The GLP-1R agonist liraglutide limits hepatic lipotoxicity and inflammatory response in mice fed a methionine-choline deficient diet

Nonalcoholic fatty liver disease (NAFLD) is the most common hepatic disorder related to type 2 diabetes (T2D). The disease can evolve towards nonalcoholic steatohepatitis (NASH), a state of hepatic inflammation and fibrosis. There is presently no drug that effectively improves and/or prevents NAFLD/NASH/fibrosis. GLP-1 receptor agonists (GLP-1Ra) are effective in treating T2D. As with the endogenous gut incretins, GLP-1Ra potentiate glucose-induced insulin secretion. In addition, GLP-1Ra limit food intake and weight gain, additional beneficial properties in the context of obesity/insulin-resistance. Nevertheless, these pleiotropic effects of GLP-1Ra complicate the elucidation of their direct action on the liver. In the present study, we used the classical methionine-choline deficient (MCD) dietary model to investigate the potential direct hepatic actions of the GLP-1Ra liraglutide. A four-week infusion of liraglutide (570μg/kg/day) did not impact body weight, fat accretion or glycemic control in MCD-diet fed mice, confirming the suitability of this model for avoiding confounding factors. Liraglutide treatment did not [...] SOMM, Emmanuel, et al. The GLP-1R agonist liraglutide limits hepatic lipotoxicity and inflammatory response in mice fed a methionine-choline deficient diet. Translational Research


Brief Commentary
Background: Nonalcoholic fatty liver disease (NAFLD) and steatohepatitis (NASH) are common hepatic disorders related to type 2 diabetes (T2D). Presently, no drug effectively improves and/or prevents NAFLD/NASH/liver fibrosis. GLP-1 receptor agonists (GLP-1Ra) are effective in treating T2D but their direct action on the liver remains elusive.
Translational significance: We presently show that infusion of the GLP-1Ra liraglutide prevents accumulation of C16 and C24-ceramides/ sphingomyelins species, inflammation and initiation of fibrosis in liver from mice on a methionine-choline deficient diet, suggesting beneficial hepatic actions independent of weight loss and global hepatic steatosis. GLP-1Ra may present added-value in hepatic complications of T2D.

INTRODUCTION
Nonalcoholic fatty liver disease (NAFLD) is the most common hepatic disorder in industrialized countries. 1,2 Obesity and type 2 diabetes (T2D) are frequent causes of NAFLD. 1,2 Nevertheless, NAFLD can occur in lean patients due to various factors, including overconsumption of specific nutrients (such as fructose or cholesterol), accumulation of visceral fat, genetic predisposition, dyslipidemia, or lipodystrophy. 1,2 A significant proportion of patients with NAFLD develop a state of hepatic inflammation [nonalcoholic steatohepatitis, NASH], which can lead to fibrosis and cirrhosis, potentially resulting in hepatocellular carcinomaD 2 5 X X. 3 Whilst lifestyle changes, including a modification in eating habits, weight loss or physical activity, have beneficial effects on liver steatosis, there is currently no efficient long-term treatment available to improve and/or avoid the development and progression of NAFLD, NASH, and fibrosis. [1][2][3] Studying NAFLD is challenging in humans due to the invasive nature of liver biopsies and the slow progression rate of the disease. This situation has led to the development of many animal models, 4 of which the methionine and choline deficient (MCD) diet is most frequently used. 4 The MCD diet is rich in sucrose and devoid of methionine and choline, D 2 6 X X2 nutrients required for mitochondrial b-oxidation and very-low density lipoprotein synthesis in the liver. 5,6 As a consequence, the MCD diet induces hepatic steatosis, inflammation, and fibrosis without weight gain. 4 In addition, MCD diet-fed mice present liver-specific, but not global, insulin-resistance. 7 Introduced about 15 years ago, GLP-1Ra are considered effective drugs for the treatment of T2D, with the advantage of limiting hypoglycemia. As does endogenous GLP-1 secreted by the enteroendocrine L-cells in the intestine, GLP-1Ra primarily potentiate glucose-induced insulin secretion. 8 In addition, GLP-1Ra protect pancreatic islets, reduce food intake, and stimulate energy expenditure, 8 all additional beneficial outcomes in the context of obesity/insulin-resistance. Interestingly, in vitro studies suggest that GLP-1Ra act directly on hepatocytes to limit endoplasmic reticulumD 2 7 X X stress, 9 activate AMP-activated protein kinaseD 2 8 X X, 10-12 dampen lipogenesis 10 or interact with the insulin signaling pathways. [11][12][13] Nevertheless, the pleiotropic effects of GLP-1Ra on the pancreas and hypothalamus complicate the elucidation of their direct action on the liver in vivo. In fact, the antisteatotic action of GLP-1Ra has been mainly observed in situations of weight loss and/or insulin sensitization in both animal models [14][15][16][17][18][19][20] and humans. 21,22 To investigate the liver-specific action of GLP-1Ra, we infused liraglutide to mice fed with the MCD diet, hypothesizing that liraglutide might improve NAFLD/ NASH by reducing steatosis, inflammation and fibrosis independently of weight loss.

MATERIAL AND METHODS
Animals, diets, treatments. All experimental protocols were performed in accordance with the Swiss animal welfare laws. Fourteen-week-old C57BL/6J male mice (Charles River Laboratories) were acclimatized for D 2 9 X X1 week in standard conditions (22˚C, 12h-light/12h-dark cycle, free access to water and food). Thereafter, mice were fed with a regular chow diet (RM3, SDSD 3 0 X X; Chow group) or a D 3 1 X XMCDD 3 2 X X diet (A02082002BR, Research DietsD 3 3 X X; MCD and MCD/Lira groups) for 7 weeks. During the last D 3 4 X X4 weeks of the experiment, mice were either infused with isotonic saline solution (Chow and MCD groups) or with liraglutide (0.1 ml/hD 3 5 X X of Victoza, Novo Nordisk [6 mg liraglutide/ml]D 3 6 X X; MCD/Lira group) via micro-osmotic pumps (Model 1004D, Alzet). Food intake and body weight were monitored weekly during the experiment. A schematic representation of the study protocol is illustrated in Fig 1. At the end of the experiment, mice were fasted 4 hours, anesthetized with isoflurane and immediately sacrificed. The osmotic pumps were retrieved and emptying was checked to confirm liraglutide release. Blood samples were collected, and organs were dissected and weighed before fixation or cryopreservation in liquid nitrogen.
Blood analyses. Blood samples were collected in EDTA-coated tubes and plasma was stored at ÀD 3 7 X X80˚C. Plasma levels of glucose, ALT, AST, total cholesterol, triglycerides and free fatty acids were assessed using a Cobas C111 robot and appropriate reagents (Roche Diagnostics). Plasma levels of insulin were measured using an ultrasensitive mouse insulin ELISA kit (Mercodia).
Gene expression. Total RNA was isolated from 50D 3 8 X X to 100 mg liver samples using TRI Reagent Solution (ThermoFisher). Five hundred ng of total RNA were reverse transcribed using the PrimeScript RT reagent kit (Takara) and gene expression levels were assessed by quantitative PCR using the Power SYBR Green master mix (Thermofisher) and a Light-Cycler 480 (Roche Diagnostics). Normalization was done using the 40S ribosomal protein S29 (RPS29) gene.
Histology, morphometry, pathological score. Liver and the distal part of the intestine (ileum) were either fixed overnight in 10% formalin before dehydration and embedded in paraffin or immediately embedded in OCT medium (Cell path LTD) and frozen on dry ice before storage at ÀD 3 9 X X80˚C. Paraffin-embedded sections were stained with hematoxylin-eosin (H&E), Sirius red, Periodic Acid Schiff'sD 4 0 X X solutions and Ki67 antibody using classical procedures. Frozen sections were stained with Oil red O solution using classical procedures. Pictures were acquired using a VS120 microscope (Olympus). Steatosis and fibrosis quantifications were achieved with Oil red O and Sirius red D 4 1 X Xlabeled slides using the area measurement tool of the ImageJ software (NIH). For fibrosis quantification, vessels, exhibiting a massive confounding collagen staining, were excluded from the analysis which represents only collagen fibers emerging in the liver parenchyma. Two independent liver lobes per animal were used for these quantifications. Ileal villi and crypt sizes were measured on H&E stained sections with the strait segment tool of ImageJ software. The pathological score, including steatosis and activity (ballooning and lobular inflammation) was evaluated by a trained pathologist (CDV) blinded to the diets as previously described. 23 Lipidomic analysis. Lipid extraction and lipidomic analyses were performed as previously described. 24  Lipids were isolated from 20 mg of liver tissue. After extraction, the organic lipid containing phase was dried in a vacuum concentrator (CentriVap, Labconco). Lipids were then dissolved in chloroform/methanol and divided into D 4 2 X X3 aliquots. One aliquot was treated by alkaline hydrolysis to enrich for sphingolipids 26 and the other D 4 3 X X2 aliquots were used for glycerophospholipids and phosphorus assay, respectively. Mass spectrometry analysis for the identification and quantification of phospho-and sphingolipid species was performed on a TSQ Vantage Triple Stage Quadrupole Mass Spectrometer (Thermo Fisher Scientific) equipped with a robotic nanoflow ion source (Nanomate HD, Advion Biosciences), using multiple reaction monitoringD 4 4 X X. Each individual ion dissociation pathway was optimized with regard to collision energy. Lipid concentrations were calculated relative to the relevant internal standards and normalized to the phosphate content of each lipid extract to account for variability in the amount of starting material and to correct for sample loss during the extraction procedure.
Microbiota analysis. DNA extraction. DNA was extracted from about 30 mg colon content using Quick-DNA Fecal/Soil Microbe Miniprep Kit (Zymo). Purified DNA was quantified using the Qubit dsDNA BR Assay Kit (Thermo Fisher Scientific) and stored at À20˚C. Two negative controls were performed by extracting DNA using the same extraction procedure but omitting the addition of colon content.
Amplicon Sequencing. The V3À4 region of the bacterial 16S rRNA genes (Escherichia coli positions 341À805) was amplified using 1 ng of DNA extracted from colon content samples or 5 mL of the DNA extract obtained from negative controls in a 25 D 4 5 X XmL volume of ZymoBIOMICS PCR PreMix (Zymo Research) containing each of 0.4 mM forward primer 5'-barcode-CCTACGGGNGGCWGCAG-3' and reverse primer 5'barcode-GACTACHVGGGTATCTAATCC-3' tagged with 8-nt barcodes at their 5'ends (Fasteris, Plan-les-Ouates, Switzerland). The PCRs were carried out with an initial denaturation at 95˚C for 3 minutes, followed by 30 cycles of denaturation at 95˚C for 30 seconds, annealing at 51˚C for 30 seconds, and extension at 72˚C for 60 seconds, and a final extension at 72˚C for 10 minutes. Duplicate PCRs of each sample were combined and run (1 mL) on a 2100 Bioanalyzer (Agilent Technologies) for quality analysis and quantification. The sequencing library was constructed using the MetaFast protocol (Fasteris, Plan-les-Ouates, Switzerland). The sequencing was carried out for 2 £ 300 cycles on an Illumina MiSeq instrument using MiSeq Reagent Kit v3.
Bioinformatics analysis. Libraries were demultiplexed using the Fasteris internal software. No mismatches were allowed in a barcode and a maximum of 2 mismatches were allowed per 16S primer sequences. Trimmomatic v.0.32 27 was used to remove Illumina adapters and to trim reads, by cutting them once the average quality within the 4-base window fell below a quality score of 15. Reads with a final length <60 bases were discarded. Forward and reverse reads were further quality-filtered and paired-end-joined with PEAR v.0.9.11 28 using the following settings: maximum assembly length (Àm) 470; minimum assembly length (ÀD 4 6 X Xn) 390; minimum overlap (ÀD 4 7 X Xv) 20; minimum read size after trimming (ÀD 4 8 X Xt) 240; P value (ÀD 4 9 X XP) 0.0001; maximal proportion of uncalled bases (ÀD 5 0 X Xu) 0; and quality score threshold in trimming (ÀD 5 1 X Xq) 33. The merged sequences were subjected to the UPARSE pipeline (USEARCH v.8.1.1861) 29 for clustering into operational taxonomic units (OTUs) at 97% identity. Taxonomic assignment of the representative OTUs was performed using MOTHUR v.1.39.5 30 classify.seqs command (with method = wang, ÀD 5 2 X Xcutoff = 80) and the EzBioCloud 16S rRNA gene sequence database 31 downloaded on 4 January 2018. We removed OTUs that (D 5 3 X X1) had an average relative abundance higher in the negative controls than in colon content samples or (D 5 4 X X2) had <90% identity to reference prokaryotic EzBio-Cloud 16S sequences as revealed by USEARCH (-usearch_global -id 0.9 -query_cov 0.99). 32 Normalization of OTU counts for sequencing depth (110,000) was performed using the "rrarefy" command from the R vegan package. Sequencing data were submitted to the European Nucleotide Archive (ENA; www.ebi.ac. uk/enaD 5 5 X X; study number: PRJEB36797).

Statistical analyses Q3
X X. Statistical analyses of the data using D 5 6 X X1-way ANOVA was performed using the Graph-Pad Prism version 7.02 software. Graphs represent mean + standard error of the mean (SEM). Clustered heat map was generated in R using the heatmap.2 function of the "gplots" package. A D 5 7 X XPvalue < 0.05 was considered statistically significant.

RESULTS
Liraglutide does not change energy homeostasis or carbohydrate metabolism in mice on an MCD dietD 5 8 X X. First, we investigated the metabolic action of liraglutide in the nutritional context of the MCD-diet. Previous studies have shown that GLP-1Ra, like liraglutide, reduce food intake in rodents. 33 Here, we confirmed that mice fed an MCD-diet eat approximately 30% less than mice fed a chow diet (Fig 2A). Nevertheless, liraglutide did not potentiate the anorexic effect of the MCD-diet. In fact, MCD diet-fed mice infused with liraglutide ate the same amount of food as those infused with saline solution (Fig 2A). Similarly, body weight ( Fig 2B), epididymal white adipose tissue mass (Fig 2C), interscapular brown adipose tissue mass (Fig 2D), glycemia (Fig 2E), and insulinemia ( Fig 2F) were all significantly decreased by the MCD-diet consumption but unaffected by liraglutide administration. The homeostasis model assessmentD 5 9 X X index, considered as a good marker of whole body insulin sensitivity, was greatly improved in the MCD-diet group but unchanged by liraglutide (Chow: 28.9 § 1.8, MCD: 2.5 § 0.5, MCD/ Lira: 4.3 § 1.4, D 6 0 X XP= 0.22 for MCD/Lira vsD 6 1 X X MCD groups). Taken together, these results showed that liraglutide infusion did not impact energy homeostasis or basal glycemic control in mice on the MCD-diet, allowing us to investigate directly the hepatic action of liraglutide independently of potentially confounding metabolic effects.
Liraglutide changes hepatic lipid composition qualitatively but not quantitativelyD 6 2 X X. Relative liver weight was not statistically different between Chow, MCD and MCD/Lira groups ( Fig 3A). Circulating levels of alanine aminotransferase (Fig 3B), aspartate aminotransferase ( Fig 3C) and free fatty acids (Fig 3D), increased significantly by the MCD-diet, were not further modified by liraglutide treatment. Similarly, hypocholesterolemia ( Fig 3E) and hypotriglyceridemia ( Fig  3F) were comparable in MCD diet-fed mice infused with saline solution or liraglutide. As expected, histological hepatic sections stained with H&E and Oil-red O revealed that livers of MCD diet-fed mice displayed significant steatosis and lobular inflammation, while their hepatocytes presented only mild ballooning (Fig 3G, J). Lipids, representing more than 20% of the liver area in MCD diet-fed mice (Fig 3G,  H), mostly accumulated as single droplets (iD 6 3 X Xe,D 6 4 X X macro-steatosisD 6 5 X X; Fig 3G, I). Liraglutide infusion did not quantitatively affect hepatosteatosis in MCD diet-fed mice ( Fig 3GD 6 6 X XÀJ), in accordance with unchanged gene expression of central regulators of lipid metabolism ( Fig 3K). Nevertheless, lipid species differ in terms of lipotoxicity. 34,35 To evaluate potential changes in lipid composition, we analyzed the hepatic content of over 900 lipid species by targeted mass spectrometry. Hierarchical clustering analysis based on fold-change compared to the Chow group ( Fig 4A) highlighted lipid species significantly changed between the MCD and the MCD/Lira groups. Interestingly, the marked increases in C16 and C24-ceramide ( Fig 4B) and sphingomyelins ( Fig 4C) levels observed in the liver of MCD diet-fed mice were partly alleviated in the liver of MCD diet-fed mice treated with liraglutide. Different biochemical pathways (de novo ceramide biosynthesis vsD 6 7 X X degradation of sphingomyelin into ceramide) converge towardD 6 8 X X the generation of ceramide (Fig 4D 6 9 X XD). Gene expression of key enzymes involved in ceramide production was increased in MCD diet-fed mice Beyond their structural function, the degree of fatty acid saturation of lipid species also impacts the hepatic status. [36][37][38] To analyze the degree of fatty acid saturation/desaturation, we measured the levels of phospholipids harboring (D 7 1 X X1) only 1 double bond in both fatty acyl chains combined (considered as monounsaturated fatty acids, MUFA); (D 7 2 X X2) between 2À6 double bonds in their D 7 3 X X2 fatty acyl chains (considered as polyunsaturated fatty acids, PUFA); and (D 7 4 X X3) no double bonds in their D 7 5 X X2 fatty acyl chains (considered as saturated fatty acids, SFA). Hepatic PUFA lipids showed a trend to decrease in MCD/Lira mice compared to chow diet-fed mice while MUFA lipids were similarly decreased in the MCD and MCD/Lira groups compared to the Chow group ( Fig  4F). Saturated fatty acid containing lipids were specifically decreased in the MCD/Lira group compared to the Chow and MCD groups (Fig 4F). Among qualitative changes in lipid composition, the length of the fatty acid chain is another important feature for cell membrane permeability/fluidity. 35 Phospholipids with long chains (LC), harboring a combined carbon number of 28À34 in both fatty acyl chains, were reduced in both the MCD and MCD/Lira groups (Fig 4G). In contrast, phospholipids with very long chains (VLC), harboring 38D 7 6 X XÀ44 carbons, remained unchanged in the Chow, MCD, and MCD/Lira groups (Fig 4G).

Liraglutide limits inflammation and initiation of fibrosis in
mice on an MCD-dietD 7 7 X X. To further explore the pathology of hepatic inflammatory infiltrates in the MCD/Lira and the MCD groups of mice, we extended the analysis to expression of immune cell markers by means of quantitative real-time PCR. In the liver, macrophages are considered key players, even initiators, of the inflammatory process and are classified according to their tissue of origin or their activation state. 39,40 Molecular markers of both liver-resident Kupffer cells (F4/80, Tlr4, Tlr9) and monocyte-derived macrophages (Ly6c, Ccr2) were all overexpressed in the MCD group and normalized in the MCD/Lira group (Fig 5A). Similarly, markers of classically activated (pro-inflammatory) M1 macrophages (Cd14, Tnfa) were up-regulated in the MCD group and significantly reduced in the MCD/Lira group (Fig 5A). In contrast, no difference was observed between the MCD and the MCD/Lira group regarding the expression of alternatively activated (anti-inflammatory) M2 macrophage markers (Arg1, Cd206, Retnla, Cd163D 7 8 X X; Fig 5D 7 9 X XA). Markers of myeloid dendritic cells (Cd11c, MHCII, Cd86D 8 0 X X; Fig 5, B) and activated hepatic stellate cells (Acta1 and Vim, involved in the deleterious epithelial-mesenchymal transition, Fig 5D 8 1 X XC) were significantly reduced by liraglutide treatment. In accordance, the expression of several pro-fibrogenic genes was lower in the MCD/Lira group compared to the MCD group (Fig 5D 8 2 X XD), including Timp1, Serpine1, Mmp13. Even more strikingly, Tgfb1 and genes encoding collagen proteins showed the same basal levels of expression in the MCD/Lira as in the Chow group ( Fig 5D 8 3 X XD). Our experimental set-up (7 weeks of MCD-diet consumption) allowed initiating but not observing a massive fibrosis. Nevertheless, histological staining of hepatic sections with Sirius red showed that the administration of liraglutide reduced the number and the size of collagen fibers in livers of MCD diet-fed mice (Fig 5D 8 4 X XE). Quantitatively, MCD diet-fed mice exhibited a D 8 5 X X7-fold increase in the Sirius red positive area of liver parenchyma compared to mice fed a chow diet (Fig 5D 8 6 X XF). Per comparison, MCD diet-fed mice treated with liraglutide exhibited a significantly attenuated D 8 7 X X4-fold increase in Sirius red positive area compared to chow diet-fed mice (Fig 5D 8 8

X XF).
Liraglutide impacts the gut-liver axisD 8 9 X X. As an enterokine, GLP-1 production is regulated by the gut microbiota. 41,42 In turn, GLP-1R agonism can modify intestinal bacteria populations. 14,43 In this context, we analyzed and compared the changes in gut microbiota composition in our D 9 0 X X3 experimental groups based on sequencing of 16S rRNA gene amplicons. The overall bacterial diversity, measured by the Shannon diversity index, was lower in MCD diet-fed mice, whether treated or not treated with liraglutide (3.37 § 0.46 and 2.98 § 0.23, respectively) than in control mice (4.70 § 0.33). At the phylum level, MCD-diet consumption increased the relative proportion of Firmicutes and massively decreased that of Bacteroidetes ( Fig 6D 9 1 X XA). Among Firmicutes, MCD-diet promoted an overall increase in the relative abundance of the family Erysipelotrichaceae, mostly due to the increase in genera Turicibacter and Allobaculum while the genus ASBT decreased (Fig 6D 9 2 X XBD 9 3 X X, C). MCD-diet was also associated with increased relative proportions of Clostridiaceae (notably represented by Clostridium), Lactobacillaceae (Lactobacillus) and Peptostreptococcaceae (Romboutsia) and reduced proportions of Lachnospiraceae (Eisenbergiella, Clostridium g21, and KE159538D 9 4 X X; Fig  6D 9 5 X XC). The global decrease in Bacteroides under MCDdiet was mainly due to a drop in the relative abundance of the family S24-7 and genera belonging to it.
The main effects of liraglutide administration in the MCD nutritional context was the relative increase in Allobaculum vs Turicibacter and the D 9 6 X Xnormalization of the Bacteroides content ( Fig 6D 9 7 X XC). Extending the investigations to the gut environment, we observed that gross ileal morphology was markedly impacted by the MCD diet. Villi and crypts were significantly shorter in MCD diet-fed mice compared to chow diet-fed mice ( Fig 6D 9 8 X XDD 9 9 X X, E). Absolute and relative numbers of proliferative (Ki67-positive) cells were significantly reduced in MCD diet-fed mice whilst liraglutide treatment partially restored the relative proportion of proliferative cells ( Fig  6D 1 0 0 X XFD 1 0 1 X X, G). Finally, the MCD-diet reduced the number of goblet (mucus-producing) cells by D 1 0 2 X X2-fold, which was attenuated by liraglutide infusion (Fig 6D 1 0 3 X XHD 1 0 4 X X, I).

DISCUSSION
Liraglutide is an efficient antiD 1 0 5 X Xdiabetic drug, recently approved to treat obesity. 44 However, the clinical impact of liraglutide on NAFLD/NASH remains elusive. Some studies reported an anti-steatotic action in diabetic patients, 22,45-51 but this was not confirmed by other studies. 52,53 In vivo rodent studies also reported anti-steatotic activity of GLP-1Ra, but most often secondary to weight loss or insulin sensitization in situations of positive energy balance, such as in ob/ob mice, 14 diet-induced obese (DIO) mice [14][15][16] and DIOrats. [17][18][19][20] In vitro studies suggested that GLP-1Ra exert diverse actions on hepatocytes, including prevention of endoplasmic reticulumD 1 0 6 X X stress, 9 activation of AMP-activated protein kinaseD 1 0 7 X X, 10-12 inhibition of lipogenesis 10 or modulation of insulin signaling. [11][12][13] The MCD diet has long been considered a gold standard to study nutritional NASH. This diet also causes an increase in energy expenditure linked to suppression of hepatic expression of stearoyl-coenzyme A desaturase-1 (SCD-1), the enzyme that catalyzes the rate-limiting step in the formation of MUFAs. 54 As MCD dietfed mice did not respond to peripheral signals (hypolipidemia, hypoglycemia, hypoinsulinemia, hypoleptinemia) aimed at increasing their food intake, 7,54-55 they developed NASH in a state of negative energy balance and weight loss.
Here, we observed that liraglutide does not influence body weight and glycemic parameters in the MCD-diet context, allowing us to dissociate its direct hepatic action from confounding, indirect metabolic effects. First, we observed no change in hepatic neutral lipid storage in response to liraglutide infusion in MCD dietfed mice. Partitioning of lipid droplets (macro vs micro steatosis) as well as gene expression of lipid enzymes and transporters were also unaffected by liraglutide. Together, these data indicate no direct lipid-lowering action of liraglutide in the liver of this lean NASH model, suggesting that the anti-steatotic action previously observed could be secondary to weight loss and/ or improved insulin sensitivity. Nevertheless, beyond the total hepatic lipid content, enrichment in some lipid species/intermediates is linked to liver insulin-resistance and progression of NASH. [56][57][58][59] Thus, we extended our investigations to the hepatic lipid composition through LC-MS. This approach revealed that liraglutide could limit the MCD-diet-induced accumulation of C16 and C24-ceramides/sphingomyelins, confirming previous studies implicating these lipid intermediates in metabolic disorders. In fact, CerS2 +/À mice presenting high C16-ceramides levels are prone to diet-induced steatohepatitis and insulin resistance. 60 CerS6 deficiency 61 or silencing, 62 which reduces C16ceramides levels, allows DIO resistance and improvement in insulin sensitivity and hepatic steatosis. Different pharmacological interventions have also shown a concomitant reduction in ceramide species and amelioration of NASH. Antagonism of peripheral can-nabinoidÀ1 receptor reversed the HFD-induced increase in CerS6 expression and ceramide C16:0 synthase activity. 63 Myriocin, an inhibitor of ceramide synthesis, reduced C16 and C24 ceramides in mice fed a western diet, improving several metabolic parameters including liver fibrosis. 64 To our knowledge, the effect of GLP-1Ra on hepatic ceramide accumulation has not been previously reported. Furthermore, GLP-1Ra Exendin-4 has been shown to prevent ceramide accumulation in cardiac progenitor cells 65 as well as in the heart of Dgat À/À mice. 66 Different biochemical pathways converge towards the generation of ceramide, including de novo ceramide biosynthesis and degradation of sphingomyelin into ceramide. 67 In our study, concomitant upregulation of C16/C24-ceramide and C16/C24-sphingomyelin species, as well as gene expression patterns of Sptlc2 and CerS6, suggest that different sphingolipid pathways are also involved in both the MCD diet-induced ceramide enrichment and, accordingly, ceramide suppression induced by liraglutide. This observation is in accordance with studies showing that sphingolipid metabolism leading to ceramide synthesis is upregulated in NAFLD. 68,69 Ceramides play an important role in NASH via their crosstalk with inflammatory mediators. Chronic inflammation in the liver (in particular the rise in TNFa levels) can initiate ceramide production in hepatocytes and, in turn, ceramides can increase cytokine secretion. 70 In humans, sphingolipid species also correlate with hepatic inflammation, suggesting a role for these metabolites during progression from steatosis to NASH. 71 Kupffer cells, the liver resident macrophages, are early responders to hepatocyte injuries, driving TNF-a production. 72 This initial inflammatory event seems to be targeted by liraglutide. In fact, the MCD diet-induced overexpression of Kupffer cell markers (F4/80, Tlr4) and Tnfa are, respectively, fully normalized and consistently alleviated by liraglutide infusion. Macrophage polarization is another important process in the control of inflammation. In fact, depending on multiple factors (for example presence of microbes or damaged tissues), macrophages can oscillate between a classically (pro-inflammatory/M1) activated state or an alternatively (anti-inflammatory/M2) activated state. 73 A previous study reported that liraglutide modulates Kupffer cell polarization leading to an anti-inflammatory-M2 status. 74 Nevertheless, even if we presently observe that liraglutide attenuates MCD-diet induced expression of M1 macrophage markers, we did not detect any enrichment in M2 macrophage markers in our liraglutide-treated mice. Taken together, our results show that liraglutide administration promotes hepatic anti-inflammatory action in MCD-diet-fed mice, in line with anti-inflammatory properties previously described in adiponectin/ApoE deficiency, 75 high-fat/high-cholesterol diets 18,76 and diabetic patients. 21,46,[77][78][79] A prolonged pro-inflammatory status can trigger fibrosis, a physio-pathological process with an imbalance between extracellular matrix deposition and reabsorption. Importantly, the severity of hepatic fibrosis is an independent predictor of disease evolution and mortality. 80 In addition to anti-inflammatory actions, we observed that liraglutide presents an anti-fibrotic activity. In fact, the expression of the master regulator of fibrosis initiation (Tgfb1) as well as downstream genes mediating epithelialÀmesenchymal transition and extracellular matrix remodeling, all increased by the MCD-diet consumption, were alleviated or fully normalized by liraglutide administration. Histologically, our experimental set-up (7 weeks of MCD-diet exposure) allowed us to observe the early stage of collagen accumulation. Collagen fiber deposits in liver parenchyma appear limited by approximately 40% in the liraglutide-treated group. In accordance, a previous study showed suppression of fibrotic phenotypes in MCD diet-fed mice treated with the GLP-1Ra exenatide. 81 In humans, anti-fibrotic properties of liraglutide are still debated. Liraglutide reduced liver fibrosis in obese women with polycystic ovary syndrome 82 and in some cohorts of type 2 diabetic patients 21,77 but not in others. 53 By what means liraglutide could dampen the fibrosis initiation remains an open question. Activation of hepatic stellate cells (HSCs) into proliferative, fibrogenic myofibroblasts is a key step in hepatic fibrosis progress both in experimental models and human liver NASH. 83 Interestingly, liraglutide can directly de-activate HSCs in vitro. 84 Further work is now required to evaluate the role of liraglutide on HSCs biology in vivo, in particular regarding their proliferation, differentiation and survival. In this way, it has been recently shown that GLP-2R is expressed in HSCs and that HSCs activation and fibrosis were increased in livers of Glp2r À/À mice. 85 Recently, signals from the gut were identified as important contributors to the pathogenesis of liver disorders. 86,87 In particular, the relationship between gut microbiota and inflammatory status of the liver has been widely studied. Our results showed modulation of hepatic gene expression of Tlr4 and Tlr9 (involved in the coupling between recognition of bacterial compounds and production of pro-inflammatory cytokines), which led us to analyze gut microbiota composition through NGS in our different experimental groups. The most impressive changes in gut microbiota composition, as a result of the MCD-diet consumption, were the reduction in the Bacteroidales S24-7 family, Lachnospiraceae and Erysipelotrichaceae ASTB. In contrast, the MCD-diet led to massive increases in the relative proportions of Clostridium, Turicibacter, Allobaculum, and Lactobacillus genera. These observations only partially recapitulate changes previously reported in the same nutritional context. A rise in Clostridium and Turicibacter proportions were in agreement with a previous study in MCD diet-fed mice. 88 In contrast, changes in the relative abundance of Lactobacillus showed an inverse trend (increase) as compared to earlier studies. 89,90 Liraglutide treatment allowed for a full recovery of the relative proportion of Bacteroides which was decreased by the MCD-diet. Interestingly, several interactions between Bacteroides species and the host immune system have been reported, [91][92][93] suggesting that an adequate presence of this bacterial genus is required in the microbiota in order to control the inflammatory balance. Among the Erysipelotrichi family, liraglutide treatment also favors the enrichment in Allobaculum at the expense of Turicibacter. How these microbiome changes contribute to the observed action of liraglutide remains to be elucidated. In fact, technical limitations in the culture of anaerobic bacteria as well as the absence of genera selective antibiotics hinder experimental fine-tuning of gut microbiota composition. Nevertheless, from a translational point of view, it is interesting to note that the relative proportion of Erysipelotrichi also correlates with the percentage change in the liver fat content during low choline diet consumption in human clinical studies. 94 Of note, the liraglutide-induced shift in gut microbiota composition appears to be highly diet-dependent. In fact, in response to liraglutide, the main changes observed in HFD-fed mice (concerning Proteobacteria and Akkermansia muciniphila) 14 are not observed in our MCD diet-fed mice.
By expanding our investigations to the intestine, which represents both a host/microbiota interface and the tissue source of endogenous GLP-1, we observed that liraglutide infusion does not restore villus or crypt height in MCD diet-fed mice. However, liraglutide increased the level of proliferative cells in the crypt, in line with the reversible intestinotrophic action of exendin-4 previously observed in rats 95 and that of liraglutide in mice. 96 This trophic action seems effective under physiological conditions, but not under pathophysiological conditions. In the Apc Min/+ mice model of colon cancer, GLP-1Ra did not increase crypt cell proliferation but rather stimulated crypt fission. 97 The preservation of mucus-producing cells, conferred by liraglutide, in the MCD-diet deserves further investigation, in particular to understand if this process could explain the observed change in gut microbiota.
In conclusion, liraglutide treatment led to a concomitant decrease in deleterious ceramide/sphingomyelin species, inflammation and initiation of fibrosis in the liver of MCD diet-fed mice, independently of weight loss and global liver steatosis. These results confirm a direct beneficial hepatic action of the GLP-1Ra liraglutide, with potential translational relevance for T2D patients affected by NASH.