Introduction

Non alcoholic fatty liver disease (NAFLD) has become a major public health concern worldwide1. NAFLD ranges from benign steatosis to non alcoholic steatohepatitis (NASH), which may progress to irreversible damage, such as fibrosis or hepatocarcinoma. The hallmark of NAFLD is an elevated level of neutral lipids, which accumulate as lipid droplets in hepatocytes2. Although the aetiology of the disease is not fully understood, it is strongly associated with obesity and type 2 diabetes (T2D). In human NAFLD, the fatty acids that accumulate in hepatocytes originate from dietary fat3,4, adipose tissue lipolysis and hepatic de novo lipogenesis3. In T2D, adipose tissue insulin resistance promotes lipolysis, whereas hyperglycaemia combined with hyperinsulinemia sustains hepatic de novo lipogenesis5.

Given the burden of the NAFLD epidemic, identifying molecular players that can be targeted is a rather important issue6,7. Moreover, finding drugs that may be used to treat NASH and its progression to irreversible liver disease is a so far unmet medical need to be solved8,9. Among drugs currently being tested in clinical trials are a number of molecules that activate the peroxisome proliferator activated receptors (PPARs)8,10. Three PPAR isotypes are known (α, β/δ, and γ), and they are members of the nuclear receptor family, which act as fatty acid sensors that orchestrate transcription in response to a variety of endogenous ligands10, such as fatty acids11, fatty acid derivatives12 and phospholipids13. Once activated by the binding of these lipids, PPARs may either induce or repress the expression of their specific target genes. PPARs are influential regulators of genes involved in metabolism in different tissues14. Therefore, several pharmacological agonists have been developed, tested in preclinical models of NAFLD15,16, and are currently being either used or tested in clinical trials for the treatment of metabolic diseases, and especially NAFLD16.

PPARα is the most abundant PPAR isotype in the healthy liver17 and in hepatocytes, PPARα regulates the expression of thousands of genes and contributes to the remarkable metabolic flexibility of the liver18,19,20,21. PPARα is particularly active during suckling22,23,24 and fasting19,20,25,26,27,28, two conditions in which fatty acids are a preferred source of energy for the organism. PPARα is also expressed in many other tissues, including skeletal muscle29, adipose tissues30,31,32,33, intestine34, heart35, and kidney36. Germline deletion of Pparα renders mice susceptible for many metabolic defects including obesity37, steatosis37,38,39, hepatic inflammation40 and steatohepatitis41, but not diabetes42,43. We have shown recently that a hepatocyte-specific deletion of Pparα induces spontaneous steatosis in aging mice and blunts fasting-induced ketogenesis19,20. Moreover PPARα is required for the expression of fibroblast growth factor 21 (FGF21)44,45, a liver-derived hormone with many endocrine46 and hepatoprotective functions47,48.

In the present study, we evaluated the importance of hepatocyte PPARα in steatosis associated with diet-induced obesity. We provide evidence that in mice fed a high fat diet (HFD), Pparα deletion in hepatocytes is sufficient to promote NAFLD. In addition, analysis of the hepatic transcriptome, lipidome, and metabolome, demonstrated that extrahepatic PPARα activity significantly contributes to metabolic homeostasis in response to HFD consumption.

Results

Hepatic and total Pparα deficiencies dissociate HFD-induced obesity and fatty liver from glucose intolerance

Male mice from different genotypes, namely wild-type (WT), germline Pparα-null (Pparα−/−) and hepatocyte-specific Pparα-null (Pparαhep−/−), were fed a low-fat diet (10% fat, CTRL) or a HFD (60% fat) at 8 weeks of age for 10 weeks at thermoneutrality (30 °C). At the beginning of the experiment, the Pparα−/− mice were already significantly heavier than the WT and Pparαhep−/− mice (Fig. 1a). All mice, independently of the genotype became overweight and gained approximately 15 g in response to HFD consumption (Fig. 1a,b). Moreover, unlike WT and Pparαhep−/− mice, Pparα−/− mice on CTRL diet also gained significant body weight. Therefore, Pparα−/− mice became more overweight than Pparαhep−/− and WT mice at thermoneutrality. In CTRL mice, oral glucose tolerance (OGTT) tested after 10 weeks of HFD feeding was similar regardless of the genotype (Fig. 1c,d). In the HFD-fed groups, WT mice became glucose intolerant whereas Pparαhep−/− and Pparα−/− mice were protected against this intolerance (Fig. 1c,d). These results are consistent with fasted glucose levels that increased in response to HFD only in WT mice, but not in Pparαhep−/− or Pparα−/− mice (Fig. 1e). Therefore, HFD feeding leads to fasting hyperglycaemia and glucose intolerance in WT mice, but not in Pparαhep−/− and Pparα−/− mice.

Figure 1
figure 1

Hepatic and total Pparα deficiency does not promote glucose intolerance in HFD- induced obesity. WT, Pparαhep−/−, and Pparα−/− mice were fed a control diet (CTRL) or a HFD for 10 weeks at 30 °C (thermoneutrality). (a) Body weight gain determined every week during the experiment. (b) Body weight at the end of the experiment. (c) Blood glucose measured during the oral glucose tolerance test (2 g/kg of body weight). (d) Area under the curve obtained after the oral glucose tolerance test. (e) Quantification of fasted glycaemia. (f) Plasma cholesterol (total, HDL, and LDL) and triglyceride plasma levels. Data represent mean ± SEM. #, significant diet effect and *, significant genotype effect. # or *p ≤ 0.05; ## or **p ≤ 0.01; ### or ***p ≤ 0.001.

Full size image

Different biochemical analyses were performed in plasma from fed animals (Fig. 1f). Total cholesterol, LDL-cholesterol, and HDL-cholesterol tended to increase in response to HFD diet in all three genotypes. However, we found that levels of the 3 lipid parameters were higher in the plasma of Pparαhep−/− mice than plasma from WT and Pparα−/− mice. Triglycerides were elevated in Pparα−/− mice fed the CTRL diet and the HFD diet. Triglycerides were elevated in response to HFD only in Pparαhep−/− mice, but were lower in HFD-fed Pparα−/− mice compared to control diet-fed Pparα−/− mice. Taken together, our results show that both hepatocyte-specific and whole-body deletions of Pparα promote obesity, which is dissociated from glucose intolerance in mice housed at thermoneutrality and fed a HFD.

Hepatic and total Pparα deficiencies promote liver steatosis, inflammation, and injury in HFD-induced obesity

First, we performed histological analysis in order to investigate whether the lack of Pparα either globally or liver specific was associated with changes in liver integrity (Fig. 2a). We observed that Pparαhep−/− and Pparα−/− mice developed steatosis upon CTRL diet feeding. In HFD, steatosis in Pparαhep−/− and Pparα−/− mice was much more severe than for WT mice, which is in agreement with their respective liver weight (Fig. 2a,b). To better characterize liver injury, we used the NAFLD activity scoring (NAS)49 based on the severity of NAFLD and the degree of inflammation (Fig. 2c). This NAS revealed that Pparαhep−/− and Pparα−/− mice fed a HFD exhibited increased lipid droplet deposition in the liver (Fig. 2a,c), which is confirmed by measurement of triglyceride liver content (Fig. 2e). NAS and inflammation scoring also revealed that HFD did not significantly increase hepatic inflammation in WT mice contrarily to both Pparαhep−/− and Pparα−/− mice for which at least 75% of mice presented a NAS higher or equal to 7 (Fig. 2c,d). In agreement with increased inflammation, HFD significantly increased plasma markers of liver injury (ALT and AST) in Pparα−/− and Pparαhep−/− mice (Fig. 2f).

Figure 2
figure 2

Hepatic and total Pparα deficiency promote liver steatosis and inflammation in HFD-induced obesity. WT, Pparαhep−/−, and Pparα−/− mice were fed a control diet (CTRL) or a HFD for 10 weeks at 30 °C (thermoneutrality). (a) Representative pictures of Haematoxylin and Eosin staining of liver sections. Scale bars, 100 µm. (b) Liver weight as a percentage of body weight. (c) NAFLD activity score (NAS). (d) Histological scoring of inflammation foci in 10 distinct areas at 200×. (e) Quantification of hepatic triglycerides (f) Plasma ALT and AST. Data represent mean ± SEM. #, significant diet effect and *, significant genotype effect. # or *p ≤ 0.05; ## or **p ≤ 0.01; ### or ***p ≤ 0.001.

Full size image

Gene expression profile in WT, Pparα hep−/−, and Pparα −/− mice in response to HFD-induced obesity

Next, we evaluated the hepatic transcriptome expression pattern in response to HFD using microarrays (details of the experiment and results from the microarray are available in Gene Expression Omnibus database by the name GSE123354). Overall, we identified a total of 8860 probes corresponding to 7173 HFD sensitive genes in at least one of the three genotypes (based on adjusted p-value; FDR < 5%, Fig. 3). Hierarchical clustering of genes highlighted 12 clusters showing specific gene expression profiles according to the experimental conditions (Fig. 3a). Four of them (clusters 1, 4, 11 and 12) showed a typical pattern in Pparα−/− mice compared to Pparαhep−/− and WT mice regardless of diet. Genes from cluster 1 and 4 were up-regulated in Pparα−/− in comparison to Pparαhep−/− and WT mice. On the contrary, genes from cluster 11 and 12 were down-regulated in Pparα−/−. The analysis of gene categories differentially expressed in Pparα−/− as compared to Pparαhep−/− and WT mice highlight the down-regulation of genes associated with PPAR signalling and metabolic homeostasis (Supplementary Figure 1).

Figure 3
figure 3

Analysis of the liver transcriptome in WT, Pparαhep−/−, and Pparα−/− mice in response to HFD. WT, Pparαhep−/−, and Pparα−/− mice were fed a control diet (CTRL) or a HFD for 10 weeks at 30 °C (thermoneutrality). A transcriptomic analysis performed with liver samples from WT, Pparαhep−/−, and Pparα−/− exposed or not exposed to HFD (n = 8 mice/group) revealed 7173 differentially regulated genes (FDR < 5%). (a) Heat map of microarray expression data from 7173 regulated genes. Red and green indicate values above and below the mean averaged centred and scaled expression values (Z-score), respectively. Black indicates values close to the mean. According to the probe clustering (left panel), 12 gene clusters exhibited specific gene expression profiles. (b) sPLS-DA representing the 7173 differentially expressed genes from the microarray. (c) Correlation analysis between either WT, Pparαhep−/−, and Pparα−/− mice for the effect of HFD. (d) Relative hepatic expression of Vnn1, Fmo3, Col1a1 and Pparγ2 quantified by qPCR. Data represent mean ± SEM. #, significant diet effect and *, significant genotype effect. # or *p ≤ 0.05; ## or **p ≤ 0.01; ### or ***p ≤ 0.001. (e) Venn diagrams comparing the number of genes significantly regulated under HFD in the livers of WT, Pparαhep−/−, and Pparα−/− mice at adjusted p-value < 0.05.

Full size image

Cluster 8 highlighted genes common between Pparα−/− and Pparαhep−/− mice but distinct from WT mice regardless of diet. Interestingly, HFD influences the liver transcriptome in a genotype-specific manner (clusters 2, 5, 6, 9 and 10). The sparse Partial Least Square Discriminant Analysis (sPLS-DA) shows that, although Pparα−/− mice liver gene expression profile is clearly different from others, there is also a marked effect of the diet in mice from the 3 genotypes (Fig. 3b). Moreover, the gene expression profile in the liver of HFD-fed mice revealed that: i) the response in WT mice is very different from Pparα−/− and Pparαhep−/− mice; ii) Pparα−/− and Pparαhep−/− mice have a closer response to HFD (Fig. 3c). Representative qPCR measurements of different gene expression are in line with the microarray analysis (Fig. 3d). The expression of Vnn1, encoding a liver-enriched oxidative stress sensor involved in the regulation of multiple metabolic pathways50, was identified as an HFD-responsive gene specific to the WT mice. Fmo3, involved in trimethylamine N-oxide (TMAO) production, was identified as an HFD-induced gene specifically in Pparα−/− mice. HFD increases the expression of collagen Col1a1 in the absence of hepatocyte-specific or whole-body Pparα but not in WT mice. Lastly, Ppar-γ2 was identified as an HFD-responsive gene common to the three mouse genotypes. We counted 354 hepatic genes responsive to HFD common to the three mouse genotypes (Fig. 3e). These genes are mostly involved in metabolic responses to HFD which do not depend on PPARα (Supplementary Figure 2).

PPARα-dependent changes in hepatic gene expression profiles in response to HFD-induced obesity

We next analysed the genes dependent on Pparα. A large group of 1749 Differentially Expressed Genes (DEGs) includes 922 genes significantly up-regulated and 827 significantly down-regulated by HFD feeding only in WT mice (Fig. 4a). Examples of these genes include well-established PPARα targets, such as Cyp4a14, Acot3, Acot2, and Fitm1 (Fig. 4b), and 8 categories of genes involved in metabolism (Fig. 4c) with up-regulated expression in response to a HFD only in WT mice. However, we also identified five KEGG categories down-regulated in response to HFD specifically in WT mice. These categories relate to RNA transport, ribosome activity and protein processing (Fig. 4d).

Figure 4
figure 4

PPARα-independent changes in hepatic gene expression profiles in response to HFD. WT, Pparαhep−/−, and Pparα−/− mice were fed a control diet (CTRL) or a HFD for 10 weeks at 30 °C (thermoneutrality). (a) Venn diagram presenting the number of hepatic genes over-expressed (bold) and down-regulated (regular) in response to HFD in WT, Pparαhep−/−, Pparα−/− mice (FDR < 5%) (b) Grey bars represent the top 15 specifically induced and repressed genes between WT exposed to CTRL diet and WT exposed to HFD. Red and blue bars represent the profile in Pparαhep−/− and Pparα−/− mice, respectively. (c) Gene Ontology (GO) enrichment analysis (p ≤ 0.01) of KEGG categories based on functional interactions specifically down-regulated in WT mice fed a HFD using the String database. (d) Gene Ontology (GO) enrichment analysis (adjusted p-value; p ≤ 0.01) of KEGG categories (based on functionally interactions) up-regulated in WT mice fed a HFD the using string database.

Full size image

Then, we questioned whether hepatocyte and wholebody deletion of Pparα (Pparα−/− and Pparαhep−/− mice) induces overlapping responses in HFD-induced obesity. We identified a group of DEGs including 337 and 349 genes significantly up-regulated or down-regulated, respectively, by HFD feeding in both Pparα−/− and Pparαhep−/− mice (Fig. 5a). Gene category analysis did not reveal any functions related to the 349 down-regulated genes by HFD feeding in Pparα−/− or Pparαhep−/− mice. However, gene category analysis highlighted the functions related to the 337 genes significantly up-regulated by HFD feeding in both Pparα−/− and Pparαhep−/− mice (Fig. 5b), suggesting that these genes are negatively regulated by PPARα. Most of these categories relate to the inflammatory process, including the NF-kappa B, TNF, and TLR signalling pathways. We selected the genes directly related to these pathways using the KEGG database and the gene database network (Supplementary Figure 3) and confirmed a marked up-regulation of genes belonging to NF-kappa B, TNF, and TLR in the hepatocyte-specific or whole-body absence of Pparα (Fig. 5c), in accordance with inflammatory markers measured (Fig. 2c,d).

Figure 5
figure 5

Hepatocyte PPARα prevents liver inflammatory gene expression in response to HFD. WT, Pparαhep−/−, and Pparα−/− mice were fed a control diet (CTRL) or a HFD for 10 weeks at 30 °C (thermoneutrality). (a) Venn diagram highlighting the number of hepatic genes over-expressed (bold) and down-regulated (regular) in response to HFD specifically in both Pparαhep−/− and Pparα−/− mice (FDR < 5%). (b) Gene Ontology (GO) enrichment analysis (adjusted p ≤ 0.01) of KEGG categories specifically up-regulated in both Pparαhep−/− and Pparα−/− mice fed a HFD. (c) Gene expression profile of genes identified as being involved in NF-kappa B, TNF and TLR signalling pathways in KEGG.

Full size image

Specific effect of hepatocytic PPARα and whole body PPARα deletion on liver gene expression in response to HFD-induced obesity

The venn diagram (Fig. 3e) also reveals a large group of DEGs modulated only in Pparαhep−/− mice fed a HFD compared to CTRL diet which encompassed 630 and 865 significantly up or down-regulated genes, respectively (Supplementary Figure 4). Figure 3e also defines another large group of DEGs including 1143 and 1257 genes significantly up-regulated and down-regulated, respectively, by HFD feeding only in Pparα−/− mice (Supplementary Figure 5). This suggests that the hepato-specific and whole-body deletions of Pparα have distinct and specific consequences in the hepatic response to HFD-induced obesity.

To further investigate their specific response, we first indentified the enriched terms across hepatic genes significantly responsive to HFD feeding in Pparαhep−/− mice only (Fig. 6a). This analysis highlights metabolic pathways as significantly enriched (Fig. 6a) and a main network (Fig. 6b) including genes related to isoprenoid metabolism and cholesterol synthesis. Second, we analyzed the enriched terms across hepatic genes significantly responsive to HFD feeding in Pparα−/− mice but not in WT nor in Pparαhep−/− mice (Fig. 6c). This analysis highlight pathways related to metabolism, hemostasis and inflammation as significantly enriched (Fig. 6c) and a main network (Fig. 6d) mostly including genes related to inflammation as well as cell morphology, adhesion and migration.

Figure 6
figure 6

Specific effect of hepatocytic PPARα and whole body PPARα expression on liver gene expression in response to HFD-induced obesity. WT, Pparαhep−/−, and Pparα−/− mice were fed a control diet (CTRL) or a HFD for 10 weeks at 30 °C (thermoneutrality). (a) Enriched terms across HFD sensitive genes (p < 0.05) only in Pparhepα−/− mice (b) Network of enriched terms across HFD sensitive genes (p < 0.05) only in Pparhepα−/− mice (c) Enriched terms across HFD sensitive genes (p < 0.05) only in Pparα−/− mice (d) Network of enriched terms across HFD sensitive genes (p < 0.05) only in Pparα−/− mice.

Full size image

Metabolic and lipidomic profiling of PPARα-dependent regulation of hepatic homeostasis in response to HFD

We performed unbiased hepatic metabolomic profiling of aqueous metabolites using proton nuclear magnetic resonance (1H-NMR). We used a projection to latent structures for discriminant analysis (PLS-DA) to investigate whether there was a separation between experimental groups of observations. A valid and robust PLS-DA model was obtained that discriminated HFD-fed Pparα−/− mice from all other groups (Supplementary Figure 6), further supporting the role of non-hepatocytic PPARα activity in liver homeostatic response to HFD.

We also performed a targeted analysis of 75 lipid species including neutral lipids (cholesterol, cholesterol esters, and triglycerides), phospholipids, and sphingolipids (Fig. 7). The relative abundance of each species in the livers of WT, Pparα−/−, and Pparαhep−/− mice fed one of the two diets (CTRL and HFD) was evaluated to determine the contribution of hepatocyte and whole body PPARα activity to hepatic lipid homeostasis. The results are presented as a heatmap with hierarchical clustering (Fig. 7a), in which we observed that the samples first clustered according to the diet, demonstrating that HFD-feeding was the main discriminating factor for hepatic lipid content. We identified four main clusters of lipids with distinct profiles relative to the different experimental conditions. Lipids in cluster 1, such as the ceramides d18:1/C18:1, d18:1/C18:0, and d18:1/C26:0 (Fig. 7b), exhibit increased relative abundance in HFD-fed Pparα−/− mice, suggesting that extra-hepatocytic PPARα contributes predominantly to lipid remodelling during HFD-feeding. Cluster 1 also contains linoleic acid (C18:2n-6), which exhibits increased abundance in HFD-fed Pparα−/− mice, but also in HFD-fed Pparαhep−/− mice. Lipids in cluster 2, such as the phospholipids PC36:3, PC28:6, PE38:4, and triglyceride TG C57 (Fig. 7c), are less abundant in HFD Pparα−/− mice. Lipids in cluster 3, such as the palmitoleic acid (C16:1n-7) and PE32:1 (Fig. 7d), are less abundant in HFD mice from the three genotypes. Lipids in cluster 4 (Fig. 7e), such as the polyunsaturated fatty acids C20:4n-6 and C22:5n-3 are more abundant in WT mice from the CTRL diet group and reduced in the livers of mice fed a HFD.

Figure 7
figure 7

PPARα-dependent regulation of hepatic lipid homeostasis in response to HFD. WT, Pparαhep−/−, and Pparα−/− mice were fed a control diet (CTRL) or a HFD for 10 weeks at 30 °C (thermoneutrality). (a) Heat map of data from hepatic lipid profiling in WT, Pparαhep−/−, and Pparα−/− mice exposed to HFD. Hierarchical clustering is also shown and defines four main lipid clusters. Representation of characteristic lipid species defining cluster 1 (b), 2 (c), 3 (d), and 4 (e). Data represent mean ± SEM. #, significant diet effect and *, significant genotype effect. # or *p ≤ 0.05; ## or **p ≤ 0.01; ### or ***p ≤ 0.001. Cer: ceramide; SM: sphingomyelin; TG: triglyceride; PC: phosphatidyl choline; PE: phosphatidyl ethanolamine; CE: cholesterol ester.

Full size image

Overall, this lipidomic profiling highlights that the hepatic lipidome depends on both the genotype and diet. Therefore, both hepatic Pparα and whole body Pparα deletions result in a specific lipidomic response to HFD feeding. However, whole body Pparα deficiency has a stronger influence on the effect of HFD-induced obesity on liver metabolic homeostasis.

Discussion

NAFLD is the hepatic manifestation of the obesity epidemic and represents a major public health issue worldwide51. NAFLD ranges from benign steatosis to NASH and may promote liver fibrosis and cancer. Therefore, there is a great interest in drugs that could be used to cure NAFLD or reverse NASH before it promotes irreversible damage8. PPARα, and other PPAR isotypes represent targets currently being tested in clinical trials10,52,53. PPARα is a ligand-activated nuclear receptor that plays a key role in the regulation of metabolic homeostasis by modulating the expression of rate-limiting enzymes involved in fatty acid degradation18,19,20. Preclinical studies in Pparα-null mice54 have shown that PPARα protects from steatosis19,54. Moreover, several clinical lines of evidence indicate that PPARα is also influential in human NASH55.

Most studies performed in vivo in Pparα-null mice have suggested that the mechanisms by which PPARα protects from steatosis and NASH involve its ability to transactivate genes required for fatty acid catabolism38 and to repress a number of inflammatory genes39,41. Because PPARα is expressed in many cell types and tissues with high fatty acid oxidation activity29,30,31,34, it is interesting to define in vivo the specific contribution of hepatocytic PPARα in preventing NAFLD.

When fed a regular diet, Pparα−/− mice are steatotic and overweight as previously reported26,38. When fed a HFD, they become even more steatotic and develop further liver inflammation39,41. Importantly, although Pparα-null mice develop steatosis, they do not exhibit reduced glucose tolerance compared to WT mice43. The data we obtained in Pparαhep−/− mice fed a HFD indicate that the deletion of Pparα in hepatocytes is sufficient to promote steatosis and inflammation. Therefore, hepatocyte-specific Pparα deletion promotes steatosis and liver inflammation while dissociating steatosis from glucose intolerance as observed in Pparα−/− mice42,43.

The mechanisms involved in the susceptibility to steatosis and protection from glucose intolerance likely involve the well-established role of PPARα in the control of fatty acid transport and degradation18. The mechanisms by which hepatocyte Pparα deficiency promotes NASH likely involve lipotoxic fat accumulation, including linoleic acid (C18:2n-6), which was recently identified as promoting NAFLD etiology in vivo56. Moreover, we extend previous observations and confirm the role of hepatocyte PPARα in repressing the expression of inflammatory genes, such as those involved in the NF-kappa B pathway57,58. However, by comparing the HFD response in Pparα−/− and in Pparαhep−/− mice we observed differences in the expression of genes involved in inflammation (cell morphology, adhesion and migration) strongly suggesting that PPARα activity might be important for liver immune cells activity in the context of obesity.

This study not only shows a specific effect of hepatocyte PPARα activity, but also identifies several roles of non-hepatocytic PPARα. First, we confirmed our previous observations that, unlike Pparαhep−/− mice, Pparα−/− mice gain weight when fed a regular diet19. Moreover, because the current observations were made at thermoneutrality, this weight gain is not likely due to defective PPARα expression and activity in the brown adipose tissue. In addition, by combining different large or medium scale analyses of the liver transcriptome and metabolome, we showed that the response to HFD was different in Pparα−/− mice compared to Pparαhep−/− mice, showing that extra-hepatocyte PPARα mediates at least a part of the adaptive response to HFD. Since we observed significant differences in lipid metabolism between Pparα−/− and Pparαhep−/− mice fed a HFD, our results are consistent with a possible extra-hepatic influence of PPARα on whole body fatty acid and cholesterol homeostasis.

Taken together, our data demonstrate that hepatocyte-specific deletion of Pparα promotes steatosis and inflammation in HFD-induced obesity and provide further pre-clinical evidence that hepatocyte PPARα is a relevant direct target in NAFLD. Our data also suggest that extra-hepatic PPARα plays a major homeostatic role in the control of the metabolic response to HFD. These data are in agreement with a recent study providing evidence that extrahepatic PPARα activity such as in skeletal muscle and heart, may contribute to whole body fatty acid homeostasis59. Further research is required to investigate in which cells, other than hepatocytes, PPARα regulates lipid metabolism in health and disease.

Material and Methods

Mice

All experiments were approved by the relevant animal care and use committee (CEEA-86, Ministry of Research and Higher Education, France); notification TOXCOM108, and conducted in accordance with the European directive 2010/63/UE.

Pparαhep−/− animals were created at INRA’s rodent facility (Toulouse, France) by mating the floxed-Pparα mouse strain with C57BL/6 J albumin-Cre transgenic mice (a gift from Prof. Didier Trono, EPFL, Lausanne, Switzerland) to obtain albumin-Cre+/−Pparαflox/flox mice (i.e., Pparαhep−/− mice as described in19). The Pparα deletion was confirmed with PCR and HotStar Taq DNA Polymerase (5 U/μl, Qiagen) using the following primers: forward: 5′- AAAGCAGCCAGCTCTGTGTTGAGC-3′ and reverse, 5′-TAGGTACCGTGGACTCAGAGCTAG-3′. The amplification conditions were as follows: 95 °C for 15 min, followed by 35 cycles of 94 °C for 1 min, 65 °C for 1 min, and 72 °C for 1 min, and a final cycle of 72 °C for 10 min. This reaction produced 450-bp, 915-bp, and 1070-bp fragments, which represented the Pparα sequence with an exon 4 deletion, the wild-type allele, and the floxed allele, respectively. The albumin-Cre allele was detected by PCR using the following primer pairs: CreU, 5′-AGGTGTAGAGAAGGCACTTAG-3′ and CreD, 5′- CTAATCGCCATCTTCCAGCAGG-3′; G2lox7F, 5′-CCAATCCCTTGGTTCATGGTTGC- 3′ and G2lox7R, 5′-CGTAAGGCCCAAGGAAGTCCTGC-3′).

Pparα-deficient C57BL/6 J mice (Pparα−/−) were bred at INRA’s transgenic rodent facility. Age-matched C57BL/6 J (provided by Charles River) were acclimated for 2 weeks to the local animal facility conditions prior to the experiment.

Mouse housing was controlled for temperature and light (12-h light/12-h dark). All mice were placed in a ventilated cabinet at the specific temperature of 30 °C (thermoneutrality) throughout the experiment. All animals used in these experiments were male mice.

Diet

WT, Pparα−/− and Pparαhep−/− mice were fed a standard diet (Safe 04 U8220G10R) until 8 weeks old, when the mice were fed a high fat diet (D12492, Research Diet) thought to induce mild hepatic lesions60. This HFD contains 60% calories from fat (lard), 20% calories from carbohydrates (7% sucrose) and 20% calories from proteins; or a chow diet (D12450J, Research Diet) containing 10% calories from fat, 70% calories from carbohydrates (7% sucrose) and 20% calories from protein during 10 weeks (until 20 weeks old). The estimated cholesterol content is: 0.014 g/kg (chow diet) and 0.228 g/kg (HFD diet). Experimental groups were designed as follows: WT CTRL, 8 mice; WT HFD, 8 mice; Pparαhep−/− CTRL, 10 mice; Pparαhep−/− HFD, 9 mice; Pparα−/− CTRL, 10 mice; Pparα−/− HFD, 10 mice.

Oral glucose tolerance test

Mice were fasted for 6 h and received an oral (2 g/kg body weight) glucose load. Blood glucose was measured at the tail vein using an AccuCheck Performa glucometer (Roche Diagnostics) at −15, 0, 15, 30, 45, 60, 90, and 120 minutes.

Blood and tissue samples

Prior to sacrifice, blood was collected in EDTA-coated tubes (BD Microtainer, K2E tubes) from the submandibular vein. All mice were killed in a fed state. Plasma was collected by centrifugation (1500 × g, 10 min, 4 °C) and stored at −80 °C. Following killing by cervical dislocation, the organs were removed, weighted, dissected and used for histological analysis or snap-frozen in liquid nitrogen and stored at −80 °C.

Gene expression

Total cellular RNA was extracted with Tri reagent (Molecular Research Center). Total RNA samples (2 µg) were then reverse-transcribed with the High-capacity cDNA Reverse Transcription Kit (Applied Biosystems) for real-time quantitative polymerase chain reaction (qPCR) analyses. The primers for Sybr Green assays are presented in Supplementary Table 1. Amplifications were performed on a Stratagene Mx3005P (Agilent Technology). The qPCR data were normalized to the level of the TATA-box binding protein (TBP) messenger RNA (mRNA) and analysed by the LinRegPCR v.11 software61.

Gene expression profiles were performed at the GeT‐TRiX facility (GénoToul, Génopole Toulouse Midi-Pyrénées) using Agilent Sureprint G3 Mouse microarrays (8 × 60 K, design 077809) following the manufacturer’s instructions. For each sample, Cyanine-3 (Cy3) labeled cRNA was prepared from 200 ng of total RNA using the One-Color Quick Amp Labeling kit (Agilent) according to the manufacturer’s instructions, followed by Agencourt RNAClean XP (Agencourt Bioscience Corporation, Beverly, Massachusetts). Dye incorporation and cRNA yield were checked using Dropsense™ 96 UV/VIS droplet reader (Trinean, Belgium). 600 ng of Cy3-labelled cRNA were hybridized on the microarray slides following the manufacturer’s instructions. Immediately after washing, the slides were scanned on Agilent G2505C Microarray Scanner using Agilent Scan Control A.8.5.1 software and fluorescence signal extracted using Agilent Feature Extraction software v10.10.1.1 with default parameters.

Microarray data and experimental details are available in NCBI’s Gene Expression Omnibus62 and are accessible through GEO Series accession number GSE123354 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE123354).

Histology

Formalin-fixed, paraffin-embedded liver tissue was sliced into 3 μm sections and stained with haematoxylin and eosin (H&E) for histopathological analysis. The staining was visualized with a Leica microscope DM4000 B equipped with a Leica DFC450 C camera. The H&E-stained livers sections were analysed blindly for the steatosis and the inflammation (NAFLD activity score or NAS) according to Kleiner et al.49, Steatosis was measured depending on i) the percentage of liver cells containing fat (Grade 0 to 3); ii) the localisation of steatosis (Grade 0 to 3); and iii) the presence of microvesicular steatosis (Grade 0 to 1). The degree of inflammation was appreciated by counting the inflammatory foci into 10 distinct area at 200X for each liver slice (Grade 0 to 3). Values represent the mean of 10 fields/liver slice.

Biochemical analysis

Aspartate transaminase (AST), alanine transaminase (ALT), total cholesterol, LDL and HDL cholesterols were determined from plasma samples using a COBASMIRA + biochemical analyser (Anexplo facility).

Analysis of liver neutral lipids

Lipids were analysed as previously described20. Tissue samples were homogenized in methanol/5 mM EGTA (2:1, v/v), and lipids (corresponding to an equivalent of 2 mg tissue) extracted according to the Bligh–Dyer method63, with chloroform/methanol/water (2.5:2.5:2 v/v/v), in the presence of the following internal standards: glyceryl trinonadecanoate, stigmasterol, and cholesteryl heptadecanoate (Sigma). Triglycerides, free cholesterol, and cholesterol esters were analysed by gas-liquid chromatography on a Focus Thermo Electron system equipped with a Zebron- 1 Phenomenex fused-silica capillary column (5 m, 0.25 mm i.d., 0.25 mm film thickness). The oven temperature was programmed to increase from 200 to 350 °C at 5 °C/min, and the carrier gas was hydrogen (0.5 bar). The injector and detector temperatures were 315 °C and 345 °C, respectively.

Liver fatty acid analysis

To measure all hepatic fatty acid methyl ester (FAME) molecular species, lipids that corresponded to an equivalent of 1 mg of liver were extracted in the presence of the internal standard, glyceryl triheptadecanoate (2 μg). The lipid extract was transmethylated with 1 ml BF3 in methanol (14% solution; Sigma) and 1 ml heptane for 60 min at 80 °C, and evaporated to dryness. The FAMEs were extracted with heptane/water (2:1). The organic phase was evaporated to dryness and dissolved in 50 μl ethyl acetate. A sample (1 μl) of total FAME was analysed by gas-liquid chromatography (Clarus 600 Perkin Elmer system, with Famewax RESTEK fused silica capillary columns, 30-m × 0.32-mm i.d., 0.25-μm film thickness). The oven temperature was programmed to increase from 110 °C to 220 °C at a rate of 2 °C/min, and the carrier gas was hydrogen (7.25 psi). The injector and detector temperatures were 225 °C and 245 °C, respectively.

Liver phospholipid and sphingolipid analysis

Chemicals and reagents

The liquid chromatography solvent, acetonitrile, was HPLC-grade and purchased from Acros Organics. Ammonium formate (>99%) was supplied by Sigma Aldrich. Synthetic lipid standards (Cer d18:1/18:0, Cer d18:1/15:0, PE 12:0/12:0, PE 16:0/16:0, PC 13:0/13:0, PC 16:0/16:0, SM d18:1/18:0, SM d18:1/12:0) were purchased from Avanti Polar Lipids.

Lipid extraction

Lipids were extracted from the liver (1 mg) as described by Bligh and Dyer in dichloromethane / methanol (2% acetic acid) / water (2.5:2.5:2 v/v/v). Internal standards were added (Cer d18:1/15:0, 16 ng; PE 12:0/12:0, 180 ng; PC 13:0/13:0, 16 ng; SM d18:1/12:0, 16 ng; PI 16:0/17:0, 30 ng; PS 12:0/12:0, 156.25 ng). The solution was centrifuged at 1500 rpm for 3 min. The organic phase was collected and dried under azote, then dissolved in 50 μl MeOH. Sample solutions were analysed using an Agilent 1290 UPLC system coupled to a G6460 triple quadripole spectrometer (Agilent Technologies). MassHunter software was used for data acquisition and analysis. A Kinetex HILIC column (Phenomenex, 50 × 4.6 mm, 2.6 μm) was used for LC separations. The column temperature was maintained at 40 °C. Mobile phase A was acetonitrile and B was 10 mM ammonium formate in water at pH 3.2. The gradient was as follows: from 10% to 30% B in 10 min, 100% B from 10 to 12 min, and then back to 10% B at 13 min for 1 min to re-equilibrate prior to the next injection. The flow rate of the mobile phase was 0.3 ml/min, and the injection volume was 5 μl. An electrospray source was employed in positive (for Cer, PE, PC, and SM analysis) or negative ion mode (for PI and PS analysis). The collision gas was nitrogen. Needle voltage was set at +4000 V. Several scan modes were used. First, to obtain the naturally different masses of different species, we analysed cell lipid extracts with a precursor ion scan at 184 m/z, 241 m/z, and 264 m/z for PC/SM, PI, and Cer, respectively. We performed a neutral loss scan at 141 and 87 m/z for PE and PS, respectively. The collision energy optimums for Cer, PE, PC, SM, PI, and PS were 25 eV, 20 eV, 30 eV, 25 eV, 45 eV, and 22 eV, respectively. The corresponding SRM transitions were used to quantify different phospholipid species for each class. Two MRM acquisitions were necessary, due to important differences between phospholipid classes. Data were treated with QqQ Quantitative (vB.05.00) and Qualitative analysis software (vB.04.00).

Metabolomic analyses by 1H nuclear magnetic resonance (NMR) spectroscopy

1H NMR spectroscopy was performed on aqueous liver extracts prepared from liver samples (50–75 mg) homogenized in chloroform/methanol/NaCl 0.9% (2/1/0.6) containing 0.1% butyl hydroxytoluene and centrifuged at 5000 × g for 10 min. The supernatant was collected, lyophilized, and reconstituted in 600 μl of D2O containing 0.25 mM TSP [3-(trimethylsilyl)propionic-(2,2,3,3-d4) acid sodium salt] as a chemical shift reference at 0 ppm. All 1 H NMR spectra were obtained on a Bruker DRX-600 Avance NMR spectrometer operating at 600.13 MHz for 1 H resonance frequency using an inverse detection 5 mm 1H-13C-15N cryoprobe attached to a CryoPlatform (the preamplifier cooling unit). The 1 H NMR spectra were acquired at 300 K with a 1D NOESY-presat sequence (relaxation delay – 90°-t-90°-tm-90°-acquisition). A total of 128 transients were acquired into a spectrum with 20 ppm width, 32 k data points, a relaxation delay of 2.0 s, and a mixing delay of 100 ms. All 1 H spectra were zero-filled to 64 k points and subjected to 0.3 Hz exponential line broadening before Fourier transformation. The spectra were phase and baseline corrected and referenced to TSP (1 H, d 0.0 ppm) using Bruker Topspin 2.1 software (Bruker GmbH, Karlsruhe, Germany). Multivariate analysis of metabolomic data was performed.

Statistical analysis

Biochemical, qPCR and phenotypic data were analysed using and graphpad software. Differential effects were assessed on log2 transformed data by performing ANOVA followed by Sidak post-hoc tests. p-values < 0.05 were considered significant.

Hierarchical clustering of lipid quantification data was performed using R (R Development Core Team, 2018) with the heatmap.2 function from the package, gplots. Data were log2 transformed, then centred and scaled by lipid. Hierarchical clustering was applied to the samples and the lipids using 1-Pearson correlation coefficient as distance and Ward’s criterion (Ward.D2) for agglomeration. All the data represented on the heat map had adjusted p-values <0.05 for one or more comparisons performed with an analysis of variance.

Microarray data were analyzed using R and Bioconductor packages (www.bioconductor.org, v 3.0), as described in GEO accession GSE123354. Raw data (median signal intensity) were filtered, log2 transformed, corrected for batch effects (microarray washing bath) and normalized using quantile method64.

A model was fitted using the limma lmFit function65 considering array weights using arrayWeights function. Pair-wise comparisons between biological conditions were applied using specific contrasts. A correction for multiple testing was applied using Benjamini-Hochberg procedure66 for False Discovery Rate (FDR). Probes with FDR ≤ 0.05 were considered to be differentially expressed between conditions.

In addition to the differential analysis, a multivariate exploratory analysis was performed. A Sparse Partial Least Squares Discriminant Analysis67 (sPLS-DA) was conducted using mixOmics package68 under to select the most discriminative variables (genes) that help classify the samples according to their experimental conditions among their expression values. Twenty iterations of 5-fold cross-validation was used to evaluate the model performance for the selection of the most informative components (6 components chosen) using all the variables (PLS-DA). Then a “sparse” PLS-DA model was parametrized selecting the first 100, 120, 80, 40, 100 and 20 (chosen according the performance results of 20 iterations of 5-fold cross-validations) most discriminant variables on the components 1 to 6 respectively.

Hierarchical clustering was applied to the samples and the differentially expressed probes using 1-Pearson correlation coefficient as distance and Ward’s criterion for agglomeration. The clustering results are illustrated as a heatmap of expression signals. Gene network and enrichment of KEGG pathways was either performed using the online software STRING V.1169 or Metascape70. Correlation graphic chart was generated using the chart correlation function from the Performance Analytics package.