Embryonic stem cells (ESCs) have the potential to be used in many therapeutic applications due to their ability to differentiate into cells of any of the three primary germ layers (ectoderm, endoderm, and mesoderm). There are several different routes by which pluripotent ESCs can be guided to differentiate into specific germ layers, including genetic manipulation, chemical cues, and mechanical stimuli. With the latter, it has been observed that substrate stiffness plays a significant role in governing phenotype specific differentiation.
In our previous study we investigated the effect of mechanical stiffness of fibrin, selected as a model natural polymer based substrate, on ESC differentiation. Our results indicated that while ectoderm and mesoderm germ layers responded weakly to the change in fibrin substrate stiffness in the chosen range (2 – 250KPa), endoderm markers were strongly responsive, with the softer substrates up-regulating endoderm specific markers. Although cells respond differently to substrates of varying stiffness, this macroscopic property is perhaps not what the cell truly experiences; modification of stiffness arises from changes in the gel microstructure which also directly interacts with the cells. However, it is not clear what specific microstructural features are the most influential in inducing cellular differentiation, and how they affect ESC behavior.
In the current work, we have developed an integrated experimental and computational approach to investigate the effect of microstructural features of fibrin gels on the differentiation of mouse ESCs. Twelve different fibrin gels were fabricated by varying the fibrinogen concentration (1, 2, 4, and 8 mg/mL) and fibrin to thrombin crosslinking ratio (10, 2.5, and 1.25). The fibrin gels were used to induce ESC differentiation employing both 2D and 3D cultures. For the former, cells were seeded on top of the fibrin gel; for the 3D culture, cells were suspended within a fibrinogen solution before the addition of thrombin to induce subsequent polymerization. After the differentiation protocol the ESCs were analyzed for phenotypic commitment by performing qRT-PCR for the specific germ layer markers. Each of the 12 different fibrin gels was imaged with Scanning Electron Microscopy (SEM). Microstructural features of each of these gels were quantified using an image analysis tool developed at the University of Pittsburgh  for the characterization of fibrous scaffolds. Specific features which were characterized include fiber diameter, fiber orientation, pore aspect ratio, porosity, and pore size distribution. Such characterization enabled correlation of gel macroscopic properties with its microstructural features. In the next step the gel microstructural features were correlated with the ESC differentiation pattern using regression analysis. Two variants of least-squared regression, the LASSO and Ridge Regression, were utilized in which penalties are imposed on regression coefficients with high values, a method known as shrinkage. The variability of the system is handled by bootstrapping which creates the feature space for each regression. Performing regression at each of the bootstrap points results in a range of correlation coefficients from which confidence intervals of the parameters were obtained. This analysis reveals the sensitivity of cellular phenotype commitment and differentiation patterning to each of the microstructural features and paves the way for conducting further mechanistic study. Moreover, such information can be used to help guide the design of scaffolds with specific properties for tissue engineering applications.
 A. D’Amore et al. “Characterization of the complete fiber network topology of planar fibrous tissues and scaffolds.” Biomaterials, Vol. 31, 2010, 5345-5354.