obesity and its associated comorbidities provide powerful examples of the complexity of gene-environment interactions in which the multiplicity of phenotypic responses that result from environmental challenges are significantly influenced by underlying genetic variations (1, 12, 13). While the development of obesity is closely linked to the consumption of diets high in calories and fat, not everyone exposed to such a diet becomes obese. Furthermore, since obesity can co-occur with insulin resistance, dyslipidemia, and hypertension often enough to warrant the term “metabolic syndrome,” these additional phenotypic responses provide important distinctions of otherwise similarly obese individuals (7–9). This phenotypic diversity suggests that understanding obesity susceptibility alone will not be sufficient to uncover the genetic variants that are predictive of its comorbid diseases nor of the optimal or tolerable biological states for genetically diverse individuals. Inbred strains of mice are emerging as a valuable resource to dissect genetic interactions that mediate sensitivity to diet-induced obesity from those that drive susceptibility to diseases like atherogenesis and insulin resistance. Mice, like humans, exhibit varying sensitivity and resistance to both diet-induced obesity and comorbidities (5). Using genetically controlled animal models, Shockley et al. (11) have exploited a set of 10 inbred strains with established differences in atherogenic and obesogenic responses to a Western-style high-fat diet and here have profiled diet-induced changes in hepatic gene expression. Diet-induced gene expression patterns ranged from being largely independent of genetic variation to those that were highly modified by strain, sex, as well as tightly correlated to quantitative phenotypic responses. The results provide a powerful new resource to explore the relative covariance of genetic variations, genomic expression pattern signatures, and a battery of pathologic and physiologic processes. Exploiting these kinds of resources should catalyze the discovery of underlying pathways, disease mechanisms, and translational implications for individuals with variant genetics attempting to optimize their health in challenging environments.
By characterizing diet-induced patterns of gene expression and correlating these to genetic variation of strain, sex, and phenotypic susceptibilities and quantitative responses of circulating triglycerides, cholesterol, HDL, and glucose as well as body mass, Shockley et al. (11) have provided large-scale definition of gene regulatory associations associated with diet responses. Both males and females were included in the study, creating a total of 40 strain*diet*sex experimental groups. Microarray hybridizations were performed on a total of 120 liver RNA samples representing three mice/experimental group, producing a wealth of data that the authors mined to identify robust and strain-, sex-, and strain*sex-dependent responses to the diet. Global analysis of the expression data by ANOVA revealed extensive impact of both strain and sex on dietary responses in liver. The main effect of diet was apparent in the differential expression of almost 57% of genes on the arrays. Among the top co-regulated of these, the authors showed using Gene Ontology (GO) enrichment analyses that the expected effects of diet-mediated suppression of cholesterol biosynthesis occurred in all strains and that other specific biological systems functional involvements were also strongly implicated. Valuable data emerge from the ability to identify sets of genes that are reflective of strain-, sex-, and strain*sex-dependent effects of diet, attributable in part from the statistical power of the study design in which the underlying genetic variations imparted systematic effects at the level of biological pathways and processes. When strain is taken into account, the majority of significantly enriched GO categories reflected a powerful correlation between effects of diet on immune function, such as antigen processing and presentation. The authors suggest that these changes may be related to diet-induced liver damage, which was indicated by changes expression of damage-related genes and by an increase in circulating levels of glutamate dehydrogenase. However, these changes may also broadly reflect genetic correlation between pathways that mediate immune responses and the development of diet-induced atherogenic plaques, the formation and propagation of which are driven by immune responses. Integrating the expression data with information about the incidence, severity, and phenotypic details of atherosclerotic lesions in this same set of strains should shed additional light on this possibility. Other important follow-on studies may also determine if strain-specific patterns of immune gene expression in liver are paralleled by qualitatively similar effects on lymphocyte populations, atherogenic lesions, and insulin pathway-associated tissues.
Importantly, the authors have also provided a publicly accessible database that can be mined in a number of ways. For example, this resource could be used to select strains with divergent responses to diet (e.g., sensitive and resistant to liver damage) or to identify coexpression networks centered around a gene(s) of specific interest. It could also be used to extract supportive evidence for genes implicated in processes related to lipid metabolism or atherogenesis. The authors demonstrate its utility in narrowing lists of quantitative trait locus (QTL) candidate genes to those most plausibly related to the phenotype. As proof of principle, a set of 91 potential candidate genes in a QTL from a previous study was narrowed to two genes by iterative filtering based on differential expression, correlation with relevant phenotypes, tissue specificity, and block haplotype mapping.
As outside users of the database, we used the website to extract probe sets correlated with either plasma cholesterol level and/or total body mass. By analyzing the most highly correlated probe sets using Geneset Enrichment and network analysis methods [for example by using the Toppgene Server (2)], one could easily confirm and extend the authors' results to identify extensively regulated biological processes that correlated with the observed physiological and pathophysiological traits and parameters. The network shown in Fig. 1 demonstrates that progressively increased cholesterol levels are highly correlated to genomic responses that are tightly linked to dysregulation of inflammation, immunity, and response to xenobiotic agents.
When the potential opportunities for mining this database are considered, it is also important to keep in mind its limitations. As is typical for large-scale platforms such as microarrays, the gene expression profiles are captured in a single tissue and at a single time point after the diet intervention. Whether differences in expression at this specific time point are reflective of disease susceptibility or downstream of and secondary to those that initiate pathogenesis is difficult or impossible to determine without the addition of longitudinal data collected across multiple time points. Similarly, the extent to which expression changes in liver predict distal disease endpoints in the vasculature remains open for study. Nonetheless, the data do provide the ability to stratify the strain panel in a number of ways for many different kinds of future studies. The relatively tight clustering of individuals within a strain compared with between strains suggests that the heritable influence on expression is significant, which bodes well for future selection of strains, pathways, and biomarkers that could reflect specific therapeutically addressable pathways of physiologic and pathophysiologic significance.
Inbred strain surveys of environmental responses present a powerful means to advance the discovery of gene-environment interactions relevant to disease. The finding that an atherogenic diet unmasked a spectrum of genetic differences that were relatively silent in mice consuming control chow highlights the need for population-based models in studies of environmental exposures. Population-based models, such as the Mouse Phenome strain collection (6) (of which the panel used by Shockley et al. is a subset) and recombinant inbred strain sets [e.g., BXD (10), Collaborative Cross (3, 4)], offer the means to detect both robust effects of environment and those that vary between individuals and potentially mediate differential sensitivity. Integration of genetic mapping resources in strain panels that are sufficiently large and diverse adds the further opportunity to define the genetic architecture that underlies variable environmental responses between individuals. Although the inclusion of multiple strains proportionately increases time, budget, and labor, these types of studies provide a rich new fuel for discovering and understanding population variations, heritable differences, and relevant pathways that shape individual responses to environmental factors such as diet. Exploration, evaluation, and translation of this emerging network of data and knowledge have the potential to advance greatly the emerging field of personalized medicine. Our challenge is to use and translate from these models and ultimately to identify the most instructive biomarkers, therapeutics, and preventives for individuals who share critical responses to variant genetics and environmental factors so as to best implicate underlying disease etiologies, treatments, and preventions. As we build and characterize multifactorial disease models over deeper panels of genetic diversity, we should acquire increasingly more powerful abilities to translate our understanding to improve disease treatments and preventions.
- Copyright © 2009 the American Physiological Society