Modeling Heart Disease on Chip by Deep Learning-Based Image Analysis – MoHeDis

The advent of human induced pluripotent stem cell (hiPSC) technology revolutionized the entire field of therapy development for cardiac diseases such as dilated cardiomyopathies. Cardiomyocytes derived from hiPSCs (hiPSC-CMs) are now at the basis of a growing number of platforms dedicated to performing high-throughput and high-content screening (HTS/HCS). The continuous development of these platforms raises remarkably high expectations in both academia and industry, although there are still challenges to overcome such as the well-documented immaturity of hiPSC-CMs compared to adult cardiac myocytes. To improve the maturity of these cells, several methods exist but most of them are not compatible with HTS. We hypothesize that the culture of hiPSC-CMs on micro-patterns with controlled rigidity and surface chemistry will improve the structural features of these cells while remaining compatible with HTS.

Our first objective is to differentiate cardiomyocytes from hiPSC of different sources (healthy control, patients with dilated cardiomyopathy). In a perspective of drug screening, the features and behavior of the biological material used must be highly reproducible to extract robust information and knowledge from HTS/HCS. Presently, the differentiation of hiPSC into cardiomyocytes is not totally controlled and lead to strong batch-to-batch variations. In addition to differences into the efficacy of differentiation, these variations are also due to differences into the cardiomyocyte’s subtypes from one differentiation to another and also to the maturation state of hiPSC-CMs. Hence, we will combine monolayer differentiation, amplification and ventricular specification of cardiomyocytes to set-up a robust protocol for differentiation and maturation of cardiomyocytes from hiPSC toward their use for HTS/HCS. Specific methods will be used to measure the homogeneity of the hiPSC-CMs.

Our second objective is dedicated to design and fabricate a new biomimetic cell culture platform as a commercial product. To mimic cardiac micro-environment, we propose to develop substrates constituted of spatially tunable rigidities which mimic heart’s rigidity. Cell&Soft which is a small French company, has developed a pioneering method that allows the elaboration of polymerized hydrogel substrates exhibiting cellular and subcellular physiological µ-patterns of rigidity. This technology is very robust and suits to the requirements of an HTS/HCS platform in terms of handling and imaging: the µ-pattern polyacrylamide substrates are optically clear and fabricated on a glass slide so cells can be visualized with most types of microscopic imaging.
Our third objective is to develop proprietary software tools to quantitatively analyze the morphology and function of hiPSC-CMs cultured onto the biomimetic platform. These tools will be based on deep learning approaches and after image acquisition we will first adapt an already existing feature extractor for the analysis of hiPSC-CMs. In a second step, we will create a discriminant model to better understand the characteristics of mutated cell lines compared to controls.

Then, for our last objective we propose to first, use atomic force microscopy to mechanically characterize distinct cardiac cells (healthy and mutant), and in a second time, to generate large datasets to be exploited by deep-learning techniques that can further help distinguishing healthy and mutant cells. This will allow us to correlate morphology, function, and mechanical properties of hiPSC-CMs to validate the cell model and the biomimetic platform and to gain insights onto the pathophysiological mechanisms of DCM. The final aim of this project is the development of several commercial products, i.e. 96- and 384- well plates with micro-patterned cell culture substrates for the culture of hiPSC-CMs compatible with HTS/HCS applications and software tools for the analysis of hiPSC-CMs for HTS/HCS applications.