Motivation: Primary purpose of modeling gene regulatory networks for developmental process is to reveal pathways governing the cellular differentiation to specific phenotypes. Knowledge of differentiation network will enable generation of desired cell fates by careful alteration of the governing network by adequate manipulation of cellular environment.

Results: We have developed a novel integer programming-based approach to reconstruct the underlying regulatory architecture of differentiating embryonic stem cells from discrete temporal gene expression data. The network reconstruction problem is formulated using inherent features of biological networks: (i) that of cascade architecture which enables treatment of the entire complex network as a set of interconnected modules and (ii) that of sparsity of interconnection between the transcription factors. The developed framework is applied to the system of embryonic stem cells differentiating towards pancreatic lineage. Experimentally determined expression profile dynamics of relevant transcription factors serve as the input to the network identification algorithm. The developed formulation accurately captures many of the known regulatory modes involved in pancreatic differentiation. The predictive capacity of the model is tested by simulating an in silico potential pathway of subsequent differentiation. The predicted pathway is experimentally verified by concurrent differentiation experiments. Experimental results agree well with model predictions, thereby illustrating the predictive accuracy of the proposed algorithm.

Contact: ipb1@pitt.edu

Supplementary information: Supplementary data are available at Bioinformatics online.

Embryonic stem cell-derived endoderm is critical for the development of cellular therapies for the treatment of disease such as diabetes, liver cirrhosis, or pulmonary emphysema. Here, we describe a novel approach to induce endoderm from mouse embryonic stem cells (mES) using fibronectin-coated collagen gels. This technique results in a homogenous endoderm-like cell population, demonstrating endoderm-specific gene and protein expression, which remains committed following in vivo transplantation. In this system, activin, normally an endoderm inducer caused an 80% decrease in the Foxa2 positive endoderm fraction, while follistatin increased the Foxa2 positive endoderm fraction to 78%. Our work suggests that activin delays the induction of endoderm through it transient precursors, the epiblast and mesendoderm. Long term differentiation, displays a two-fold reduction in hepatic gene expression and three-fold reduction in hepatic protein expression of activin-treated cells compared to follistatin-treated cells. Moreover, subcutaneous transplantation of activin-treated cells in a syngeneic mouse generated a heterogeneous teratoma-like mass, suggesting these were a more primitive population. In contrast, follistatin-treated cells resulted in an encapsulated epithelial-like mass, suggesting these cells remained committed to the endoderm lineage. In conclusion, we demonstrate a novel technique to induce the direct differentiation of endoderm from mES cells without cell sorting. In addition, our work suggests a new role for activin in induction of the precursors to endoderm, and a new endoderm-enrichment technique using follistatin.