To guide the development of a yeast strain with the ability to sense inflammation biomarkers we included modeling of different biosensor designs. In the dry lab, we focused on comparing the three designs in terms of their contribution to biosensor sensitivity and behavior. This resulted in identifying the most applicable design for our biosensor. Furthermore, special effort was spent on modeling and evaluating different versions of the modified GPA1 protein (advanced design), by employing computational methods for protein engineering.
Sensitivity Comparison
We compared the three designs of engineered signaling pathways in S. cerevisiae in terms of sensitivity through modeling with ordinary differential equations (ODEs). The models revealed the additional benefits of employing the yeast pheromone cascade in signal amplification (~7 orders of magnitude), thus rendering one of the designs as a clear candidate for the application in the biosensor (fig. 5).
Fig. 5. Comparison of the dynamic ranges of our designs. Plotting reporter signal strength against interleukin concentration in uM reveals a significant difference in sensitivity between the three designs.
Impact of Adverse Effects
We modeled the effects of hypothetical cellular scenarios (e.g. reporter toxicity) on the pathways within the framework of stochastic differential equations (SDEs). There, we explored various expected and unexpected behaviors in the models, which suggested that specific failures of pathway components may lead to characteristic statistics of reporter concentrations (fig. 6). This tool has the potential to improve our troubleshooting in the future.
Fig. 6. The effect of reporter toxicity on the reporter concentration. The amount of noise applied to all variables increases with reporter concentration, which leads to highly variable reporter concentrations (illustrative)
Protein Modeling
As the most applicable design required utilization of the yeast pheromone cascade, we had to engineer a novel GPA1 protein that would allow for signal transduction from our designed receptor system. Guided by several iterations of simulations with Rosetta Software Suite, we identified multiple regions suitable for inserting cleavage sites.
However, the predictions suggested that the post-cleavage protein fragments did not exhibit the properties we expected (fig. 7). Based on these findings, we articulated a refined framework for engineering signal transduction in our biosensor.
Fig. 7. Change in Gibbs free energy of fragments produced after cleavage of GPA1 mutant m124. The positions of each fragment in the original protein is denoted on the x-axis. Negative values suggest increased affinity to the beta subunit of the yeast G protein.