* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download Directed Enzyme Evolution and High
Promoter (genetics) wikipedia , lookup
Western blot wikipedia , lookup
Gene expression wikipedia , lookup
History of molecular evolution wikipedia , lookup
Ancestral sequence reconstruction wikipedia , lookup
Gene expression profiling wikipedia , lookup
Biochemistry wikipedia , lookup
Genome evolution wikipedia , lookup
Protein moonlighting wikipedia , lookup
Vectors in gene therapy wikipedia , lookup
Endogenous retrovirus wikipedia , lookup
Gene regulatory network wikipedia , lookup
Biosynthesis wikipedia , lookup
Genomic library wikipedia , lookup
Cre-Lox recombination wikipedia , lookup
Community fingerprinting wikipedia , lookup
Silencer (genetics) wikipedia , lookup
List of types of proteins wikipedia , lookup
Two-hybrid screening wikipedia , lookup
Deoxyribozyme wikipedia , lookup
3 Directed Enzyme Evolution and High-Throughput Screening Michael J. McLachlan,1 Ryan P. Sullivan2 and Huimin Zhao3 1 Center for Biophysics and Computational Biology, University of Illinois at Urbana-Champaign, Urbana, IL 61801, USA 2 Department of Chemical and Biomolecular Engineering, University of Illinois at Urbana-Champaign, Urbana, IL 61801, USA 3 Departments of Chemical and Biomolecular Engineering, and Chemistry, University of Illinois at Urbana-Champaign, Urbana, IL 61801, USA 3.1 Introduction Early enzyme products developed for industrial applications were produced by native hosts via fermentation and consisted of a complex mixture of secreted enzymes produced at low yields. Now, over 90% of industrial enzymes are produced recombinantly for maximal purity and productivity [1]. The integration of enzymes into a commercial process relies on the availability of those with high activity and stability under process conditions, desired substrate specificity, and high selectivity. More often than not, naturally occurring enzymes do not fulfill the requirements of these harsh industrial conditions, and optimization is necessary to obtain a suitable enzyme catalyst for production needs. This tailoring of enzymes can be accomplished through two experimental routes. The first is rational design, which targets specific residues of a protein for mutagenesis to predetermined amino acid mutations, and is only applicable when there is detailed knowledge of the relationships between the enzymes structure and mechanism/function. And while an increasing number of enzymes are being characterized, the majority do not have this depth of information readily available, as it requires considerable effort to obtain. In the absence of this information, the tailoring of an enzyme can still be accomplished through the second route: directed evolution. Biocatalysis for the Pharmaceutical Industry : Discovery, Development, and Manufacturing edited by J. Tao, G.-Q. Lin, and A. L. © 2009 John Wiley & Sons Asia (Pte) Ltd. ISBN: 978-0-470-82314-9 46 Biocatalysis for the Pharmaceutical Industry Directed evolution is the general term applied to the combined techniques of generation of a library of protein mutants (or variants) and selection of a protein with desirable function from within that library [2]. It is an iterative Darwinian optimization process, whereby the fittest variants are selected from an ensemble of mutants [3]. Directed evolution can be used to target a number of enzymatic characteristics, including activity, substrate specificity, thermal and oxidative stability, enantioselectivity or enantiospecificity, pH optima or range, and tolerance to solvent [4]. While a typical directed evolution experiment focuses on a single enzymatic trait, there are some examples of improving several traits simultaneously. Developed primarily in academic laboratories, directed evolution practices in industry are still in their infancy and present an imposing challenge for process scientists today. Choosing the appropriate methods of library generation and screening or selection is paramount to the success of any directed evolution experiment. Library diversity can be created through either mutagenesis (random or semi-rational) or gene recombination, and which of these methods is chosen depends on many factors, such as the availabilities of homologous genes, structural knowledge, and characteristic data of the enzyme of interest. The library size created is typically very large ( > 104–6), and close evaluation of each variant is not feasible. The need for a method to find the improved ‘needle in a haystack’ enzyme becomes evident, with several strategies, including selection, enrichment, and high-throughput screening, offering ways to sift through the library clutter and find a variant with the desired enzymatic trait. However, once an evolved enzyme has been found that exhibits improved characteristics, the artificial conditions in which the selection method was carried out may result in an enzyme whose properties may not carry over to the real biocatalytic process. Therefore, the more similar a screening system is to the actual application process, the more likely it is to find an improved enzyme that will be complementary to the application. The aims of this chapter are to familiarize the reader with the various techniques used in library creation and high-throughput screening, as well as representative examples in the pharmaceutical industry and the technology used to adapt the laboratory-based techniques to industrial settings. The topics covered do not represent a complete listing of methods utilized, as more strategies for library creation and high-throughput screening are constantly introduced. 3.2 Directed Evolution Library Creation Strategies Library creation strategies generally fall into three main categories: random mutagenesis, semi-rational design, and gene shuffling (Figure 3.1). Random mutagenesis introduces mutations throughout a target gene encoding for the industrial enzyme of interest. These mutations may be in the form of point mutations (either transitions or transversions), insertions, deletions, inversions, or frame-shift mutations. Semi-rational design is a combination of random mutagenesis and site-directed mutagenesis, in which specific residue positions are rationally determined to play important roles in the enzymes function, and are subsequently randomized to all 20 amino acids. Gene shuffling involves exchanging fragments of genes with one another to create a library of chimeric progeny. In directed evolution studies, this is typically accomplished by homologous recombination for sequences with high similarity or by nonhomologous recombination for those with low similarity. Directed Enzyme Evolution and High-Throughput Screening 47 Figure 3.1 Overview of DNA library creation strategies. Random mutagenesis introduces mutations at positions throughout the gene sequence. Semi-rational design randomizes only the specific position(s) of interest. Gene shuffling brings existing sequence diversity from different parental DNA sequences together to form a chimeric library 3.2.1 Random and Semi-Rational Mutagenesis The most commonly used method for random mutagenesis library creation is error-prone polymerase chain reaction (epPCR), which introduces mutations in a gene product by lowering the fidelity of DNA polymerase. This can be achieved through many different reaction conditions, including use of nucleotide base analogs and alkylating agents into the polymerase chain reaction (PCR) reaction mixture, but severe biases typically result from such modification. The more favored protocols utilize an Mg2þ -dependent polymerase that lacks exonuclease activity, substitute Mn2þ for Mg2þ , and extend DNA products with an uneven mixture of the four precursor deoxynucleotidetriphosphates [5]. The mutation rate can be easily adjusted by varying the concentration of Mn2þ in the reaction mixture or by changing the number of cycles of amplification. Typically, a library is created with a low mutation rate (one to five base pairs, or one to three amino acids per gene) to prevent disruption of enzymatic activity and generation of a library too large to screen comprehensively. However, where there is an efficient selection scheme or robust high-throughput screen available, higher mutation rates have been successful in gaining improved function [6,7]. epPCR is considered as a random mutagenesis technique, but the term random should be interpreted loosely. Ideally, a protein library should satisfy the requirement of having an equal probability of substitution of any of the 20 amino acids randomly into any residue position of the protein. However, the library creation is performed at the level of the gene encoding the protein; thus, an ideal distribution is improbable due to the redundancy of the genetic code. Mutations at the wobble position (third base of the codon) typically result in a neutral mutation, and the frequency of two adjacent bases being mutated is considerably low, which results in only six possible amino acids on average at each position. As a result, the DNA sequences of selected progeny with improved function are determined, and positions at which mutations have been incorporated are sometimes subjected to an additional round of screening using saturation mutagenesis. Saturation mutagenesis involves creating a library by designing 48 Biocatalysis for the Pharmaceutical Industry degenerate primers for the residue position in question and screening the resultant library to determine which of the 20 amino acids results in the most dramatic improved effect at that position. Gene site saturation mutagenesis (GSSM) takes this concept to its fullest extent by creating a library consisting of all 20 possible amino acid substitutions for each position in a protein [8,9]. Sequence saturation mutagenesis (SeSaM) [10] is an alternative four-step approach to epPCR that overcomes several barriers faced by its predecessor. First, a library is created with standard nucleotides with the presence of a-phosphothionate nucleotide, an alkaline labile analog. The incorporated analog is hydrolyzed, thus creating a library of fragments of random length. The fragments are then treated with a terminal deoxynucleotidyl tranferase (TdT) to incorporate a random number of universal bases at the 30 -termini, followed by elongation of labeled gene product to full length. A concluding PCR then replaces the universal base substitutions with standard nucleotides. The mutations incorporated into the SeSaM library are randomly distributed throughout the parent gene template and independent of the DNA polymerase, thus avoiding any potential mutational bias and resulting in the exchange of an amino acid at any position to all other 19 possible amino acids. The SeSaM protocol has also been modified to enrich the transversion rate and allow for adjustable mutational biases [11]. However, the use of different base analogs, biotinylated primers, and the additional time-consuming steps for library creation are drawbacks to this method. Insertion and deletion mutagenesis expands the reach of random mutagenesis to include alteration of the size of the gene of interest when generating a library. This process requires the use of transposable elements that insert randomly into the target and are subsequently excised, leaving behind an in-frame fragment. Pentapeptide scanning mutagenesis results in the insertion of a 15-bp fragment, but the amino acids incorporated are predetermined due to the cognate target site of the transposon and preset restriction site sequence [12]. This results in a less than optimal library with less diversity and more bias than other methods. It is particularly suited towards obtaining mutants whose loop structures dictate activity enhancements. An improved technique of the same style is random insertion/deletion, which makes it possible to delete a multitude of consecutive bases from random locations in a gene and insert either a specific or randomized sequence of arbitrary length in its place [13]. The diversity obtained from this method far outstretches that of conventional epPCR, but it comes at the cost of several additional processing steps. 3.2.2 Gene Shuffling Although random mutagenesis techniques offer a relatively simple scheme to approach a desired enzymatic trait, the frequency of gaining beneficial mutations is generally low. Typically, only one or two amino acid changes are generated in each round, with higher random mutation rates leading to either a library too large to screen successfully or to incorporation of mutations that lead to loss of function. Gene shuffling can overcome this limitation by allowing a large number of beneficial mutations from multiple genes to incorporate at a single step of the library creation process. There are a variety of shuffling methods available, which can be divided into categories of homology dependent and homology independent. Directed Enzyme Evolution and High-Throughput Screening 49 3.2.2.1 Homology-Dependent, Primer-Independent Assembly The pioneering technique of DNA shuffling [14] was a breakthrough for in vitro recombinant mutagenesis and diversity generation. The method is centered on taking a number of homologous parental template genes, randomly digesting them with DNase I and reassembling the pieces back together into full-length genes through a primerless PCR reaction (Figure 3.2a). In this extension process, pieces of different genes can anneal to each other at homologous regions and extend to full-length fragments through the aid of DNA polymerase. DNA shuffling can be applied to combining beneficial mutations found from random mutagenesis screening, which can reduce the processing steps required to determine which mutations screened for are actually necessary for improved function. A single gene variation of DNA shuffling involves incorporation of mutations at the extension level, screening for improved progeny, and subsequent shuffling of these mutants for further screening. In contrast, family shuffling [15] begins the process with naturally occurring homologous genes (typically > 60% identical). Because the parental templates in this process have already been subjected to natural selection, much larger numbers of mutations are tolerated, leading to a library with a broader sequence space that maintains a higher percentage of functional enzymes. These methods all have similar limitations, however, as they rely on sequence homology for recombination events to occur. Gene fragments will tend to anneal to each other only at areas of Figure 3.2 Examples of gene shuffling methods used for DNA library creation in directed evolution. (a) Homology-dependent primer-independent DNA shuffling; (b) homology-dependent primerdependent StEP; (c) homology-independent SHIPREC 50 Biocatalysis for the Pharmaceutical Industry high sequence similarity, leaving out the possibility of crossovers in other regions with low similarity. Also, the chimeric genes typically only contain one to four crossover events, which strongly limits the accessible sequence space. This low crossover rate of DNA shuffling was addressed with a method called random chimeragenesis on transient templates (RACHITT) [16]. The parental template in this case contained uracil and was single stranded to serve as a scaffold for hybridization of second-strand fragments from homologous genes. The crossover rate was dramatically increased (around 14 per chimera product), leading to much greater diversity in the library. However, this method is more labor intensive and is subject to optimization of reaction times and temperatures to obtain full-length genes for screening. 3.2.2.2 Homology-Dependent, Primer-Dependent Assembly In contrast to the methods described above, several techniques have been developed that require end-primers or addition of oligonucleotides to encourage frequent crossover between templates. Random priming recombination [17] involves a parental template gene as a scaffold, with random primers used in place of DNase I fragments to anneal randomly on the template and extend to full-length genes. A staggered extension process (StEP, [18]) starts the extension step of shuffling only with end-primers. Instead of extending the gene to full length in one cycle, the process is carried out through very short extension and annealing steps, resulting in template switching throughout the course of full-length gene assembly, thereby producing multiple crossovers (Figure 3.2b). However, the limitation in combining close-proximity mutations is still present. Synthetic shuffling is accomplished without any full-length gene templates, but rather with synthetic degenerate oligonucleotides designed with bioinformatic information on the gene of interest [19]. The oligonucleotides contain overlapping ends that anneal and extend into full-length composite genes. This method opens up the capabilities of introducing site-directed mutations (via semi-synthetic shuffling) or codon optimization into the gene library. Biased mutation-assembling [20] takes wild-type and mutant genes with distinct beneficial mutations, defines blocks within them containing only single mutations, and amplifies the regions with small overlaps to adjacent blocks. The blocks are then reassembled by overlap extension PCR, but are combined in ratios which favor incorporation of mutations which displayed the most improvement. 3.2.2.3 In Vivo Assembly All of the gene shuffling methods discussed so far involve the application of in vitro techniques for generation of the gene library. Another possibility to create low-complexity chimeric libraries is through in vivo recombination methods. In combinatorial libraries enhanced by recombination in yeast [21], parental genes are again randomly fragmented with DNase I and reassembled in a primerless reaction. However, the annealing step is repeated at lower and lower temperatures as the reaction progresses (‘progressive hybridization’), facilitating the annealing of low-homology genes. Amplification of full-length genes is then carried out with primers designed with overlapping regions on the expression vector. The linearized vector and library of genes are then transformed into yeast to promote in vivo homologous recombination. Because there is typically more than one gene hybrid transformed into a yeast cell, recombination results in an increased diversity. Directed Enzyme Evolution and High-Throughput Screening 51 Heteroduplex recombination involves preparing single-stranded DNA from two different homologous genes and mixing them to form heteroduplexes. These heteroduplexes are then transformed into a host that then creates hybrid homoduplexes through in vivo mismatch repair mechanisms [22]. Another technique takes advantage of recA-mediated homologous recombination in a recBC sbcA Escherichia coli mutant [23]. In this method, the parental genes flank a linearized plasmid, and transformation along with in vivo homologous recombination results in single crossover progeny on circularized plasmids. Repeating this process produces one crossover per iteration, and can be modified to include a different parental gene per round. 3.2.2.4 Homology-Independent Assembly As stated earlier, one of the major drawbacks to DNA shuffling and techniques derived from this method is that crossovers occur only at homologous regions. If low-homology parent genes are used in these methods, then the majority of product genes tend to be parental themselves instead of hybrids [24]. In an attempt to dissociate the creation of hybrid enzymes from DNA sequence homology, a technique called incremental truncation for the creation of hybrid enzymes (ITCHY) was created [25]. In this procedure, two parent enzymes are incrementally truncated with exonuclease III under controlled conditions, and the various generated 50 -fragments and 30 -fragments are ligated back together to form a library of chimeric sequences. The resulting enzymes screened are fusions of the amino-terminal portion of one parent and the carboxy-terminal end of the other parent. However, careful optimization and control of the fragment generation is required for a successful library, and makes the method difficult and time consuming. In addition, the fusion enzymes created contain only one crossover region, are not necessarily full-length genes, and can be conjoined at places that are not at structurally related sites [26]. A modified version of this protocol, called THIO-ITCHY, includes the random incorporation of a-phosphothioate nucleotide analogs into the parent genes [27,28]. Exonuclease III activity is inhibited at sites of analog incorporation, which relieves the efforts of producing incremental truncation aliquots. In combination with epPCR, the diversity of the fusion libraries created can be further expanded. Yet another variant of the ITCHY method, fittingly called SCRATCHY, was developed through the combination of ITCHY with DNA shuffling to create multiple crossover libraries independent of homology [27,28]. Further development of this technique produced enhancedcrossover SCRATCHY, in which amplification of ITCHY hybrids is carried out in defined blocks with skewed primers (a forward primer from one parent and a reverse primer from the other). These amplified fragments are then pooled and subjected to DNA shuffling, resulting in a process that selectively enriches hybrids that contain multiple crossovers [29]. Another method that produces hybrid genes independent of homology is sequence homology independent protein recombination (SHIPREC) [30]. Parental genes are first fused together via a linker containing a unique restriction site to form heterodimers. These heterodimers are then randomly digested with DNase I in a controlled reaction. The fragments corresponding to the length of either parental gene are then isolated, blunt-end digested, and ligated to create circularized gene hybrids. These hybrids are then linearized with the unique restriction site introduced originally into the linker (Figure 3.2c). The library is still limited to one crossover event, but is also more likely to contain structurally conserved crossover events. 52 Biocatalysis for the Pharmaceutical Industry 3.2.2.5 Multiple-Parent, Nonhomologous Assembly All of the previous homology-independent shuffling methods are limited by the use of only two parental genes. However, methods to recombine DNA from multiple parents do exist. In cases where gene templates are from eukaryotic sources, exon shuffling can be implemented to interchange domain-encoding exons [31]. In this way, functional domains are connected through design of oligonucleotides used to amplify permitted crossovers, and the random aspect of assembly is avoided. Random multi-recombinant PCR (RM-PCR) is a more generalized scheme of exon shuffling in which reassembly of blocks into full-length genes is possible through overlap-extension PCR, with oligonucleotides containing the crossovers, and can be used for prokaryotic and eukaryotic genes alike [32]. Degenerate homoduplex recombination is similar to synthetic shuffling, and utilizes a set of degenerate top-strand primers that are designed to encompass the diversity information of the genes to be shuffled. Gaps in between these top-strand primers are filled in using dephosphorylated bottom-strand oligonucleotides as templates [33]. Nonhomologous random recombination assembles DNase I-digested fragments that have been blunt-end polished prior to ligation. The addition of hairpin oligonucleotides that cap only the ends of the full-length genes results in preferential ligation of intermolecular pieces [34]. To address limitations with creating frame-shift mutants with premature stop codons, the addition of a chloroamphenicol acetyltransferase (CAT) fusion allowed for preferential selection against truncated or insoluble proteins [35]. There is no single library creation method that will satisfy the ideal requirements of highdiversity quality, large library size, and ultimately the feasibility of comprehensive screening of the generated library. Evaluation of each case is necessary to determine which library creation method is suitable for the projects goals. 3.3 Directed Evolution Library Screening/Selection Methods The options for analyzing a library will be guided in part by the properties of the protein being studied. Whether the target is a binding protein or an enzyme, what substrates or derivatives are available and how these are linked to metabolism will help determine the appropriate screening technique. The analysis involves finding a method that will link the desired activity with the gene variant that encodes it, and can be split into two approaches: screening and selection. In a screening method, variants in the library are assayed individually. For example, by measuring the enzymatic activity in a cell lysate, both active and inactive clones will be discovered. Alternatively, selection methods deal with the entire library at once. For example, by applying antibiotics to bacteria on an agar plate, only resistant clones will grow. A selection method is desirable, since it increases the size of the library that can be practically assayed. The methods used can also be classified as either in vivo (where intact cells are used) or in vitro (where isolated cellular components are utilized). The following will give an overview of these methods, with more details available in published reviews [36,37]. 3.3.1 In Vivo Methods: Genetic Complementation The classic microbiological approach of genetic complementation can be very useful as a directed evolution selection [38,39]. The activity being investigated is intrinsically linked to Directed Enzyme Evolution and High-Throughput Screening 53 the growth of a microbe such as E. coli or Saccharomyces cerevisiae, as it either provides an advantage to the wild-type strain or the library can complement a particular mutant strain. By creating an E. coli strain that was unable to utilize glucose as a carbon source, an overexpression library was screened to discover genes with latent glucokinase activity [40]. When the desired reaction is not essential, creative thinking must be employed. Adding a selectable moiety may create derivatives of a natural substrate that allow for selection. Hwang et al. [41] used this approach to derive a screen for enantioselective hydrolases. By linking the antibiotic chloramphenicol to either the R- or S-enantiomer of 2-phenylbutyric acid, they showed that Exiguobacterium acetylicum could hydrolyse the R-form (as the released chloramphenicol inhibited growth) but not the S-form. However, one must be aware that activity on a derivative may not always correspond to the activity on the actual desired substrate. The degree of selection can be adjusted as needed by the use of different vectors or promoters to alter the expression level of the enzyme within the cell. Alternatively, the stringency of selection can be increased by channeling the substrate away by adding a competing pathway [42]. 3.3.2 In Vivo Methods: Chemical Complementation The drawback of traditional genetic complementation approaches is that they are specific for the gene being studied, and new screens must be devised for each enzyme being investigated. Chemical complementation methods aim to be more general in their approach, by using the substrate of the targeted reaction to influence an unrelated reporter system. This allows one to use well-studied reporters, such as b-galactosidase, or the amino-acid-selectable markers of yeast if they can be linked in some manner to the substrate. This has been accomplished through the use of the yeast three-hybrid assay [43]. Here, the phenotypic readout is the blue color produced by the action of b-galactosidase on X-gal or onitrophenol-b-D-galactopyranoside. The transcription of this reporter gene is determined by the dimerization of a methotrexate-binding DNA binding domain to a dexamethasone-binding activation domain. The two components are linked by a composite molecule containing the substrate of interest, methotrexate, and dexamethasone. If the library variant cleaves the substrate, then there is no dimerization of the transcriptional components, b-galactosidase will not be produced, and there will be no blue color. By changing the reporter gene, this type of screen could be converted to a selection, as was demonstrated with a glycosynthase, where bond formation controlled the transcription of a leucine biosynthetic reporter [44]. Other approaches could use transcriptional regulators that directly bind the substrate or product of the reaction and activate the reporter gene. For instance, a mutant transcriptional activator from Pseudomonas putida, NahH, was used that can bind various benzoic acids to develop a screening/selection method to detect the action of benzaldehyde dehydrogenase [45]. A transcriptional regulator may need to be engineered to bind the desired compound before it can be used in such a manner [46]. 3.3.3 In Vivo Methods: Surface Display The ability to display proteins on the surface of an organism has been exploited as a screening method, typically in conjunction with fluorescence-activated cell sorting (FACS) [37,47,48] 54 Biocatalysis for the Pharmaceutical Industry Figure 3.3 Cell surface display. Proteins displayed on the cell surface of organisms bind a fluorescent molecule. Cells are passed through a FACS machine, allowing separation into populations that do or do not bind the target (Figure 3.3). The approach is particularly well suited for engineering binding proteins such as antibodies, although it has also been applied to enzymes. One of the first widely used surface display methods was through the use of phage [49]. By inserting a target protein into one of the coat proteins of filamentous phage (initially protein pIII, but others such as pVIII have also been used), very large libraries (1010) of variants can be expressed in E. coli and subsequently packaged into active phage. Antibodies are used to enrich the desired variants in vitro, and the phage can then reinfect bacteria in multiple rounds of selection. The selection strategy is useful for increasing the binding affinity of a protein. Stability can also be engineered through the use of proteases, which will cleave a poorly folded protein more rapidly and reduce the infectivity of the phage [50]. Another method is yeast display, which consists of fusing the engineering target to the C-terminus of the Aga2 cell surface agglutinin protein, allowing exposure to fluorescently labeled substrates in the media [51]. Bacterial display systems have also been produced using a number of different scaffold proteins, such as the outer membrane proteins OmpA or OmpX, the thioredoxin protein within the bacterial flagella, or autotransporters such as AIDA-I [37]. Proteins may express differently depending on which scaffold they are fused with, or whether bacteria or yeast are used. The use of yeast as an expression host may be beneficial if the protein requires post-translational modifications. Directed Enzyme Evolution and High-Throughput Screening 55 3.3.4 In Vitro Methods: Lysate Assay Perhaps the most direct way of analyzing a library is to perform the functional assay on the protein itself. Individual clones from cells transformed with the library can be grown and made to express the protein of interest. This can then be isolated and the relevant assay performed. A cell lysate will often contain enough of the protein such that further purification is not necessary for screening purposes. Using microplates increases the throughput of the screening and is useful where the progression of a reaction can be monitored visually, either by the distinct absorbance of the substrate or product or of a coupled reaction. The throughput of this type of screen is comparatively low, on the order of 104, but the approach is flexible and easily implemented. Reaction products from such assays could also be detected directly through the use of gas chromatography, high-performance liquid chromatography, or mass spectrometry. 3.3.5 In Vitro Methods: Ribosome Display Ribosome display is an in vitro display method by which one can link the protein of interest to the gene coding for it. As demonstrated with a single-chain scFv antibody, by using in vitro transcription and translation on DNA lacking a stop codon, a complex consisting of the mRNA, ribosome, and translated protein is generated [52]. The variants of interest are then selected by exposing the complex to the proteins binding partner and washing away nonbinding complexes. The bound complexes are dissociated and the mRNA molecules can be collected and converted back into DNA via reverse transcription, resulting in a pool of sequences enriched for the desired binding activity. A similar approach is used in the technique of mRNA display, but the protein of interest is linked to its mRNA by the use of a puromycin molecule [53,54]. Both techniques have the advantage of being completely in vitro, increasing the potential size of the library (to 1013) and the speed of screening by avoiding transformation steps into cells. 3.3.6 In Vitro Methods: In Vitro Compartmentalization Inspired by the natural linkage of genotype and phenotype observed with cells, Tawfik and Griffiths [55] introduced the idea of in vitro compartmentalization (IVC) by coupling transcription and translation within a water-in-oil emulsion. By forming aqueous droplets of around 2 mm in diameter, genes, their proteins, and more importently substrates and products of the protein are all contained, which allows for the selection of catalytic activities. The concept was initially demonstrated on enzymes that act on DNA, such as HaeIII methyltransferase, where the catalytic activity acted directly on the genotype. The use of antibodies and streptavidin-coated microbeads allowed attachment of a gene, its encoded protein, and biotintagged substrate which was converted to product. An antibody to the product allowed the isolation of improved variants of the enzyme phosphotriesterase through flow cytometry [56]. Extending the emulsion to a water-in-oil-in-water mixture allowed further refinement of the IVC concept. Compartmentalization of E. coli containing serum paraoxonase variants allowed the accumulation of fluorescent product to a point where it could be detected by FACS [57]. This approach was also used with in vitro transcription and translation to evolve b-galactosidase activity from the Ebg gene [58]. 56 Biocatalysis for the Pharmaceutical Industry 3.3.7 Equipment/Automation Directed evolution relies on the analysis of large numbers of clones to enable the discovery of rare variants with improved function. In order to analyze these large libraries, methods of screening or selection have been developed, many of which use specialized equipment or automation. These range from the use of multichannel pipettes, all the way up to robotics, depending on the level of investment [59]. Specialized robotic systems are available to perform tasks such as colony picking, cell culture, protein purification, and cell-based assays. A very versatile piece of equipment that is affordable for individual laboratories is the microplate reader. This allows multiple samples to be analyzed at once, commonly in a 96-well format, although 384- and 1536-well formats are available. Typical measurements that can be performed include UV–Vis absorbance, fluorescence, or luminescence, allowing a range of assays to be performed, such as cell growth, enzyme kinetics, enzyme stability, or enzymelinked immunosorbent assay [60–62]. Functionality can be increased by the use of liquid dispensing systems or automatic plate handling. Flow cytometry is increasingly used in high-throughput screening [63]. Here, a cell solution is pressurized into a stream and directed through various lasers, with information on forward and side scatter and on fluorescence being collected. Based on this data, the cells can be directed into separate pools. The throughput of these machines is up to 40 000 cells per second, making it feasible to screen libraries containing on the order of 109 variants. 3.4 Selected Industrial Examples One of the main caveats that come with the integration of enzymatic processes into industrial settings is that enzymes have evolved through Darwinian evolution for the purpose of survival, not for the purpose of overproducing valuable pharmaceutical products in harsh industrial settings. While an enzyme may have the desired catalytic function for a particular processing step, it most likely will not have sufficient catalytic properties, stability, specificity, or enantioselectivity when simply exchanged with a developed chemical route of synthesis. In addition, the production of the enzyme itself may be economically infeasible due to low expression or solubility. Therefore, there are now many examples of utilizing library creation and screening tools to enhance enzymes for industrial applications. A small sampling of recent examples is listed below for the improvement of various enzymatic functions, with further references for more in-depth study provided for the readers benefit. 3.4.1 Activity The activity of an enzyme is crucial to the productivity of an industrial process. Many pharmaceutically relevant enzymatic processes utilize enzymes with multiple domains that catalyze different reactions. One example is a hybrid, nonribosomal peptide synthase–polyketide synthase (NRPS-PKS) from Pantoea agglomerans [64], which produces the broadspectrum antibiotic andrimid [65,66], an acetyl-CoA carboxylase inhibitor. Based on the proposed pathway for andrimid biosynthesis, the valine-specific A domain of the AdmK protein was replaced with CytC1, a 2-aminobutyrate-incorporating A domain from the Streptomyces sp. RK95-74 cytotrienin NRPS-PKS. The CtyC1 domain was chosen due to its broad specificity profile. While andrimid was still produced with AdmK-CtyC1 hybrid, the Directed Enzyme Evolution and High-Throughput Screening 57 Figure 3.4 Improvement of the activity of chimeric NRPSs using directed evolution. (1) A heterologous A domain is swapped into an NRPS, typically resulting in a significant loss of synthetase activity. (2) A library of chimeric synthetase mutants is constructed in which the heterologous A domain has been diversified (for example, by error-prone PCR). (3) The library is subjected to an in vivo screen for production of the unnatural nonribosomal peptide derivative. (4) Clones showing improved production are characterized and subjected to further rounds of diversification and screening level was reduced by 32-fold compared with the wild-type AdmK. Several rounds of errorprone PCR were performed, and the resulting mutant AdmK-CtyC1 enzyme produced 10.7fold more andrimid than the wild-type AdmK-CtyC1 and only threefold less than the wild-type AdmK. The screen was also extended to test for the increased production of andrimid derivatives produced by the new CtyC1 A domain (Figure 3.4). The results indicated that chimeric NRPSs can be functionally improved, as well as reach activity levels approaching those of the wild-type counterparts [67]. Cephalosporins are a class of antibiotic produced via the intermediate 7-aminocephalosporanic acid (7-ACA), or 7-aminodesacetoxycephalosporanic acid (7-ADCA). Directed evolution has been used to improve the activity of cephalosporin acylases to produce these intermediates from adipyl-7-ACA or cephalosporin C [68]. Using site-directed saturation mutagenesis and a selection system whereby the E. coli host is dependent on leucine liberated from derivatives of the cephalosporin side-chains, a mutant was found that increased the catalytic efficiency toward adipyl-7-ADCA by 36-fold. Oxidoreductases are a family of enzymes that catalyze a number of industrially important reactions, but they often require additional nicotinamide (NADH or NADPH) cofactors which 58 Biocatalysis for the Pharmaceutical Industry are too expensive to supply stoichiometrically. Regeneration of the reduced cofactor form was accomplished by Woodyer et al. [69] by linking an oxidoreductase reaction to phosphite dehydrogenase, an enzyme that oxidizes inert phosphite to phosphate with concomitant reduction of NAD(P) to NAD(P)H. However, the wild-type enzyme had relatively low activity. Through several rounds of random mutagenesis, the activity of the wild-type enzyme was improved by sixfold. The advantage of the wild-type enzyme was demonstrated using the model industrial bioconversion reaction of trimethylpyruvate to L-tert-leucine, an unnatural amino acid derivative. 3.4.2 Thermostability An enzymes thermostability is generally one of the more difficult characteristics to rationally improve. Beneficial mutations that lead to increased thermostability can be involved in many different mechanisms, including solvent interactions, structural support, and electrostatic balance. The majority of attempts to enhance the thermostability of an enzyme have, in some way or another, incorporated random mutagenesis techniques to generate improved enzymes. However, once improved mutants have been screened for and isolated, combining the different mutations together can be labor intensive. Through biased mutation-assembling, which simplified and reduced the workload of incorporating multiple beneficial mutations, Hamamatsu et al. [20] were able to improve the thermostability of a prolyl endopeptidase from Flavobacterium meningosepticum, resulting in a 1200-fold improvement of the enzymes half-life at 60 C. Another example involved a combination of computational data, rational design, and directed evolution. Iterative saturation mutagenesis [70] was used on a model enzyme lipase from Bacillus subtilis to improve its thermostability. Appropriate amino acid residues were chosen based on X-ray data, in particular the B factors (or B values). These factors reflect smearing of atomic electron densities with respect to their equilibrium positions as a result of thermal motion and positional disorder. Amino acids which display large B factors correspond to those which have pronounced flexibility and are targets for stabilization. Each position was subjected to saturation mutagenesis, to generate as many single position libraries as were suitable for the experiment. After screening, the improved mutants found were ranked by position on their contribution to the enzymes improved thermostability. Consecutive rounds of saturation mutagenesis were then performed, first starting with the position of highest improvement, followed by the second highest, and so on until all positions had been explored. The B. subtilis lipase was evolved through this method to two mutants (five and seven total mutations each) that had half-lives of more than 15 h at 55 C, compared with 2 min for the wild-type enzyme. The principle of additive mutational effects played an important role in this design. Also, the selection of residues was limited to having crystal structures available, although the same process can be carried out with template mutants found by other random mutagenesis techniques. 3.4.3 Substrate Specificity The cholesterol-lowering drug atorvastatin, marketed as Lipitor, is an example where biocatalysis research has been applied extensively and is in industrial use. The enzyme 2-deoxyribose5-phosphate aldolase (DERA) has been a target of directed evolution for the production of atorvastatin intermediates [8,9,71]. DeSantis and coworkers [8,9] used structure-based Directed Enzyme Evolution and High-Throughput Screening 59 mutagenesis in an attempt to expand the substrate specificity of the native enzyme. The enzyme variants were fully purified and assayed in vitro by monitoring the decrease of NADH at 340 nm using a spectrophotometer. One variant, S238D, showed new activity towards 3-azidopropinaldehyde to form an azido pyranose which is an intermediate in atorvastatin synthesis. Jennewein et al. [71] also studied this wild-type enzyme, creating a library with error-prone PCR and recombining positive mutants. By screening mutants with a microplate reader or with gas chromatography, they managed to increase the synthesis of the intermediate (3R,5S)-6chloro-2,4,6-trideoxyhexapyranoside by 10-fold. 3.4.4 Product Specificity Polyketides are a structurally diverse group of compounds and one of the richest sources of pharmaceuticals. The biosynthesis of the polyketide avermectin-analog doramectin (commercially sold by Pfizer as Dectomax) in a mutant Streptomyces avermitilis strain was hampered by the product specificity of one of the key intermediate enzymes, aveC, in the doramectin production pathway. Although the particular function of aveC was unknown, the occurrence of undesirable product CHC-B2 was found to be a result of its dual-product specificity. In an effort to improve the production ratio of doramectin:CHC-B2, site-specific and error-prone mutagenesis libraries were screened and resulted in several unrelated beneficial mutations that improved the production ratio up to fourfold [72]. However, much of the screened library had mutations that adversely affected the production ratio of doramectin. Continuing the efforts to improve aveC for doramectin production, a new approach was established to curb the complex and time-consuming initial screens, due in large part to the long growth periods of the S. avermitilis host strain and involved the electrospray ionization tandem mass spectrometry method to detect the resultant products. A semi-synthetic shuffling method [73] was utilized to allow for screening of the diversity of mutations generated in the initial rounds while reducing the library size to be screened. After three rounds of semi-synthetic shuffling, an improved mutant containing 10 mutations was inserted into production strains, and doramectin ratios were found to be improved more than 23-fold over the wild type [74]. 3.4.5 Enantioselectivity The chiral molecule (R)-4-cyano-3-hydroxybutyric acid is another intermediate in the synthesis of atorvastatin (Figure 3.5), and its production enzymatically has been targeted by a number of groups [8,9,75]. DeSantis and coworkers [8,9] used a route involving the hydrolysis of 3-hydroxyglutaryl nitrile by a nitrilase enzyme. GSSM was used to increase the enantioselectivity under high substrate conditions. A chiral substrate was synthesized whereby one nitrile group contained the nitrogen isotope 15 N, meaning that the R- and S-products differed by one mass unit, which was detected by mass spectrometry. Screening of over 30 000 variants indicated the best mutant contained the Ala190His mutation, which increased the enantiomeric excess (ee) to 98.1% from 87.8% and had complete conversion in 15 h as opposed to 24 h when 2.25 M hydroxyglutaryl nitrile was used. Fox et al. [75] used a halohydrin dehydrogenase to produce ethyl (R)-4-cyano-3-hydroxybutyrate, a starting point for atorvastatin synthesis. The enzyme from Agrobacterium radiobacter acts on ethyl (S)-4-chloro-3-hydroxybutyrate, forming the product but initially at a low rate. A statistical model based on protein structure–activity relationships was incorporated, and 18 rounds of screening were carried 60 Biocatalysis for the Pharmaceutical Industry Figure 3.5 Directed evolution applied to enzymes for the production of atorvastatin synthesis intermediates. The substrates and products of the evolved enzymes are shown out. The final mutant showed a 4000-fold improvement in volumetric production, with the ethyl (R)-4-cyano-3-hydroxybutyrate product being 99.5% pure and with an ee of over 99.9%. Epothilones are a class of molecules that show anticancer activity. Production of a synthetic intermediate was investigated through the action of an esterase on various sterically hindered 3-hydroxy esters [76]. No initial activity was observed, so a Pseudomonas fluorescens esterase was transformed into a mutator strain Epicurian coli and screened using an indicator in the growth plates that would produce a red color if hydrolysis occurred. An ee of 25% was achieved from a variant containing two mutations. 3.5 Conclusions and Future Directions The complexity of todays pharmaceutical compounds and an increasing awareness of the environmental impact of traditional chemical syntheses have opened the door to biocatalysis. Directed evolution is an integral tool in the development of synthetic enzymes, ensuring they are suitable for use in an industrial setting. The past success of this approach indicates that it will continue to provide many examples of safe and efficient production of chemical intermediates and medical compounds. Directed Enzyme Evolution and High-Throughput Screening 61 References [1] Sen, S., Venkata Dasu, V. and Mandal, B. (2007) Developments in directed evolution for improving enzyme functions. Applied Biochemistry and Biotechnology, 143, 212–223. [2] Yuan, L., Kurek, I., English, J. and Keenan, R. (2005) Laboratory-directed protein evolution. Microbiology and Molecular Biology Reviews, 69, 373–392. [3] Roodveldt, C., Aharoni, A. and Tawfik, D.S. (2005) Directed evolution of proteins for heterologous expression and stability. Current Opinion in Structural Biology, 15, 50–56. [4] Hibbert, E.G., Baganz, F., Hailes, H.C. et al. (2005) Directed evolution of biocatalytic processes. Biomolecular Engineering, 22, 11–19. [5] Leung, D.W., Chen, E. and Goeddel, D.V. (1989) A method for random mutagenesis of a defined DNA segment using a modified polymerase chain reaction. Technique, 1, 11–15. [6] Daugherty, P.S., Chen, G., Iverson, B.L. and Georgiou, G. (2000) Quantitative analysis of the effect of the mutation frequency on the affinity maturation of single chain Fv antibodies. Proceedings of the National Academy of Sciences of the United States of America, 97, 2029–2034. [7] Kunichika, K., Hashimoto, Y. and Imoto, T. (2002) Robustness of hen lysozyme monitored by random mutations. Protein Engineering, 15, 805–809. [8] DeSantis, G., Liu, J., Clark, D.P. et al. (2003) Structure-based mutagenesis approaches toward expanding the substrate specificity of D-2-deoxyribose-5-phosphate aldolase. Bioorganic and Medicinal Chemistry, 11, 43–52. [9] DeSantis, G., Wong, K., Farwell, B. et al. (2003) Creation of a productive, highly enantioselective nitrilase through gene site saturation mutagenesis (GSSM). Journal of the American Chemical Society, 125, 11476–11477. [10] Wong, T.S., Tee, K.L., Hauer, B. and Schwaneberg, U. (2004) Sequence saturation mutagenesis (SeSaM): a novel method for directed evolution. Nucleic Acids Research, 32, e26. [11] Seng Wong, T., Roccatano, D., Loakes, D. et al. (2007) Transversion-enriched sequence saturation mutagenesis (SeSaM-Tv þ ): a random mutagenesis method with consecutive nucleotide exchanges that complements the bias of error-prone PCR. Biotechnology Journal, 3, 74–82. [12] Hallet, B., Sherratt, D.J. and Hayes, F. (1997) Pentapeptide scanning mutagenesis: random insertion of a variable five amino acid cassette in a target protein. Nucleic Acids Research, 25, 1866–1867. [13] Murakami, H., Hohsaka, T. and Sisido, M. (2002) Random insertion and deletion of arbitrary number of bases for codon-based random mutation of DNAs. Nature Biotechnology, 20, 76–81. [14] Stemmer, W.P. (1994) DNA shuffling by random fragmentation and reassembly: in vitro recombination for molecular evolution. Proceedings of the National Academy of Sciences of the United States of America, 91, 10747–10751. [15] Crameri, A., Raillard, S.A., Bermudez, E. and Stemmer, W.P. (1998) DNA shuffling of a family of genes from diverse species accelerates directed evolution. Nature, 391, 288–291. [16] Coco, W.M. (2003) RACHITT: gene family shuffling by random chimeragenesis on transient templates. Methods in Molecular Biology (Clifton, NJ), 231, 111–127. [17] Shao, Z., Zhao, H., Giver, L. and Arnold, F.H. (1998) Random-priming in vitro recombination: an effective tool for directed evolution. Nucleic Acids Research, 26, 681–683. [18] Zhao, H., Giver, L., Shao, Z. et al. (1998) Molecular evolution by staggered extension process (StEP) in vitro recombination. Nature Biotechnology, 16, 258–261. [19] Ness, J.E., Kim, S., Gottman, A. et al. (2002) Synthetic shuffling expands functional protein diversity by allowing amino acids to recombine independently. Nature Biotechnology, 20, 1251–1255. [20] Hamamatsu, N., Aita, T., Nomiya, Y. et al. (2005) Biased mutation-assembling: an efficient method for rapid directed evolution through simultaneous mutation accumulation. Protein Engineering Design & Selection, 18, 265–271. [21] Abecassis, V., Pompon, D. and Truan, G. (2000) High efficiency family shuffling based on multi-step PCR and in vivo DNA recombination in yeast: statistical and functional analysis of a combinatorial library between human cytochrome P450 1A1 and 1A2. Nucleic Acids Research, 28, E88. [22] Volkov, A.A., Shao, Z. and Arnold, F.H. (1999) Recombination and chimeragenesis by in vitro heteroduplex formation and in vivo repair. Nucleic Acids Research, 27, e18. [23] Xu, S., Ju, J., Misono, H. and Ohnishi, K. (2006) Directed evolution of extradiol dioxygenase by a novel in vivo DNA shuffling. Gene, 368, 126–137. 62 Biocatalysis for the Pharmaceutical Industry [24] Joern, J.M., Meinhold, P. and Arnold, F.H. (2002) Analysis of shuffled gene libraries. Journal of Molecular Biology, 316, 643–656. [25] Ostermeier, M., Shim, J.H. and Benkovic, S.J. (1999) A combinatorial approach to hybrid enzymes independent of DNA homology. Nature Biotechnology, 17, 1205–1209. [26] Sieber, V., Martinez, C.A. and Arnold, F.H. (2001) Libraries of hybrid proteins from distantly related sequences. Nature Biotechnology, 19, 456–460. [27] Lutz, S., Ostermeier, M. and Benkovic, S.J. (2001) Rapid generation of incremental truncation libraries for protein engineering using alpha-phosphothioate nucleotides. Nucleic Acids Research, 29, E16. [28] Lutz, S., Ostermeier, M., Moore, G.L. et al. (2001) Creating multiple-crossover DNA libraries independent of sequence identity. Proceedings of the National Academy of Sciences of the United States of America, 98, 11248–11253. [29] Kawarasaki, Y., Griswold, K.E., Stevenson, J.D. et al. (2003) Enhanced crossover SCRATCHY: construction and high-throughput screening of a combinatorial library containing multiple non-homologous crossovers. Nucleic Acids Research, 31, e126. [30] Udit, A.K., Silberg, J.J. and Sieber, V. (2003) Sequence homology-independent protein recombination (SHIPREC). Methods in Molecular Biology (Clifton, NJ), 231, 153–163. [31] Kolkman, J.A. and Stemmer, W.P. (2001) Directed evolution of proteins by exon shuffling. Nature Biotechnology, 19, 423–428. [32] Tsuji, T., Onimaru, M. and Yanagawa, H. (2001) Random multi-recombinant PCR for the construction of combinatorial protein libraries. Nucleic Acids Research, 29, E97. [33] Coco, W.M., Encell, L.P., Levinson, W.E. et al. (2002) Growth factor engineering by degenerate homoduplex gene family recombination. Nature Biotechnology, 20, 1246–1250. [34] Bittker, J.A., Le, B.V. and Liu, D.R. (2002) Nucleic acid evolution and minimization by nonhomologous random recombination. Nature Biotechnology, 20, 1024–1029. [35] Bittker, J.A., Le, B.V., Liu, J.M. and Liu, D.R. (2004) Directed evolution of protein enzymes using nonhomologous random recombination. Proceedings of the National Academy of Sciences of the United States of America, 101, 7011–7016. [36] Boersma, Y.L., Droge, M.J. and Quax, W.J. (2007) Selection strategies for improved biocatalysts. FEBS Journal, 274, 2181–2195. [37] Daugherty, P.S. (2007) Protein engineering with bacterial display. Current Opinion in Structural Biology, 17, 474–480. [38] Francis, J.C. and Hansche, P.E. (1972) Directed evolution of metabolic pathways in microbial populations. I. Modification of the acid phosphatase pH optimum in S. cerevisiae. Genetics, 70, 59–73. [39] Hall, B.G. (1981) Changes in the substrate specificities of an enzyme during directed evolution of new functions. Biochemistry, 20, 4042–4049. [40] Miller, B.G. and Raines, R.T. (2004) Identifying latent enzyme activities: substrate ambiguity within modern bacterial sugar kinases. Biochemistry, 43, 6387–6392. [41] Hwang, B.Y., Oh, J.M., Kim, J. and Kim, B.G. (2006) Pro-antibiotic substrates for the identification of enantioselective hydrolases. Biotechnology Letters, 28, 1181–1185. [42] Kleeb, A.C., Edalat, M.H., Gamper, M. et al. (2007) Metabolic engineering of a genetic selection system with tunable stringency. Proceedings of the National Academy of Sciences of the United States of America, 104, 13907–13912. [43] Baker, K., Bleczinski, C., Lin, H. et al. (2002) Chemical complementation: a reaction-independent genetic assay for enzyme catalysis. Proceedings of the National Academy of Sciences of the United States of America, 99, 16537–16542. [44] Lin, H., Tao, H. and Cornish, V.W. (2004) Directed evolution of a glycosynthase via chemical complementation. Journal of the American Chemical Society, 126, 15051–15059. [45] Van Sint Fiet, S., van Beilen, J.B. and Witholt, B. (2006) Selection of biocatalysts for chemical synthesis. Proceedings of the National Academy of Sciences of the United States of America, 103, 1693–1698. [46] Mohn, W.W., Garmendia, J., Galvao, T.C. and de Lorenzo, V. (2006) Surveying biotransformations with a la carte genetic traps: translating dehydrochlorination of lindane (gamma-hexachlorocyclohexane) into lacZ-based phenotypes. Environmental Microbiology, 8, 546–555. [47] Becker, S., Schmoldt, H.U., Adams, T.M. et al. (2004) Ultra-high-throughput screening based on cell-surface display and fluorescence-activated cell sorting for the identification of novel biocatalysts. Current Opinion in Biotechnology, 15, 323–329. Directed Enzyme Evolution and High-Throughput Screening 63 [48] Gai, S.A. and Wittrup, K.D. (2007) Yeast surface display for protein engineering and characterization. Current Opinion in Structural Biology, 17, 467–473. [49] Smith, G.P. (1985) Filamentous fusion phage: novel expression vectors that display cloned antigens on the virion surface. Science, 228, 1315–1317. [50] Paschke, M. (2006) Phage display systems and their applications. Applied Microbiology and Biotechnology, 70, 2–11. [51] Boder, E.T. and Wittrup, K.D. (1997) Yeast surface display for screening combinatorial polypeptide libraries. Nature Biotechnology, 15, 553–557. [52] Hanes, J. and Pluckthun, A. (1997) In vitro selection and evolution of functional proteins by using ribosome display. Proceedings of the National Academy of Sciences of the United States of America, 94, 4937–4942. [53] Nemoto, N., Miyamoto-Sato, E., Husimi, Y. and Yanagawa, H. (1997) In vitro virus: bonding of mRNA bearing puromycin at the 30 -terminal end to the C-terminal end of its encoded protein on the ribosome in vitro. FEBS Letters, 414, 405–408. [54] Roberts, R.W. and Szostak, J.W. (1997) RNA-peptide fusions for the in vitro selection of peptides and proteins. Proceedings of the National Academy of Sciences of the United States of America, 94, 12297–12302. [55] Tawfik, D.S. and Griffiths, A.D. (1998) Man-made cell-like compartments for molecular evolution. Nature Biotechnology, 16, 652–656. [56] Griffiths, A.D. and Tawfik, D.S. (2003) Directed evolution of an extremely fast phosphotriesterase by in vitro compartmentalization. The EMBO Journal, 22, 24–35. [57] Aharoni, A., Amitai, G., Bernath, K. et al. (2005) High-throughput screening of enzyme libraries: thiolactonases evolved by fluorescence-activated sorting of single cells in emulsion compartments. Chemistry & Biology, 12, 1281–1289. [58] Mastrobattista, E., Taly, V., Chanudet, E. et al. (2005) High-throughput screening of enzyme libraries: in vitro evolution of a beta-galactosidase by fluorescence-activated sorting of double emulsions. Chemistry & Biology, 12, 1291–1300. [59] Blow, N. (2008) Lab automation: tales along the road to automation. Nature Methods, 5, 109–112. [60] Aucamp, J.P., Cosme, A.M., Lye, G.J. and Dalby, P.A. (2005) High-throughput measurement of protein stability in microtiter plates. Biotechnology and Bioengineering, 89, 599–607. [61] Geddie, M.L., Rowe, L.A., Alexander, O.B. and Matsumura, I. (2004) High throughput microplate screens for directed protein evolution. Methods in Enzymology, 388, 134–145. [62] Huang, K.S., Mark, D. and Gandenberger, F.U. (2006) High-throughput measurements of biochemical responses using the plate::vision multimode 96 minilens array reader. Methods in Enzymology, 414, 589–600. [63] Ibrahim, S.F. and van den Engh, G. (2007) Flow cytometry and cell sorting. Advances in Biochemical Engineering/Biotechnology, 106, 19–39. [64] Jin, M., Fischbach, M.A. and Clardy, J. (2006) A biosynthetic gene cluster for the acetyl-CoA carboxylase inhibitor andrimid. Journal of the American Chemical Society, 128, 10660–10661. [65] Fischbach, M.A. and Walsh, C.T. (2006) Assembly-line enzymology for polyketide and nonribosomal peptide antibiotics: logic, machinery, and mechanisms. Chemical Reviews, 106, 3468–3496. [66] Pohlmann, J., Lampe, T., Shimada, M. et al. (2005) Pyrrolidinedione derivatives as antibacterial agents with a novel mode of action. Bioorganic & Medicinal Chemistry Letters, 15, 1189–1192. [67] Fischbach, M.A., Lai, J.R., Roche, E.D. et al. (2007) Directed evolution can rapidly improve the activity of chimeric assembly-line enzymes. Proceedings of the National Academy of Sciences of the United States of America, 104, 11951–11956. [68] Otten, L.G., Sio, C.F., Reis, C.R. et al. (2007) A highly active adipyl-cephalosporin acylase obtained via rational randomization. FEBS Journal, 274, 5600–5610. [69] Woodyer, R., van der Donk, W.A. and Zhao, H.M. (2006) Optimizing a biocatalyst for improved NAD(P)H regeneration: directed evolution of phosphite dehydrogenase. Combinatorial Chemistry & High Throughput Screening, 9, 237–245. [70] Reetz, M.T., Carballeira, J.D. and Vogel, A. (2006) Iterative saturation mutagenesis on the basis of B factors as a strategy for increasing protein thermostability. Angewandte Chemie (International Edition in English), 45, 7745–7751. [71] Jennewein, S., Schurmann, M., Wolberg, M. et al. (2006) Directed evolution of an industrial biocatalyst: 2-deoxyD-ribose 5-phosphate aldolase. Biotechnology Journal, 1, 537–548. [72] Stutzman-Engwall, K., Conlon, S., Fedechko, R. et al. (2003) Engineering the aveC gene to enhance the ratio of doramectin to its CHC-B2 analogue produced in Streptomyces avermitilis. Biotechnology and Bioengineering, 82, 359–369. 64 Biocatalysis for the Pharmaceutical Industry [73] Crameri, A., Stemmer, W.P.C., Minshull, J. et al. (2000) Oligonucleotide mediated nucleic acid recombination. PCT International Patent Application WO 00/42561. [74] Stutzman-Engwall, K., Conlon, S., Fedechko, R. et al. (2005) Semi-synthetic DNA shuffling of aveC leads to improved industrial scale production of doramectin by Streptomyces avermitilis. Metabolic Engineering, 7, 27–37. [75] Fox, R.J., Davis, S.C., Mundorff, E.C. et al. (2007) Improving catalytic function by ProSAR-driven enzyme evolution. Nature Biotechnology, 25, 338–344. [76] Bornscheuer, U.T., Altenbuchner, J. and Meyer, H.H. (1998) Directed evolution of an esterase for the stereoselective resolution of a key intermediate in the synthesis of epothilones. Biotechnology and Bioengineering, 58, 554–559.