ACAT-related enzyme 2 required for viability 1 (ARV1) is a putative lipid transporter of the endoplasmic reticulum that is conserved across eukaryotic species. The ARV1 protein contains a conserved N-terminal cytosolic zinc ribbon motif known as the ARV1 homology domain, followed by multiple transmembrane regions anchoring it in the ER. Deletion of ARV1 in yeast results in defective sterol trafficking, aberrant lipid synthesis, ER stress, membrane disorganization and hypersensitivity to fatty acids (FAs). We sought to investigate the role of Arv1 in mammalian lipid metabolism.
Homologous recombination was used to disrupt the Arv1 gene in mice. Animals were examined for alterations in lipid and lipoprotein levels, body weight, body composition, glucose tolerance and energy expenditure.
Global loss of Arv1 significantly decreased total cholesterol and high-density lipoprotein cholesterol levels in the plasma. Arv1 knockout mice exhibited a dramatic lean phenotype, with major reductions in white adipose tissue (WAT) mass and body weight on a chow diet. This loss of WAT is accompanied by improved glucose tolerance, higher adiponectin levels, increased energy expenditure and greater rates of whole-body FA oxidation.
This work identifies Arv1 as an important player in mammalian lipid metabolism and whole-body energy homeostasis.
ACAT (Acyl-CoA cholesterol acyl transferase) related enzyme 2 required for viability 1 (ARV1) is a transmembrane protein of the endoplasmic reticulum (ER) that is conserved between plants, yeast and mammals.1, 2 ARV1 was originally identified in a screen for genes required for viability in the absence of ARE1 and ARE2—the yeast homologs of Acyl-CoA cholesterol acyl transferase 1 and 2.2 Yeast deficient in ARV1 accumulate sterols in the ER, and show reduced sterol content in the plasma membrane based on biochemical membrane fractionation,2 as well as filipin staining.3, 4, 5, 6 Concomitant with ER lipid retention, yeast lacking ARV1 also harbor major defects in phospholipid, sphingolipid7 and glycosylphosphatidylinositol synthesis,4 and are particularly susceptible to sterol-binding antibiotics such as nystatin2, 8 and amphotericin B.6 Lipid accumulation in the ER in ARV1-deficient yeast is accompanied by severe ER stress and activation of the unfolded protein response.5 Most recently, Ruggles et al. performed a genome-wide screen for palmitoleate-induced fatty acid (FA) toxicity in yeast.9 Yeast strains deleted for ARV1 are viable but highly sensitive to unsaturated FAs. In this same study, overexpression of human ARV1 increased lipid synthesis in HEK293 cells and mouse liver, whereas knockdown of Arv1 in MIN6 pancreatic β-cells and HEK293 cells inhibited lipogenesis and caused lipoapoptosis.9
The ARV1 protein contains a conserved N-terminal zinc ribbon motif known as the ‘Arv1 homology domain,’ (AHD) which is followed by several transmembrane regions. The predicted and experimentally defined topology of the yeast Arv1p indicated the AHD is cytosolic. This is followed by a transmembrane region, a large loop region located inside the ER lumen, and two more transmembrane domains with the extreme C-terminus of the protein extending back into the ER lumen10 (Figure 1a). The murine Arv1 gene encodes a 266 amino acid (AA) protein. In contrast to its 321 AA yeast counterpart, murine Arv1 differs in that the AHD is preceded by a N-terminal 28 amino acid leader sequence (5–6). Expression of human ARV1 functionally complements the loss of ARV1 in yeast, rescuing all of the known growth and viability defects.2, 7 Despite this functional conservation, very little is known about the biochemical activity of Arv1 in any organism. In the current study, we sought to test the involvement of Arv1 in mammalian lipid metabolism by generating mice with a germline deletion of the gene. To our surprise, Arv1 knockout (Arv1 KO) mice have a striking metabolic phenotype including major reductions in white adipose tissue (WAT) mass, improved glucose tolerance and increased energy expenditure. These findings are consistent with an important role for Arv1 in FA homeostasis, and identify it as a novel regulator of body composition and energy expenditure in mice.
Materials and methods
Generation of Arv1 knockout mice
A genomic library derived from female 129S6/SvEvTac (Taconic, Hudson, NY, USA) mice was screened with a cDNA probe corresponding to the Arv1 gene to identify positive BAC clones. Two BAC clones were used to generate the targeting vector illustrated in Figure 1. Positive embryonic stem cell clones (R1 cells, SCRC-1011, American Type Culture Collection, Manassas, VA, USA) were selected with G418, and genomic integration was confirmed by Southern blotting with an AseI digest. Mycoplasma testing was not performed on embryonic stem cells before injection. Correctly targeted embryonic stem cells were injected into blastocysts from C57Bl/6J mice. Male chimeras were bred to C57BL6/J mice to obtain F1 progeny. These mice were then crossed with FLP recombinase transgenic mice (C57BL6/J) to remove the neomycin resistance gene, creating a conditional Arv1 allele with exons 2 and 3 flanked by loxP sites. The mice harboring the conditional Arv1 allele were then crossed with CMV-Cre transgenic mice on a C57BL6/J background from Jackson Labs to delete Arv1 in the germline (B6.C-Tg(CMV-cre)1Cgn/J, Jax stock number 006054), and backcrossed an additional two to four times before experiments.
The Cre transgene was detected by PCR using published primers—Cre-For: 5′-ACGACCAAGTGACAGCAATG-3′, Cre-Rev: 5′-CTCGACCAGTTTAGTTACCC-3′; 350 bp.11 Primers from Taconic for mouse chromosome 11 were included as a positive control (1260_1: 5′-GAGACTCTGGCTACTCATCC-3′, 1260_2: 5′-CCTTCAGCAAGAGCTGGGGAC-3′; 585 bp). The WT Arv1 allele was detected using the primers WTArv1For: 5′-CTTTGTTATGTTGTACAGGGTTAC-3′, WTArv1Rev: 5′-CCCGTGGACATGGACATGGGAAC-3′; 218 bp, where the forward primer binding site is unique to the WT allele in Intron 1. The floxed Arv1 allele was detected using the primers FlxFor: 5′-GCAGCCCAATTCCGATCATATTC-3′, FlxRev: 5′-CAGGAACAGAAATCACAGGCAG-3′, to amplify a 162 bp fragment of intron 1 where the forward primer binds to the loxP site. In the reactions to detect the WT and floxed alleles, primers that detect the murine IL2 genomic sequence from JAX were included as a positive control (JAX-oIMR7338: 5′-CTAGGCCACAGAATTGAAAGATCT-3′, JAX-oIMR7339: 5′-GTAGGTGGAAATTCTAGCATCATCC-3′; 324 bp). The deleted (null) Arv1 allele was detected with the primers: GENO3WTARV1F: 5′-CATTTGAGAGGGTAGCACAACTAC-3′, GENO-WTARV1-Rev: 5′-GACTTCCTCTCAGACTCCTCTACCGTGAG-3′ and GENO3DELARV1R: 5′-CCTGGCTTAAGGCAAACGCTTCTGAAG-3′. The first two primers flank the loxP site in intron 1, while the third primer binds to intron 3, which is brought close enough for amplification when the allele is deleted (Floxed 409 bp, WT 391 bp, null 182 bp). Annealing temperatures for PCR were as follows: Cre—60 ºC, WT—58 ºC, Floxed—58 ºC, Null—55 ºC. To detect the WT allele, a 2:1 molar ratio of target primers to internal control primers was optimal.
Animal care and treatment
All mice were housed in a pathogen-free animal facility with a daylight cycle from 0700 to 1900 h. Animals were allowed free access to food and water, and were maintained on a standard rodent chow diet (Purina 5010). One cohort of animals were fed a western-type high-fat diet containing 0.2% cholesterol and 45% kcal per fat (Research Diets D12079B) from weaning (Supplementary Data). Non-fasted body weights were measured once weekly on the weekday of birth beginning at the time of weaning. Fasting for lipid and glucose phenotyping was performed for 4 h to allow for clearance of intestinally derived triglycerides (TGs). Blood glucose was measured via the tail vein with the OneTouch Ultra glucometer following a 4-h fast (Lifescan; Johnson & Johnson, Milpitas, CA, USA). Immediately after tail glucose measurement, plasma was obtained for lipid analysis by retro-orbital bleeding with heparinized natelson collecting tubes under the influence of isofluorane anesthesia. Plasma insulin levels were measured with the Ultra Sensitive Mouse Insulin ELISA (Crystal Chem, Downers Grove, IL, USA).12 All the procedures were performed according to the regulations and with the prior approval of the Institutional Animal Care and Use Committees of the University of Pennsylvania (protocol 704534) and Baylor College of Medicine (protocol AN-6243).
RNA isolation and real-time reverse transcriptase PCR
Total RNA was isolated from liver using TriZol (Invitrogen, Grand Island, NY, USA). The QIAGEN RNeasy mini kit was used to isolate RNA from nonhepatic tissues, with the exception of adipose and brain, in which cases the QIAGEN RNeasy lipid tissue mini kit proved superior. RNA (1 μg) was reverse transcribed into cDNA using the SuperScript III Kit (Invitrogen) or the iScript kit (Bio-Rad, Hercules, CA, USA). The cDNA products were diluted 1:20 into a final reaction volume of 12.5 μl in the wells of a 384-well reaction plate containing 300 nm of each forward and reverse primer, and 6.25 μl of 2X SYBR green master mix (Applied Biosystems, Grand Island, CA, USA). The following primers were used to amplify exons 2 and 3 of mouse ARV1: forward 5′-TCAGGAGCTGTACCGGGACTA-3′, reverse 5′-CTGCAGCTGCCACCACCGCAGGTATGCTTCA-3′. Beta actin was amplified with the following primers: forward 5′-TTGGGTATGGAATCCTGTGG-3′, reverse 5′-CTTCTGCATCCTGTCAGCAA-3′. A Taqman assay to exons 2 and 3 was also used to measure Arv1 expression (Mm01253489_m1, Applied Biosystems) with Premix Ex Taq (RR390L, Clontech, CA, USA). Relative quantities were calculated using the delta delta Ct method.13
WAT and brown adipose tissue (BAT) were harvested and fixed in 10% buffered formalin overnight, dehydrated in a series of ethanol washes and embedded in paraffin. Tissue sections were stained with the hematoxylin and eosin by the Penn Cardiovascular Institute Histology Core and the Texas Medical Center—Digestive Diseases Center, Cellular Morphology Core.
Plasma total and high-density lipoprotein cholesterol (HDL-C) were measured enzymatically on a Cobas MIRA II (Roche Diagnostic Systems, Indianapolis, IN, USA) using reagents from Diagnostic Chemicals Limited. Plasma adiponectin levels were measured using mouse adiponectin ELISA kits from Millipore Corporation (Billerica, MA, USA). Hepatic TG levels were measured by homogenizing liver in three volumes of phosphate-buffered saline. The liver lysates were diluted 1:5 in phosphate-buffered saline, and 20 μl was transferred to the well of a microplate. TGs were solubilized with the addition of 20 μl of 1% deoxycholate to each well for 5 min at 37 °C. TGs were then measured using the Infinity triglyceride reagent (Thermo, Grand Island, NY, USA).
Glucose tolerance tests
Intraperitoneal glucose tolerance tests were performed on mice fasted overnight (16 h). Mice were given an intraperitoneal injection of 2 g glucose per kg body weight (BW), and blood glucose levels were measured at 0, 15, 30, 60, 90 and 120 min later via the tail vein.
Metabolic phenotyping and activity measurements
Body composition was measured by DEXA (dual-energy X-ray absorptiometry) scan and magnetic resonance imaging by the Mouse Phenotyping, Physiology and Metabolism Core at the University of Pennsylvania School of Medicine as previously described.14 Metabolic rates were measured by indirect calorimetry in open-circuit oxymax chambers that are a component of the Comprehensive Lab Animal Monitoring System (Columbus Instruments, Columbus, OH, USA). Mice were housed singly and maintained at 24 °C under a 12-h light–dark cycle (light period 0700–1900 h). Food and water were available ad libitum. All the mice were acclimated to monitoring cages for 24 h before beginning the physiological recordings. To calculate oxygen consumption (VO2), carbon dioxide production (VCO2) and RER (ratio of VCO2 to VO2), gas concentrations were measured at the inlet and outlet of the sealed chambers. VCO2, VO2 and calculated heat were normalized to lean mass determined by magnetic resonance imaging scan for each animal.
Oral fat tolerance and whole-body FA oxidation
Female mice were fasted for 14 h before oral fat tolerance testing. Mice received an oral gavage of 15 μl of olive oil per gram of body weight. Plasma was collected via retro-orbital bleeding under anesthesia before gavage, and then at 1, 3, 5 and 7 h later. Plasma TGs were measured by using the Infinity Triglycerides enzymatic assay kit from Thermo Fisher (Grand Island, NY, USA), using human control serum as a standard from 0 to 30 μg per well (CS1 human serum, Wako Diagnostics, VA, USA). Whole-body FA oxidation was measured on the basis of previously published methods15, 16 by the Mouse Metabolic Phenotyping Core at Penn. Briefly, mice were fasted 24 h before the experiment, injected intraperitoneally with 500 μl of [3H]-oleic acid (10 μCi, 2 mm oleic acid: 2% BSA) and killed exactly 30 min later for plasma. An aliquot of plasma (15 μl) was transferred to a new tube and then diluted with 50 μl of 10% BSA and 0.5 ml water. Next, 50 μl of perchloric acid (HClO4) were added to the sample to precipitate albumin-bound FAs. The aqueous phase (containing [3H]-H2O resulting from FA oxidation) was measured by scintillation counting. The counts were corrected for the specific activity of the total [3H] in the plasma and normalized to the fasting body weight of each animal, reported as microgram oleate oxidized per kilogram body weight per hour.
Animal numbers were estimated on the basis of previous experience with each specific assay and expected effect size. As comparisons were made across genotypes, no pre-randomization was performed. Researchers were not blinded to genotype. Animals were sex-matched and age-matched for all the experiments as detailed in the figure legends. Data were normally distributed and similar variances were observed in both the groups. All the data are presented as means±s.d. Differences between the two groups were assessed with a two-tailed, unpaired Student’s t-test. Comparisons involving multiple groups were tested by one-way analysis of variance and Tukey’s post test. P-values <0.05 were considered statistically significant (*).
To test the role of Arv1 in mammalian lipid metabolism, we first examined Arv1 RNA expression in mouse tissues. Arv1 messenger RNA was present at a low level in all the tissues with greatest expression observed in the kidney, testes and BAT (Figure 1b, Supplementary Figure 1). We next used gene targeting to generate mice with a global deletion of Arv1 (Figures 1c and d). Loss of critical exons 2 and 3 produces a nonfunctional message, which encodes the first 53 AAs of the 266 AA ARV1 protein followed by 18 novel frameshifted AAs and a premature stop codon. This truncated protein fragment (53/266 total AAs) would lack the last cysteine residue in the putative zinc ribbon motif6 and all transmembrane domains. Liver Arv1 messenger RNA levels were reduced by >35% in the heterozygous and 95% in the homozygous knockout mice (Figure 1e, P<0.05). Arv1 KO mice were born in the expected Mendelian ratio. Female KO mice were found to be infertile, so mice for experiments were generated by breeding heterozygotes.
Wild-type (WT) and Arv1 KO mice on a chow diet were weighed weekly to generate growth curves (Figures 2a and b). Arv1 KO mice were not significantly lighter than WT animals at the time of weaning, but failed to gain body weight in adulthood (>8 weeks of age), reaching an average maximum of around 17 g for females and 20 g for males (similar results were also observed on a western-type high-fat diet; Supplementary Figure 2). Arv1 KO mice showed substantial reductions in subcutaneous WAT and perigonadal WAT relative to WT mice at 12 weeks of age. In contrast, BAT, liver, heart, lungs, spleen, kidney and brain weights were not significantly different from WT animals (Supplementary Figure 3). To gain further insight into the difference in body weight, we examined 4–6-month-old male Arv1 KO mice by DEXA scan. Male Arv1 KO mice had body lengths identical to those of WT mice, whereas females were modestly shorter (Supplementary Figures 3j and 4j). Total bone mineral content was also not different (Figure 2c), indicating that alterations in the bone size or density were not major contributors to the difference in body weight. Arv1 KO mice showed significant reductions in lean mass (20%, P=0.001), as well as fat mass (44%, P=0.001; Figures 2d and e). The decrement in fat mass was greater than the difference in lean mass, and the animals consequently had a much lower percentage of total body fat (adiposity; Figure 2f). We also noticed that Arv1 KO mice appeared to be more susceptible to spontaneous death in adulthood (3–6 months of age) than WT. This effect was not completely penetrant- manifest in ~30% of the mice by 5 months of age (Supplementary Figure 5). The Arv1 KO mice also did not cope well with overnight fasting, which was required in some of the subsequent experiments (12–16 h), with a fraction of them (approximately one out of six mice) appearing cold to the touch, hypoglycemic and moribund the next morning.
WAT was noticeably depleted in Arv1 KO mice (Figure 3a, Supplementary Figure 6) with very little subcutaneous or perigonadal WAT obtained upon dissection. In contrast, livers and BAT pads appeared grossly normal (Figure 3b). At a cellular level, hematoxylin and eosin staining of paraffin sections revealed markedly different white adipose morphology. Arv1 KO adipose exhibited more intense cytoplasmic staining and dramatically smaller white adipocytes in all of the depots examined—namely inguinal subcutaneous, perigonadal, mesenteric and perirenal WAT (Figure 3c).
Arv1 KO mice had significantly lower plasma lipids. Separation of plasma lipoproteins by gel filtration chromatography revealed a difference in the HDL-C peak (Figure 4a, Supplementary Figure 7). Male Arv1 KO mice had 15% lower total cholesterol (P=0.003) and 17% lower HDL-C (P=0.0008), whereas non HDL-C and TGs were not significantly different between the groups (Figure 4b). Despite the major reductions in WAT mass, we did not observe an increase in the hepatic TG content (Figure 4c). No difference was observed in fasting glucose in female mice (Supplementary Figure 7), however, male Arv1 KO mice had lower blood glucose after a 4 h fast (Figure 4d), and plasma insulin levels were not significantly different. (Figure 4e). Consistent with decreased WAT mass, Arv1 KO mice had significantly reduced leptin (Figure 4f), but interestingly had increased plasma adiponectin relative to WT controls (Figure 4g). Male Arv1 KO mice had a dramatic improvement in glucose tolerance compared with WT animals (AUC reduced by 45%, Figure 4h).
The Arv1 KO mice were examined using metabolic cages to investigate the basis for the decreased fat mass and body weight. Arv1 KO mice consumed more food during the light cycle than WT mice (Figure 5a). No differences were observed in water intake (Figure 5b). Consistent with the increase in food consumption during the light hours, Arv1 KO mice were also more active during the light period than WT animals (Figure 5c). As expected based on the reduced WAT mass, we found that Arv1 KO animals did have increased oxygen consumption (VO2) and carbon dioxide production (VCO2), which reached statistical significance during the light hours (Figures 5d and e). The respiratory exchange ratio (RER) did not vary by genotype in the light phase, and the KO was only modestly lower than WT in the dark phase (WT: 0.905±0.025 versus KO 0.864±0.019, −4.5%, P<0.05), suggesting there was no major shift in carbohydrate versus fat utilization. Arv1 KO mice also expended more energy (calculated heat) than WT controls during the light cycle (Figure 5f). We hypothesized that the lack of TG stores in the WAT may be related to increased FA clearance and disposal. To test this, we measured fat tolerance in female Arv1 KO mice after an oral bolus of olive oil. Arv1 KO mice started out with a lower basal plasma TG level following an overnight fast, and had a nearly flat TG excursion after the olive oil gavage (Figure 6a). To test whether an increased rate of FA disposal could explain the difference in oral fat tolerance, we measured whole-body oxidation of a [3H]-oleic acid tracer. Consistent with the increase in energy expenditure, Arv1 KO mice also showed a robust 66% elevation in the rate of FA oxidation (Figure 6b).
Mice lacking Arv1 exhibit a dramatic lean phenotype with severely reduced WAT stores in all depots on a chow diet. We focused on male mice for the majority of studies, but it should also be noted that female mice lacking Arv1 also had similarly reduced body weight, fat mass, plasma lipids and higher adiponectin (Supplementary Figure 4). The lean phenotype included increased energy expenditure, improved oral fat tolerance and a faster rate of whole-body FA oxidation. Arv1 KO mice actually consumed more food (Figure 5a, Supplementary Figure 8) and were more active than WT mice (Figure 5c). Standard animal housing conditions are below thermoneutrality for mice, and in this setting spontaneous cage activity is generally not a major contributor to the differences in energy expenditure.17 However, further work is needed to determine the contributions of activity, nonshivering thermogenesis, and basal metabolic rate to the increased energy expenditure.
The global decrease in WAT in the Arv1 KO mice is best described as a lean phenotype, as it was not accompanied by other features of congenital lipodystrophy. The best examples for comparison are the Bscl2 KO,18, 19, 20 fatty liver dystrophy21 (Fld), Agpat2 KO22 and A-ZIP23 transgenic mice. In all the cases, cell autonomous defects in adipocyte differentiation, viability or function prevent effective storage of FA as TG, and impair adipokine secretion. Without functional adipose tissue to sequester excess FA, severe insulin resistance, fatty liver and splenomegaly ensue. In contrast, Arv1 KO mice have improved glucose tolerance, increased adiponectin, normal liver TG content and a complete lack of ectopic lipid deposition in other tissues. The Arv1 KO mice have comparatively more adipose than these lipodystrophic mice (−44% subcutaneous WAT mass, −50% perigonadal WAT relative to WT controls) as well as normal BAT mass (Figure 2e). This suggests that adipocytes and fat reserves in the Arv1 KO mice are smaller, but fundamentally healthy, and the global loss of fat mass is more likely owing to Arv1 loss in other tissues.
Mutations in ARV1 were initially identified as a consequence of inviability in the absence of sterol esterification in yeast. Beyond this context, loss of ARV1 in any organism does indeed compromise growth but is not lethal. ARV1 KO mice of both sexes appeared to be more susceptible to spontaneous death in adulthood (3–6 months of age) for reasons that are unclear. It is possible that ARV1 KO mice are less able to cope with fluctuations in food availability without the buffer of additional TG stores in WAT. The lack of WAT TG stores for adequate insulation and thermogenesis could be a factor in their increased mortality. Although improved glucose tolerance is generally considered beneficial, a major defect in glucose production or mobilization may be detrimental in this case. In accordance with observations in yeast where ARV1-deficient cells were mating deficient, we observed that Arv1 KO animals, particularly females, also had reproductive impairments. It remains to be determined whether these phenotypes reflect localized variation in sterol or FA esterification.
In earlier studies, we used antisense oligonucleotides (ASOs) to transiently knock down Arv1 in adult C57BL6/J male mice.24 ASOs primarily targeted the liver, although significant knockdown of Arv1 was also observed in adipose and intestine. ASO-mediated knockdown of Arv1 resulted in a doubling of plasma cholesterol, a decrease in liver TG content, an increase in liver:body weight ratio, higher bile acid concentrations in the plasma and elevated transaminase levels (ALT). These lipid changes were accompanied by activation of the FXR regulatory pathway in the liver. In contrast, mice with a germline deletion of Arv1 described in the current manuscript had reduced total and HDL-cholesterol (24% and 27% respectively, Figure 4b), unchanged liver:body weight ratios, and normal liver morphology and TG content. It is possible that the ASOs may have toxic and/or off-target25, 26, 27, 28 effects, or that a germline deletion may be distinct from the transient downregulation achieved with ASOs. The current data with whole-body deletion suggests an overwhelming and dominant role of ARV1 in extrahepatic tissues regulating energy metabolism; perhaps the brain, skeletal muscle or adipose—which merit further exploration.
A unifying model for these results may reflect a role for ARV1 in FA homeostasis. The ER is the site of FA ligation to CoA, as well as subsequent esterification in cholesteryl esters and TGs. ARV1 may be directly involved in lipid synthesis or storage, or may indirectly alter these processes through changes in ER lipid content. Indeed, in the yeast model of ARV1 deficiency, we observed a severe UFA-sensitivity; among 152 strains that were hypersensitive to 0.5 mm palmitoleate loading, ARV1-deficient yeast were the most liposensitive.9 Studying ARV1 in isolated mammalian cell culture systems derived from the Arv1 KO mice will shed light on these important questions. Further work is needed to define the physiological function of ARV1 in energy homeostasis by refining its site of action to specific tissues, such as the central nervous system, skeletal muscle and adipose. A better understanding of ARV1 and its downstream effectors may identify new targets for the treatment of obesity and insulin resistance in humans.
This work was supported by a National Scientist Development grant from the American Heart Association—11SDG7210077 (WRL), a mentor-based postdoctoral fellowship from the American Diabetes Association (ADA)—7-09-MN-41 (DJR), an undergraduate summer research fellowship from ADA (MS) and the Bristol Myers Squibb ‘Freedom to Discover’ Unrestricted Cardiovascular Research Grant (to DJR), the University of Pennsylvania Diabetes Research Center (DRC) Mouse Phenotyping, Physiology and Metabolism Core (P30-DK19525), the Penn Cardiovascular Institute Histology and Gene Expression Core and the Texas Medical Center Digestive Diseases Center ‘Core B—Cellular and Molecular Morphology’ (P30 DK56338)
About this article
Supplementary Information accompanies this paper on the Nutrition & Diabetes website (http://www.nature.com/nutd)