Application of Combined 'omics Platforms to Accelerate Biomedical Discovery in Diabesity

Resources

Software programs mentioned in the presentations

Cytoscape
Open source bioinformatics software platform for visualizing molecular interaction networks and integrating these interactions with with annotations, gene expression profiles and other state data.

Metscape 2
NCIBI Metscape 2 is a plugin for Cytoscape used to visualize and analyze metabolomic data.

Clinical studies mentioned in the presentations

STRRIDE
Slentz CA, Aiken LB, Houmard JA, et al. Inactivity, exercise, and visceral fat. STRRIDE: a randomized, controlled study of exercise intensity and amount. Am. J. Cardiol. 2007;100(12)_1759-1766.

Framingham Study

CATHGEN
Repository of DNA, RNA, and plasma samples with clinical information to investigate the relationships between genes and cardiovascular disease.

Generating Testable Hypotheses from 'Omics Data

Karnovsky A, Weymouth T, Hull T, et al. Metscape 2 bioinformatics tool for the analysis and visualization of metabolomics and gene expression data. Bioinformatics 2012;28(3):373-80.

Subramanian A, Tamayo P, Mootha VK, et al. Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proc. Natl. Acad. Sci. USA 2005;102(43):15545-15550.

Sartor MA, Leikauf GD, Medvedovic M. LRpath: a logistic regression approach for identifying enriched biological groups in gene expression data. Bioinformatics 2009;25(2):211-217.

Xu J, Gowen L, Raphalides C, et al. Decreased hepatic futile cycling compensates for increased glucose disposal in the Pten heterodeficient mouse. Diabetes 2006;55(12):3372-3380.

Xu J, Trujillo C, Chang V, et al. Peroxisome proliferator-activated receptor alpha (PPARalpha) influences substrate utilization for hepatic glucose production. J. Biol. Chem. 2002;277(52):50237-50244.

Yang L, Vaitheesvaran B, Hartil K, et al. The fasted/fed mouse metabolic acetylome: N6-acetylation differences suggest acetylation coordinates organ-specific switching. J. Proteome Res. 2011;10(9):4134-4149.

Unraveling the Link between Obesity and Diabetes

Newgard CB, An J, Bain JR, et al. A branched-chain amino acid-related metabolic signature that differentiates obese and lean humans and contributes to insulin resistance. Cell Metab. 2009;9(4):311-326.

Laferrere B, Reilly D, Arias S, et al. Differential metabolic impact of gastric bypass surgery versus dietary intervention in diabetic subjects despite identical weight loss. Sci. Transl. Med. 2011;3(80):80re2.

Ferrara CT, Wang P, Neto EC, et al. Genetic networks of liver metabolism revealed by integration of metabolic profiling. PLoS Genet. 2008;4(3):e1000034.

Unraveling the Link between Diabetes and Cardiovascular Disease

Shah SH, Bain JR, Muehlbauer MJ, et al. Association of a peripheral blood metabolic profile with coronary artery disease and risk of subsequent cardiovascular events. Circ. Cardiovasc. Genet. 2010;3(2):207-214.

Matsumoto M, Han S, Kitamura T, Accili D. Dual role of transcription factor FoxO1 in controlling hepatic insulin sensitivity and lipid metabolism. J. Clin. Invest. 2006;116(9):2464-2472.

Unraveling the Link between Diabetes and the Central Nervous System

Kuhajda FP, Aja S, Tu Y, et al. Pharmacological glycerol-3-phosphate acyltransferase inhibition decreases food intake and adiposity and increase insulin sensitivity in diet-induced obesity. Am. J. Physiol. Regul. Integr. Comp. Physiol. 2011;301(1):R116-R130.

Mansouri A, Aja S, Moran TH, et al. Intraperitoneal injections of low doses of C75 elicit a behaviorally specific and vagal afferent-independent inhibition of eating in rats. Am. J. Physiol. Regul. Integr. Comp. Physiol. 2008;295(3):R799-R805.

Aja S, Landree LE, Kleman AM, et al. Pharmacological stimulation of brain carnitine palmitoly-transferse-1 decreases food intake and body weight. Am. J. Physiol. Regul. Integr. Comp. Physiol. 2008;294(2):R352-361.

Kim EK, Miller I, Aja S, et al. C75, a fatty acid synthase inhibitor, reduces food intake via hypothalamic AMP-activated protein kinase. J. Biol. Chem. 2004;279(19):19970-19976.

Sponsors

Presented by

Bronze Sponsor

Agilent Technologies

Academy Friends

Bristol-Myers Squibb

Metabolon, Inc.

Thermo Fisher Scientific

It's no secret that diabetes is an increasing concern, not just for Western countries, where diet and lifestyle promote an expanding waistline and insulin resistance, but also for developing countries, in which the effects of a changing diet on the health of a population are already visible. In the U.S., diabetes affects approximately 11% of the population over age 20, and there are an additional 79 million adults with prediabetes, a condition that often precedes diabetes in which glucose levels are higher than normal.

Diabetes is an impairment in the body's ability to switch between glucose and fat as energy sources. Normally, when a person has not eaten recently (a fasting state), the muscles preferentially oxidize fat over glucose to ensure a supply of glucose for the brain. After a person eats, however, there is excess glucose in the system, and the muscles switch their primary energy source and start oxidizing glucose and storing fats. Individuals with diabetes are unable to make this switch, and their muscles continue to oxidize glucose in a fasting state, resulting in low blood glucose levels, and burn rather than store fats when a person has eaten, resulting in high blood lipid and glucose levels.

As diabetes and its associated comorbidities, such as cardiovascular disease, kidney disease, and neurological disorders, rise in epidemic proportions, it is now more important than ever to develop new tools to understand the complex metabolic mechanisms and pathways involved in this disease and to find new therapeutic targets. Scientists are turning to large-scale 'omics data to answer this call. In April 2012, leaders in this field met at the New York Academy of Sciences to discuss how various types of 'omics data (metabolomic, proteomic, transcriptomic, and lipidomic) can be integrated to reveal a more complete picture of these mechanisms.

The primary focus of the seminar was diabesity—a form of diabetes that typically develops in later life and is associated with being obese. Not all obese people have diabetes, and not all people with diabetes are obese, but there is definitely a connection between the two conditions. One of the main questions throughout the day was how to use 'omics data to create a profile of different disease states to understand why some individual develop diabetes and its associated complications while others do not.

'Omics data can be difficult to analyze based on the sheer size of the datasets. Charles Burant gave an introduction to two programs being developed in his lab to visualize and analyze several types of 'omics data. The first, Metscape, generates networks from genomic and metabolomic data, whereas the second, CoolMap displays the correlation between thousands of data points in a heat map-like format. The hope is that, by using these tools, researchers can generate hypotheses about the metabolic networks that respond to particular types of intervention, which can then be tested for their therapeutic value.

Providing a more concrete framework for how 'omics data can be used to generate hypotheses, Irwin Kurland presented a tiered model to use in mouse models in which commonly used measures of metabolism, such as calorimetry, low-resolution nuclear magnetic resonance (NMR) imaging, and glucose tolerance tests, can be used to inform which 'omics experiments to do next. As examples of this paradigm, Kurland presented data from his lab on two mouse models, a model of insulin resistance and a fatty acid amide hydrolase (FAAH)-knockout mouse, which has reduced fatty acid hydrolysis.

Next, Christopher Newgard and Barbara Kahn spoke about the role of branched chain amino acids in obesity and diabetes and about how 'omics-based analysis of genetically-induced obesity in mice with different susceptibilities for diabetes can reveal how these two states are linked. Newgard and Domenico Accili also shared their data on the link between diabetes and one of its most common comorbidities, heart disease. Newgard identified a set of metabolites uniquely linked to cardiovascular events in individuals with heart disease, whereas Accili focused on the role of the transcription factor FoxO1 in regulating lipid synthesis in the liver. The final section of the symposium focused on how the brain affects diabetes and obesity by regulating food intake and energy utilization. Again, Accili showed how, in addition to being important in the liver, FoxO1 also plays an active role in the brain's regulation of food intake. Gabriele Ronnett expanded on the role of the brain by focusing on how lipid metabolism in the brain regulates body weight and energy use throughout the body. Both speakers posed hypotheses for interventional approaches that may prove useful for treating diabetes and obesity in the future.

Speakers:
Irwin Kurland, Albert Einstein College of Medicine
Charles Burant, University of Michigan Medical School

Highlights

• Using the right tools and experimental framework, 'omics data can be used to generate testable hypotheses.
• Metscape 2 is a tool that integrates several types of 'omics data to visualize complex networks.
• CoolMap is a tool that displays correlations between data points in a heat map-like format.
• Using a tiered 'omics approach can provide insight into the mechanisms of metabolism and diabetes.

'Omic techniques generate a vast quantity of data, and, alone, they generally only provide correlations between variables. The usefulness of 'omics data, stressed by several speakers throughout the day, is its ability to generate hypotheses and to identify new targets to inform subsequent experiments.

New tools in visualizing 'omics data

One of the challenges of interpreting 'omics data is being able to separate information from background. Charles Burant of the University of Michigan Medical School presented two tools being developed in his lab to aid in the visualization and integration of large 'omics datasets and, ultimately, to help generate testable hypotheses from those datasets.

The first tool, Metscape 2, is a plugin for the program Cytoscape, a common platform for visualizing complex networks. Currently, Metscape 2 can incorporate gene expression and metabolomics data across different time points or different experimental conditions. Based on input data, Metscape 2 creates an interaction map that allows researchers to visualize the changes in gene expression and in metabolite levels and to link these changes to disease states.

The second tool that Burant discussed was CoolMap, which enables researchers to visualize large, two-dimensional data and to interpret correlations between datasets. CoolMap can manage datasets of 8000 x 8000 data points and shows the Pearson's correlation coefficient in a heat map-like format.

As an example, Burant showed two CoolMap plots of various clinical parameters before and after weight loss. By visualizing the changes in these parameters, researchers can focus on the relationships that differ between the two states and can make hypotheses that can be tested in further experiments.

The usefulness of CoolMap, Burant argued, will manifest in its ability to identify known unknowns, a term Burant used to describe unidentified, reproducible features in mass spectrometry data generated from untargeted high-throughput metabolomic studies. To demonstrate this point, Burant used a CoolMap showing the correlation between various metabolites (fatty acids, amino acids, acetyl CoA, etc.) from a group of 25 people after the subjects had lost 22.5% of their body weight. CoolMap can cluster the metabolites to reveal groups of metabolites that are highly related. By exporting a group of highly-related metabolites into Metscape 2, Burant showed that these metabolites were all part of a specific pathway Once a particular pathway is suspected of being important, researchers can hypothesize what other metabolites they should be able to see in their data and can go back to their original mass spectrometry data and identify some of their known unknowns.

CoolMap showing the correlations between various clinical parameters (fatty acids, amino acids, acetyl CoA, etc.) of 25 people after losing 22.5% of their body weight. The data have been clustered to find areas of high-relatedness. (Image courtesy of Charles Burant)

While Burant and colleagues have already made Metscape 2 and CoolMap available to researchers, they are also constantly improving the platforms. Future versions of Metscape 2 should be able to integrate proteomic, phosphoproteomic, and acetelomic data to understand the relationship between genes, proteins, and metabolites in various states, and Burant is working to integrate CoolMap with Metscape 2 and with other 'omics programs to provide a suite of tools to integrate various types of directed metabolomics, in which specific metabolites of interest are identified and quantified against stable isotope standards, and undirected metabolomics, in which researchers are not looking for specific metabolites but are instead performing an unbiased survey of which metabolites are sensitive to various conditions.

A model for using 'omics to generate testable hypotheses

High-throughput 'omics techniques can generate a lot of data, but how do you use those data to generate testable hypotheses? Irwin Kurland of the Albert Einstein College of Medicine provided a tiered approach for using 'omics techniques in mice.

Tiered approach to using 'omics data to analyze diabetes and obesity in mice. (Image courtesy of Irwin Kurland)

The first tier consists of common tests used to examine metabolism: indirect calorimetry, low-resolution nuclear magnetic resonance (NMR) imaging, and the stable isotope glucose tolerance test. Indirect calorimetry can provide information on what type of fuel the body is burning, carbohydrates or lipids. Low-resolution NMR imaging measures body composition, and the stable isotope glucose tolerance test can measure overall and hepatic glucose uptake. The results of each of these tests can inform subsequent experiments. For example, if the low-resolution NMR reveals changes in body fat, one could follow up by measuring lipogenesis or lipolysis. Changes in lipogenesis and/or lipolysis then provide enough evidence to follow up with lipidomic analyses to monitor which lipids are being produced and/or broken down. Similarly, if changes in insulin resistance are observed by the stable glucose tolerance test, other isotope tests can be done to monitor hepatic glucose production, Cori cycling (also called the lactic acid cycle), and lactate production. The results of these tests can inform metabolomic and lipidomic profiling efforts.

Kurland presented work on mouse models of two metabolic disorders, increased insulin sensitivity and prediabetes, to demonstrate his approach. The first was the PTEN^+/− mouse in which one copy of the PTEN gene is knocked out, a model of insulin resistance. PTEN normally inhibits insulin signaling. Because insulin signaling stimulates hepatic glucose uptake, Kurland expected that the PTEN^+/− mouse would show increased hepatic glucose uptake. However, contrary to this hypothesis, the stable isotope glucose tolerance test revealed virtually no hepatic glucose uptake in the PTEN^+/− mouse whereas hepatic glucose production was the same as in the wild type mouse. To explain these counterintuitive results, Kurland hypothesized that, to ensure that enough glucose is supplied to the brain, hepatic glycolysis—the breaking down of glucose—is dramatically suppressed in the fasted state, whereas hepatic glucose production occurs as normal.

The second model that Kurland presented was the fatty acid amide hydrolase (FAAH)- knockout mouse, a model of prediabetes. These mice are fat, but otherwise normal. The stable isotope glucose tolerance test indicated that the FAAH-knockout mouse had higher blood glucose levels after feeding even though glucose uptake in the periphery was the same. Therefore, the extra glucose is likely due to an increase in hepatic glucose production. This knowledge paved the way for investigating lipid levels and adipose tissue. The break down of lipids in the adipose tissue creates glycerol. Administering 2-¹³C glycerol to mice and monitoring its dilution can provide information on how actively the adipose tissue is breaking down lipids. In the FAAH mouse, Kurland observed an increase in the production of glycerol in the adipose tissue and a corresponding increase in the use of glycerol to produce glucose in the liver. Overall, the FAAH-knockout mouse showed changes in TCA cycle (also known as the citric acid cycle) metabolites that affect the malate-aspartate shuttle, one of the main conduits for transferring energy from glycolysis into the mitochondria. In particular, citrate levels were increased, indicating perturbations in acetyl CoA levels.

Based on these results, Kurland proceeded to analyze the acetylome of the FAAH-knockout mouse in the food-deprived and re-fed states from seven tissues. The acetylome consists of proteins whose activities are regulated by acetylation. In particular, acetylation of malate dehydrogenase (MDH) was increased in the liver of the re-fed mice. Activation of MDH by acetylation promotes oxidation through the TCA cycle. These studies show how, by beginning with gross measurements, such as calorimetry, one can eventually work towards understanding a mechanism at the molecular level.

Speakers:
Christopher Newgard, Duke University Medical Center
Barbara Kahn, Beth Israel Deaconess Medical Center & Harvard Medical School

Highlights

• 'Omics techniques can generate profiles of disease and can inform interventional strategies.
• Branch chain amino acids are strongly associated with insulin resistance.
• Mouse models of obesity and diabetes can shed light on why some obese individuals are protected from diabetes.

The role of branched chain amino acids in diabetes and obesity

One of the goals of 'omics techniques, as described by Christopher Newgard of Duke University Medical Center, is to create metabolic signatures of human diseases that can be used as prognostic factors, to monitor disease progression and to inform interventions. Part of Newgard's talk focused on a comparison between the metabolic profile of obese individuals with prediabetes and insulin resistance to that of lean individuals.

While previous studies have also looked at metabolic differences in obese and lean individuals, these studies have primarily focused on one or a small number of metabolites based on the particular hypothesis of each study. In contrast, Newgard's study was more of a fishing expedition, using targeted metabolomics to measure over one hundred analytes.

After grouping the analytes of interest by principal component analysis, Newgard focused on one group—branch-chain amino acids (BCAAs), including glutamate, 3- and 5-carbon acyl carnitines, phenylalanine, and tyrosine. Grouping these compounds into one principal component makes sense both metabolically and biochemically. Most of these compounds are metabolically linked. The BCAAs go through a similar set of reactions during catabolism, which reactions all generate glutamate during the first step and all produce 3- and 5-carbon acyl carnitines. The aromatic amino acids, phenylalanine and tyrosine compete with the BCAAs for the same transporters to enter cells.

Other studies have shown an association between BCAA levels and insulin resistance; however, the advantage of Newgard's study is that, because it was done on a global scale, the researchers were able to show that the whole pathway related to BCAA catabolism is affected and that the BCAA profile was the profile most strongly associated with insulin resistance, even more so than the lipid profile.

BCAAs contribute to insulin resistance

Hypothesizing that BCAAs play a major role in insulin resistance, Newgard showed that interventions to treat diabetes, such as gastric bypass surgery, weight loss, and drug therapeutics, decrease BCAA levels and that that decrease corresponds to the efficacy of the interventions. In addition, BCAAs were shown to play a causative role in insulin resistance. Rats fed a high-fat diet supplemented with BCAAs spontaneously ate less food and weighed less than rats fed a normal high-fat diet, but rats on both diets were equally insulin resistant.

Newgard proposed a mechanism for the role of BCAAs in insulin resistance. In an excess of nutrients and calories, which often occurs in a Western diet, the normal catabolism of BCAAs in the adipose tissue is overwhelmed, and BCAAs spill out into the bloodstream. They find their way to the muscle where they generate species that enter the TCA cycle and impair the ability of the mitochondria to completely oxidize fat. In the presence of these excess nutrients, the fuel-switching ability of the cell is impaired, and glucose becomes almost superfluous as a fuel source, which could lead to the high blood glucose levels observed in prediabetes and in diabetes even in the absence of impaired insulin signaling.

Model for impaired BCAA metabolism in the adipose tissue and muscle during obesity. (Image courtesy of Christopher Newgard)

Barbara Kahn followed up on the investigation of impaired BCAA metabolism in adipose tissue using a branched-chain aminotransferase (BCAT)-knockout mouse as a model. Knocking out BCAT impairs the mouse's ability to metabolize BCAAs and results in high serum BCAA levels. This state mimics characteristics of obesity, in which enzymes involved in BCAA metabolism are often downregulated, leading to high levels of circulating BCAAs. Replacing the adipose tissue in the BCAT-knockout mice with normal adipose tissue decreased the circulating levels of BCAAs, demonstrating that adipose does indeed play a major role in regulating the levels of BCAAs.

Teasing the link between diabetes and obesity with mouse models

Newgard also presented data on using mouse models to understand the link between obesity and diabetes. Starting with two common laboratory strains of mice, B6 and BTBR, Newgard, in collaboration with Alan Attie of the University of Wisconsin, introduced the ob gene into these mice to create two strains of genetically-induced obese mice. While both strains are insulin resistant, only one progresses to diabetes. Breeding these strains together and performing genomic and metabolomic profiling can reveal gene–transcript–metabolite and gene–metabolite–transcript networks. Specifically, glutamine and glutamate were shown to regulate PEPCK, a gluconeogenic enzyme, via alanine glyoxaminotransferase and arginase-1. These studies have the potential to uncover metabolic networks involved in the pathogenesis of diabetes.

Speakers:
Christopher Newgard, Duke University Medical Center
Domenico Accili, Columbia University

Highlights

• Dyslipidemia, a common feature of diabetes, is also a key risk factor for cardiovascular disease.
• 'Omics approaches can be used to determine why some people with coronary artery disease progress to cardiac events while others do not.
• The transcription factor FoxO1 regulates the bile acid profile, which in turn regulates lipid synthesis.

Using 'omics to profile different disease states in cardiovascular disease

Nearly one half of patients with diabetes will die of cardiovascular-related events. Both Newgard and Domenico Accili of Columbia University provided insight into the link between cardiovascular disease and diabetes.

Why do some people with coronary artery disease experience cardiovascular events while others do not? To address this question, Newgard turned to targeted metabolomic profiling. Newgard showed that one set of metabolites—small- to medium-chain dicarboxy acylcarnitines—were associated with cardiovascular events but not with coronary artery disease. Newgard is now undertaking a two-pronged approach to identify these metabolites and to understand how they are associated with cardiovascular events. Using a chemical approach involving mass spectrometry, Newgard was able to identify one of the metabolites of interest as 3-methylglutarate. Future studies can now focus on pathways, such as hypoxia and inflammation, that produce this metabolite. Using a human genetic approach, Newgard is doing metabolomic and genomic profiling of patients in the CATHGEN repository, which contains DNA and serum samples from people undergoing cardiac catheterization. To date, Newgard has profiled approximately 3500 individuals, 70% of whom have coronary artery disease, and 30% of whom have diabetes. The genomic and metabolomic profiles of these patients have implicated genes involved in endoplasmic reticulum stress and in the unfolded protein response pathway. These genes are believed to play a role in mediating the concentrations of the small to medium chain acyl carnitines.

The role of FoxO1 in regulating lipid synthesis

Model for how regulation of bile acids by FoxO1 regulates cholesterol, triglyceride, and free fatty acid production in the liver. (Image courtesy of Domenico Accili)

To better understand the link between diabetes and heart disease, Accili investigated the regulation of lipoproteins by the liver. Low levels of high-density lipoprotein (HDL) and high levels of low-density lipoproteins (LDL) are associated with an increased risk for cardiovascular disease. Accili showed that the dyslipidemia that occurs in diabetes may result from a change in the bile acid profile. Selectively knocking out the gene (Foxo1) for the transcription factor FoxO1 in the liver of mouse models resulted in a lipid profile similar to that seen in humans with cardiovascular disease. Metabolomic analyses of these mice revealed changes in the types, but not in the overall levels, of bile acids. These bile acids were less robust in activating the farnesoid X receptor (FXR, also known as the bile acid receptor), resulting in decreased cholesterol absorption, increased cholesterol synthesis, and increased triglyceride synthesis. Interestingly, all of these characteristics are also observed in individuals with diabetes.

Accili postulated that by modulating FoxO1's activity insulin regulates the balance between two pathways that produce bile acids. Under normal insulin activity, bile acids provide robust stimulation of FXR in the liver, which decreases triglyceride, free fatty acid, and cholesterol synthesis. Deletion of Foxo1 in the liver, however, shifts the formation of bile acids to an alternative pathway. This pathway generates bile acids that are not able to activate FXR as well, resulting in increased triglycerides, free fatty acids, and cholesterol. In support of this model, activation of FXR or addition of the normal bile acids decreased liver triglyceride levels. Accili speculated that dyslipidemia in diabetes could be treated by targeting components of the bile acid synthetic pathway or by providing missing bile acids.

Speakers:
Domenico Accili, Columbia University
Gabriele Ronnett, The Johns Hopkins University School of Medicine

Highlights

• Fatty acid metabolism in the brain plays a key role in food intake and energy utilization.
• The transcription factor FoxO1 regulates neuronal activity that regulates food intake.
• 'Omics techniques and small molecule regulators of enzymes involved in fatty acid metabolism can tease out these pathways and can suggest druggable targets for metabolic disorders.

The role of FoxO1 in regulating appetite control in the brain

Both Accili and Gabriele Ronnett of the Johns Hopkins University School of Medicine discussed how metabolism in the brain affects food intake, energy utilization, and insulin sensitivity. The hormones insulin and leptin activate signaling pathways in the brain that decrease food intake and increase energy expenditure. However, under conditions of insulin resistance, these pathways are dysregulated. Studying how the brain regulates appetite and energy utilization could identify druggable targets to modify these processes during the course of diabetes.

In addition to knocking out Foxo1 in the liver, as described in the previous section, Accili also selectively knocked out Foxo1 in AgRP neurons—neurons that produce the agouti-related protein neuropeptide—in the brain. AgRP neurons are found in the arcuate nucleus of the hypothalamus and their activation promotes food intake while decreasing energy expenditure. Under normal conditions, insulin and leptin inhibit the activity of these neurons, which decreases food intake and increases energy expenditure. Previous attempts to identify druggable targets in these pathways have proven unsuccessful because selective knockouts of either insulin or leptin in AgRP neurons do not significantly affect food intake. Because FoxO1 is downstream of both insulin and leptin signaling pathways, Accili investigated whether genetic manipulation of Foxo1 could identify druggable targets in AgRP neurons.

Knocking out Foxo1 inhibits AgRP neuronal activity in a manner that mimics insulin and leptin activity. The knockout mice ate less food, had less body fat, and experienced decreased hepatic glucose production, which also mimics insulin activity. Accili used a genomic approach to show that transcription of genes involved in the mitochondrial electron transport chain, in the synthesis of mitochondrial ribosomal proteins, and in the production of ribosomal proteins increased in the Foxo1-knockout mice.

After validating by immunohistochemistry that their transcriptomic approach was indeed identifying genes that were dysregulated in the Foxo1-knockout mice, Acilli sifted through the transcriptomic data to identify a druggable target. Accili focused on a particular hit, the G-protein coupled receptor Gpr17. Gpr17 is of interest because it is already the target of several drugs, such as Singulair, an asthma medication, and such as anti-platelet aggregation agents; however, until recently Gpr17 has not been associated with metabolism. Accili showed that activation of Gpr17 increased food intake and that this effect was eliminated in the Foxo1-knockout mouse.

Model for regulation of food intake and heptatic glucose activity by FoxO1 and Gpr17. (Image courtesy of Domenico Accili)

Based on these results, Accili put forth a model in which, during energy deprivation, AgRP neurons are stimulated via the activation of Gpr17 by an unknown ligand. The neurons' activity promotes food intake and hepatic glucose production. After feeding, AgRP neurons are inhibited by the action of insulin and leptin, which promotes FoxO1-dependent inhibition of Gpr17. Accili speculated that antagonists of Gpr17 may prove to be useful for decreasing appetite.

Altering lipid oxidation in the brain can regulated appetite and weight

Also interested in the role of the brain in regulating energy usage, Gabriele Ronnett described her efforts to alter fatty acid metabolism in the brain. Fatty acid metabolism sits at the fulcrum of energy surplus and energy deficit. To tease out the mechanism of fatty acid metabolism and energy utilization, Ronnett focused on three key enzymes: fatty acid synthase (FAS), glycerol-3-phosphate-O-acyltransferase (GPAT), and carnitine palmitoyltransferase-1 (CPT-1). FAS generates long-chain acyl-CoA compounds, which can either be stored via the activity of GPAT or oxidized via the activity of CPT-1. Ronnett hypothesized that either FAS inhibition or CPT-1 stimulation in the central nervous system would decrease food intake and body weight. She tested this hypothesis using three small molecules: C75, an inhibitor of FAS and activator of CPT-1, FSG67, an inhibitor of GPAT, and C89b, an activator of CPT-1.

The role of FAS, GPAT, and CPT-1 in fatty acid metabolism. (Image courtesy of Gabriele Ronnett)

Both C75 and FSG67 induced weight loss in obese mice. Examining the genetic effects of these compounds revealed that the synthesis of enzymes involved in fatty acid storage was downregulated, whereas the synthesis of enzymes involved in fat disposition was upregulated. FSG67 is currently in preclinical safety tests. C89b, the CPT-1 stimulator, also decreased food intake and induced weight loss, consistent with the role of CPT-1 in promoting lipid oxidation. The effects seen with C89b, however, were more dramatic and longer lasting than those seen with C75 or with FSG67.

To investigate the mechanisms of these compounds' effects, researchers are conducting metabolomic studies in neurons in vitro. So far, they have seen that C75 and FSG67 increase reactive oxygen species while reducing the secretion of inflammatory cytokines. Based on these data, Ronnett speculated that C75 and FSG67 are not just altering fatty acid metabolism in the neurons but may also have long-term effects on inflammation in the brain. To understand what other pathways are affected by alteration of fatty acid flux and to elucidate what metabolic changes are affecting the observed changes in inflammatory signals, Ronnett is undertaking a full metabolomic profile in primary hypothalamic and cortical neurons treated with palmitate, C75, and FSG67.

Charles Burant presented to tools from his lab that enable researchers to visualize 'omics data. How can these visualization and analysis tools help researchers use 'omics data? And what are the best ways to visualize 'omics data?

Irwin Kurland presented one model for using 'omics tools to generate hypotheses in mouse models. What strategies or models can help researchers use 'omics techniques efficiently to generate testable hypotheses?

Both Christopher Newgard and Barbara Kahn showed the importance of BCAAs in diabetes. Do enzymes involved in BCAA synthesis and metabolism represent new targets for the treatment of diabetes and obesity?

Domenico Accili speculated that the bile acid profile is an important regulator of lipid synthesis. Can replacing the missing bile acids in patients with diabetes help normalize their lipid profile and reduce their risk for heart disease?

Both Accili and Gabriele Ronnett describe how the brain regulates food intake and energy utilization. Will inhibition of Gpr17, shown to reduce appetite, pan out as a therapeutic strategy in humans? Will the small molecules used by Ronnett display any therapeutic benefit in humans? Finally, what other strategies, besides CPT-1 stimulation and FAS inhibition, could modulate fatty acid metabolism in the brain?