Deletion of the nuclear receptor RORα in macrophages does not modify the development of obesity, insulin resistance and NASH

Retinoic acid receptor-related orphan receptor-alpha (RORα) is a transcription factor from the nuclear receptor family expressed by immune cells and involved in the development of obesity, insulin resistance (IR) and non-alcoholic steatohepatitis (NASH). It was recently reported that mice deficient for RORα in macrophages develop more severe NASH upon high fat diet (HFD) feeding due to altered Kupffer cell function. To better understand the role of RORα in obesity and IR, we independently generated a macrophage RORα-deficient mouse line. We report that RORα deletion in macrophages does not impact on HFD-induced obesity and IR. Surprisingly, we did not confirm an effect on NASH development upon HFD feeding nor in the more severe and obesity-independent choline-deficient, L-amino acid-defined diet model. Our results therefore show that RORα deletion in macrophages does not alter the development of obesity and IR and question its role in NASH.

During the last few decades, the prevalence of obesity has dramatically increased worldwide, arising from excessive dietary intake and a sedentary lifestyle 1 . Obesity is a strong risk factor for type 2 diabetes (T2D) and nonalcoholic fatty liver disease (NAFLD). T2D is characterized by hyperglycemia resulting from insulin resistance (IR) and relative insulin deficiency 2 . The chronic and low-grade systemic inflammation developing during obesity is considered as one critical step in the pathogenesis of IR and represents a valuable therapeutic target 3 . In obesity, enlarged adipocytes promote the recruitment and activation of immune cells leading to the accumulation of monocytes, pro-inflammatory macrophages, neutrophils, T cells and B cells in adipose tissue (AT) 4 . AT macrophages (ATM) are the most abundant leukocytes in obese AT and among the most important immune cell types mediating inflammation and IR 5,6 . ATM secrete several pro-inflammatory cytokines such as IL-1β, TNF-α and IL-6, all interfering with the insulin signaling pathway 7 . Expansion of ATM mainly results from tissue recruitment of circulating monocytes followed by their differentiation into pro-inflammatory macrophages 8 .
NAFLD comprises a liver phenotype spectrum ranging from simple steatosis, also called non-alcoholic fatty liver (NAFL), to non-alcoholic steatohepatitis (NASH). NAFLD diagnosis is based on histology with steatosis, characterized by triglyceride accumulation in hepatocytes, being the first stage 9 . NAFL progresses into NASH with the appearance of lobular inflammation and hepatocyte ballooning. NASH predisposes to fibrosis and may ultimately evolve into cirrhosis and hepatocellular carcinoma 10 . NAFLD is strongly associated with obesity, IR and the capacity of AT to store lipids 11 . The AT-liver crosstalk is essential in NAFLD pathophysiology. Indeed, development of IR in AT increases free fatty acids (FFAs) release, leading to FFAs uptake and esterification into triglycerides in the liver. Excessive lipid accumulation within hepatocytes induces lipotoxicity and the release of damage-associated molecular patterns (DAMPs) that ultimately activate immune cells and favor leukocyte recruitment. Immune cells, and particularly the resident liver macrophages (Kupffer cells, KCs), play key roles in NAFLD by regulating hepatocyte metabolism, inflammation and fibrosis 12,13 .

Results
Characterization of RORα MKO mice. We generated Rora floxed mice by introducing loxP sites flanking exon 3 (Fig. S1A). Exon 3, the first common exon of isoforms 1 and 4, encodes for the DBD and includes the first zinc finger motif. Deletion of exon 3 leads to a frameshift resulting in a premature stop codon and the absence of functional domains (Fig. S1B,C). To achieve RORα deletion in macrophages, these mice were subsequently crossed with the LysM-Cre mice. Rora fl/fl Lyz2 Cre/+ (MKO) mice and their littermate controls Rora +/+ Lyz2 Cre/+ (WT), exhibiting comparable expression of Lyz2 and Cre (Fig. 1A), were further studied.
To validate and characterize the RORα deletion in MKO mice, we demonstrated the excision of exon 3 in DNA from sorted splenic macrophages (Fig. 1B). Since the LysM-Cre mouse model mainly targets macrophages and neutrophils, but also a fraction of monocytes and DCs 30 , we sorted these populations from spleen and quantified the deletion of exon 3 in Rora mRNA by RT-qPCR (Fig. 1C). MKO mice displayed an important deletion of exon 3 in macrophages and to a lesser extent in monocytes and DCs, but surprisingly not in neutrophils. No deletion of exon 3 in Rora mRNA was observed in T cells and B cells. To confirm the absence of deletion in neutrophils, we also analyzed the bone marrow (BM) cells which mainly contain neutrophils. Neither total BM cells nor purified neutrophils from BM showed a deletion of exon 3, while bone marrow-derived macrophages (BMDM) did (Fig. 1D). Collectively, our results show that macrophages, monocytes and DCs, but not neutrophils, exhibit a LyzM-Cre-induced deletion of Rora exon 3.
Rorc expression was not affected by Rora deletion (Fig. 1E) and Rorb was undetectable in both WT and MKO immune cells (data not shown), suggesting no compensatory regulation of other ROR genes. The expression of a key ROR target gene Arntl, encoding for BMAL1, was significantly lower in RORα-deficient splenic macrophages and BMDM (Fig. 1F,G). A tendency to a lower Arntl expression was also observed in monocytes, but not in DCs, probably due to the high expression of Rorc (Fig. 1E). Thus, despite the significant LysM-Cre-induced deletion of exon 3 in monocytes and DCs, MKO mice foremost displayed a phenotype in macrophages. As macrophages display some tissue-specific characteristics, we sorted ATM (CD45 + CD3ε − TCRβ − CD19 − CD20 − F4/80 + CD64 + ) from epididymal AT to evaluate the efficiency of Rora deletion in these cells. Similar to the splenic macrophages and BMDM, MKO mice showed a significant deletion of exon 3 and a significant decrease of the ROR target gene Arntl in ATM (Fig. 1H). No significant deletion of exon 3 occurred in total epididymal AT, liver nor skeletal muscle (Fig. 1I), suggesting the absence of an unwanted recombination in parenchymal cells of metabolic tissues.
RORα deletion in macrophages does not affect obesity, IR nor steatosis. To investigate the role of macrophages RORα in metabolic functions, in particular obesity and IR, we fed WT and MKO mice with a 60% high fat diet (HFD) for 12 weeks. WT and MKO mice under HFD significantly gained weight, AT mass and increased leptin expression similarly in both genotypes ( Fig. 2A-C). Fasting serum glucose and insulin concentrations increased significantly upon HFD feeding although no difference between genotypes was observed (Fig. 2D,E). Furthermore, we observed no difference in glucose levels between WT and MKO mice during IPGTT and IPITT (Fig. 2F,G). Moreover, insulin signaling in skeletal muscle, epididymal AT and liver was similar in both genotypes ( Fig. 2H and S2A). AT inflammation is a central driver for obesity-induced IR 4 . No difference in Tnf, Il1b, Il6, and Il10 expression was observed in epididymal AT between WT and MKO mice (Fig. 2I). Likewise, plasma cholesterol and triglyceride levels were not different between WT and MKO mice (Fig. S2B,C). Taken together, our results show that RORα deletion in macrophages has no impact on HFD-induced obesity and IR. www.nature.com/scientificreports/ www.nature.com/scientificreports/ In interconnection with obesity and IR, HFD feeding also induces liver steatosis. Liver weight, steatosis level determined by histology and hepatic triglyceride content were increased upon HFD feeding to the same extent in WT and MKO mice ( Fig. 3A-E). Histology analysis did not reveal cell infiltrates upon HFD feeding despite the significantly increased expression of inflammatory genes such as Tnf, Il1b, Il6 and Il10 (Fig. 3F). Likewise, we did not observe any significant differences in hepatic cytokine expression between WT and MKO mice besides a tendency to a lower Il10 expression in MKO mice (p = 0.07). HFD feeding induced a similar mild increase of plasma transaminase activity in both WT and MKO mice (Fig. 3G). Liver fibrosis and expression of collagen genes slightly increased upon HFD feeding for 12 weeks at comparable level between WT and MKO mice ( Fig. S3A-C). www.nature.com/scientificreports/ RORα deletion in macrophages does not affect NASH. While HFD feeding efficiently promotes obesity and IR, the liver pathology is mainly characterized by triglyceride accumulation corresponding to the NAFL stage 9 . To investigate whether the deletion of RORα in macrophages might play a role in advanced NAFLD, WT and MKO mice were fed with the choline-deficient, L-amino acid-defined (CDAA) diet supplemented with sucrose and 2% cholesterol for 8 weeks, leading to a pronounced hepatic steatosis, inflammation and fibrosis, but dissociated from obesity and IR allowing the analysis of liver-centric responses 9 . As expected and unlike HFD, the CDAA diet did not induce obesity, hyperglycemia nor AT inflammation, but led to a 10% weight loss and a drop of glycemia of 30% in both WT and MKO ( Fig. S4A-C). The CDAA diet produced a massive hepatic steatosis, but RORα deletion in macrophages did not impact on liver weight nor steatosis level ( Fig. 4A-E). Histological analysis showed evidences of cell infiltrates upon CDAA diet feeding but no manifest difference between WT and MKO mice was observed ( Fig. 4C & S4D). Further analysis of the expression of inflammatory genes revealed no difference in Tnf and Il6 expression whereas Il1b and Il10 mRNA significantly decreased in livers of MKO mice (Fig. 4F). Despite this modest decrease in cytokine expression, we did not observe any significant differences in plasma transaminase activity nor fibrosis, while a strong effect of the CDAA diet was found on these parameters (

Effect of RORα deletion in KC is offset by high RORγ expression.
Contrasting with these results in 2 independent and complementary models of NAFLD, it was reported that RORα deletion in macrophages, achieved by using a similar strategy (Rora fl/fl Lyz2 Cre ), predisposes mice to NASH after 12 weeks of HFD feeding 31 . NASH exacerbation was attributed to a key role of RORα in Kupffer Cell (KC) function 31 . To determine a possible cause for this discrepancy, we first assessed the extent of Rora exon 3 deletion in KC from chow diet fed mice. KC and other liver immune cell populations, sorted to a high level of purity ( Fig. S5A-G), revealed efficient and specific deletion of Rora exon 3 in KC (Fig. 5A). Interestingly, and similarly to publicly available transcriptomic data (Fig. S5H), we observed that KC expressed lower levels of Rora than other macrophage populations, with lung macrophages expressing the highest level (Fig. 5B). www.nature.com/scientificreports/ Despite a significant Rora exon 3 deletion, expression of the ROR target gene Arntl did not decrease in KC, unlike lung macrophages (Fig. 5C), splenic macrophages (Fig. 1F), BMDM (Fig. 1G) and ATM (Fig. 1H). The lack of effect of RORα deletion on Arntl expression in KC suggests a compensatory mechanism by another ROR protein. Rorb was not expressed in KC, similarly to other macrophage populations tested (data not shown), but Rorc expression was detected (Fig. 5D). No compensation of Rorc expression was found in any macrophage subsets from MKO mice compared to littermate controls. However, basal Rorc expression was between 5-and www.nature.com/scientificreports/ tenfold higher in KC compared to other macrophage subsets. The low Rora expression and the high Rorc expression in KC led to a Rorc/Rora ratio 10-to 50-fold higher than in other macrophage populations (Fig. 5E). Collectively, these results suggest that RORα deficiency is unlikely to induce major transcriptional effect in KC due to a concomitant high RORγ expression.

Discussion
Over the past years, a considerable number of studies contributed to our understanding about the role of nuclear receptors, including RORs, in the regulation of lipid and glucose metabolism, circadian rhythm as well as the development and function of the immune system. Due to the pleiotropic effects and wide cell distribution of RORα and considering the key role of macrophages in metabolic diseases, we investigated the impact of RORα deletion in these cells on obesity, IR and NASH by using the LysM-Cre mice. Evaluation of deletion efficiency and decreased expression of a major RORα target gene in several myeloid subsets showed that macrophages were mostly impacted by RORα deletion in basal conditions although we cannot formally exclude that other myeloid subsets are also affected upon metabolic challenge. Yet, importantly, we found no impact of LysM-Cre-mediated RORα deletion neither on HFD-induced obesity, IR and steatosis nor on CDAA diet-induced NASH. These findings contrast with an earlier report showing that LysM-Cre-mediated RORα deletion increases the susceptibility to HFD-induced NASH 31 , while both studies used the same HFD reference and duration of treatment. As feeding a HFD for 12 weeks is not per se leading to NASH, but rather to NAFL 9 , only a more pronounced steatosis (higher liver weight and triglyceride content) was reported in MKO mice in the earlier study 31 . Indicators of increased hepatic lipotoxicity such as higher plasma transaminase activity, increased expression of pro-inflammatory cytokines (Tnf, Il1b and Il6) and decreased expression of Il10 in liver were also observed 31 . It was proposed that RORα directly regulates M2 polarization of KCs leading to increased IL-10 production which would protect hepatocytes against lipid accumulation 31 . While we also observed a tendency to lower Il10 expression in livers from MKO mice upon HFD feeding (Fig. 3F) and a significant decrease upon CDAA diet feeding (Fig. 4F), we did not observe any exacerbation of steatosis or NASH. As the putative protective role for IL-10 was only based on in vitro findings, we believe that the reduced www.nature.com/scientificreports/ Il10 expression in liver of MKO mice upon HFD is unlikely to contribute to aggravated steatosis. Further supporting this finding, IL-10-deficient mice fed a HFD for 12 weeks developed increased liver inflammation, but decreased steatosis and transaminase activity 32 , suggesting that beyond its well described anti-inflammatory function, endogenous IL-10 is not protective against HFD-induced steatosis. Several factors might account for the discrepancy between our results and these earlier findings 31 . The microbiological status of the animal facility and housing conditions are two underestimated factors that might account for major differences in experimental outcome and it is not uncommon for this information to be missing. Specific pathogen free (SPF) status indicates that mice are free of defined pathogens including the mouse hepatitis virus and Helicobacter hepaticus which may interfere with liver function. Gut microbiota also impacts on NAFLD development 33 . The microbiota composition is highly variable between animal facilities, mouse lines and even between cages, especially when two genotypes are bred separately. Generation of littermate animals is critical to insure that both deficient and control mice harbor identical gut microbiota while post-weaning cohousing is ineffective 34 . Our mice were housed in SPF facility and bred to obtain WT and MKO littermate animals. No information was provided in the earlier study about the animal facility and littermate status.
Two additional factors that may also account for the observed discrepancies between studies are the floxed mice and the breeding strategy used. We generated mice with floxed Rora exon 3 while exon 4 was targeted in the earlier study 31 . Deletion of exon 4 may still result in the translation of a truncated protein containing a portion of the DBD that includes the exon 3-encoded zinc finger motif (Fig. S1C). This difference in deletion strategy is however unlikely to account for the observed phenotypic differences. The selected control animals represent a second major difference between the studies. The LysM-Cre mice carry an insertion of Cre recombinase into the Lyz2 gene, leading to Cre expression under the control of the Lyz2 promoter and enhancers, but abolishing endogenous Lyz2 expression. While we intentionally maintained similar Cre and Lyz2 expression between WT and MKO by using only hemizygous animals for this locus (comparing Rora +/+ Lyz2 Cre/+ with Rora fl/fl Lyz2 Cre/+ ), floxed mice (Rora fl/fl Lyz2 +/+ ) were used as WT control and compared with MKO mice missing information about the Lyz2 locus (Rora fl/fl Lyz2 Cre/? ) in the earlier study 31 . This latter strategy results in differential Cre and Lyz2 expression between the WT, expressing no Cre, and the MKO mice expressing less or no Lyz2. Even though mammalian genomes possess no loxP sites, active Cre recombinase recognition sites, called pseudo loxP sites, have been reported 35 and Cre expression mediates DNA damage, cell toxicity and apoptosis even in the absence of floxed alleles 36 . In addition to RORα deletion and exogenous Cre expression in macrophages, MKO mice generated with this strategy possess also at least one inactivated Lyz2 allele and possibly a homozygous (whole body) Lyz2 deletion. Indeed, based on the described genotyping strategy using a single PCR reaction to only analyze the Cre gene 31 , it is highly likely that MKO mice were actually also Lyz2 deficient.
Lysozyme is an antimicrobial protein that catalyzes the hydrolysis of peptidoglycan between N-acetylglucosamine and N-acetylmuramic acid, contributing to the degradation of the Gram-positive bacterial cell wall. In humans, lysozyme is encoded by a single gene LYZ, whereas two genes exist in mouse, Lyz1 and Lyz2 encoding for lysozyme P and lysozyme M respectively. Lyz2 is the most expressed lysozyme gene in mice with a high expression in the myeloid lineage, similar to the human LYZ gene. Expectedly, Lyz2-deficient mice develop more severe bacterial infections [37][38][39] . In addition to its antimicrobial properties, lysozyme is also an anti-inflammatory factor. Lysozyme inhibits serum complement activation 40 and possesses a LPS-binding ability that reduces LPS-related inflammation 41 . Peptides derived from lysozyme cleavage inhibit production of pro-inflammatory cytokines, including TNF-α and IL-1β, by macrophages 42 . Moreover, lysozyme improves the antioxidant capacity of hepatocytes in vitro and in vivo, leading to a protection against oxidative stress in liver 43 . Lysozyme expression and activity is affected in various liver diseases [44][45][46] , but, to our knowledge, the role of Lyz2 in NASH was never investigated. These roles in infection, inflammation and probably in liver homeostasis, suggest that Lyz2 may play a role in NASH development. Thus, caution should be exerted when comparing mouse lines with different Lyz2 copy numbers.
By using a 12-week HFD that induces liver steatosis (NAFL) or an 8-week CDAA diet that leads to NASH, we observed no differences between WT and MKO mice for any tested hepatic parameters including liver weight, triglyceride content, inflammatory and fibrosis gene expression, histology and transaminase activity. We also found no impact on obesity and glucose homeostasis. Investigations on KC showed that, despite an effective deletion of RORα, KCs were the only macrophage population without transcriptional effect on the ROR target gene Arntl. We found no compensatory expression of other ROR genes in MKO, but we evidenced a high level of Rorc expression specifically in KC that might offset the RORα deletion. The redundancy between ROR proteins is well established 17,[47][48][49] . Hepatocytes express high levels of RORγ in addition to RORα 48 . It was shown that a single deletion of RORα or RORγ in hepatocytes affects only 2 and 6 transcripts respectively in whole liver transcriptome, while RORα/γ deletion leads to a broader effect with 299 genes affected, unambiguously demonstrating the redundancy between RORα and RORγ in hepatocytes 49 . Such a redundancy might be at play in KCs and further investigation with a double RORα/γ deficient mouse model might address this question. Moreover, usage of the new KC-specific Clec4f-Cre mouse line would improve the specificity towards KC and may result in a better deletion efficiency than the one reached with the LysM-cre mice 50 .
In conclusion RORα deletion in macrophages using the LyzM-Cre system has no impact on the development of obesity, IR and NASH. We suggest that the previously reported impact of RORα deletion in macrophages on NASH 31 likely does not result from a specific effect of RORα deletion, but rather from a different Cre or Lyz2 copy number between WT and MKO.

Methods
Generation of RORα MKO Mice. We generated mice harboring a floxed allele of Rora (Rora fl/fl ) by flanking exon 3 with two loxP sequences. The strategy and outcome on RORα1 and RORα4 are summarized in figures S1A and S1B respectively. Briefly, a targeting vector containing loxP sites, a FRT-floxed neomycin cassette and homologous regions surrounding exon 3 was constructed and transfected in embryonic stem (ES) cells derived from 129/Sv. After screening for homologous recombination by southern blot and PCR, ES cells were injected into C57BL/6 blastocysts. Neomycin cassette was removed in vivo by using FLP deleter mice in C57BL/6 background. Finally, mice were backcrossed with C57BL/6 J mice for at least six generations. To achieve the deletion of RORα in macrophages, Rora fl/fl mice were crossed with LysM-Cre transgenic mice (Jackson laboratory) in which Cre recombinase is expressed under the control of endogenous Lyz2 promoter. Littermates RORα-deficient (MKO, Rora fl/fl Lyz2 Cre/+ ) and WT (Rora +/+ Lyz2 Cre/+ ) mice were generated by crossing Rora fl/+ Lyz2 Cre/Cre with Rora fl/+ Lyz2 +/+ mice. All mice were genotyped twice.
Mouse studies. Mice were kept on a 12-h light/dark cycle in the SPF animal facility from the Institut Pasteur de Lille with ad libitum access to food and water. Littermate WT and MKO mice were maintained all along the experiment procedures in 904 cm 2 cages (Green line GR900, Tecniplast) with 6-12 mice per cage and a ratio WT:MKO tending to 1:1. Only male mice were used for the experiments. All animal procedures were approved by the ethical committee for animal experimentation of the Nord-Pas-de-Calais Region (CEEA75) (APAFIS#7160-2017040313471173) in accordance with European guidelines on the protection of animals used for scientific purposes (2010/63/UE).
Ten-week-old RORα WT and their littermates RORα MKO mice were fed with a 60% high-fat diet (HFD, Research Diet, D12492) for 12 weeks or maintained under chow diet (SAFE, #A04). To induce a NASH-like disease, ten-week-old RORα WT and MKO mice were fed with a choline-deficient, L-amino acid-defined (CDAA) diet with 35% sucrose, 21% fat and 2% cholesterol (Ssniff, custom diet) for 8 weeks. In addition to CDAA diet, mice also received monosaccharides in the drinking water (42 g/L, fructose:glucose ratio of 55:45). Weight of mice was measured weekly. Before sacrifice, mice were fasted for 5 h. Mice were sacrificed at ZT3 (10 am) for HFD and at ZT7 (2 pm) for CDAA experiment.
Mouse genotyping. DNA was extracted from tail with REDExtract-N-Amp Tissue PCR Kit (Sigma, #XNAT-1000RXN). Floxed Rora was detected by PCR with the Rora genotyping primers (Supplementary Table 1) and the following cycling conditions: 1 cycle at 94 °C for 3 min; 35 cycles at 94 °C for 30 s, 55 °C for 30 s, 72 °C for 1 min; and 1 cycle at 72 °C for 10 min; hold at 4 °C. Samples were separated by gel electrophoresis on a 1.5% agarose gel. Rora +/+ gave a single band at 250 bp, Rora fl/fl at 340 bp and Rora fl/+ had both bands. Endogenous Lyz2 and Cre were detected with the Lyz2 genotyping and Cre genotyping primers respectively (Supplementary Table 1) and the following cycling conditions: 1 cycle at 94 °C for 3 min; 35 cycles at 94 °C for 1 min, 63 °C for 1 min, 72 °C for 90 s; and 1 cycle at 72 °C for 10 min; hold at 4 °C. Samples were separated by gel electrophoresis on a 1.5% agarose gel. Lyz2 +/+ gave a single band at 350 bp for Lyz2 PCR while Lyz2 Cre/Cre gave a single band at 700 bp for Cre PCR. Lyz2 Cre/+ gave bands for both Lyz2 and Cre PCR. Agarose gels were acquired with a Gel Doc XR system (Bio-Rad) and the Image Lab software verion 2.0 build 8 for PC (Bio-Rad, https ://www.bio-rad.com/ fr-fr/produ ct/image -lab-softw are?ID=KRE6P 5E8Z). The following acquisition settings were chosen: application: SYBR Safe; Image exposure: automatically optimized.
Bone marrow-derived macrophages. Bone marrow was isolated from tibia and femur of mice. Bone marrow cells were culture in RPMI 1640 with Hepes and L-glutamine supplemented with 10% fetal bovine serum, 20% L929-conditioned medium and 25 µg/mL gentamycin for one week. After three days of culture, fresh medium was added. After one week of culture, supernatant was discarded and adherent bone marrowderived macrophages (BMDMs) were washed two times with PBS. BMDMs were collected by using a cell scraper, counted and plated at a concentration of 10 6 cells/ml. After 24 h, BMDMs were treated with 100 nM dexamethasone (Sigma, #D1756) for 2 h to synchronize cells, washed and maintained in complete medium for 32 h to reach the pic of RORα activity corresponding to ZT0 in vivo.
Insulin and glucose tolerance tests. After 10 weeks of HFD, a Glucose Tolerance Test (GTT) was performed by intraperitoneal injection of glucose (1 g/kg). Tail blood sample was collected before glucose injection to measure fasting insulin by ELISA (Mercodia #10-1247-10) according to manufacturer instruction. After 11 weeks of HFD, an Insulin Tolerance Test (ITT) was performed by intraperitoneal injection of human insulin (1 IU/kg) (Actrapid, Novo Nordisk). Glycemia was measured from tail before and 15, 30, 60, 90 and 120 min after glucose or insulin injection by using a glucose meter (Accu-Check performa, Roche). Before GTT and ITT, mice were fasted for 5 h at ZT2 (9 am) and the tests were performed at ZT7 (2 pm).

Metabolic parameters.
Before sacrifice and after 5 h of fasting, blood samples were collected from the retro orbital sinus of mice. Plasma alanine aminotransferase (ALAT), aspartate aminotransferase (ASAT), total cholesterol and triglycerides were measured on a Konelab 20 (Thermo Fisher) with reagents from Thermo Scientific for ALAT (#981769) and ASAT (#981771) and reagents from DiaSys for cholesterol (#113009910026) and triglycerides (#157109910026). The plasma ALAT and ASAT measurement under CDAA diet was performed without fasting for the time point T0, 2, 4 and 6 weeks.
Measurement of liver triglycerides. Lipids were extracted from the liver caudate lobe. A weighted piece of tissue was homogenized with T10 Ultra-Turrax (Ika) in PBS 1% Triton. Samples were transferred into glass tubes and mixed with a 2:1 chlorofrom:methanol mixture. After centrifugation, upper-and inter-phase were discarded. The lower organic phase was evaporated under nitrogen flow and reconstituted in 1% Triton X100. Triglyceride content was measured with Triglycerides FS kit (DiaSys, #157109910026).

Statistical analyses.
All statistical analyses were carried out using GraphPad Prism 8 for Windows (Graph-Pad Software) and presented as means ± SEM. The study was done blinded for genotype. Only one mouse (WT) was excluded from the analysis because it did not reach an appropriate weight gain under HFD feeding (Body weight: 31.2 g; ingAT: 0.165 g; epiAT: 0.348 g). No mice were excluded in CDAA diet group. Data were analyzed with 2-way ANOVA and Sidak's multiple comparisons post-hoc test or Student's t-test. Values with P < 0.05 were considered as significant. All statistical details including statistical test used and exact value of n are described in each figure legend. Each data point represents genuine replication, also called true replicate, and were obtained from a single measurement or multiple measurements illustrated by the mean, such as gene expression by RT-qPCR that was assessed in duplicate or histology quantification made on five random fields per mice. For sake of clarity, results of statistical comparisons between WT and MKO mice upon Chow feeding were not illustrated on the figures but were not significant for all the parameters. In Figs. 2, 3, and 4, the diet effect was analyzed by 2-way ANOVA and was significant for all the parameters with the exception of liver weight (Fig. 3B) that was not statistically significant between chow and HFD (p = 0.06). The datasets used in this study are available from GEO under the accession number GSE104342 51 and GSE56682 52 and from data assembled by the ImmGen consortium (http://www.immge n.org/) 53 .

Data availability
Further information and requests for resources and reagents should be directed to and will be fulfilled by David Dombrowicz (david.dombrowicz@inserm.fr).