Felice C. Lightstone (16-FS-007)
Abstract
The purpose of this study was to determine if it is feasible to simulate the permeability of drugs and toxic agents across human-tissue membranes using computer-generated models of various human tissue types. Recent studies postulated the existence of microscopic domains of specific lipid compositions within the cell membrane. However, due to the absence of critical supporting data about the actual lipid compositions, we were unable to conduct the computer-based simulations needed to construct the membrane models. When necessary data to construct these simulations becomes available in the near future, the resulting tissues models may aid in the understanding of how these lipid compositions, and thus membrane behavior, vary throughout the body. We believe that this knowledge can have a significant impact on the drug design and development process.
Background and Research Objectives
One of the most critical cellular processes that controls almost every aspect of life is the signaling that occurs from one side of a cell membrane to the other. However, lipid composition can alter the membrane fluidity, thickness, flexibility, and electrostatic properties. Different tissue types contain different compositions of lipids in their cell membranes; thus, cell behavior (and response to external stimulus) can vary in different parts of the body. Recent studies postulated the existence of microscopic domains of specific lipid compositions within the cell membrane. It is vitally important to understand how these lipid compositions (and, by extension, membrane behavior) vary throughout the body. In addition, the toxicological outcome of the administration of a drug or the intake of a toxicant is determined by its concentration in the body over time. Accurately predicting the distribution of these substances is largely controlled by membrane permeability through the different tissues and organs.
The purpose of this project was to explore the feasibility of using coarse-grained and atomistic computer-based simulations to model human-tissue membranes by changing the lipid compositions. We planned to model lipid membranes to mirror the lipid composition of organs in the body and simulate the permeability of drugs and toxicants across those membranes. The simulations could then be compared to experimental results to validate the minimal lipid compositions that accurately model the tissue types. This knowledge may have a significant impact on the drug design and development process, which is currently estimated to take as long as fifteen years. During this process, 90 percent of drug candidates fail in clinical trials because of toxicological effects or by proving to be ineffective in treating the target disease.
Scientific Approach and Accomplishments
In FY16 we (1) created two models of a human brain-cell membrane composed of sixteen and sixty different lipid types, respectively, (2) ran these models at a coarse-grained resolution using the MARTINI coarse-grain force field for biomolecular simulations, (3) completed the coarse-grained simulations for human-brain cell membranes, and (4) converted the coarse-grained representation to an atomistic level for the 16-lipid model. We projected that the atomistic simulations would continue to run into the next year.
In FY17 we planned to (1) continue the atomistic simulations of sixteen lipid types in human brain-cell membranes, (2) build the models for the heart and liver, (3) run these new models at both the coarse-grained and atomistic level, and (4) select one drug for permeability simulations in brain, heart, and liver tissues. However, for reasons cited below, we were unable to realize these goals.
The feasibility of creating lipid compositions of different human organs has been quite challenging due to the limited amount of data currently available for human plasma membranes for specific organs. The progress of our work was hampered by the limited availability of critical supporting experimental data: specifically, the absence of actual lipid compositions of human tissues placed critical limitations on this project. To create the models for each human tissue type, the lipid composition of the inner and outer leaflets of the plasma membranes for each tissue must be known. In most current studies in this area, the lipid composition of the tissues and cells are reported for the whole cell, not just the plasma membrane. The whole cell includes other membranes (such as the nuclear and mitochondrial membrane), so the lipid composition is not just of the plasma membrane.
In addition to the difficulty of obtaining a mixture of membrane compositions, the outer and inner tissue leaflets are extremely difficult to separate and determine the lipid compositions quantitatively. New methods of creating artificial lipid bilayers wherein lipid compositions are quantifiable have yet to be designed for each leaflet, so quantifying the lipids in each leaflet is still difficult.
Finally, data about the lipid compositions of plasma membranes for the ten major human organs simply are not available currently. Given that lack of information, the only models we were able to create were of the human neural-plasma membrane because the lipid composition was reported in currently available literature. We were able to create both complex and simple models of just the human neural-plasma membrane using the MARTINI coarse-grained model. The complex model was composed of fifty-eight different lipid types, and the simple model was composed of sixteen lipid types. Each system was run for 40 milliseconds, showing the systems to be stable and mixing. As the systems are run for a longer period of time, we expect that some of the lipids would begin clustering together and represent a true biological system.
Impact on Mission
Future work developing the human-tissue membrane models would support Lawrence Livermore National Laboratory's core competencies in bioscience and bioengineering by determining the feasibility of computational testing and prediction to identify better drug candidates faster. The membrane models developed by future work may be used to simulate exposure to unknown chemicals to predict human outcomes and would be relevant to chemical and biological security. Support of high-performance computing, simulation, and data science at the Laboratory would be realized through the development of new models and protocols that could be applied to other high-performance computing efforts. The anticipated large size of the simulations will require newly designed workflows for simulations of molecular dynamics.
Conclusion
The stated goal of our feasibility study proved to be too ambitious, given the current level of research in the field of lipidomics. The difficulties of compiling a suitable mix of membrane compositions and in determining the lipid compositions of tissues quantitatively, as well as the absence of data about the compositions of plasma membranes for all of the major human organs, were significant factors in our inability to realize all of our project's goals. However, with the continued growth in the field of lipidomics, methods for growing excess plasma membranes from living cells are being developed, so that new experimental data will be available in the future.
