Almost all common diseases, from diabetes to schizophrenia to arthritis to cancer, are complex. That is, these disorders are due to interactions of multiple genetic and environmental factors. Although the medical research community has been very successful in identifying the underlying causes of "simple" or Mendelian disorders, success in understanding complex disorders has been very limited. The few important successes primarily represent "low lying fruit", such as genes that were obvious candidates or that had unusually large effects. It is now clear that a vast majority of common diseases will result from interactions of genetic factors and environmental factors that each has modest effects. Given the enormous genetic heterogeneity and environment diversity of the human population, how can these factors best be identified? One approach is to try to simplify the problem using mouse models. The mouse is a key organism for understanding gene function in mammals. In particular, the production of highly penetrant mutations using gene targeting or mutagenesis is a well-established technique for understanding gene function. However, the aspects of gene function most relevant to common diseases very frequently only become apparent from the analysis of low penetrance mutations, whose effects may not be related to those of fully penetrant alleles. In order to dissect the contributions of low penetrance mutations to atherosclerosis, we have utilized naturally occurring genetic variation among inbred strains of mice. Beginning about 100 years ago, biologists have created genetically homogenous defined stocks of mice by inbreeding. At the present time, well over 100 different inbred strains are available and about 50 of these have now been completely sequenced. The range of naturally occurring variations among these strains is enormous. For example, inbred mice differ in physical and behavioral characteristics, susceptibility to many diseases, including cancer and atherosclerosis, and resistance to infection. As with the corresponding human traits, these variations in mice are usually determined by multiple genetic factors. Because the traits are frequently quantitative, the genetic loci contributing to these traits are called ''quantitative trait loci" or QTL. The effect of each QTL is usually quite small and, thus, they have been quite difficult to identify at the molecular level. Clearly, if we can identify the genes that underlie the complex traits at each of these loci, we would have a new way of understanding the biology of many phenotypes of medical importance. To tackle the very difficult problem of relating particular sequence variations to these complex phenotypes, we have developed approaches that integrate genetic segregation, complex trait phenotypes, and whole genome expression array analyses (Drake et al., 2006). Using these combined data, we have modeled biologic networks and made predictions about the involvement of novel genes in certain traits relevant to atherosclerosis (Doss et al., 2005; Schadt et al., 2005; Ghazalpour et al., 2006). We then tested these predictions using a series of transgenic experiments. The results have been highly encouraging, as a high fraction of the genes predicted to be involved in these traits have been confirmed. In these studies, gene expression serves as a kind of intermediary between genotype and the complex phenotypes. Using this combined approach, we have successfully identified several genes underlying QTLs. However, the focus of our efforts is not on uncovering single genes for the disorder, but rather on the interactions of genes operating in a complex multicellular biological system, as discussed below. The complex trait that we are studying is atherosclerosis, a disease of the large arteries that is the primary cause of heart disease of stroke (reviewed in Lusis, 2000). Atherosclerosis is the underlying cause of more than 50% of all deaths in the United States, and to understand fully the genetic and environmental factors that contribute to this disease is a very important problem. The disease involves an inflammatory response characterized by the accumulation of monocytes, macrophages, and lymphocytes just underneath the endothelial lining of the blood vessels. A number of systemic factors, including elevated cholesterol levels and high blood pressure, contribute to the development of this inflammatory process. Family studies and twin studies have suggested that about 50% of the variance in the disease can be explained by genetic factors and about 50% is environmental. Although some fairly effective therapies for atherosclerosis, such as the statins, which decrease blood cholesterol levels, are available, cardiovascular disorders associated with atherosclerosis remain, by far, the major cause of death in the US and other western countries. Indeed, atherosclerosis has also become the major cause of death in developing countries throughout the world. Much of what we have learned about atherosclerosis has come from studies of a number of relatively rare Mendelian forms of a disease, such as familial combined hypercholesterolemia, characterized by mutations of the low density lipoprotein receptor that is responsible for removal of the pro-atherogenic cholesterol carrying particles in the blood. Epidemiological studies have also taught us about the multitude of environmental factors that are important in the disease. These include well-known factors such as elevated levels of low-density lipoproteins, reduced levels of high-density lipoproteins, elevated blood pressure, lack of exercise, excess body fat, smoking, and a high fat diet. Other recent factors have also come to light, including maternal hypercholesterolemia during fetal development and air pollution. In a recent collaborative study with our colleagues Jesus Araujo and Andre Nel in the Department of Medicine, we have recently shown that Los Angeles air can significantly impact on the development of atherosclerotic lesions in our mouse models. Over the last 15-20 years, transgenic and gene targeting studies in mice have examined hundreds of "candidate genes" for their possible involvement in atherosclerosis (for example, Warden et al., 1993; Shih et al., 1998; Brennan et al., 2001). These studies have clearly shown that genes contributing to both systemic factors, such as blood cholesterol and blood pressure, as well as local factors, including inflammatory signals in the vessel wall, are important to the development to the disease. And these studies have provided a rough picture of the events that can lead to a myocardial infarction. In the early stages of lesion formation, the accumulation of lipoproteins in the vessel wall, followed by oxidative processes, results in formation of certain oxidized phospholipids that are highly pro-inflammatory (Rajavashisth et al., 1990). These oxidized phospholipids cause endothelial cells in the vessels to produce adhesion molecules for monocytes and lymphocytes, as well as a variety of chemotactic and growth factors (Liao et al., 1994). This, then, appears to result in recruitment of these blood cells to the vessel wall, where they take up cholesterol and produce a variety of cytokines that, among other things, result in the migration and proliferation of smooth muscle cells. Smooth muscle cells form a fibrous cap overlying the core of inflammatory cells, cholesterol, and necrotic debris. A heart attack is usually the result of the rupture of such a lesion, resulting in the formation of a clot, which blocks the flow of blood. Many important questions remain. Among those that we are particularly interested in studying are the following: How do oxidized lipids trigger inflammation in endothelial cells? Are there specific receptors for oxidized lipids, and what are the pathways by which inflammatory genes are induced? How do high-density lipoproteins protect against the development of atherosclerosis? How does diabetes accelerate the development of atherosclerotic lesions? What contributes to the stability of lesions? Why are males much more prone to coronary artery disease than females? Why are some very large lesions quite stable whereas much smaller lesions sometimes rupture to trigger a heart attack? A major effort in our lab at present is to construct biologic networks for each of the major cell types involved in atherosclerosis: macrophages, endothelial cells, and smooth muscle cells. We are constructing these networks by isolating these cells from selected inbred and recombinant inbred mouse strains and interrogating these using expression array analyses. The cells will be examined under basal conditions and following treatment with oxidized phospholipids, bacterial lipopolysaccharide, or other conditions relevant to atherosclerosis. The genetic variations among the inbred strains will result in perturbations in gene expression, allowing the grouping of genes which are similarly perturbed into "co-expression" networks. These networks will then be integrated with traits relevant to atherosclerosis, including lesion size and composition, vascular calcification, and metabolic parameters. The predicted networks will be validated and refined in vitro using small interfering RNA and in vivo transgenic approaches. Our studies, thus far, have suggested that pathways relating to inflammation and oxidation are particularly relevant (Wang et al., 2006). To test these pathways, we have created conditional knockout mice for key proteins, including macrophage colony stimulating factor and heme-oxygenase. At the same time that we are carrying out these studies in mice, we will also carry out certain parallel studies using human cells and with human families (Allayee et al., 1998; Aouizerat et al., 1999; Lusis et al., 2004). Recently, for example, we constructed biologic networks for human endothelial cells obtained from heart transplant donors (Gargalovic et al., 2006). These studies allowed us to construct biologic networks for the induction of inflammatory genes by oxidized phospholipids and they revealed that the unfolded protein response (UPR) pathway contributes importantly to inflammation in the context of atherosclerosis. We are now examining this by constructing conditional transgenic mice for key transcription factors in the UPR pathway. We are also constructing biologic networks for macrophages isolated from the blood of individuals that have been assessed for various parameters related to atherosclerosis, including intimal medial thickness, a measure of cardiovascular disease. Through collaborations, we are also pursuing genetic analyses of atherosclerosis in families and in populations. For example, using these studies in mice, we originally identified the gene for 5-lipoxygenase, an enzyme in the leukotriene biosynthetic pathway, as being involved in atherosclerosis development. Then, in collaboration with a group at the University of Southern California, we showed that variations of this gene are strongly associated with the development of atherosclerosis (Dwyer et al., 2004). In conclusion, traditional approaches, including candidate gene studies and positional cloning, have worked very well for Mendelian diseases, but they have not, in general, been successful for linking genomic variation to complex clinical phenotypes. Our approach to complex diseases emphasizes the utilization of functional genomic data resources. An important part of our strategy is to use mouse animal models, which are amenable to experimentation, alongside human studies. Another key aspect is the use of statistical modeling to generate biologic networks. Such integrative approaches should also be very valuable in the development of novel therapeutic strategies for the treatment of complex diseases. Since complex diseases such as atherosclerosis generally involve perturbations of multiple pathways, we suspect that useful strategies in the future will simultaneously target elements in different pathways to provide the greatest specificity and efficacy.