Article Text

Download PDFPDF
Original research
Integrative study of diet-induced mouse models of NAFLD identifies PPARα as a sexually dimorphic drug target
  1. Sarra Smati1,2,
  2. Arnaud Polizzi1,
  3. Anne Fougerat1,
  4. Sandrine Ellero-Simatos1,
  5. Yuna Blum3,4,
  6. Yannick Lippi1,
  7. Marion Régnier1,
  8. Alexia Laroyenne1,
  9. Marine Huillet1,
  10. Muhammad Arif5,
  11. Cheng Zhang5,
  12. Frederic Lasserre1,
  13. Alain Marrot1,
  14. Talal Al Saati6,
  15. JingHong Wan7,8,
  16. Caroline Sommer1,
  17. Claire Naylies1,
  18. Aurelie Batut2,
  19. Celine Lukowicz1,
  20. Tiffany Fougeray1,
  21. Blandine Tramunt2,
  22. Patricia Dubot9,10,
  23. Lorraine Smith1,
  24. Justine Bertrand-Michel2,
  25. Nathalie Hennuyer11,
  26. Jean-Philippe Pradere2,
  27. Bart Staels11,
  28. Remy Burcelin2,
  29. Françoise Lenfant2,
  30. Jean-François Arnal2,
  31. Thierry Levade9,10,
  32. Laurence Gamet-Payrastre1,
  33. Sandrine Lagarrigue12,
  34. Nicolas Loiseau1,
  35. Sophie Lotersztajn7,8,
  36. Catherine Postic13,
  37. Walter Wahli1,14,15,
  38. Christophe Bureau16,
  39. Maeva Guillaume16,
  40. Adil Mardinoglu5,17,
  41. Alexandra Montagner2,
  42. Pierre Gourdy2,18,
  43. Hervé Guillou1
  1. 1 Toxalim (Research Center in Food Toxicology), INRAE, ENVT, INP- PURPAN, UMR 1331, UPS, Université de Toulouse, Toulouse, France
  2. 2 Institut des Maladies Métaboliques et Cardiovasculaires (I2MC), UMR1297, INSERM/UPS, Université de Toulouse, Toulouse, France
  3. 3 CIT, Ligue Nationale Contre Le Cancer, Paris, France
  4. 4 IGDR UMR 6290, CNRS, Université de Rennes 1, Rennes, France
  5. 5 Science for Life Laboratory, KTH-Royal Institute of Technology, Solna, Sweden
  6. 6 Experimental Histopathology Department, INSERM US006-CREFRE, University Hospital of Toulouse, Toulouse, France
  7. 7 INSERM-UMR1149, Centre de Recherche sur l'Inflammation, Paris, France
  8. 8 Sorbonne Paris Cité, Laboratoire d'Excellence Inflamex, Faculté de Médecine, Site Xavier Bichat, Université Paris Diderot, Paris, France
  9. 9 Laboratoire de Biochimie Métabolique, CHU Toulouse, Toulouse, France
  10. 10 INSERM U1037, CRCT, Université Paul Sabatier, Toulouse, France
  11. 11 Univ. Lille, INSERM, CHU Lille, Institut Pasteur de Lille, U1011-EGID, F-59000, Lille, France
  12. 12 INRAE, INSTITUT AGRO, PEGASE UMR1348, 35590, Saint-Gilles, France
  13. 13 Université de Paris, Institut Cochin, CNRS, INSERM, Paris, France
  14. 14 Lee Kong Chian School of Medicine, Nanyang Technological University, Singapore
  15. 15 Center for Integrative Genomics, University of Lausanne, Le Génopode, Lausanne, Switzerland
  16. 16 Hepatology Unit, Rangueil Hospital Toulouse, Paul Sabatier University Toulouse 3, Toulouse, France
  17. 17 Centre for Host-Microbiome Interactions, Faculty of Dentistry, Oral & Craniofacial Sciences, King’s College London, London, UK
  18. 18 Endocrinology-Diabetology-Nutrition Department, Toulouse University Hospital, Toulouse, France
  1. Correspondence to Dr Hervé Guillou, Toxalim, INRAE, Toulouse, Occitanie, France; herve.guillou{at}inrae.fr; Professor Pierre Gourdy; pierre.gourdy{at}inserm.fr

Abstract

Objective We evaluated the influence of sex on the pathophysiology of non-alcoholic fatty liver disease (NAFLD). We investigated diet-induced phenotypic responses to define sex-specific regulation between healthy liver and NAFLD to identify influential pathways in different preclinical murine models and their relevance in humans.

Design Different models of diet-induced NAFLD (high-fat diet, choline-deficient high-fat diet, Western diet or Western diet supplemented with fructose and glucose in drinking water) were compared with a control diet in male and female mice. We performed metabolic phenotyping, including plasma biochemistry and liver histology, untargeted large-scale approaches (liver metabolome, lipidome and transcriptome), gene expression profiling and network analysis to identify sex-specific pathways in the mouse liver.

Results The different diets induced sex-specific responses that illustrated an increased susceptibility to NAFLD in male mice. The most severe lipid accumulation and inflammation/fibrosis occurred in males receiving the high-fat diet and Western diet, respectively. Sex-biased hepatic gene signatures were identified for these different dietary challenges. The peroxisome proliferator-activated receptor α (PPARα) co-expression network was identified as sexually dimorphic, and in vivo experiments in mice demonstrated that hepatocyte PPARα determines a sex-specific response to fasting and treatment with pemafibrate, a selective PPARα agonist. Liver molecular signatures in humans also provided evidence of sexually dimorphic gene expression profiles and the sex-specific co-expression network for PPARα.

Conclusions These findings underscore the sex specificity of NAFLD pathophysiology in preclinical studies and identify PPARα as a pivotal, sexually dimorphic, pharmacological target.

Trial registration number NCT02390232.

  • nonalcoholic steatohepatitis
  • liver metabolism
  • lipid metabolism
  • gene expression

Data availability statement

Data are available in a public, open access repository. Data are available on reasonable request. All data relevant to the study are included in the article or uploaded as supplementary information. All of the liver gene expression profiling are deposited on public database (Gene expression omnibus). Any other data will be made available on request.

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Significance of this study

What is already known on this subject?

  • The liver is a sexually dimorphic organ and non-alcoholic fatty liver disease (NAFLD) is sexually dimorphic.

  • Peroxisome proliferator-activated receptor α (PPARα) is a nuclear receptor expressed in many tissues and involved in the transcriptional control of metabolic and inflammatory responses.

  • PPARα is a target for hypolipidaemic drugs in the fibrate family that are under investigation for NAFLD treatment.

What are the new findings?

  • An integrated study of NAFLD in mice revealed sex-specific metabolic and inflammatory responses to four different dietary challenges.

  • Transcriptome analysis across these diet-induced NAFLD models identified robust sex-biased genes in mouse liver, including PPARα-dependent genes and correlation networks.

  • In human NAFLD, liver molecular signatures provide evidence of sexually dimorphic gene expression profiles and a sex-specific co-expression network for PPARα.

  • Hepatocyte-restricted deletion of Pparα highlights sex-specific and PPARα-dependent responses to fasting, to pemafibrate and to diet-induced non-alcoholic steatohepatitis.

How might it impact on clinical practice in the foreseeable future?

  • This work emphasises the importance of considering sex specificity in NAFLD treatment, such as targeting hepatocyte PPARα.

Introduction

Non-alcoholic fatty liver disease (NAFLD) is the most common chronic liver disease worldwide and is associated with obesity and type 2 diabetes.1 The prevalence of NAFLD is estimated to be 25% worldwide.2 NAFLD ranges from steatosis to non-alcoholic steatohepatitis (NASH), which can further progress to fibrosis, cirrhosis and hepatocellular carcinoma.3 Diagnosis at early stages and monitoring of disease progression largely rely on histological analysis of liver biopsies. Although NAFLD is a major public health problem, no medication has yet been approved for its treatment, but several drugs are currently being tested in clinical trials.4 In the era of personalised medicine, taking into account the molecular basis of sex differences is essential for comprehending the pathophysiology, advances in individualisation of NAFLD management and drug development.5 6 Epidemiological studies have revealed that premenopausal women are protected against cardiovascular diseases,7 as well as NAFLD,8–11 and a recent study demonstrated different molecular signatures in men and women with NAFLD.12 However, the potential sexually dimorphic traits leading to female protection against the progression of fibrosis progression are still unclear.13–15

The molecular basis of sex-specific susceptibility to NAFLD has been investigated in rodents. Oestrogen deficiency promotes steatosis in ovariectomised mice fed a high-fat diet (HFD), and they are protected by 17β-oestradiol administration,16 indicating that oestrogens largely contribute to female protection. One of the mechanisms involves the peroxisome proliferator-activated receptor gamma coactivator 1-alpha (PGC1A), a transcriptional coactivator of nuclear receptors. To modulate the expression of proteins involved in protecting against oxidative stress in hepatocytes, oestrogens signal through a mechanism dependent on oestrogen receptor alpha (ERα) and PGC1A.17 In mice, part of the liver sexual dimorphism is not dependent on hepatocyte ERα.18 Glucocorticoid-dependent signalling has also been suggested to play a significant role in liver sexual dimorphism.19 20 In the present study, we further investigated the basis of sex-specific NAFLD susceptibility through an integrative study of diet-induced mouse models of NAFLD.

Several models have recently been developed to investigate more advanced stages of NAFLD, including NASH in mice.21 22 Although it is difficult to observe significant hepatocyte ballooning and fibrosis in mice, several groups have identified diets with specific formulations that promote NASH in male mice.23–25 We took advantage of these different dietary challenges to identify influential pathways in the sexual dimorphism of the pathology. Using male and female mice, we determined diet-induced phenotypic responses using a combination of approaches integrating multiple layers of information across these different metabolic challenges. This allowed us to identify hepatocyte peroxisome proliferator-activated receptor α (PPARα) as being associated with sex-dependent susceptibility to NAFLD. Accordingly, sex specificity in hepatocyte PPARα activity was characterised in response to fasting and to pharmacological activation by pemafibrate. Finally, we provide evidence of sex-specific molecular signatures in human NAFLD, including a sexually dimorphic co-expression PPARα network.

Methods

Mice and diets

Male and female C57BL/6J mice aged 13 weeks (n=12 per group and per sex) were fed a chow diet (CTRL, D12450J, Research Diets), an HFD (D12492, Research Diets), choline-deficient HFD (CDHFD, D05010402, Research Diets), Western diet (WD, TD.88137, Envigo) or WD with glucose (18.9 g/L) and fructose (23.1 g/L) in drinking water (WD GF)24 for 15 weeks.

Human liver samples from patients with NAFLD

Liver samples were selected from a cohort of patients with biopsy-proven NAFLD established in the Hepatology Department of the Centre Hospitalier Universitaire de Toulouse. The 80 biopsies were from 48 men and 32 women (78% menopausal).

See online supplemental file 1 for further methods and experimental details regarding NAFLD classification in patients.

Supplemental material

Results

Sex-specific features in response to different dietary challenges

First, we investigated the responses of male and female C57BL/6J mice aged 13 weeks to five different dietary challenges (figure 1A). They were fed either CTRL or one of four different hypercaloric diets for 15 weeks. All mice fed an hypercaloric diet significantly gained weight, except for females fed a CDHFD (figure 1B). With all diets, males gained more body weight than females (figure 1C). The food intake was higher in females regardless of the diet (figure 1D). Perigonadal (PG) and subcutaneous white adipose tissue (WAT) increased with all hypercaloric diets, especially in males with the exception of the PG WAT, which was higher in females fed an HFD (figure 1E). The relative liver weight was increased in males in response to HFD, WD and WD GF. Only WD and WD GF led to increased liver weight in females (figure 1E).

Figure 1

Sex-specific phenotype in response to different dietary challenges. (A) C57BL/6J female and male mice were fed a control diet (CTRL), high-fat diet (HFD), choline-deficient HFD (CDHFD), Western diet (WD) or a WD provided with glucose and fructose in drinking water (WD GF) for 15 weeks (n=12/group). (B) Change in body weight in response to each diet compared with CTRL in male (blue) and female (red) mice (n=12). (C) Relative weight gain after 15 weeks of diet. (D) Food intake expressed in grams based on mouse body weight for each diet. Food consumption was measured weekly. (E) Relative perigonadal (PG), subcutaneous (SC), white adipose tissue (WAT) and relative liver weight expressed in grams per gram body weight at the end of the experiment. (F) Oral glucose tolerance test (OGTT) assessed after 13 weeks of diet in males and females (n=12/group). Blue and red curves represent male and female glycaemia (mg/dL), respectively. (G) Area under the curve representing OGTT results. (H) Insulinaemia was measured after 6 hours of fasting. Results are the mean±SEM. #diet effect, *sex effect. * or #P<0.05, ** or ##p<0.01, *** or ###p<0.001. Differential effects were analysed by analysis of variance followed by Student’s t-tests with a pooled variance estimate.

Glucose tolerance, as assessed by the oral glucose tolerance test (OGTT), was impaired by all hypercaloric diets. Males were more affected than females by HFD and CDHFD, whereas glucose intolerance was similar between males and females in the WD groups (figure 1F and G). Fasting insulinaemia was higher in males fed any of the hypercaloric diets but not in females (figure 1H). Plasma cholesterol, but not plasma triglycerides, increased with all hypercaloric diets, more severely in males (online supplemental figure 1).

Supplemental material

Sex-specific gene expression profile in response to different dietary challenges

Next, we analysed hepatic gene expression. Principal component analysis of the transcriptome revealed a difference between male and female liver gene expression (figure 2A) regardless of diet. However, differences seemed to be stronger in males than in females (figure 2A). Differentially expressed genes (DEGs) were subjected to hierarchical clustering, further highlighting clusters of genes that are consistently sexually dimorphic with all diets (clusters 4, 5, 8 and 10) (figure 2B). However, we identified four clusters that exhibited sex-dependent responses to the diets (clusters 1, 2, 3 and 7). We performed Gene Ontology analysis (figure 2C) and revealed the upregulation of gene expression relative to fibrogenesis and PPAR signalling pathways (clusters 1 and 7) in males fed an HFD, WD or WD GF. The higher expression of genes representative of these two clusters, such as Col1a1, Pparγ2, Pparα, Vnn1 and Fgf21 was confirmed by RT-quantitative PCR (RT-qPCR) (figure 2D). Consistent with hepatic Fgf21 expression, circulating FGF21 was robustly increased in response to the different diets in males but not in females (figure 2E). In addition, we observed that genes related to platelet activation and immune response (cluster 3) were consistently expressed at high levels in females for all five diets but were upregulated in males fed a WD or WD GF (figure 2C). Finally, we noticed that WD and WD GF led to the upregulation of genes associated with the proteasome and biosynthesis of unsaturated fatty acids (cluster 2) in both sexes (figure 2C). The hepatic expression of the rate-limiting enzyme in monounsaturated fatty acid synthesis (Scd1) was assessed by RT-qPCR (figure 2F). Finally, cluster 10 relates to genes related to complement and coagulation cascade as well as cytochrome P450s (Kyoto Encyclopedia of Genes and Genomes (KEGG) function ‘chemical carcinogenesis’), such as Cyp4a14 (figure 2G), which are more expressed in females liver whatever the diet. Overall, this liver gene expression profiling evidences some robust differences between sexes and sex-specific responses to diets enriched in metabolic pathways.

Figure 2

Sex-specific hepatic gene expression profiles associated with different dietary challenges. (A) Principle component analysis (PCA) score plots of the whole transcriptomic dataset in the liver (n=6/group). Each dot represents an observation (animal) projected onto first (horizontal axis) and second (vertical axis) PCA variables. (B) Heatmap presenting data from a microarray experiment performed with liver samples (n=6/group). Hierarchical clustering is also shown, which allows the definition of 12 gene clusters (p≤0.05). (C) Gene Ontology analysis of clusters 1, 2, 3 and 7. The false discovery rate is provided for each category. (D) Gene expression in the liver derived from complementary quantitative PCR (qPCR) experiments (n=12/group) illustrating the selected clusters (Col1a1, Pparα, Pparγ2, Pparα, Vnn1 and Fgf21). (E) Plasma FGF21 was measured on samples collected at the end of the experiment. Gene expression in the liver derived from complementary qPCR experiments (n=12/group) illustrating the selected clusters, Scd1 (F) and Cyp4a14 (G). Results are the mean±SEM. #diet effect, *sex effect. * or #P<0.05, ** or ##p<0.01, *** or ###p<0.001. CDHFD, choline-deficient HFD; CTRL, control diet; ECM, extracellular matrix; HFD, high-fat diet; KEGG, Kyoto Encyclopedia of Genes and Genomes; WD, Western diet; WD GF, a WD provided with glucose and fructose in drinking water.

Sex-specific susceptibility to diet-induced NAFLD

To assess the correlation between changes in gene expression and the progression of liver steatosis, we performed histological analysis and hepatic lipid profiling. Males developed liver steatosis with all hypercaloric diets, but especially with HFD and WDs (figure 3A,B). Compared with males, females were protected against steatosis regardless of the diet, but greater liver lipid accumulation was observed in females fed the WDs compared with HFDs (figure 3A,B). The histological scores were confirmed by quantification of hepatic triglycerides (figure 3C). Free and esterified cholesterol were similarly increased in both sexes in response to WDs (figure 3C).

Figure 3

Sex-specific susceptibility to diet-induced steatosis and steatohepatitis. (A) Representative histological sections of liver stained with H&E from males and females in each group. Magnification 100×. (B) Liver steatosis estimated on histological liver sections. Scoring: parenchymal involvement by steatosis <5%, 0; 5%–33%, 1; 33%–66%, 2; >66%, 3 (n=12/group). (C) Neutral lipids (triglycerides, free and esterified cholesterol) were extracted from the livers and analysed by gas-liquid chromatography. (D) Inflammatory score for histological liver sections. Each tissue section was analysed for 10 microscopic fields (magnification 200×) to determine the mean number of inflammation foci per field (n=12/group). (E) NAFLD activity score (NAS) score and (F) alanine aminotransferase (ALT) determined from plasma samples collected at the end of the experiment. (G) Table summarising diet-induced non-alcoholic fatty liver disease (NAFLD) model characteristics and limitations. Results are the mean±SEM. #diet effect, *sex effect. * or #p<0.05, ** or ##p<0.01, *** or ###p<0.001. Differential effects were analysed by analysis of variance followed by Student’s t-tests with a pooled variance estimate. Histological scores were analysed by a non-parametric test (Kruskall-Wallis). CDHFD, choline-deficient HFD; CTRL, control diet; HFD, high-fat diet; WD, Western diet; WD GF, a WD provided with glucose and fructose in drinking water.

Next, we assessed liver inflammation. Lobular inflammatory foci were only detectable in male mice fed a WD or WD GF (figure 3D). Females were robustly protected against inflammatory cell infiltration of the liver. The NAFLD activity score (NAS), including liver steatosis level, was sexually dimorphic with all hypercaloric diets (figure 3E). In addition, the plasma alanine aminotransferase activity was elevated with all hypercaloric diets. This elevation was more severe in males (figure 3F). Males developed significant liver fibrosis with all hypercaloric diets, whereas, in females, significant fibrosis was only detected in response to the CDHFD (online supplemental figure 2A,B). Therefore, the most dimorphic phenotype concerning fibrosis and inflammation was obtained with WD and WD GF (online supplemental figure 2B), and these data were confirmed by pathological scoring (online supplemental figure 2C). The main results obtained in mice of both sexes in response to the different diets are summarised in figure 3G.

Supplemental material

Taken together, our data suggest that HFD can be considered a representative model of sexually dimorphic steatosis, and WDs as a representative model of sexually dimorphic steatosis with early signs of NASH. Females are less prone to damages induced by either an HFD or WD, which we further investigated.

Sex differences in the liver metabolome associated with susceptibility to NAFLD

We performed reporter metabolite analysis using the genome-scale metabolic model for liver tissue in the HFD and in the WD models, which are representative of sexually dimorphic steatosis and early NASH, respectively. This analysis is based on gene and predicts changes in hepatic metabolism. These predictions were sexually dimorphic and diet-dependent (figure 4A). Their hierarchical classification highlighted sexually dimorphic soluble metabolites specific for HFD (cluster 1) and a large cluster combining soluble metabolites and lipids for WD (cluster 2). These predictions were validated by liver nuclear magnetic resonance profiling. In males fed an HFD, we identified an increase in the relative abundance of reduced glutathione (GSH) associated with a decrease in glutathione precursors (figure 4B). Such markers of oxidative stress were not detected in females (figure 4C). Moreover, the relative abundance of trimethylamine and hypotaurine was lower in males fed a WD compared with CTRL (figure 4D), whereas the hepatic glucose concentration was significantly increased in females in response to the WD (figure 4E). Lactate, another marker of redox homeostasis was also significantly reduced in females when compared with males on both HFD and WD (figure 4F).

Figure 4

Sex-specific changes in the hepatic metabolome associated with susceptibility to steatosis and non-alcoholic steatohepatitis (NASH). (A) Reporter metabolites in the liver gene expression profile analysis of male and female mice fed a high-fat diet (HFD) or Western diet (WD) versus control (CTRL) (n=6/group) were investigated by referring to the liver genome-scale metabolic model (iHepatocytes2322). The adjusted p values for each reporter metabolite were calculated for upregulated and downregulated genes, and the −log10 of the p values is presented. c, cytoplasm; FAME, fatty acid methyl ester; m, mitochondria; n, nucleus; p, peroxisome; r, endoplasmic reticulum. (B) Coefficient plots related to the O-PLS-DA models discriminating between CTRL and HFD in males and females (C) derived from the liver extract 1H-nuclear magnetic resonance (NMR)-based spectra of mice in the HFD versus CTRL group (n=12/group). The figure shows the discriminant metabolites that are higher or lower in the HFD versus CTRL group. Metabolites are colour-coded according to their correlation coefficient: red indicating a very strong positive correlation (R2 >0.65). The direction of the metabolite indicates the group with which it is positively associated, as labelled on the diagram. (D) Coefficient plots related to the O-PLS-DA models discriminating between CTRL and WD in males and females (E) derived from the liver extract 1H-NMR-based spectra of mice in the WD versus CTRL group (n=12/group). The figure shows the discriminant metabolites that are higher or lower in the WD versus CTRL group. Metabolites are colour-coded according to their correlation coefficient: red indicating a very strong positive correlation (R2 >0.65). The direction of the metabolite indicates the group with which it is positively associated, as labelled on the diagram. (F) Area under the curve of the 1H-NMR spectra was integrated for the lactate signals. (G) Plasma levels of long chain acylcarnitines measured after 6 hours of fasting. Relative abundance of C20:1n-9 (H) and Cer18:1/20:0 (I) in the liver measured by GC-FID (n=6/group). (J) Gene expression profile of genes involved in fatty acid synthesis, desaturation and elongation, traffic and storage in droplets, glycerolipid and sphingolipid metabolism. Data represent mean±SEM, #diet effect, *sex effect. * or #P<0.05, ** or ##p<0.01, *** or ###p<0.001.

Lipids such as acyl-coAs were also predicted to be impacted significantly by both diet and sex (figure 4A). To investigate whether fatty acid β-oxidation may be sensitive to either sex or diet, we measured plasma acylcarnitines of various chain lengths (figure 4G and online supplemental figure 3D). While medium chain acylcarnitines were not significantly modified, short chain acylcarnitines and long chain acylcarnitines were diet-sensitive but not significantly different between sexes. Cluster 2 of metabolites predicted to be increased in males fed a WD also included eicosanoyl-coAs, suggesting remodelling of long chain fatty acids (figure 4A). Therefore, we measured the relative abundance of hepatic lipids (online supplemental figure 3A) that revealed a marked effect of both the diet and the sex on liver lipidome (online supplemental figure 3B). Consistent with metabolic modelling, lipidomic analysis revealed marked changes in the relative abundance of long chain fatty acids (online supplemental figure 3C) such as eicosenoic acid (C20:1n-9) (figure 4H) and eicosapentaenoic acid (C20:5n-3) (online supplemental figure 3E). Changes in the relative abundance of long chain fatty acids correlates with marked changes in the relative abundance of complex lipids containing long acyl chains such as sphingolipids (figure 4I and online supplemental figure 3F) and phospholipids (online supplemental figure 3C). Since long chain acylcarnitines were not different between sexes, we assessed whether changes in the expression of genes encoding key enzymes in fatty acid synthesis, desaturation, elongation, storage, glycerolipid and sphingolipid homeostasis were sexually dimorphic (figure 4J). Our analysis revealed major sex-specific profiles of genes such as Elovl3 and Cidec that are overexpressed in males while genes such as Pnpla3 and Acot3 are overexpressed in females.

Supplemental material

Altogether, these results suggest that steatosis induced by either HFD or WD is associated with sex-specific rewiring of liver gene expression and lipid metabolism.

Sex-biased hepatic transcriptome profile

Beyond liver metabolism, we next analysed the sexually dimorphic changes in gene expression associated with susceptibility to NAFLD (HFD vs CTRL and WD vs CTRL). Only a few genes presented similar regulation in both sexes in response to HFD (online supplemental figure 4A–C) and WD (online supplemental figure 4D–F). We further analysed the sexually dimorphic gene expression profiles from CTRL, HFD and WD-fed mice to define sex-biased genes, that is, DEGs between males and females in these three conditions. First, we identified the genes that are differentially expressed between males and females by taking into account the magnitude of change in their expression in each diet group (figure 5A). Genes highly expressed in females compared with males included CYP450 (Cyp2b13, Cyp3a44), sulfotransferases (Sult2a1, Sult2a3, Sult2a4, Sult3a1), Slc22a26 and Fmo3 regardless of the diet. Genes highly expressed in males compared with females include CYP450 (Cyp4a12, Cyp2d34), Elovl3, Slco1a1 and Hsd3b5, regardless of the diet. Moreover, KEGG categories associated with the main physiological activities of the liver were found to be sexually dimorphic (online supplemental figure 4G). As expected, steroid hormone biosynthesis was the most sex-biased category (50.7%). However, metabolism of xenobiotics, retinol metabolism and complement and coagulation cascades were categories in which >30% of genes were sex-biased. In addition, we used a set of genes associated with either hepato-specific functions or NAFLD in databases (KEGG) and the literature (online supplemental table 1 and online supplemental figure 5A). Normalised expression of these genes revealed a clustering of genes (online supplemental figure 5B) regulated by both sex and diet (online supplemental figure 5C). We also presented the hierarchical clustering of the different groups based on the hepatic gene expression profile of this set of liver-specific genes (online supplemental figure 5D). Males and females were highly segregated.

Supplemental material

Supplemental material

Supplemental material

Figure 5

Sex-biased hepatic gene expression profiles in control (CTRL), high-fat diet (HFD) and Western diet (WD). (A) Volcano plot of differences in gene expression between males and females in CTRL, HFD and WD. Colours indicate p<0.05 and log (base 2) fold change >2 (red), p<0.05 and log (base 2) fold change <2 (blue) and non-significant (NS) (black). (B) Venn diagram representing the number of genes significantly upregulated in females compared with males receiving CTRL, HFD and WD (p<0.05). Enrichment analysis of the 1017 genes consistently overexpressed in females compared with males in the three experimental groups (female-biased genes). (C) Venn diagram representing the number of genes significantly upregulated in males compared with females in CTRL, HFD and WD (p<0.05). Enrichment analysis of the 1128 genes consistently overexpressed in males compared with females in the three experimental groups (male-biased genes). (D) Networks of the 50 genes showing the highest absolute correlation with each gene of interest (red node). Pparα, Pparβ/δ and Pparγ in males from the CTRL, HFD and WD groups are presented as circle plots (n=6/group). The edges corresponding to significant correlations are presented (Bonferroni-adjusted p<0.05). Another network circle plot based on these 50 genes is presented for female mice. Magenta nodes correspond to genes that significantly correlate with the gene of interest. Red and blue were used for positive and negative correlations, respectively.

Next, we investigated the sex-biased genes robustly upregulated in the female (figure 5B) or male liver (figure 5C). Gene Ontology analysis performed on the 1017 female-biased genes revealed KEGG categories related to complement and coagulation cascades, steroid hormone biosynthesis, linoleic acid metabolism, platelet activation, chemokine signalling and bile secretion (figure 5B). Gene Ontology analysis performed on the 1128 male-biased genes revealed KEGG categories related to metabolism, including xenobiotic and steroid hormone homeostasis, as well as the peroxisome (figure 5C). This analysis further highlights the major differences between males and females in most of the physiological activities of the liver, which are not restricted to steroid hormone biosynthesis. Linoleic acid metabolism and peroxisome are related to PPARs, which are transcription factors activated by fatty acids and involved in the regulation of genes that determine hepatocyte peroxisome biogenesis. We performed a correlation network analysis of the three PPAR isotypes, revealing dense networks, with liver PPARα being strongly sexually dimorphic (figure 5D) and suggesting that PPARα contributes to hepatic sexual dimorphism.

PPARα-associated hepatic gene network is sexually dimorphic in human NAFLD

Because our data indicate that PPARα activity is sexually dimorphic in mice, we also questioned the influence of sex on PPARα activity in human disease. Therefore, we compared the liver gene expression profiles by microarray analyses in liver samples from 32 women and 48 men (mean age 58.1±10.4 and 56.1±11.9 years, respectively) included in a cohort of patients with biopsy-proven NAFLD (online supplemental table 2). Both in mouse (online supplemental figure 6A) and in human (online supplemental figure 6B), correlation matrix analysis of gene expression revealed sexually dimorphic pattern.

Supplemental material

Supplemental material

Gene Ontology biological function analysis revealed that the most significant functional categories differentiating men and women in regard to the liver were related to inflammation and immune response, and metabolism (figure 6A). We analysed sexual differences in gene expression that correlated with PPARα and found that genes correlating with PPARα expression were more abundant in men (n=598) than in women (n=248) (figure 6B). Focusing on the 50 best correlations with PPARα expression (figure 6C) further highlighted differences between men and women. The correlation network of FGF21, a well-recognised PPARα target gene (figure 6D), was also sexually dimorphic. In addition, the correlation network of several coregulators (NCOR2, TBL1X, TBL1XR) of PPARα was dimorphic (online supplemental figure 7). In contrast, the correlation network of other PPAR isotypes were not dimorphic (online supplemental figure 7). Finally, our analysis revealed differences in the enrichment networks of genes that significantly correlated with PPARα expression in men (figure 6E) and women (figure 6F). Taken together, these results suggest that liver PPARα activity contributes to the sexual dimorphism of human NAFLD.

Supplemental material

Figure 6

Gene expression profiling in human non-alcoholic fatty liver disease (NAFLD) highlights peroxisome proliferator-activated receptor α (PPARα) as a sexually dimorphic target. Gene expression profiles were analysed by microarrays in 48 men and 32 women with biopsy-proven NAFLD. (A) Enriched terms across differentially expressed genes (p<0.05) between liver samples from men and women. (B) Venn diagram presenting the number of hepatic genes that correlated with PPARα mRNA expression in each sex (adjusted p value <0.05). (C) Networks of the 50 genes showing the highest absolute correlation with PPARα and FGF21 (D). The edges corresponding to significant correlations are presented (Bonferroni-adjusted p<0.05). The network circle plot based on the top 50 genes selected in men is presented on the left while the corresponding circle plot in women is on the right. Magenta nodes correspond to genes that significantly correlate with the gene of interest. Red and blue were used for positive and negative correlations, respectively. (E) Network of enriched terms across genes showing mRNA expression highly correlated with PPARα mRNA in men (adjusted p value <0.05). (F) Network of enriched terms across genes showing mRNA expression highly correlated with PPARα mRNA in women (adjusted p value <0.05).

Sexually dimorphic responses to fasting and drug treatment dependent on hepatocyte PPARα

We next tested whether physiological and pharmacological challenges that acutely regulate hepatocyte PPARα activity may induce sexually dimorphic responses. Fasting robustly regulates hepatocyte PPARα activity in vivo in male mice.26–28 Therefore, we evaluated hepatocyte PPARα-dependent responses to fasting in both sexes. On fasting, male and female Pparαhep−/− mice were hypoglycaemic compared with Pparα hep+/+ mice (figure 7A), and the fasting-induced increase in ketonaemia was blunted in Pparαhep−/− mice of both sexes (figure 7B). Compared with Pparαhep+/+ mice, fasted Pparαhep−/− males, but not females, had increased hepatic triglycerides (figure 7C). We also analysed hepatic gene expression in mice fasted for 16 hours. Our results revealed that, in the absence of hepatocyte PPARα, the fasting response was different between male and female mice (figure 7D). PPARα invalidation induced the overexpression of 2438 hepatic genes (p<0.01), including 255 exclusively in males, 1737 exclusively in females and 446 common to both sexes (figure 7D). This result indicated that the deletion of hepatocyte PPARα induces more extensive genomic response in females than males on fasting. In females, these 1737 upregulated genes in Pparαhep−/− fasted mice were mainly associated with functions related to nuclear factor kappa B or tumour necrosis factor signalling (figure 7E,F) suggesting higher inflammation.

Figure 7

Sexually dimorphic hepatocyte Pparα activation by fasting and by pharmacological agonist. Male and female C57BL/6J Pparαhep+/+ and Pparαhep−/− mice were fasted for 24 hours. (A) Glycaemia (mg/dL) and (B) ketonaemia (mmol/L) evolution in each group (n=12/group). Mice were either sacrificed after 18 hours of fasting or fed ad libitum. (C) Hepatic triglycerides were extracted from the livers and hepatic gene expression assessed by microarray analysis (n=6/group). (D) Venn diagram representing hepatic genes significantly upregulated (up) or downregulated (below) during fasting in Pparα hep−/− compared with Pparαhep+/+ in female (red circle) and male (blue circle) mice (p<0.01). (E) Gene Ontology analysis of upregulated and downregulated genes in the absence of hepatocyte Pparα during fasting, exclusively in females. The false discovery rate is provided for each category. (F) Gene expression in the liver derived from complementary quantitative PCR (qPCR) experiments (n=6/group) illustrating Ccl2 and Cxcl10. Results are given as the mean±SEM. #diet effect, *sex effect. * or #P<0.05, ** or ##p<0.01, *** or ###p<0.001. Male and female C57BL/6J Pparαhep+/+ and Pparαhep−/− mice were treated with pemafibrate (0.1 mg/kg) or vehicle (CMC) for 14 days by daily oral gavage. (G) Body weight, glycaemia and relative liver weight were measured at the end of the experiment. Hepatic gene expression was assessed by microarray analysis (n=6/group). (H) Volcano plot of differences in gene expression between pemafibrate and vehicle in male and (I) female mice. Colours indicate p<0.05 and log (base 2) fold change >2 (red), p<0.05 and log (base 2) fold change <2 (blue) and non-significant (NS) (black). (J) Venn diagram representing hepatic genes significantly upregulated (up) or downregulated (below) by pemafibrate compared with vehicle in Pparαhep+/+ female (red circle) and male (blue circle) mice (p<0.01). (K) Gene Ontology analysis of genes with dimorphic response to pemafibrate. The false discovery rate is provided for each category.

Finally, male and female Pparαhep+/+ and Pparαhep−/− mice fed a standard diet were treated with pemafibrate, a selective PPARα agonist. Body weight and glycaemia were not influenced by pemafibrate treatment (figure 7G). However, pemafibrate treatment induced significant hepatomegaly in both male and female wild-type mice but not in Pparαhep−/− mice (figure 7G). Gene expression influenced by pemafibrate was then analysed in males (figure 7H) and females (figure 7I). Regulation of liver gene expression in response to pemafibrate was strictly dependent on hepatocyte Pparα in both males and females (online supplemental figure 8A–D). In Pparαhep+/+ mice, the hepatic response to pemafibrate was sexually dimorphic (figure 7J). Several genes were only sensitive to pemafibrate in females (1000 upregulated and 802 downregulated genes), suggesting that the female liver is much more responsive to pemafibrate than the male liver. We analysed the interaction between sex and pemafibrate treatment and found that the sex-specific response was mostly related to gene functions relative to lipid homeostasis (figure 7K).

Supplemental material

Sexually dimorphic effect of hepatocyte PPARα in diet-induced NAFLD

To further investigate the potential sexual dimorphism of PPARα activity in the liver, we fed male and female Pparαhep−/− mice and their Pparαhep+/+ littermates a WD for 15 weeks. Male, but not female Pparαhep−/− mice were protected against body weight gain induced by the WD (figure 8A and B) while food intake, circulating leptin, adipose tissue weight, circulating free fatty acids were not modified by WD feeding (online supplemental figure 9A), only adiponectin was increased in response to WD feeding in female Pparαhep−/− mice (online supplemental figure 9A). In contrast to changes in body weight gain, glucose tolerance measured by the OGTT was not influenced by genotype either in males or females (figure 8C,D).

Supplemental material

Figure 8

Sexually dimorphic effect of hepatocyte proliferator-activated receptor α (PPARα) in diet-induced non-alcoholic fatty liver disease (NAFLD). Male and female C57BL/6J Pparαhep+/+ and Pparαhep−/− mice were fed a control diet (CTRL) or Western diet (WD) for 15 weeks (n=6–12/group). (A) Change in body weight in response to WD compared with CTRL male (blue) and female (red) mice (n=6–12). (B) Relative weight gain after 15 weeks of diet. (C) Oral glucose tolerance test (OGTT) assessed after 11 weeks of diet in males and females (n=6–12/group). Blue and red curves represent male and female glycaemia (mg/dL), respectively. (D) Area under the curve representing the OGTT results. (E) Representative histological sections of liver stained with H&E assessed in males and females from each group. Magnification 100×. (F) Liver steatosis estimated on histological liver sections. Scoring: parenchymal involvement by steatosis <5%, 0; 5%–33%, 1; 33%–66%, 2; >66%, 3 (n=12/group). (G) Hepatic triglycerides extracted from the livers. After extraction, lipids were analysed by gas-liquid chromatography. Results are given as the mean±SEM. #diet effect, *sex effect. * or #P<0.05, ** or ##p<0.01, *** or ###p<0.001. Differential effects were analysed by analysis of variance followed by Student’s t-tests with a pooled variance estimate. (H) Inflammatory score for histological liver sections. Each tissue section was analysed for 10 microscopic fields (magnification 100×) to determine the mean number of inflammation foci per field (n=6–12/group). (I) Histological fibrosis score in CTRL and WD groups (males and females). Sirius red staining was used to evaluate fibrosis as follows: 0, no fibrosis; 1, pericellular and perivenular fibrosis; 2, focal bridging fibrosis. Histological scores were analysed by a non-parametric test (Kruskall-Wallis). (J) Gene expression in the liver derived from complementary quantitative PCR (qPCR) experiments illustrating Pparα, Cyp4a10 and Fgf21. (K) Plasma FGF21 was measured on samples collected at the end of the experiment. Results are given as the mean±SEM. #diet effect, *sex effect. * or #P<0.05, ** or ##p<0.01, *** or ###p<0.001. Differential effects were analysed by analysis of variance, followed by Student’s t-tests with a pooled variance estimate. ND, not detectable.

Next, we evaluated liver steatosis by histological and liver lipid profile analyses. As expected, Pparα deletion in hepatocytes was sufficient to induce steatosis in male mice fed CTRL diet, but female Pparαhep−/− mice did not develop spontaneous steatosis (figure 8E and online supplemental figure 9B). WD-induced liver steatosis was not exacerbated by Pparα invalidation in either sex (figure 8E–G). Although males and females exhibited the same level of steatosis and alanine aminotransferase (online supplemental figure 9C), female Pparαhep−/− mice developed more severe inflammation (figure 8H and online supplemental figure 9D) and fibrosis (figure 8I) than Pparα hep+/+ females. This difference was not observed in males, except for the alpha smooth muscle actin (αSMA) staining (online supplemental figure 9E).

Associated with these changes, we observed several sex-specific regulations of genes well-known to be direct PPARα targets such as Cyp4a10 and Fgf21 (figure 8J) as well as Cyp4a14 and Cidec (online supplemental figure 9F).28 Interestingly, the expression of genes involved in lipid metabolism such as Elovl3, Pnpla3 and Mogat1 previously identified as sexually dimorphic were also dependent on hepatocyte PPARα (online supplemental figure 9F). However, other PPARα target genes such as Ehhadh or Vnn1 showed similar expression pattern in both sexes (online supplemental figure 9F). Importantly, we observed that circulating FGF21 levels increased in response to WD feeding through a mechanism requiring hepatocyte PPARα in both sexes (figure 8K). Circulating concentrations of plasma FGF21 was twice higher in male mice than in females.

Taken together, these results indicate sexually dimorphic activity of hepatocyte PPARα that controls liver lipid homeostasis on standard diet. Moreover, in response to WD sexually dimorphic activity of hepatocyte PPARα influences body weight gain, hepatic inflammation and the regulation of genes involved in lipid homeostasis as well as FGF21 level.

Discussion

In order to investigate the sex-specific susceptibility to steatosis in both mice and humans,29 30 we performed a large-scale integrative systems analysis to investigate the effect of diets promoting NAFLD in male and female mice. Our analysis revealed distinct metabolic and hepatic phenotypes with consistent sexual dimorphism. In agreement with this, steatosis severity was estimated by combining histological analysis and biochemical measurements, revealing that steatosis is consistently more severe in males than females. None of the preclinical models we used fully recapitulate the human disease as even the highest fibrosis score in WD-fed males was moderate and no ballooning was observed. In addition, we cannot exclude that, in mice, the differences observed between sexes are due to different time courses in the development of the disease that may take much longer time in females than in males.

However, this large set of experiments allowed us to investigate the interaction between diet and sex in the regulation of gene expression in the liver. Beyond the direct effects of micronutrients and macronutrients on metabolism and the progression of steatosis,31 the composition of the diet may also lead to changes in gut microbiota composition, with distinct dysbiosis between HFD and WD, and subsequently influencing hepatic homeostasis through bacterial metabolites and endotoxins.32 33 Our analysis suggests the existence of complex biological regulatory levels controlling sex-specific liver homeostasis that is highly applicable to the network of genes associated with or involved in NAFLD. Using network analysis, we identified hepatocyte PPARα as a sexually dimorphic player in mouse liver homeostasis. PPARα is a nuclear receptor highly expressed in hepatocytes that plays a central role in fatty acid homeostasis. It is activated by long chain fatty acids and controls hundreds of genes involved in metabolism and inflammation. It is particularly active during adipose tissue lipolysis,34 for instance, in response to starvation.26 During adipose tissue lipolysis, PPARα is also essential to drive systemic responses through the transcriptional control of FGF21,35 a liver-derived hormone controlling many aspects of metabolism and behaviour.36

Our observation agrees with a previous study37 showing that whole-body deletion of Pparα leads to sexually dimorphic changes in lipid metabolism and the sex-specific role of liver PPARα in protection from oestrogen-induced intrahepatic cholestasis.38 Preclinical39–42 and clinical43 studies have shown that PPARα is a pharmacological target in NAFLD.44

To further explore the relationship between PPARα activity and sexual dimorphism in the liver, we used mice lacking hepatocyte PPARα in different experimental settings. Our work confirms previous findings27 showing that male mice develop steatosis in the absence of hepatocyte PPARα, but this was not observed in female mice. In addition, hepatocyte PPARα influences both body weight gain in response to WD, as well as FGF21 production and liver fibrosis, in a sex-specific manner.

To investigate whether hepatocyte PPARα may also control sexually dimorphic responses to an acute physiological challenge,26 28 we investigated the hepatic response to fasting and provide evidence that hepatocyte PPARα determines sex-specific regulation of liver gene expression. In NAFLD, a large part of hepatic lipid accumulation are derived from adipose tissue.45 Moreover, several lines of evidence suggest that adipose tissue lipolysis controls liver activity through mechanisms including PPARα-dependent responses. During fasting, this adipose-to-liver response is challenged. Therefore, our data support the possibility that part of sex-specific responses involving PPARα might relate to modification of this interorgan communication.

To question the pharmacological relevance of our findings, we determined the changes in hepatic gene expression in response to an acute challenge with pemafibrate,42 a PPAR agonist that is currently being tested in clinical trials for the treatment of dyslipidaemia and NAFLD (ClinicalTrials.gov Identifier: NCT04079530 and NCT03350165, respectively). Our analysis revealed the very high specificity of the molecule for PPARα and that pemafibrate induces a sex-specific genomic response in the liver. Whether these sex-specific responses would be sufficient to alter the drug response to pemafibrate and other drugs currently in clinical trials remain to be investigated. Given the protection from NAFLD observed in female mice, experiments assessing the efficiency of drugs in both sexes cannot easily been tested.

In our experimental settings, the severity of steatosis observed in males depended on the type of diet. Both HFD and WD induced obesity and steatosis. In line with several previous reports, we found that females were protected from steatosis induced by HFD.46 47 In males fed an HFD, metabolomic profiling revealed changes in hepatic glutathione metabolism that were not observed in females. These results agree with studies highlighting that oxidative stress contributes to the progression of NAFLD.48 49 The hepatoprotective action of oestradiol, involving ERα and PGC1A in the transcriptional control of antioxidant proteins, has been demonstrated in hepatocytes from female mice.17 Moreover, supplementation of precursors of glutathione and NAD+ has been shown to decrease steatosis in mice.50 Finally, metabolic modelling of human samples showed that altered GSH and NAD+ metabolism is a prevailing feature in NAFLD.50 Therefore, it would be important to consider the potential sexually dimorphic differences in the oxidative stress response that account for the protection of women from steatosis and whether it declines with menopause.

Mice fed a WD developed more severe steatosis than those exposed to an HFD. In addition, WDs increased liver inflammation and induced significant fibrosis in males. Moreover, although modest steatosis occurred in WD-fed female mice, they remained protected from inflammation and fibrosis. Our metabolic modelling and lipidome analysis highlighted additional sexually dimorphic metabolites including a marked shift in the relative abundance of long chain fatty acids and complex lipids containing long acyl chains such as ceramides and phospholipids. Since we did not detect sex-specific change in the relative abundance of plasma acylcarnitines, we thought that our observations result from sex differences in lipid anabolism rather than catabolism. Our data support this hypothesis, as genes involved in fatty acid homeostasis are highly sexually dimorphic. Hepatic lipogenesis is known to be transcriptionally regulated51 and different between males and females.52 Our analysis highlights that the core of lipogenic genes and a set of genes involved in fatty acid elongation, traffic, fatty acid storage, acyl-coA synthesis and complex lipid synthesis (glycerolipids, sphingolipids) are differentially regulated in males and females. Importantly, some of these genes such as Elovl3,53 Cidec 54 and Pnpla3 55 have previously been linked to steatosis. Moreover, these data are in line with the finding that PPARα is an important regulator of hepatic sexual dimorphism since PPARα activity might be sensitive to the metabolism of fatty acids which are recognised ligands for this nuclear receptor. In addition, some of these sexually dimorphic genes have been described as directly regulated by PPARα in hepatocytes.27 28

Finally, we also questioned the relevance of our finding in human NAFLD. A recent study reported sexually dimorphic gene expression profiles in human NAFLD,12 and our analysis performed in an independent cohort extends and further confirms distinct molecular signatures in men and women with NAFLD. Using correlation analysis, we more specifically report sexually dimorphic PPARα-related and FGF21-related gene networks as well as PPARα-related functions in patients with NAFLD. Some limitations must be acknowledged in the interpretation of these data due to phenotypic differences between men and women with NAFLD. First, as observed in our cohort, body mass index (BMI) is usually higher in women than in men in individuals with metabolic diseases such as NAFLD.15 Second, since most of the women in our cohort are postmenopausal, it is likely that the sexual dimorphism observed here does not rely on the direct influence of female gonadal hormones. Analysis of larger cohorts are thus needed to further examine the influence of BMI as well as the hormonal status in sexually dimorphic liver gene expression in human NAFLD.

Taken together, our data provide a resource for investigating hepatic genes and pathways involved in the development of steatosis that are useful for preclinical research on sex differences in NAFLD. Moreover, this study identifies hepatocyte PPARα as a relevant sexually dimorphic target in NAFLD, further supporting the need to consider sex as a crucial determinant of NAFLD progression, biomarker discovery and therapeutic responses.

Data availability statement

Data are available in a public, open access repository. Data are available on reasonable request. All data relevant to the study are included in the article or uploaded as supplementary information. All of the liver gene expression profiling are deposited on public database (Gene expression omnibus). Any other data will be made available on request.

Ethics statements

Patient consent for publication

Ethics approval

This cohort was approved by the National Agency for the Safety of Medicines and Health Products (ANSM) and local ethics committee in 2015, and by the Minister of Higher Education, Research, and Innovation in 2017 (DC-2017-2984). In vivo studies were performed in accordance with European guidelines for the use and care of laboratory animals and approved by an independent ethics committee (authorisation 17430-2018110611093660 v3).

Acknowledgments

We thank all members of the EZOP staff for their help during this project. We thank the staff from Genotoul (Anexplo, GeT-TRiX, Metatoul-Axiom and Metatoul-Lipidomic) facilities for their help during this project. We thank the Phenotypage service—US006/CREFRE INSERM/UPS, and more specifically Laurent Monbrun for his technical assistance.

References

Footnotes

  • PG and HG contributed equally.

  • Contributors SS designed experiments, performed experiments, analysed the data and wrote the paper. AP, AF, SE-S, YB, YL and MR contributed to design experiments, performed experiments and analysed the data. AL, MH, FL, AMarr, TAS, JW, CS, CN, AB, CL, TF, BT, PD, LS and JBM performed experiments and contributed to data analysis. MA and CZ analysed the data. NH, J-PP, BS, RB, FL, J-FA, TL, LGP, SLa, NL, SLo, CP, WW, CB and AMard provided critical materials, supervised experiments and contributed to data analysis. MG provided critical reagents, contributed to design the project and to data analysis. AMon and PG designed the project, supervised experiments, analysed the data and wrote the paper. HG designed the project, supervised experiments, performed experiments, analysed the data and wrote the paper.

  • Funding SS was supported by a PhD grant from INSERM. AF was supported by a postdoctoral fellowship from Agreenskills. AL was supported by a grant from SNFGE. PG, BT and AMon were supported by grants from the “Société Francophone du Diabète”, the “Société Française d’Endocrinologie” and the “Société Française de Nutrition”. This work was funded by JPI Fatmal (RB, AMon, PG, SES, HG). WW, AMon, NL, PG and HG were supported by grants from the Région Occitanie. MG, CB, AMon, PG and HG were supported by a grant from AFEF. CP, AMon, PG and HG were supported by a grant from ANR (Hepatomorphic).

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.