Increased NEFA levels reduce blood Mg²⁺ in hypertriacylglycerolaemic states via direct binding of NEFA to Mg²⁺

Three hundred and two individuals, aged 55–80 years, were enrolled in the 300-Obesity cohort study (an observational cross-sectional cohort study within the Human Functional Genetics Projects) at the Radboud university medical center between 2014 and 2016 [28]. The study was carried out in accordance with the Declaration of Helsinki and all participants provided informed consent. All participants had a BMI above 27 kg/m². Individuals with a recent cardiovascular event (myocardial infarction, transient ischaemic attack, stroke <6 months ago), a history of bariatric surgery or bowel resection, inflammatory bowel disease, renal dysfunction or increased bleeding tendency, or who used oral or subcutaneous anticoagulant therapy or thrombocyte aggregation inhibitors (other than acetylsalicylic acid and carbasalate calcium) were excluded. Blood samples were taken in the morning, following an overnight fast. Blood glucose, triacylglycerols, total cholesterol and HDL-cholesterol were measured using standard laboratory procedures (Cobas C8000; Roche Diagnostics, Risch-Rotkreuz, Switzerland). The HOMA-IR was calculated using the standard formula, as published by Matthews et al [29]. High-throughput nuclear magnetic resonance (NMR) metabolomics platform (Nightingale’s Biomarker Analysis Platform, Nightingale Health, Helsinki, Finland) [27] was used for the quantification of 231 lipid and metabolite measures. The metabolites were measured in a single experiment, set-up for the quantification of different metabolite groups. In this article we focus on lipoproteins: total lipid concentrations of 14 lipoprotein subclasses, lipoprotein particles sizes, apolipoproteins and cholesterol. The NMR metabolomics platform used in this study has previously been used in various epidemiological studies [30, 31]. Details of the NMR-based metabolomics experimentation have been described previously [27]. In the present study, serum Mg²⁺ levels were measured in all individuals. However, 17 measurements had substantial duplo errors (large deviations [≥0.05 mmol/l from the average] between duplicate measurements) and, therefore, only 285 individuals were included in the analysis.

All procedures for this experiment were conducted in compliance with the UK Animals Scientific Procedures Act (1986) and University of Oxford ethical guidelines. Kir6.2-p.Val59Met mice were generated in our laboratories, as previously published [32]. All mice were 12–16 weeks old. Mice were housed with a maximum of five animals per cage (ventilated), separated by gender and with ad libitum access to food and standard pellet chow, in a 12 h light/dark cycle, at 21°C. In four male and four female C57BL6 mice, expression of a Kir6.2-p.Val59Met transgene was induced using subcutaneous injection of 400 μl tamoxifen (0.02 g/ml corn oil). This inducible mouse model recapitulates the phenotype of neonatal diabetes and these mice develop diabetes due to impaired insulin secretion [32, 33].

Three days after tamoxifen injection, mice were fasted overnight from 17:00 to 09:00 hours. Successful induction of the Kir6.2-p.Val59Met transgene was validated by measuring fasted blood glucose levels using a StatStrip Xpress glucose meter (Nova Biomedical, Waltham, MA, USA). One female mouse did not have elevated fasting glucose levels and was excluded from subsequent analyses. The remaining mice then underwent oral gavage and blood sampling using an identical experimental set up to that described above for C57BL6/J wild-type mice.

BSA and fatty-acid-free BSA (FF-BSA) (both Sigma Aldrich) were separately dissolved at several concentrations in a 1 mmol/l MgCl₂ solution (Merck Millipore, Darmstadt, Germany) or a physiological buffer, both set at pH 7.5 by adding NaOH. The physiological buffer contained the following (all from Merck Millipore unless stated otherwise): 27 mmol/l NaHCO₃, 112 mmol/l NaCl, 5 mmol/l KCl, 1 mmol/l MgCl₂, 1 mmol/l Na₂HPO₄ (VWR International, Radnor, PA, USA) and 2.5 mmol/l CaCl₂ dissolved in Milli-Q (Radboud Institute for Molecular Life Sciences, Nijmegen, the Netherlands).

To extract endogenous NEFAs, BSA (0.2 g/ml) dissolved in Milli-Q was mixed (1:2 vol./vol.) with ice-cold ethanol–diethyl ether (3:1 vol./vol.). The lipid phase was evaporated overnight at room temperature. To remove trace amounts of BSA, the solution was centrifuged three times with an Amicon 50 kDa filter (Merck Millipore) for 15 min at 2500 g to clog the protein in the filter. Extracted NEFAs were added to 250 μl FBS (Biowest, Nuaillé, France), 1 mmol/l MgCl₂, or 0.5 mmol/l BSA in the amounts of 0, 25, 50, 100, 200, 350, 500 and 700 μl. Dilution factors were accounted for when measuring the concentrations of Mg²⁺ and NEFA.

Protein (Pierce, Thermo Scientific, Waltham, MA, USA), NEFA (WAKO Diagnostics, Delfzijl, the Netherlands), triacylglycerols (Roche Molecular Biochemicals, Indianapolis, IN, USA) and Mg²⁺ (Roche/Hitachi, Tokyo, Japan) concentrations were measured using a spectrophotometric assay according to the manufacturer’s protocols. The Mg²⁺ colourimetric assay was based on a Xylidyl Blue-I method and the absorbance was measured at 600 nm. NEFAs were measured at 546 nm, triacylglycerols at 500 nm and protein at 562 nm on a Bio-Rad Benchmark plus microplate spectrophotometer (Bio-Rad laboratories, Hercules, CA, USA). For inductively coupled plasma mass spectrometry (ICP-MS) analyses, serum samples were dissolved in HNO₃ (>65%, Sigma) and diluted prior to being subjected to ICP-MS (X-series; Thermo Scientific).

As part of a study (ClinicalTrials.​gov registration no. NCT01967459), which aimed to examine the effect of vitamin D supplementation on postprandial leucocyte activation, 24 female volunteers underwent an oral fat-loading test. The study was approved by the Institutional Review Board of the Franciscus Gasthuis & Vlietland Rotterdam and the regional independent medical ethics committee of the Maasstad Hospital Rotterdam [34]. All participants provided informed consent. Samples from the baseline oral fat load were used in this study to measure serum Mg²⁺, NEFA and triacylglycerol concentrations.

Inclusion criteria were an age above 18 years, a premenopausal status, BMI ≥25 kg/m² and vitamin D deficiency. Exclusion criteria were the use of any kind of medication (except for oral contraceptives), smoking, pregnancy, participation in a clinical study less than 6 months before inclusion and the use of vitamin supplements.

Participants visited the hospital after a 10 h overnight fast and venous blood samples were obtained after an oral fat load (before Vitamin D treatment). Each participant received an oral fat load using fresh cream (Albert Heijn, Zaandam, the Netherlands) at a dose of 50 g of fat per m² body surface calculated by the Mosteller formula [35]. During the oral fat loading, test participants were not allowed to eat or to drink (except water) and they were asked to refrain from physical activity. Venous blood sampling was repeated at 2 h intervals until 8 h after oral fat loading, and serum Mg²⁺ and NEFA levels and plasma triacylglycerol levels were measured (as detailed in the ‘Analytical measurements’ section above). Two individuals were excluded due to insufficient sample availability.

Results are presented as mean ± SEM, unless stated otherwise. Variables of overweight individuals were correlated univariately to serum Mg²⁺ levels using Pearson’s correlation analyses (SPSS for Windows v22.0.0.1, RRID:SCR_002865; IBM, New York, NY, USA). Based on the initial experiment in wild-type mice, the sample size for the experiment in Kir6.2-p.Val59Met mice was calculated using a one-way ANOVA statistic (with Dunnett’s correction for multiple comparison); to detect an effect size of 0.3 (SD 0.13) with a power of 80% and α level of 5%: a total of six mice were required per group. The sample size of the oral lipid (cream) load study in healthy individuals was assessed using a one-way ANOVA (with Dunnett’s correction for multiple comparison); to detect an effect size of 0.1 (SD 0.1) with a power of 80% and an α level of 5%: a total of 23 individuals were required. Significance of Mg²⁺, triacylglycerol and NEFA concentrations at 1, 2, 4, 6 and 8 h compared with those at 0 h was evaluated using a one-way ANOVA with Dunnett’s correction for multiple comparisons. Direct correlations between Mg²⁺ and triacylglycerol or NEFA concentrations were assessed using linear regression analyses. A p value of ≤0.05 was considered statistically significant. All statistical analyses were performed using Graphpad Prism v7 (RRID:SCR_002798; Graphpad software, La Jolla, CA, USA).

Factors affecting blood Mg²⁺ levels were evaluated in 285 overweight individuals (BMI >27 kg/m²) from the 300-Obesity cohort (ESM Table 1). The mean serum Mg²⁺ concentration in this cohort was 0.89 ± 0.09 (SD) mmol/l, with only 2% of the individuals having hypomagnesaemia (serum Mg²⁺ <0.7 mmol/l, see ESM Table 1 and ESM Fig. 1). Despite the fairly healthy serum Mg²⁺ levels in these individuals, serum Mg²⁺ concentrations were inversely correlated with triacylglycerol levels, predominantly those in VLDL particles (Table 1). The serum Mg²⁺ concentration was also inversely correlated with HOMA-IR. As HOMA-IR was strongly correlated with plasma triacylglycerol concentration (ESM Tables 2, 3), we questioned whether insulin resistance modulated the inverse correlation between serum Mg²⁺ and plasma triacylglycerol levels and triacylglycerols in VLDL particles. However, HOMA-IR did not influence these correlations in multivariable regression analyses (ESM Tables 4 and 5).

To further investigate the relationship between lipoproteins and serum Mg²⁺ level, the composition of lipoprotein particles was investigated and the correlation of these particles with serum Mg²⁺ concentration was analysed (Table 1). The concentration of the larger VLDL particles showed the strongest inverse correlations with serum Mg²⁺ levels (Table 1). Interestingly, the concentration of smaller HDL particles was positively correlated with serum Mg²⁺. There was no correlation between serum Mg²⁺ levels and the concentration of any of the intermediate-density lipoprotein and LDL particles (Table 1). No significant correlation was observed between serum Mg²⁺ and the diameter of the VLDL, LDL or HDL particles (Table 1).

To unravel the underlying mechanism that explains how blood triacylglycerols are associated with blood Mg²⁺ concentrations, mice were subjected to an oral gavage of olive oil following an overnight fast. Plasma triacylglycerols and NEFA levels both peaked at 4 h post-gavage (Fig. 1a, b, Δ plasma Mg²⁺ −0.31 mmol/l at 4 h post-gavage). Interestingly, the plasma Mg²⁺ concentration showed a reciprocal decrease, reaching a nadir at 4 h post-gavage (Fig. 1a, b). At basal levels (0 h), no significant correlation was observed between plasma Mg²⁺ levels and NEFA concentrations (Fig. 1c, p = 0.27). However, when plasma NEFA levels increased (at 4 and 6 h), there was a clear inverse correlation between plasma Mg²⁺ and NEFA concentrations (Fig. 1d, e, p ≤ 0.05). When plasma NEFA levels decreased and reached a concentration approaching levels at baseline (8 h), there was no longer a significant correlation between plasma NEFA and Mg²⁺ concentrations (Fig. 1f, p = 0.54). Similar correlations were observed between plasma Mg²⁺ and triacylglycerol levels (ESM Fig. 2a–d).

Increased blood lipids can enhance glucose-induced insulin secretion [36]. Insulin induces a compartmental shift of Mg²⁺, decreasing the blood concentration of Mg²⁺ and increasing intracellular levels [37]. Hence, to rule out insulin-dependent effects, the same oral gavage experiment was performed in inducible Kir6.2-p.Val59Met mice, which develop hypoinsulinaemia and hyperglycaemia [38]. In these mice, a similar reduction in plasma Mg²⁺ was observed in response to the oral gavage of olive oil as compared with wild-type mice (Fig. 1g). However, while the reduction in the plasma Mg²⁺ concentration was numerically similar to that observed in wild-type mice, it did not reach statistical significance in Kir6.2-p.Val59Met mice due to a high variation between the animals (which was substantially larger than in the initial experiment using wild-type mice). However, a significant inverse correlation between plasma Mg²⁺ and triacylglycerols was still observed in these hypoinsulinaemic mice (Fig. 1h, p ≤ 0.05).

As ~30% of Mg²⁺ is bound to albumin, which is the predominant carrier of NEFAs, we investigated whether the binding of Mg²⁺ to albumin is dependent on NEFAs [14]. With increasing concentrations of BSA in MgCl₂, we found that the Mg²⁺ concentration declined as BSA concentrations increased (Fig. 2a). Mg²⁺ concentrations decreased from 0.85 ± 0.02 mmol/l at 0 mmol/l BSA to 0.64 ± 0.01 mmol/l at a near-physiological concentration of BSA (0.5 mmol/l or 33.25 g/l), a 25% reduction (Fig. 2a). Interestingly, the absence of fatty acid (i.e. the use of FF-BSA instead of BSA in the MgCl₂ solution) abrogated this effect, in line with the theory that binding of Mg²⁺ to albumin is NEFA-dependent (Fig. 2b). Linear regression analyses showed a significant inverse correlation between Mg²⁺ and BSA concentration (Fig. 2c); the regression constant was approximately four times stronger for BSA than for FF-BSA (Fig. 2c). To exclude the possibility that other cations present in blood compete with the binding of Mg²⁺ to BSA, BSA was also dissolved in a physiological buffer, which mimicked the concentration of other abundant blood electrolytes. This approach resulted in a similar decrease in the Mg²⁺ concentration with increasing BSA concentrations (Fig. 2d). Again, FF-BSA had no significant effect on Mg²⁺ levels (Fig. 2e). In the physiological buffer, BSA and FF-BSA displayed correlations similar to those observed in the MgCl₂ buffer (Fig. 2f).

We then set out to modify the binding of Mg²⁺ to albumin by increasing the concentration of NEFA. Elevating the concentration of NEFAs in a BSA–MgCl₂ solution directly reduced the Mg²⁺ concentration (linear regression: y = −0.12x + 0.83, p ≤ 0.05; Fig. 3a). To resemble the in vivo setting, the experiment was repeated using FBS instead of BSA; increasing NEFA levels in FBS reduced Mg²⁺ concentrations to a similar extent as when BSA was used (linear regression: y = −0.10x + 1.36, p ≤ 0.05; Fig. 3b). Interestingly, the NEFA-induced reduction in Mg²⁺ concentration was protein-independent, as increasing NEFA levels in a MgCl₂ solution also lowered Mg²⁺ concentration (linear regression: y = −0.08x + 1.00, p ≤ 0.05; Fig. 3c).

To discover whether triacylglycerols and NEFAs also affect blood Mg²⁺ concentration in humans, we examined samples obtained from 24 healthy female individuals, with a BMI >25 kg/m², who received an oral fat load [34]. Serum NEFA and Mg²⁺ levels and plasma triacylglycerol concentrations were measured over a period of 8 h. Plasma triacylglycerol levels increased significantly, from 1.18 ± 0.10 mmol/l at 0 h to a peak of 2.13 ± 0.18 mmol/l at 4 h (Fig. 4a). Serum NEFA concentrations also significantly increased, from 0.38 ± 0.03 mmol/l at 0 h to a peak of 0.76 ± 0.04 mmol/l at 6 h (Fig. 4b). In accordance with our previous findings, serum Mg²⁺ levels dropped from 0.82 ± 0.01 mmol/l at 0 h to a nadir of 0.75 ± 0.02 mmol/l at 6 h (Fig. 4a, b, Δ plasma Mg²⁺ −0.07 mmol/l at 6 h post-intake). The total serum Mg²⁺ concentration, as measured by ICP-MS, was not affected by fat (cream) intake; hence, these findings suggest that only the free Mg²⁺ levels were affected by lipid loading (Fig. 4c).