A first line of attack on diabetes and cardiovascular disease is a reduction in body mass. A consistent finding across many different diet studies is that carbohydrate (CHO) restriction has a central role in facilitating weight loss and improving features of metabolic syndrome 12, due to mechanisms related to metabolic efficiency 34 and factors related to appetite regulation 56.

Genetic factors interact with dietary nutrients to impact the development of obesity and the outcome of weight loss therapies. The etiology of obesity is complex and can result from a disruption in functioning of diverse but interconnected pathways. Researchers have shown that polymorphisms in several different genes play a role in determining weight loss or weight maintenance in response to various pharmacological and non-pharmacological therapies. The approach taken in these genetic studies has been to separate individuals based upon allelic variation is a candidate gene and determine if weight loss or better maintenance of body mass varies as a function of the polymorphism. For example, body mass responses to various therapies have been linked to genes coding for products involved in the sympathetic nervous system 7891011, appetite regulating hormones 1213, adipose tissue transcription factors 1415, and proteins regulating fat digestion, deposition and mobilization 151617.

In this study we use physiogenomics 18, a medical application of sensitivity analysis and systems engineering. Sensitivity analysis is the study of the relationship between input and output from a system as determined by each system component. Physiogenomics utilizes the genes as the components of the system. The gene variability, measured by single nucleotide polymorphisms (SNPs), is correlated to physiological responses of a population, the output. Physiogenomics determines how the SNP frequency varies among individuals similarly responding to the input over the entire range of the response distribution.

Scrutiny of weight loss responses for subjects who have consumed CHO restricted diets in our laboratory revealed a rather large amount of variability in the magnitude of weight loss 1920. This variability is not readily explained by standard covariates such as caloric intake, gender, age, activity, etc. Therefore a physiogenomic approach was undertaken using families of candidate genes, selected from those hypothesized to be involved in the metabolic adaptations induced by CHO restriction in the treatment of obesity. The results indicate that the magnitude of weight loss induced by a CHO restricted diet is, in part, explained by polymorphisms in specific genes among those we selected to study: genes that regulate lipases, intracellular glucose metabolism, HDL homeostasis, and appetite hormones.

Table 1 summarizes the weight loss data available for the study population. The distribution of weight loss was non-gaussian (Figure 1). Of the potential covariates listed in Table 3, length of diet, body mass index, and baseline total cholesterol were significantly associated with weight loss. Length of diet, in particular, accounted for an increased weight loss of 0.55 kg with each additional week of diet. Increased baseline body mass index (BMI) was also correlated with weight loss. There was a relationship between total cholesterol level and weight loss, with an additional 19 grams of weight lost for every mg/dl of total cholesterol at the beginning of the study.

After adjusting for the associations in Table 3, each SNP in Table 2 was tested for association with the residual variable (body mass adjusted for the covariates) (Table 4). Of the 27 SNPs, four had a statistically significant association with the residual body mass, namely gastric lipase (LIPF, SNP rs814628), cholesteryl ester transfer protein (CETP, SNP rs5883), hepatic glycogen synthase 2 (GYS2, SNP rs2306179), and galanin (GAL, SNP rs694066). The first three results remained significant when adjustment was made for the testing of multiple SNPs using the FDR method.

Figures 2, 3, 4, 5 show a detailed representation of the genetic association tests for all genes. The overall distribution of change in body mass is shown along with the individual genotypes and a LOESS fit of the allele frequency as a function of body mass. The bell curve shows the fitted distribution of body mass phenotype in the clinical studies. The LOESS curve shows the localized frequency of the least common allele for sectors of the distribution. For SNPs with a strong association, the marker frequency will be significantly different between the high end and the low end of the distribution. Conversely, if a marker is neutral, the frequency will be independent on the body mass and the LOESS curve will be essentially flat.

For example, the first panel in Figure 2 shows the LOESS curve for SNP rs814628, which is located in the gene for gastric lipase (LIPF). The frequency of the minor allele approaches 60% at the highest amount of weight loss (left of the distribution) whereas it is 5% in subjects with little body mass change (right of the distribution). The overall frequency of this SNP in the study population is 10% (Table 2). This marked dependence of SNP frequency with the phenotype is indicative of a strong association between the gene marker and body mass, as attested by a p value of 0.0002 (Table 4).

The principal metabolic adaptations contributing to weight loss induced by CHO restriction are unknown. This study used physiogenomic analysis to examine the relations between genes regulating target proteins impacting the intake and metabolism of dietary nutrients and weight loss in subjects on a very low CHO diet. The change with length of diet is a good confirmation that it is indeed the low CHO intervention that caused the weight loss, although it is unknown if similar changes would be found if weight loss were brought about by other types of diets. Genetic polymorphisms also significantly associate with weight loss when these factors are used as a covariate indicating that genetic variation may be a valuable tool to predict individual variability in weight loss to dietary CHO restriction. The results indicate that common genetic markers in gastric lipase, glycogen synthase, CETP, and galanin have a substantial effect on weight loss response to CHO restricted diets.

Very low CHO diets are typically higher in fat. We therefore hypothesized that polymorphisms in various lipases may account for variability in weight loss. We tested SNPs in genes for gastric, hepatic, lipoprotein, hormone-sensitive, lysosomal acid, and endothelial lipase. Unexpectedly, we discovered the gastric isoenzyme was the most significant genetic association to weight loss, whereas other lipases were not. Gastric lipase is secreted by the mucosa of the stomach and hydrolyzes the ester bonds of dietary triglycerides in the gastrointestinal tract. Pancreatic lipase is generally the dominant enzyme in the hydrolysis of gastrointestinal lipids, but gastric lipase can contribute significantly especially in instances where pancreatic lipase is deficient 34. Thus a significant locus of variability is the ability to perform first pass breakdown of dietary fat. Individuals with the least common version of the enzyme had the most weight loss, indicating that impairment of gastric fat breakdown on a low CHO diet enhances weight loss.

The second category of genes we surveyed was related to glycogen synthesis. Glycogen synthase catalyzes the formation of glycogen from glucose. A defect in this pathway in skeletal muscle has a dominant role in the insulin resistance that occurs in diabetes 35. Polymorphisms in glycogen synthase kinase beta (GSK3B), a regulator of glycogen synthase activity, and skeletal muscle glycogen synthase 1 (GYS1) have been examined in several studies, but have generally failed to associate with diabetes or measures of insulin resistance 3637. In addition to GSK3B and GYS1, we examined hepatic glycogen synthase 2 because recent work also indicates that insulin stimulated hepatic glycogen synthesis is impaired in diabetics 38. We discovered that a polymorphism in hepatic, but not skeletal muscle, glycogen synthase was associated with weight loss. The results suggest that the hepatic response to carbohydrate restriction may influence the weight loss response to a low CHO diet.

In addition to weight loss, our prior work has shown that low CHO diets result in reliable and dramatic changes in lipoprotein metabolism characterized by decreased triglycerides and remodeling of LDL and HDL cholesterol to form larger particles 122. Since weight loss has similar effects, we surveyed various apolipoproteins and enzymes regulating triglyceride and lipoprotein metabolism including CETP, apolipoprotein A-I, and apolipoprotein C-III. We found that a specific polymorphism in plasma CETP was significantly associated with weight loss. The major function of CETP is the net mass transfer of cholesterol esters from HDL to triglyceride-rich lipoproteins and LDL-C and of triglyceride-rich lipoproteins to HDL-C and LDL-C 39, thereby providing a mechanistic link to explain the triglyceride lowering and remodeling effects of LDL and HDL observed with low CHO diets 2240. Several studies have linked polymorphisms in the CETP gene to lipoprotein responses and risk for cardiovascular disease, and it has been hypothesized that these relations may be altered by weight loss 41. Our study is the first study to show an association of a polymorphism in CETP gene to weight loss. The finding suggests that the weight loss response to CHO restriction may be mechanistically linked to the intravascular processing of lipoproteins.

Hormonal regulation of food intake was hypothesized to be one mechanism by which CHO restricted diets affect weight loss. We examined polymorphisms in galanin, neuropeptide Y, and ghrelin. Galanin was the only hormone significantly associated with weight loss. Galanin stimulates food consumption, particularly fat intake. A prior study that measured polymorphisms in galanin failed to find an association with fat intake or obesity 42. Prior work has shown that galanin in the para-ventricular nucleus is stimulated by a fat feeding and increased circulating triglycerides, which in turn promotes further fat consumption in a non-homeostatic feed-forward manner 43. The finding in this study that a polymorphism surveyed in the galanin gene was associated with weight loss provides evidence for a role of fat-mediated appetite hormones in determining the response to carbohydrate restriction.

Physiogenomics introduces a new paradigm in the genetic analysis of complex phenotypes. Historically, a candidate gene approach identified one specific hypothesis. However, such focused hypotheses are often unrealistic given the number of overlapping pathways at the organismic, cellular, and molecular levels. Array technologies provide efficient methods to simultaneously probe large numbers of genes using general hypotheses about entire pathways and systems. As a previous example of this approach, we had demonstrated a strong association between CK activity during statin treatment and variability in genes related to vascular function, angiotensin II Type 1 receptor (AGTR1) and nitric oxide synthase 3 (NOS3) 44. This finding had led us to suggest the novel hypothesis that vascular smooth muscle function may contribute to the muscle side effects of statins.

Similarly, we believe that novel hypotheses have been generated in this study. The approach to select gene families or functionally related genes generates positive and negative results for physiogenomic analysis. It is the contrast in statistical significance levels within each of the four functional groups pursued in this study that provides the mechanistic insight. The associated gene markers can be combined into SNP ensembles harnessing their combined predictive power to estimate weight loss attainable from carbohydrate restriction for each individual. The SNP ensemble can then be tested retrospectively or prospectively to assess its predictive diagnostic power in populations separate from the ones used to generate the model. We believe this approach is pivotal to the discovery of multi-gene effects determining human dietary response. Applications to the management of obesity and diabetes include individualized counseling and dietary choice based on innate capacity to react to various nutritional regimens. We foresee the translation of these findings to diagnostic systems for personalized diet.

Dr. Ruano, Mr. Kocherla and Dr. Windemuth are full-time employees and shareholders of Genomas Inc.

JSV, MLF, WJK were involved with the acquisition of funding, design, conception and execution of the dietary intervention and biochemical analysis portion of the study, as well as acquisition, analysis and interpretation of data. CEF and RJW were primarily involved in data acquisition, analysis and interpretation of data. GR, AW, and MK performed the genetic analysis. GR, TH, and AW performed the statistical analysis. All authors were involved in the drafting and revision of the manuscript and give their final approval of this version to be published.