Computational Design and Experimental Characterization of Proteins with Increased Stability and Solubility

Open Access
Author:
Schweiker, Katrina Lynn
Graduate Program:
Integrative Biosciences
Degree:
Doctor of Philosophy
Document Type:
Dissertation
Date of Defense:
February 26, 2009
Committee Members:
  • George Makhatadze, Dissertation Advisor
  • George I Makhatadze, Committee Chair
  • Judith S Bond, Committee Member
  • Thomas E Spratt, Committee Member
  • Philip C. Bevilacqua, Committee Member
Keywords:
  • protein stability
  • protein engineering
  • computational design
  • protein folding
Abstract:
The ability to design proteins from first principles will provide an efficient way to develop stabilized proteins, which could have a profound impact on a variety of biotechnological industries. For example, a biosensor made out of stable proteins would be able to be functional in harsh environmental conditions, such as the desert, where sensors made from less stable proteins would not be effective. Another example is that life-saving vaccines made from stable proteins could be stored at ambient temperatures, making it possible to distribute them more effectively to developing nations where refrigeration is not always an option. In addition to addressing the question of what forces govern thermodynamic stability, the field of protein design can also provide insight into the intramolecular interactions that are important for kinetic stability and solubility. In the first part of this dissertation, the SH3 domain of the Fyn tyrosine kinase (FynSH3) was stabilized by the rational design of surface charge-charge interactions. Analysis of the computationally optimized distributions of surface charges showed that the increase in favorable energy per substitution begins to level off after five substitutions. One of the sequences with five substitutions (four charge reversals and one introduction of a new charge) was selected for experimental characterization. Nine additional variants were also characterized to explore the stepwise contributions of these substitutions to the stability of FynSH3. The designed sequence was found to have an increased thermostability of 12 °C and an increase in the free energy of unfolding (ΔG) of 8 kJ/mol, relative to the wild-type protein. These results suggest that a significant increase in stability can be achieved through a small number of amino acid substitutions, and argue for a seminal role of surface charge-charge interactions in modulating protein stability. The second part of this dissertation addresses the question of how important the unfolded state of a protein is for determining its stability and whether it needs to be considered in the design approach. Some of the first attempts to address this issue tried to explain the pH-dependent changes in stability (ΔG) for several different proteins, where it was found that, in order to reproduce experimental data, a statistical (Gaussian polymer chain, GPC) representation of the unfolded state needed to be included in the calculations of ΔG. However, incorporation of this model into our design approach did not significantly improve our predictions. To determine whether this was due to an inability of the Gaussian model to accurately describe the distance distributions, and therefore the energies, observed in structural representations of the unfolded state, the distance distributions for a GPC were compared to those observed in the excluded volume limit (EV) structural libraries of two proteins: ubiquitin and NTL9. For residues that were close in sequence, where the unfolded state energies are the largest, it was found that these distributions were markedly different between the GPC and EV methods. A possible explanation for this observation is that the EV limit does not consider charge-charge interactions when creating the large-scale structural libraries. Molecular dynamics (MD) simulations were performed on the 2,000 structures in the EV libraries to model the unfolded state in the presence of charge-charge interactions, yet the Gaussian model was still unable to accurately reproduce the distance distributions of the structural library. However, very little difference was observed in the charge-charge interaction energies calculated by the Gaussian model versus directly calculating the energies in the post-MD unfolded state structural libraries, suggesting that the statistical model may be sufficient for describing the behavior of the unfolded state. Since (1) the overall charge-charge interaction energies in the unfolded state are small and (2) our design approach focuses on the differences in energies (ΔΔG) rather than absolute energies (ΔG) for selecting more stable variants, the overall effect of unfolded state can most likely be ignored without adversely affecting the predictive ability of the algorithm. The implication of these results for a protein that has previously been thought to have specific residual interactions in the unfolded state is discussed. In the third part of this dissertation, the question of how the thermodynamic stabilization of proteins redesigned by our approach affects the kinetics of the folding and unfolding reactions is addressed. The folding and unfolding kinetics of the wild-type and designed variants of a bacterial cold shock protein (CspB), FynSH3, tenascin, and procarboxypeptidase were examined. Since the hydrophobic collapse of the protein core is the first step in protein folding, the rate of hydrophobic collapse should drive the folding rate. All of the proteins designed in this study contain substitutions on the protein surface, while the core residues remain unchanged. Therefore, one would intuitively predict that the folding rates of the wild-type and designed variants of each protein should remain the same, so the observed increases in stability must come from much slower rates of unfolding. This is a logical conclusion because the designed proteins contain more favorable surface charge-charge interactions than the wild-type proteins, meaning that it would take more energy to break these favorable interactions once they had been formed, thus decreasing the unfolding rate. For CspB, this was indeed shown to be the case. However, the increased stability of the FynSH3, tenascin, and procarboxypeptidase variants appears to be due to a faster folding rate, while the unfolding rate remains unchanged relative to the wild-type. Based on φ-value analysis data from the literature, it appears that this affect is due to the substitutions being made at positions that have native-like structure in the transition state. The results of these experiments show that while proteins can be thermodynamically stabilized by the same method, the kinetic mechanisms of stabilization can be vastly different. By incorporating the results of existing φ-value analyses into the design algorithm, it should be possible to select for residues that would decrease the unfolding rate, rather than increase the folding rate. This means that one could potentially design a protein that is not only thermostable, but also kinetically stable, which would have profound implications for the development of protein therapeutics. The fourth part of the dissertation explores the role of surface charges in making proteins less susceptible to aggregation. A few recent reports suggest that adding a large number of charged moieties to proteins (supercharging) increases solubility and decreases aggregation due to thermal denaturation. While this approach seems to be an effective way to combat protein aggregation, nothing is known about the thermodynamic effects of supercharging. A supercharged variant of ubiquitin was designed by introducing charges at positions that were not predicted to have a significant impact on the thermodynamic stability. Not only was the supercharged variant of ubiquitin more soluble than the wild-type at neutral pH, but it also showed reversible thermal denaturation under conditions where wild-type ubiquitin aggregates. Interestingly, this protein was destabilized relative to the wild-type protein. While the supercharged ubiquitin was predicted to have similar thermodynamic stability to the wild-type, it is possible that our design approach cannot accurately predict charge-charge interaction energies in a highly charged molecule. Further studies on more supercharged proteins should help develop a foundation by which we can further understand the thermodynamic mechanisms, and therefore, more accurately predict, the effects of supercharging on protein both protein stability and protein aggregation. In the fifth, and final, part of the dissertation, the effects of pressure on protein denaturation are examined. Pressure perturbation calorimetry (PPC) is a new experimental method that is being used to study the volumetric properties of proteins. PPC measures the coefficient of thermal expansion (α) of a protein in dilute solution when subjected to changes in pressure (ΔP ~ 80 psi) under isothermal conditions. By measuring α as a function of temperature, it is possible to measure the volumetric changes (ΔV/V) in proteins upon unfolding. A novel method for analyzing the data using a thermodynamic two-state model of unfolding was developed, and was used to analyze PPC data for five model proteins: lysozyme, ribonuclease A, ubiquitin, cytochrome c, and eglinC. It was observed that the volumetric changes upon unfolding of all proteins, except cytochrome c, converged at high temperature. The anomalous behavior of cytochrome c is most likely due to the imperfect packing of the protein around the heme group. The results discussed in this chapter set a foundation for exploring how the alteration of intramolecular interactions such as packing interactions or surface charge-charge interactions will affect the volumetric properties of proteins.