Expediting cryo-EM sample preparation using design of experiments.

Abstract

Determining cryo-EM grid freezing parameters to attain optimal particle distribution and ice thickness for data collection is a highly sample-specific process which frequently requires several iterative rounds of freezing and screening grids. Researchers relying on national cryo-EM centers and similar facilities for screening services can be subject to lengthy wait times for each screening session, which may delay research considerably if many rounds are needed. This provides a strong motivation for a more strategic and efficient approach to grid optimization. Here, we present a method of using Design of Experiments to more efficiently optimize the parameter space to produce data-collection-ready grids in as little as two rounds of freezing and screening. We employ a fractional factorial design to probe a wide range of parameters including blotting conditions, sample characteristics, and grid characteristics through a relatively small set of test conditions. Quantitative screening results based on ice thickness and particle distribution are used to generate a least squares regression model, which provides a set of optimized freezing conditions. We tested this approach with apoferritin on both the Vitrojet and Leica GP2, and with glutamate dehydrogenase on the Vitrobot. For all three attempts, grids were frozen with the optimized conditions given by the model. Optimized grids in each case had ice thickness within an ideal range for data collection and a near-monolayer of particles, with enough imageable area to support multiple days of data collection. Resulting grids were of consistent quality when the procedure was performed by different researchers on different vitrification instruments and on different days. Based on the quality of the optimized grids produced in these experiments, we conclude that this is a robust and reliable method for expediting and improving the overall success of the cryo-EM grid optimization process.