INTEGRATED ANALYSIS OF TRANSCRIPT-LEVEL REGULATION OF HUMAN ADIPOGENESIS AND CELL TYPE-SELECTIVE DISEASE ASSOCIATION OF HIGH REGULATORY LOAD GENES

Most diseases are characterized by altered epigenetic and metabolic states, pointing to the need of a global and combined understanding of mechanisms underlying epigenetic and metabolic changes as an important piece to enable disease eradication. Adipocytes impact systemic homeostasis and their differentiation encompasses a phenotypic change whose function becomes impaired with diseases such as obesity and metabolic syndrome. Following an integrative systems biology approach, we combined different omics data from the differentiation of Simpson-Golabi-Behmel syndrome (SGBS) adipocytes with a human metabolic model to observe key metabolic changes upon differentiation, their regulation and relevance for disease.
Pursuing the link to disease, we used public data from the genome-wide binding of TFs and location of active enhancers to test for disease association in function of the regulatory load, revealing a cell type-selective enrichment for disease of the high regulatory load genes.
Diverse experimental data were collected, covering a gene expression time-course during adipogenesis, with identification of miR-27a, miR-29a and miR-222 target genes, the genome-wide binding profiles of PPARg, C/EBPa and LXRa , and the H3K4me3 histone modification mark for actively transcribed transcription start sites (TSSs).
Metabolic genes showed a highly dynamic expression pattern during adipogenesis, most being targeted by PPARg and C/EBPa . Lipid metabolism pathways including triacylglyceride synthesis showed extensive and combinatorial regulation by TFs and miRNAs, converging on known dyslipidemia genes. For data visualization, we developed a web portal that interactively renders
metabolic pathways with omics data overlaid (IDARE, http://systemsbiology.uni.lu/idare.html).
Public ChIP-seq data revealed a general principle of higher disease association of genes under higher regulatory control in a cell type-selective manner. First, data from the genome-wide binding of 10 TFs in HUVEC cells showed an enrichment for vascular diseases on metabolic genes targeted by > 6 TFs. Second, data from the binding of a total of 93 TFs confirmed the enrichment for disease association of genes with the top TF load in 8 additional cell lines.
Finally, active enhancer data from 139 samples spanning 96 cell types and tissues demonstrated the cell type-selective disease enrichment of the genes with the highest active enhancer load.
High regulatory load genes enriched for disease association beyond genetic variation, including association types like “altered expression” and “biomarker”, among others.
Additionally, high regulatory load genes appeared on average in more KEGG pathways and had higher betweenness centrality in a liver disease network than other genes, showing longer 3’UTRs harboring more binding sites for diverse microRNA families, suggesting also a higher post-transcriptional regulatory load and a role as signal integrators within biological networks.
Our results point to the pertinence of including high regulatory load genes for unbiased prioritization of novel candidate genes for disease association.