DIRECTED ENZYME EVOLUTION
Frances H. Arnold

INTRODUCTION

Staunch Darwinists attribute all the complexity of living things to an algorithm of mutation and natural selection. The exquisite products of this evolution algorithm are apparent at all levels, from the amazing diversity of life all the way down to individual protein molecules. Scientists and engineers who wish to redesign these same molecules are now implementing their own versions of the algorithm. Directed evolution allows us to explore enzyme functions never required in the natural environment and for which the molecular basis is poorly understood. This bottom-up design approach contrasts with the more conventional, top-down one in which proteins are tamed `rationally' using computers and site-directed mutagenesis. I will describe how molecular evolution can be directed in the test tube in order to produce useful biocatalysts. It is not possible to provide a complete account of all the methods proposed for in vitro evolution; I will therefore introduce methods and strategies used successfully in our laboratory for directing the evolution of enzymes. Some alternative methods applied to biocatalyst evolution in other laboratories will also be discussed. The reader should also be aware that there is a rather substantial and largely separate literature on combinatorial approaches to engineering binding molecules. Recent advances in the ability to create genetic diversity and to screen or select for improved functions in large libraries of enzyme variants are being combined in a robust approach to solving difficult molecular design problems. With directed evolution we now have the ability to tailor individual proteins as well as whole biosynthetic and biodegradation pathways for biotechnology applications.

Why directed evolution?

The limitations of nature's biocatalysts present insurmountable challenges to rational design
When natural enzymes are recruited for industrial applications--from serving as catalysts in chemicals synthesis to additives for laundry detergents--we discover that they are often not well suited to these tasks. Due to poor substrate solubility, breakdown of unstable products, or competing chemical reactions, the conditions for an enzyme reaction may be unsuitable for large-scale applications. Reflecting their participation in complex biochemical networks inside living cells, enzymes are often inhibited by their own substrates or products, either of which may severely limit the productivity of a biocatalytic process. Evolution is usually the culprit: enzymes are optimized and often highly specialized for specific biological functions within the context of a living organism. Biotechnology, in contrast, needs enzymes which are stable and active over long periods of time (a feature that might clash with the need for rapid protein turnover inside a cell), enzymes which are active in nonaqueous solvents (a feature probably not required in most biological milieu), and enzymes which can accept different substrates (substrates not present in nature).

It is possible to produce new enzymes in recombinant organisms, altering the amino acid sequence and therefore the properties through appropriate modifications at the DNA level. We are hampered, however, by near complete ignorance of how the amino acid sequence affects every aspect of enzyme performance, from its ability to be expressed in a heterologous host to its catalytic activity in nonnatural environments. Numerous protein engineering experiments have demonstrated that changes in protein properties are brought about by the cumulative effects of many small adjustments, many of which are distributed or propagated over significant distances. Furthermore, proteins are usually teetering on the brink of instability, with folded structures that are more stable than unfolded--and therefore inactive--ones by the equivalent of a few hydrogen bonds out of the hundreds that form. Superimposing upon the need to retain this relatively fragile folded state the additional requirements of having to fold in the first place and of maintaining or even reengineering a catalytic site that is affected at some level by virtually any modification yields a design problem of such complexity that any rational design effort will require enormous inputs of structural, mechanistic, and dynamic information. Information that is available for but a tiny fraction of interesting catalysts. Even if one trait is successfully designed (e.g. enhanced stability), it is virtually impossible to predict the cost to another (e.g. catalytic activity or expression level). The relatively few examples where 'rational' design has yielded useful enzymes do not negate the fact that rational protein design is often a fruitless exercise.

Extending natural diversity by laboratory evolution
All these hurdles to the rational design of enzymes are bypassed by evolution. The power of evolution as an algorithm for molecular design is perhaps best appreciated by studying its products. By constructing the evolutionary histories of today's proteins we have learned that they are highly adaptable molecules, at least on evolutionary time scales. Well illustrated by the panoply of α/β barrel enzymes [Holm & Sander, 1997], enzymes catalyzing very different reactions have evolved divergently from a common ancestral protein of the same general structure, acquiring diverse capabilities by processes of random mutation, recombination, and natural selection. We also know that enzymes sharing a common function (for example, all catalyzing a particular step in a metabolic pathway) in addition to three-dimensional structure can exhibit widely different properties (stability, solubility, tolerance to pH, etc.) depending on where they are found.

Enzymes evolve, and adapt, at the molecular level. The structures of protein modules are conserved (although the modules themselves are often shuffled to create new, multifunctional proteins). Function, however, can vary. Specific features such as substrate specificity or thermostability vary significantly. Amino acid sequences can vary to such an extent that evolutionary relationships may no longer be apparent from sequences alone [Holm & Sander, 1997].

Evolution is a powerful algorithm with proven ability to alter enzyme function and especially to `tune' enzyme properties. It is also an algorithm that can be implemented in the laboratory for redesign. The challenge is to collapse the time scale to months, or even weeks.

Choosing an evolutionary strategy
Evolutionary mechanisms at work in nature assure adaptability to ever changing environments. Evolution does not work towards any particular direction, nor is there a goal; the underlying processes occur spontaneously during reproduction and survival of the whole organism. In contrast, a directed molecular evolution experiment has a defined goal, and the key processes--mutation, recombination and screening or selection--are controlled by the experimenter. Although there may be multiple ways to reach a defined goal (i.e. a desired enzyme function), the approach that minimizes the effort is preferred.

The major steps in a typical directed enzyme evolution experiment are outlined in Figure 1. The genetic diversity for evolution is created by mutagenesis and/or recombination of one or more parent sequences. These altered genes are cloned back into a plasmid for expression in a suitable host organism (bacteria or yeast). Clones expressing improved enzymes are identified in a high-throughput screen, or in some cases, by selection, and the gene(s) encoding those improved enzymes are isolated and recycled to the next round of directed evolution. Approaches for carrying out these key steps will be discussed in some detail in the Methods section. Here we will focus on more fundamental considerations that help to define workable strategies for directed evolution.

Figure 1. Key steps of a typical directed enzyme evolution experiment.

To appreciate the challenge of designing and carrying out a successful directed protein evolution experiment, it is important to underscore the powerful combinatorial features of this system. A typical enzyme is a linear chain of N amino acids (N is usually several hundred), and there are 20 possible amino acids at each position in the chain. Thus the `sequence space' of possible proteins is huge beyond the imagination (20 ^N ). Even in 4 billion years, nature has had a chance to explore but a tiny fraction of these possibilities. A laboratory exploration of this vast space of sequences and their corresponding functions must obviously be severely limited and carefully guided [Arnold, 1998a]. Because much of sequence space will be devoid of the desired function and probably even folded proteins, it is best to direct the evolution of one (or more) existing enzymes rather than look for function in random peptide libraries.

Evolution is often referred to as a hill-climbing exercise in the fitness landscape of sequence space [Eigen, 1986; Kauffman, 1993]. The fitnesses (performance, for laboratory evolution) of the proteins in sequence space make up this landscape, whose most basic features are still quite unknown. The landscape for laboratory evolution will be different for each property or collection of properties undergoing evolution. I have argued [Arnold, 1996; Arnold, 1998a] that an uphill climb in a landscape is more likely to be successful if it can take place in small steps (one or two amino acid substitutions). The high dimensionality of the surface (there are 19N one-mutant neighbors of any given sequence) offers many opportunities to find improved mutants. While we may never reach the 'global optimum', the improvements achieved by taking even a simple random up-hill walk via single amino acid mutations often yields useful results. A widely effective evolutionary strategy, illustrated in Figure 2, is one in which the steps are small (preferably single amino acid substitutions in each generation) and multiple such mutations are accumulated, either sequentially or by recombination [Stemmer, 1994; Stemmer, 1997], to acquire the desired function [Arnold, 1996; Arnold, 1998a].Such an approach is compatible with a low level of random mutagenesis over the entire gene. An alternative approach is to direct a much higher level of random mutation to a relatively small region of the gene [Black et al., 1996]. Both approaches have their advantages. Mutagenesis over the entire gene allows discovery of unanticipated solutions (a common experience). More intense, directed mutation, however, may yield novel combinations of amino acid substitutions, combinations that would be inaccessible by single step walks because the intermediates are unfavorable. Details of the evolutionary exploration will ultimately be dictated by a combination of 1) the power of the search tool, 2) the frequency of beneficial mutations (usually small!), and 3) the choice of starting point(s).

Figure 2. An effective strategy for directed enzyme evolution is one in which small changes associated with 1-2 amino acid substitutions are accumulated sequentially (a) or by recombination of improved genes (b).

Yet another approach to creating diversity for directed evolution is the in vitro shuffling of homologous genes, or "family shuffling" [Crameri et al., 1998], illustrated in Figure 3. Here, recombination of two or more parent genes yields a chimeric gene library for evolution of the desired features. Because the recombined sequences are related through divergent evolution from a common ancestor of similar structure and function--and therefore the sequence differences are to some extent neutral with respect to structure and function--it appears that very large jumps in sequence space can yield functional proteins [Crameri et al ., 1998; Arnold, 1998b]. In vivo recombination can also yield interesting new chimeric enzymes [van Kampen et al., 1998; Okkels, J. S., 1997; Gray, 1992], but the diversity may be more limited than that obtained by in vitro methods.

Figure 3. Homologous proteins are descended by divergent evolution from a common ancestral protein and share its overall 3D structure. Recombining homologous genes creates chimeras, some fraction of which should also fold into that structure. Such `family shuffled' libraries could be rich in novel function.

Good problems for directed enzyme evolution
There are four requirements for successful directed evolution. 1) The desired function must be physically possible. 2) The function must also be biologically, or evolutionarily, feasible. In practice, this means that there exists a mutational pathway to get from here to there through ever-improving variants (see above). While we cannot know a priori that the path exists, a good experiment will maximize the likelihood. 3) You must be able to make libraries of mutants complex enough to contain rare beneficial mutations. This usually means functional expression in a suitable microorganism such as E. coli or S. cerevisiae . 4) You must have a rapid screen or selection that reflects the desired function. Just how rapid the screen must be depends on how rare mutations leading to the desired property are and how many must be accumulated to achieve the desired result.

Whether directed evolution will solve a particular problem depends to some extent on how hard natural evolution has already worked at it. If a particular trait is already under selective pressure (e.g. catalytic activity on the biologically relevant substrate), it is unlikely that further improvements can be obtained in the laboratory by small mutational steps. However, if biological function has imposed additional constraints, for example the trait is coupled to another trait that is also under selective pressure (e.g. high thermostability), then this balance can be altered during laboratory evolution [Giver et al., 1998]. While selected traits are often difficult to improve, they should be relatively easy to remove (e.g. product inhibition). Many traits are not under selective pressure; they may be changing as a result of random genetic drift, they may be vestigial--reflecting the enzyme's history, or they may be coupled to selected traits [Benner, 1989]. In general, nonselected traits are easier to improve, but may be more difficult to remove. It is especially easy to improve traits never required for biological function, such as stability or activity in a nonnatural environment or activity towards a new substrate, where there is often much room for improvement and small changes in sequence and function can be accumulated. As expected for any hill-climbing exercise, the number of pathways leading uphill diminishes as a peak is approached. Thus the ease with which improved mutants are identified (the frequency of improved clones) should eventually decrease as the sequences move closer to an optimum.

The preceding discussion focused only on altering existing enzyme traits. 'Evolving' a completely new function is a risky venture, because we rarely know how far in sequence space we have to go in order to create the new function or how frequent the solutions will be. If there is good reason to believe that a new function, for example activity towards a substrate not accepted at all by the wild type enzyme, can be obtained (at a level measurable during a high throughput screen) by making 1-2 amino acid substitutions, then evolution is feasible. However, if the new function requires the simultaneous placement of multiple new amino acid residues, it is unlikely to appear in a random library of mutants. Such a problem is probably a better candidate for a combination of rational design and combinatorial tuning. 

Table I summarizes the key features of selected directed enzyme evolution experiments. In nearly all cases, the desired trait(s) was at a measurable, albeit low, level in the starting enzyme(s). The problems can be roughly divided into a few major categories: improving function in nonnatural or extreme environments (where activities or stabilities are low), improving activity towards a new substrate, tuning specificity (enantioselectivity), and increasing functional expression in a heterologous host.

Acknowledgments
F. A. wishes to thank the many talented undergraduate, graduate and postdoctoral students who have contributed to directed evolution research at Caltech. The financial support of the Office of Naval Research, the Department of Energy, and the Army Research Office is gratefully acknowledged.

References
Akanuma, S., Yamagushi, A., Tanaka, N. and Ooshima, T.Protein Sci.7, 698-705 (1998).
Allen, S. J. and Holbrook, J. J., Protein Eng. 13 , 5-7 (2000).
Arnold, F. H. Acct. Chem. Research 31, 125-131 (1998a).
Arnold, F. H. Nat. Biotechnol.16, 617-618 (1998b).
Arnold, F. H. Chem. Eng. Sci.51, 5091-5102 (1996).
Arnold, F. H. and Chen, K. Subtilisin Variants Suitable for Hydrolysis and Synthesis in Organic Media" U.S. Patent 5,316,935. (1994).
Arnold, F. H. and Moore, J. C. Para-Nitrobenzyl Esterases with Enhanced Activity in Aqueous and Nonaqueous Media, U.S. Patents 5,741,691 (1998) and 5,906,930 (1999).
Benner, S. A. Chem Rev. 89, 789-806 (1989).
Black, M. E., Newcomb, T. G., Wilson, H. M. P. and Loeb, L.A. Proc. Natl. Acad. Sci. USA, 93, 3525-3529 (1996).
Bornscheuer, U.T., Altenbuchner, J., Meyer, H. H. Biotechnol Bioengineering 58, 554-55 (1998).
Bruhlmann, F. and Chen, W. Biotechnol. Bioeng., 63 , 544-551 (1999).
Buchholz, F., Angrand, P. O. and Steward, A. F. Nat. Biotechnol. 16, 657-662 (1998).
Chen, K. and Arnold, F. H. Proc. Natl. Acad. Sci. USA 90 , 5618-5622 (1993).
Cherry, J. R., Lamsa, M. H., Schneider, P., Vind, J., Svendsen, A., Jones, A., and Pedersen, A. H. Nature Biotechnology, 17, 379-384 (1999).
Cheong, T. K. and Oriel, P. J. Enzyme & Microbial Technology 26, 152-158 (2000).
Christians, F. C. and Loeb, L. A. Proc. Natl. Acad. Sci. USA 93 , 6124-6128 (1996).
Crameri A., Raillard S. A., Bermudez E. and Stemmer W. P. C . Nature391,288-291 (1998).
Crameri, A., Dawes, G., Rodrigues, E. Jr, , Silver, S. and Stemmer, W. P. C. Nat. Biotechnol.15, 436-438 (1997).
Eigen, M. Chemica Scripta26B, 13-26 (1986).
Giver, L., Gershenson, A., Freskgard, P. O. and Arnold, F. H., Proc. Natl. Acad. Sci. USA 95, 12809-12813 (1998).
Gray, G. L. Hybrid prokaryotic polypeptides produced by in vivo homologous recombination. US Patent 5,093,257 (1992).
Hansson, L.O., Bolton-Grob, R., Massoud, T. and Mannervik, B. J. Mol. Biology (1999).
Holm, L. and Sander, C. Proteins:Struct. Funct. Genet . 28, 72-82 (1997).
Hoseki, J., Yano, T., Koyama, Y., Kuramitsu, S. and Kagamiyama, H. J. Biochem. 126, 951-956 (1999).
Joo, H., Lin, Z. and Arnold, F. H. Nature399, 670-673 (1999).
Kauffman, S. A.The Origins of Order, Oxford University Press, Inc., 1993.
Kumamaru, T., Suenaga, H., Mitsuoka, M., Watanabe, T. and Furukawa, K. Nat. Biotechnol.16, 663-666 (1998).
Lanio, T., Jeltsch, A. and Pingoud, A. J. Mol. Biol.283 , 59-69 (1998).
Li, Q. S.., Schwaneberg, U., Fischer, P., and Schmid, R. D., Chem. Eur. J.6, 1531-1536 (2000).
Liao, H., McKenzie, T. and Hageman, R. Proc. Natl. Acad. Sci. USA 83, 576-580 (1986).
Liu, D. R., Magliery, T. J., Pasternk, M. and Schultz, P.G. Proc Natl Acd Sci USA94, 10092-10097 (1997).
MacBeath, G., Kast, P. and Hilvert, D. Science 279 , 1958-1961 (1998).
Matsumura, I., Wallingford, J. B., Surana, N. K., Vize, P. D., and A. D., Nat. Biotechnol. 17, 696-701 (1999).
Matsumura, I. and Ellington, A. D., Protein Science 8 , 731-740 (1999).
May, O., Nguyen, P. and Arnold, F. H., Nat. Biotechnol. 18 , 317-320 (2000).
Miyazaki, K. and Arnold, F. H., J. Mol. Evolution, 49 , 716-720 (2000).
Moore, J. C. and Arnold, F. H. Nat. Biotechnol.14 , 458-467 (1996).
Moore, J. C., Jin, H. M., Kuchner, O. and Arnold, F. H. J. Mol. Biol. 272, 336-347 (1997).
Naki, D., Paech, C. Ganshaw, G. and Schellenberger, V. Appl. Nicrobiol. Biotechnol. 49, 290-294 (1998).
Okkels, J. S. (1997) Method for preparing polypeptide variants. PCT application WO 97/07205.
Okkels J. S., Thellersen M., Petersen D.A., Svendsen A., Patkar S. A., and Borch K. (1997) Novel lypolytic enzymes. PCT application WO 97/07202.
Oue, S., Okamoto, Yano, A., T., and Kagamiyama, H., J.Biol Chem. 274, 2344-2349 (1999).
Reetz M. T., Zonta, A., Schimossek, K., Liebeton, K. and Jaeger, K. E. Angew. Chem.Int. Ed. Engl. 36,2830-2832 (1997).
Schmidt-Dannert, C., Umeno, D. and Arnold, F. H., Nat. Biotechnol ., 18, 750-753 (2000).
Song, J. K. and Rhee, J. S., Appl. Environ. Microbiol . 66 , 890-894 (2000).
Stemmer, W. P. C. Nature370, 389-391 (1994).
Stemmer, W. P. C. Methods for in vitro recombination. US Patent 5,605,793 (1997).
Strausberg, S. L., Alexander, P. A., Gallagher, D. T., Gilliland, G. L., Barnett, B. L. and Bryan, P. N. Biotechnology13 , 669-673 (1995).
Taguchi, S., Ozaki, A. and Momose, H. Appl. and Envir. Microbiol . 64, 492-495 (1998).
van Kampen, M. D., Dekker, N., Egmond, M. and Verheij, H. Biochemistry 37:3459-3466 (1998).
Wan, L., Twitchett, M. B., Eltis, L. D., Mauk, A. G., and Smith, M. PNAS 95, 12825-12831 (1998).
Yano, T., Oue, S. and Kagamiyama, H. Proc. Natl. Acad. Sci. USA 95, 5511-5515 (1997).
You, L. and Arnold, F. H. Protein Eng.9, 77-83 (1996).
Zaccolo, M. and Gherardi, E., J. Mol. Biology285, 775-783 (1999).
Zhang, J-H., Dawes, G. and Stemmer, W. P. C. Proc. Natl. Acad. Sci. USA94, 4504-4509 (1997).
Zhao, H.and Arnold, F. H., Protein Eng. 12, 47-53 (1999).
Zhao, H., Giver, L., Shao, Z., Affholter, J. A. and Arnold, F. H. Nat. Biotechnol. 16,258-235 (1998).

Target enzyme	Target function	Change effected	Approach	Organism	Reference
kanamycin nucleotidyltransferase	thermostability	>200-fold increase in half-life at 60-65 oC	mutator strain + selection	B. stearothermophilus	Liao et al. (1986)
subtilisin E	activity in organic solvents	~ 170-fold increase in 60% dimethylformamide	error-prone PCR + screening	B. subtilis	Chen & Arnold (1993); Arnold & Chen (1994)
β-lactamase	activity towards new substrate	32,000-fold greater resistance to cefotaxime	DNA shuffling + selection	E. coli	Stemmer (1994)
subtilisin BPN'	stability in the absence of Ca2+	1000-fold increase in half-life	loop removed +cassette mutagenesis + screening	B. subtilis	Strausberg et al. (1995)
para-nitrobenzyl esterase	activity towards pNB esters; activity in organic solvent	60-150 fold increase	error-prone PCR and DNA shuffling + screening	E. coli	Moore & Arnold (1996); Moore et al. (1997); Arnold & Moore (1998)
thymidine kinase	substrate specificity (gene therapy)	43-fold increase in sensitivity to gancyclovir in hamster cells	cassette mutagenesis + selection and screening	E. coli	Black et al. (1996)
β-galactosidase	activity towards new substrate; substrate specificity	66-fold increased activity; 1000-fold increase in substrate specificity	DNA shuffling + screening	E. coli	Zhang et al. (1997)
subtilisin E	expression level; activity in organic solvents	500-fold increase in total activity	error-prone PCR + screening	B. subtilis	You & Arnold (1996)
O6-alkylguanine-DNA alkyltransferase	protection against alkylating agents (gene therapy)	10-fold increased protection against toxic methylating agent	cassette mutagenesis+selection	E. coli	Christians & Loeb (1996)
arsenate detoxification pathway	arsenic resistance	12-fold increased rate of arsenate reduction	DNA shuffling + screening	E. coli	Crameri et al. (1997)

Target enzyme	Target function	Change effected	Approach	Organism	Reference
aminoacyl-tRNA synthetase	aminoacylation of a modified tRNA	55-fold increase in activity	DNA shuffling + selection	E. coli	Liu et al. (1997)
aspartate aminotransferase	activity towards β -branched amino and 2-oxo acids	105 increase in activity	DNA shuffling + selection	E. coli	Yano et al. (1997)
lipase	wash performance	improved performance in one-cycle wash	mutagenesis and in vivo recombination + screening	S. cerevisiae	Okkels et al. (1997)
lipase	enantioselectivity in hydrolysis of p-nitrophenyl 2-methyldecanoate	increase in enantiomeric excess from 2% to 81%	error-prone PCR + screening	E. coli	Reetz et al. (1997)
lipases	activity towards long-chain p-nitrophenyl esters	3-fold increase	in vivo recombination of homologous genes + screening	E. coli	van Kampen et al. (1998)
pNB esterase	thermostability	14 oC increase in Tm + increased activity at all temperatures	error-prone PCR, DNA shuffling + screening	E. coli	Giver et al. (1998)
esterase	enantioselectivity of hydrolysis of a sterically hindered 3-hydroxy ester	increase in enantiomeric excess from 0% to 25%	mutator strain + selection	E. coli	Bornscheuer et al. (1998)
subtilisin E	thermostability	17 oC increase in Tm + increased activity at all temperatures	error-prone PCR, DNA shuffling + screening	B. subtilis	Zhao & Arnold (1999)
subtilisin E	thermostability	50-fold  increase in half-life at 65 oC	DNA shuffling + screening	B. subtilis	Zhao et al. (1998)

Target enzyme	Target function	Change effected	Approach	Organism	Reference
B. lentus subtilisin	expression level (total activity of secreted enzyme)	50% increase	error-prone PCR + enrichment in hollow fibers	B. subtilis	Naki et al. (1998)
subtilisin BPN’	activity at 10oC	2-fold increase	chemical mutagenesis + screening	B. subtilis, E. coli	Taguchi et al. (1998)
3-isopropylmalate dehydrogenase	thermostability	3.4-fold increase in activity at 70 oC	spontaneous mutations + selection	T. thermophilus	Akanuma et al. (1998)
cephalosporinases	activity towards moxalactam	270-540-fold increased resistance	DNA shuffling of homologous genes + selection	E. coli	Crameri et al. (1998)
chorismate mutase	conversion to monomeric enzyme (solubility)	functional monomeric enzyme	oligonucleotide directed codon mutagenesis + selection	E. coli	MacBeath et al. (1998)
biphenyl dioxygenases	degradation of polychlorinated biphenyls (PCBs)	gained activity towards substrates poorly degraded by native enzymes, improved activity towards various substrates	DNA shuffling of homologous genes + screening	E. coli	Kumamaru et al. (1998)
FLP recombinase	in vivo recombination efficiency at elevated temperatures in E.coli and mammalian cells; in vitro thermostability	improved recombination efficiency in E. coli and mammalian cells	error-prone PCR and DNA shuffling + screening	E. coli	Buchholz et al. (1998)
EcoRV endonuclease	extend recognition site	becomes 10 bp cutter	targeted random mutagenesis + screening	E. coli	Lanio et al. (1998)
cytochrome P450cam	increased activity in peroxide shunt pathway, towards naphthalene	5-20-fold increase	error-prone PCR+ StEP shuffling + screening	E. coli	Joo et al. (1999)
myoglobin	peroxidase activity	25-fold increase	error-prone PCR + screening	E. coli	Wan et al. (1998)
aspartate aminotransferase	substrate specificity	2.1x106-fold increase in cat. efficiency toward valine, 17 mutations	DNA shuffling + selection	E. coli	Oue et al. (1999)
TEM-1 β-lactamase	activity towards cefotaxime	20,000 fold increase	high frequency random mutagenesis	E. coli	Zaccolo et al. (1999)
glutathione transferase	substrate specificity	found range of specificities	DNA (family) shuffling+screening	E. coli	Hansson et al. (1999)
biphenyl dioxgenase	extended substrate range	yes	error-prone PCR + Shuffling (StEP)	E. coli	Bruhlmann et al. (1999)
Coprinus cinereus peroxidase	stability to peroxide, thermostability	110-fold greater thermal stability, 2.8-fold oxidative stability	mutagenesis, in vivo recombination + site-directed mutants	yeast	Cherry et al. (1999)
beta-glucuronidase	retention of function after glutaraldehyde cross-linking	more resistant to glutaraldehyde and formaldehyde	error-prone PCR and DNA shuffling + screening	E. coli	Matsumura et al. (1999)
catalase I of B. stearothermophilus	stability	increased a little	random elongation mutagenesis	E. coli	Matsumura et al. (1999)
subtilisin S41	improved thermostability	100-fold increase in half life	error-prone PCR + saturation mutagenesis and screening	B. subtilis	Miyazaki & Arnold (1999)
hydantoinase	enantioselectivity + total activity	inverted enantioselectivity, 3 x increase in total activity	error-prone PCR + screening	E. coli	May et al. (2000)
subtilisins	various properties	improved activity stability	DNA "family" shuffling + screening	B. subtilis	Ness et al. (1999)
esterase	enantioselectivity	more selective	error-prone PCR + screening	E. coli	Henke & Bornscheuer (1999)
thymidine kinase	substrate specificity	7-44 fold improved specificity	DNA "family" shuffling + screening	E. coli	Christians et al. (1999)
catechol 2,3-dioxygenases	thermostability	13-26 x more thermostable	DNA "family" shuffling + screening	E. coli	Kikuchi et al. (1999)
B. stearothermophilus lactate dehydrogenase	remove fructose 1,6 bisphosphate requirement (FBP)	active without FBP	random mutagenesis + screening	E. coli	Allen & Holbrook (2000)
kanamycin nucleotidyl transferase	thermostability	increase 20oC	DNA shuffling + screening/selection	T. thermophilus	Hoseki et al. (1999)
B. stearothermophilus amidase	increase expression in E. coli	increased expression (23x)	error prone PCR + screening	E. coli	Cheong & Oriel (2000)
phospholipase A1	thermostability	increase Tm by 11oC without compromising activity	error prone PCR + screening	E. coli	Song & Rhee (2000)
horseradish peroxidase	increase activity/expression in S. cerevisiae	increase total activity 40 x	error prone PCR + screening	S. cerevisiae	Morawski et al. (2000)
L-2-hydroxyacid dehydrogenase	co-factor (fructose 1,6 bisphosphate) requirement	70-fold activation witout cofactor (fully active in the absence of cofactor)	DNA shuffling + screening	E. coli	Allen et al. (2000)
phytoene desaturase and lycopene cyclase	new carotenoid pathyway (substrate and reaction specificity)	bacteria synthesize new carotenoids	DNA family shuffling + screening	E. coli	Schmidt-Dannert et al. (2000)
cytochrome P450 BM3	substrate specificities	hydroxylates indole	saturation mutagenesis + screening	E. coli	Li et al. (2000)