Proteins are the chemical workhorses of living organisms, acting as sensors, performing regulatory functions, and catalysing reactions. This functional diversity is being harnessed for a growing number of applications in industry and medicine. Ever more powerful engineering strategies are emerging that allow proteins to be customised in the laboratory for carrying out new tasks, further expanding their utility.
by Dr Kenneth J. Woycechowsky
PROTEIN ENGINEERING: MOTIVATIONS AND CHALLENGES
For functions such as binding to a target molecule or promoting a chemical reaction, proteins are tough to beat. Consequently, proteins have found use as drugs, analytical tools, additives in consumer goods, reagents for food processing, catalysts for organic synthesis and components of many other processes and products. However, proteins are not necessarily useful outside their native biological contexts. Therefore, some tinkering is often required.
The past 20 years have seen tremendous advances in our ability to manipulate protein sequences; the trick is to identify which manipulations will lead to a desired function. This problem can be likened to finding a needle in a haystack, where the haystack (i.e., the number of possible amino acid sequences) is astronomically large. For example, there are ~10130 (20100) ways to make a 100 residue protein using the set of 20 standard amino acids. It is simply not feasible to sample this much sequence diversity. Thus, protein engineers must decide which small slice of sequence space to explore in the quest for desirable variants. This article aims to summarise recent advances in the development of strategies for protein engineering, with an emphasis on the generation of new catalytic activities. The large leaps in activity that have been attained suggest that the usefulness and accessibility of customised proteins for science, industry and medicine will continue to grow.
SITE-DIRECTED MUTAGENESIS
Modern molecular biology techniques have made routine the use of synthetic oligonucleotides to substitute one codon for another in a DNA sequence as a means to introduce defined changes into the encoded amino acid sequence. This type of site-directed mutagenesis can be combined with knowledge of structure-function relationships to design point mutations that redirect the properties of proteins. The conversion of lactate dehydrogenase into a malate dehydrogenase [1] provides a powerful example of what can be accomplished with only a single point mutation, as the replacement of a glutamine by an arginine caused a complete inversion of substrate specificity. More recently, alanine racemase was converted into an aldolase via a single tyrosine-to-alanine mutation [2]; such a catalyst could be a useful tool for organic synthesis. Additional point mutations can further extend the range of accessible activities, and a steadily growing list of examples attests to the power of site-directed mutagenesis to tailor the properties of proteins [3]. Still, this type of rational redesign via point mutations requires a blend of structural information, mechanistic understanding and chemical intuition that varies on a case-by-case basis. The approaches described below are potentially more generic.
DIRECTED EVOLUTION
Evolution has provided the diverse array of proteins found in nature. The process of evolution involves the introduction of (random) mutations followed by the selection and amplification of new sequences with beneficial properties. The functional scope of proteins can be further expanded by directed evolution, which mimics natural evolutionary processes in the laboratory. Directed evolution [Figure 1] has been particularly successful for applications such as improving stability, changing specificity or stereoselectivity, and optimising activity under non-physiological conditions [4]. Typically, large collections (libraries) of DNA encoding protein variants that contain random mutations are constructed using error-prone PCR, DNA shuffling or cassette mutagenesis. Upon transcription and translation, variants displaying desirable properties are identified from the resulting protein libraries either through selection, in which parallel processing allows only active variants to appear, or through high-throughput screening, in which a library is deconvoluted and all variants are examined one at a time. The number of variants evaluated in a single round of laboratory evolution usually range from 103 to 1013, and the subset of mutants that meets the standards specified by the assay can be amplified by either cell growth or PCR techniques. It is therefore crucial that a link be maintained between the protein function and the genetic information encoding it. A variety of methods have been developed to establish this linkage, from co-compartmentalisation (by transformation into cells or the formation of water-in-oil emulsions) to forging direct physical connections between a protein and the genetic material that encodes it (using phage display, ribosome display or mRNA display).
Directed evolution involves iterations of the mutagenesis-selection-amplification cycle. Each cycle usually yields only small improvements, but these improvements can add up to give large effects after many iterations. One important application of directed evolution has been in the engineering of proteases as additives for detergents [5]. By screening large libraries of protease variants for activity under appropriate conditions (high temperature and detergent concentration), novel enzymes were generated that are commercially successful because of their high catalytic activities inside washing machines. One important obstacle for directed evolution is obtaining an initial measurable level of activity that can be used as the basis for further improvements. Recent reports of the conversion of glyoxalase II into an antibiotic degrading enzyme [6] and of the installation of RNA ligase activity into a previously inactive zinc finger protein [7] should help to widen the scope of directed evolution to include both the establishment of new activities and the optimisation of function. The development of improved selection systems and high-throughput screens with greater dynamic ranges is a particularly important challenge for maximising the effectiveness of directed evolution as a strategy for engineering new and improved proteins.
COMPUTATIONAL DESIGN
Ultimately, much of protein engineering will become automated. Such advances will entail the development of both robotic systems to facilitate the construction of protein variants and computational methods to identify which mutations to make [Figure 2]. The field of computational design has made particularly impressive strides in recent years. In general, computational methods amount to virtual screening of very large amino acid sequence libraries (often >1020 variants). Detailed structure predictions are made for each member of these libraries and then a calculation is made of the energetics associated with a specific trait (structural stability, ligand binding). The output of such a computational design program is a list of amino acid sequences ranked by their associated energies. If it works, only one or a few of the top-ranked sequences need to be physically produced and tested. The Rosetta program developed by David Baker’s group at the University of Washington has provided a steady stream of breakthroughs in recent years, including the stabilisation of an enzyme and the redesign of a protein-protein interface that maintain high affinity and high specificity.
These successes have also extended to catalyst design, yielding novel enzymes from initially inactive precursors. For this application, several residues that define a new potential active site intended to stabilise a given transition state are grafted onto a protein in silico, the mutation of neighbouring residues is then simulated, the structures of the resulting virtual variants are predicted and interaction energies are then calculated. In one case, aldolase activity was generated with a 104-fold rate enhancement over background, and the accuracy of the predicted structures was verified by X-ray crystallography for two active variants [8]. Nonetheless, the designed aldolases are not very efficient, displaying kcat/Km values that are several orders of magnitude lower than those of natural enzymes. The combination of computational design and directed evolution could provide a route for engineering proteins with new and optimised properties. Indeed, another example of computational enzyme design led to weakly active catalysts for the Kemp elimination, a non-biological proton transfer reaction [9], and the outcomes of the computational design were then used as starting points for directed evolution. Seven rounds of evolution led to a >200-fold increase in activity, demonstrating how rational and random elements can be productively combined in the design process.
One of the great advantages provided by computational design is that this strategy is, in principle, completely generic. While nature will continue to provide useful precedents and starting points for protein engineering, the limitations imposed by biology in terms of the structures and functions accessible for design are being removed. One big step in this direction was the design of a protein with a non-natural fold [10]. This feat was accomplished by first constraining the polypeptide backbone to a specified, but unprecedented, topology and then searching for amino acid sequences that would stably occupy that structure. The output of the computational design was then synthesised and found to adopt the predicted structure with high stability and atomic-level accuracy. These recent advances pave the way for the engineering of proteins with virtually any structure for virtually any function. However, computational design is still far from a routine and completely reliable procedure. Important future improvements will include decreasing the protracted computational times currently required, more accurately predicting the electrostatics of buried residues and developing better descriptions of backbone motions.
PERSPECTIVES
Natural evolutionary processes have yielded a set of proteins that provide insights, inspiration and starting points for generating customised variants. Mutagenesis techniques, high-throughput screening and selection systems, and computational methods have combined to improve our ability to sift through different sequences and identify new proteins with interesting and useful properties. In the future, the scope of protein engineering strategies will likely expand to include the incorporation of nonstandard amino acids side chains and alternative backbone structures (such as β-peptides). The combination of computational design, directed evolution and chemical intuition should continue to spur rapid progress in the field, ushering in an era where the design and production of customised proteins that act as sensors, circuits, therapeutics and industrial catalysts is routine.
REFERENCES
1. Wilks HM, et al. Science 1988; 242: 1541-1544.
2. Seebeck FP, Hilvert D. J. Am Chem Soc 2005; 125: 10158-10159.
3. Toscano MD, et al. Angew Chem Int Ed 2007; 46: 3212-3236.
4. Kuchner O, Arnold FH. Trends Biotechnol 1997; 15: 523-530.
5. Cherry JR, Fidantsef AL. Curr Opin Biotechnol 2003; 14: 438-443.
6. Park H-S, et al. Science 2006; 311: 535-538.
7. Seelig B, Szostak JW. Nature 2007; 448: 828-831.
8. Jiang L, et al. Science 2008; 319: 1387-1391.
9. Röthlisberger D, et al. Nature 2008; 453: 190-195.
10. Kuhlman B, et al. Science 2003; 302: 1364-1368.
THE AUTHOR
Kenneth J. Woycechowsky, Ph.D.
Laboratory of Organic Chemistry,
ETH Zürich,
Hönggerberg HCI F330,
CH-8093 Zürich, Switzerland
tel. +41 44 632 4430
e-mail: woycechowsky@org.chem.ethz.ch
