Ultrastructure Analysis of Cardiomyocytes and Their Nuclei

Abstract

Cardiomyocytes are elongated and densely packed in the mammalian heart and connected end on end to achieve a functional syncytium. Qualitative accounts describe their nuclei as being elongated in their long axis direction, which might help to better distribute mechanical load and reduce mechanical stress during contraction. Alterations of nuclear orientation and shape have also been known to be associated with certain cardiomyopathies. Yet, to date, the alignment of cardiomyocytes and their nuclei at the cellular (micron) scale has not been assessed in a quantitative fashion. To examine this we developed 3D computer vision methods to segment myocytes and their nuclei in cleared and membrane stained thick 3D tissue sections from a wild type mouse heart, imaged using confocal microscopy. We extended a geometric flow based superpixel algorithm to 3D and then adaptively merged the resulting supervoxels to recover individual myocytes. In parallel we also applied recent popular deep learning based cell segmentation methods to the same data. Our experiments revealed a close alignment of myocyte orientation with nucleus orientation, with a median difference of approximately 10◦, and also showed that most cardiomyocytes contain only one or two nuclei. These findings pave the way for future investigations of the effect of specific cardiac diseases on nuclear shape, elongation and number.