High-fat diets (HFDs) result in increased body weight. However, this is not uniform and determining the factors that make some animals or individual more susceptible to this diet-induced weight gain is a critical research question. The expansion of white adipose tissue (WAT) associated with weight gain requires high rates of angiogenesis to support the expanding tissue mass. We hypothesized that diet-induced obese (DIO) mice have a greater capacity for WAT angiogenesis and remodeling than diet-resistant (DR) mice at a young age, before age or DIO.
We measured body weight and body composition by nuclear magnetic resonance. We compared the expression of genes related to lipid metabolism, angiogenesis and inflammation by real-time, quantitative PCR and PCR arrays. WAT morphology and distribution of adipocyte size were analyzed. The level of hypoxia and vascular density was assessed by immunohistochemistry in WAT of young mice.
C57Bl/6 mice were DIO and FVB/N (FVB) mice DR after 8 weeks on a low-fat diet or HFD. However, C57Bl/6 mice had lower body weight, lower adiposity, smaller adipocytes and decreased leptin and lipogenic genes expression in adipose tissue than FVB mice at 9 weeks of age on a chow diet. Despite having smaller adipocytes, the level of hypoxia and the expression of pro-angiogenesis genes were higher in WAT of young C57Bl/6 mice than young FVB mice. In addition, expression of genes related to macrophages and their recruitment, and to proinflammatory cytokines, was significantly higher in WAT of young C57Bl/6 mice than young FVB mice.
These data suggest that the potential for WAT remodeling in early period of growth is higher in C57Bl/6 mice as compared with FVB mice, and we hypothesize that it may contribute to the increased susceptibility to DIO of C57Bl/6 mice.
Obesity is a complex and multifactorial disease. Early identification of obesity-prone humans may provide a strategy for preventing or reducing the incidence and severity of the disorder. Although it is clear that environmental factors such as consumption of calorically dense, high-fat diets (HFDs) can lead to increased adiposity, this response is not uniform. Rodent experiments make it clear that even when exposed to the same obesogenic diet and housed under the same conditions, there are important genetic differences that contribute to the susceptibility to gain weight on a HFD.
Modulation of genes related to thermogenesis, such as uncoupling protein 1, affects the development of diet-induced obesity (DIO) in mice,1 and hypothalamic neural circuits involved with regulation of energy balance are altered in offspring of DIO rats.2 Many hypotheses have been offered to explain the differences between DIO and diet-resistant (DR) animals. One common hypothesis is that differences in the key central nervous system circuits that regulate energy balance differ between DIO and DR animals. Consistent with this, there is an inborn genetic disposition to develop leptin or insulin resistance, altered glucose sensing and disturbed metabolism of norepinephrine, serotonin and dopamine in the hypothalamus of DIO rats.2
An alternative hypothesis is that a key difference between DIO and DR animals relates to the ability of white adipose tissue (WAT) to accommodate the storage of increased calorie intake. Analogous to tumors, WAT has the capability to quickly expand and proliferate throughout adulthood. Also similar to tumors, such expansion is highly dependent on the ability to generate new blood vessels3, 4 and as a result, WAT is highly vascularized and has angiogenic properties5 that are critical to its ability to expand.3, 4 Interestingly, treatments that inhibit WAT angiogenesis potently (but reversibly) reduce body weight and fat mass in rodents and non-human primates, indicating that WAT integrity and sustained weight gain depends on maintaining WAT vasculature.4, 6, 7
WAT angiogenesis is highly correlated with adipogenesis, inflammation and the need for ongoing tissue remodeling. Many different cell types in WAT, including mature adipocytes, preadipocytes, pericytes, endothelial cells (ECs) and inflammatory cells, are involved in the capacity for WAT to remodel itself under different circumstances.3, 8 Obese rodents and humans have elevated levels of angiogenic factors and macrophages in their WAT.9, 10 Such low-grade inflammation is progressively increased in WAT of mice fed a HFD and these changes have been hypothesized to contribute to obesity and insulin resistance.11 The surface area of adipocytes reflects their volume and is positively correlated with the number of macrophages.11, 12 As obesity develops, there is a switch to a more inflammatory phenotype in macrophages in WAT.13
In the present work, we hypothesized that at a young age and before exposure to a HFD, the WAT of mice that are susceptible to weight gain on a HFD have increased capacity for tissue expansion, including the capability of rapid angiogenesis, and that this capacity may contribute to the enhanced susceptibility to develop obesity. To this end, we compared the ability of two mouse strains to gain weight and undergo WAT remodeling when fed a HFD. While C57Bl/6 (C57) mice gained substantially more weight and fat on the HFD, FVB/N (FVB) mice were almost completely resistant to weight gain on the same HFD. Consistent with this, the capacity for WAT tissue remodeling was far greater in C57 compared with FVB mice, and this was the case before their exposure to the HFD.
Materials and methods
Male C57 mice and FVB mice obtained from Jackson Labs (Bar Harbor, MA, USA) at 8 weeks of age were housed individually in standard mouse cages with a 12-h light/12-h dark schedule. Sixteen of each genotype had ad lib access to a HFD providing 40% calories as fat (HFD, D03082706, Research diet, Inc., New Brunswick, NJ, USA) or a low-fat diet with 4% fat (LFD, D03082705) for 8 weeks. In two separate experiments, a total of 16 mice of each genotype were fed a chow diet and killed at 9 weeks of age after adaptation to the cage for 1 week for gene expression or histological studies. All animal protocols were approved by the University of Cincinnati Institutional Animal Care and Use Committee.
Body composition was analyzed by an nuclear magnetic resonance (Echo Medical Systems, Houston, TX, USA) on weeks 0, 8 and 12 in the first experiment, and at 9 weeks of age in the second study.
Adipocyte tissue morphometry
To determine adipocyte size, the cross-sectional area of adipocytes was determined on paraffin-embedded hematoxylin and eosin-stained sections of epididymal fat at a magnification of × 200 by image processing with custom-designed software. The method of determining adipocyte size was modified from previous papers using a computer image analysis.14 The distribution of adipocyte size was determined by relative frequencies of adipocytes having a size within a regular interval. The number of adipocytes was determined by dividing the total fat mass by the mean adipocyte size. A total 500 to 1200 adipocytes were used for the analysis of the mean adipocyte area and the distribution of adipocyte size for each mouse.14
To assess the degree of hypoxia in WAT, a Hypoxyprobe-1 Plus Kit (Hypoxyprobe, Inc., Burlington, MA, USA) was used. Five mice in each group were intraperitoneally injected with 60 mg kg−1 of a Hypoxyprobe-1 (pimonidazole HCl) 30 min before being killed. Epididymal fat was fixed in alcoholic zinc formalin (Anatech Ltd, Battle Creek, MI, USA) overnight. The adipose tissue was embedded in a paraffin in pairs containing one C57 and one FVB mouse each per block to minimize any differences that might arise during the process of immunohistochemistry and sectioned at the thickness of 5 μm. The immunohistochemistry was performed and a total of 200 images (five images per section) were captured at a magnification of × 100 (AxioVision 4.6 microscope; Carl Zeiss, Thornwood, NY, USA). To correlate hypoxia with adipocyte size, the brown color signifying hypoxia was selected by using HSL color system (Hue 0–25 and 253–255; Saturation 25–255; and Luminescence 0–225) and the number and the cross-sectional area of adipocytes were concurrently measured in the same field with the software. The degree of hypoxia was determined by dividing the area of brown color by the number or the averaged cross-sectional area of adipocytes in the same field.
For the detection of PECAM1, sections were incubated overnight at 4 °C with a rat anti-mouse CD31 antibody (1:500; BD Pharmingen, San Diego, CA, USA). After incubation with biotinylated anti-rat IgG (1:200; Jackson ImmunoResearch Laboratories, West Grove, PA, USA), sections were revealed with avidin-biotin complex (1:500; Vector Laboratories, Burlingame, CA, USA). Sections were counterstained with hematoxylin and eosin. The numbers of adipocytes and vessels were analyzed by three independent individuals blinded to the experimental conditions.
Mice (8 per condition) were killed after a 1-h fast. The epididymal fat was quickly frozen in 2-methyl butane on dry ice. An RT2 First Strand Kit and RT2 Profiler PCR Array System (SABiosciences Corporation, Frederick, MD, USA) was used to perform PCR arrays related to angiogenesis. Total RNA was isolated using RNAeasy columns (Qiagen, Valencia, CA, USA) after processing with Tri-reagent (Molecular research center, Inc., Cincinnati, OH, USA). The first strand cDNA and the cocktail solutions of cDNA including RT2 qPCR Master Mix were prepared for each mouse. PCR arrays (eight in each group) were performed using an iCycler (Bio-Rad Laboratories, Hercules, CA, USA). The quality control tests for genomic DNA control, reverse transcription control and positive PCR control of each array plates were all passed below the criteria. The array data were analyzed using RT2 Profiler PCR Array Data Analysis. The names of all genes are presented using official symbol of NCBI in this study (Figures 2, 3, 4, 5 and Supplementary Table 2). Three housekeeping genes not influenced by the experimental condition (Hsp90ab1, Gapdh and Actb) were selected as an endogenous control. The normalization and analysis of data were performed according to the instructions (SABioscience Corporation).
Real-time quantitative PCR
Total RNA was isolated from epididymal fat of each mouse in the same way as above, and cDNA was synthesized using iScript (Bio-rad). PCR was performed in triplicate using an iCycler and the iQ SYBR Green Supermix (Bio-rad Laboratories) or 7900HT Fast Real-Time PCR system (Applied Biosystems, Carlsbad, CA, USA; Supplementary Table 1).14 The expressions of each gene were normalized to constitutively expressed L32 and relative expression was quantified.14 The names of all genes were presented using the official symbol of NCBI in this study (Figures 2, 3, 4, 5 and Supplementary Table 2).
All data are expressed as mean±s.e.m. Data were analyzed by Student’s t-test or two-way repeated measures analysis of variance followed by Bonferoni’s multiple comparisons using Prism 5.0 (La Jolla, CA, USA).
C57 mice are prone to DIO, whereas FVB mice are resistant
To test whether FVB mice are resistant to DIO relative to C57 mice, we compared the change in body weight and energy intake by feeding both genotypes a LFD or HFD for 8 weeks. The body weight of C57 mice was lighter relative to FVB mice at 9 weeks of age at the start of the experiment (Figure 1a). C57 mice increased their weight by 20% after 8 weeks on the HFD compared with those on the LFD. By contrast, the FVB mice fed the HFD gained comparable weight as those on the LFD (Figure 1b). Likewise, the fat mass of C57 mice on a HFD was greater than that on a LFD, whereas that of FVB mice on a HFD was not significantly different from that of LFD mice after both 5 and 12 weeks (Figure 1c). The difference in cumulative energy intake between LFD and HFD was significantly greater in C57 mice compared with FVB mice (Figures 1d and e). Given the lower energy expenditure in C57 mice reported in previous studies,15, 16 these data support a greater energy surplus in C57 mice relative to FVB mice. These data confirm the differential susceptibility to a HFD between DIO C57 and FVB mice.
Young C57 mice fed chow have smaller adipocytes and lower adiposity than young FVB mice
The body weight of C57 mice was actually lower due to both lower adiposity and less lean mass relative to FVB mice at 9 weeks of age on a chow diet (Figures 2a and b). The cross-sectional area of individual adipocytes was smaller in C57 mice than FVB mice and there were more small adipocytes and fewer large adipocytes in C57 mice (Figures 2c and d). There was also a trend toward a decrease in the number of adipocytes in C57 mice relative to FVB mice (P=0.06). Consistent with the lower adiposity and smaller adipocytes, the level of gene expression of Lep was lower in WAT of C57 mice relative to FVB mice (Figure 2e).
To determine whether the difference in adipocytes between young C57 mice and FVB mice is associated with differential adipose lipid metabolism, we compared the expression of genes involved in adipogenesis, utilization of fatty acid, lipogenesis and lipolysis in WAT of both the groups. There was no significant difference between genotypes in expression of genes for transcriptional factors involved in adipogenesis such as Pparg2 and Srebf. The expression of genes involved in lipogenesis, including Fasn and Acaca, were significantly lower in C57 mice relative to FVB mice. Lpl and Cd36 are associated with release of fatty acids from circulating lipoproteins and uptake by WAT.17, 18 These genes involved in fatty acid utilization were significantly lower in WAT of C57 mice relative to FVB mice (Figure 2g). The expression level of Adrb3 (beta 3-adrenergic receptor) is associated with lipolysis and sympathetic activity in WAT,19, 20, 21 and is lower in WAT of obese rodents.22 Consistent with the lower adiposity of C57 mice, Adrb3 gene expression was higher in WAT of C57 mice relative to FVB mice (Figure 2g).
Young C57 mice on a chow diet have higher angiogenic activity in adipose tissue than young FVB mice
The growth or regression of vasculature depends on the net balance between angiogenic stimulators and angiogenic inhibitors, and the process of angiogenesis comprises modulation of migration and proliferation of ECs, conversion of ECs to pericytes, recruitment of pericytes and smooth muscle cells and stabilization of vessels.23, 24, 25 Because the expandability of WAT highly depends on its vasculature,4 we hypothesized that C57 mice, with their tendency to develop DIO when fed a HFD, have a higher angiogenic potential in WAT at a young age, before exposure to a HFD. To test this, we determined the expression of 84 genes involved in the process of angiogenesis by real-time PCR array. A total of 32% of angiogenic genes were differentially expressed in WAT between C57 mice and FVB mice, with 78% of the significantly changed genes being more highly expressed in C57 relative to FVB mice.
Genes for growth factors that increase migration and proliferation of ECs or stimulation of vascular endothelial growth factor (VEGF) such as Vegfa, Egf, Fgf1, Figf, Hgf and Igf1 were all expressed at higher levels in WAT of C57 mice relative to FVB mice, whereas Pgf, which stimulates angiogenesis only in pathological conditions (Figure 3a),25, 26 was lower. Angpt1, which maintains a normalized state of vasculature, was highly expressed in C57 mice, while Tek, a receptor for Angpt1 and Angpt2, and Angpt2, an antagonist ligand for Tie-2, were not different, indicating that stabilization of WAT vasculature and pro-angiogenesis was increased in WAT of C57 mice relative to FVB mice.26, 27 However, Ctgf, which is associated with fibrosis, proliferation of mesenchymal cells, including pericytes and vascular smooth muscle cells, and inhibition of VEGF-induced angiogenesis,28, 29 and Pdgfa, which is related to proliferation of mesenchymal cells and an autocrine regulator of fibroblast growth factor (FGF)-2 and VEGF, were lower in WAT of C57 mice (Figure 3a).30 This may imply that the process of supporting actively sprouting vessels is less active in WAT of C57 mice than FVB mice.
Transcriptional factors such as Epas1(Hif-2alpha), Smad5, Tbx1 and Tbx4 were highly expressed in WAT of C57 mice, whereas Hif-1alpha was not different between genotyes (Figure 3b). Epas1 promotes angiogenesis in response to hypoxia and Smad5 is associated with ECs proliferation.31, 32 Tbx1 and 4 are associated with FGF signaling, angiogenesis or morphogenesis during development.33, 34 The difference in expression of these transcription factors is consistent with a more pro-angiogenic environment in WAT of C57 mice (Figure 3b).
The expression of Thbs1 and Thbs2, endogenous anti-angiogenic factors, was lower in WAT of C57 mice (P=0.07 and P<0.05, respectively; Figure 3c). Other anti-angiogenic factors including Col18a1 and Col4a3, a major constituent of basement membrane cleaved by MMP-9, tended to be lower (P=0.07) or were significantly lower in WAT of C57 mice relative to FVB mice (Figure 3c).25, 35 These lower anti-angiogenic factors are also consistent with the greater pro-angiogenesis in WAT of young C57 mice. In addition, genes regarding degradation of the extracellular matrix such as Mmp9 and Mmp19 were higher in C57 mice, favoring angiogenesis, but Mmp2 was lower (Figure 3c).
Stab1 is expressed by sinusoidal endothelium or alternatively by activated macrophages and acts as a scavenger receptor for endocytosis of SPARC (secreted protein, acidic, cysteine-rich), a potent inhibitor of angiogenesis.36 Expression of Stab1 was higher in WAT of C57 mice. Ptgs1, which is related to enhanced tube formation or vasodilation, was also higher in WAT of C57 mice (Figure 3d).37
Although most of the significantly changed genes were pro-angiogenic, some of the genes were either anti-angiogenic or have mixed function. Flt-1 is increased by hypoxia, but it is not clear whether it mediates VEGF signaling in its intact form, or has a role as a vascular permeability receptor, or sequesters VEGF and subsequently antagonizes VEGF action in a soluble form under some circumstances.24, 38, 39 In the present study, WAT of C57 mice had higher Flt-1 gene expression relative to FVB mice. Npr1 is related to vasodilatation via increased guanylyl cyclase activity and inhibits synthesis and function of VEGF activities.40 Lect1 contributes to growth of chondrocytes and inhibits angiogenesis (Figure 3d).41, 42 Npr1 and Lect1were higher in WAT of C57 mice.
Genes involved in modulation of cell-to-cell contacts, such as Cdh5, Efna1, Efnb2, Ephb4 and Pecam1, were not different between C57 mice and FVB mice. In addition, the expression of Itgav and Itgab3 genes that are highly expressed in activated ECs25 was not different. Consistent with no difference in expression of these genes, the number of vessels per adipocyte was not different although the numbers of adipocytes or vessels per field were higher in WAT of young C57 mice relative to FVB mice (Figures 4a and b). These data are consistent with the hypothesis that the status of ECs does not differ between C57 mice and FVB mice maintained on a chow diet.
In sum, the analysis of these gene expression differences indicates little difference in the state of WAT ECs interacting with other ECs or pericytes between C57 mice and FVB mice. However, WAT of C57 mice had considerably greater expression of angiogenic factors compared with FVB mice. In particular, growth factors involved in proliferation and migration of ECs were mostly higher in WAT of C57 mice, and anti-angiogenic factors such as Thbs1 and Thbs2 were lower.
WAT of young C57 mice is more hypoxic than that of FVB mice
Within a 100-μm distance from capillaries, the exchange of oxygen or nutrients is the result of diffusion.5 Hypoxia is a key stimulator of angiogenesis. We therefore hypothesized that WAT hypoxia contributes to more pro-angiogenesis in WAT of C57 mice relative to FVB mice. To test this hypothesis, we compared the hypoxic area in WAT between C57 and FVB mice using immunohistochemistry. Because the staining area in WAT is highly dependent on the number and size of adipocytes, we also compared the hypoxic area after adjusting for number and size of adipocytes in the same field. The absolute area of hypoxia was significantly increased in WAT of C57 mice relative to FVB mice. Importantly, the area of hypoxia adjusted for either adipocyte number or size was still significantly increased in C57 mice relative to FVB mice (Figure 4c). These data suggest that WAT hypoxia may contribute to the increased expression of pro-angiogenic genes in WAT of C57 mice relative to FVB mice.
Young C57 mice on a chow diet have more inflammatory adipose tissue relative to young FVB mice
Because angiogenesis contributes to tissue remodeling, we hypothesized that low-grade inflammation may be associated with the greater weight gain on the HFD in C57 mice. To test this hypothesis, we compared the expression of genes related to inflammatory cytokines and macrophages in WAT between young C57 and FVB mice. A macrophage marker, Emr1, as well as genes related to recruitment of macrophages such as Ccr2 and Itgax, were significantly increased in WAT of C57 mice relative to FVB mice. Proinflammatory cytokines, including Il1b, Il6, Ccl3 and Cxcl2, were significantly increased in WAT of C57 mice relative to FVB mice, whereas genes related to alternative activation of macrophages such as Jag1, Tgfb1 and Mgl2 were not significantly different. However, Arg1 was significantly increased in WAT of C57 mice relative to FVB mice (Figure 5).
The goal of these studies was to identify potential factors that might contribute to the differences observed in weight gain among animals exposed to a HFD. Long-term maintenance on a HFD increased body weight and adiposity of C57 mice, whereas those of FVB mice were not significantly increased. The energy surplus in C57 mice was likely the result of both increased energy intake (Figure 1) and decreased energy expenditure.15, 16 This is consistent with previous data on these strains of mice (Figure 1).16 However, the difference of body weight between these strains at a younger age was not consistent with what was observed after 8 weeks on the HFD. Specifically, the body weight and adiposity were significantly lower in C57 mice relative to FVB mice at 9 weeks of age before exposure to a HFD. Consistent with the lower adiposity, leptin gene expression was lower in WAT of C57 mice relative to FVB mice and their adipocyte size was smaller. In addition, the amount of WAT correlated with expression of genes linked to lipid metabolism, including lower Lpl, Cd36, and Fasn, and higher Adrb3 in C57 compared with FVB mice.
The key finding in this study is that young C57 mice have an expression profile consistent with greater inflammation and angiogenesis in WAT despite having less body fat than young FVB mice. Such data suggest that young C57 mice may already have a higher capacity for WAT remodeling.43, 44 The number of adipose tissue macrophages is positively correlated with increased size of individual adipocytes.12 Interestingly, C57 mice have increased expression of inflammatory markers in WAT despite having reduced adipocyte size. WAT remodeling dynamically occurs both after acute weight loss or when WAT rapidly grows such as occurs in C57 mice exposed to a HFD.43 As expression of genes associated with WAT lipid metabolism correlated with the increased adiposity of young FVB mice (Figure 2g), neither fatty acid metabolism nor adipocyte death appears to contribute to increased macrophage and pro-inflammatory cytokine markers in young C57 mice.43, 45
Numerous cell types in WAT, including immune cells, ECs, preadipocytes and mature adipocytes, are involved in the process of tissue remodeling as WAT expansion and adipocyte differentiation require an adjusted vascular network and angiogenesis precedes adipogenesis.3, 5, 8, 44, 46 In addition, angiogenesis accompanies inflammation, and increased macrophage infiltration has been hypothesized to promote angiogenesis in WAT.8, 47 During this process, hypoxia is a critical factor to initiate angiogenesis and it induces inflammation in WAT.43 Larger adipocytes suffer from lack of oxygen and nutrient transfer because oxygen exchange usually occurs by diffusion through adipocytes.5 Given the smaller adipocytes of C57 mice relative to FVB mice, it is an interesting finding that levels of hypoxia evidenced by increased pimonidamole adduct staining and gene expression of Hif-2alpha are increased in WAT of C57 mice consistent with higher expression of pro-angiogenic and pro-inflammatory genes. WAT hypoxia in obesity has been observed in previous studies.48 Those studies suggest that the increased WAT hypoxia in young DIO mice is important in initiating increased angiogenesis and the mild inflammation that contributes to the necessary tissue remodeling that allows C57 mice to gain more weight when exposed to a HFD.
In this study, several important pro-angiogenic factors, including Vegfa, Fgf1, Egf, Hgf, Igf1, Figf1 and Angpt1, were highly expressed and anti-angiogenic factors such as Thbs1, Thbs2, Col18a1 and Col4a3 were lower in WAT of young C57 mice, suggesting an important difference in angiogenic potential between C57 and FVB mice. For example, Vegfa is a key target for Hif-1alpha and has a critical role in the angiogenic response to hypoxia. Vegfa stimulates angiogenesis in WAT and VEGF receptor-2 signaling contributes to both WAT angiogenesis and adipogenesis.3, 24 Hepatocyte growth factor is another WAT pro-angiogenic factor that does not directly promote adipocyte differentiation.49 Both VEGF and hepatocyte growth factor synergistically stimulate neovascularization, endothelial migration and proliferation.50 Angiopoietin 1, which stabilizes vessels, also has a role in vascular remodeling by promoting angiogenesis in conjunction with VEGF.51
Of note, the gene expression level of Fgf1 was dramatically upregulated in young C57 mice relative to young FVB mice, whereas there was no difference of Fgf2, a potent and common angiogenic stimulator. FGF1 simulates adipogenesis in WAT of humans,52 and FGFs are well known to stimulate angiogenesis in tumors by increasing VEGF-driven neovascularization.25, 26 Although a role of each of these genes in angiogenesis has been reported, many of these gene products have also been reported to regulate inflammation and adipogenesis.23, 43
Given the growing prevalence of obesity, an important question is why individual’s susceptibility to obesity varies. The present study used C57 mice and FVB mice to explore differences in WAT that may contribute to this variable response to obesogenic environments. Although energy balance clearly involves central nervous system processes that regulate both food intake and energy expenditure, growing evidence points to an important role in signals from WAT that are related to the status of WAT vasculature.6 Consequently, the one way to view the current data is that C57 and FVB mice differ in the capacity of their WAT to expand and remodel and that this contributes to the propensity for increased intake and weight gain when mice are exposed to a HFD. Further work will be needed to assess the causal relationships between adipose tissue remodeling and energy balance (Figure 6).
This work was supported by National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) Grant DK093848 and Basic Science Research Program of the National Research Foundation of Korea (NRF) funded by the Ministry of Education, Science and Technology (2012-0003537).
About this article
Supplementary Information accompanies the paper on International Journal of Obesity website (http://www.nature.com/ijo)
Oxygen restriction as challenge test reveals early high-fat-diet-induced changes in glucose and lipid metabolism
Pflügers Archiv - European Journal of Physiology (2015)