Why design new protein function?
Nature provides us with myriads of examples of exquisitely selective proteins that function as binders and enzymes. These proteins, however, are often formidably complex. For instance, a typical enzyme or the antigen-binding domain of an antibody comprise more than 200 amino acids and fold into complex three-dimensional conformations that depend critically on thousands of atomic interactions. By changing the protein sequence and structure or designing completely new proteins, we may generate desirable new enzymes for green chemistry, binders for research, diagnostics and therapeutics, and exquisite molecular sensors to measure metabolite concentrations. Designing proteins with new or enhanced properties is therefore a fundamental challenge of immense importance.
Further reading:
-
Fleishman, S. J.; Baker, D. Role of the Biomolecular Energy Gap in Protein Design, Structure, and Evolution. Cell 2012, 149, 262–273.
Our approach
We study families of natural proteins, such as antibodies and enzymes, to learn how function is encoded in their structure and sequence. We then develop computer algorithms in the Rosetta software suite for atomistic modeling and design to encode this understanding, design new proteins that do not exist in nature and test whether they work as expected in experiments. We therefore combine cutting-edge computational methods development with high-throughput experimental screening and stringent biochemical and structural analysis of designed proteins.
Further reading:
-
Fleishman, S. J.; Whitehead, T. A.; Ekiert, D. C.; Dreyfus, C.; Corn, J. E.; Strauch, E.-M.; Wilson, I. A.; Baker, D. Computational Design of Proteins Targeting the Conserved Stem Region of Influenza Hemagglutinin. Science 2011, 332, 816–821.
-
Whitehead, T. A.; Baker, D.; Fleishman, S. J. Computational Design of Novel Protein Binders and Experimental Affinity Maturation. Methods Enzymol. 2013, 523, 1–19.
Evolution-guided atomistic design — a new paradigm for design of function
Our initial research focus was on antibodies, because they are the most versatile binders in nature, able to bind both small molecules and large proteins, including therapeutic targets. At the very start of our antibody-design project, we noticed that designing large and complex proteins, such as antibodies, is complicated by their marginal stability, and our research therefore expanded to questions on the fundamental principles that ensure stability and foldability in large proteins of complex fold. A practical challenge that we faced was the sheer size of conformation and sequence space of such large proteins. To address this challenge, we developed a unique strategy that starts with a structural-bioinformatics analysis of natural proteins and uses the resulting statistics as constraints for Rosetta atomistic design calculations. Using this evolution-guided atomistic design approach, we were able, for the first time, to design atomically accurate antibodies, despite more than 50 mutations from any natural antibody.
Further reading:
-
Baran, D.; Pszolla, M. G.; Lapidoth, G. D.; Norn, C.; Dym, O.; Unger, T.; Albeck, S.; Tyka, M. D.; Fleishman, S. J. Principles for Computational Design of Binding Antibodies. Proc. Natl. Acad. Sci. U. S. A. 2017, 114, 10900–10905.
-
Khersonsky, O.; Fleishman, S. J. Why Reinvent the Wheel? Building New Proteins Based on Ready-Made Parts. Protein Sci. 2016, 25, 1179–11
Membrane proteins energetics & design
Serving as portals between a cell and its environment, membrane proteins are central to cell signaling and metabolism. However, they reside in a complex physical environment, surrounded by water, hydrophobic acyl chains, and polar lipid headgroups. Due to this complexity, our understanding of the design principles of membrane proteins falls far behind that of soluble proteins; despite their importance, for instance, we only have atomic structures
of a few hundred membrane proteins, compared to >100,000 for soluble proteins. For modeling and design of membrane proteins to be robust, we must improve our understanding
of membrane-protein energetics. We therefore established a high-throughput experimental assay to probe the energetics of every amino acid at every position across the bacterial inner membrane. The results of this assay resolved long-standing questions and provided quantitative understanding on membrane-protein stability; furthermore, our measurements are accurate enough to serve in energy-based prediction of membrane-protein topology. We will
use this new energetics to probe specificity, affinity, and stability of select membrane proteins involved in cell signaling, and to design new membrane protein inhibitors.
Further reading:
-
Elazar, A., Weinstein, J., Biran, I., Fridman, Y., Bibi, E., and Fleishman, S.J. (2016). Mutational scanning reveals the determinants of protein insertion and association energetics in the plasma membrane. eLife 5:e12125
-
Elazar, A., Weinstein, J., Prilusky, J., and Fleishman, S.J. (2016). The interplay between hydrophobicity and the positive-inside rule in determining membrane-protein topology. Proc. Natl. Acad. Sci. in press.
Automated design of function
The high accuracy and predictability of evolution-guided atomistic design enabled breakthroughs in a number of very challenging design of function problems. Importantly, the resulting methods were general enough to enable the first fully automated, web-enabled design platforms. For instance, we developed the Protein Repair One Stop Shop (PROSS), a web-based design platform to stabilize proteins. PROSS has resulted in large improvements in thermal stability and orders-of-magnitude improvement in protein-expression levels, including in a malaria vaccine immunogen. PROSS is now routinely used by hundreds of labs around the world to overcome challenging problems in protein production. Since stability is of paramount important in designing new protein activities, our lab uses PROSS routinely, leading to much higher success in the design of functional proteins. Extending the evolution-guided design strategy to protein active sites, we developed the Functional Library (FuncLib), a web server for the design of repertoires of functionally diverse enzymes. FuncLib has resulted in as much as 4,000-fold higher catalytic efficiency than the natural starting enzyme, including the first enzymes with therapeutic potential to treat poisoning by nerve-agents such as soman and sarin. Similarly, a method called the Affinity Library (AffiLib), has enabled one-shot orders-of-magnitude improvement in protein-protein interaction affinity. Furthermore, we developed a strategy called modular backbone assembly and design to generate stable and highly efficient enzymes by recombining backbone fragments of natural ones and to generate ultrahigh specificity binding pairs. In all of these methods, only a small set of designed proteins needs to be experimentally tested in order to identify stable and highly efficient enzymes and binders. These are concrete examples of the application of computational protein design to “real-world” challenges in protein engineering that have so far either required laborious iterative experimental optimization or been abandoned for being too difficult.
Further reading:
-
Goldenzweig, A.; Goldsmith, M.; Hill, S. E.; Gertman, O.; Laurino, P.; Ashani, Y.; Dym, O.; Unger, T.; Albeck, S.; Prilusky, J.; et al. Automated Structure- and Sequence-Based Design of Proteins for High Bacterial Expression and Stability. Mol. Cell 2016, 63, 337–346.
-
Campeotto, I.; Goldenzweig, A.; Davey, J.; Barfod, L.; Marshall, J. M.; Silk, S. E.; Wright, K. E.; Draper, S. J.; Higgins, M. K.; Fleishman, S. J. One-Step Design of a Stable Variant of the Malaria Invasion Protein RH5 for Use as a Vaccine Immunogen. Proc. Natl. Acad. Sci. U. S. A. 2017, 114, 998–1002.
-
Goldenzweig, A.; Fleishman, S. J. Principles of Protein Stability and Their Application in Computational Design. Annu. Rev. Biochem. 2018, 87, 105–129.
-
Khersonsky, O.; Lipsh, R.; Avizemer, Z.; Ashani, Y.; Goldsmith, M.; Leader, H.; Dym, O.; Rogotner, S.; Trudeau, D. L.; Prilusky, J.; et al. Automated Design of Efficient and Functionally Diverse Enzyme Repertoires. Mol. Cell 2018 in press.
-
Lapidoth, G.; Khersonsky, O.; Lipsh, R.; Dym, O.; Albeck, S.; Rogotner, S.; Fleishman, S. J. Highly Active Enzymes by Automated Combinatorial Backbone Assembly and Sequence Design. Nat. Commun. 2018, 9, 2780.
Design of vast functional repertoires
Automated computational design of functional proteins is enabling exciting opportunities. We can now automatically design millions of different binders or enzymes, each individually optimized to exhibit high stability, activity, and specificity. We are now developing a fully integrated repertoire design-build-test-learn strategy where millions of enzymes or antibodies are individually designed in each iteration and genetically encoded. We then use high-throughput screening and deep sequencing strategies to accurately characterize active designs and employ machine-learning methods to extract features of successful designs. We then encode these features in future repertoires until we reach general and reliable design methods. We anticipate that through the application of such repertoire-design cycles, we will learn new rules for designing completely new and highly efficient binders and enzymes in one shot.
Where are we headed?
Despite much recent progress, computational design is still a highly specialized field exercised by only a few expert labs. Our vision is to transform protein engineering so that all its major challenges can be addressed through computational protein design. Instead of relying on genome mining, animal immunization, repertoire screening, and in-vitro evolution — which can be time consuming and result in proteins exhibiting sub-optimal properties — future diagnostics, therapeutics, enzymes, and biosensors will be computationally designed to exhibit desired activity, stability, and specificity.