Machine-learned Design and Bioxolography of Functional 3D Skeletal Muscle Tissues

Background. Engineered 3D skeletal muscle tissue (SMT) is an important tool to study muscle physiology and disease. SMT engineering has applications in regenerative medicine, in vitro drug screening, bio-hybrid robotics, and cultured meat. However, state-of-the-art engineered muscle tissue does not effectively mimic the cellular heterogeneity, architecture, and performance of biological muscle. To address this, researchers are developing techniques to bioprint, differentiate, and mature functional muscle architectures. Machine Learning (ML) approaches could streamline SMT design by rapidly exploring the ideal conditions for muscle biofabrication, computationally capturing the complex interplay between biofabrication parameters (bioprinting, differentiation, and mechanical and electrical maturation), and the resulting functionality of the contractile SMT. Objectives. This project consists of four key objectives: (1) Adaptation of xolography1 to muscle bioprinting. We will develop the bioxolography technique to fabricate highly aligned and anisotropic cell-laden hydrogels without constraints on the achievable shapes. (2) Biofabrication and differentiation of 3D heterocellular muscle constructs. We will develop a protocol to bioprint, culture, and differentiate arrays of muscle bundles which contain multiple cell types, thereby mimicking the natural muscle. (3) Maturation and characterization of engineered muscle tissues. We will mechanically stimulate our engineered muscle tissues for maturation and characterize their cellular morphology and contractile performance. (4) Development of an ML pipeline to guide the biofabrication and design of muscle actuators. Finally, we will develop an ML pipeline which maps biofabrication and tissue engineering parameters to performance metrics of engineered muscle (structure and function), leveraging a differentiable simulation approach originally developed to model soft material and actuator deformation in soft robotics. We will demonstrate proof-of-concept of this model by designing a centimeter-scale asymmetric, antagonistically actuated skeletal muscle construct.