have been achieved by artificial metathases. It has also been shown that such metathases under certain conditions can also work inside cells, which raises fascinating perspectives for their potential use in vivo. Yet, some major hurdles have to be overcome for such hybrid catalysts to be applicable as real alternatives to their organometallic counterparts. Since these systems rely on GH-type cofactors, they remain fragile toward reactive cell metabolites, especially thiols including GSH. This problem could be solved by either chemical engineering of more stable metathesis catalysts or by biological engineering of highly shielded active sites. The second challenge is the modest reactivity, compared to established commercial catalysts. However, taking into account the powerful tools offered by enzyme engineering, it is only a matter of time until highly efficient metathases can be developed for in vivo applications.
As noted in section 5.4, several heme enzymes have been reported to catalyze olefin cyclopropanation via carbene insertion into olefin substrates. Metal substituted heme proteins also catalyze this reaction. Indeed, (PPIX)Ir(Me)-substituted myoglobin (see active site in Figure_15) catalyzed cyclopropanation reactions of an internal alkene 174 and an unactivated aliphatic olefin 175, neither of which has been reported for natural heme enzymes (Scheme 56 a), b). To facilitate in vivo generation of ArMs containing heme cofactors, Brustad and Snow recently reported an orthogonal P450-Fe-DPIX pair and demonstrated that the resulting ArMs could catalyze carbene insertion into olefins to generate cyclopropanes (Scheme 56 c), Figure 18). G265F-T269V-L272W-L322I-F405M-T268A) to preferentially bind a synthetic heme derivative, DPIX-Fe, which bears two methyl groups instead of the bulkier vinyl groups. 350 The ArM is a cyclopropanation catalyst.
Prior to any reported studies on carbene insertion chemistry using metal-substituted heme enzymes, Lewis reported that ArMs containing a dirhodium cofactor could catalyze cyclopropanation of 4-methoxystyrene and insertion into the Si-H bond of diphenylmethylsilane using diazoacetate carbene precursors (Scheme 57). The dirhodium ArMs were constructed by genetically encoding a p-azidophenylalanine residue at different sites within tHisF and phytase scaffolds, followed by strain-promoted azidealkyne cycloaddition (SPAAC) of bicyclononyne (BCN)-substituted dirhodium cofactor (177). 10 Neither of the carbene insertion reactions investigated proceeded with significant enantioselectivity, likely as a result of the dirhodium center projecting out of the scaffolds investigated. Significant levels of formal carbene insertion into the O-H bond of water were also observed, but these ArMs accepted donor acceptor carbenes, which has not yet been reported using heme protein scaffolds.
To determine if a protein scaffold could be used to correct the chemo-and enantioselectivity problems encountered by initial dirhodium ArM constructs, bioconjugation of 177 to several additional scaffolds was pursued. Gratifyingly, incorporating 177 into a prolyl oligopeptidase (POP) 351 scaffold from Pyrococcus furiosus, which has a large active site capable of completely encapsulating the bulky dirhodium cofactor, led to modest enantioselectivity for styrene cyclopropanation (38% ee). 352 Targeted mutagenesis of residues in the POP active site led to dirhodium ArMs that catalyzed styrene cyclopropanation with up to 92% ee and improved selectivity for cyclopropanation over competing formal insertion into the O-H bond of water (Scheme 57 b)). Key residues introduced into the POP scaffold included a histidine (H328), believed to coordinate to one of the Rh centers in 177, and two phenylalanines (F99, F594) across the active site relative to the coordinating histidine (Figure 19). 353 The hydrolase domain is shown in green, the propeller domain is shown in grey and cofactor 177 linked at Z477 is shown in red. In magenta are highlighted alanine mutations introduced to widen the pore and thereby enable access of the cofactor to the active site. Additional mutations introduced into Pfu POP are shown as coloured spheres. Adapted with permission from ref. 352 Copyright 2015, Rights Managed by Nature Publishing Group.
The stereochemistry, molecular weight (Mw), and polydispersity (Mw/Mn) of polymers are critical factors that can be influenced by the secondary coordination sphere of metal catalysts. Researchers have also explored the potential to control polymer properties using interactions provided by either a protein scaffold or a protein assembly. The welldefined interiors of these systems could provide a hydrophobic environment to favor polymerization, orient substrates, and influence polymer size and polydispersity. ArM polymerases that lead to the formation of C-C bonds via a variety of mechanisms, including ATRP, acetylene polymerization, and ROMP have been developed. The latter metathesis-type polymerization is presented in section 5. 23, entries 1 and 2).
Hayashi and coworkers covalently linked a maleimide-substituted Cp*Rh(cod) complex (185) to the interior of apo-nitrobindin (Nb) to perform the polymerization of phenyl acetylene 182 (Scheme 58). 199 Nb contains an Fe-protoporphyrin IX 222 cofactor in a cavity formed by eight β-strands. In order to covalently anchor the pianostool cofactor, Q96, located at the entrance of the cavity, was mutated to a cysteine. The resulting artificial metalloenzyme was tested for alkyne polymerization. The free cofactor, 185, yielded the thermodynamically preferred cis-poly(phenylacetylene) 183 with trans : cis = In 2011, Bruns and coworkers reported that horseradish peroxidase (HRP) was able to catalyze atom transfer radical polymerization (ATRP) using N-isopropyl acrylamide (NIPAAm, 187) as the monomer and alkylbromide (186) as initiator (Table 23, entry 6). 355 In 2014, Bruns and coworkers constructed an ArM ATRPase comprised of copper complex 189 and the group II chaperonin thermosome (THS) from the archaea Thermoplasma acidophilum as a protein scaffold. 356 THS is a hexadecameric protein complex and can switch between an open and a closed state by consumption of ATP. In its closed state, the protein possesses two cavities of 130 nm 2 each that can host a polymerization event. In its open state, the pores are large enough allow the polymer to exit the capsule. The ATRPase, 189•THS C363A-K316C-C364A catalyzed the polymerization of 187 under ATRP conditions (Scheme 59). This ATRPase yielded Poly-NIPAAm (PNIPAAm, 188) with a lower Mw/Mn and Mn compared to the catalyst bound on the surface of bovine serum albumin, a globular protein (BSA-189). These results indicate that the protein cage not only improves the polymerization process, but also limits the diffusion of monomers into the active site of the polymerase (Table 23, entries 7 and 8).
Diels-Alder reactions are widely used for the construction of six-membered rings due the predictable regio-and stereospecificity of these reactions. 357 Copper(II)-catalyzed Diels-Alder reactions of azachalcones with cyclopentadiene (Scheme 60 a)) have been extensively employed in fundamental studies on the design of a wide range of catalysts, 358 including ArMs. An early example, particularly notable for its high enantioselectivity, was reported by Reetz. ArM Diels-Alderases were constructed via non-covalent binding of a Cu(II) phthalocyanine complex (Figure 21, 190) to various serum albumins. Up to 91% conversion, 91:9 endo/exo ratio, and 98% ee could be obtained using this system (Table 24, entry 1-6), although the rate of the ArM-catalyzed reaction was slower than that catalyzed by cofactor alone. In a separate effort, Reetz also demonstrated that a His2Asp
Cu(II) binding site could be introduced into tHisF, the thermostable synthase subunit of the glutaminase synthase enzyme complex from Thermotoga maritima, to generate an ArM Diels-Alderase. 359 EPR spectroscopy was used to confirm Cu(II) binding in the designed site, and eliminating potential metal-binding cysteine and histidine residues from the scaffold surface improved the enantioselectivity of this system (up to 46% ee, Table 24, entry 7 and Figure 20. Scaffold-specific interactions have also been used to anchor catalysts into protein scaffolds to generate ArMs that catalyze the Diels-Alder reaction displayed in Scheme 60 a). For example, Niemeyer reconstituted apo myoglobin using a heme cofactor substituted with both ssDNA and a Cu(II)-bipy complex (Figure 21, 191). 360 The resulting ArM catalyzed the Diels-Alder reaction with improved enantioselectivity relative to the heme-ssDNA cofactor alone (18% ee for endo product and 10% for exo, Table 24, entry 8). More recently, Mahy and Ricoux demonstrated that a testosterone-substituted Cu(II)phenanthrene complex (Figure 21, 192) could be incorporated into a neocarzinostatin variant engineered to bind testosterone with improved affinity relative to the WT enzyme. 361 The scaffold exhibited tighter binding affinity to the Cu(II)-phenanthrolinetesterone cofactor (Kd = 3 μM) than toward testosterone (Kd = 13 μM), and docking experiments suggested that the added phenanthroline ligand enhanced the complementarity of the cofactor to the protein. The resulting ArM catalyzed the reaction displayed in Scheme 60 a) with an increased endo : exo ratio compared to cofactor alone, but with lower conversion and no enantioselectivity (Table 24, entry 9).
Several approaches to generate ArM Diels-Alderases via covalent modification of scaffold proteins have also been reported. For example, Salmain constructed a Diels-Alderase by covalently linking a chloroacetamide-substituted [(η 6 -arene)ruthenium(II)]-complex to papain via cysteine alkylation (Figure21, 193). 362 The TOF of the Diels-Alder reaction between cyclopentadiene and acrolein (a rare example not involving aza-chalcones, Scheme 60 b)) was increased more than threefold with the ArM relative to the free cofactor.
Kamer 351 and later Okuda and Hayashi 363 have demonstrated that bidentate and tridentate nitrogen ligands (Figure 22, 194-202) can be introduced into protein scaffolds to generate ArM Diels-Alderases for the reaction shown in Scheme 60 a) following metallation with Cu(II). Modest enantioselectivity was observed in the former case (Table 24, entries 10-11) 351 while improved conversion relative to free cofactor was observed in the latter(Table 24, entry 12) 363 . A similar tridentate ligand 196 was used to generate an improved ArM Diels-Alderase via bioconjugation to an engineered variant of the transmembrane protein ferric hydroxamate uptake component A (FhuA). 348 The endo-selectivity was improved to 98% by the 196•FhuA compared to 54% and 66% obtained for the free copper ion and 196 respectively (Table 24, entry 13).
Filice showed that a similar approach could be used to generate immobilized ArM Diels-Alderases. A lipase from G. thermocatenulatus (GTL) containing cysteine mutation in a cleft distal to the active site was immobilized on hydrophobic sepharose. A Cu(II) cofactor derived from ligand 200 was used to alkylate the distal cysteine in GTL to generate an ArM, which catalyzed the reaction shown in Scheme 60 a) with up to 92% ee (Table 24, entries 14-15). Both the support agent and immobilizing orientation had a large impact on the selectivity of the reaction, and the native activity of the lipase was still remained after cofactor incorporation in the distal site. Reactions were carried out in a solution of 1 : 1 tert-butylalcohol/MOPS (2mM, pH 7.5) at 5°C.
Although the Friedel-Crafts reaction was introduced more than 130 years ago, it still represents an attractive tool to alkylate aromatic compounds. 369 Initially, it required ≥ 1.0 equivalent of Lewis acid, and alkyl halides. Today however, alkyl alcohols, ethers, tosylamides, or styrenes can be used instead of alkyl halides, and catalytic asymmetric Friedel-Crafts reactions have been developed. [369][370][371] Artificial metalloenzymes that catalyze the Friedel-Crafts reaction have been developed
by Roelfes et al. A BpyAla residue was genetically encoded into LmrR to eliminate the need for covalent modification of the scaffold. 372 Metallation with Cu(II) provided an ArM that catalyzed the Friedel-Crafts alkylation of a variety of indoles (Scheme 62) with conversions as high as 94% and high enantioselectivity (up to 83% ee, Table 26, entries 2-7). Interestingly, when the ligand was placed outside of the hydrophobic pocket, the opposite product enantiomer was formed (Table 26, entry 3). More recently, Roelfes demonstrated that the hydrophobic cavity of dimeric LmrR was capable of binding an unsubstituted Cu(II) phenanthroline cofactor directly to form ArMs through supramolecular assembly. 373 The binding affinity of the cofactor was measured by tryptophan fluorescence quenching, which showed the dissociation constant (Kd) was much higher when an interfacial tryptophan residue (W96A) was removed (45 μM compared to 2.6 μM).
Fluorescence decay also showed decreased lifetime when the scaffold was saturated with Cu(II), suggesting cofactor binding at the hydrophobic pore. The enantioselectivity of these ArMs toward Friedel-Crafts alkylations was significantly higher (up to 94% ee) (Table 26, entries 8-11) than the covalently linked systems noted above. These studies constitute one of the few cases in which covalent scaffold modification, metal binding to a scaffold residue (albeit an unnatural amino acid), and non-covalent scaffold binding have all be used to generate ArM catalysts, all of which provided high enantioselectivity.
Scheme 62. Friedel-Crafts reaction catalyzed by LmrR-Cu ArMs.
Dioxygen is widely used as a terminal oxidant in biological processes. Although the reaction of organic compounds with dioxygen is generally thermodynamically favorable, kinetic barriers slow the oxidation of most organic compounds. This can be traced back to spin considerations: organic compounds exist mostly as spin-paired species whereas dioxygen has a triplet ground-state, thus rendering the reaction spin-forbidded. 374 To overcome the high energetic barrier, nature typically generates metal-O2 species in the active site of metalloenzyme including: such as cytochrome P450, methane monooxygenase and tyrosinase etc. Inspired by these metalloenzymes relying on dioxygen for oxygenation and oxidation reactions, heme analogues or copper complexes have been exploited as a cofactors to create numerous artificial oxidases. This section summarizes oxidation reactions catalyzed by artificial oxidase using peroxides and molecular oxygen as oxidant. Oxygen-atom insertion reactions are described in the following sections.
Shortly after the initial reports of catalytic antibodies, Schultz and Cochran described a class of metallo-catalytic antibodies which they coined "hemoabzymes". 133,375 Shortly thereafter, Imanaka and Mahy joined these efforts. [376][377] This term "hemoabzyme" stems from their composition: a porphyrin ("hemo-"), inserted into an anti-body ("-ab-") that has been elicited in mice against a protein-conjugated heme hapten. Progress achieved relying on this strategy has been reviewed recently. [62][63] In 1990, Schultz and coworkers used N-methyl-mesoporphyrin IX (19) as the hapten. This molecule was selected as a transition-state analogue for porphyrin metalation. The metallation of the mesoporphyrin IX was accelerated by monoclonal anti-19 to afford the Scheme 66. Supplementation of a synthetic o-carboxyphenyl-porphyrin 214 or 215 with their cognate antibodies affords artificial peroxidases.
Harada and coworkers elicited monoclonal antibodies 03-1 and 12E11G against the anionic porphyrin 213 and the cationic porphyrin 216, respectively (Scheme 67). 134,380 The dissociation constant between the antibody 03-1 and Fe-meso-tetrakis(4carboxyphenyl)porphyrin (TCPP) 213 was 1.5 × 10 -7 M and the antibody 12E11G bound to Fe-meso-tetrakis(4-N-methylpyridyl)porphyrin (TMPyP) 216 with Kd of 2.6 × 10 -7 M. In the presence of hydrogen peroxide, the resulting ArMs catalyzed the oxidation of pyrogallol 212 with higher catalytic performance than the free cofactor (Table 27, entries 10-13).
octapeptide via two thioether linkages and provides a histidine ligated to iron. The IgG1 monoclonal antibody Ab-3A3 binds 217 with a dissociation constant KD ~ 10 -7 M. 217 alone catalyzes peroxide degradation as well as monooxygenation reactions; however, its performance is limited by dimerization and by its oxidative degradation during catalysis.
Upon integration within the antibody, the authors hoped to alleviate these challenges and to improve catalytic efficiency by providing a well-defined second coordination sphere environment for enantioselective monooxygenation. They investigated Scheme 69. Supramolecular assembly of 217 with Ab-3A3 yields an ArM for the nitration of phenol. (The catalyst loading is reported vs. the hydrogen peroxide concentration)
In 2004, Mahy and coworkers reported a "Trojan Horse" strategy to construct ArMs. In the previous hemoabzymes, antibodies were raised against a molecule that is directly involved in the reaction (e.g. the catalyst, a structural analog of a reaction intermediate or transition state). In the "Trojan Horse" strategy, an anchor moiety was used as hapten.
This constitutes a versatile means to elicit an host-anchor couple for assembling ArMs, alleviating the requirement of natural protein-cofactor specific interactions. As proof of principle, they tethered a porphyrin to estradiol (220, Scheme 70) and incorporated this cofactor into Ab-7A3, an antibody raised against estradiol. [384][385][386] The resulting antibody•porphyrin conjugate 220•Ab-7A3 had an affinity KD = 4 x 10 -7 M and displayed peroxidase activity twofold improved compared to the free cofactor 220. 385 Scheme 70. A Trojan Horse anchoring strategy affords an "Hemoabzyme" with peroxidase activity.
Table 27. Selected results for the peroxidase activity of "hemoabzymes". Many hemoproteins, including myoglobin, cytochrome b5, horseradish peroxidase, and cytochromes P450 possess a Fe-protoporphyrin IX cofactor (FePPIX). Since FePPIX is with His93. This strong electron σ-donation also increases the peroxidase activity of the ArM. In the presence of hydrogen peroxide, the rate of guaiacol oxidation catalyzed by 225•Mb is eleven-fold higher than that observed with native Mb (222•Mb) (Table 28, entries 8 and 9). (see active site in Figure_15). The same group also reconstituted HRP with Fe-porphycene 225 since the active species responsible for the peroxidase activity has a longer lifetime in HRP compared to Mb (Figure 23). 389 In the presence of hydrogen peroxide, 225•HRP oxidized the guaiacol 221 with a comparable rate to native HRP (222•HRP). (Table 28, entries 10 and 11).
Finally, Fe-corrole 226 was also used to reconstitute both Mb and HRP. Hayashi and coworkers hypothesized that the high-valent oxidation states of this cofactor might be better stabilized due to the trianionic character of the corrole (Figure 23). The resulting Mb-and HRP-based ArMs were used to catalyze guaiacol oxidation in the presence of hydrogen peroxide. The catalytic activity decreased in the following order: De novo designed proteins have also been used as scaffolds to generate ArM oxidases and oxygenases via metallation with both metal ions and metal porphyrin complexes.
DeGrado and Kaplan, for example, designed a phenol oxidase based on DFtet, a fourchain heterotetrameric helix bundle from the family of the dueferri (DF) proteins. 391 The de novo-designed metalloenzyme contains two Fe-ions in its active site and its catalytic activity depends on the presence of dioxygen. Phenol oxidase activity occurs via formation of an oxo-bridged di-Fe(III) species (Scheme 72). Combinations of different monomeric units and introduction of mutations in the active site (i.e. increasing the size of cavity in the active site) led to highly active variants. The engineered metalloenzymes catalyze the oxidation of 4-aminophenol (227) to the corresponding benzoquinone monoimine (228).
Catalytic efficiencies (kcat/KM) of up to 1540 M -1 • min -1 along with a ~1000-fold rate enhancement relative to the background reaction were achieved, and the best variant (G4-DFtet) provided over 100 turnovers.
Scheme 72. Oxidation of 4-aminophenol (227) using an engineered di-iron metalloenzyme G4-DFtet.
Lombardi et al. also engineered an ArM oxidase using the DF1 scaffold (PDB entry 1EC5), a well-characterized metalloprotein of the dueferri protein family. 392 DF1 is a dimeric protein in which each monomer adopts an helix-loop-helix structure. 393 The active site consists of two Fe-ions coordinated by two glutamates and one histidine each (Figure 24).
Introducing mutations beneficial for the catalytic activity of the G4-DFtet analogue led to a destabilization of the protein fold, but this issue was solved by mutating the interhelical turn. An NMR structure of the new DF3 variant was determined (PDB entry 2KIK), and the designed artificial metalloenzyme displayed oxidase activities towards various substrates (Table 29). 30, entries 1 and 2). 397 This artificial metalloenzyme catalyzed the oxidation of catechol under aerobic conditions. Based on molecular dynamics simulations, they redesigned the copper binding site to more closely mimic the dicopper active site of natural catechol oxidases (Figure 25). The resulting triple mutant displayed 87-fold higher kcat/KM values for the 4-tert-buthylcatecol 233 oxidation compared to the wild type β-lactamase (Scheme 74 and Table 31). Highlighted, in yellow, the amino acid position mutated to convert β-lactamase in catechol oxidase (D88G, S185H, P224G). In magenta the five coordinating histidine residues (displayed as spheres).
In 2015, Ward and coworkers constructed artificial copper oxidase by mining the pdb to identify proteins containing latent facial triad motifs, i.e. proteins would contain a metal binding site following a single point mutation. Following mutagenesis and metallation, the resulting ArMs might be expeted to possess radically different catalytic activity compared to the wild-type proteins. The algorithm "Search of Three-dimensional Atom Motifs in Protein Structure" (STAMPS) 398 was used to identify thirteen proteins with a putative facial triad motifs. 399 Among the identified proteins, only 6-phosphogluconolactonase (6-PGLac hereafter), afforded peroxidase activity toward o-dianisidine in the presence of H2O2 upon metallation CuSO4 (Scheme 75). Using glucose oxidase with glucose and O2 in the presence of Cu×6-PGLac led to slightly improved activity. A dramatic increase in activity was observed using t-BuOOH as oxidant. Michaelis-Menten kinetics were determined KM = 11 µM, kcat = 78×10 3 •s -1 , and the kcat obtained was only four orders of magnitude lower than natural horseradish peroxidase (7.7×10 7 M -1 s -1 ). To identify the exact copper localization within 6-PGLac, crystals soaked with excess Cu 2+ were subjected to X-ray analysis. Three copper binding sites were identified, one of which was the predicted triad (His67, His104, Asp131). However, only the two histidines were bound, while the Asp131 was rotated away (O••••Cu = 6.6 Å) Mutation of the two surface histidines to alanines (H9A or H95A) revealed that, despite the copper electron density found in the X-ray analysis, these were not catalytically active.
A single point mutation of 6-phosphogluconolactonase yields a facial triad (H67-H104-N131D). Addition of CuSO4 leads to an artificial peroxidase. 399 a) A colorimetric assay was applied to test peroxidase activity of Cu-PGLac in the presence of t-BuOOH; b) the crystal structure of the artificial peroxidase (PDB code 4tm7) highlights copper coordination to only two histidines H67 and H104; c) mutagenesis highlights the requirement of both histidines but not of the aspartate D121.
In addition to supramolecular anchoring and dative anchoring, a non-natural ligand was directly introduced into the amino acid sequence. In 2008, Schultz and Lee incorporated the unnatural amino acid (2,2'-bipyridine-5-yl)alanine (Bpy-Ala, 237, Figure 26) into the catabolite activator protein (CAP), a dsDNA binding protein of E. coli (Figure 27). 400 The engineered construct (CAP K26Bpy-Ala) showed a high affinity towards a 50-bp doublestranded DNA containing the 22-bp CAP recognition sequence (Kd = 1.90 nM vs. 0.85 nM for the wild type CAP). In the presence of Cu(II) or Fe(II), cAMP and 3-mercaptopropionic acid or ascorbic acid as a reducing agent, site specific DNA cleavage activity was observed. Thus, a DNA binding protein was repurposed into an artificial metalloenzyme.
Additional DNA-cleaving ArMs are summarized in section 6.8. and a reducing agent cleavage of the DNA substrate was achieved.
Enantiopure sulfoxides are important building blocks for organic synthesis and have found applications as chiral auxiliaries and ligands in enantioselective catalysis. [401][402][403] Enantioselective oxidation of sulfides has gained considerable attention as an easy way of generating enantiopure sulfoxides. 402,[404][405] Most commonly used sulfoxidation catalysts rely on titanium, manganese, iron, or vanadium. 402,[406][407] The mechanism of asymmetric induction in these processes resembles that of several natural enzymes as the prochiral sulfide does not coordinate to the metal prior to its oxidation. Instead, a metalactivated oxygen moiety (e.g. a metal-oxo, -peroxo, etc.) is delivered to a suitablypositioned sulfide. Accordingly, sulfoxidation is an attractive reaction to scrutinize the role of second coordination sphere interactions that could be imparted by a host protein containing a suitable sulfoxidation catalyst. In this context, the asymmetric sulfoxidation of thioanisole-type substrates has become a benchmark reaction to evaluate the performance of artificial sulfoxidases (Scheme 76). Most of the ArMs previously described for their peroxidase activity have also been tested as sulfoxidases.
Scheme 76. Thioanisole 27 is used as a benchmark substrate to evaluate the performance of artificial sulfoxidases. For reaction conditions and results, please refer to Table 32.
The first examples of ArM sulfoxidases were reported by the groups of Sheldon in 1998 and Keinan in 1999 (chapter 3 and Table 32, entries 1-2), and a similar antibody-based strategy was later used by Mahy and coworkers. Among the hemoabzymes that were investigated, 217•Ab-3A3 displayed the highest peroxidase activity (Scheme 68). Mahy and coworkers later showed that this ArM catalyzes the sulfoxidation of thioanisole 27 in the presence of H2O2. 408 The ArM 217•Ab-3A3 performed better than 217 alone, with 82 and 38 TON respectively (Table 32, entries 3-4). The ArM was also more enantioselective, with 45 % ee vs. 23 % ee in favor of the (R)-28. Those results were obtained in presence of 5 % t-BuOH, which significantly improved both activity and enantioselectivity of 217, with and without Ab-3A3.
The hemoabzymes with anti-estradiol antibodies were also tested for the sulfoxidation of thioanisole 27 in presence of H2O2. First, 220•Ab-7A3 was twice as active as 220 alone, and afforded for (S)-28 with 8 % ee. (Table 32, entries 5-6). 384 Substitution of the Nmethylpyridinium groups by p-sulfonatophenyl groups yielded an 25% less active artificial peroxidase affording similar enantioselectivity: up to 10 % ee for (S)-28. 409 Upon incorporation of the sulfonated cofactor within neocarzinostatin (NCS), the enantioselectivity slightly increased (i.e. 13 % ee), but was accompanied with a decrease in activity (i.e. 6 TON). 410 Better results were obtained with the ArM resulting from incorporation of an iron porphyrin within xylanase, 213•Xln instead of an antibody. Up to 40 % ee of (S)-28 were achieved ( 12). 413 The oxidation of the thioanisole 27 was investigated in the presence of hydrogen peroxide. The catalytic efficiency was enhanced tenfold in the presence of (gp27-gp5)3, compared to the free cofactor 252 (Table 32, entries 9-10), which was attributed to the hydrophobic environment of the cup structure.
ArM sulfoxidases have also been generated using metal-salen and -salophen cofactors. 32, entries 11-12). A crystal structure of 244 • Mb A71G was obtained, and based on this structure, the salophen ligand was replaced with a salen ligand to create a larger substrate access pathway. 414 Modeling suggested that the location of the metal complex in Mb depends on the nature of the 3, 3'-substituents of the salen moiety. Gratifyingly, the enantioselectivity for the thioanisole 27 oxidation was indeed influenced by the nature of the complexes 245-247. The highest ee (32%) was obtained with 245•Mb (Table 32, entry 13).
Lu later reported on a dual anchoring strategy to fix the orientation of a Mn(salen) cofactor bearing two methane thiosulfonate groups (248, Scheme 76) within an apo myoglobin scaffold containing two cysteine mutations L72C-Y103C. The resulting ArM catalyzed the sulfoxidation of thioanisole with higher enantioselectivity (51% ee) than an analogous system generated using the Y103C myoglobin mutant (12% ee) (Table 32, entries 14-15). 327 In both of these systems, sulfoxidation was accompanied by overoxidation to the corresponding sulfone 242. By selecting alternate anchoring sites within Mb T39C-L72C, Lu also showed that ArMs could enhance the chemoselectivity of sulfoxidation. 328 Increased activity (52% yield), enantioselectivity (60% ee), and chemoselectivity (100% sulfoxide) were obtained for reactions catalyzed by 249•Mb T39C-L72C (Scheme 76 and Table 32, entry 16 and Figure 15). Subsequent studies on this system suggested that polar residues in the ArM active site led to its improved activity, and the need for substrates to enter the active site via a hydrophobic region of the scaffold was believed to favor entry of sulfide over sulfoxide to impart chemoselectivity. In addition, residues in the secondary coordination sphere of the ArM active site were shown to impact the pH dependence of ArM catalysis relative to cofactor alone. 415 In 2009, Ménage and coworkers inserted Mn-salen complexes (250, Scheme 76) into human serum albumin (HSA hereafter) to catalyze the oxidation of sulfide 27 to the corresponding sulfoxide 28. 416 Serum albumins have been used in the past as host to catalyze a variety of reactions. 27,91,93 The choice of HSA as a host for ArM formation was nanomolar range. In terms of catalysis, however, the best results were obtained with bovine serum albumin (BSA). In presence of hydrogen peroxide, Artificial sulfoxidases based on non-porphyrin iron complexes have been investigated by Ménage and coworkers. For example, Fe complexes 268, 254 (Scheme 79) were incorporated into NikA, a periplasmic nickel-binding protein. 422 Ménage and coworkers had previously reported on the incorporation of small iron complexes into the pocket of NikA: they characterized the complex • protein assembly as well as investigated the ironmediated oxygen insertion in an aromatic C-H bond of the ligand. [423][424] The affinity of In contrast, HSA gradually lost its activity and selectivity upon recycling. 416 Despite the better defined localization of the cofactor within HSA, no enantioselectivity was observed for the sulfoxidation. Mutation of residues close to the "drug site 2" may allow to genetically improve the catalytic performance of this artificial sulfoxidase.
Finally, the groups of Mahy and Banse explored a covalent anchoring strategy for the generation of an artificial non-heme ArM sulfoxidase. β-lactoglobulin (β-LG) was used as a scaffold to covalently anchor the iron-cofactor 255 to Cys121, Scheme 76. 426 This ArM catalyzed the sulfoxidation of thioanisole, with a TON = 5.6 and a 20 % ee in favour of (R)-28. The selectivity eroded over the course of the reaction, suggesting a denaturation of the ArM.
Olefin epoxidation is a powerful approach for introducing oxygen atoms into organic compounds due the stereospecificity of this reaction and the potential for subsequent ring opening reactions of epoxides. 431 While a wide range of olefin epoxidation catalysts have been developed, 432 high levels of enantioselectivity remain challenging for a number of olefin classes. 433 Several different ArMs have therefore been explored with the hope of controlling epoxidation selectivity.
For example, Kazlauskas 434 and Soumillion 435 independently reported that the active site Zn of carbonic anhydrase could be replaced with Mn to generate artificial epoxidases.
Specifically, incubating apo hCA II with Mn(II) salts (chloride or acetate) led to the formation of an hCA • Mn ArM that catalyzed the epoxidation of styrenes in the presence of hydrogen peroxide with a moderate conversion and enantioselectivity (up to 67% ee).
The Kazlauskas system also provided modest enantioselectivity (52% ee) on an aliphatic olefin, albeit with low conversion (Scheme 80).
Scheme 80. Epoxidation catalyzed by Mn-substituted hCA II.
Manganese complexes with multidentate nitrogen ligands are robust catalysts for a range of oxygenation reactions. Reetz incorporated a maleimide-substituted Mn-salen cofactor 272 into papain via cysteine alkylation; the resulting ArM catalyzed the olefin epoxidation, but low enantioselectivity was observed (< 10% ee). 164 Lewis and coworkers also generated Mn-terpyridine ArMs by linking a maleimide-substituted terpyridine cofactor (273) to cysteine residues in nitrobindin and tHisF. Both benzylic oxidation and olefin epoxidation were catalyzed by these ArMs (Scheme 81). 200
Scheme 82. Styrene epoxidation catalyzed by ArMs assembled with synthetic Mn(III)porphyrins via a Trojan horse anchoring strategy and supramolecular incorporation. The catalyst loading is reported vs. the oxidant.
Catalytic asymmetric dihydroxylation of olefins with osmium tetroxide and chiral ligands was discovered by Sharpless in the 1980s. As discussed in chapter 3, an ArM dihydroxylase was reported Kokubo. 116 Inspired by this report, Ward and coworkers envisaged that the classic cinchona alkaloid ligands typically used for asymmetric dihydroxylation could be replaced with amino acid side chain functionality provided by Sav.
X-ray analysis of a Sav crystal soaked with excess K2[OsO2(OH)4] revealed metal binding on most surface-exposed basic amino acid residues, including H87 and H127. 437 Incubating WT Sav with 1-2 eq. K2[OsO2(OH)4] yielded a highly enantioselective catalyst for the dihydroxylation of vinyl substrates, Table 33 and Scheme 83. (R)-1,2-dihydroxy-2phenyl propane was obtained from substrate a-methylstyrene 10 with 95 % ee. To identify the position of the catalytic site, residues that had osmium-bound in the crystal structure were mutated to alanine. Surprisingly none of these mutants displayed a significantly eroded ee suggesting that the active osmium is not bound to any of these residues. In stark contrast, mutations within the biotin binding vestibule (e.g. S112, K121) and deep within the biotin binding pocket (e.g. D128) had a significant impact on the enantioselectivity of the artificial dihydroxylase. The ee for 1,2-dihydroxy-3-thiophenyl propane increased from 2 % (R)-to 71 % (S)-product when catalysis was performed with mutant S112Y. Mutant D128A afforded the same product with 71 % (R)-selectivity, Table 33. One of the most stringent requirements for ArMs is the preferred use of water as a reaction medium. 27 A few reports, however, outline efforts to create ArMs that operate in the presence of large amounts of organic solvents. 20,174,436,[438][439] To overcome the inherent tendency of proteins to denature in the presence of organic solvents, Tiller and co-worker have investigated the potential of organosoluble polymer enzyme conjugates (PECs hereafter,294) using poly(2-oxazoline) 291 for the creation of ArMs.
Lu designed and engineered a biosynthetic model of cytochrome c oxidase (CcO) by introducing a heme-copper center into Mb (CubMb). 290 A tyrosine mutant in the active site (CubMb F33Y or CubMb G65Y) enabled its catalytic activity for the reduction of O2 to H2O and different reactive oxygen species (ROS) with high activity (TON > 1000). 444 Multiple approaches were used to fine-tune the ArM and thus increase the corresponding reduction rates. A point mutation (S92A) was introduced into CubMb F33Y; this removed hydrogen bonds to the proximal His ligand and a heme propionate, created a more hydrophobic heme pocket, and increased the heme E o from 95 to 123 mV. Reconstituting the scaffold with different heme derivatives (Figure 29, 222, 298-301) led to the identification of ArMs with even higher E o (up to 320 mV with 300) and better activity (more than threefold TONs after 9 h compared to CubMb F33Y). 445 A series of different tyrosine analogs (Figure 30,
were also introduced into the enzymes in E. coli via stop codon suppression methodology. 302 mimicked the Tyr-His crosslink in CcO, and the resulting ArM catalyzed the reduction of oxygen to water with eightfold more selectivity and threefold higher TON than CubMb F33Y. 446 Tyrosine residues bearing different halogen substituents with decreased pKa also improved the selectivity and TON (up to 1200) of the ArM, suggesting the enhancement of proton donating ability of the phenol ring of the Tyr was beneficial. 447 More recently, 3-methoxytyrosine (306), a tyrosine analog with similar pKa but lower E o , was incorporated into the ArM, the resulting ArM also showed higher oxidase activity (15.0 vs 6.5 μM min -1 ) and selectivity (82% vs 51% to water). 448
Figure29. Various heme derivatives used to fine-tune the redox potential of the ArMs.
Various tyrosine analogs incorporated in the artificial oxidase to optimize the activity.
As noted in section 5.4, selective functionalization of C-H bonds remains a key challenge in catalysis, and ArMs have great potential to improve control over such processes.
Toward this end, a Mn(terpy) cofactor (307) was covalently linked to the nitrobindin (Nb) scaffold, a heme protein with a β-barrel structure, as noted in section 5.5 (Scheme 81) to generate an ArM that catalyzed the oxidation of a broad range of hydrocarbons with good to excellent yields (up to 97%). Unfortunately, however selectivity identical to that of the free cofactor was observed (Scheme 86). 200 Scheme 86. Selected results for Mn(terpy) ArM catalyzed C-H oxidation.
In order to scrutinize the role of both the tetradentate ligand and the metal in Mb, Hayashi and coworkers tested the hydroxylation of ethylbenzene 308 by Mb reconstituted by Feporphycene 225 by Mn-protoporphyrin IX 309 and a Mn-porphycene 310 (Scheme 87). 449 Using hydrogen peroxide as oxidant, only 224•Mb yielded phenylethanol 50 with TON = 13 and an ee = 14% (S))(Table 36, entries 1-4). No overoxidation to the corresponding acetophenone could be detected.
ArMs have also been designed to catalyze site-selective cleavage of proteins and nucleic acids. Indeed, many of the earliest ArMs developed via covalent modification of protein scaffolds involved modification of protein-and DNA-binding scaffold proteins for this purpose. Unlike most of the systems described above, surface modification of the scaffold is sufficient for selective biopolymer cleavage since the scaffold need only situate cofactors proximal to bound DNA. For example, following iminothiolane treatment, the surface lysine residues on the E. coli Trp repressor protein (trp) were used to install an iodoacetamide-substituted phenanthroline cofactor (Figure 22, 200) that was metallated to generate an ArM nuclease. 125 Substrate DNA that contained the aroH transcription unit naturally recognized by the trp scaffold was selectively cleaved in the presence of tryptophan and 3-mercaptopropionic acid. A similar approach was used to link an Fe(III)-EDTA cofactor (311) to the σ 70 subunit of E. coli RNAP complex. The resulting ArM catalyzed cleavage of nucleic acid and protein components of the RNAP complex near the σ 70 binding sites. 450 Haloacetamide-substituted phenanthroline (Figure 22, 200) 451 or EDTA ligands 452 and an Fe(III)-EDTA cofactor (Figure 31 311) 453 have been used to generate ArMs for selective biopolymer cleavage via alkylation of cysteine residues on scaffold proteins. 15 Related systems have been generated via disulfide exchange 454 and transesterification. 455 Finally, in order to bypass the need for the covalent scaffold modification, BpyAla was genetically encoded into catabolite activator protein (CAP) near the the DNA binding site of this scaffold. 400 The BpyAla 237 residue was selectively metallated with Fe(II) and Cu(II) to generate ArM nucleases.
Building on these studies, Nolte recently constructed an ArM containing a Mn-porphyrin cofactor ((Figure 31, 312) that selectively and processively oxidized tri-adenine sites in double stranded DNA (dsDNA). 456 A trimeric, ring-shaped, DNA-binding clamp protein (gp45) was used as the scaffold, and cysteine residues on the clamp were alkylated with a maleimide substituted manganese-phorphyrin complex (312). Supercoiled DNA was quantitatively oxidized by 312 and the reactive position of dsDNA was turned into an aldehyde after oxidation, which was further modified with hydroxylamine-derived biotin.
By the biotin-streptavidin interaction, the reaction site was then visualized in AFM. When the gp45 was closed with an octapeptide, the activity of the ArM was decreased (55% conversion) and the site-selectivity was no longer existed, indicating the binding of gp45 to the DNA was the key to processive catalysis. The reaction direction of the ArM could also be guided by clamper-loader complex (gp44/62) like the native polymerase.
A wide range of natural enzymes catalyze an equally diverse range of hydration reactions.
Around one-third of the enzymes classified by the Nomenclature Committee of the International Union of Biochemistry and Molecular Biology (NC-IUBMB) are involved in the addition-or removal of a water molecule from a specific substrate, thus illustrating their importance in nature. Most of these enzymes are hydrolases (EC 3), but examples can also be found among the lyases (carbon-oxygen bond cleavage, EC 4.2). 457 Some of these enzymes (e.g. carbonic anhydrase, acetylcholinesterase or fumarase) belong to the fastest and most efficient enzymes known today, operating at the limit of substrate/product diffusion (with kcat/KM ≈ 10 8 -10 9 M -1 • s -1 ). 458 In addition to the large variety of natural hydrolytic (metallo)-enzymes, researchers have created ArMs to hydrolyze small model compounds (4-nitrophenyl acetate/phosphate), antibiotics (cefotaxime, ampicillin), or an analogue of the nerve agent cyclosarin. ArMs have also been applied for the hydration of CO2 or α,β-unsaturated 2-acyl pyridines and were engineered to catalyze epoxide ring-opening reactions, Kemp eliminations, transesterifications, and RNA ligations. The applied metalloenzymes were created by placing a catalytically active metal center (often Zn(II)) either in existing proteins or in de novo-designed scaffolds. The catalytic performance and the selectivity of the constructs were subsequently improved by means of genetic engineering. The artificial hydrolases are presented below.
In 2006, Kim and Benkovic et al. engineered an artificial β-lactamase for the hydrolysis of the antibiotic cefotaxime (313). 459 They repurposed the catalytic activity of an existing protein scaffold by applying the SIAFE approach (simultaneous incorporation and adjustment of functional elements) in conjunction with directed evolution. Insertion, deletion, and substitution of active site loops led to β-lactamase activity in the αβ/βα metallo-hydrolase scaffold of glyoxalase II. Subsequent modification of the protein by point mutations resulted in a highly active engineered β-lactamase (evMBL8, PBD entry 2F50).
The designed enzyme completely lost its native activity (hydrolysis of the thioester bond of (S)-D-lactoylglutathione), in favor of the hydrolysis of cefotaxime (313) (Scheme 88).
The engineered enzyme, whose metal content was 1.63 ± 0.43 mol zinc and 0.46 ± 0.14 mol iron per mol of enzyme, displayed a catalytic efficiency (kcat/KM) of 184 M -1 • s -1 .
Scheme 88. An engineered β-lactamase evMBL8 catalyzes the hydrolysis of cefotaxime (313).
Beside antibiotics, small model compounds have also been examined as substrates for designed ArM hydrolases. For example, Pecoraro has engineered a variety of ArMs that catalyze hydrolysis of 4-nitrophenyl acetate (315). 460 The de novo-designed scaffolds consist of a coiled-coil motif with an engineered His3 Zn(II) binding site. Metallation of this scaffold with Zn(II) led to an ArM that hydrolyzed 315 with a catalytic efficiency only ~100fold lower than that of the natural human carbonic anhydrase II (hCAII) (Table 37). More Zastrow and Pecoraro subsequently investigated the influence of the active site location on the catalytic activity of TRI-based ArM hydrolases. 463 Four coiled-coil motifs were designed in which i) the positions of the Zn(II)-site and the Hg(II)-site were inverted (TRIL9CL23H and TRIL9HL23C, Table 38, entries 1-2 and 3-4), ii) the Zn(II)-site was moved closer to the N-terminus (TRIL9CL19H, Table 38, entries 5-6), iii) the Hg(II)-site was omitted (TRIL2WL23H, Table 38, entries 7-8). Analysis of these constructs revealed that the maximal rate, the metal binding affinity and the substrate access depended on the position of the Zn(II)-site, whereas all designs showed similar maximal catalytic efficiencies for the hydrolysis of 315. [Zn(II)(H2O/OH -)](TRIL2WL23H)3 n+ 9.0 0.016 1.8 8.9 0.24 a Reaction conditions: 10-20 μM active Zn(II)-bound peptide complex; values at pH 7.5 determined in 50 mM HEPES; values at pH 9.0 determined in 50 mM CHES; 25°C. The error range of the kinetic parameters is not displayed for clarity. b Determined at optimized pH. c Data reported by Pecoraro et al. 460 Efficient and selective hydrolysis of organophosphates, especially those in nerve agents, has become increasingly important due the serious biological threat that these compounds pose. In 2012, Baker et al. repurposed a mouse adenosine deaminase to catalyze the hydrolysis of such an organophosphate. 464 A coumarinyl analogue 314 of the nerve agent cyclosarin was efficiently hydrolyzed by the engineered zinc dependent enzyme PT3.3 (Scheme 89), and a catalytic efficiency (kcat/KM) of ~10 4 M -1 • s -1 was determined. This ArM was engineered via a combination of computational design and directed evolution.
Initially, the protein data base (PDB) was screened for enzymes containing mononuclear zinc sites. These enzymes are known to catalyze hydrolysis reactions in nature. A calculated model of the transition state for phosphate hydrolysis was subsequently docked into these scaffolds, beneficial hydrogen bonds in the active sites were engineered, and steric clashes were minimized. Out of a set of 12 designed enzymes, one showed a moderate catalytic activity (kcat/KM = 4 M -1 • s -1 ). Saturation mutagenesis of 12 residues surrounding the active site led to the identification of a variant with a ~40-fold improved catalytic efficiency (PT3.1, PDB entry 3T1G). Random mutagenesis using error-prone PCR and a point mutation in the active site (based on analysis of the crystal structure) further increased the activity. Overall, the catalytic efficiency towards the hydrolysis of diethyl 7-hydroxycoumarinyl phosphate (314) was improved by a factor of 10 In 2012, Kuhlman reported an Zn(II)-dependent ArM that catalyzed hydrolysis of 4nitrophenyl acetate/phosphate (315/316). 468 A 5 kDa helical hairpin monomer (Rab4binding domain of rabenosyn) was used as a starting point to engineer a zinc-mediated homodimer (MID1-zinc). The construct bears two Zn(II)-binding sites at the protein interface. Analysis of the crystal structure (PDB entry 3V1C) revealed that only three out of the four designed histidine residues coordinate to zinc, thus suggesting a possible coordination site that may promote esterase activity. The Zn(His)3 motif, located in a 6 • 4 Å protein cavity, indeed displayed hydrolase activity towards 4-nitrophenyl acetate (315, kcat/KM = 630 M -1 • s -1 , max. TON = 50) and 4-nitrophenyl phosphate (316, kcat/KM = 14 M - 1 • s -1 ) (Scheme 90). Mutagenesis of the three zinc-coordinating histidines to alanines led to a complete loss of esterase activity, suggesting that indeed the Zn(His)3 motif is required for catalysis. Scheme 90: Hydrolytic activity of MID1-zinc towards 4-nitrophenyl acetate (315) and 4nitrophenyl phosphate (316).
Similar to the previous report from Kim and Benkovic et al., Tezcan engineered an ArM βlactamase to hydrolyze an antibiotic. The Tezcan system, however, involved Zn(II)mediated self-assembly of four engineered cytochrome cb562 units (Figure 33). 469 To stabilize the resulting D2 symmetric tetramer, a total of 16 surface mutations including the introduction of zinc-coordination sites, hydrophobic interactions, and disulfide bonds were introduced. The designed supramolecular protein assemblies harbor eight Zn(II)-ions (4 structural zinc sites and 4 catalytic zinc sites). Saturation mutagenesis of four individual amino acid positions surrounding the catalytic zinc sites led to the identification of a highly active variant (Zn8: A104/G57 AB34, PDB entry 4U9E). This ArM β-lactamase catalyzed the hydrolysis of ampicillin (317) and 4-nitrophenyl acetate (315) with catalytic efficiencies (kcat/KM) of 35090 M -1 • min -1 and 296 M -1 • s -1 respectively (Scheme 91). In vivo βlactamase activity was demonstrated by expression and secretion of this construct into the periplasm of E. coli. Colonies successfully grew on LB-agar plates containing ampicillin (0.8 mg/l). catalytic activity or knowledge of the reaction mechanism was required to engineer this ligase activity. Structure determination of a highly active variant (ligase 10C, PDB entry 2LZE) revealed substantial differences compared to the initial protein (hRXRα). 472 The helical DNA recognition domain was replaced by a long unstructured loop and the protein adopted a cyclic structure. The artificial ligase still contained two zinc-finger domains, although four of the eight initial cysteine coordination sites were replaced by aspartate, glutamate or histidine. The enzyme maintained its high activity even at temperatures up to 65°C (Figure 34). Scheme 92. In vitro selection of artificial ligase enzymes by mRNA display. A DNA library is transcribed and the corresponding mRNA modified with puromycin (P) followed by in vitro translation. The zinc-finger proteins (linked to their genotype) catalyze the ligation of a 5'-triphosphate-activated RNA (PPP-substrate) with the terminal 3'-hydroxyl group of a second RNA (HO-substrate) bearing a biotin anchor (B). The mRNA is reverse-transcribed and the linked strands (Product) are immobilized on streptavidin (Sav) coated beads. After selection, the constructs are released from the beads by UV-irradiation of a photocleavable linker (PC). DNA is amplified/mutated by PCR and subjected to the next round of selection. Adapted with permission from ref. 470 Copyright Macmillan Publishers Limited 2007.
In 2013, Roelfes et al. applied their previously designed LmrR-Cu(phen) Diels-Alderase (section 5.8) in the hydration of α,β-unsaturated 2-acyl pyridine (Scheme 93). 366 Slightly lower conversion and enantiomeric excess (up to 84% ee) were observed compared to the Diels-Alder reaction. The point mutants of this ArM also suggested that the two residues (F93, D100) at the front entrance had a crucial impact on both the activity and the selectivity. 39, entry 1). In order to generalize their approach of de novo-designed ArMs, they switched from a three-stranded coiled-coil motif to a single-stranded antiparallel three-helix bundle (α3D-motif, originally designed by DeGrado et al. 474 ). Being single-stranded, this 73 amino acid protein allows the introduction of unique mutations (rather than mutations that are repeated with threefold symmetry). The potential to improve the catalytic performance of ArMs derived from this scaffold is therefore expected to be higher. Indeed, Pecoraro et al. engineered CO2 hydration activity into this scaffold by mutagenesis (4 mutations) and elongation of the Cterminus (4 residues). 475 Three leucine residues were replaced by histidines to form the catalytically active Zn(II)(His)3O-site and an existing histidine was replaced by valine to avoid undesired Zn(II)-binding. The engineered protein (α3DH3) bound Zn(II) with nanomolar affinity (Kd = 59 ± 9 nM at pH 9.0) to form a highly active ArM (Table 39, entries 2-5 and Figure 35).
In 2011, DeGrado et al. engineered Kemp-eliminase activity into the C-terminal domain (last 75 residues) of calmodulin (CaM), a Ca 2+ binding EF-hand regulatory protein. 478 Although the artificial enzyme does not contain a catalytically active metal center, it will be highlighted in this review, since its activity is allosterically regulated by the addition of Ca 2+
ions. Upon binding of Ca 2+ , CaM undergoes a structural rearrangement and opens a hydrophobic pocket. De Grado et al. placed a general base (mutation F92E) at the bottom of the hydrophobic pocket (Figure 36), guided by computational design (i.e. docking of the substrate). This mutation afforded an active Kemp-eliminase (cCaM-F92E, PDB entry 2KZ2, Scheme 94). A catalytic efficiency (kcat/KM) of 5.8 ± 0.3 M -1 • s -1 was determined for from intact ArM. Denaturation of the PYP scaffold in the presence of the organic solvent required for substrate solubility was suggested as an explanation for the lack of enantioselectivity in this transformation. These authors later reported a two-step procedure for incorporating phosphines into proteins, though no catalysis was reported. 480
The studies highlighted in this review showcase the wide range of metal complexes, protein scaffolds, and linkage strategies that have been used to generate ArMs to date.
This diversity of components and methodologies has led to systems that catalyze an impressive range of challenging chemical reactions, many of which are not found in nature. Moreover, a range of approaches has been developed to optimize ArMs for chemo-, regio-, site-, and enantioselective catalysis. Early efforts toward this end involved altering cofactor linkage sites within scaffold proteins, 144 altering cofactor structure, 153 and exploring different scaffold/cofactor combinations to enable chemogenetic ArM optimization 176 . Point mutations were later introduced into putative ArM active sites of scaffold proteins to improve ArM selectivity, 172,414 and iterative targeted mutagenesis was subsequently reported 483 . Despite these advances and many more since, 39 ArM catalysis remains rather underdeveloped when compared to homogeneous transition metal catalysis or enzyme catalysis. Because of this, Whitesides' pronouncement regarding the practicality of ArMs 2 remains true to this day.
As noted in the introduction, fundamental advances in protein engineering 484 and aqueous organometallic chemistry 485 contributed to a resurgence in ArM research around the year 2000. Further improvements in ArM efficiency will require innovations that specifically address and exploit the hybrid structures of ArMs. 87 For example, while many studies have demonstrated the potential for protein scaffolds to control the selectivity or organometallic catalysts, the catalysts used are typically well-known from the organometallic literature and remain unmodified upon incorporation into protein scaffolds. The activity and selectivity or cofactors designed to interact synergistically with scaffold proteins via ligand exchange, 319 electron transfer, 294 or other means could be more extensively tuned or even uniquely activated via incorporation into proteins. Similarly, while a range of bioconjugation techniques have been used for cofactor incorporation, only a handful of these are compatible with complex media, including cell lysate (in vitro) or cytosol (in vivo).
Transitioning to efficient bioorthogonal methods 352,486 for covalent 352 , non-covalent 349 , dative 469 , and substitutive 350 cofactor incorporation will ensure that ArM libraries can be readily generated as efficiently as for natural enzymes.
Just as methodology must be adapted to suit different aspects of the unique composition or ArMs, so too must our understanding of the structure and function of these catalysts, which currently lags far behind that of transition metal catalysts and enzymes. ArM characterization typically involves a battery of spectroscopy, mass spectrometry, and, in some cases, X-ray crystallography to ensure the desired composition. Evidence for specific scaffold-cofactor interactions (e.g. dative ligation of metal centers) is often presented, 255,352 and many examples in which scaffold accelerated catalysis have been reported. 44,319 On the other hand, the exact contributions to catalysis of the former and the origins of the latter remain unclear. Substrate binding interactions within ArM active sites remain almost completely unexplored. These are precisely the types of phenomena that natural enzymes have evolved to harness, 487 so understanding how they have emerged in ArMs (albeit to only a primordial extent at this point) could improve our ability to design and evolve comparably efficient systems. Given the importance of protein scaffolds to the activity of natural metal cofactors, 488 it is reasonable to hypothesize that large improvements in the activity and selectivity of synthetic catalysts could be achieved using properly designed protein scaffolds. Toward this end, far deeper biophysical studies into the nature of ArM catalysis, mirroring efforts in enzymology, are required. Similarly, significant improvements in computational approaches to design and simulate the structure and conformational dynamics or proteins containing synthetic metal catalysts will be required to provide insight into the molecular level events that occur during ArM catalysis. 272,481,489 Advances in cofactor design, activation, and bioconjugation, along with improved understanding of ArM catalysis would facilitate development of what is perhaps the most important methodological hurdle limiting ArM catalysis: directed evolution. 490 It is difficult to overstate the impact that directed evolution has had on biocatalysis using natural enzymes, 162 and yet such efforts toward ArMs remain in their infancy. Reports to date have demonstrated that iterative targeted mutagenesis of residues proximal to metal cofactors (or at least their presumed location during catalysis) can improve ArM selectivity and activity. 7 While similar approaches have of course been used to optimize natural enzymes, the true power of directed evolution and the great improvements and surprises in the activity of evolved enzymes have been realized not solely via mutation of active site residues but from functional screens and selections conducted on diverse populations of enzymes containing mutations throughout their structure. 490 Unfortunately, random mutagenesis methods remain unexplored in part due to the small ArM library sizes that can be generated using existing cofactor incorporation methods, 491 while targeted mutagenesis (outside of the putative ArM active site) is hindered by a lack of fundamental understanding of how distal scaffold mutations impact cofactor selectivity.
Toward the latter point, for example, it is not obvious that phylogenetic tools used to generate targeted "smart libraries" of natural enzymes are applicable to ArMs given that the function of the two are completely unrelated. 162 Nonetheless, adapting the full arsenal of library methodologies that are available for enzyme evolution, ranging from random mutagenesis and gene shuffling 484 to combinatorial codon mutagenesis 492 and chimeragenesis 493 , to ArMs will ensure that the latter can be evolved with the efficiency of the former. Genome mining to identify diverse homologues of relevant scaffolds and computations tools to identify new scaffolds based on sequence but (predicted) structural homology could also be highly beneficial for identifying new ArM scaffolds. [494][495] Based on the success of evolved enzymes for practical applications, this would go a long way to establish whether ArMs can in fact have practical utility.
Despite the need for significant advances in our ability to engineer and understand ArMs, hints of their potential utility have already begun to emerge. The many enantioselective transformations highlighted above have certainly proven their worth for fine chemical synthesis. The few studies reported on the activity of immobilized ArMs suggest that significant improvements in the practicality of ArM catalysis could be realized using different immobilization procedures. 257,260,314,332 A key goal of future ArM research must be to develop such transformations that cannot be accomplished using small molecule or enzyme catalysts. In such cases, the added complexity of ArM formation relative to other systems might be offset by savings in other aspects of a process. Tandem processes involving enzyme and ArM catalysis have also been developed, 261 and these illustrate the unique ability of ArMs to shield transition metals from species that would otherwise lead to catalyst death. Taking this possibility to the extreme, ArM catalysis in vivo 349,496 could ultimately be used to augment metabolic pathways with synthetic reactions to expand the scope of biosynthesis 497 . Obviously much work remains before these possibilities can be realized, but recent progress in ArM catalysis and other areas of protein engineering and organometallic chemistry should encourage those willing to take on the challenges that lie ahead.