A Computational Model For The Pi3k/Akt/Mtor Pathway Predicts Pten As A Negative Feedback Loop

The PI3K/Akt/mTOR pathway regulation is of paramount importance in numerous homeostatic physiologic processes. While it is often seen to be hyperactivated in many cancers including a majority of breast cancers, it is repressed in type-2 diabetes. In cancers this hyperactivation could be due to an overexpression or a mutation of a growth factor receptor, mutation of the PI3K or any activating mutation in the pathway leading to increased cell growth and tumorigenesis. PTEN is the negative regulator of this pathway and is also often mutated or is lower in abundance. The PI3K pathway has multiple feedback loops that serves to return the pathway output to its basal value following a perturbation. We asked in this context if PTEN could be controlled through a negative feedback loop.We assembled a collection of breast cancer cell lines (Her2+, ER/PR+, TN), plotted their normalized PTEN vs Akt and ran an unsupervised k-means clustering algorithm on the data points. We found that the PTEN high and low clusters showed a positive correlation between Akt and PTEN implying a positive regulation of PTEN by Akt. We then built a computational model for PI3K pathway with the output of the pathway connected directly back to PTEN in a feedback loop. We trained the computational model on BT474 cells (Her2+ system) and found that just 2 flavors of the model were sufficient to represent the Akt, PTEN values of all the cell lines in the PTEN-high cluster. We then expanded the model to learn the dynamics of the system during PI3K inhibition (by selective drugs) by including the Foxo driven growth factor receptor upregulation feedback loop. For training we modeled PTEN as a sum of two exponentially decaying terms and learnt the parameters of the model by fitting the experimental data to model output. By simulating different PTEN decay rates we showed that the Foxo driven feedback loop along with the rate of PTEN decay played an important role in determining the extent of the Akt rebound. However, the computational model with the Akt connected directly to PTEN failed to reproduce the PTEN decay seen during PI3K inhibition suggesting the presence of further downstream mechanisms. We then expanded the model to include mTOR and its substrates 4EBP1 and S6K, which are negative and positive regulators of PTEN translation, respectively. We trained the model on the phosho-4EBP1 and phosho-S6K time profiles obtained during PI3K inhibition and found that these additional mechanisms completely captured the PTEN dynamics seen during PI3K inhibition. As a prediction we also showed that the model captures the abrogation of PTEN decrease in 4EBP1-KO cells along with a concomitant lower rebound in Akt.Thus, here we develop an experimentally validated model of the PI3K/Akt pathway with multiple feedback loops and show that it accurately models the PTEN dynamics during PI3K inhibition and provides quantitative support for PTEN regulation through a negative feedback loop. Citation Format: Kiran Vanaja, Radha Mukherjee, Neal Rosen, Andre Levchenko. A computational model for the PI3K/Akt/mTOR pathway predicts PTEN as a negative feedback loop [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 5506.