Magnesium deficiency prevents high-fat-diet-induced obesity in mice

This study was approved by the animal ethics board of the Radboud University Nijmegen (RU DEC 2015-0073) and the Dutch Central Commission for Animal Experiments (CCD, AVD103002015239). Forty-eight male C57BL6/J mice (Charles River Laboratories, Sulzfeld, Germany), aged 9–10 weeks, were randomly allocated to four experimental groups of n = 12 mice. Experimental diets consisted of 10% or 60% energy from palm oil plus 0.03% or 0.21% wt/wt magnesium oxide. Researchers and animal caretakers were blinded for Mg²⁺ content. On days −1, 84 and 112, mice were housed individually in metabolic cages for 24 h. Blood was collected via cheek puncture at days −1, 28, 56 and 84. At weeks 14 and 15, ITT and GTT, respectively, were performed. After 17 weeks on the diets, mice were anaesthetised by 4% vol./vol. isoflurane and exsanguinated via orbital sinus bleeding. See electronic supplementary material (ESM) for full methods.

After 14 weeks on the diets, mice underwent an intraperitoneal ITT. After 6 h of fasting, from 08:00 to 14:00, mice were injected with 0.75 U/kg body weight of human insulin (Novorapid, Novo Nordisk, Bagsværd, Denmark). Blood glucose levels were measured at 0, 20, 40, 60, 90 and 120 min. After 15 weeks on the diets, mice underwent an IPGTT. After an overnight fast, from 18:00 to 09:00, mice were injected with 2 g/kg body weight of d-glucose (Invitrogen, Bleiswijk, the Netherlands). Blood glucose was measured at 0, 15, 30, 60 and 120 min. See ESM for full methods.

Total RNA was extracted using TRIzol reagent (Invitrogen, Paisley, UK), subjected to DNase (Promega, Fitchburg, WI, USA) treatment and measured using the Nanodrop 2000c spectrophotometer (Thermo Scientific, Waltham, MA, USA). RNA was reverse transcribed using Moloney murine leukaemia virus (M-MLV) reverse transcriptase (Invitrogen, Bleiswijk, the Netherlands). Gene expression levels were quantified by SYBR-Green (Bio-Rad, Veenendaal, the Netherlands) on a CFX96 real-time PCR detection system (Bio-Rad) and normalised for Gapdh. Primer sequences for Acadl, Adrb3, Atgl (also known as Pnpla2), Cact, Cd36, Cpt1-l (also known as Cpt1a), Cpt1-m (also known as Cpt1b), Cpt2, Fbp1, G6pase (also known as G6pc), Gapdh, Glut1 (also known as Slc2a1), Glut2 (also known as Slc2a2), Glut4 (also known as Slc2a4), Gs, Gyk, Hmgcs1, Hsl (also known as Lipe), Mgll, Pepck1, Pklr, Ppar-α (also known as Ppara), Ppar-γ (also known as Pparg), Srebp1 (also known as Srebf1) and Ucp1 are provided in ESM Table 1.

Epididymal fat and liver tissues were fixed in 10% vol./vol. neutral-buffered formalin (KLINIPATH, Duiven, the Netherlands) in PBS. Samples were dehydrated through alcohol, embedded in paraffin and cut into 4 μm sections. Sections were stained with H&E using standard procedures. The average cell size of 100–300 cells per mouse was determined manually using ImageJ software (v1.48, NIH, Bethesda, MD, USA, RRID:SCR_003070). Liver samples were snap frozen in liquid nitrogen, cut into 10 μm sections, stained with Oil Red O (Sigma-Aldrich, St. Louis, MO, USA) and counterstained with haematoxylin.

Five randomly selected samples of each group were analysed, with no technical replicates, by RNA sequencing. Quality control and RNA sequencing were performed by the Beijing Genomics Institute (BGI), Hong Kong, China. Per sample, 13 million reads were sequenced using the Hiseq 4000 platform (Illumina, San Diego, CA, USA) using a 50 bp single-end module. Clean reads were mapped to Mus musculus transcriptome (GRCm38/mm10) using the HISAT/Bowtie2 tool (RRID:SCR_005476) [24, 25]. RSEM software v1.2.31 (RRID:SCR_013027) was used to quantify gene expression levels (fragments per kilobase million [FPKM] values) [26]. FPKM values were log₂ transformed and further analysed in R (www.​r-project.​org, v3.4.1., RRID:SCR_001905). Heatmaps for individual GO terms were created using the ggplot2 library (r-project) [27]. See ESM for full method details.

Serum Mg²⁺ was determined using a spectrophotometric assay at 600 nm (Roche/Hitachi, Tokyo, Japan) according to the manufacturer’s protocol. Liver samples were weighed and lysed in lysis buffer (10% wt/vol.) containing 50 mmol/l Tris-HCl pH 7.5, 1 mmol/l EGTA, 1 mmol/l EDTA, 1% vol./vol. Triton X-100, 10 mmol/l glycerophosphate, 1 mmol/l sodium orthovanadate, 50 mmol/l sodium fluoride, 10 mmol/l sodium pyrophosphate and 150 mmol/l sodium chloride. Triacylglycerol concentrations in serum and liver lysate were assayed using an enzymatic kit (Roche Molecular Biochemicals, Indianapolis, IN, USA), according to the manufacturer’s protocol. Serum NEFA (NEFA-C kit, WAKO Diagnostics, Delfzijl, the Netherlands), cholesterol (Human Diagnostics, Wiesbaden, Germany), glucose (Instruchemie, Delfzijl, the Netherlands), leptin (R&D Systems, Minneapolis, MN, USA) and adiponectin (R&D Systems, Minneapolis, MN, USA) concentrations were determined according to manufacturers’ protocols.

3-Methoxytyramine and normetanephrine were analysed by a 6490 LC-MS/MS (Agilent Technologies, Amstelveen, the Netherlands) after solid phase extraction (SPE) Oasis WCX μElution sample cleanup (Waters, Etten-Leur, the Netherlands). A calibration curve was used with 3-methoxytyramine-HCl and normetanephrine-HCl (Sigma-Aldrich, St. Louis, MO, USA) as calibrators. 3-Methoxytyramine-d₄-HCl and normetanephrine-d₃-HCl (Medical Isotopes, Pelham, NH, USA) were used as internal standards. An ethylene bridged hybrid (BEH) Amide 1.7 μm 100A, 2.1 × 100 mm column (Waters) was used as an analytical column.

All experimental procedures were conducted in compliance with the UK Animals Scientific Procedures Act (1986) and University of Oxford ethical guidelines. Thirty-nine male C57BL6/J mice (Medical Research Council [MRC], Harwell, UK) were randomly allocated to four groups of n = 10 mice (n = 9 in the low Mg²⁺[LowMg]-LFD group). At 8 weeks old, mice were put on experimental diets identical to the Radboud university medical center experiment for 9 weeks. At day 14, mice were housed individually in metabolic cages (Tecniplast, Buguggiate, Italy). Blood was collected via tail bleed at days −1 and 14. Respiration metabolic cages (TSE PhenoMaster Cages, Bad Homburg, Germany) were used at days 28 and 56 and body temperatures were measured by rectal probe (ATP-instrumentation, Ashby, UK). Data were averaged per hour and plotted from 18:30 to 09:30 h. See ESM Methods for full details.

Differentiated 3 T3-L1 cells (mycoplasma-free, ATCC, Manassas, VA, USA) were incubated for 20 h in DMEM, containing 0 or 1 mmol/l MgCl₂. Aliquots of 50 μl medium were taken hourly and heated for 8 min at 65°C. The concentration of NEFA was assessed using the WAKO NEFA-C kit (Instruchemie, Delfzijl, the Netherlands). See ESM Methods for full details.

For the animal experiments, a two-way ANOVA was used to look for a significant interaction effect between the two main variables (dietary fat and Mg²⁺ content). If there was none, significant differences between the groups were assessed using a two-way ANOVA approach with a Tukey’s multiple comparisons test. If the two-way ANOVA demonstrated a significant interaction effect between the two main variables, an unpaired multiple t test approach using the Holm–Sidak method for multiple comparisons was used. Statistical significance was determined using Graphpad Prism v7 (La Jolla, CA, USA, RRID: SCR_002798). For the lipolysis assays, an unpaired Student’s t test was used.

The mice were fed an LFD or HFD containing either a low (0.03% wt/wt) or normal (0.21% wt/wt) Mg²⁺ concentration for 17 weeks (Fig. 1a). There was no difference in body weight between low and normal Mg²⁺ groups on the LFD, but mice on the LowMg-HFD gained significantly less weight than those on the normal Mg²⁺(NormMg)-HFD (47.00 ± 1.53 g vs 38.62 ± 1.51 g in mice given a NormMg-HFD and LowMg-HFD, respectively, p < 0.05, Fig. 1a,b). The lower body weight of the LowMg-HFD group could not be explained by differences in dietary intake, as shown by similar food intake and faeces production between the two HFD groups (Fig. 1c,d). There was also no difference in water intake or urinary volume between the HFD groups (Fig. 1e,f). Hypomagnesaemia was detected in both the LowMg groups, but was significantly more pronounced in mice that were concomitantly fed an HFD (Fig. 1g).

To explore glucose metabolism in more detail, beta cell function and insulin sensitivity were determined by IPGTT and IPITT. In the IPGTT (a measure for beta cell dysfunction and insulin resistance), glucose clearance was reduced in both HFD groups (Fig. 2a). Glucose clearance was not significantly different between NormMg-HFD-fed mice and LowMg-HFD-fed mice (2.58 ± 0.08 vs 2.26 ± 0.13 mol/l × min in NormMg-HFD- and LowMg-HFD-fed mice, respectively, p = 0.07, Fig. 2c). LowMg-HFD-fed mice required significantly less insulin than NormMg-HFD-fed mice to clear the glucose, consistent with LowMg-HFD-fed mice being more insulin sensitive (Fig. 2b). Fasting blood glucose and insulin concentrations were significantly increased in the HFD-fed mice, in accordance with the increased body weight (Fig. 2d,e). Compared with the NormMg-HFD-fed mice, fasting blood glucose was lower in the LowMg-HFD-fed mice (Fig. 2d). The effect of dietary Mg²⁺ on fasting insulin was not statistically significant (Fig. 2e, two-way ANOVA for dietary Mg²⁺ effect, p = 0.07).

Both HFD-fed groups demonstrated increased insulin resistance in the ITT compared with their respective controls (Fig. 2f,g). Low dietary Mg²⁺ content resulted in a significantly lower AUC of the ITT (Fig. 2g, two-way ANOVA for dietary Mg²⁺ effect p < 0.05). In the LowMg-HFD group compared with the NormMg-HFD-fed mice, the AUC of the ITT was not significantly different (0.91 ± 0.05 vs 0.72 ± 0.05 mol/l × min in NormMg-HFD-fed and LowMg-HFD-fed mice, respectively, Tukey’s test p = 0.07, Fig. 2f,g).

Insulin resistance is often accompanied by hyperlipidaemia, in particular, high triacylglycerol and NEFA levels. As expected, the HFD-fed mice had higher serum triacylglycerol and NEFA levels than LFD-fed mice (Fig. 2h,i). Interestingly, despite their lower body weight and increased insulin sensitivity, LowMg-HFD-fed mice also had high serum triacylglycerol and NEFA (Fig. 2h,i). In contrast, serum leptin levels correlated with body weight; hence, reduced leptin levels were observed in the LowMg-HFD-fed mice compared with NormMg-HFD-fed mice (Fig. 2j). No difference between the two HFD groups was observed in serum adiponectin and cholesterol (ESM Fig. 1a,b).

Liver function is often impaired in type 2 diabetes as a consequence of insulin resistance and hepatic steatosis [28]. Feeding mice a NormMg-HFD resulted in a significantly heavier liver. However, this effect was abrogated in mice fed a LowMg-HFD (Fig. 3a). In Mg²⁺-deficient mice, the HFD did not increase liver triacylglycerol content (Fig. 3b). In line with the triacylglycerol measurements, H&E and Oil Red O staining showed reduced hepatic lipid accumulation in the Mg²⁺-deficient HFD-fed mice (Fig. 3c,d, respectively). Hepatic mRNA expression of Cd36, a long-chain fatty acid transporter, was reduced in the LowMg-HFD-fed mice compared with the NormMg-HFD-fed mice (ESM Fig. 2a). Low dietary Mg²⁺ increased hepatic mRNA expression of sterol regulatory element-binding protein 1c (Srebp1c), which is involved in cholesterol and fatty acid metabolism, and phosphoenolpyruvate carboxykinase 1 (Pepck1), essential for gluconeogenesis, glyceroneogenesis and fatty acid re-esterification (ESM Fig. 2b,c, two-way ANOVA for dietary Mg²⁺ effect p < 0.05). No differences were observed between the two HFD-fed groups in the expression of key genes involved in hepatic glycolysis, ketogenesis and β oxidation (ESM Fig. 2d–l).

Our results show that mice fed a LowMg-HFD diet exhibit reduced body weight and high triacylglycerol levels compared with their NormMg-HFD-fed littermates. Interestingly, the LowMg-HFD group had decreased mass of epididymal and inguinal white adipose tissue (eWAT and iWAT, respectively) (Fig. 4a,b), which may point towards defective lipid handling in white adipose tissue (WAT). The HFD increased adipocyte size (Fig. 4c,d), but no significant difference was observed between the two HFD groups (Fig. 4c,d). Nevertheless, quantitative PCR showed that mRNA expression of Srebp1c, Pepck1 and genes involved in β oxidation was increased in the eWAT of the LowMg-HFD group compared with the NormMg-HFD group (ESM Fig. 3a–f).

To determine the consequences of low Mg²⁺ on lipid metabolism, we performed RNA sequencing on eWAT. A principal component analysis using the log₂ transformed expression values shows that the samples from both LFD groups cluster closely together, indicating the absence of a strong Mg²⁺ effect, whereas there is a clear separation of NormMg-HFD vs LowMg-HFD gene expression profiles (Fig. 4e). To investigate the effect of Mg²⁺ on specific biological pathways, the fold changes for groups of genes belonging to the same gene ontology (GO) were analysed. GO term analysis indicated that processes associated with adiposity (e.g. inflammation) were downregulated in LowMg-HFD-fed vs NormMg-HFD-fed mice, in accordance with decreased adipose tissue mass (ESM Table 2). Interestingly, despite the increased insulin sensitivity of the LowMg-HFD-fed mice, several key genes involved in the triacylglycerol catabolism pathway (lipolysis) were upregulated in the LowMg-HFD vs the NormMg-HFD group, which may explain the reduced lipid storage as well as the high serum NEFA levels (Fig. 4f). A modest increase in acyl-CoA dehydrogenase dependent β oxidation was observed in the LowMg-HFD-fed mice vs the NormMg-HFD-fed mice (Fig. 4g). The metabolic effects of Mg²⁺ in eWAT appear to be specific to lipid homeostasis, as there was no clear effect on glycolysis (ESM Fig. 3g). Although serum cholesterol levels were not different between the experimental groups, cholesterol biosynthesis was greatly reduced in the LowMg-HFD-fed vs the NormMg-HFD-fed mice (ESM Fig. 3h, ESM Table 2).

To investigate whether hypomagnesaemia has a direct effect on lipolysis in eWAT, we examined the effect of Mg²⁺ on lipolysis in differentiated 3 T3-L1 cells in vitro. Unstimulated lipolysis was not different at 0 or 1 mmol/l Mg²⁺, indicating that Mg²⁺ deficiency does not directly induce lipolysis in adipocytes (ESM Fig. 4a).

β₃-Adrenergic receptors (ADRB3) are essential regulators of lipid metabolism, increasing brown adipose tissue (BAT) activity and reducing WAT lipid storage via activation of lipolysis [29‐32]. We therefore explored whether enhanced β-adrenergic signalling could explain the high triacylglycerol levels, increased lipolysis and reduced body weight of Mg²⁺-deficient HFD-fed mice. Expression of Adrb3 was significantly increased by 2.5-fold in the eWAT of LowMg-HFD-fed compared with NormMg-HFD-fed mice (Fig. 5a). Additionally, both HFD-fed groups showed elevated mRNA expression of Adrb3 in BAT, but this upregulation was more pronounced in the LowMg-HFD group (Fig. 5b). To determine whether this was the result of enhanced adrenaline (epinephrine) release, serum levels of the dopamine metabolite 3-methoxytyramine and the adrenaline metabolite normetanephrine were measured. However, no significant increase was observed (Fig. 5c and ESM Fig. 5a). mRNA expression of the lipolysis genes adipose triacylglycerol lipase (Atgl), hormone-sensitive lipase (Hsl) and monoglyceride lipase (Mgll) was significantly increased in eWAT of Mg²⁺-deficient HFD-fed mice compared with the NormMg-HFD group (Fig. 5d–f).

Expression of Ucp1 in BAT, which is essential for non-shivering thermogenesis, was upregulated in NormMg-HFD-fed mice compared with NormMg-LFD-fed mice (Fig. 5g). In line with increased β₃-adrenergic signalling, Ucp1 expression was further increased in BAT of LowMg-HFD-fed mice. BAT thermogenesis is strongly regulated by fatty acid availability [33]. Indeed, genes involved in NEFA metabolism of BAT are upregulated (Fig. 5h–j) (Atgl, Cpt1-m and Acadl). In contrast, mRNA levels of glucose transporters 1 and 4 (Glut1/4) in BAT were unchanged in LowMg-HFD-fed compared with NormMg-HFD-fed mice (ESM Fig. 5b,c). mRNA expression of the fatty acid transporter Cd36 and of the important metabolic transcription factors peroxisome proliferator-activated receptor alpha (Pparα) and gamma (Pparγ) remained unchanged in BAT between the NormMg-HFD-fed and LowMg-HFD-fed mice (ESM Fig. 5d–f).

To investigate the energy metabolism in Mg²⁺-deficient HFD-fed mice, the dietary intervention study was repeated with respiratory cages. Respiration, body temperature and activity were measured at week 8, which was when the weight differences developed in our first experiment. In line with the previous experiment, no differences were observed in food and water intake between the two HFD-fed groups (ESM Fig. 6a–d); and the Mg²⁺-deficient HFD-fed mice had reduced body weight compared with the NormMg-HFD-fed mice (Fig. 6a). Lean body mass was not different between the two HFD groups, indicating that the weight difference depends on adipose tissue mass (Fig. 6b). As with our previous experiment, the HFD caused a reduction in serum Mg²⁺ levels (Fig. 6c). A significant increase was observed in serum triacylglycerol when the animals were killed (after 9 weeks) in the LowMg-HFD group compared with the NormMg-HFD group (Fig. 6d). Low dietary Mg²⁺-fed mice had decreased non-fasted serum glucose (Fig. 6e; two-way ANOVA for dietary Mg²⁺ effect p < 0.05), while the difference between the two HFD-fed groups was not significant (NormMg-HFD vs LowMg-HFD, Tukey’s test p = 0.07). Hypomagnesaemia and HFD decreased core body temperature (Fig. 6f). In contrast, the body temperature of LowMg-HFD-fed mice was higher than NormMg-HFD-fed mice (Fig. 6f; 35.8 ± 0.1 vs 36.4 ± 0.2°C in NormMg-HFD and LowMg-HFD, p < 0.05), in line with increased BAT activity. Moreover, mice fed a Mg²⁺-deficient diet showed increased locomotor activity (ESM Fig. 6e,f; two-way ANOVA for dietary Mg²⁺ effect p < 0.05). However, energy expenditure was not different between the two HFD-fed groups (Fig. 6g,h). In both HFD groups, the respiratory exchange ratio (RER) was approximately 0.7, indicating that fatty acids are the main energy source (Fig. 6i,j). Interestingly, in the LFD groups, hypomagnesaemia resulted in a reduction of the RER (Fig. 6i,j), suggesting an increased use of fatty acids as an energy substrate over glucose. However, this reduction was not statistically significant (Fig. 6j; 0.82 ± 0.01 vs 0.79 ± 0.01 average RER in NormMg-LFD and LowMg-LFD, respectively; p = 0.10).