Towards Fast, Reliable, and Affordable Image-Based Characterisation of Isolated Cardiomyocytes

Joachim Greiner (Freiburg im Breisgau)1, J. Madl (Freiburg im Breisgau)1, R. Peyronnet (Freiburg im Breisgau)1, P. Kohl (Freiburg im Breisgau)1, E. Rog-Zielinska (Freiburg im Breisgau)1

1Universitäts-Herzzentrum Freiburg - Bad Krozingen Institut für Experimentelle Kardiovaskuläre Medizin Freiburg im Breisgau, Deutschland


Isolating high-quality primary cardiomyocytes is essential for basic cardiac research, and the quality of the isolated cells profoundly affects experimental outcomes, throughput, and reproducibility. Commonly used Langendorff-based preparations require substantial expertise and skills, and cannot be applied to non-perfusable tissue fragments. Optimisation of the Langendorff-based protocols for project-specific needs is often lengthy and requires sacrificing a substantial number of animals (since only one protocol at a time can be tested). Especially in the view of the 3R initiatives (Replace, Reduce, Refine), the large numbers of animals often needed to establish successful cardiomyocyte isolation protocols raise ethical concerns. We recently developed a novel tissue slice-based isolation protocol which can yield high-quality adult cardiomyocytes from the same heart over multiple consecutive days, thus reducing animal use and experimental variability.1 Access to slices from the same tissue block creates a highly controlled environment where researchers can test multiple conditions on tissue from the same source in parallel. However, the laborious manual evaluation of cell after isolation is currently a bottleneck for developing new next-generation isolation protocols. This is particularly critical for human tissue, where tailoring the protocol to suit each donor and specific disease conditions is likely advantageous.

To resolve this bottleneck, we developed a deep learning-assisted workflow that automates the time-consuming cell quality assessment and provides extensive information about the preparation: cell counts and morphology (length, width), classification (rod-shaped vs rounded; responsiveness to electrical stimuli), and sarcomere length analysis. We imaged isolated cardiomyocyte suspensions using bright-field microscopy and automatically segmented individual cardiomyocytes using convolutional neural networks. This approach is robust even under challenging conditions such as cell crowding, presence of debris, and uneven illumination. Segmented cells can be quantitatively analysed with high throughput (100 cells/min vs hours of manual assessment). The automated pipeline requires no human input post-image capture and is compatible with low-cost equipment. 

Our workflow enables fast and robust improvements in cardiomyocyte isolation protocols, from both animal and human tissue. Better control over cell quality ensures more reliable and reproducible results, and a drastic decrease in the number of experimental animals. Our cell quality evaluation method is cost-effective, accessible, and versatile – and it can be adapted to various research areas, ranging from pharmacological to cell culture studies.

Figure: Exemplary readouts: automated segmentation, cell counts, and sarcomere length assessment. A: Automated cardiomyocyte counting and classification in challenging bright field images (yield rod-shaped cells: 65%). Cyan dots indicate individual cardiomyocytes, which are then colour-coded based on morphology (yellow: rod-shaped, red: not rod-shaped). B: Automated cardiomyocyte segmentation in bright field images. Colours indicate individual cells. C: Colour-coded regional sarcomere length distribution within a single cardiomyocyte, with quantification of regional intra-cardiomyocyte heterogeneities in sarcomere length in D. Scalebar: 200 µm.


1 Greiner, J. et al. (2022). Cells 11, 233.
Diese Seite teilen