Many bioinspired transition-metal catalysts have been developed over the recent years. In this review the progress in the design and application of ligand systems based on peptides and DNA and the development of artificial metalloenzymes are reviewed with a particular emphasis on the combination of phosphane ligands with powerful molecular recognition and shape selectivity of biomolecules. The various approaches for the assembly of these catalytic systems will be highlighted, and the possibilities that the use of the building blocks of Nature provide for catalyst optimisation strategies are discussed.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Advancing the understanding of catalytic systems has proven very fruitful for the development of effective homogeneous catalysts. For instance, the growing insight in organometallic chemistry employing phosphorus donor ligands is facilitating the design of systems with optimised electronic and steric ligand parameters to create very effective transition metal catalysts for industrial chemical reactions.1 Integrating the rational ligand design process with combinatorial catalyst optimisation techniques has proven a powerful strategy.2 Nevertheless, despite these developments, efficient transition-metal catalysts are still lacking for various chemical conversions with demanding selectivity and reactivity problems.
Enzymes often outperform synthetic catalysts by employing highly efficient substrate recognition and orientation, stabilisation of reactive intermediates and other second-sphere interactions that are still not fully understood. Studies of synthetic mimics of enzyme active sites have indicated that second-sphere substrate interactions are very important for the catalytic proficiency of enzymes.3
Enzymes are increasingly applied in industrial-scale synthetic processes.4 Such processes are very effective, but natural enzymes are not available for many important chemical conversions performed in industry, creating the demand for the development of synthetic catalysts that rival the performance of natural enzymes. Increasing insight in the mode of action of enzymes and the advances in computational modelling techniques have led to the de novo design of artificial enzymes with unnatural catalytic activities.5 Alternatively, much effort has been made to develop synthetic catalysts that mimic the characteristics of enzymes. For example, shape-selective heterogeneous catalysts have been developed by controlling the pore size in zeolites, leading to increased selectivity for C1 oxidation of n-alkanes.6 Another approach is the use of transition-metal complexes with resins containing hydrogen bonding groups to mimic enzyme structures.7 In homogeneous catalysis sophisticated modular ligand systems have been designed by using bioinspired supramolecular interactions8 or by introducing bulky organic scaffolds like cyclodextrins or calixarenes for substrate encapsulation.9 In addition, ligand systems with specific hydrogen-bonding interactions have been designed to enforce steric interactions or to favourably orientate substrates towards key reaction intermediates.10
A field that aims to combine the best features of natural and “man-made” catalysts is the design of transition-metal catalyst systems based on the building blocks of Nature.11 This approach makes use of ligand-functionalised peptide chains, DNA, RNA and proteins, seeking to create enzyme-like transition-metal catalysts for reactions not found in Nature such as allylic substitution, olefin hydrogenation, hydroformylation, metathesis and cycloadditions. A synergy is envisioned between rational ligand design and the molecular recognition properties of biomolecules, providing catalytic systems, which can overcome selectivity issues for demanding substrates and avoid the need for protecting groups. In addition to the potential of the chirality of these natural building blocks in asymmetric reactions,11c, 12 the potential for second-sphere ligand–substrate interactions may lead to the induction of enzyme-like regio-, chemo-, enantio- and substrate selectivities.13
An additional advantage of this approach is the role separation between the metal centre, which typically dominates the chemical activity, and the biomolecular environment, which is used to induce selectivity in the reaction (Figure 1). This division of functionalities greatly simplifies the application of modular combinatorial synthesis approaches.
This review highlights various bioinspired approaches towards transition-metal catalyst systems and their applications. In line with our own research programme, the emphasis will lie on systems employing phosphorus-containing ligands. Also the new opportunities for catalyst development these systems provide, using traditional synthetic tools combined with combinatorial assembly and/or advanced biomolecular methods will be discussed as well as future prospects of these systems for solving challenging reactivity and selectivity problems encountered in homogeneous catalysis.
Peptide-Based Ligands for Transition-Metal Catalysis
Nature has always been an important source of inspiration for synthetic chemists. This also applies to ligand design in homogeneous catalysis. For example amino acids constitute an easily accessible chiral pool and can be used directly as ligands in transition-metal-catalysed chemical conversions14 or be used as the source for the synthesis of chiral ligand scaffolds.[15 This section will provide an overview of catalyst systems containing amino acid building blocks to create protein-like structures for well-known transition-metal-catalysed reactions.
Hayashi and co-workers already reported in 1982 the modification of a diphosphane moiety with leucine to obtain a chiral diphosphane ligand that was used in the palladium-catalysed asymmetric allylation of 2-acetylcyclohexanone (Scheme 1).16 The observed 52 % enantiomeric excess (ee) was ascribed to the second-sphere interaction of the leucine part of the ligand with the sodium enolate intermediate. This represents one of many examples where the modification of phosphane ligands with amino acid residues is exploited to obtain chiral ligands for asymmetric transition-metal-catalysed reactions.11c Another example of the potential of amino acid functionalised phosphanes is the use as water-soluble ligands.17
Oligopeptide-Modified Phosphane Ligands
For the purpose of connecting phosphane ligands to longer peptide structures, either phosphane-containing peptide building blocks or selective strategies for the modification of peptides with phosphanes are required. Several phosphane-containing peptide building blocks have been reported that have not been incorporated into larger peptidic structures.18 Hoveyda et al. reported the synthesis of a small library of dipeptides containing a diphenylphosphane-modified valine with either glycine or phenylalanine as the second amino acid.19 These ligands were successfully applied in the copper-catalysed conjugate addition of dialkyl zinc reagents to various electrophiles. High enantioselectivities (>90 %) were typically obtained and in general single amino acid variants outperformed dipeptide structures.
Breit and co-workers created a similar class of phosphane ligands based on small peptide chains consisting of D and L valine units with a terminal phosphane or phosphite moiety.20 These systems were applied in copper-catalysed conjugate additions, achieving selectivities similar to those reported by Hoveyda et al. Breit and co-workers also demonstrated that these phosphane ligands can form diphosphane ligands by homodimeric self-assembly of the peptides into helical structures through hydrogen bonding and π-stacking (Scheme 2a).21 These assemblies were applied in the rhodium-catalysed hydrogenation of protected amino acid precursors and dimethylitaconate reaching over 95 % ee for each substrate used. This concept was extended to β-turn mimics using assembled hetero combinations of peptide-modified phosphanes containing glycine and alanine in addition to the valine residues (Scheme 2b).22 These systems were applied in the rhodium-catalysed hydroformylation of styrene, in which high branched selectivity (>85 %) was obtained with nearly all combinations.23 The ee observed by using these catalysts was low (<38 %).
The application of longer oligopeptides was pioneered by Gilbertson et al. who synthesised various phosphane-modified amino acids that were introduced at different positions in synthetic oligopeptides containing β-turn and helical structures.24 For this purpose several sulfur-protected phosphane-containing alanine derivatives compatible with solid-phase peptide synthesis were developed (Scheme 3a).25 In addition a phosphane-sulfide functionalised proline as alternative phosphane-peptide building blocks (Scheme 3b) and a procedure to introduce phosphane moieties into aromatic amino acids after peptide synthesis were reported (Scheme 3c).26 Several phosphane-containing oligopeptides with a helical motif were synthesised. This small library was evaluated on solid support as ligands for the rhodium-catalysed hydrogenation of amino acid precursors reaching up to 18 % ee (Scheme 3d).27 In addition to helical motifs, phosphane peptides containing a proline-β-turn were prepared both on solid support and in solution and applied as ligands in the palladium-catalysed asymmetric allylic substitution of 3-acetoxycyclopentene with dimethyl malonate (Scheme 3e).28 The product was obtained in up to 95 % ee in solution and 88 % ee on solid support. Interestingly, low ee was obtained when oligopeptides lacking the β-turn structure were used, showing this motif is essential for the induction of enantioselectivity in this system.
Inspired by the research of Gilbertson and co-workers, several other groups have created resin-supported phosphane-containing oligopeptides. Wills et al. reported the functionalisation of free amino groups of supported oligopeptides with para-diphenylphosphinobenzoic acid (Scheme 4a).29 The thus obtained ligands were tested in palladium-catalysed allylic substitution of diphenylallyl acetate, but no ee was observed. Meldal et al. reported the phosphinomethylation of primary and secondary amino groups of resin-bound oligopeptides to create diphosphane and P,S ligand systems (Scheme 4b–d).30 These systems were employed in the palladium-catalysed allylic substitution of diphenylallyl acetate obtaining up to 21 % ee for the P,P systems and up to 60 % ee for the P,S systems. Similar systems containing N-heterocyclic carbene ligands were successfully applied in palladium-catalysed Sonogashira and Suzuki cross-coupling reactions.31 The group of Landis created a series of resin-supported 3,4-diazaphospholanes modified with amino acids (Scheme 4e), resulting in a β-turn like structure.32 High selectivity was obtained when a small library of these supported phosphanes was employed in the palladium-catalysed allylic substitution of diphenylallyl acetate affording up to 92 % ee.
The previous examples show that the ligand systems using well-defined hydrogen bonding motifs gave generally higher selectivities compared to those with random chains, highlighting the importance of ligands with a well-defined three-dimensional structure. Lammertsma et al. reported biphosphane analogues of Gramicidin S, a cyclic decapeptide containing a rigid structure enforced by two β-turns.33 These peptides were synthesised on solid support by replacement of two ornithines with phosphane-containing amino acids. X-ray crystal structures of two of these cyclic peptides containing sulfur-protected phosphanes show that the phosphane amino acids are in a favourable conformation for diphosphane complex formation (Scheme 5a). Similar systems were prepared by modification of the two ornithine (Orn) amino acids with phosphane carboxylic acids through amide bond formation. (Scheme 5 b).34 Rhodium complexes of these cyclic peptides were characterised and used as hydrogenation catalysts, resulting in moderate enantioselectivities (up to 52 % R) using the meta-substituted cyclic peptide. Interestingly the opposite enantiomers were obtained in low ee when the para-substituted cyclic peptide was used. Low ee values (<15 %) were obtained in the palladium-catalysed allylic substitution of diphenylallyl acetate.
Other Oligopeptide Design Strategies
Another approach is to mimic protein structures using peptide-containing dendrimers. Arya et al. reported an amino acid based dendrimer containing several free amines, anchored on a solid support (Scheme 6).35 These amines were allowed to react with diphenylphosphinomethanol to create multiple diphosphane sites on the dendrimer. The created phosphane-containing dendrimers were used in rhodium-catalysed hydroformylation of a variety of olefins. Good conversions and branched selectivities were obtained. Rhodium leaching was significantly reduced by positioning the phosphane ligands at different sites in the dendrimer interior, which allowed several recycling cycles in hydroformylation.36
In the above sections several methods for mimicking enzymes using short synthetic peptides have been described. Larger peptide structures can also be obtained by using solid-phase peptide synthesis. Roelfes et al. utilised this to introduce a copper binding site in short (31 amino acids) peptides based on bovine pancreatic polypeptide (Scheme 7).37 This polypeptide was synthesised on solid support with strategically placed histidine and unnatural 3- and 4-pyridylalanine amino acids to create artificial copper binding sites. The resulting copper adducts showed good enantioselectivities in Diels–Alder reactions and Michael additions of 83 % and 86 % ee respectively.
Artificial Metallo DNAzymes
Oligonucleotides are another attractive class of biopolymers for catalyst development.38 Not only does the sugar–phosphate backbone provide a source of chirality, oligonucleotides have been found to exhibit molecular-recognition properties that rival those of proteins and can function as catalysts in their own right (DNA- and RNAzymes)39 Moreover, the well-known base-pairing rules allow the rational engineering of their secondary and tertiary structure, affording fascinating prospects in nanotechnology.40 These features make oligonucleotides well-suited for engineering sophisticated catalysts.39c–f Developments in the synthesis of hybrid transition-metal-containing DNA catalysts can be divided in two main approaches: supramolecular assembly and covalent anchoring.
Roelfes and co-workers pioneered this field with the development of copper dinitrogen complexes that bind to the DNA double helix by intercalation and/or groove binding. (Scheme 8).41 Assembled systems of various nitrogen copper complexes and commercially available salmon testes and calf thymus DNA served as catalysts for asymmetric Diels–Alder,42 Michael addition,43 and Friedel–Crafts alkylation44 reactions. By using azachalcone or α,β-unsaturated 2-acyl imidazoles as substrate, high ee values (>80 %) were reported for these reactions. Remarkably, all reactions were performed in water and a later study showed that the additions of small amounts of organic solvents can be beneficial to the reaction up to a level where the DNA starts to precipitate.45 Using different synthetic oligonucleotide strands it was shown that the enantioselectivity of the Diels–Alder reaction is sequence dependent.46L1 (n=2 R=3,5-dimethoxybenzene) provided high enantioselectivity with alternating guanidine–cytosine (G–C) sequences, whereas L2 gave high enantioselectivity when using short guanidine sequences (GGG, Scheme 8). Interestingly, high selectivity was still obtained with salmon testes and calf thymus DNA, although these sources contain an enormous number of different binding sites. These results indicated that the most selective catalytic species are also the most active. The same system was applied by the group of Roelfes in the copper-catalysed hydrolytic kinetic resolution of pyridyloxiranes47 and by Toru and co-workers in copper-catalysed electrophilic fluorination reactions.48 In a very recent study, the first non-enzymatic enantioselective and diastereospecific syn hydration of α,β-unsaturated ketones was achieved by using this system (Scheme 8).49 This hydration reaction has no equivalent in conventional chiral transition-metal catalysis and thus clearly demonstrates the potential of bioinspired approaches to provide transition-metal catalysts with unprecedented reaction profiles.
The alternative approach to obtain artificial metallo-DNAzymes is by covalent anchoring of catalytically active transition-metal complexes to DNA. This strategy fixes the location of the transition-metal catalysts which allows a more specific design of the ligand environment. The covalent attachment of ligands is however more challenging compared to supramolecular anchoring, as synthetic modification of the DNA is required. Due to incompatibility between DNA synthesis conditions and ligand synthesis, transition-metal binding ligands are generally introduced after DNA synthesis and purification to an incorporated unnatural nucleotide.
The group of Roelfes introduced a bipyridine–copper system covalently into DNA.50 They used an amine linker at a terminal position of an oligonucleotide, which was modified by amide bond formation with a carboxylic acid modified bipyridine (Scheme 9). The resulting oligonucleotide was then hybridised to a longer complementary scaffolding strand along with a second oligonucleotide. The assemblies were used as ligands in the copper-catalysed Diels–Alder reaction between azachalcone and cyclopentadiene. The conversion and enantioselectivity were found to be dependent on the oligonucleotide sequences used. Catalyst containing oligonucleotide 1 with a complementary strand gave 22 % ee, which could be improved to 93 % ee by addition of a second oligonucleotide and optimisation of the base sequences.
Jakobsen et al. reported the application of a polyaza-crown ether-containing DNA in the copper-catalysed Diels–Alder reaction of azachalcone and cyclopentadiene,51 but the obtained enantioselectivities were low (<10 % ee). Bipyridine–copper and salen–nickel complexes have also been introduced into RNA for sequence-specific DNA cleavage.52
The group of Jäschke synthesised several diene-modified oligonucleotides of 19 bases by modification of an introduced 4-triazolyldeoxyuridine (Scheme 10).53 These modified oligonucleotides were tested in the iridium-catalysed allylic amination of phenyl allyl acetate. Several complementary DNA and RNA strands were used including some that form loops in either the complementary or the ligand-containing strand. Both the ee of the substrate and the product were recorded as the conversion was found to be around 45 % for all systems. The reactions gave ee values up to 27 % and interestingly it appeared that the configuration of the major enantiomer was dependent on the complementary strand used. The influence of the complementary strands on the catalyst performance outlined in the examples above holds great promise for combinatorial screening and catalyst optimisation as the sequence of the complementary strand can easily be varied using automated DNA synthesis.
Kamer et al. reported the synthesis of phosphane-modified nucleotides based on 5-iodouridine, which were applied in palladium-catalysed allylic amination of diphenylallyl acetate (Scheme 11 a).54 A significant influence of the solvent was found on the enantioselectivity, with ee changing from 80 % (S) in tetrahydrofuran (THF) to 16 % (R) in a 50:50 mixture of acetonitrile and THF for R1=R2=H. For R1=R2=Ac the ee switched from 8 % (S) in THF to 23 % (R) in dichloromethane. Disappointingly, a trimer containing the phosphane-modified nucleotide gave lower enantioselectivities (<12 % ee). Functionalisation of longer DNA strands proved troublesome due to reduced rates of the palladium-catalysed cross-coupling step involved, which prompted the development of alternative modification strategies.
Jäschke and co-workers disclosed the successful modification of longer DNA strands with mono- and diphosphanes by introduction of a free amine in DNA strands followed by modification via amide bond formation with carboxylic acids containing phosphanes.55 Kamer et al. used a similar strategy to obtain phosphane-modified oligonucleotides (Scheme 11 b).56 Several phosphane-modified 15 base long oligonucleotides were synthesised and a palladium complex was reported. Initial catalysis experiments using the phosphane-modified mononucleotides resulted in full substrate conversion for the palladium-catalysed substitution of diphenyl acetate, but no ee. This was attributed to the large distance between the phosphane moiety and the chiral backbone in these nucleotides. Catalysis using either of the above phosphane-modified oligonucleotides systems developed by Kamer or Jäschke has not been reported so far.
These new synthetic strategies for the introduction of phosphane ligands into nucleotides allow many more opportunities for metallo DNAzyme applications. In combination with all the possibilities for molecular recognition and three-dimensional shaping of DNA these systems hold much promise for the future.
In spite of the pioneering work by Kaiser and Whitesides already published in the late 1970s, research on artificial metalloenzymes has flourished only in the last decade.57 Many strategies can be found in literature for incorporation of catalytic transition-metal centres into protein structures, each with their distinct features and (dis)advantages.58 These strategies can be divided into two main categories: non-covalent anchoring and covalent modification. The non-covalent anchoring approach can be divided into two subcategories: dative anchoring, employing the affinity of a protein for a transition metal, and supramolecular anchoring, using high-affinity protein–substrate interactions. For both subcategories a high complexation constant is a prerequisite to ensure precise and stable localisation of the transition-metal centre in the macromolecule. This condition is also a direct advantage compared to the covalent approach as catalyst assembly is in principle easily achieved by mixing the components. Applications of artificial metalloenzymes falling into these categories will be discussed to give an overview of the achievements so far in this exciting field of bioinspired catalysis.
The dative approach for the creation of artificial metalloenzymes involves direct coordination of transition metals to protein hosts in defined cavities or on the outer sphere. In this way a suitable transition-metal centre for the desired catalytic reaction can be introduced.59 Kaiser et al. were the first to report on this approach by replacing zinc in the active site of carboxypeptidase A (CPA) with copper, creating an oxidation catalyst for ascorbic acid.57b However, as proteins can have multiple sites for metal binding, selective creation of catalysts with a defined transition-metal centre can be problematic.60
The groups of Kazlauskas61 and Soumillion62 simultaneously reported the replacement of the zinc atom in the active site of carbonic anhydrases with manganese, leading to an artificial metalloenzyme for asymmetric epoxidation reactions. Similar enantioselectivities were reported (up to 67 % ee) for the epoxidation of styrene analogues (Scheme 12 a). In addition both groups reported the application of site-directed mutagenesis to identify several key amino acid residues, which control the enantioselectivity in these reactions.
More recently, the group of Kazlauskas also introduced rhodium in carbonic anhydrases to create a stilbene hydrogenation catalyst with high selectivity for the cis substrate (Scheme 12 b).59 The significant amount of trans-stilbene formed by isomerisation due to surface-bound rhodium was reduced by removal of surface lysine and histidine residues using a combination of site-directed mutagenesis and chemical modification. The same procedure resulted in a catalyst that gave surprisingly high linear selectivity in the hydroformylation of styrene (Scheme 12 c).63 More examples of oxidation,64 hydrogenation65 and hydroformylation66 catalysis using dative systems have been reported.
Reetz and co-workers recently reported the creation of an artificial copper-binding site in the synthase subunit of imidazole glycerol phosphate synthase from Thermotoga maritima (tHisF) (Scheme 12 d).67 The introduced motif consisted of two histidine and one aspartic acid residue located at a strategic position to bind a copper atom. The system was applied in the copper-catalysed Diels–Alder reaction between azachalcone and cyclopentadiene leading to 35 % ee with improved activity compared to the native protein. The removal of several native histidines, which could act as alternative copper binding sites, led to an increase of the enantioselectivity up to 46 %.
These examples of artificial metalloenzymes using dative protein–transition-metal interactions illustrate how such systems can lead to catalysts showing interesting enantio- and chemoselectivity for non-native reactions. Engineering of the protein sequences using mutagenesis techniques proved to be a useful catalyst optimisation tool.
Supramolecular Anchoring Using Biotin–Avidin Technology
Wilson and Whitesides reported already in 1978 the formation of an artificial metalloenzyme using a rhodium–diphosphane complex modified with biotin in combination with avidin (Scheme 13).57a The interaction between avidin and biotin is one of the strongest protein–substrate interactions known (Ka=1015M−1) and is considered to be irreversible. This supramolecular assembly was applied in asymmetric hydrogenation of the prochiral amino acid precursor N-acetamidoacrylic acid, affording full conversion to the product with an ee of 41 % (S) (Scheme 13 a).This seminal work by Whitesides and co-workers inspired several groups to further exploit the potential of this system, using combined synthetic and biochemical techniques for catalyst optimisation. Several studies were published in which the system was changed synthetically by varying the linker using different aliphatic and aromatic inserts and changing the diphosphane motif.68 The real power of this system, however, was demonstrated using complementary biochemical tools for optimising the protein scaffold alongside these synthetic chemical modifications. Ward et al. found that changing the protein host to streptavidin improved the system drastically, which was explained by the deeper binding pocket and less cationic character of this protein.68b This catalyst system performed the hydrogenation of N-acetamidoacrylic acid and N-acetamidocinnamic acid with over 90 % ee (R). Using a small library of different linkers and catalytic rhodium centres in combination with a selected set of mutants of streptavidin, the selectivity could be inverted to 57 % ee (S) for N-acetamidoacrylic acid.68c Further optimisation using saturation mutagenesis at the site of the mutation in the most promising mutant scaffold resulted in over 60 % ee (S) in the hydrogenation of N-acetamidoacrylic acid and over 80 % (S) for N-acetamidocinnamic acid.69 This could be further enhanced for both substrates to up to 95 % ee (S) and a threefold rate increase using a combination of enantiopure amino acid linkers and saturation mutagenesis.70 These so-called second-generation hybrid catalysts also showed improved organic solvent tolerance, which is important as substrate solubility in pure water is one of the main limitations of using protein structures as macromolecular catalyst scaffolds. Reetz et al. used directed evolution as a Darwinian approach for catalyst optimisation.71 By applying several rounds of “evolution” and saturation mutagenesis the enantioselectivity provided by Whitesides’ catalytic system could either be improved from 23 % ee (R) to 65 % ee (R) or inverted to 7 % ee (S) for the hydrogenation of N-acetamideacrylic acid.
The group of Ward successfully extended this approach to other transition-metal-catalysed reactions. Using biotin modified with different aminosulfonamide ruthenium η6-arenes Ward performed asymmetric transfer hydrogenation of prochiral ketones, achieving up to 94 % ee (R) using acetophenone as substrate (Scheme 13 b).72 Aromatic and aliphatic ketones were hydrogenated with good enantioselectivity, provided that the ketone contained two substituents that varied sufficiently in bulk size.73 Using an X-ray crystal structure obtained of one of these hybrid catalysts, the system was further improved by using designed evolution.74 Systems using either benzene or p-cymene as arene could both be converted into R- and S-selective 4-phenyl-2-butanone transfer hydrogenation catalysts. After further optimisation using mutagenesis p-methylbenzophenone could be hydrogenated with 98 % ee (R).75 A d6 piano stool ruthenium complex used in hybrid transfer hydrogenation catalysts was also found to afford metalloenzymes that catalyse alcohol oxidation using tert-butylhydroperoxide as oxidant.76 Rhodium and iridium analogues were also created but were found to be significantly less active.
More recently, Ward et al. evaluated the previously established set of phosphane-modified biotin ligands in palladium-catalysed allylic alkylation of diphenyl allyl acetate (Scheme 13 c).77 Also in this catalytic system over 90 % ee (R) and 80 % ee (S) could be achieved by the introduction of a variety of linkers and mutations. The biotin–streptavidin system was also used by the group of Ward for the sulfoxidation of thioanisoles using biotin-modified with different manganese–salen complexes (Scheme 13 d).78 These systems only gave low enantioselectivities (<15 % ee).
Other Supramolecular-Assembled Artificial Metalloenzymes
Additional supramolecular anchoring methodologies have been applied by using different protein–substrate/cofactor interactions, mostly for asymmetric oxidation reactions.79 One successful example of such a system was reported by Gross et al. who used iron- and manganese-containing corroles anchored to human serum albumin (HSA) as artificial metalloenzymes (Scheme 14 a).80 Up to 74 % ee was obtained for the oxidation of several aryl methyl sulfides using these complexes.81
Aided by computational methods, the group of Watanabe rationally designed active site mutants of myoglobin for reconstitution with chromium–salphen complexes (Scheme 14 b).82 The sulfoxidation of thioanisole using the obtained catalysts led to inversion of the enantioselectivity from 4.6 % ee (R) for the wild-type protein to 13 % ee (S) for the most successful mutant. The system was further improved by a combination of mutant design and the introduction of manganese instead of chromium to create catalysts that gave selectivities ranging from 33 % ee (S) to 27 % ee (R).83
Supramolecular-assembled systems have also been applied in Diels–Alder reactions. Reetz and co-workers applied copper–phthalocyanine complexes in combination with different serum albumins (Scheme 14 c).84 Enantioselectivities ranging from 85 % to 95 % and endo-selectivities ranging from 91 % to 95 % were found. In addition to interactions between the ligands and the proteins, catalyst anchoring in the above systems is likely to involve the coordination of at least one amino acid to the transition metal, and therefore several of these systems have also been categorised as dative artificial metalloenzymes.58b
The biotin–avidin and other above-mentioned supramolecular assembled systems clearly demonstrate how artificial metalloenzymes can benefit from optimisation opportunities that are unprecedented in either traditional transition-metal catalysis or enzyme catalysis alone. These catalyst systems can be tuned for specific reactions and substrates by using combinations of both synthetic organic and biomolecular techniques.
Supramolecular Anchoring Using Antibody Complexes
The non-covalently assembled artificial metalloenzymes discussed above employ protein binding of artificial cofactors which resembles naturally occurring substrate–protein interactions. Supramolecular protein–catalyst interactions can also be engineered by raising monoclonal antibodies against a transition-metal complex of interest. Keinan et al. created monoclonal antibodies using an α-naphthoxy tin porphyrin as transition state mimic for the sulfoxidation of thioanisole.85 Obtained antibody SN37.4 in combination with a ruthenium porphyrin performed the sulfoxidation of thioanisole and analogues with up to 43 % ee (S) (Scheme 15 a). To make up for the lack of coordinating amino acids like histidine and cysteine in antibodies microperoxidase 8 (MP8), a heme attached to an octapeptide containing a histidine, was developed by the group of Mahy (Scheme 15 b).86 Using this heme in combination with antibody 3A3 improved the rates for the oxidation of thioanisole and the obtained selectivity was 45 % ee (R).87 More recently, Mahy and co-workers covalently attached a porphyrin to estradiol, which, in combination with estradiol antibody 7A3, catalysed the oxidation of 2,2′-azino-bis(3-ethylbenzothiazoline-6-sulfonate) (Scheme 15 c).88 This complex did show activity for asymmetric oxidation of thioanisole but only low ee of 11 % (R) was found.89
Harada and co-workers demonstrated that this approach can be extended to phosphane transition-metal complexes by raising monoclonal antibodies against the complex show in Scheme 16.90 Several obtained antibodies combined with the diphosphane–rhodium complex were tested for hydrogenation of amino acid precursors. One combination showed 98 % ee (S) in the hydrogenation of N-acetamidoacrylic acid and also gave the best product yield (23 %).
The true power of enzymes, that is, their tremendous rate acceleration, high substrate selectivity and selective transformation of multifunctional unprotected substrates, is dependent on highly organised protein–substrate interactions. Therefore, combining the shape selectivity and powerful substrate recognition of proteins with optimal catalyst performance achieved via rational ligand design requires the use of protein structures that can encapsulate the substrate to enforce optimal orientation with respect to the artificial cofactor (Figure 2). Covalent anchoring allows exploitation of protein structures, which have the desired structural properties for efficient substrate–protein interactions, because, unlike for supramolecular assembled artificial metalloenzymes, the binding site is not occupied for the catalyst assembly. This can lead to the full exploitation of the shape-selectivity of promising protein structures.
A variety of strategies are available for site-selective covalent modification of proteins for the creation of artificial metalloenzymes.91 Several reported methods rely on specific reactivity of certain amino acids residues in a specific protein environment92 or on changing the reactivity of such residues.93 However, these methods share the disadvantage of supramolecular systems that these only allow the use of select protein structures. A more flexible method is the bioconjugation to cysteine, which is very common in biochemistry. Because of the relatively low abundance of this amino acid (1.7 %), proteins containing a unique cysteine can typically be easily obtained.
Kaiser et al. were the first to create an artificial enzyme by covalent modification of an amino acid by alkylation of the cysteine 25 in papain (Scheme 17 a) with bromoacetyl-functionalised flavins.94 This seminal report was an inspiration for much future work for the creation of artificial metalloenzymes using the same covalent attachment procedure.92e, 95 De Vries and co-workers attached a phosphite cofactor to papain (Scheme 17 b).96 After rhodium complexation this modified protein was used for hydrogenation of N-acetamidoacrylic acid methyl ester showing full conversion but no ee. Recently Salmain and co-workers attached an η6-arene ruthenium phenantroline complex to papain (Scheme 17 c).97 The obtained artificial metalloenzyme showed a significant rate enhancement in the Diels–Alder reaction between cyclopentadiene and acrolein, but no ee was observed.
Both Reetz and De Vries have suggested that papain is an unsuitable protein host for application in asymmetric catalysis due to large conformational flexibility. The importance of conformational rigidity was also identified by the group of Lu when they improved their myoglobin manganese salen catalyst system by using an intriguing dual anchoring strategy (Scheme 18).98 This artificial metalloenzyme showed increased activity and improved ee of 51 % (S) compared to 21 % (S) found for the single-point anchored system. It thus appears that structural rigidity is an important factor in artificial metalloenzyme performance, likely by reducing the number of accessible structures of the catalytic species by conformational dynamics of the modified proteins and/or increasing the functional interactions between the host and the metal-complex.
Distefano et al. clearly showed that application of a site-specific covalent modification strategy makes it possible to rationally choose a suitable protein host based on specific structural and/or functional characteristics. The adipocyte lipid binding protein (ALBP) was specifically selected as host protein for its ability to truly encapsulate catalysts in its 600 Å3 cavity, and its capacity to sequester a variety of apolar substrates.99 The unique cysteine at position 117 was modified by alkylation using 5-iodoacetamide-1,10-phenantroline, a cofactor previously linked by the same covalent approach to DNA-binding proteins for specific DNA cleavage.100 Copper conjugates were created and used in aryl amide hydrolysis and selective ester hydrolysis via kinetic resolution, achieving up to 86 % ee (L-specific, Scheme 19 a). X-ray crystallography provided structures of the modified protein and fluorescence studies confirmed copper complexation.101 The selectivity of the ester hydrolysis reaction could be further improved to 94 % ee (L-specific) by employing three mutant proteins of intestinal fatty acid binding protein (IFABP) with varying position of the cysteine as protein scaffold. Kamer et al. used ALBP for modification with an η6-p-cymene ruthenium phenantroline complex via a maleimide linker (Scheme 19 b).103 The same cofactor was also coupled to photoactive yellow protein (PYP) Application of these complexes in the transfer hydrogenation of acetophenone afforded only trace amounts of product.
Bioconjugation of Phosphane Ligands
Whereas the supramolecular anchoring of diphosphane-based catalysts has evolved over three decades into one of the most successful approaches in artificial metalloenzyme development to date, the covalent site-specific modification of proteins with phosphanes has only recently been achieved. The attachment of a diphosphane to the active site serine of a lipase via a phosphonate inhibitor was reported by Reetz and co-workers but the linkage proved hydrolytically unstable.92e The strong nucleophilic character of phosphanes makes cysteine-selective bioconjugation of such ligands a non-trivial task since most common electrophiles used for cysteine modification (e.g. iodoacetamides, maleimides) will undergo reaction with the phosphane. Protected phosphines (e.g. by sulfur or BH3 protection) can be introduced, however deprotection methods were found incompatible or ineffective for the obtained conjugates.104
Recently our group reported the modification of the native cysteine of PYP using 1,1′-carbonyldiimidazole(CDI)-activated carboxylic acid derivatives of mono- and diphosphanes (Scheme 20 a).103, 105 The similar anchoring of palladium complexes afforded metalloenzymes active as allylic amination catalysts. Several rhodium complexes were also created by using the same conjugation strategy, affording an active catalyst for the hydrogenation of dimethyl itaconate using a coupled diphosphane ligand (Scheme 20 b).105 Unfortunately no ee was found in allylic substitution and only up to 6 % ee in the hydrogenation of dimethylitaconate. The denaturation of the protein due to the presence of organic co-solvents, required to obtain reproducible conversion, was thought to be the cause for the lack of enantioselectivity of these catalyst systems.
The same method was also applied to other protein templates, but this led to unselective or incomplete protein modification and consequently the development for effective bioconjugation techniques for phosphanes was continued.104 By combining two well-developed and straightforward methods for protein modification we recently achieved highly site-selective and efficient functionalisation of three structurally different proteins with phosphane ligands. (Scheme 21).106 By first introducing a hydrazide by cysteine-selective maleimide addition followed by a reaction with an aldehyde containing phosphane, a variety of mono- and diphosphane moieties were selectively introduced. In addition, the formation of rhodium complexes was demonstrated by mass spectrometry and 31P NMR spectroscopy.
These developments open the way to create artificial metalloenzymes by highly chemoselective modification of proteins with transition-metal-coordinating phosphane ligands. Akin to the powerful strategy established for the biotin–avidin system, these systems can be developed in a modular fashion. Structural diversity of the synthetic catalyst moiety can be achieved by varying the nature of the metal, the phosphane ligand and hydrazide components. But importantly, the structure space of the host can now be extended beyond (strept)avidin, only constrained by the presence of a unique cysteine as selective nucleophilic modification site.
Genetic Encoding of Transition-Metal Binding Ligands
Recent developments in biochemistry demonstrated the introduction of additional amino acids beyond the natural 20 in eukaryotic and prokaryotic in systems.107 This technique allows site-selective introduction of transition-metal binding ligands in virtually any protein structure without challenging chemical modification. Schultz and co-workers demonstrated the effectiveness of such a system by introducing bipyridine-substituted alanine in DNA binding catabolite activator protein (CAP).108 The resulting artificial metalloenzyme containing copper as active metal was used as a sequence-selective DNA cleavage catalyst (Figure 3). Such systems show much promise for future application as artificial metalloenzymes, although development and use of these systems requires significant know-how and is not straight forward.
Optimisation Techniques for Artificial Metalloenzymes
The various approaches outlined above illustrate how artificial metalloenzymes provide unique opportunities for catalyst optimisation. In addition to traditional synthetic tools for optimisation of the various components of the metal complex, the spacer linking the catalyst to the protein and the structure of the protein scaffold can be optimised, a strategy for which the term “chemogenetic” approach has been coined by Distefano and Häring.109
Structural variants of protein scaffold can nowadays easily be generated using advanced mutagenesis strategies. Whereas most of the systems to date have only been subjected to limited mutagenesis to tune their performance, the studies of Ward and Reetz on the avidin–biotin system highlight the potential of evolving systems by the combinatorial screening of large numbers of mutant proteins.110
Directed evolution is a very powerful tool for the optimisation of the properties of proteins, which is increasingly applied to improve enzyme performance and protein properties.111 This approach uses repeating cycles of random mutagenesis followed by screening for the best hits in a Darwinian fashion.
This methodology has been successfully applied to tailor the structure of the protein scaffold in artificial metalloenzymes based on the rhodium diphosphane modified biotin-streptavidin system by the group of Reetz (Figure 4).71 To reduce the screening effort, Reetz et al. applied a method for mutagenesis named Combinatorial Active-site Saturation Test (CAST),112 in which a selection of appropriate amino acids is subjected to saturation mutagenesis.113 Although successful, the procedure was reported to be labor-intensive.
Ward et al. have implemented an alternative strategy, whereby “evolution” of the enzyme is achieved by chemogenetic optimisation guided by rational design, an approach which they called “designed evolution”.114 In contrast to directed evolution, in this process the mutagenesis in every optimisation cycle is steered by design choices based on structural information. This significantly reduces the number of mutants employed in the procedure, while the optimisation process is still very effective.
Since synthesis of artificial metalloenzymes involves an assembly between the host and the catalytic metal (complex), purified proteins will generally be required. Therefore, to facilitate screening efforts, efficient methodology for the parallel expression and purification of protein variants needs to be developed, which has been found to be a limiting factor in several systems.95
Regarding the functionalisation of proteins with synthetic organometallic complexes, the use of generally applicable covalent anchoring methodology, such as cysteine modification, has a significant advantage of increased structural diversity compared to supramolecular systems. As covalent assembly is much less dependent on the protein structure, this brings much more flexibility to the design of the systems, and offers the possibility to make rational choices concerning the protein scaffold(s) to be employed. For example, Reetz and co-workers set out to develop a platform for artificial metalloenzymes based on a robust protein specifically chosen for its suitability to develop en masse purification and bioconjugation procedures.67, 115 Also particular protein structures may be rationally matched with the desired (shape) selectivity in a target reaction. However, applications of covalently assembled systems have not been as successful as their supramolecular counterparts, whose straightforward assembly greatly facilitates combinatorial screening efforts. The synthesis and screening of libraries of covalently modified proteins requires the development of assembly procedures which involve minimal purification and can be performed in a parallel fashion. Part of the current efforts in our laboratory is directed towards optimising the phosphane-conjugation methodology in this direction.
Structural information obtained from X-ray crystallography, NMR or modelling studies can be used to guide the design of optimal protein structure. Effective design still requires improved tools for the prediction of tertiary structure of proteins and a better understanding of all factors involved in protein dynamics has to be obtained. New computational methods have been developed which may aid artificial metalloenzyme optimisation. Using ORBIT (Optimisation of Rotamers By Iterative Techniques)116 the group of Mayo created a p-nitrophenyl acetate hydrolase from a catalytically inert 108-amino acid residue of Escherichia coli thioredoxin.117 Hellinga and co-workers developed a computational method called DEZYMER for the redesign of binding sites in proteins, to make them bind other substrates.118 This program was successfully applied to design binding sites for trinitrotoluene, L-lactate, serotonin and pinacolyl methyl phosphonic acid.119 More recently the group of Baker developed a computational method that screens protein structures to find optimal active site replacements named RosettaMatch.120 This program was used to design a novel enzyme for the retro-aldol condensation of an unnatural substrate,5 and to create new enzymes for the not naturally occurring Diels–Alder reaction121 and Kemp elimination.122 The last example involved both in silico enzyme design and directed evolution for catalyst development resulting in a 200-fold rate increase.
The use of advanced biochemical and computational methods in combination with synthetic organic optimisation of cofactors provide a powerful set of tools for further development of new artificial metalloenzyme catalyst systems.
Summary and Outlook
The last decade has witnessed an exponential growth of interdisciplinary research aimed at the construction of bioinspired ligand systems for transition-metal-catalysed reactions. The concepts underlying biomolecular assembly and biocatalysis have motivated the development of a variety of intriguing approaches. Peptides are used as a source of chirality but also elegantly exploited for supramolecular assembly. Natural and synthetic DNA is explored as scaffold for the synthesis of hybrid catalysts, leading to the discovery of a catalytic activity hitherto not seen for any transition-metal catalyst. The power of peptide and DNA synthesis and the elegance of supramolecular chemistry offer fascinating prospects for the field of transition-metal catalysis.
The merging of transition-metal catalysts with natural protein scaffolds leads to enzymes with unnatural catalytic activities. It is clear that the choice of the host protein is a major challenge in the field. Not only should the host be amenable to modification, it should have a structure that is appropriate to accommodate the desired catalyst and be compatible with the envisioned reaction conditions. For efficient chirality transfer from the scaffold to the catalytic reaction the catalytic centre should be in close proximity to the protein structure. On the other hand, the structure of the synthetic catalyst and the chemistry used for its introduction should not adversely affect the stability and tertiary structure of the scaffold. Finding the right balance between these factors is crucial for the development of successful systems.
Whereas most of the systems described above show at least moderate levels of activity and selectivity, providing good starting points for further optimization, some perform very poorly. Low selectivity in catalytic reactions may arise from too little interaction of the catalytic moiety with the host. This probably underlies the poor performance of papain as host, whose cysteine used for catalyst anchoring lies in a very shallow groove on the protein surface. Decreasing the conformational freedom of the catalyst by designing additional interactions between host and catalyst (e.g. by a dual anchoring strategy) may improve the performance in such cases.
Alternatively, the introduced catalyst may engage in adverse interactions with the host, leading to a system in which the catalytic centre becomes inaccessible or inactive. This scenario probably underpins the poor results we have obtained using PYP as host.
Despite the various design challenges, the diversity of systems and applications reported to date underline the potential of this approach, and it is evident that the application of advanced biomolecular techniques such as site-directed mutagenesis and directed evolution in combination with traditional synthetic tools is a very powerful approach for optimising the performance of these systems. Continuous progress in the computational design and modelling techniques for both DNA and polypeptide structures will lead to further insight and possibilities of these systems.
Additional major challenges that will need to be addressed are the miniaturisation and parallelisation of the synthesis of the hybrid systems to truly exploit the possibilities these systems offer for combinatorial screening.
With the powerful tools available and considering the increasing research efforts in the field we anticipate that further advances in the development of these hybrid catalyst systems will lead to effective solutions for demanding transition-metal-catalysed chemical reactions.
We are indebted to the postdoctoral researchers, graduate and undergraduate students as well as the research groups we collaborate with, for their valuable contributions to this research. We thank the European Union (Marie Curie excellence grant MEXT-2004-014320; NEST Adventure STREP Project artizymes contract No. FP6-2003-NEST-B3 15471; Network of Excellence Idecat (Idecat-CT-2005-011730; COST action CM0802 PhoSciNet), EASTCHEM, the National Research School Combination-Catalysis NRSCC and Sasol Technology UK Ltd. for funding.
Peter Joseph Deuss was born in Amsterdam (The Netherlands). He received his B.Sc. and M.Sc in 2004 and 2006, respectively, both from the University of Amsterdam. After completing his master’s degree he moved that same year to Scotland to start his PhD research at the University of St. Andrews (UK) under supervision of Professor Paul C. J. Kamer on a project in collaboration with SASOL. His research interests focus on the development of artificial metalloenzymes for application in demanding transition-metal-catalysed reactions.
René den Heeten studied chemistry at Leiden University and received his PhD in 2009 at the University of Amsterdam. His PhD research was done in the group of Prof. P. W. N. M. van Leeuwen and Prof. P. C. J. Kamer and dealt with the development of new bioinspired transition-metal catalysts and artificial transition metalloenzymes. He attended the group of Prof. R. V. A. Orru at the VU University Amsterdam as a postdoctoral fellow. Currently he is working at Huntsman Polyurethanes in Rozenburg.
Wouter Laan was educated at the University of Amsterdam, where he obtained his M.Sc.in Chemistry in 2000 and his PhD in molecular microbiology in 2005 under the supervision of Prof. Klaas Hellingwerf. He is currently working as a post-doctoral research fellow in the group of Prof. Paul Kamer at the University of St Andrews. His current research interests lie in the development of bioinspired transition-metal catalysts, with a focus on the synthesis of protein-based artificial metalloenzymes.
Paul Kamer obtained a degree in biochemistry at the University of Amsterdam and did his PhD in physical organic chemistry at the University of Utrecht. As a postdoctoral fellow of the Dutch Cancer Society (KWF) he carried out postdoctoral research at the California Institute of Technology and the University of Leiden. He was appointed Lecturer at the University of Amsterdam and full Professor of homogeneous catalysis in 2005. In the same year he received a Marie Curie Excellence Grant and moved to the University of St Andrews. His current research interests are (asymmetric) homogeneous catalysis, biocatalysis, combinatorial synthesis, and artificial metalloenzymes.