Protein engineering in the 21st century.

Abstract

Protein engineering, the process by which novel proteins with desired properties are developed, has grown by leaps and bounds since the first examples of protein mutagenesis were described over three decades ago. In the last 30 years, protein engineers have successfully created a wide range of proteins tailored to specific industrial, medical, and research applications. Yet challenges remain that prevent the engineering of complex protein functions on demand. With the aim of creating a forum for the protein engineering community, the inaugural Protein Engineering Canada Conference was held on June 20–22, 2014, in Ottawa, Canada. Over the course of the 2-day conference, 115 protein scientists from over 30 institutions representing 5 countries shared ideas, networked with colleagues, and learned about exciting new research covering a broad array of topics such as enzyme engineering, computational protein design, X-ray crystallography, protein NMR, molecular modeling, and protein evolution. This special issue highlights the breadth of research that was covered during the Protein Engineering Canada Conference by including previously unpublished research articles submitted by speakers as well as some selections that were submitted directly to Protein Science. The first set of articles included in this special issue deal with enzyme catalysis, one of the most frequently engineered protein functions due to the many applications of designed enzymes in industrial and biomedical processes. While the protein engineering field has made great strides towards the development of new biocatalysts, these efforts still rely heavily on structure/function studies of known enzymes. Several informative examples of this are presented in this issue. In the first contribution, Ebert et al. studied the tolerance of Type II R67 dihydrofolate reductase to combinatorial active site mutations.1 They discovered that this tetrameric enzyme can maintain native-like activity across a large active-site sequence space and various quaternary structures. These results illustrate the high potential for continued functional evolution of this enzyme conferring antibiotic resistance to bacteria. In a second article by Egesborg et al.,2 a similar mutagenesis approach was employed to identify drug-resistant mutants of BlaC β-lactamase from M. tuberculosis, allowing for the discovery of double mutants with several thousand fold greater resistance to the commonly-used antibiotic clavulanate. This finding provides new insights into mechanisms that give rise to antibiotic resistance. Finally, Foo et al.3 probed the effect of local environment on the activity of a membrane-embedded rhomboid protease by examining the consequence of using membranes with different alkyl chain lengths to reconstitute the enzyme. The results highlight the effect of hydrophobic mismatch between protein and environment on catalytic turnover numbers, suggesting a potential role for lipids in the regulation of enzyme activity in vivo. Another important objective of protein engineering is the design of peptides that can bind selectively to target proteins. Two articles that highlight the utility of this approach are presented here. In an article by Kaplan et al.,4 surface residues of peptides that bind to basic leucine-zipper (bZIP) transcription factors were redesigned in order to increase affinity and selectivity to their bZIP partner. They found that incorporation of residues with strong helical propensity increased binding affinity to the target bZIP but that alterations to electrostatic interactions made by surface residues instead lead to changes in binding specificity. This information may be exploited in the future design of peptide-based inhibitors of transcription factors that contain these structural elements. In another article, Clinton et al. developed peptide mimics of the Ebola virus N-trimer, a highly conserved region of the GP2 fusion protein.5 These peptide mimics show great potential as drug targets in the development of Ebola virus entry inhibitors, a timely discovery given the recent outbreaks of this deadly virus. Over the years, protein engineering has greatly benefitted from the continued development of computational techniques that facilitate the rational design of proteins, as illustrated by two papers included in this special issue. In the first article, Frappier and Najmanovich describe how an elastic network atomic contact model for coarse-grained normal mode analysis can be used to predict the effect of mutations on thermostability.6 The utility of this technique was demonstrated by applying normal mode analysis to correctly distinguish between sequences originating from thermophilic and mesophilic species. In another article, Zhou and Grigoryan describe the development of a method to search for specific tertiary structural elements within the Protein Data Bank.7 Their method enables the rapid identification of backbone substructures matching the desired tertiary fragment within a user-defined RMSD limit. Both of these computational techniques may find use in protein engineering by helping researchers predict the effect of mutations on protein structure and stability. Among the computational tools available to protein engineers, computational protein design methods are becoming increasingly valuable due to their ability to evaluate amino acid sequences in silico on a scale that is experimentally impossible to achieve. Three articles that use computational protein design methodologies to engineer proteins are included in this issue. In the first study, Murphy et al. show how computational protein design can be used to design de novo a four-helix bundle that adopts its predicted structure with atomic-level accuracy.8 A solution NMR structure of their designed protein showed excellent structural agreement with the predicted structure for 3 out of 4 helices, with the fourth being shifted by 3 Å along the bundle axis. In a second article, Raymond et al. used computational design to generate an allosterically regulated enzyme catalyzing the retroaldol reaction from a protein scaffold devoid of catalytic activity.9 Their designed retroaldolase is activated in the presence of Ca2+ ions but displays modest catalytic efficiency and rate acceleration. A third article by Borgo and Havranek describes the engineering of an enzyme capable of catalyzing substrate-assisted Edman degradation.10 Using a computational approach that included quantum calculations, docking, and computational protein design, they were able to develop an Edmanase displaying high catalytic efficiencies towards several substrates but only a modest rate acceleration. The three articles described above illustrate how computational protein design methods can be applied to various protein engineering problems but also highlight the limitations of current approaches. For example, the accurate prediction of protein stabilities has proven to be a challenging task in computational protein design. In their contribution, Davey and Chica explored how alterations to the input template structure affect stability predictions made with computational protein design.11 They show that there exists a bias caused by the input template rotameric configuration that skews predictions towards sequences displaying either increased or decreased stability. Their findings suggest that many failed computational protein design predictions may result from scoring of sequences on incorrect templates for that sequence and highlight the importance of using alternate backbone templates for the scoring of mutant sequences. Although protein engineering has yielded many useful proteins for industrial and biomedical applications, the identification of protein sequences that will display desired activities remains challenging in part due to our incomplete understanding of the properties that dictate protein folding and dynamics. Three contributions that attempt to address this problem are included in this special issue. Using a dataset of 108 proteins from various structural classes, Broom et al. discovered that both folding and unfolding rates correlate strongly with structural complexity.12 They also found that while there is a strong correlation between folding and unfolding rates, single mutations have on average an approximately 15-fold larger absolute effect on unfolding rates than on folding rates, suggesting that proteins can be engineered to display fast folding and slow unfolding. There is also growing evidence that dynamics play crucial roles in protein function and that they should thus be considered in the protein engineering process. Gagné et al. used NMR to characterize the global effect on dynamics upon ligand binding in human angiogenin.13 They identified two clusters of long-range amino-acid networks that respond in a correlated fashion upon binding, suggesting that they may be involved in controlling enzyme function. In another article, Axe et al. studied the effect of mutation on dynamics of amino-acid networks in the alpha subunit of tryptophan synthase.14 They observed changes in the structural dynamics of long-range amino-acid networks caused by a single mutation. The results from these two studies demonstrate the importance of considering long-range impacts on protein structure and dynamics when engineering proteins. Taken together, these articles offer a glimpse of the exciting research that was presented at the Protein Engineering Canada Conference and serve to illustrate how the field of protein engineering has matured over the past decades. However, protein engineering remains a challenging enterprise, and only with a complete understanding of the link between protein structure, dynamics, and function will it become routine. Nevertheless, we expect that continuing advances in structural biology, high-throughput screening methodologies, and computational modeling will allow design of an increasing number of proteins required to tackle important challenges of the 21st century in medicine, agriculture, and industry.