Comparative functional genomics of energy metabolism and insulin resistance in mammalian systems

Type 2 diabetes mellitus is a highly prevalent human disease often preceded by a period of hyperinsulinemia and insulin resistance. In both pathologic states, tissues involved in energy storage and metabolism demonstrate impaired physiological responses to insulin. Several cell and animal models have been developed to study the action of insulin and its effects in key metabolic tissues. In order to identify commonalities and differences between these model systems and relate these back to human disease, we have undertaken a unique functional genomics approach. Gene expression microarrays allow the simultaneous measurement of thousands of transcripts. To demonstrate normal physiology in a variety of systems, we apply expression profiling to characterize the effects of insulin in the tissues of C57/BL6 mice, lean Zucker rats, and lean human subjects. We further compare these results with previously published data in 3T3-L1 adipocytes. Comparative analysis across several organisms provides novel insights into the transcriptional responses of three key tissues: skeletal muscle, adipose, and liver. In addition, we apply a similar approach to study models of insulin resistance. In 3T3-L1 adipocytes, we revisit TNF- alpha-induced resistance and compare this with several whole-organism models. In order to interpret the volumes of data collected from these experiments, however, we need a reliable and efficient tool to effectively manage this data. Interpretation of high-throughput expression data is a bioinformatics problem requiring careful integration of statistics, mathematical modeling, computational infrastructure, and biology. As gene expression and promoter microarrays become increasingly prevalent, there is a greater need for precise and accessible analytical tools. These tools not only need to encapsulate mathematically robust methods, but also be simple enough so that statistical details can be sufficiently separated from biological interpretation. Utilizing a Bayesian framework, we have developed such an approach, dubbed VAMPIRE, and implemented it as a web-based, database- driven application. The tools presented here represent a unique solution to the interpretation of gene expression data. These advances have subsequently allowed us to piece together a comprehensive picture of the physiological responses to prolonged hyperinsulinemia, and to suggest novel targets for further study of type 2 diabetes