Search

Hyper proliferation of vascular smooth muscle cells (VSMCs) contributes to neointima formation in atherosclerosis and the response to vascular injury. Understanding how to control VSMC proliferation would advance the effort to treat vascular disease. Drug responses are often different among patients with the same vascular disease condition, making it difficult to cater patient-specific treatments (existence of heterogeneity). Thus, we examined variations in response to drug treatments using VSMC spheroids that mimic vascular disease condition in vivo. FAK and its downstream Rho GTPases (Rac, Rho, and Cdc42) control cell-cell contact and play a key role in vascular pathology. Here, we tested the importance of FAK and Rho GTPases in spheroid formation. VSMC spheroids were made using a hanging-drop method with either inhibitors or vehicle control. Changes in morphology were used as a key indicator of spheroid response to drug treatment. A machine learning (ML) image segmentation (VGG16-U-net) was used to segment the spheroid images. After the morphological features were extracted from the segmented images, a two-level cluster framework was used to cluster VSMC spheroids into different morphologies. We found that FAK and Rho GTPases are required for normal spheroid formation. Next, we analyzed the various morphologies of disrupted spheroids resulting from drug treatment using our ML pipeline. The first-level clustering analysis showed the presence of 4 clusters of spheroids with rounded and disrupted morphologies. The subsequent second-level clustering analysis identified the presence of four distinct morphological clusters among these disrupted spheroids, and they exhibited differential responses to FAK and Rho GTPases inhibition. Particularly, we found FAK and Rho specific morphological phenotypes, thus suggesting that there may be two distinct pathways governing VSMC spheroid formation. Collectively, we revealed there are significant heterogeneities in the drug responses of spheroid formation, which were overlooked in previous analyses. This is our first step towards developing the ML-based method that can be used to assess the effects of different drugs on the VSMC spheroid model for better characterization of pathologic progression of vascular disease.

Quantitative live cell imaging has been widely used to study various dynamical processes in cell biology. Phase contrast microscopy is a popular imaging modality for live cell imaging since it does not require labeling and cause any phototoxicity to live cells. However, phase contrast images have posed significant challenges for accurate image segmentation due to complex image features. Fluorescence live cell imaging has also been used to monitor the dynamics of specific molecules in live cells. But unlike immunofluorescence imaging, fluorescence live cell images are highly prone to noise, low contrast, and uneven illumination. These often lead to erroneous cell segmentation, hindering quantitative analyses of dynamical cellular processes. Although deep learning has been successfully applied in image segmentation by automatically learning hierarchical features directly from raw data, it typically requires large datasets and high computational cost to train deep neural networks. These make it challenging to apply deep learning in routine laboratory settings. In this paper, we evaluate a deep learning-based segmentation pipeline for time-lapse live cell movies, which uses small efforts to prepare the training set by leveraging the temporal coherence of time-lapse image sequences. We train deep neural networks using a small portion of images in the movies, and then predict cell edges for the entire image frames of the same movies. To further increase segmentation accuracy using small numbers of training frames, we integrate VGG16 pretrained model with the U-Net structure (VGG16-U-Net) for neural network training. Using live cell movies from phase contrast, Total Internal Reflection Fluorescence (TIRF), and spinning disk confocal microscopes, we demonstrate that the labeling of cell edges in small portions (5∼10%) can provide enough training data for the deep learning segmentation. Particularly, VGG16-U-Net produces significantly more accurate segmentation than U-Net by increasing the recall performance. We expect that our deep learning segmentation pipeline will facilitate quantitative analyses of challenging high-resolution live cell movies.