The initial resistance that the high-content/high-throughput community had to face has been overcome, and now most in the field acknowledge that flow cytometry is enormously high throughput and very high content.
One of those converts is J. Paul Robinson, Ph.D., SVM professor of cytomics and professor of biomedical engineering at Purdue University. At CHI’s recent “High-Content Flow Cytometry” meeting, Dr. Robinson talked about the ability to automate flow cytometry-based high-content screens and data processing to perform multiplexed high-content analysis in real-time, running multiple assays and samples in parallel, and to miniaturize the technology to conserve resources and reduce costs.
Dr. Robinson also discussed how data-extraction issues have historically been a stumbling block in the use of flow cytometry in drug discovery. Dr. Robinson described the main challenge in streamlining multiparametric data processing as eliminating operator input and avoiding manual gating. Automation provides a solution to managing the huge datasets produced with multiparametric high-content analysis.
To overcome the inefficiencies of traditional approaches to data analysis, Dr. Robinson’s group developed a technology based on a sequential gating process. The method involves combining all of the data files acquired into one file for analysis. This not only accelerates the process but also facilitates data comparison.
Using this approach, a biological sample can be analyzed in parts—exploiting the ability to separate individual cells for analysis using flow cytometry—with the data then combined to piece together the complete picture. Automation of the entire process—from sample preparation, to running complex assays, to data analysis—makes it possible to perform functional phenotypic profiling of cell populations efficiently.
The most important factors, he explained, are to embed multiple computational algorithms and analytical tools into the data analysis software and to extract the feature space correctly. He described the graphical interface the group developed—a logic map designed to analyze the data from functional live-cell assays that incorporate fluorescent markers.
In addition, Dr. Robinson presented an example of multicolor flow analysis to demonstrate the ability to perform cytomics studies for pathway analysis. Overall the use of flow cytometry to analyze individual cell populations and evaluate the effects of drugs on cell function allows you “to ask the what-if questions” and to begin to integrate pathways, he said.
Peter Krutzik, Ph.D., a senior scientist in Gary Nolan’s group at the Baxter Lab in genetics, pharmacology, microbiology, and immunology at Stanford University, described the use of phosphor flow cytometry to profile perturbations in signaling networks when peripheral blood cells are exposed to drug compounds.
This method allows for multiplexed analysis of cell types and signaling pathways, and has applications in drug target identification and throughout the drug development workflow, including in the analysis of patient samples during clinical testing.
The increasing use of automation and miniaturization is lowering the cost of flow cytometry-based high-content analysis in primary and secondary drug screening, making it possible to run larger screening campaigns.
Describing the benefits of using flow cytometry to study disease specificity as it relates to cell types and signaling pathways, Dr. Kritzik reported that it is possible to measure at least 10 parameters in 5,000–10,000 single cells per second. This capability is particularly important when studying heterogeneous cell populations and a variety of different phosphoproteins.
Dr. Kritzik and colleagues are using phosphor-flow cytometry to study three phosphoproteins—all linked to disease and all potential drug targets—in subsets of peripheral blood cells. Taking advantage of the capacity for multiplexing, they can perform 14 assays per well, generating reproducible data with low intra- and inter-run variance.
The screens have demonstrated cell-type specific druggability; for example, a particular drug might inhibit kinase activity in T cells but not in monocytes. The researchers were able to correlate cell populations with pathway activity. Grouping the data according to a specific phosphoprotein yields information on selectivity patterns.
Furthermore, Dr. Kritzik presented data from an automated secondary dose-response screen in which they were able to generate more than 20,000 data points per day and to create IC50 fingerprints for various Jak kinase inhibitors that are currently in development in the pharmaceutical industry.
Optimizing Data Analysis
In his presentation, Dan Ehrlich, Ph.D., research professor of biomedical engineering at Boston University, emphasized the problems associated with segmentation in data analysis and the “noise” it can introduce due to boundary drawing errors.
Various algorithms can be used to perform cell segmentation in the analysis of data from flow cytometry-based imaging studies. In segmentation, individual pixels are classified as belonging either to a particular region of the cell or to the background, and misclassification errors can affect a variety of measurements and calculations.
Dr. Ehrlich’s team has developed a method for one-dimensional high-content screening in a flow system using a parallel microfluidic cytometer (PMC), which incorporates a high-speed scanning photomultiplier-based detector, for low-pixel-count, one-dimensional imaging. He described published work using the system to generate six-pixel 1-D images in studies of protein localization in a yeast model for human protein misfolding diseases and in nuclear translocation assays in CHO cells expressed in an NFκB-EGFP reporter.
His group is developing a next-generation, simpler 1-D flow system and experimenting with modifications such as increased scanner resolution, various ways of implementing flow-based high-content analysis, and the potential for clinical applications of PMC in a disposable chip format.