No sequence homology required High percentage of folded proteins Maintains structural domains Can use varied parent genes
Combines useful mutations and loses harmful mutations Can use multiple parent genes to increase the diversity and fitness of a library Crossovers occur in regions of short (0-5 bases) sequence homology Single hybridization event reduces the mismatching sometimes seen in PCR-based methods
May reduce parental sequence reproduction and favor the formation of chimeras
Comparable diversity to other DNA-shuffling formats Simple, single-tube method with no fragment purification
Useful for screening proteins expressed in yeast
Creates hybrid library of two unrelated genes
No sequence homology required so junctions are randomly distributed Researchers preselect full-length hybrids
Same pros as for SHIPREC
Researchers preselect functional hybrids with auxotrophic E. coli
Same pros as for ITCHY
More efficient and easier than ITCHY
Restricted to intron-containing genes (not applicable [S0]
to prokaryotic genes)
Relies on sequence homology 
Incomplete degradation by X exonuclease affects quality , 
of ssDNA substrate, which is critical Additional steps require upfront phagemid cloning and preparation of ss gene fragments and template DNA Additional steps in phage compared with family shuffling ,  May be unnecessary with most proteins or in experiments with effective screens Relies on sequence homology 
Project-specific optimization of PCR conditions can be time-consuming and limit robustness of the method Additional steps 
No advantage to in vitro family shuffling unless proteins are to be screened in yeast Only one crossover per hybrid per round (low diversity  library, but may iterate or combine with homologous recombination methods to improve crossovers) Limited to two parents of equal length Low-fitness hybrids (two thirds may contain frame-shift) May induce aa deletions or duplications at junctions Same cons as for SHIPREC 
Parent gene length is not conserved in hybrids
dNTP analogs remain in hybrid and may interfere with other DNA-binding proteins
Note: Methods 1-6 exploit homologous recombination; Methods 7-9 are based on ligation. ssDNA : Adapted from Kurtzman, A.L. et al., Curr. Opin. Biotechnol., 12, 361-370, 2001.
version of ITCHY was performed via timed exonuclease digestions, and proved difficult to optimize; it subsequently led to the development of THIO-ITCHY, in which initial templates are created with phosphothioate linkages, incorporated at random points along the length of the gene [86,87]. More recently, a nonhomologous recombination technique, termed SCRATCHY, which combines features of ITCHY and DNA shuffling, has been described [88,89]. Although these techniques do not require much detailed knowledge of the protein structure, such information can be utilized in the design of crossover points, as illustrated by the recently described technique known as structural-based combinatorial protein engineering (SCOPE) .
Compared with oligonucleotide-based or in vivo random mutagenesis, the key advantage of in vitro recombination methods is the ability to accumulate beneficial mutations while simultaneously removing deleterious mutations [27,33,81]. Since the description of the first DNA recombination-based techniques, a multitude of techniques based on the recombination model have been developed, with many industrial and academic researchers striving to improve or modify the basic techniques. Some excellent reviews of new patented and widely available techniques that have emerged in recent years have been reported [37,57,58], including techniques based on oligonucleotide-driven randomization, whole-gene randomization, and homology-dependent and homology-independent recombination techniques. Also, numerous reports of computational methods seek to complement laboratory-based techniques by identifying useful sequence space to be probed during directed evolution experiments. This concept of in silico design and screening has proven very useful in reducing the redundancy of enzyme libraries and enabling more useful sequence space to be explored. Some examples include computer programs that identify specific aa residues in a protein more tolerant to mutagenesis , and others such as Xencor's patented protein design automation (PDA®) technology [92,93]. It is difficult to be comprehensive, because new techniques are constantly emerging; for more in-depth reviews of the growing area of computational methods to aid protein engineering experiments, see [31,36,52,53,57,58,94] and references therein.
It is also worth noting that the various recombination-based DNA shuffling techniques are not just limited to creating diversity in single protein structures. DNA shuffling has been used to adapt whole organisms to excel under foreign conditions associated with commercial process conditions , to evolve murine leukemia viruses with broader cell tropisms , to create retroviruses with improved stability and yields under processing conditions , in the evolution of new metabolic pathways , and in the de novo molecular evolution of an Archael DNA fragment with unknown function (lacking ampi-cillin resistance) to confer ampicillin resistance on an E. coli strain .
The previous section briefly outlined some of the numerous methods that are available for the first step of any directed evolution experiment — creating the initial genetic diversity. Perhaps the most challenging step in any directed evolution experiment is finding those proteins that perform the desired function according to specified criteria among the created library of mutant sequences . In parallel with the development of methods to create diversity, numerous technological advances have been made in HTS to enable rapid screening of libraries of mutant proteins.
Some of the key features associated with any effective screening method are that they (1) should be high throughput (to enable detection of useful variants), (2) are highly sensitive to the desired function, (3) are reproducible, (4) are robust, and (5) enable easy detection . Typically, screening an enzyme library involves analysis of variants that are arranged on microtiter plates (usually 96- or 384-well plates), on Petri dishes, or by other means and are then analyzed by functional assays [38,100]. HTS can be performed with whole cells, cell lysates, or partially purified proteins. Some recent innovative developments in enzyme screening and assay methods are reviewed elsewhere [38,101-109].
Alternatively, selection of an improved enzyme involves coupling the desired trait to those variants possessing the specific trait. Selections offer the advantage of isolating functional proteins from very large libraries (frequently > 107 clones) simply by growing a population of cells under selective conditions. However, the expressed enzyme must confer a significant biological advantage, which limits the range of enzymes and properties that can be assessed [100,110]. In vivo selection protocols are the most convenient but require that the expressed enzyme directly confers a significant biological advantage [38,100,111]. Consequently, numerous in vitro selection protocols have been developed which directly link the phenotype and genotype. Phage display [112-114] remains the most common technique for in vitro selection, while other tools, such as water-in-oil in vitro compartmentaliza-tion, have emerged as very powerful alternative selection tools linking phenotype with genotype for the directed evolution of enzyme properties [115-118]. Comprehensive reviews of recent developments in in vitro selection techniques have been reported [31,38,101,111].
2.2.3 Combining Rational Design and Directed Evolution
The two principal strategies employed for genetic manipulation of an enzyme's primary aa sequence, rational design and directed evolution, which approaches the goal of engineering from different perspectives, are by no means mutually exclusive. In recent years, a number of papers have appeared in the literature which employ rational design and directed evolution strategies in combination, particularly in cases where a certain amount of knowledge of the enzyme's structure is known, or where a region of the protein structure is known to affect a particular property such as substrate binding for instance.
One of the first examples illustrating the power of using this combined approach used directed evolution to augment a rational design attempt to engineer a novel function into an a/p-barrel enzyme by completely converting the activity of indole-3-glycerol-phos-phate synthase (IGPS) into that of an efficient phosphoribosyl-anthranilate isomerase (PRAI). Structure-based design was employed to modify the IGPS a/p-barrel by incorporating the basic design of the PRAI loop system, yielding a hybrid variant with very low PRAI activity. This variant served as a scaffold for subsequent directed evolution engineering using DNA shuffling and StEP. Genetic selection (agar plate-based) was then employed to select a variant with increased PRAI activity. The final engineered variant obtained displayed six-fold higher activity than wild-type PRAI and had no IGPS activity, while retaining 28% identity to PRAI and 90% identity to IGPS .
Since then, an ever-increasing number of reports combining rational design and directed evolution approaches to alter enzyme performance have appeared in the literature. These have included combined approaches to improve overall catalytic activity [119-121], altering the substrate [122,123] and product specificity  of various enzymes, enhanced thermostability [125-129], and enhanced resistance to natural enzyme inhibitors  (see Section 2.3). It has been predicted that future enzyme engineering strategies will employ a combination of both directed evolution and rational design strategies [25-29,35,39,41,125-127,131,132].
Manipulations of protein properties at the genetic level, by directed evolution and rational design techniques, have led to dramatic advances in engineering enzymes. Although overshadowed to some extent by these advances, both physical and chemical means of modifying enzymes still remain very useful strategies for the optimization of enzyme properties, as does the addition of stabilizers. Physical modification, for instance through immobilization onto solid supports, has proven very effective at enhancing enzyme stability (some examples are listed in Section 2.3.2). One major goal of chemical modification is to increase enzyme stability and activity (Sections 2.3.1 and 2.3.2), and has proven very effective in achieving these goals, particularly in nonaqueous environments (Section 188.8.131.52). As this review focuses primarily on recent advances in enzyme engineering by manipulation at the genetic level, the interested reader is referred to some recent reviews on chemical modification of proteins [50,133-138]. For a review of physical and chemical approaches to engineering lipases, which are widely used in the detergent industry and increasingly so in the production of optically pure compounds, see [139,140].
Properties of enzymes that have been improved include increased catalytic performance (increase activity), thermostability and stability in unusual environments (e.g., organic solvents or extremes of pH), altered selectivity/specificity, evolution of new activities, susceptibility to proteases, and solubility. The following section highlights recent examples in some of these areas.
Improving the catalytic efficiency of an enzyme is a common goal in protein engineering. Rational design, directed evolution, and other techniques have been used successfully to improve biocatalyst activity and turnover properties, and some examples are given below.
Notable rational design successes in improving catalytic turnover have been reported (reviewed in [25,29,100]). For instance, the turnover of superoxide dismutase (one of the fastest known enzymes) was improved using rational design approaches , while the turnover of papain for nitrile hydratase was increased 104-fold relative to the wild-type by replacement of an active site residue (Q19E) . Knowledge of aa residues involved in catalysis can aid the design of new, improved active sites by means of rational design. For example, deacetoxycephalosporin C synthase (DAOCS) from Streptomyces clavuligerus catalyzes the ring expansion of penicillin to form cephalosporin, and has potential industrial uses in synthesizing new cephalosporin antibiotics. The C-terminus was known to be critical to DAOCS activity. Chin et al.  explored the role of asparagine residue 304 (N304) by substituting N304 with other aas. One mutant, N304L, displayed increased activity for penicillin and penicillin analogs ranging from 130 to 420%, relative to the wildtype. Other mutants such as N304K and N304R exhibited even more improved activity (up to 730% that of the wild-type). Replacement of N304 with aas bearing aliphatic or basic side-chains is thought to incorporate favorable hydrophobic or charged interactions between the variant enzymes and their substrates, thus enhancing enzyme activity .
Directed evolution techniques have also proven adept at enhancing enzyme activity (reviewed in [14,32,36]). Directed evolution studies have frequently concluded that beneficial mutations do not necessarily occur in the enzyme's active site, and that residues outside the active site can play important roles in determining catalytic activity . For example, amylosucrase from Neisseria polysaccharea is a glucosyltransferase that catalyzes the formation of glucose polymers from sucrose . Unique among enzymes belonging to its class, amylosucrase can produce an amylase-like glucan consisting solely of a-1,4-
linked glucose residues. Unlike other similar enzymes, amylosucrase does not require the addition of expensive activated sugars. It has numerous potential industrial applications as a polysaccharide-synthesizing or -modifying agent. Its potential has been limited by its low catalytic efficiency on sucrose, low stability, and side reactions producing sucrose isomers. Using ep-PCR, DNA shuffling, and appropriate selection, Van der Veen et al.  generated two amylosucrase variants with vastly improved properties more suited to industrial synthesis conditions compared to the wild-type enzyme. The specific activities were over five-fold greater than the wild-type, and both displayed greatly improved catalytic efficiency. Both variants had two aa substitutions outside the active site of the enzyme, which would have been difficult to predict using a rational design approach.
Some impressive examples of using directed evolution and high-throughput screening strategies to improve catalytic turnover have been reported. Using ep-PCR and DNA shuffling, a variant of the Pseudomonas diminuta MG organophosphohydrolase, a nerve agent degrader, was selected using cell surface display and a colorimetric assay. The variant showed a 25-fold increase in turnover . Selection of a large mutant library by in vitro compartmentalization yielded a phosphotriesterase variant exhibiting a 63-fold higher kcat than the parent enzyme .
Rational design and directed evolution have been combined to improve catalytic activity. A five-fold increase in the activity of a carbon-carbon bond-forming side reaction of benzoylformate decarboxylase was reported using both strategies . Ep-PCR generated a mutant (L476G), with a five-fold greater carboligase activity than the parent. Saturation mutagenesis was then applied to Leu476 that yielded an additional eight variants with both improved catalytic activity and enantioselectivity than the parent.
Enzymes are increasingly used for a wide range of industrial applications (reviewed in [4,5,9,14,15,32]). The increasing use of biocatalysts in industry is dependent on developments that improve enzyme performance under process conditions. Enzymes have evolved to catalyze specific reactions under certain environmental conditions. From an industrial point of view, many enzymes are not readily amenable to industrial-scale use, due to the extremes of temperature, pressure, pH, and other conditions used in such processes. Thus, the ability to engineer increased stability into enzymes such that they can be readily applied to industrial processes is desirable.
Many industrial processes occur at elevated temperatures. Enzyme thermostability is of major industrial importance, as the use of biocatalysts in industry is dependent, in many instances, on the enzyme's ability to tolerate or function at elevated temperatures. Intensive efforts have been made in academic and industrial laboratories in recent years to improve enzyme thermostability (reviewed in [41,50,110,127,128,146]). Strategies used to improve protein thermostability include directed evolution approaches, rational design methods, consensus sequence approaches, and chemical modification and addition of stabilizers. The following section highlights some of the more recent successes reported using these techniques.
Studies of hyperstable enzymes detected in extreme environmental conditions have revealed that enzymes have utilized numerous strategies to evolve thermostability. A multitude of long-range and local interactions determine enzyme stability at elevated temperatures [127,146]. Two important general stabilizing strategies have been observed in hyperstable enzymes: (1) the presence of large surface networks of electrostatic interactions and (2) a tendency of hyperstable enzymes to be multimeric. In practice, however, such properties are difficult to engineer into a protein.
Attempts to design enzyme thermostability rationally have utilized several approaches with some successes, including mutations aimed at reducing the entropy of unfolded proteins (by introducing disulfide bridges or by increasing Pro residues), increasing the number of a-helices through Gly^Ala substitutions, and engineering of salt bridges [30,41,127]. Table 2.2 highlights some of the successes achieved using rational design principles to improve enzyme stability at elevated temperatures. Despite these successes, no simple set of rules for predicting thermostability have been devised as yet, thereby limiting the ability of rational design to engineer thermostability.
Directed evolution techniques have proven very adept at modifying the thermostability of commercially important enzymes. The advances in techniques for generating diversity (Section 184.108.40.206), coupled with advances in high-throughput screening for thermostability [14,147,148], has led to the development of many biocatalysts with improved thermostability (reviewed in [12,32,41,50]). For instance, a variant of subtilisin, widely used as a detergent protease, was developed displaying a 1200-fold increase in half-life at 60°C [149,150]. A peroxidase developed for use in the detergent industry was improved by directed evolution to display a 174-fold increase in thermostability at 50°C . Adaptation of enzymes to colder temperatures using directed evolution have been reported [152-154].
Many of the earlier attempts at improving enzyme thermostability using directed evolution or rational design strategies came with a concomitant reduction in catalytic activity. One explanation for this trade-off is that during evolution, enzymes have adjusted the strength and number of stabilizing interactions to optimize the balance between rigidity (for stability) and flexibility (for activity) . Some recent examples illustrate that it is now possible to alter thermostability without any detrimental effect on catalytic activity (Table 2.3).
Lehmann et al.  and Lehmann and Wyss  have introduced a semirational strategy, the consensus approach, for engineering thermostability into proteins. This approach hypothesizes that at a given position in a protein sequence, conserved aa residues contribute more than average to stability than nonconsensus aa residues. Thus, substitution of the nonconserved aas may be a feasible course of action for increasing thermostability. Using this approach, an improved phytase, a commercially important animal feed additive, was designed . By aligning homologous sequences of 13 wild-type phytases from fungal sources, key residues were identified that were thought to contribute to thermostability. A synthetic consensus phytase was constructed from this data that was 15 to 26°C more stable than all the parent phytases. In addition, gains in thermostability did not compromise the catalytic activity of the phytase. Other examples of utilizing the consensus approach have been reported, highlighting the potential of using this approach as an additional tool for engineering and understanding thermostability in enzymes [155,156].
Manipulation of protein properties at the genetic level by directed evolution, rational design, and other techniques, has led to dramatic advances in engineering enzyme ther-mostability, as outlined previously. Although overshadowed by these advances, chemical modification remains a very useful technique, as does the technique of addition of stabilizers (reviewed in ). Protein structures can be stabilized by forming intramolecular crosslinks, such that the internal crosslink can prevent protein unfolding under stressful conditions. One form of this involves crosslinking small protein crystals with gluteralde-hyde. These crosslinked enzyme crystals (CLECs) display increased operational and storage stabilities relative to untreated enzymes. Glucoamylase hydrolyzes starch and is used extensively in the starch, glucose, and fermentation industries. Typically, immobilized glucoamylase preparations are used to reduce operation costs and downtime, but have the disadvantage of reduced thermostability over nonimmobilized forms. To overcome these
Was this article helpful?