Differential effects of calorie restriction and exercise on the adipose transcriptome in diet-induced obese mice.

We tested the hypothesis that obesity reversal by calorie restriction (CR) versus treadmill exercise (EX) differentially modulates adipose gene expression using 48 female C57BL/6 mice administered a diet-induced obesity (DIO) regimen for 8 weeks, then randomized to receive for 8 weeks either: (1) a control (AIN-76A) diet, fed ad libitum (DIO control); (2) a 30% CR regimen; (3) a treadmill EX regimen (with AIN-76A diet fed ad libitum); or (4) continuation of the DIO diet. Relative to the DIO controls, both CR and EX reduced adiposity by 35-40% and serum leptin levels by 80%, but only CR increased adiponectin and insulin sensitivity. Gene expression microarray analysis of visceral white adipose tissue revealed 209 genes responsive to both CR and EX, relative to the DIO group. However, CR uniquely altered expression of an additional 496 genes, whereas only 20 were uniquely affected by EX. Of the genes distinctly responsive to CR, 17 related to carbohydrate metabolism and glucose transport, including glucose transporter (GLUT) 4. Chromatin immunoprecipitation assays of the Glut4 promoter revealed that, relative to the DIO controls, CR significantly increased histone 4 acetylation, suggesting epigenetic regulation may underlie some of the differential effects of CR versus EX on the adipose transcriptome.

1. Introduction

More than two thirds of all adults in the USA are either overweight or obese [1]. Obesity is associated with an increased risk of developing several chronic diseases, including atherosclerosis, type 2 diabetes and many types of cancer [2-4]. At the crux of obesity-related diseases is metabolic dysregulation characterized by insulin resistance and elevated levels of circulating insulin, glucose, and several other metabolic factors directly linked to excess adiposity. In the context of low adiposity, insulin activates signaling through the insulin receptor, resulting in translocation of the glucose transporter 4 (Glut4) to the cell membrane to increase glucose uptake into the adipocyte [5]. In contrast, high levels of adiposity are marked by enlarged adipocytes which are unresponsive to insulin levels even under conditions of hyperinsulinemia [6]. In the insulin-resistant state, adipose tissue secretes adipokines and proinflammatory factors that reduce insulin sensitivity in peripheral tissues, thereby affecting whole-body glucose homeostasis [7, 8]. Unfortunately, the mechanisms underlying these changes in insulin responsiveness in adipocytes are poorly understood. Furthermore, mechanism-based lifestyle strategies for effectively offsetting obesity-induced insulin resistance are lacking.

Increased energy expenditure and decreased energy intake are the two most commonly recommended lifestyle changes to reduce adiposity and restore insulin sensitivity [9]. Calorie restriction (CR) and exercise (EX) are both effective at improving insulin sensitivity and decreasing both body weight and percent body fat [10, 11], although the differential effects of these two antiobesity interventions on weight reduction, body composition, and chronic disease risk are well established [9, 11]. There are also conflicting reports within the human literature concerning the effectiveness of EX in improving body fat distribution and adipokine secretion, two key predictors of insulin resistance [10, 12–14]. Furthermore, studies on the beneficial effects of EX have focused mainly on molecular changes in the skeletal muscle and liver, while considerably less is known about changes in adipose tissue [11, 15, 16]. Therefore, the aim of this study was to compare the effect of CR and EX on visceral white adipose tissue (VWAT) gene expression, along with changes in body composition and insulin resistance, in diet-induced obese mice. We utilized an animal model of postmenopausal obesity because postmenopausal women are especially at risk for developing diseases associated with obesity such as type 2 diabetes [17] and breast cancer, the second leading cause of cancer death in women [4].

In the present study, we show that, despite comparable reductions in adiposity in obese mice, CR was more effective than EX at increasing adiponectin, improving insulin sensitivity, and altering the adipose transcriptome. Although both CR and EX qualitatively affected a shared set of genes related to metabolism, CR had a stronger quantitative effect on these genes. Furthermore, CR induced a dramatic change in expression of an additional set of genes related to carbohydrate metabolism and transport in VWAT that was not observed in the EX mice.

2. Materials and Methods

2.1. Animal Study Design

All animal protocols were approved by the University of Texas at Austin Institutional Animal Care and Use Committee. To model the postmenopausal state, 6-week-old ovariectomized C57BL/6 mice were used (Charles River Labs, Inc. Frederick, Md, USA). Ovariectomized mice exhibit characteristics of the postmenopausal state in humans: decreased levels of circulating estrogen, loss of bone mineral density, and cessation of estrous cycles [18]. Upon arrival, mice had ad libitum access to water and chow diet and were on a 12 : 12 h light/dark cycle.

To compare the effects of CR and EX on reversal of obesity and insulin resistance, and other metabolic perturbations, 48 mice were singly housed upon receipt and put on a diet-induced obesity (DIO) regimen for 8 wks consisting of ad libitum access to a 60 kcal% fat diet (D12492; Research Diets, Inc, New Brunswick, NJ, USA), beginning one week after arrival. At week 9, the mice were randomized into the following treatment groups (n = 12/group): (1) DIO control (AIN-76A diet fed ad libitum); (2) 30% CR; (3) treadmill exercise regimen, fed AIN-76A diet ad libitum (EX); or (4) continuation on the DIO regimen. In animal models, CR diet regimens, typically involving a 20–40 reduction in carbohydrate calorie intake and designed to limit total energy intake while insuring adequate nutrition, represent the most potent dietary approach to prevent and/or reverse obesity and inhibit tumor growth [7]. The DIO control and EX groups were switched from the DIO regimen to a modified AIN-76A diet (D12450B, that is 10 kcal% fat and is the base diet of our CR regimen; Research Diets, New Brunswick, NJ, USA) consumed ad libitum. The DIO control group was used as a feeding control for determining CR feed intake and to ensure that EX mice were not overeating to compensate for increased energy expenditure. We have previously shown that switching DIO mice to the control (AIN-76A) diet maintains adiposity near the peak level achieved during the 8 weeks of DIO [19], and this was confirmed in the current study. Since body weight and body composition data on this DIO control diet were comparable to continuous DIO, the DIO control group was used as the comparator for all analyses. This also provided control for changes in expression due to differences in diet composition/fat consumption, since the diets for the DIO control, CR, and EX groups all based on the AIN-76A diet. The CR group consumed a modified diet (D0302702, administered in daily aliquots) providing 30% fewer calories from carbohydrates compared to the control diet, with all other components being isonutrient when intake was limited to 70% of mean kcal consumption of the diet control group. The EX group were run on a variable speed treadmill 5 days/wk on a 5% grade, beginning with 10 min/day at 12 m/min. Time and intensity were increased gradually over the next two weeks until the EX group reached 40 min/day at a maximum rate of 20 m/min. The DIO control, continuous DIO and CR mice were all placed on the treadmill but not run. Body weights and feed intake were measured weekly.

At the beginning of week 17, when the CR and EX mice achieved comparable reductions in adiposity relative to the DIO controls, mice were euthanized. In the morning the mice were killed, all mice received their respective dietary or exercise treatment, followed by a 6-hr fast. Mice were anesthetized with isofluorane for terminal blood collection via the retro-orbital venous plexus, and then killed by cervical dislocation. Whole blood was allowed to clot at room temperature for 30 min prior to centrifugation at 1000 ×g for 10 min. The serum was removed and stored at −80°C for analyses. A 1-gram sample of VWAT was collected from each mouse and stored at −80°C until further analyses. Carcasses were stored at −20°C. Percent body fat and lean mass were determined using dual energy X-ray absorptiometry (DXA) (GE Lunar Piximus II, Madison, WI, USA) as described previously [20].

To further characterize the effect of CR on the histone code (which required different tissue processing procedures than the gene expression microarray analysis), an additional group of 15 mice received the AIN-76A control diet (labeled overweight mice), CR diet (labeled lean mice), or DIO (labeled obese mice) for 8 weeks (n = 5/diet group). Body composition on these mice was determined using quantitative magnetic resonance (Echo Medical Systems, Houston, TX, USA). Animals were then killed after an 8-hr fast, serum collected as described above, and tissues (including VWAT, liver and mammary glands 4 and 9) were excised, formadehyde treated to crosslink proteins, and immediately flash frozen for analysis by the chromatin immunoprecipitation assay described below.

2.2. Glucose Tolerance Test

To determine the effects of CR and EX on glucose regulation following weight loss, we conducted a glucose tolerance test (GTT) on the 48 mice on study at week 15. GTT was performed after a 6-hour fast by administration of 20% glucose (2 g/kg body weight IP). Blood samples were taken from the tail and analyzed for glucose concentration using an Ascencia Elite XL 3901G glucose analyzer (Bayer Corporation, Mishawaka, Ind). Glucose levels were determined at baseline, 15, 30, 60, and 120 min after injection of glucose.

2.3. Serum Hormones

Leptin, insulin, and adiponectin were measured in serum collected at the terminal bleed, using mouse adipokine LINCOplex Multiplex Assays (Millipore, Inc., Billerica, MA, USA) analyzed on a BioRad Bioplex 200 analysis system (Biorad, Inc. Hercules, CA, USA).

2.4. Gene Expression Microarray Analysis

Total RNA was isolated from VWAT tissues using an organic extraction and precipitation protocol with a DNAseI treatment step (Asuragen Inc., Austin, TX, USA). Biotin-labeled targets were prepared using modified MessageAmp-based protocols (Ambion Inc., Austin, TX, USA) and hybridized to MOE 430A 2.0 arrays (Affymetrix, Santa Clara, CA, USA). The arrays were scanned on an Affymetrix GeneChip Scanner 3000 7G. A summary of the image signal data, detection calls, and gene annotations for every gene interrogated on the array was generated using Affymetrix Statistical Algorithm MAS 5.0 (GCOS v1.3), with all arrays scaled to 500. Sample normalization was carried out using the Robust Multichip Average (RMA) followed by multiple group analysis comparison using ANOVA. Pairwise comparisons were performed to identify expression fold differences with false discovery rate (FDR) set at 0.05. Genes with expression differences equal or greater than 2-fold compared to DIO controls, were selected to be analyzed using the Database for Annotation, Visualization and Integrated Discovery (DAVID; [21]). The resulting Gene Ontology (GO) analysis was used to identify genes relevant to the different effects of CR and EX in reversing obesity, some of which were selected for further analysis. In the DAVID analysis, genes that were represented more than once in the microarray output were filtered. Some of the genes in the Gene Ontology analysis belonged to more than one functional category and are tabulated accordingly. Expression changes were verified in VWAT from a separate cohort of mice that underwent CR or EX following DIO, as described above, using Taqman Gene Expression Assay (Applied Bioystems Inc., Carlsbad, CA, USA). Gene expression data were normalized to the housekeeping gene β-actin.

2.5. Chromatin Immunoprecipitation (ChIP) Assay

ChIP assays were performed per manufacturer's instructions (Millipore). Briefly, proteins from VWAT were formaldehyde crosslinked to DNA. After homogenization, lysis, and sonication, proteins were incubated overnight with antibodies to acetyl-histone H4 or trimethyl histone H4 (Millipore). The DNA-protein complexes were washed, DNA was eluted, and crosslinking was reversed by heating to 65°C overnight. DNA was purified using QIAGEN PCR purification kit (QIAGEN, Valencia, CA, USA). Quantitative, real-time PCR was performed using SYBR Green (ABI) with the following Slc2a4 primers: forward primer 5′-CCCTTTAAGGCTCCATCTCC-3′ and reverse primer 5′-TGTGTGTATGCCCCGAAGTA-3′ (ABI). GAPDH was used as the internal control for analysis of acetylation with the following primers: forward primer 5′-CATGGCCTTCCGTGTTCCTA-3′ and reverse primer 5′-CCTGCTTCACCACCTTCTTGAT-3′. For analysis of methylation, p16 was used as the internal control with the following primers: forward primer 5′-ACACTCCTTGCCTACCTGAA-3′ and reverse primer 5′-CGAACTCGAGGAGAGCCATC-3′.

2.6. Statistics

Values are presented as mean ± standard error (SE). One-way analysis of variance (ANOVA) followed by Tukey's Honestly Significant Differences test was used to assess the effects of diet on mean weekly body weight at weeks 8 and 16, body composition data at week 16, serum adipokine levels, and fasting glucose levels. Repeated measures analysis was used to evaluate glucose tolerance tests. For serum insulin, mRNA levels (as measured by qRT-PCR), and relative quantification of Glut 4 in ChIP experiments means, were compared using Student's t-test. For all tests SPSS software was used (SPSS Inc., Chicago, IL, USA), and P ≤ .05 was considered statistically significant.

3. Results

3.1. Both CR and EX Decrease Adiposity, Insulin and Leptin Levels, but Only CR Increases Adiponectin and Restores Insulin Sensitivity in DIO Mice

During the first 8 weeks, the DIO regimen increased mean body weight of the 48 mice on study from 20.3 ± 0.5 g to 30.7 ± 0.5 g, and % body fat to 52.3%. As shown in Table 1, one week after randomization (week 9 of the study), the DIO control, EX, CR, and continuous DIO groups did not differ in body weight. However, by week 16 of the study, the DIO control group (30.8 ± 1.6 g) was significantly heavier than the EX mice (26.0 ± 0.9 g) and the CR mice (19.9 ± 0.5 g), but not the continuous DIO group (33.2 ± 1.5 g). The body weight data closely correlated with calorie intake (for weeks 9–16: 709 ± 10.0 kcal for DIO controls; 556 ± 4.6 kcal for the EX mice; 413.0 ± 4.3 kcal for the CR; and 722 ± 11.6 kcal for the continuous DIO group) and % body fat (Table 1). Although the CR mice weighed significantly less than the EX mice (primarily due to the increase in lean mass in the EX group relative to the CR mice), there was no difference in percent body fat, with both groups exhibiting >25% reductions in % body fat compared to DIO control mice. Achieving meaningful reductions in adiposity in obese mice via CR and EX was a goal of the study design, given that percent body fat is associated with insulin resistance and other key metabolic changes associated with DIO [22]. Since body weight, kcal consumption and body composition data for mice on the DIO control diet were comparable to mice on the continuous DIO diet, the DIO control group was used as the comparator for all analyses. This allowed us to eliminate the possibility that any changes observed in hormones and gene expression could have been due to differences in diet composition/fat consumption, since the diets for the DIO control, CR, and EX groups were all based on the same AIN-76A diet composition. As shown in Figure 1(a), this also allowed us to limit our hormone and microarray analyses to 3 groups (DIO control, CR, and EX), without the continuous DIO group, thus increasing the number of mice per group analyzed within our budget constraints.

Table 1

Body composition after 8 weeks of DIO followed by 8 weeks of control diet, exercise, or calorie restriction.

Group	Body weight week 9 (g)	Body weight week 16 (g)	Percent body fat (%)	Lean mass (g)
DIO Control	30.1 ± 0.9a	30.8. ± 0.6a	51.1 ± 3.8a	12.6 ± 0.2a
Exercise	29.3 ± 0.6a	26.0 ± 0.9b	38.9 ± 2.7b	13.6 ± 0.2b
Calorie Restriction	31.3± 0.6a	19.9 ± 0.5c	33.7 ± 1.4b	10.5 ± 0.1c
ContinuousDIO*	30.3 ± 1.2a	33.2± 1.5a	57.3 ± 2.9a	11.8 ± 0.2d

Data are presented mean ± SEM. Significant differences (P < .05) beween data within a column are indicated by different superscripts; n = 12/group.

Figure 1

Effect of calorie restriction or exercise in diet-induced obese mice on serum hormones and glucose tolerance. (a) Animal study design for gene expression microarray experiments. (b) Serum leptin levels, (c) serum adiponectin levels, and (d) serum insulin levels after 8 weeks of intervention, (n = 11 for DIO group; n = 10 for CR group; n = 10 for EX group). (e) Blood glucose concentrations during a glucose tolerance after 7 weeks of intervention. Data shown are mean ± SE. DIO (●), EX (), CR (□), n = 12/group. Significance (P ≤ .05) between groups is denoted by different letters.

At the end of the study we also measured circulating leptin and adiponectin levels, two adipokines that are positively and negatively correlated with adiposity, respectively [23]. Consistent with decreased adiposity, leptin levels were roughly 80% lower in the CR and EX mice (Figure 1(b)). However, only CR increased adiponectin levels compared to DIO control mice (Figure 1(c)), even though percent body fat in CR and EX mice did not statistically differ. The higher levels of adiponectin observed in the CR mice were associated with decreased fasting insulin levels (Figure 1(d)), decreased fasting glucose levels (Figure 1(e)), and increased insulin sensitivity as indicated by significantly lower blood glucose levels at every time point following glucose challenge (Figure 1(e)). In contrast, the EX mice did not display increased insulin sensitivity or decreased fasting insulin levels compared to sedentary DIO. Taken together, these data demonstrate that CR and EX differentially affected adipose tissue metabolism.

3.2. Transcriptional Changes Common to CR and EX in VWAT

Gene expression microarray analysis was performed on VWAT collected following the 8-wk weight-loss phase after DIO. Pairwise comparisons of DIO versus CR and DIO versus EX revealed that 725 transcripts were significantly altered (±2.0 fold, P < .05, Figure 2(a)). Of those 725 transcripts, 209 were common to CR and EX (Figure 2(a)), possibly representing a suite of genes most sensitive to energy balance. GO analysis was used to categorize these genes according to function and revealed that the majority of genes altered both by CR and EX were related to metabolic process, immune response, and stress response (Figure 2(b)). Within the metabolic process category, 24 of the genes were related to lipid metabolism, and overall the response of the genes to CR and EX was qualitatively similar. More specifically, a number of genes involved in fatty acid synthesis and transport were upregulated (Figure 2(c)). These included stearoyl-CoA desaturase (Scd1), fatty acid synthase (Fasn), carnitine palmitoyltransferase 1 (Cpt1), and elongation of long chain fatty acids 3 and 6 (ELOVL3 and ELOVL6). In addition, 9 genes related to glucose metabolism were affected by CR and EX, including pyruvate dehydrogenase E1 alpha 1(Pdha1), leptin (Lep), and glycerol phosphate dehydrogenase 2 (Gpd2). As in lipid metabolism, genes related to carbohydrate metabolism were qualitatively responsive to both CR and EX, although CR had a stronger quantitative effect.

Figure 2

Effect of weight loss induced by calorie restriction or exercise on mRNA expression in VWAT. (a) Venn Diagram of genes differentially expressed by CR and EX compared to DIO controls. (b) Classification of genes targeted by both CR and EX. (c) Heat map of genes related to metabolic processes affected by both CR and EX. (d) Classification of genes targeted uniquely by CR (n = 6/group).

Reductions in adiposity are accompanied by lower levels of immune cell infiltrates into adipose tissue, which mediate the proinflammatory state associated with obesity [24]. As expected, the reduced adiposity in CR and EX mice was associated with decreased expression of genes related to immune response (Figure 2(b)). These immune-related genes also comprised the majority of the genes in the stress response category, including downregulation of transcripts that code for chemokines that attract and are produced by monocytes and macrophages, specifically Chemokine (C-C motif) ligand (Ccl) 2, 6, 7, and 9.

3.3. Unique Transcriptional Changes in Response to CR or EX in VWAT

CR uniquely affected expression of 496 genes, whereas a mere 20 genes were responsive only to EX (Figure 2(a)). GO analysis of the genes uniquely responsive to EX revealed that only the grouping of genes related to mitochondrial transport was significant. Specifically, uncoupling proteins Ucp1 and Ucp2 were both upregulated by EX. Given the robust transcriptional response to CR, we focused our analysis on those genes whose expression was affected by CR but not EX (Table 2). GO analysis showed that in every category of genes altered by both CR and EX, CR impacted an additional set of genes unaffected by EX. For example, in genes relating to cellular lipid metabolic processes, which was the largest subset of transcripts uniquely altered by CR, soluble carrier family 27 (Slc27a1) and Acetyl-Coenzyme A carboxylase alpha (Acaca) were upregulated. CR also uniquely increased expression of sterol regulatory element binding transcription factor 1 (Srebp1), a master regulator for lipid metabolism in adipocytes. With respect to immune response and stress response, CR resulted in a downregulation of gene expression, whereas expression of genes related to biosynthesis of steroids was upregulated.

Table 2

Transcriptional changes in response to calorie restriction or exercise in visceral white adipose tissue.

Gene Symbol	Gene title	Fold change
Cellular lipid metabolic process		CR × DIO
Slc27a1	Solute carrier family 27 (fatty acid transporter)	2.53	up
Fads2	Fatty acid desaturase 2	2.18	up
Ces3	Carboxylesterase 3	2.51	up
Sult1a1	Sulfotransferase family 1A	2.22	up
Ptges	Prostaglandin E synthase	2.25	up
Sgpp1	Sphingosine-1-phosphate phosphatase 1	2.73	down
Echs1	Enoyl Coenzyme A hydratase	2.00	up
Hsd11b1	Hydroxysteroid 11-beta dehydrogenase 1	3.28	up
Apoc3	Apolipoprotein C-III	5.71	up
Srebf1*	Sterol regulatory element binding transcription factor 1	2.98	up
Aldh1a7	Aldehyde dehydrogenase family 1	2.18	up
Hpgd	Hydroxyprostaglandin dehydrogenase 15 (NAD)	2.09	down
Rdh11	Retinol dehydrogenase 11	4.59	up
Rarres2	Retinoic acid receptor responder (tazarotene induced) 2	2.61	down
Nsdhl*	NAD(P) dependent steroid dehydrogenase-like	2.65	up
Gpam	Glycerol-3-phosphate acyltransferase	2.06	up
Hmgcs1*	3-hydroxy-3-methylglutaryl-Coenzyme A synthase 1	3.01	down
Abat	4-aminobutyrate aminotransferase	2.88	down
Sorl1	Sortilin-related receptor	2.77	up
Pip4k2a	Phosphatidylinositol-5-phosphate 4-kinase	2.02	down
Acaca	Acetyl-Coenzyme A carboxylase alpha	4.60	up
Tm7sf2*	Transmembrane 7 superfamily member 2	3.69	up
Sc5d*	Sterol-C5-desaturase	2.11	up
Fdft1*	Farnesyl diphosphate farnesyl transferase 1	2.68	up
Hsd17b12*	Hydroxysteroid (17-beta) dehydrogenase 12	2.17	up
Pcx*	Pyruvate carboxylase	2.88	up
Cellular carbohydrate metabolic process
Fn3k	Fructosamine 3 kinase	4.05	up
Chst1*	Carbohydrate (keratan sulfate Gal-6) sulfotransferase 1	2.00	up
Dlat	Dihydrolipoamide S-acetyltransferase	2.59	up
Pkm2	Pyruvate kinase	2.37	up
Pmm1	Phosphomannomutase 1	2.72	up
Ppp1r1a	Protein phosphatase 1	2.61	up
Pgd	Phosphogluconate dehydrogenase	2.40	up
Agl	Amylo-1	2.16	up
Oxct1	3-oxoacid CoA transferase 1	2.17	down
Pdk1	Pyruvate dehydrogenase kinase	3.07	up
Gpd1	Glycerol-3-phosphate dehydrogenase 1 (soluble)	2.02	up
Taldo1	Transaldolase 1	2.03	up
Glucose transport
Sh2b2*	SH2B adaptor protein 2	3.26	up
Slc2a4	Solute carrier family 2 (facilitated glucose transporter)	3.43	up
Pcx*	Pyruvate carboxylase	2.88	up
Klf15	Kruppel-like factor 15	2.36	up
Immune response		CR × DIO
Malt1	Mucosa associated lymphoid tissue Lymphoma translocation gene 1	2.55	down
Bcl6*	B-cell leukemia/lymphoma 6	2.90	down
Clec7a*	C-type lectin domain family 7	2.89	down
Cfb	Complement factor B	2.53	down
Cd55*	CD55 antigen	2.25	Down
Thy1	Thymus cell antigen 1	2.44	down
Biosynthesis of Steroids
Lss	Lanosterol synthase	2.13	up
Hmgcs1*	3-hydroxy-3-methylglutaryl-Coenzyme A synthase 1	3.01	down
Tm7sf2*	Transmembrane 7 superfamily member 2	3.69	up
Sc5d*	Sterol-C5-desaturase	2.11	up
Fdft1*	Farnesyl diphosphate farnesyl transferase 1	2.68	up
Hsd17b12*	Hydroxysteroid (17-beta) dehydrogenase 12	2.17	up
Nsdhl*	NAD(P) dependent steroid dehydrogenase-like	2.65	up
Sh2b2*	SH2B adaptor protein 2	3.26	up
Stress response
Thbs1	Thrombospondin 1	2.29	down
Tfpi2	Tissue factor pathway inhibitor 2	4.32	down
Gp1bb	Glycoprotein Ib	2.06	up
Taok3	TAO kinase 3	2.21	down
Sod3	Superoxide dismutase 3	2.17	down
Dusp10	Dual specificity phosphatase 10	2.62	down
Adrb3	Adrenergic receptor	2.09	up
F2r	Coagulation factor II (thrombin) receptor	2.22	down
Ccnd1	Cyclin D1	2.91	down
Evl	Ena-vasodilator stimulated phosphoprotein	2.32	down
Ctsb	Cathepsin B	2.00	down
Ly86	Lymphocyte antigen 86	2.65	down
Fabp4	Fatty acid binding protein 4	2.03	up
Rad50	RAD50 homolog (S. cerevisiae)	2.27	down
Tsc22d2	TSC22 domain family 2	2.41	down
Ptger3	Prostaglandin E receptor 3 (subtype EP3)	2.21	up
Lcp1	Lymphocyte cytosolic protein 1	3.06	down
Pros1	Protein S (alpha)	2.09	down
Hspa12a	Heat shock protein 12A	2.40	down
Anxa2	Annexin A2	2.01	down
Uhrf1	Ubiquitin-like	2.84	down
Cdkn1a	Cyclin-dependent kinase inhibitor 1A (P21)	2.28	down
Srebf1*	Sterol regulatory element binding transcription factor 1	2.98	up
Chst1*	Carbohydrate (keratan sulfate Gal-6) sulfotransferase 1	2.00	up
Bcl6*	B-cell leukemia/lymphoma 6	2.90	down
Clec7a*	C-type lectin domain family 7	2.89	down
Cd55*	CD55 antigen	2.25	down
Exercise unique
Mitochondrial Transport		EX × DIO
Ucp1	Uncoupling protein	8.94	up
Ucp2	Uncoupling protein 2	2.09	down

*Genes represented in two different categories.

In addition to affecting more genes in each functional category than EX, CR affected the transcription of genes in another category not modulated by EX, specifically 4 genes related to glucose transport. Complementing this increase in transcription of glucose transport genes, CR resulted in upregulation of another 14 genes related to carbohydrate metabolism processes. Taken together, these data are suggestive of increased glucose flux into the adipose tissue, which may underlie the enhanced insulin sensitivity observed in response to CR.

3.4. Real-Time RT-PCR Confirmation of Microarray Results

Given that the DIO mice were on a high-fat diet, and the CR and EX consumed a low-fat diet, we were concerned that the observed differences in expression of metabolic genes might be due to differences in the macronutrient contents of the diets and not energy balance per se. To address this concern, confirmatory analysis of mRNA expression was done using the diet control mice as the reference group. A gene that was responsive to both CR and EX (Lep), two genes uniquely responsive to EX (Ucp1 and Ucp2), and three genes uniquely responsive to CR and relating to carbohydrate metabolism and transport (Slc2a4, Acly, and Sh2b) were selected for validation. RT-PCR analysis verified that Ucp1, but not Ucp2, was significantly increased by EX only (Figure 3). Although according to the microarray analyses, Lep was reduced by both CR and EX, RT-PCR analyses revealed that only CR significantly reduced Lep expression (Figure 3). All three genes relating to carbohydrate metabolism and transport, Acly, Slc2a4, and Sh2b2, were indeed significantly increased by CR only (Figure 3). Importantly, Slc2a4 codes for the insulin-responsive glucose transporter, Glut4. Translocation of Glut4 from the cytosol to the plasma membrane in response to insulin signaling is the rate-limiting step of glucose transport into the adipocyte. Furthermore, downregulation of Glut4 at the messenger RNA and protein levels has been implicated in obesity and insulin resistance. Although we lacked sufficient VWAT samples for an extensive protein analysis, due to the use of these tissues for genomic and other analyses, Western blot analyses for Glut4 protein expression on 3 VWAT samples/group showed similar trends as observed with the mRNA analyses. Specifically, the lowest Glut 4 protein expression was observed in VWAT from a control mouse, the highest expression was in a CR sample, and the samples from the exercise group were similar to the controls (data not shown). Finally, increases in the enzyme ATP-citrate lyase (Acly), which was also upregulated by CR but not EX, has recently been linked to increases in Glut4 mRNA levels [25]. For these reasons, we focused our analyses on elucidating how CR resulted in increased transcription of Glut4.

Figure 3

Confirmation of microarray data analysis of mRNA expression in VWAT. Expression of mRNA transcripts in VWAT from DIO control, CR and EX mice (n = 7/group). (a) Leptin (Lep), (b) Uncoupling protein 1 (Ucp1), (c) Uncoupling protein 2 (Ucp2), (d) Solute carrier 2, family 4 (Slc2a4) (e) ATP-citrate lyase (Acly), and (f) SH2B adapter protein 2 (Sh2b2) (Data shown are mean ± SE, *(P ≤ .05)).

3.5. CR Results in Acetylation of Histone 4 at the GLUT4 Promoter

Since regulation of acetylation of histones has been shown to be nutrient dependent [25], we hypothesized that increased Glut4 mRNA expression in CR mice may be the result of histone modifications at the Glut4 promoter. To test this hypothesis, we generated obese (DIO) and lean (CR) mice (relative to overweight control mice consuming AIN-76A diet ad libitum) through an 8-week diet intervention. Body weight and percent body fat were positively associated with fasting blood glucose levels and inversely related to Glut4 mRNA levels in the VWAT (Figures 4(b), 4(c), and 4(d)). Modifications to the histone code such as methylation, which can result in decreased transcription [26] or acetylation, which can result in increased transcription [25] may account for the differences in Glut4 transcription in VWAT, so both forms of epigenetic alteration were assessed. There were no differences in trimethylation of histone 4 at the Glut4 promoter (Figure 5(a)). However, CR significantly increased histone 4 acetylation at the Glut4 promoter compared to control mice (Figure 5(b)), which was associated with higher levels of Glut4 mRNA and increased insulin sensitivity.

Figure 4

Lean phenotype is associated with lower blood glucose levels and elevated levels of Glut4 mRNA relative to control mice. (a) Study design for chromosomal immunoprecipitation experiments. (b) Average body weight of mice on DIO, overweight (control AIN-76A) diet, or a lean (CR) regimen to generate obese, overweight or lean mice (n = 5/group). Data shown are mean ± SE. (c) Fasting blood glucose levels of mice after 8 weeks of respective dietary regimen (n = 5/group). Data shown are mean ± SE. Significance (P < .05) between groups is denoted by different letters. (d) Relative mRNA expression of Glut4 in VWAT (n = 5/group). Significance (P < .05) between groups is denoted by different letters.

Figure 5

Calorie restriction increases histone H4 acetylation at the GLUT 4 promoter. (a) Relative quantification of Glut4 DNA immunoprecipitated with anti-trimethyl H4 antibody. (b) Relative quantification of Glut4 DNA immunoprecipitated with anti-acetyl H4 antibody (n = 3/group). Data shown are mean ± SE, *(P ≤ 0.05).

4. Discussion

With over two thirds of American adults classified as overweight or obese [1], increased understanding of how best to reverse the harmful effects of obesity is urgently needed. Given the critical role of adipose tissue in regulating glucose homeostasis and other aspects of metabolism, analysis of the changes that occur in adipose tissue after weight loss could reveal novel targets for prevention or treatment of obesity-related diseases. To our knowledge, this is the first study to compare the effects of CR and EX (the two most commonly recommended lifestyle modifications to prevent or reverse obesity) on gene expression in adipose tissue in a model of DIO. The direct comparison of these two obesity reversal interventions revealed the following novel findings: (1) CR led to altered expression of more than 20 times the number of genes in the adipose tissue than were uniquely affected by EX; (2) alteration of expression of carbohydrate transport genes (particularly GLUT4) was uniquely affected by CR and correlated with the increased insulin sensitivity exhibited by CR; (3) upregulation of Glut 4 by CR may be explained in part by our finding that CR increased acetylation of histone 4 at the Glut4 promoter.

CR and EX both resulted in significant weight loss compared to sedentary DIO controls, which remained obese with a % body fat >50%. Although CR, and EX groups displayed comparable levels of percent body fat at the end of the intervention, only CR significantly improved insulin sensitivity. Exercise has been shown to increase insulin sensitivity in mice and humans [27, 28], although these effects are less clear in obese individuals such as the DIO mice used in this report. The relatively short intervention in our study may explain why EX was not as affective as CR at altering indices of insulin resistance. In other rodent studies that showed a significant affect of EX, the intervention was either more than 10 weeks long [10, 28, 29] or the intervention period was longer than the period of diet-induced obesity [15]. These differences in study design suggest that, in the short term, EX may not be as effective as CR in restoring insulin sensitivity.

CR has been shown to decrease expression of genes related to aging and tumorigenesis in multiple tissues [30]. However, there is a paucity of studies examining the effect of CR on adipose tissue following weight loss. More importantly, there are no studies, to our knowledge, directly comparing the effect of CR and EX on gene expression in adipose tissue. To our knowledge there are only two microarray studies comparing CR to EX; one was performed by our group in the mouse mammary gland [31] and the other by Lu et al. in mouse skin [32]. In these reports CR and EX exhibited distinct effects on gene expression, with CR impacting more than 4 times the number of genes than EX. In the present study, we found that this differential impact was more pronounced in adipose tissue, with CR affecting more than 20 times the number of genes altered by EX. Not only did CR induce a stronger quantitative effect than EX on genes that were qualitatively similar in their response to both CR and EX, but CR affected an additional 48 genes related to metabolism that were unaffected by EX. Of those genes, there was an overall upregulation of genes related to carbohydrate metabolism and glucose transport, including Glut4.

We also found that DIO downregulates multiple genes that play a role in lipid metabolism and upregulates a profile of genes related to immune/inflammatory response [33, 34]. Furthermore, many of the lipid metabolism genes shown to be decreased by DIO were increased by CR and EX in our study [33]. Likewise, immune response genes that have been shown to be increased by DIO were decreased after weight loss induced by CR or EX [33]. Together these data support previous findings that in the obese state, there is diminished fatty acid synthesis and transport, characteristic of insulin-resistant adipose tissue rich in immune cell infiltrates. Importantly, our data show that these processes are sensitive to both CR and EX interventions.

Many of the transcripts related to lipid and carbohydrate metabolism that were affected by both CR and EX in the present study were also shown by Shankar et al. to be induced by a high-carbohydrate diet [35]. Increased transcription of these genes is consistent with increased uptake of glucose and fatty acids into the adipose tissue. In the study by Shankar et al. these transcriptional changes were measured in rats fed a high-carbohydrate diet for 4 weeks, during which time the rats gained weight and the adipocytes hypertrophied, whereas the mice in our study first underwent DIO but then lost weight for 8 weeks before analysis. The similarities between the two studies are indicative of increased signaling through the insulin receptor/phosphatidylinositol 3-kinase (IR/PI3K) pathway that mediates glucose uptake and the lipogenic effects of insulin in adipose tissue.

Glucose uptake into adipose tissue is mediated by two different Glut isoforms: Glut1 and Glut4. Glut1 mediates basal uptake of glucose into adipocytes. Although others have reported that Glut1 mRNA increases with obesity [36], we did not observe any changes in Glut1 mRNA expression in the microarray. Translocation of the GLUT4 transporter from the cytosol to the membrane is the rate-limiting step in insulin-mediated glucose uptake in adipocytes and skeletal muscle [37]. The importance of Glut4 function in adipose tissue is underscored by the finding that overexpression of Glut4 in adipocytes rescues insulin resistance in mice with muscle-specific knockout of Glut4 [38]. However, expression of Glut4 in the muscle does not compensate for lack of Glut4 activity in adipose tissue [39], further implicating adipose tissue as a key metabolic organ in the etiology of insulin resistance. There is considerable evidence that Glut4 mRNA levels in adipose tissue decrease with obesity [40], and that increases in Glut4 mRNA in adipose tissue can ameliorate insulin resistance [41, 42]. Indeed, our finding that Glut4 mRNA levels were significantly increased by CR, but not by EX, and that this increase was associated with improved insulin sensitivity, supports this idea. Therefore, increased transcription of Glut4 in VWAT during weight loss may be a critical event in reversing insulin resistance.

Studies into the transcriptional regulation of Glut4 in skeletal muscle implicate a histone deacetylase (HDAC5) as a crucial mediator of changes to Glut4 mRNA levels in response to exercise [43]. Raychaudhuri et al. have also described a series of histone modifications mediated by histone deacetylases and histone methyltransferases that culminate in a metabolic knockdown of the Glut4 gene in the skeletal muscle of rats that had experienced intrauterine growth restriction [44]. Collectively, these studies suggest that transcriptional regulation of the Glut4 gene is highly responsive to changes in energy balance. This led to our hypothesis that Glut4 mRNA levels in adipose tissue could be subjected to similar transcriptional regulation. In support of this hypothesis, Wellen et al. recently discovered that during adipocyte differentiation, levels of global histone acetylation are dependent on glucose availability [25]. More specifically, acetylation of histones 3 and 4 at the Glut4 promoter is linked to increased Glut4 mRNA expression in response to higher concentrations of glucose during differentiation. Our in vivo ChIP data extend the in vitro findings and show that increased acetylation of histone 4 at the Glut4 promoter, which was associated with higher levels of Glut4 mRNA, occurred in lean mice that were highly insulin sensitive as indicated by significantly decreased fasting glucose levels. Taken together, these data suggest that insulin-responsive adipose tissue maintains H4 acetylation. This leads to increased transcription of Glut4 to facilitate continued glucose uptake. However, as adiposity increases so does insulin resistance [3]. Deregulation of signal transduction downstream of the insulin receptor results in decreased trafficking of Glut4 to the cell membrane [45, 46] and a decline in glucose flux into the adipocyte [47]. According to the findings of Wellen et al., limited glucose availability results in diminished histone acetylation and decreased Glut4 mRNA expression [25]. Therefore, in the context of obesity and insulin resistance, the lower levels of Glut4 mRNA expressed in adipose tissue may be a consequence of decreased insulin-mediated uptake of glucose that results in diminished histone acetylation at the Glut4 promoter. Further analyses are required to determine if other modifications to the histone code at the Glut4 promoter may be contributing to transcriptional repression of Glut4 mRNA in obesity.

In conclusion, these findings show that obesity reversal by CR versus EX results in many shared, but also many differential, changes in the adipose transcriptome. In particular, CR has specific and significant effects on the expression of key metabolic genes and pathways associated with obesity-related disease. In addition, some of the effects of these antiobesity interventions on VWAT gene expression and metabolism may result from chromatin remodeling, as illustrated by CR's effect on histone acetylation of the GLUT4 promoter. Taken together, these studies provide insights regarding new targets, including potential epigenetic-related regulation of key metabolic genes, such as Glut4 acetylation, for preventing or treating obesity-related diseases.