Loss of hepatic Flcn protects against fibrosis and inflammation by activating autophagy pathways

Non-alcoholic fatty liver disease (NAFLD) is the most frequent liver disease worldwide and can progress to non-alcoholic steatohepatitis (NASH), which is characterized by triglyceride accumulation, inflammation, and fibrosis. No pharmacological agents are currently approved to treat these conditions, but it is clear now that modulation of lipid synthesis and autophagy are key biological mechanisms that could help reduce or prevent these liver diseases. The folliculin (FLCN) protein has been recently identified as a central regulatory node governing whole body energy homeostasis, and we hypothesized that FLCN regulates highly metabolic tissues like the liver. We thus generated a liver specific Flcn knockout mouse model to study its role in liver disease progression. Using the methionine- and choline-deficient diet to mimic liver fibrosis, we demonstrate that loss of Flcn reduced triglyceride accumulation, fibrosis, and inflammation in mice. In this aggressive liver disease setting, loss of Flcn led to activation of transcription factors TFEB and TFE3 to promote autophagy, promoting the degradation of intracellular lipid stores, ultimately resulting in reduced hepatocellular damage and inflammation. Hence, the activity of FLCN could be a promising target for small molecule drugs to treat liver fibrosis by specifically activating autophagy. Collectively, these results show an unexpected role for Flcn in fatty liver disease progression and highlight new potential treatment strategies.


Results
Hepatic loss of Flcn protects against high fat diet stimulated NAFLD. To study the role of FLCN in the liver, we generated a liver-specific Flcn KO (liver-Flcn −/− ) mouse model by crossing Flcn lox/lox mice with albumin-cre +/− mice ( Fig. 1A and supplemental Fig. 1A,B) 22 . Liver-Flcn −/− mice were born at the expected Mendelian frequency, displayed no developmental defects, survived without difficulty, were fertile, and had similar weights compared with wild-type mice at weaning (4 weeks of age).
We initially established whether Flcn loss in the liver affects whole-body metabolism by feeding them either normal chow or a high fat diet (HFD) containing 60% kcal from fat for 6 weeks. Control mice (Cre mice) gained significantly more weight than liver-Flcn −/− mice (~ 50% vs ~ 20%, respectively) on a HFD challenge (Fig. 1B). Hematoxylin and eosin (H&E) staining revealed a strong reduction in fat droplets in livers of Flcn KO mice when challenged with HFD compared to Cre mice (Fig. 1C). To understand how liver-specific loss of Flcn protected against HFD-induced obesity, we then examined if it improved systemic energy metabolism. First, glucose and insulin tolerance tests showed that loss of Flcn in hepatocytes significantly increased glucose uptake and enhanced insulin responsiveness, respectively, under both chow and a HFD regimen ( Fig. 1D and supplemental Fig. 1C). Secondly, we employed the use of metabolic cages capable of simultaneously measuring animal food intake, physical activity and energy expenditure. Mice fed either chow or HFD were examined individually in metabolic cages. Liver-Flcn −/− mice were found to be more active during the dark phase of the circadian cycle concomitant with increased maximal oxygen consumption (VO 2 ) compared to Cre controls independent of diet ( Fig. 1E and supplemental Fig. 1D). Together, the data suggest that the increased energy expenditure and metabolic activity of liver-Flcn −/− mice are underlying factors in the resistance of these animals to HFD-induced obesity. Hence, loss of Flcn specifically in the liver improved metabolic homeostasis and prevented weight gain in a diet-induced obesity model.

Flcn loss protects against MCD diet-induced fibrosis and inflammation.
Since liver phenotypes associated with a HFD are highly associated with obesity and whole-body insulin resistance, the effects observed in this experimental setting may have been an indirect consequence of reduced adiposity. To test whether protection against steatosis and NAFLD was due to cell-autonomous effects of Flcn KO in hepatocytes, we investigated whether loss of hepatic Flcn was also protective in a model of liver fibrosis that does not cause weight gain or insulin resistance. To this end, we subjected Cre and liver-Flcn −/− mice to an MCD diet for six weeks. The MCD diet is a classical dietary model of liver fibrosis. Although the diet comprises high sucrose (46%) and fat (10%), www.nature.com/scientificreports/ it lacks methionine and choline, which are indispensable for hepatic mitochondrial β-oxidation and very lowdensity lipoprotein (VLDL) synthesis. This diet is a model of a metabolic challenge and results in a significant and rapid onset of a NASH-like phenotype with fibrosis, inflammation, oxidative stress, and liver cell death 27 . As previously described 28 , we show that this diet induced significant weight loss (Supplemental Fig. 2A).
As seen for mice under HFD (Fig. 1), hepatic loss of Flcn protected against hepatic lipid accumulation in mice under the MCD diet, as shown with the H&E staining ( Fig. 2A). Indeed, specific quantification of the total triglycerides in the liver tissue also showed a significant reduction in fat accumulation in Flcn KO livers (Fig. 2B). There was no significant difference in serum triglycerides, serum total cholesterol, serum free cholesterol, and serum cholesterol esters between Cre and liver-Flcn −/− mice ( Fig. 2C and Supplemental Fig. 2B-D). These results suggest that liver-specific loss of Flcn reduces fat accumulation not through increased fat export into the blood. Levels of alanine transaminase (ALT), a hepatic enzyme released following liver damage due to hepatocyte death, increased significantly in Cre mice under MCD diet, an effect blunted in liver-Flcn −/− mice, thus supporting their protection against liver damage (Fig. 2D).
One hallmark of NASH is fibrosis, which generates liver damage and eventually leads to hepatocyte death and liver dysfunction 7 . In our mouse model, relative mRNA transcript levels of genes related to fibrosis were induced in Cre mice fed the MCD diet, but generally to a lower extent in liver-Flcn −/− mice (Fig. 3A). Sirius Red staining also revealed that fibrosis was aggravated in Cre mice challenged on the MCD diet, but this effect was diminished in liver-Flcn −/− mice (Fig. 3B, quantified in C). Immunohistochemistry staining of ⍺-smooth muscle actin (⍺-SMA), a fibrosis-related marker, was significantly increased in Cre mice challenged with the MCD diet and again loss of Flcn was protective (Fig. 3D, quantified in E). Finally, immunohistochemistry staining of 4-Hydroxynonenal (4-HNE), a lipid peroxidation marker, was significantly increased in Cre mice challenged with the MCD diet but diminished in liver-Flcn −/− mice (Fig. 3F, quantified in G). Taken together, these results demonstrate that liver-specific loss of Flcn protects against fibrosis.
NASH development is conjointly characterized by increased inflammation that contributes to liver damage and disease progression. In our mouse model, relative mRNA transcript levels of genes involved in inflammation weeks (mean ± SEM, two-way ANOVA, *P < 0.05; **P < 0.01; ****P < 0.0001 compared to Cre-chow, # P < 0.05; ## P < 0.01; #### P < 0.0001 compared to Cre-HFD; n = 6 mice per condition). (C) H&E stainings in liver sections of mice fed either chow or high-fat diet (HFD) over a period of 6 weeks. Images are representative of 6 mice per condition. (D) Blood glucose during a GTT in chow or HFD-fed mice following a 16-h fast and intraperitoneal glucose administration of 2 g per kg of body weight (mean ± SEM, two-way ANOVA, *P < 0.05; ***P < 0.001; n = 6 mice per condition). (E) Quantification of metabolic cage analysis of mice fed chow or a HFD for 2 months. Circadian VO 2 consumption levels during a 12 h light: 12 h dark cycle (mean ± SEM, two-way ANOVA, **P < 0.01; ****P < 0.0001; n = 6 mice per condition). www.nature.com/scientificreports/ were lower in liver-Flcn −/− mice compared to Cre mice when fed the MCD diet (Fig. 3H). Using a mouse protein cytokine array, we determined the cytokine and chemokine secretion profiles in mouse livers. Cytokine protein levels were higher in Cre mice challenged with the MCD diet compared to liver-Flcn −/− mice (Fig. 3I,J). More specifically, some macrophage chemo-attractants (RANTES, MCP-1, IP-10) were induced in Cre mice fed the MCD diet but to a lesser extent in liver-Flcn −/− mice (Fig. 3J). Normal livers are infiltrated by specialized macrophages that contribute to inflammation when activated. Notably, immunohistochemistry staining of F4/80, a marker of murine macrophages, was increased in Cre mice fed the MCD diet, but were not increased in liver-Flcn −/− mice (Fig. 3K, quantified in L). Collectively, our data indicates that liver-specific loss of Flcn reduces fibrosis and inflammation in mice challenged with the MCD diet.
Flcn loss activates autophagy in a mouse model of liver fibrosis. This protective impact on inflammation and fibrosis following loss of Flcn prompted us to investigate the molecular signature of liver-Flcn −/− mice responsible for the protective phenotype using high-throughput whole-transcriptome sequencing (mRNA-seq). Unsupervised hierarchical clustering highlighted three distinct patterns of gene expression changes (Fig. 4A, Supplemental Table 1), clusters 1 and 3 not being affected by the Flcn status. Interestingly, cluster 2 highlighted genes whose expression was induced in Cre mice receiving the MCD diet but that were not increased in liver-Flcn −/− mice (Fig. 4A,B). Gene enrichment analysis using GO Enrichr revealed the major pathways differentially affected in this cluster by the diet and the genotypes (Fig. 4C). The most significantly affected biological pathway was extracellular matrix organization, which represents genes involved in fibrosis development typically observed in NASH, such as Acta2, Mmp2, Serpine1, and Timp1 (as validated by qRT-PCR and consistent with histology experiments in Fig. 3). Other important pathways affected include mitotic sister chromatid segrega- (C) Serum triglycerides quantification in mice fed as described in (A) (mean ± SEM, two-way ANOVA, *P < 0.05; ****P < 0.0001; n = 8 mice per condition). (D) Alanine transaminase (ALT) quantification in mice fed as described in (A) (mean ± SEM, two-way ANOVA, **P < 0.01; ****P < 0.0001; n = 8 mice per condition). Chow MCD **** **** *** * * **** **** ** * www.nature.com/scientificreports/ tion, sister chromatic segregation, and mitotic nuclear division. These genes are generally induced in regenerating livers following injury 29 . Moreover, pathways involved in inflammation such as neutrophil activation and cellular response to cytokine stimulus were differentially expressed. Overall, genes involved in fibrosis, inflammation and mitotic cell cycle were induced in mice fed the MCD diet but to a lesser extent in Flcn KO mice ( Fig. 4D-F). Concurrently, no significant changes in expression between Cre and Flcn KO mice were detected for pathways involved in lipolysis (GO:0016042) and lipogenesis (GO:0008610) (Supplemental Fig. 3A-C).
Emerging data supports a role of autophagy in the liver to regulate intracellular lipid stores and energy homeostasis. Autophagy is an important process in which the cells degrade its own content by sequestering cargoes inside the lysosomes. Degradation of these cargoes inside the lysosome is a multistep process that can n eutrophil activation involved in immune response 48 7.7e-08 7.8e-05 6 c ellular response to cytokine stimulus 45 9.7e-08 8.2e-05

GO Enrichr
Expression (scaled)  www.nature.com/scientificreports/ be regulated by master transcription factors such as Transcription Factor EB (TFEB) and Transcription Factor Binding To IGHM Enhancer 3 (TFE3) 15,30 . The best described regulator of TFEB and TFE3 is the mTOR Complex 1 (mTORC1). Under non-stress conditions, mTORC1 phosphorylates TFEB and TFE3, resulting in cytoplasmic sequestration and inhibition of their activity 15 . Conversely, nutrient starvation induced inhibition of mTORC1 removes the repressive phosphorylation on TFEB and TFE3, resulting in their nuclear translocation and activation of a panel of genes involved in autophagy 15 . We and others have shown that FLCN is a negative regulator of TFEB and TFE3 and positive regulator of mTORC1 23,[31][32][33] . Moreover, FLCN exhibits GTPase-activating protein (GAP) activity upon nutrient replenishment targeting the Ras-related GTPase C and Ras-related GTPase D (RagC/D) resulting in mTORC1 activation, inhibition of TFEB and TFE3 nuclear translocation, and impaired autophagy and lysosomal function 24,25 .
Unsupervised hierarchical clustering of autophagy-related genes from our RNA-seq data revealed an important effect of the diet (Autophagy clusters A1 and A3, Fig. 5A), but also highlighted a subset of genes induced in liver-Flcn −/− mice (Autophagy cluster A2, Fig. 5A-B). Relative mRNA transcript levels of autophagy-related genes mRNA, such as Ctsa, Depp1, and Ctsf, were significantly increased in liver-Flcn −/− mice compared to Cre mice, while the diet had no effect (Fig. 5C). These genes are generally involved in lysosome/autophagosome maturation and activation, which can reduce hepatocellular injury and inflammation by clearing damaged or misfolded proteins and suppressing transcription or maturation of pro-inflammatory cytokines 34 . Immunoblot analysis additionally revealed higher protein levels of p62 in Cre mice challenged with the MCD diet, suggesting impaired autophagy, while Flcn loss reduced p62 protein levels (Fig. 5D, quantified in E). p62 mRNA transcript levels were induced in liver-Flcn KO but were unaffected by the diet (Fig. 5F). Therefore, the change in p62 protein levels is unlikely to be due to regulation at the mRNA level of p62. To further assess autophagy induction in a Flcn KO model, we measured autophagy flux in mouse embryonic fibroblasts (MEFs) in the absence and presence of BafA1 and Chloroquine, two compounds known to pharmacologically block autophagosome-lysosome fusion.
Flcn loss in this cell line increased autophagy flux upon autophagy blockade, as measured by higher LC3-II accumulation in Flcn KO after treatment with BafA1 and Chloroquine (Fig. 5G, quantified in H). We also measured LC3 levels in liver protein lysates from mice fed either the chow or the MCD diet. We show that Flcn KO mice had higher LC3-I and LC3-II levels in both diets, suggesting functional and active autophagy (Supplemental Fig. 4A, quantified in B and C). Furthermore, the major transcription factors regulating autophagy and lysosome biogenesis, TFE3 and TFEB, were localized in the nuclei of hepatocytes, which correlated with increased autophagy in Flcn KO mice (Fig. 5I-L). Taken together, our data suggest that targeting Flcn improves resistance to fibrosis and inflammation, possibly by inducing autophagy when challenged with a liver fibrosis-inducing diet.

Discussion
In this study, we have engineered a mouse model to specifically ablate Flcn in the liver to elucidate the potential key role of this factor in the control of hepatic metabolism. Interestingly, liver specific loss of Flcn markedly enhanced whole-body energy metabolism at the baseline level as shown by the increased energy expenditure as well as improved glucose and insulin tolerance in these mice under HFD. Remarkably, deletion of Flcn in mouse liver tissue reduced fibrosis and inflammation when fed the fibrosis inducing MCD diet. Overall, our results establish a critical role of Flcn in liver-fibrosis progression.
Our work has highlighted for the first time a protective role for Flcn loss in the liver. Similarly, Flcn deletion in other tissues such as adipocytes and muscles led to improved capabilities and whole-body metabolism homeostasis 21,22 . The protective phenotype observed in the liver is comparable to previous attempts to activate autophagy and inhibit mTORC1 with rapalogs and rapamycin 13,14 , but was not accompanied by the deleterious long-term effects of these compounds on liver damage and tumorigenesis 17,18 . Indeed, we did not observe any harmful effect of Flcn deletion in livers. The difference is possibly the result of the more specific regulation of mTORC1 by FLCN on the lysosomal surface that is not affecting the mRNA translation machinery 23,26 . On the other hand, rapalogs regulate all mTOR complexes in the cell, reinforcing the potential therapeutic benefit of targeting more precisely FLCN.
Under the fibrosis-inducing diet, which mimics fibrosis as observed with NASH, RNA-seq data revealed that major pathways affected by Flcn loss specifically in the liver include inflammation, fibrosis, and mitotic cell cycle gene expression. The reduction in cell cycle gene expression is possibly the consequence of reduced liver damage and therefore of the reduced necessity to regenerate liver cells. We have previously shown that loss of Flcn in C. elegans and mammalian cells increased innate immune response in a process dependent on TFEB and TFE3 32 . In this study, we observed a decrease in inflammation gene expression in liver-Flcn KO mice fed the MCD diet. It is possible that FLCN modulates the immune microenvironment and polarizes the macrophages toward an anti-inflammatory state. Indeed, it was reported that FLCN loss in hematopoietic stem cells induced CD11c + CD206 + activated phagocytic macrophages, corresponding to the anti-inflammatory M2 macrophage subtype 35 . However, another study found that bone marrow-derived macrophages from Flcn myeloid KO mice rather promotes spontaneous M1-type polarization and enhanced baseline activation status 33 . Also, RAW 264.7 macrophages targeted with shRNA for Flcn had increased phagocytic activity toward dying/dead cells 36 . The previous effects of Flcn loss on macrophage activation were all dependent on TFEB/TFE3 activation status, highlighting the important role of autophagy in FLCN's mechanism of regulation. Therefore, the importance and significance of the inflammation observed in Flcn KO livers remains to be investigated.
FLCN seemed to be involved in regulating autophagy in a severe liver disease setting and it is possibly the primary mechanism of action by which loss of Flcn reduces liver fibrosis. Several compounds known to activate autophagy pathways such as acetylshikonin and resveratrol also reduced MCD-induced fibrosis and inflammation 37,38 . Acetylshikonin contributed to the removal of cellular lipid droplets through lipophagy in a process dependent on the AMPK/mTOR pathways 37 . Indeed, acute inhibition of mTOR and activation of autophagy  www.nature.com/scientificreports/ with rapamycin in mice reduced liver injury and inflammation similarly to our study 39 . Moreover, thioredoxin interacting protein (TXNIP/VDUP1) mediates the activation of AMPK, inhibition of mTOR, and nuclear translocation of TFEB which ultimately resulted MCD diet-induced steatosis, inflammation, and fibrosis 40 .
In conclusion, we revealed that loss of Flcn in the liver is involved in autophagy, resulting in reduction of inflammation, fibrosis, and enhanced whole-body energy metabolism. These results show an unexpected role for Flcn in NAFLD and NASH progression. In light of the recently published cryo-EM structures of FLCN 41,42 , it is envisioned that pharmacological agents targeting the pocket of the FLCN GAP enzyme could be developed to achieve the loss of function phenotype in the liver shown in this study. Such molecules would represent new possibilities for treatment of obesity and NASH through the role of FLCN in hepatocyte homeostasis.
Protein extraction and immunoblotting. Liver tissues were quickly snap frozen using a BioSpec BioSqueezer (Fisher Scientific, cat# NC1033496) cooled in liquid nitrogen following sacrifice. Approximately 100 mg of tissue were solubilized in RIPA buffer using a micropestle and sonicated 10 s. The mixture was then centrifugated at 16,000 × g for 15 min at 4 °C and the supernatant collected. Proteins were separated on SDS-PAGE gels and revealed by western blot as previously described 22 using the antibodies listed above and IRDye Figure 5. Flcn loss activates autophagy in a mouse model of liver fibrosis. (A-B) Unsupervised hierarchical clustering following RNA-seq analysis in livers of mice fed either on chow or methionine/choline deficient diet (MCD) over a period of 6 weeks. (C) Relative quantitative real-time PCR analysis of autophagy-related genes mRNA transcript levels in livers of mice fed as described in (A) (mean ± SEM of the RNA fold change of indicated mRNAs; two-way ANOVA, *P < 0.05, **P < 0.01, ****P < 0.0001; n = 8 mice per condition). (D) Immunoblot of liver protein lysates extracted from mice fed as described in (A). Data are representative of 6 mice per condition. Full-length blots are presented in Supplementary Fig. 5. (E) Quantification of the relative amount of p62 immunoblot. Data normalized to tubulin levels (mean ± SEM of the relative fold change; two-way ANOVA, *P < 0.05, **P < 0.01, ****P < 0.0001; n = 6 mice per condition). (F) Relative RNAseq analysis of p62 mRNA transcript levels in livers of mice fed as described in (A) (mean ± SEM of p62 RNA fold change; one-way ANOVA, ***P < 0.001; n = 5 mice per condition). (G) Immunoblot of protein lysates of WT and Flcn KO MEFs incubated in presence of DMSO, BafA1 (100 nM), or Chloroquine (CQ; 100 μM) for 6 h. Data are representative of three independent experiments. Full-length blots are presented in Supplementary Fig. 5. (H) Quantification of the relative amount of LC3-II/LC3-I immunoblot of MEFs treated as described in (G) (mean ± SEM of three independent experiment, one-way ANOVA, *P < 0.05). (I) TFE3 Immunohistochemistry (IHC) staining in liver sections of mice fed as described in ( ACTA2  GAC TCA CAA CGT GCC TAT C  GCA GTA GTC ACG AAG GAA TAG   AKR1A1  CTG GAG TAT TTG GAC CTC TATTT AGC CTT CCA GGT CTC TTT A   ALDOA  GTG GTG TTG TGG GCA TTA  TCG GCT CCA TCC TTC TTA T   APOA4  GAC CAC GAT CAA GGA GAA TG  CCC TTG AGC TCT TCC ATA TTC   APOE  GGA ACA GAC CCA GCA AAT AC  CTG TAT CTT CTC CAT CAG GTTTG   ASAH1  ACC GGC CAA GAA GTG TCT TNFα  GGG TGT TCA TCC ATT CTC TAC  TGG ACC CTG AGC CAT AAT   UQCRC2  TTC GTT AAA GCA GGC AGT AG  CTT CAA TCC CAC GGG TTA TC   VLCAD  Immunohistochemistry. Liver tissues embedded in paraffin were cut to 4-μm sections on slides and stained for α-SMA, 4-HNE, F4/80, TFE3 and TFEB using routine immunohistochemical protocols provided by the GCRC Histology Core using Ventana BenchMark ULTRA system (Roche). Images were acquired using Aperio Scanscope XT (Leica Biosystems) and staining quantification was performed using image analysis algorithms in Aperio ImageScope software (version 12.3.3.5048, https:// www. leica biosy stems. com/ fr/ image riepatho logiq ue/ integ rer/ aperio-image scope/).

Mouse protein cytokine array.
Approximately 100 mg of liver tissue were solubilized in RIPA buffer using a micropestle and sonicated 10 s. The mixture was then centrifugated at 16,000 × g for 15 min at 4 °C and the supernatant collected. 31 mouse cytokine/chemokine biomarkers were simultaneously quantified by using a Discovery Assay called the Mouse Cytokine Array / Chemokine Array 31-Plex (Eve Technologies Corp, cat# MD31).

RNA-seq analysis.
Total RNA was isolated and purified from approximately 25 mg of frozen liver tissue using Total RNA Mini Kit (Geneaid, cat# RT100) according to the manufacturer's instructions. Library preparation and sequencing was made at the Institute for Research in Immunology and Cancer's Genomics Platform (IRIC). 500 ng of total RNA was used for library preparation. RNA quality control was assessed with the Bioanalyzer RNA 6000 Nano assay on the 2100 Bioanalyzer system (Agilent Technologies). Library preparation was done with the KAPA mRNAseq Hyperprep kit (KAPA, cat# KK8581). Ligation was made with Illumina dualindex UMI (IDT) and 10 PCR cycles was required to amplify cDNA libraries. Libraries were quantified by QuBit and BioAnalyzer DNA1000. All libraries were diluted to 10 nM and normalized by qPCR using the KAPA library quantification kit (KAPA, cat# KK4973). Libraries were pooled to equimolar concentration. Sequencing was performed with the Illumina Nextseq500 using the Nextseq High Output 75 (1 × 75 bp) cycles kit using 2.6 pM of the pooled libraries. Around 25 M single-end PF reads were generated per sample. Following data acquisition, adaptor sequences and low quality score bases (Phred score < 30) were first trimmed using Trimmomatic 43 . The resulting reads were aligned to the GRCm38 mouse reference genome assembly, using STAR 44 , and read counts were obtained using HTSeq 45 . For all downstream analyses, we excluded lowly-expressed genes with an average read count lower than 10 across all samples, resulting in 14,057 expressed genes in total. The R package limma 46 was used to identify differences in gene expression levels between the different conditions. Nominal p-values were corrected for multiple testing using the Benjamini-Hochberg method. To assess the effect liver-Flcn KO in the response to MCD-diet, we first obtained differentially expressed genes (FDR < 0.05 and |log2FC|> 1) in Cre. MCD_diet  www.nature.com/scientificreports/ then filtered for those that show |difference in log2FC|> 1 (differentially responsive (DR) genes). Unsupervised hierarchical clustering of the 1,110 DR genes shows three distinct patterns of changes in expression. Pathway enrichment analyses were performed using Enrichr 47 .

Data availability
RNA-sequencing data has been deposited in the Gene Expression Omnibus under the accession GSE156918.