This is from a blog post I had to write for one of my classes this quarter, BIS 132: Dynamic Modeling in Biology. I hope you enjoy it and find the subject as fascinating as I do!

Instead of writing on one of the provided prompts for this week’s blog, I decided to write about something that is specifically interesting to me and, I hope, will be interesting to you as well. I did this because I felt that the 300 word summary for the homework assignment wasn’t sufficient to fully capture the research and implications. That and the fact that I’m extremely excited about this research now that I’ve read through it and have had a chance to dig into it a little bit. So let’s begin, shall we?

The article I am writing about is called “A Dynamical Systems Model for Combinatorial Cancer Therapy Enhances Oncolytic Adenovirus Efficacy by MEK-Inhibition.” It was published this past February in PLoS Computational Biology, and was authored by 4 researches from MIT and UCSF. The article discusses the use of oncolytic adenoviruses for the treatment of metastatic cancers, as metastatic cancers are very dangerous and non-surgical treatments are quite often ineffective. Then again, surgical treatments for these types of cancers are often not very effective as well. So why use viruses? Well, viruses, specifically adenoviruses, have been used as vectors to deliver recombinant DNA to targeted cells for a small number of diseases. Adenoviruses are found in vertebrates, are the cause of conjunctivitis (“Pink Eye”) among other viral diseases, and have double stranded DNA, much like Humans and most other animals. Adenoviruses lacking the E1B-55K gene, which can inhibit the tumor-suppressing protein p53, can selectively target cancerous cells, as mutations to p53 are the most common types of cancerous cells and when p53 is active it can inhibit the action of the adenovirus. Thus, they typically can only infect cancerous cells where a p53 mutation is the cause of the cancer and cannot affect healthy cells, leaving them alone.

Here’s where it gets interesting, at least to me: these oncolytic adenoviruses, including ONYX-015 which is the particular virus studied in this model, require the protein CAR (Coxackievirus-Adenovirus Receptor) to be on the surface of the cell for the virus to be able to attach to and infect the cell. The CAR protein, however, is often not expressed in cancerous cells. To induce expression of this protein requires disruption of the Mitrogen-Activated Protein Kinase Kinase (MEK/MAPK2) pathway; disrupting this pathway arrests, or freezes, the cell in the G1 phase of the cell cycle, which is the phase preceding the S, or DNA Synthesis, phase. When the cell is frozen in the G1 phase, the virus is unable to replicate, and therefore can’t spread and continue to lyse cancerous cells. So how does one kill a tumor with a virus if the virus can’t attach to the tumor cell or replicate if it can attach? This is exactly what the authors of the paper wanted to find out and describe using a dynamic model.

To model the most efficient way to combine treatments to increase CAR expression while not freezing the cell in the G1 phase, the authors constructed a dynamic model using a four-state nonlinear Ordinary Differential Equation, which operates much like a bathtub model. To do this the authors had to use experimental data to quantify CAR expression, tumor cell proliferation, adenovirus infection, cell viability, and viral replication both in the presence and absence of MEK-pathway inhibition. This led them to the four-state ODE, where the four states, which are the state variables, are cell states during the treatment process: 1) uninfected cell density, 2) G1-arrested cell density, 3) untreated and infected cell density, and 4) MEK-inhibited and infected cell density. In addition to this, the total cell population was included as a state variable. The parameters used were the rate of cell proliferation and the rate of infection, where the cells can be either treated or untreated. To simplify the model, delays caused by cell cycle phase transitions were ignored, as were dose and treatment times. Additionally, it was assumed that prolonged treatment would not increase infection and that the cancer cells were uniform.

What was predicted by this model is that two-day pretreatment of cancerous cells with an MEK-inhibitor will almost double the expression of the CAR protein. If this inhibition is stopped when the adenovirus is introduced the cell is allowed to go to the S phase of the cell cycle and the virus is replicated, resulting in cancer cell lysis. It was also predicted that increased cell density at the time of infection will reduce the efficacy of infection. Both of these predictions were shown to be correct during later in vitro experiments. What was even more significant is that the model and these experiments showed that infection during G1 phase arrest is coincident with the greatest amount of virus production and the greatest amount of cancer cell lysis. This is significant for two reasons: one, the more obvious, is that this is when the virus infection should be started to maximize treatment efficacy, and two, which isn’t as obvious initially, is that currently we don’t know much about adenovirus replication in humans, but this points to the G1-S phase transition as being critical to virus replication, which has implications throughout virology, medicine, and cellular biology. Also found is that CAR expression at the time of infection is not the only determining factor for this therapy. The other factors remain to be seen, but nonetheless this model has led to discoveries about adenoviruses and this treatment that were not expected beforehand.

The authors admit, and I agree, that more could be added to the model to make it more accurate. This added accuracy can only come from further in vitro and in vivo experiments in order to elucidate other factors that may influence this type of treatment. That being said, the model data, when compared to data gathered during experiments, is very similar, with the model data for pre-treatment, simultaneous, and post-treatment simulations being within 19% of experimental data or less (the best was within 8% for simultaneous treatment simulations). This shows that the model has an incredible amount of accurate predictive power as is, and this predictive power can only increase as more data and knowledge are accumulated. This predictive power is seen when the authors acknowledge that the simultaneous and post-treatment protocols involved experimental procedures that were not taken into account during model development; in other words, the model came very close to predicting experimental data before other factors were known. I am extremely excited about the potential and promise of further research in this area, and I hope that I was able to accurately convey that through this blog post. Hopefully when more is published I will be able to comment on that in this blog as well.

Bagheri, N., Shiina, M., Lauffenburger, D. A., & Korn, W. M. (2011). A Dynamical Systems Model for Combinatorial Cancer Therapy Enhances Oncolytic Adenovirus Efficacy by MEK-Inhibition. (C. V. Rao, Ed.)PLoS Computational Biology, 7(2), e1001085. Retrieved from