Single Cell Image Correlation Analysis and Phenotypic Screening for Diabetic Cardiomyopathy

Abstract

The “patient in a dish” paradigm postulates that more predictive information about human diseases can be obtained from in vitro models based on 3D co-culture cellular systems with a defined genetic background derived from the patient's own induced pluripotent stem cells (iPS). We have developed a model for diabetic cardiomyopathy using human iPS and a sarcomere integrity assay for medium-scale high-content screening [1]. Image correlation in combination with texture filtering allowed us to extract a streakiness score of the α-Actinin stain for each individual cell. With this score we compared the phenotype of normal cells exposed to diabetic stress with cells from type-2 diabetic patients and found similar characteristics of the streakiness score and Ca-signaling. Furthermore, the degree of cardiomyopathic phenotype of the patient-specific cells correlated with the severity of their original clinical status. These models are incorporated into successive levels of a screening platform, identifying drugs [2] preserving the healthy cardiomyocyte phenotype in vitro during diabetic stress. In addition, we were able to identify several target classes from bioactive compounds with known mechanism of action. With a tailored compound set enrichment analysis we could identify substance classes with similar chemical structure or mode of action. In summary, this work constitutes the first patient-specific iPS model of a complex metabolic condition, and the first application of image correlation spectroscopy for fully automated single cell classification and phenotypic screening.[1] Drawnel et al., “Disease modeling and phenotypic drug screening for diabetic cardiomyopathy using human induced pluripotent stem cells”. Cell reports, in press.[2] Prummer M, “Hypothesis testing in high-throughput screening for drug discovery”. J Biomol Screening. 2012;17(4):519-29.