Hossein Fazelinia
Shuang Yan Tang
Computational design of AraC protein with
novel effector specificity
Our computational approach to protein engineering (with C. Maranas) begins with the crystal structure of the AraC arabinose-binding domain in complex with arabinose (Figure 3) [11]. For a given protein sequence and conformation, the binding energy can be calculated using a physics-based energy function. Our procedure involves iterative backbone movements and optimal sequence redesign (near the binding pocket) based on potential energy scoring functions, which serve as a surrogate for binding affinity. The objective therefore is to minimize the binding energy with respect to both side chain identity and conformation. This is accomplished by identifying mutations in the protein binding pocket that minimize binding scores for novel substrates. To ensure protein stability while minimizing binding energy, we constrain the overall energy of the system.
Whereas AraC binds L-arabinose with high affinity, D-fucose can only be weakly bound and is therefore a weak inducer [13,14]. Simulation and optimization methods have been used to accurately reflect the relative strengths with which wild-type AraC binds these and other compounds. L-arabinose had a binding score of -68.5 kcal/mol, while D-fucose had a binding score of -61.3 kcal/mol. D-glucose, which is unable to induce transcription through AraC, was found to have a significantly higher binding energy (-5.56 kcal/mol). Mutations in AraC reported to improve the inducing strength of D-fucose were also tested computationally and found to significantly lower the binding energy for this ligand.
These results validate our binding models and imply the ability to computationally predict mutagenesis strategies for novel effector binding. We are now shifting our focus to model binding of novel substrates by AraC. Our initial designs are aimed at engineering AraC variants which are activated by selectively binding one of a variety of similar monoterpene molecules. In particular, we will attempt to design proteins capable of distinguishing between different oxidized forms of the bicyclic monoterpene a-pinene, such as the high-value flavonoids verbenol and verbenone, shown in Figure 4. Our interest in binding these and chemically similar target molecules stems from a need to develop biocatalysts capable of converting renewable and abundant natural resources (e.g., fatty acids, amino acids, or plant oils such as those containing a-pinene) into higher value products such as antibiotics, pharmaceutical intermediates, and chemicals for the flavor and fragrance industry [15]. Note the significant difference in structure and chemistry between the natural AraC effector molecule (L-arabinose) and our bi-alicyclic target molecules.
Shown in Figure 4, verbenol (cis and trans) and verbenone are both synthesized from a-pinene via oxidation at a single position. Oxidation of a-pinene occurs in a variety of organisms and typically results in a distribution of oxidized products [16-18]. Wild-type cytochrome P450cam monooxygenase from P. putida oxidizes a-pinene at a rate of ~19 min-1 and produces a distribution of products, the most abundant being cis-verbenol at 31% [19]. Wong and coworkers were able to alter P450cam's activity and regiospecificity on a-pinene using site-directed mutagenesis, resulting in a triple-point mutant which oxidizes a-pinene at a rate of 120 min-1 and produces 86% cis-verbenol [19]. While purely "rational" design strategies are sometimes capable of improving desired properties in enzymes for which structures are known, site-directed mutagenesis is ultimately limited compared to what is possible using directed evolution, provided mutant libraries can be assayed in a high throughput manner [20,21].