Fast protein folding: It's a wrap
by Leigh Soutter of COMSOL
How relaxed protein strands loop and lace or "fold" controls the form they take and the functions they perform in the human body. If the stretched out chains of amino acids slip into a particular knot you get hemoglobin that carries oxygen through the blood. A different 3D lace becomes an antibody to fight infection. A slight misstep, and you wind up with insoluble lumps and the chance for mad-cow disease, cystic fibrosis, sickle-cell anemia, Alzheimer's disease, and other protein folding ailments.
Naturally a detailed picture of each step in the folding process could greatly simplify prevention and therapy for these diseases. Unfortunately, the proteins, which are difficult to isolate, respond so quickly to the chemicals that trigger the folding, you can't resolve the dynamics of the mixing from the folding with conventional tools.
Dr. David Hertzog
COMSOL modeling helps researchers spy on the rapid protein folding process. Dr. David Hertzog and an interdisciplinary team from Stanford University, Lawrence Livermore Laboratories, UCLA, and Michigan State University create microfluidics devices that mix the protein with various chemical triggers and optically capture what happens in short intervals of 10 µs or less. Using COMSOL modeling with optimization scripts to minimize the mixing time, the camera gets a clear view of just the folding process. Their work recently made the cover of Analytical Chemistry (Ref. 1), with more slated for publication soon (Ref. 2). If you attended the COMSOL workshop at Stanford University this spring, you saw Dr. Hertzog present on some of their work.
The ultra-fast micro-mixing devices designed by Dr. Hertzog and his colleagues consist of channels etched into silicon chips equipped with a confocol optical detector. The protein enters with a denaturant, which relaxes the protein strands. Traveling down the lengthwise channel, the denaturant diffuses into a buffer during travel, and the proteins fold in view of the optical detector.
Watching it fold
To create one of their fast-mixing microfluidics device, Dr. Hertzog and his team start by etching intersecting channels into a silicon chip. To begin, the channels are roughly 1 µm wide and 10 µm deep. They inject proteins and a denaturant into the lengthwise channel. The denaturant relaxes the proteins into stretched-out chains. At the same time they introduce a buffer into the channel that cuts across the chip. The denaturant diffuses strongly into the buffer because it has a high diffusivity relative to the proteins. As a consequence, the proteins in the mixture concentrate, narrowing to a thin sheet approximately 0.1 µm wide. With the loss of the denaturant to the buffer, the proteins collapse and fold. When the confocal optical detector near the outlet scans down the mixer, it builds an image of the entire process. Minimizing the time needed for the denaturant to move into the buffer or "mixing time" stretches out opportunities for the tiny camera to record the proteins fold.
This team of researchers uses COMSOL to control what the optical detector sees. With COMSOL modeling they fine tune the geometry of the channels as well as flow conditions like the inflow and buffer injection velocities. The best set up minimizes the mixing or "dead" time between the first protein-buffer contact and focusing of the protein stream. The procedure applies to other fast but optically detectable reactions, where the goal is to get a clean view of a fast process.
COMSOL Modeling
With COMSOL, Dr. Hertzog and his colleagues simulate flow and transport to achieve rapid mixing. Minimizing the mixing time or time required for the denaturant to diffuse away from the protein improves how well the optical detector sees the protein fold.
What underpins the design strategy is a good COMSOL model that details the fluid flow and the focusing of the proteins within the channels. Their models combine the Navier-Stokes and the Convection and Diffusion equations already defined in COMSOL Multiphysics. With obvious symmetry in the geometry, they focus much of their work on 2D modeling and ground truth with occasional 3D tests.
Calling COMSOL Multiphysics from the command line with a parametric-optimization script allows Dr. Hertzog and his colleagues to quickly find the best geometry for the micromixers. Shown are the initial and final geometries from an optimization sequence. Notice that the protein quickly focuses to a thin sheet in the optimized geometry.
Running the simulations from the command line allows Dr. Hertzog to implement scripts that optimize the geometry. They start with simple linear channels and use a parametric-optimization routine to finds the flow conditions and the channel configurations that achieve the shortest mixing time. Typically the routine varies the length, width, and position of linear channels. However, in follow-up work (Ref. 2) Dr. Hertzog also optimizes the overall channel shape with scripting plus COMSOL simulation.
The close link between COMSOL simulation, device design, and experimental procedure proved especially successful for Dr. Hertzog and his colleagues. According to him, "The micromixer we designed with COMSOL achieved mixing such low mixing times with ssDNA, it was easy to distinguish the protein folding from the mixing kinetics." He adds, "An important benefit of the high precision results we get with COMSOL is conserving hard-to-get ingredients; our optimized micromixers require only hundreds of femtomoles of the protein and tens of microliters of solution."
The COMSOL designed micromixer achieved mixing times of about 8 ms with ssDNA, which was well within the goal range of the research team.
Ready to download
COMSOL provides a number of ready-to-run microfluidics models being used in biotechnology right now.
Of particular interest are the electrokinetic valve models, which combine fluid flow, chemical transport, and electrical effects.
About the researcher
David Hertzog conducted the micromixer studies during his PhD research with Dr. Juan Santiago in the Microfluidics Laboratory of Stanford University's Mechanical Engineering Department. The work also was supported by the Lawrence Livermore National Laboratory in the Chemistry and Materials Science Division, where he works with Dr. Olgica Bakajin. He also holds degrees in Mechanical Engineering from Cornell University and Aeronautics and Astronautics from Stanford University.
References
1. D. Hertzog, X. Michalet, M. Jager, X. Kong, J. Santiago, S. Weiss, and O. Bakajin, "Femtomole Mixer for Microsecond Kinetic Studies of Protein Folding," Analytical Chemistry, vol. 76, no. 24, 2004, pp. 7169-7178.
2. B. Ivorra, D. Hertzog, J. Santiago and B. Mohammadi, "Design of Fast Microfluidic Protein Folding Devices by Global Optimization," submitted to Int'l J. Numerical Methods in Engineering, April, 2005.