Introduction
The current pandemic of metabolic diseases, such as obesity and type 2 diabetes, cannot be completely explained by genetic alterations and the growing consumption of a Western diet [
1,
2]. Moreover, obesity is not an inevitable consequence of a fat-rich diet, since both people and mice consuming a high-fat diet (HFD) can display the opposite metabolic outcome, suggesting the existence of metabolic adaptations in some individuals [
3,
4]. Among the factors that may affect the metabolic processes on an individual basis [
5] are the gut microbiota [
6], the impact of which on host metabolism has been established [
7‐
9]. We previously found that adaptation to obesity in terms of insulin sensitivity was characterised by a specific gut microbiota profile in insulin-resistant vs insulin-sensitive obese individuals [
10]. In addition, we showed that divergent gut microbiota profiles characterise the different metabolic phenotypes developed during metabolic adaptation to an HFD in mice [
3,
4]. Recently, we reported that the periodontal microbiota profile correlates with cardio-metabolic adaptations to an HFD in mice [
11].
Furthermore, along with xenobiotics [
12], diet is considered to be the strongest modulator of gut microbiota [
13]. Evidence from hepatic transcript profiles in mice has suggested that liver pathophysiology may be affected during metabolic adaptation to HFD in mice [
14]. The existence of a gut–liver axis has been previously demonstrated and the liver is the organ in which xenobiotic metabolism occurs, especially with regard to our capacity of responding to gut microbial antigens [
15].
Moreover, the alteration of gut microbiota, termed dysbiosis, is an additional causal factor in the development of hepatic steatosis [
16], a condition that involves the accumulation of hepatic triacylglycerols, which is a common feature of metabolic disease [
17]. Indeed, the different stages of hepatic diseases, including steatosis, hepatitis and hepatocellular carcinoma (HCC), are identifiable by a precise microRNA (miRNA) signature [
18]. miRNA are pleiotropic modulators of gene expression [
19] that have been shown to be under the control of gut microbiota [
20]. Some miRNA, for example miR-181a, miR-666 and miR-21, are specifically involved in the modulation of liver pathophysiology [
18,
21].
In the present study, we aimed to elucidate the gut microbiota profiles that are associated with metabolic adaptations to HFD in mice. We also aimed to investigate the associations between specific taxa of gut microbiota and hepatic expression of miR-181a, miR-666 and miR-21 in mouse models of hepatic steatosis. In addition, we explored the link between miRNA expression levels and metabolic parameters, such as glucose tolerance, body weight and fasting blood glucose.
Methods
Animal models and dietary treatment
All animal experimental procedures were approved by the local ethical committee of Rangueil University Hospital (Toulouse, France). All experimenters were blind to group assignment and outcome assessment. No data, samples or animals were excluded from this study.
GTT and hepatic triacylglycerol measurement
After 3 months of HFD, an IPGTT or OGTT were performed. Briefly, for the IPGTT, 6 h fasted mice were injected with glucose (1 g/kg) into the peritoneal cavity, as previously described [
22]. An OGTT was performed via oral administration of glucose (1.5 mg/g) following a 6 h fast. Blood glucose levels were measured 30 min before glucose administration and at 0, 15, 30, 60, 90 and 120 min following glucose challenge.
For both IPGTT and OGTT, the glycaemic index was calculated as the sum of the blood glucose values (mmol/l) divided by the total time of the curve in min to present value in mmol/l × min, or additionally multiplied by 1000 to give value in μmol/l × min.
Liver triacylglycerol content was measured by a colorimetric assay using free glycerol and triacylglycerol reagents (Sigma Aldrich, St Louis, MO, USA) and the plate was read using the Multiskan Spectrum plate reader and the SkanIt RE software (both Thermo Labsystems, Beverly, MA, USA).
Taxonomic analysis of gut microbiota by pyrosequencing
Caecum total DNA was extracted as previously described [
4]. The whole 16S bacterial ribosomal RNA V2 region was targeted by the 28F-519R primers (designed by Research and Testing Laboratory [
www.researchandtesting.com/, accessed 1 September 2016; Lubbock, TX, USA]) and pyrosequenced by the 454 GS FLX+ system (Roche, Branford, CT, USA) at the Research and Testing Laboratory. On average, 3000 sequences were generated per mouse. The minimum number of sequences guaranteed per mouse was 1606 (
n = 62).
Preparation of murine primary hepatocytes and lipopolysaccharide stimulation
Hepatocytes were isolated by a non-recirculating collagenase perfusion through the portal vein of anesthetised 8-week-old C57BL/6 WT or
Cd14KO male mice fed a normal chow diet. Isolated cells were filtered through a 100 μm pore mesh nylon filter and cultured (2.5 × 10
6 cells per well) onto 96-well plates in DMEM (BE12-614F; Lonza, Levallois, France) supplemented with 10% (vol./vol.) FCS, 1% (vol./vol.) penicillin/streptomycin and 0.2 nmol/l
l-glutamine. After 12 h, the medium was replaced with medium plus industrially purified lipopolysaccharide (LPS; Sigma Aldrich, St Louis, MO, USA) either from the proinflammatory
Escherichia coli serotype O55:B5 [
22], or the
E. coli strain O111:B4, which stimulates human hepatocytes [
23]. Two doses of LPS were tested: 10 ng/ml (low dose) and 100 ng/ml (high dose), and cells were stimulated for 6 h. Experiments were performed in quadruplicates (control) or pooled duplicates (LPS).
miRNA-based quantitative PCR
Real-time PCR for miRNA expression was performed on total miRNA extracted from cells or livers using the miRNeasy kit (Qiagen, Courtaboeuf, France). The expression of each miRNA was normalised to U6 small nuclear RNA (snRNA) expression [
23]. For in vitro analyses, cells were directly harvested into Qiazol solution, which was provided with the miRNeasy kit. For ex vivo analyses, frozen pieces of liver were put directly into the Qiazol and total miRNA was extracted following the manufacturer’s protocol. Expression values were quantified using the
\( {2}^{-\Delta \Delta {\mathrm{C}}_{\mathrm{t}}} \) method [
4].
Hepatic microarray analysis
Total RNA was isolated from the right lobe of the liver using TRIzol (Life Technologies, Villebon sur Yvette, France) according to the manufacturer’s protocol. Preparation, labelling and hybridisations of cDNA were performed as per the manufacturer’s protocol. Samples were analysed using an Agilent SurePrint G3 Mouse GE 8 × 60K chip (design 028005; Agilent Technologies, Courtaboeuf, France). The hybridised microarrays were washed and scanned using an Agilent G2505C scanner. Data were extracted from the scanned image using the Agilent Feature Extraction software version 10.10.1.1. All of these steps were performed at the GENOTOUL GeT-TRIX facility at the French National Institute for Agricultural Research (INRA; Toulouse, France).
Statistical analysis
Statistical analyses were performed by two-way ANOVA followed by the Dunnett’s post hoc test, or using the unpaired Student’s
t test, using GraphPad Prism version 7.00 for Windows 7 (GraphPad, San Diego, CA, USA). A
p value <0.05 was considered significant. Cluster analysis was performed using PermutMatrixEN software (
http://download.cnet.com/PermutMatrix/3000-20432_4-75325452.html, accessed 20 June 2016) [
24]. Multivariate analyses were performed using the Spearman correlation coefficient and
p values were adjusted using the Benjamini–Hochberg correction (available at
www.marum.de/Binaries/Binary745/BenjaminiHochberg.xlsx, accessed 12 November 2016). A String analysis was performed to study the network of genes targeted by miR-21; these genes were identified using the software miRTarBase (
http://mirtarbase.mbc.nctu.edu.tw/, accessed 20 June 2016) [
25].
Discussion
In this study we report that, beyond the diversity already observed for blood glucose and body weight [
3,
4,
11], liver triacylglycerol content is characterised by a high heterogeneity according to the individual response of mice to a diabetogenic/non-obesogenic HFD. Furthermore, liver triacylglycerol content was positively associated with the relative abundance of Firmicutes, and negatively associated with hepatic miR-21 expression and the relative abundance of Proteobacteria and
B. acidifaciens.
Our finding that hepatic steatosis follows metabolic diversity on an individual basis, in mice fed a diabetogenic/non-obesogenic HFD, confirms our previous observations where hepatic lipid metabolism in this model, evaluated at the level of gene expression, was shown to be modulated according to the response to an HFD [
14]. Therefore, our results reinforce the reproducibility of this animal model.
The observed positive association between hepatic triacylglycerol content and the relative abundance of Firmicutes is in contrast with the results of Henao-Mejia et al [
26]. The authors found a reduction of Firmicutes in a model for inflammasome-mediated dysbiosis, regulating the progression of NAFLD and obesity [
31]. However, the murine model used in our study is very different from the one used by Henao-Mejia and colleagues; we used C57BL/6 WT mice, whereas these authors used C56BL/6
Nlrp6 KO mice. Therefore, it is likely that the different genetic backgrounds and genotypes of these mouse models account for disparity in the manifestations of dysbiosis, and are also responsible for the observed colitogenic phenotype. Differences in murine models may also explain the discrepancies found in the literature with regard to
B. acidifaciens. This bacterium has recently been shown to be associated with liver disease [
32], in contrast to our findings. Again, the murine model used in our study is very different from the mouse model used by Xie et al (streptozotocin/HFD-induced non-alcoholic steatohepatitis [NASH]/HCC C57BL/6 J mice) [
32]. Thus, it is likely that under two very different dietary conditions,
B. acidifaciens may have different associations with hepatic pathophysiology. This explanation is corroborated by another recent study by Yang et al that showed that
B. acidifaciens prevents obesity and improves insulin sensitivity in mice [
33]. Hence, based on the adaptation of both mice and microbes to a fatty environment, a divergent metabolic phenotype may arise, as we also recently reported with regard to cardio-metabolic adaptation to HFD in mice [
11].
From a molecular perspective, miRNA represent promising molecules that may link gut microbiota dysbiosis to metabolic outcomes. Gut microbiota can modulate intestinal miRNA expression [
20]. Moreover, a specific miRNA profile defines every stage of hepatic pathophysiology during the progression of disease from NAFLD (characterised by accumulated hepatic triacylglycerols) to NASH, (characterised by inflammation and fibrosis), up to HCC [
18]. Thus, taking into account the gut–liver axis with regard to our capacity to sense gut microbes [
15] and the fact that gut microbiota dysbiosis can drive NAFLD [
25], it is plausible to consider that, in our study microbes, or even their antigens (e.g. LPS), may affect the liver via the modulation of hepatic miRNA expression, as we have shown here in vitro.
Specifically, miR-21 regulates both regeneration [
34] and progression of fibrosis [
35] in the liver, and long-term (18 weeks) inhibition of miR-21 reduces both body weight and adipocyte size in aged
db/db mice [
36]. These data are in accordance with the definition of miR-21 as a ‘disease miRNA’ [
21]. However, this definition appears to be dependent on both animal model and diet. In fact, in the absence of HFD feeding, hepatic miR-21 expression is not correlated with hepatic triacylglycerols (Fig.
4;
r
2
= 0.08,
p = 0.1 [correlation not shown]). Notably, there is no clear evidence in the literature as to whether miR-21 may act as a marker of hepatic fat deposition. On one hand, Wu et al recently showed that miR-21 knockdown impairs lipid accumulation [
37], which is in contrast to our findings. In contrast, Ahn et al reported that lycopene inhibits hepatic steatosis via the upregulation of miR-21, which is in accordance with our findings [
38].
However, during metabolic adaptation to HFD, hepatic miR-21 expression was observed to be negatively associated with liver triacylglycerol content (Fig.
5f). This evidence suggests that during the adaptation to HFD, hepatic miRNA expression follows a different pattern of regulation. We confirmed this interpretation by showing that among the genes targeted by miR-21 are
Casp8 and
Fadd, which are also part of the microbial RIG-I-like receptor signalling pathway. As previously mentioned, this pathway regulates the production of proinflammatory cytokines (such as TNF-α and IL-8) in response to microbial antigens (i.e. LPS). Importantly, we recently showed that the microbial RIG-I-like receptor signalling pathway is one of the most upregulated pathways during cardio-metabolic adaptation to HFD in mice [
11], corroborating its implication in this phenomenon.
In conclusion, we propose a new triad linking gut microbiota, hepatic miRNA expression and liver triacylglycerol content. These findings may help to explain hepatic metabolism adaptation to an HFD in mice.