Novel insights into embryonic stem cell self-renewal revealed through comparative human and mouse systems biology networks

Abstract

Embryonic stem cells (ESCs), characterized by their ability to both self-renew and differentiate into multiple cell lineages, are a powerful model for biomedical research and developmental biology. Human and mouse ESCs share many features, yet have distinctive aspects, including fundamental differences in the signaling pathways and cell cycle controls that support self-renewal. Here, we explore the molecular basis of human ESC self-renewal using Bayesian network machine learning to integrate cell-type-specific, high-throughput data for gene function discovery. We integrated high-throughput ESC data from 83 human studies (~1.8 million data points collected under 1,100 conditions) and 62 mouse studies (~2.4 million data points collected under 1,085 conditions) into separate human and mouse predictive networks focused on ESC self-renewal to analyze shared and distinct functional relationships among protein-coding gene orthologs. Computational evaluations show that these networks are highly accurate, literature validation confirms their biological relevance, and reverse transcriptase polymerase chain reaction (RT-PCR) validation supports our predictions. Our results reflect the importance of key regulatory genes known to be strongly associated with self-renewal and pluripotency in both species (e.g., POU5F1, SOX2, and NANOG), identify metabolic differences between species (e.g., threonine metabolism), clarify differences between human and mouse ESC developmental signaling pathways (e.g., leukemia inhibitory factor (LIF)-activated JAK/STAT in mouse; NODAL/ACTIVIN-A-activated fibroblast growth factor in human), and reveal many novel genes and pathways predicted to be functionally associated with self-renewal in each species. These interactive networks are available online at www.StemSight.org for stem cell researchers to develop new hypotheses, discover potential mechanisms involving sparsely annotated genes, and prioritize genes of interest for experimental validation.

Keywords: Biomathematical modeling; Cell signaling, Genomics; Computational biology; Embryonic stem cells; Pluripotent stem cells.

Figure 1. Bayes Net Machine Learning Methodology for Cell-Type-Specific Comparative Networks

To produce reliable, biologically relevant comparative predictive networks for mouse and human ESCs, we adapted our approach for cell-type-specific data integration and machine learning [25] as follows: 1. Prepared a master list of protein-coding mouse and human gene orthologs. 2. Curated a set of training examples focused on ESC self-renewal in a pairwise format. Positive (known) examples were extracted from recent literature and curated pathways; negative/background examples were randomly generated from the master gene list, excluding genes involved in positive edges. 3. Collected and standardized high-throughput data from 83 human and 62 mouse ESC studies using diverse high-throughput data types (Table 1). Used distance/correlation metrics to distill data into a pairwise format. 4. Iteratively tested and validated species-specific predictive networks for comparative analysis using the same training set, but with species-specific data compendiums as input.

Figure 2. Predicted FGF Signaling Pathway Relationships Across Species

These network models show the predicted strength of relationships between all known FGF signaling ligand-receptor pairs [65]. Gene nodes are colored by SRC score (light gray = weak SRC; dark blue = strong SRC) and edge color/thickness indicates the strength of predicted functional association between ligand and receptor (yellow/thin = low probability; teal/thick = high probability). A. The hESC FGF signaling model shows that the most strongly connected ligand-receptor pairs with high-ranking SRCs are FGF2 and FGF19 (the human ortholog of Fgf15), which are most strongly associated with FGFR1 and FGFR3. FGF2 is known to activate FGF signaling in hESCs, while FGF19 has been associated with neuronal development. B. The mESC FGF signaling model showed that the most strongly connected ligand-receptor pairs with high ranking SRCs are Fgf4 and Fgf5, which are both associated with Fgfr1 and Fgfr2, and Fgf15 which is most strongly connected to Fgfr2. In mESCs, Fgf4 is known to activate FGF signaling, Fgf5 is associated with FGF signaling in the late stage blastocyst and epiblast, and Fgf15 has been shown to be involved in early neurodevelopment.

Figure 3. Predicted JAK/STAT Signaling Pathway Relationships Across Species

Analogously to Figure 2, species-specific predictions are shown for a portion of the KEGG pathway map for JAK/STAT signaling focused on the IL-6 family of cytokines. A. The hESC predicted JAK/STAT signaling pathway showed that Stat3 is the gene most correlated to self-renewal, while upstream pathway participants exhibited lower SRC scores and weaker connection probabilities, suggesting that cytokines other than the IL-6 family, or signaling cross talk, may be required for STAT3 activation in hESCs. B. The mESC predicted JAK/STAT signaling pathway showed a strong connection between Lif and its putative target, Il6st. Further, key known mouse self-renewal genes in this pathway, including Lif, Il6st, and Stat3 are strongly correlated to other self-renewal genes as indicated by SRC score.

Figure 4. Differences in mESC and hESC Threonine Metabolism

We used our StemSight Scout data visualization tool (www.StemSight.org) to create network views centered around L-threonine dehydrogenase (TDH|Tdh), which supports accelerated cell cycle kinetics in mESCs, but is not functional in hESCs. Node and edge colors are as in Figures 2 and 3, except that edges contained in the positive training set were colored orange. A. The hESC TDH-centric network shows that TDH is weakly correlated to genes in our training set and has no strong functional associations to any known self-renewal genes. B. The mESC Tdh-centric network illustrated that Tdh is strongly correlated to self-renewal genes and had strong predicted functional associations with known self-renewal genes, including Pou5f1, Sox2, Nr0b1, Klf2, Zfp42, Gdf3, and Fbx015.