Loss of angiopoietin-like 4 (ANGPTL4) in mice with diet-induced obesity uncouples visceral obesity from glucose intolerance partly via the gut microbiota

Animal studies were performed using purebred WT and Angptl4^−/− mice on a C57Bl/6 background that were bred and maintained in the same facility for more than 20 generations [11]. Angptl4^−/− mice were generated via homologous recombination of embryonic stem cells, and lack part of the Angptl4 gene, resulting in a non-functional ANGPTL4 protein [11]. The Angptl4^−/− mice were imported to our animal facility in 2006 as strain B6.129P2-Lp139tm1 N10 from Taconic (Germantown, NY, USA; a kind gift of A. Köster, Eli Lilly, Indianapolis, IN, USA). Mice were individually housed in temperature- and humidity-controlled specific pathogen-free conditions. Mice had ad libitum access to food and water.

In study 1, two groups of mice were studied; male WT and Angptl4^−/− mice at 10–13 weeks of age. These two groups were part of a larger study (the control groups of this study), which aimed to investigate the effects of different dietary fibres on non-alcoholic fatty liver disease in WT and Angptl4^−/− mice [18]. The assignment to the different fibre groups was randomised using an online randomisation tool. Liver fat levels, liver histology and hepatic expression of genes related to inflammation (e.g. F4/80 [also known as Adgre1], Cd68, Mcp-1 [Cxcl2], Il1b, Il1ra [Il1rn], Tnfa [Tnf], Itgax, Timp1) were all analysed (all were significantly higher in Angptl4^−/− mice compared with WT mice; data not included). A power calculation was performed based on plasma alanine aminotransferase (ALT) activity as a biomarker for liver health. Assuming a plasma ALT activity of 40 U/l in WT mice with a σ value of 16, using a power of 0.8, a significance level of 0.05, and an effect size of 20 U/l, the sample size was calculated as n = 11 mice per group. To allow compensation for potential loss of mice during the study, n = 12 mice were included per group. Mice were fed a high-fat/high-cholesterol/high-fructose diet, providing 45% energy as triacylglycerols (Formula D12451, Research Diets; manufactured by Research Diet Services, Wijk bij Duurstede, the Netherlands; sterilised with γ-irradiation at 9 kGy) for 18 weeks [18‐20]. The fat source of the diet was replaced by safflower oil and supplemented with 1% cholesterol (wt/wt) (Dishman, Veenendaal, the Netherlands). Fructose was administered by adding 20% fructose (wt/vol.) to the drinking water. After 17 weeks of dietary intervention, all mice underwent a glucose tolerance test, as explained below.

In study 2, male WT and Angptl4^−/− mice at 19–22 weeks of age received the same diet as in study 1 (Formula 58V8 a.k.a. D12541; manufactured by TestDiet, Saint Louis, MO, USA; sterilised with γ-irradiation at 18–50 kGy) for 18 weeks. The fat source of the diet was replaced by safflower oil and supplemented with 1% cholesterol (wt/wt). Fructose was administered by adding 20% fructose (wt/vol.) to the drinking water. A mixture of broad-spectrum antibiotics was provided in the drinking water (1 g/l ampicillin, 1 g/l neomycin sulfate and 0.5 g/l metronidazole). This antibiotic cocktail was previously shown to effectively suppress intestinal bacteria [21, 22]. Control mice received normal water. The assignment of the antibiotics/normal water was randomised using an online randomisation tool. After 15 weeks of dietary intervention, all mice underwent an intestinal permeability assay (see below for further details). After 16 weeks, all mice underwent an insulin tolerance test, and after 17 weeks a glucose tolerance test, as explained below. A power calculation was performed based on the AUC of the glucose tolerance test; assuming an AUC of 1800 mmol/l × min in WT mice with a σ value of 150, using a power of 0.8, a significance level of 0.05, and an effect size of 200 mmol/l × min, the sample size was calculated as n = 9 mice per group. To allow compensation for potential loss of mice during the study, n = 10 mice were included per group. PCR indicated that two mice in the WT group treated with antibiotics were in fact Angptl4^−/− mice. These two mice were included in the group of Angptl4^−/− mice on antibiotics.

Body weight and food intake were assessed weekly in both studies. After 18 weeks, mice were anaesthetised using isoflurane and blood was collected by orbital puncture. Mice were euthanised via cervical dislocation, after which tissues were excised and weighed and intestinal content was sampled. Samples were immediately frozen in liquid nitrogen and stored at −80°C. All animal experiments were approved by the local animal ethics committee of Wageningen University, the Netherlands. All analyses were conducted with the experimenter blinded to group assignment.

In both study 1 and 2, a glucose tolerance test was performed 1 week before mice were euthanised. The test was carried out as previously described [14]. Briefly, after 5 h of fasting, mice were injected i.p. with glucose (0.8 g/kg body weight) and blood glucose and insulin levels were measured at various time points during the test (see electronic supplementary materials (ESM) Methods for further details).

In study 2, an insulin tolerance test was performed 2 weeks before mice were euthanised. The test was carried out as previously described [14]. Briefly, blood samples were collected immediately before and at selected time points after i.p. insulin injection (0.75 U/kg body weight) (see ESM Methods for further details).

In study 2, an intestinal permeability assay was performed 3 weeks before euthanasia. FITC-dextran (4 kD) (Sigma, Houten, the Netherlands) was administered by oral gavage (600 mg/kg body weight, 40 mg/ml). After 1 h, blood was drawn, stored on ice in the dark and centrifuged for 15 min at 1000 g. The plasma obtained was diluted in PBS and fluorescence intensity was measured using a fluorescence spectrophotometer (Fluoroskan Ascent; Thermo Fisher Scientific, Breda, the Netherlands; excitation wavelength (λ_ex), 485 nm; emission wavelength (λ_em), 538 nm). FITC-dextran concentrations were determined using a standard curve generated by diluting FITC-dextran in PBS. A baseline blood sample was used to correct for autofluorescence.

Serum amyloid A was measured via ELISA (Life technologies, Bleiswijk, the Netherlands), plasma triacylglycerol by a commercially available colourimetric kit (HUMAN Diagnostics, Wiesbaden, Germany) and endotoxin levels via the Limulus Amebocyte Lysate assay (Lonza, Walkerville, MD, USA).

Mouse beta-TC6 cells (CRL11506; ATCC via LGC [Wesel, Germany]) were grown in 24-well plates in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 15% heat-inactivated FCS and 1% penicillin/streptomycin (Lonza, Verviers, Belgium) under 5% CO₂ at 37°C. Cells tested negative for mycoplasma.

Total RNA was extracted from mesenteric adipose tissue, beta-TC6 cells and mouse pancreatic islets using TRIzol reagent (Life technologies) and the RNeasy minikit (Qiagen, Venlo, The Netherlands). RNA was reverse-transcribed using the iScript cDNA synthesis kit (Bio-Rad Laboratories, Veenendaal, the Netherlands). PCR was carried out as previously described [7] for the following genes: 36b4 (also known as Rplp0), Cd68, F4/80, Mcp-1, Il6, Il1ra, Lpl and Angptl4. Further details can be found in ESM Methods.

Pancreatic islets were isolated from WT and Angptl4^−/− mice by injecting Liberase TL (Sigma) into the bile duct. Additional details can be found in ESM Methods.

Groups of 8 islets isolated from WT and Angptl4^−/− mice were fasted for 1 h, followed by stimulation with low (2 mmol/l) and high (20 mmol/l) glucose. Islets were lysed, and secreted insulin and insulin content was quantified by ELISA (Crystal Chem, Zaandam, the Netherlands) (see ESM Methods).

Short-chain fatty acid (SCFA) levels in the content of the caecum were measured using an Avance III NMR spectrometer (Bruker, Leiderdorp, the Netherlands), as previously described [18].

Paraffin-embedded pancreas sections (5 μm) were immunohistochemically stained using anti-insulin antibody (Insulin (H-86): sc-9168; Santa Cruz Biotechnology, Dallas, TX, USA; 1:800), followed by a biotinylated goat anti-rabbit antibody (BA-1000; Vector Labs, Brunschwig Chemie, Amsterdam, the Netherlands; 1:200), and an avidin-biotin-complex (ABC) coupled to peroxidase (Vector Labs). Visualisation was carried out using 3,3′-diaminobenzidine for 5 min. For negative controls, primary antibodies were omitted.

LPL activity in mesenteric white adipose tissue was quantified using a [³H]oleic acid-labelled triolein substrate, as previously described [9].

Western blotting was performed to detect LPL. Mesenteric fat pads were lysed in ice-cold Pierce IP lysis buffer (Thermo Fisher Scientific) containing protease and phosphatase inhibitors (Roche, Almere, the Netherlands). Lysates were centrifuged to remove fat droplets before protein separation and immunoblotting with goat anti-mouse LPL or rabbit anti-goat HSP90 (Cell Signaling, Danvers, MA, USA), both at 1:2000. Further processing of samples was conducted as outlined previously [7] and described in ESM Methods.

Faecal samples were resuspended in buffer and cells were lysed. DNA was subsequently extracted using either phenol:chloroform:isoamyl alcohol [25:24:1] or the Maxwell 16 System (Promega, Madison, WI, USA) (see ESM Methods).

For 16S ribosomal RNA (rRNA) gene sequencing, DNA samples were sent to the Broad Institute of MIT and Harvard (Cambridge, MA, USA) and processed as previously described [18]. For statistical significance, biological relevance and visualisation, the linear discriminant analysis (LDA) effect size (LEfSe) method was used. The sequencing data are available from the authors upon request. Further details are provided in ESM Methods.

Real-time PCR was performed on faecal DNA samples from study 2 to investigate 16S rRNA gene expression, using amplified and purified 16S rRNA to generate a standard curve (see ESM Methods for details).

Data are presented as mean ± SEM, unless otherwise indicated. Statistical analyses were performed using an unpaired Student’s t test or two-way ANOVA followed by a Bonferroni test for post-hoc analysis (GraphPad Software, La Jolla, CA, USA). A p value <0.05 was considered statistically significant.

Strikingly, despite the higher visceral fat mass and inflammation, blood glucose levels during the glucose tolerance test were markedly lower in Angptl4^−/− mice than in WT mice (Fig. 2a). The lower glucose levels in Angptl4^−/− mice were accompanied by elevated fasting plasma insulin levels (Fig. 2b). These data suggest that Angptl4^−/− mice are more glucose tolerant than WT mice, possibly owing to higher plasma insulin levels.

To explore the potential origin of the elevated plasma insulin levels in Angptl4^−/− mice, we investigated whether ANGPTL4 might be expressed in pancreatic islets, together with LPL, thereby influencing insulin secretion. LPL has been proposed as a mediator of the effect of dietary lipids on insulin secretion [23]. However, although Lpl expression was relatively high in primary pancreatic islets and in the beta cell line beta-TC6, with C_t values of ~21–22, Angptl4 expression was very low, with C_t values of ~30.5–32 (Fig. 2c, d). Hence, the effect of ANGPTL4 on glucose tolerance is unlikely to be mediated by ANGPTL4 produced by beta cells. BioGPS supports the minimal expression of Angptl4 in the mouse pancreas (www.​biogps.​org/​, accessed 11 May 2017). Furthermore, analysis of several microarray datasets (www.​ncbi.​nlm.​nih.​gov/​gds) of mouse pancreatic islets (GSE43620), the beta cell line beta-TC3 (GSE31102) and beta cells from adult mice (GSE54374) confirmed that Angptl4 is expressed at very low levels in pancreatic islets and beta cells. Consistent with the minimal expression of Angptl4 in pancreatic islets, glucose-stimulated insulin secretion was not significantly altered in isolated pancreatic islets of Angptl4^−/− mice compared with WT mice (Fig. 2e). In addition, insulin staining of the pancreas showed no difference in the abundance of pancreatic islets or insulin staining intensity in WT vs Angptl4^−/− mice (Fig. 2f).

Notably, the above data do not exclude a role for circulating ANGPTL4 in the regulation of islet LPL activity, and in insulin secretion. However, upregulation of LPL activity in beta cells has been shown to lead to hyperglycaemia during a glucose tolerance test [24]. Accordingly, it is unlikely that the observed hypoglycaemia and hyperinsulinaemia in Angptl4^−/− mice are linked to loss of inhibition of LPL activity in islet cells, whether via ANGPTL4 that is produced in the pancreas or elsewhere. Together, these findings suggest that ANGPTL4 probably does not influence plasma insulin levels via a direct effect on insulin secretion in pancreatic islets.

Considering that the gut microbiota may influence insulin secretion [25, 26], and since ANGPTL4 has been linked to the gut microbiota [27, 28], we hypothesised a potential role of the gut microbiota in the effect of loss of ANGPTL4 on insulin levels. To explore the possibility that Angptl4^−/− mice have a different gut microbiota to WT mice, in study 1 we first measured the levels of several gut-derived metabolites, including luminal succinate and SCFAs, and plasma lipopolysaccharide (LPS). In the caecum, the concentration of butyrate was significantly lower in Angptl4^−/− mice than in WT mice (Fig. 3a). Moreover, in the colon, levels of both propionate and butyrate were significantly lower, and levels of succinate were significantly higher in Angptl4^−/− mice (Fig. 3b). Also, plasma LPS levels were higher in Angptl4^−/− mice than in WT mice, even though intestinal permeability measured using FITC-dextran was similar (Fig. 3c, d). Together, these findings suggest a potential difference in gut microbiota composition between Angptl4^−/− and WT mice.

To further examine the gut microbiota in WT and Angptl4^−/− mice in study 1, we performed 16S rRNA sequencing. Principal coordinate analysis of unweighted UniFrac distance demonstrated differential clustering of microbiome sequences in Angptl4^−/− and WT mice (Fig. 3e). The majority of the difference in bacterial community between Angptl4^−/− and WT mice can be explained by the phyla Actinobacteria and Firmicutes. Indeed, only small non-significant differences were observed within the phyla Bacteroidetes, Deferribacteres and Proteobacteria. The phylum Actinobacteria was 2.5-fold more abundant in Angptl4^−/− mice, which was mainly accounted for by a significant increase in the genus Adlercreutzia. While the relative abundance of the phylum Firmicutes was comparable between the genotypes, the genera Lactobacillus and SMB53 were over-represented in Angptl4^−/− mice. In contrast, consistent with the lower colonic butyrate levels, the butyrate-producing Allobaculum was less abundant in Angptl4^−/− mice (Fig. 3f–h, ESM Table 1), whereas the abundance of other butyrate-producing bacteria was not different (members of Lachnospiraceae) or was increased (members of Clostridiaceae family). Overall, these data indicate that the gut bacterial composition is significantly different between Angptl4^−/− and WT mice, which may explain the different levels of gut-derived metabolites.

To further study the impact of the gut microbiota on the effect of Angptl4 ablation on glucose tolerance, we repeated the diet-induced obesity study in WT and Angptl4^−/− mice with or without antibiotics in their drinking water (study 2). Antibiotic supplementation significantly reduced the number of 16S rRNA gene copies in the faeces (Fig. 4a) and markedly increased caecum weight (Fig. 4b), indicating an effective suppression of gut bacteria.

Overall body weight gain was higher in Angptl4^−/− mice than WT mice, which reached significance in the absence of antibiotics (Fig. 4c). In agreement with the first study, in the absence of antibiotics, Angptl4^−/− mice had elevated mesenteric fat-pad weight. In contrast, after antibiotic treatment, the mesenteric fat mass was not significantly different between Angptl4^−/− mice and WT mice (Fig. 4d). Interestingly, in WT mice, antibiotic treatment increased Angptl4 mRNA levels in the ileum but not in mesenteric fat tissue, suggesting intestinal Angptl4 expression is suppressed by the gut bacteria (Fig. 4e, f).

Concurrent with the elevated fat mass, in the absence of antibiotics the expression of several inflammatory genes, including Cd68, Mcp1, Il6 and Il1ra, was elevated in mesenteric fat of Angptl4^−/− mice compared with WT mice. However, in the presence of antibiotics the difference in adipose tissue inflammation between the genotypes was largely abolished (Fig. 4g–j), suggesting that the increased fat mass and associated elevation of adipose tissue inflammation in Angptl4^−/− mice are partly dependent on the gut bacteria.

Consistent with data presented in Fig. 2a, in study 2 blood glucose levels during the glucose tolerance test were significantly lower in Angptl4^−/− mice than in WT mice (Fig. 5a). Suppression of the gut bacteria using antibiotics substantially reduced the differences in blood glucose levels between the two sets of mice, suggesting a role for the gut microbiota in the effect of ANGPTL4 on glucose tolerance (Fig. 5b, c). The lower glucose levels in Angptl4^−/− mice were accompanied by higher plasma insulin levels (Fig. 5d). Insulin tolerance was not different between Angptl4^−/− and WT mice (Fig. 5e, f), irrespective of antibiotic treatment (Fig. 5g, h). Overall, our findings suggest that loss of ANGPTL4 promotes (visceral) obesity yet, by raising insulin levels, reduces glucose intolerance, and that this effect is partly dependent on the gut bacteria.