The N-cadherin (N-cad) complex plays a crucial role in cardiac cell structure and function. Cadherins are adhesion proteins linking adjacent cardiac cells and, like integrin adhesions, are sensitive to force transmission. Forces through these adhesions are capable of eliciting structural and functional changes in myocytes. Compared to integrins, the mechanisms of force transduction through cadherins are less explored. α-catenin is a major component of the cadherin-catenin complex, thought to provide a link to the cell actin cytoskeleton. Using N-cad micropatterned substrates in an adhesion constrainment model, the results from this study show that α-catenin localizes to regions of highest internal stress in myocytes. This localization suggests that α-catenin acts as an adaptor protein associated with the cadherin mechanosensory apparatus, which is distinct from mechanosensing through integrins. Myosin inhibition in cells bound by integrins to fibronectin-coated patterns disrupts myofibiril organization, whereas on N-cad coated patterns, myosin inhibition leads to better organized myofibrils. This result indicates that the two adhesion systems provide independent mechanisms for regulating myocyte structural organization.

Mapping gene networks requires large amounts of transcriptomic data to learn the connections between genes, which impedes discoveries in settings with limited data, including rare diseases and diseases affecting clinically inaccessible tissues. Recently, transfer learning has revolutionized fields such as natural language understanding1,2 and computer vision3 by leveraging deep learning models pretrained on large-scale general datasets that can then be fine-tuned towards a vast array of downstream tasks with limited task-specific data. Here, we developed a context-aware, attention-based deep learning model, Geneformer, pretrained on a large-scale corpus of about 30 million single-cell transcriptomes to enable context-specific predictions in settings with limited data in network biology. During pretraining, Geneformer gained a fundamental understanding of network dynamics, encoding network hierarchy in the attention weights of the model in a completely self-supervised manner. Fine-tuning towards a diverse panel of downstream tasks relevant to chromatin and network dynamics using limited task-specific data demonstrated that Geneformer consistently boosted predictive accuracy. Applied to disease modelling with limited patient data, Geneformer identified candidate therapeutic targets for cardiomyopathy. Overall, Geneformer represents a pretrained deep learning model from which fine-tuning towards a broad range of downstream applications can be pursued to accelerate discovery of key network regulators and candidate therapeutic targets.

Truncating mutations in the sarcomere protein titin cause dilated cardiomyopathy due to sarcomere insufficiency. However, it remains mechanistically unclear how these mutations decrease sarcomere content in cardiomyocytes. Utilizing human induced pluripotent stem cell-derived cardiomyocytes, CRISPR/Cas9, and live microscopy, we characterize the fundamental mechanisms of human cardiac sarcomere formation. We observe that sarcomerogenesis initiates at protocostameres, sites of cell-extracellular matrix adhesion, where nucleation and centripetal assembly of α-actinin-2-containing fibers provide a template for the fusion of Z-disk precursors, Z bodies, and subsequent striation. We identify that β-cardiac myosin-titin-protocostamere form an essential mechanical connection that transmits forces required to direct α-actinin-2 centripetal fiber assembly and sarcomere formation. Titin propagates diastolic traction stresses from β-cardiac myosin, but not α-cardiac myosin or non-muscle myosin II, to protocostameres during sarcomerogenesis. Ablating protocostameres or decoupling titin from protocostameres abolishes sarcomere assembly. Together these results identify the mechanical and molecular components critical for human cardiac sarcomerogenesis.