Research using XSEDE supercomputers studies bioelectric effects of cells to develop new anti-cancer strategies
A very high resolution top-view of the simulation box. It illustrates size of the domain, and the coarsening of the mesh far from the aggregate all the way to the boundaries
Cell membranes are the key regulating factor for biological processes at the tissue-scale. The application of an electric field can alter the cell membrane's permeability to chemicals in their vicinity or even kill the subjected cells. The process, known as electropermeabilization or electroporation, consists of applying short, intense electric pulses.
Electroporation is currently used in the treatment of some cancers. For example, electro-chemotherapy is a cutting-edge cancer treatment technique that uses electroporation as a means to deliver chemotherapy into cancerous cells.
"Irreversible electroporation is used to destroy cancer cells," said Frederic Gibou, a faculty member in the Department of Mechanical Engineering and Computer Science at the University of California, Santa Barbara (UCSB). "Electroporation is also used to deliver chemotherapy by enhancing cell permeability to non-permeant drugs. One of the important questions is which electric pulses will maximize the efficacy of treatments."
Another interesting application is accelerating combat wound healing using electric pulsation.
"It's an exciting, but mainly unexplored area that stems from a deeper discussion at the frontier of developmental biology, namely how electricity influences morphogenesis" — or the biological process that causes an organism to develop its shape — Gibou said. "In wound healing, the goal is to externally manipulate electric cues to guide cells to grow faster in the wounded region and accelerate the healing process."
The common factor among these applications is their bioelectric physical nature. In recent years, it has been established that the bioelectric nature of living organisms plays a pivotal role in the development of their form and growth.
To understand bioelectric phenomena, Gibou's group has considered computer experiments on multicellular spheroids in 3D. Spheroids are aggregates of a few tens of thousands of cells that are used in biology because of their structural and functional similarity with tumors.
XSEDE Cyberinfrastructure Plays Key Role
In new research published in the Journal of Computational Physics, Gibou and his team delve into a new computational framework for parallel simulations that models the complex bioelectrical interaction at the tissue scale.
Supercomputer allocations on Comet at the San Diego Supercomputer Center (SDSC) and Stampede2 at the Texas Advanced Computing Center (TACC) were awarded to the researchers through XSEDE, the Extreme Science and Engineering Discovery Environment funded by the National Science Foundation (NSF). Additionally, the researchers used TACC's long-term storage system, Ranch, also an XSEDE resource.
"XSEDE is a fantastic advanced computing ecosystem that really allows us to get data that would be impossible otherwise," Gibou said.
"We started from the phenomenological cell-scale model that was developed in the research group of our colleague, Clair Poignard, at the Université de Bordeaux, France, with whom we have collaborated for several years," Gibou said.
"This model, which describes the evolution of transmembrane potential on an isolated cell, has been compared and validated with the response of a single cell in experiments," he said. "From there, we developed the first computational framework that is able to consider a cell aggregate of tens of thousands of cells and to simulate their interactions. The end goal is to develop an effective tissue-scale theory for electroporation."
"This has been, by far, the largest simulation of cell aggregate electroporation to date," said Pouria Mistani, another lead author on the paper. "It has successfully provided terabytes of high fidelity measurements of electroporation processes on tens of thousands of highly resolved cells in a 3D multicellular spheroid configuration."
At the beginning of the project, Mistani used Comet to develop and test initial versions of the code. Then, he used Stampede2 to finalize and verify the code and to perform scaling tests. Finally, all the large scale simulations and measurements were performed on Stampede2.
"We also benefited from the suite of visualization tools provided by Stampede2," Mistani said. "It's as if you're looking at cells in the human body — you can directly probe the data and gain insights about how to properly describe the system. The ability to make observations and interrogate the data helps conceive the underlying effective theory."
With large-scale simulations, transferring raw simulation results to local machines is very difficult.
"A better strategy is to keep the data on supercomputers and reduce it via some post-processing before transferring to local machines for further analysis. Ranch is an excellent resource for storage of the simulation results," Mistani continued.
One of the main reasons for the absence of an effective theory at the tissue scale is the lack of data. Specifically, the missing data in the case of electroporation is the time evolution of the transmembrane potential of each individual cell in a tissue environment. Experiments are not able to make those measurements.
"Currently, experimental limitations prevent the development of an effective tissue-level electroporation theory," Mistani said. "Our work has developed a computational approach that can simulate the response of individual cells in a spheroid to an electric field as well as their mutual interactions."
Each cell behaves according to certain rules. "But when you consider a large number of them together, the aggregate exhibits novel coherent behaviors. It is this emergent phenomenon that is crucial for developing effective theories at the tissue-scale — novel behaviors that emerge from the coupling of many individual elements," Mistani said.
The effects of electroporation used in cancer treatment, for example, depend on many factors, such as the strength of the electric field, its pulse, and its frequency. "This work could bring an effective theory that helps understand the tissue response to these parameters and thus optimize such treatments," says Mistani.
The researchers are currently mining this unique dataset to develop an effective tissue-scale theory of cell aggregate electroporation.
"Before our work, the largest existing simulations of cell aggregate electroporation only considered about one hundred cells in 3D," said Mistani, "or were limited to 2D simulations. Those simulations either ignored the real 3D nature of spheroids or considered too few cells for tissue-scale emergent behaviors to manifest."
The Importance of Computer Simulations
Computer simulations are more and more prominent in the fields of science and engineering because they enable researchers to get data that sometimes cannot be obtained otherwise. State-of-the-art computer architectures, such as Comet and Stampede2, and advanced numerical methods open up new possibilities in advancing the frontiers of science in disciplines that are of high interest to the public, such as cancer treatment, combat wound healing, or the broad field of morphogenesis.
"For us, this research would not have been possible without XSEDE because such simulations require over 2,000 cores for 24 hours and terabytes of data to reach time scales and length scales where the collective interactions between cells manifest themselves as a pattern," Gibou said. "It helped us observe a surprising structure for the behavior of the aggregate out of the inherent randomness. XSEDE provides a truly unique infrastructure for scientific discovery in the era of big data."
Moving forward, the team's research goal is to develop an effective theory that describes the simulation results.
"This is an example where simulations are not merely used as a predictive tool, but help discover new phenomena," Gibou said. "Under electroporation, cells respond in surprising synchronicity and it's beautiful to witness how such levels of order emerge out of inherent randomness."
The research of P. Mistani, A. Guittet, and F. Gibou was supported by NSF DMS-1620471 and ARO W911NF-16-1-0136 under the leadership of Dr. Joseph Myers. C. Poignard's research is supported by Plan Cancer DYNAMO (ref. PC201515) and Plan Cancer NUMEP (ref. PC201615). P. Mistani would like to thank Daniil Bochkov in the CASL group for fruitful discussions that have contributed to this research. This work used XSEDE, which is supported by National Science Foundation grant number ACI-1053575. The authors acknowledge TACC at The University of Texas at Austin for providing HPC and visualization resources that have contributed to the research results reported within this paper. This research was performed in part within the scope of the Inria associate team NUM4SEP, between the CASL group at UCSB and the Inria team. MONC. C.P.'s research is partly performed within the scope of the European Associated Laboratory EBAM on electroporation, granted by CNRS.