Probabilistic Protein Design, Comparative Modeling, and the Structure of a Multidomain P53 Oligomer Bound to DNA

Proteins are the main functional components of all cellular processes, and most of them fold into unique three-dimensional shapes guided by their amino-acid sequence. Discovering the structure of a protein, or protein complexes, can provide important clues about how they perform their function. However, the chemical, physical or architectural properties of many proteins impede traditional approaches to structure determination. Two such proteins, the tumor suppressor p53 and the cholesterol processing enzyme endothelial lipase, are prime examples of problematic proteins that defy structural investigation via crystallographic methods. Therefore, new techniques must be developed to gain valuable structural insights, such as: computationally assisted protein design strategies, more efficient crystal screening, or a combination of both. We applied a statistical computationally assisted design strategy to stabilize a p53 variant consisting of two independently folding domains. The re-engineered variant retained normal DNA-binding activities, and allowed us to experimentally determine the first structure of a physiologically active multi-domain p53 tetramer bound to a full-length DNA response element. We then demonstrated how computational methodology can be used to gain functional detail of proteins in the absence of experimentally determined structures. By creating comparative models of endothelial lipase, we discovered structural features that describe function and regulation, and gained a better understanding of the mechanisms conferring substrate specificity. Additionally, traditional methods for protein structure determination, such as X-ray crystallography, require relatively large amounts of purified sample in order to screen a sufficient variety of conditions. To improve this process, we developed a novel method for protein crystal screening using a microfluidics platform. We show how it is possible to use smaller quantities of protein to screen larger varieties of conditions, in turn increasing the probability of success in obtaining crystals. Furthermore, in contrast to current crystallographic approaches, all steps from screening to crystal growth to data collection were performed within the same reaction chamber, without any manipulation of the crystal, dramatically increasing the efficiency of both time and sample required to realize the structure. Collectively, these results demonstrate how advances in computational and experimental approaches can provide structural detail for proteins in circumstances where traditional methodology fails