ONLINE METHODS plates. For lipoprotein profiling, we used 200 ul plasma, which was separated Experimental mice and diets. All mouse experiments were approved by the by fast-performance liquid chromatography(FPLC)on a Superose 6 10/300 nimal Welfare Officers of University Medical Center Hamburg-Eppendorf GL column(GE Healthcare)with a flow rate of 0.5 ml/ min. Forty fractions (UKE)and Behorde fur Gesundheit und Verbraucherschutz Hamburg. Mice (0.5 ml each) were collected. Triglyceride and cholesterol concentrations we re purchased from Charles River Laboratories or bred in-house at UKE measured in each fraction (Hamburg)and UTSw(Dallas), respectively. Mice were housed at 24C with a day and night cycle of 12 h each. For the experiments, we used age-matched Cholesterol uptake studies. To analyze cholesterol uptake, we housed mice at (12- to 14-week old)male C57BL/6] WT mice, as well as dbldb, Mdr2-l-, thermoneutrality or 6C and fed them the cholesterol-enriched HFD described LdIr--, Ldr/- mice carrying floxed alleles of Lrpl(ref. 47)and Cyp7b1-- above 4 h before organ and blood collection, we orally administered a 200-ul (ref. 19)mice. To delete hepatic LRPl expression, we crossed floxed mice with lipid emulsion(10 mg human triglyceride-rich lipoprotein-derived lipid pe ice expressing Cre recombinase under the control of the albumin promoter ml saline)enriched with a tracer dose of C-cholesterol(185 kBq/mouse).For (Alb). We allocated mice to the groups on the basis of their body weights. For measurement of radioactivity by B-scintillation counting, we solubilized organs practical reasons, investigators were not blinded as to the group allocation. If in Solvable(PerkinElmer, 100 ul per 10 mg organ). Cholesterol uptake was not stated otherwise, mice were housed in single cages and fed a cholesterol- calculated as a percentage of the injected dose. enriched HFD containing 21% total fat and 0.2% cholesterol (Ssniff EF R/M acc. TD88137 mod )for a total period of 10 d. In one experimental setup, mice Bile acid and hydroxysterol measurements. We performed targeted bile received either a normal chow diet or a chow diet enriched with cholesterol acid and hydroxysterol analysis by HPLC coupled to electrospray ionia- (0.2%)for 10 d. Animals had ad libitum access to food and water. After at least tion tandem mass spectrometry as recently described5. In brief, sample d of adaptation to the respective diet, the mice were housed at a thermon- preparation comprised a simple methanol liquid-liquid extraction of tissues a utral temperature(30C), 22C, 16Cor 6C for 7 d in a humidity-controlled and biofluids. Quantitative measurement of bile acids was performed using g climate chamber before their organs and fecal samples were harvested For a HPLC-ESI-QqQ system in multiple reaction monitoring(MRM) mode 45 measurements of body temperature, we implemented a stepwise reduction in (HPLC: 1200 Infinity Quaternary LC System (Agilent): column: Accucore istration of the agonist Phenex 20606(Phenex Pharmaceuticals AG; 40 mg/ml Scienctific, Inc ) QqQ: API 4000 Q trap(ABSCIEX)). Peak identifica- 9 and 10 Control mice received an oral gavage of the vehicle For B3-adrener- well as MRM transitions and peak areas, respectively, to corresponding gic receptor stimulation we supplemented the cholesterol-enriched HFD with standard chromatograms. CL316243(0.001%wt/wt: Tocris). To block intestinal cholesterol absorption z treated mice with an ezetimibe-supplemented diet(10 mg ezetimibe per 167 mg Gene expression analysis. We isolated and purified total RNA from mouse E Sigma-Aldrich) to the drinking water. We anesthetized mice after a 4-h fasting ing human liver cDNA samples that were published previously 2Quantitative c) period and transcardially withdrew blood with syringes that were pre-treated real-time PCR reactions for the indicated genes were conducted on a 7900HT with EDTA(Sigma-Aldrich) to isolate plasma. Subsequently, we perfused sequence detection system (Applied Biosystems) using Taq ManAssay-ou animals with PBS and harvested the tissues, which were immediately conserved Demand primer sets(Applied Biosystems, Cyp 7al: Mm00484150_ml, Cyp8b1: either in TRIzol reagent(Invitrogen)or stored at-80C for further processing. Mm00501637-sl, Cyp7b1: Mm00484157-ml, Cyp27al: Mm00470430 al and fec Baat: Mm00476075 ml Abcbl I: Mm00445168 m1 Nr0b2 Mm00442278 ml Sc0a2:Mm00488258m1,Uepl:Mm00494069m1,Dio2:Mm00515664m1, microbiome analysis. We performed sequencing of the 16S-rRNA-encoding Elovl3: Mm00468164_ml, Ppargcla(Pgcla): Mmo0447183_ml, Prdm16 gene of fecal samples as described. We performed m using Mm00712556-ml, Slc1Oal: Mm00441421_ml, Collal: Mm00801666 R. To minimize the effect of compositional outliers, multivariate outliers were Cxcl10: Mm00445235-_ml, Emr1: Mm00802530-_ml, Tgfbl: Mm00441724_ml detectedandremovedusingthemvoutlier'package(https://cran.r-project.org/Inf:Mm00443258_ml,Tbp:Mm00446973_ml,Cyp7A1:hs00167982_ml web/packages/ mvoutlier/index. html)based ta, including all CYP7B1: Hso0191385_ml, CYP8B1: Hs00244754sl and TAF1: Hso0270322 enera with a mean relative abundance of greater than 0.5%6. Alpha diversity ml). Cycle thresholds(C values)were normalized to those of the housekee z measures, including observed number, Chaol estimator and Shannon index, genes TATA-box binding protein(Tbp)or TATA-box binding protein were calculated using the diversity and 'estimateRO' functions in the 'vegan factor I(TAFI). o package for r based on genus level and operational taxonomic unit (oTU)level abundance tables. We used pairwise Wilcoxon rank-sum tests to determine Study subjects. The study design, subjects and sample collection have been significant differences between treatment groups. The Bray-Curtis dissimilarity described in detail. In brief, liver samples were obtained from subjects who vas calculated on the OTU and genus level abundance tables using the vegdisto were undergoing bariatric surgery(obese)or abdominal surgery for early function from the vegan package. Permutationals MANOVA was used to detect ectal cancer or for benign diseases such as sigma-diverticulosis(non-obese)at differences in beta diversity between the treatment groups using the ' adonis the Department of General and visceral Surgery, University of UIm, German function and 10,000 permutations. Identification of differentially abundant gen- Individuals undergoing bariatric surgery had not been subjected to a weight- era-and species-level OTUs was done based on a core microbiota, including loss regimen before surgery, but they were commended to adhere to a regular with a mean relative abund in at least one of the treatment meal pattern to minimize the confounding effects of negative weight balance. roups. The'mvabund package was used for fitting generalized linear models We selected gender-matched subgroups including subjects who were non- with 10,000 permutations to assess which taxa deviated significantly between (obese) and subjects who were obese and had type 2 diabetes(obese T2D) groups. We adjusted P values for multiple-testing by applying the Benjamini- from the total cohort. Permission was given under the number Hochberg correction 50. In box plots, center lines represent medians, limits are the local ethics committee in Ulm, Germany, and informed consent in writing first and third quartiles (inter-quartile range: IQR), whiskers represent the lowest was obtained from each subject. and highest values within a 1.5 x IQR range the first/third quartile, and points are values outside the 1.5 x IQR range t the first/third quartile. Plasma activity of alanine aminotransferase and aspartate aminotransferase. For assessment of liver damage, we determined plasma activities of alanine Plasma parameters. We quantified plasma cholesterol and triglyceride lev. notransferase(ALT) and aspartate aminotransferase(AST)using the COBas els using commercial kits(Roche) that were adapted to 96-well microtiter Mira System(Roche) NATURE MEDICINE doi:10.1038/m.43© 2017 Nature America, Inc., part of Springer Nature. All rights reserved. nature medicine doi:10.1038/nm.4357 ONLINE METHODS Experimental mice and diets. All mouse experiments were approved by the Animal Welfare Officers of University Medical Center Hamburg-Eppendorf (UKE) and Behörde für Gesundheit und Verbraucherschutz Hamburg. Mice were purchased from Charles River Laboratories or bred in-house at UKE (Hamburg) and UTSW (Dallas), respectively. Mice were housed at 24 °C with a day and night cycle of 12 h each. For the experiments, we used age-matched (12- to 14-week old) male C57BL/6J WT mice, as well as db/db, Mdr2−/−, Ldlr−/−, Ldlr−/− mice carrying floxed alleles of Lrp1 (ref. 47) and Cyp7b1−/− (ref. 19) mice. To delete hepatic LRP1 expression, we crossed floxed mice with mice expressing Cre recombinase under the control of the albumin promoter (Alb). We allocated mice to the groups on the basis of their body weights. For practical reasons, investigators were not blinded as to the group allocation. If not stated otherwise, mice were housed in single cages and fed a cholesterol￾enriched HFD containing 21% total fat and 0.2% cholesterol (Ssniff EF R/M acc. TD88137 mod.) for a total period of 10 d. In one experimental setup, mice received either a normal chow diet or a chow diet enriched with cholesterol (0.2%) for 10 d. Animals had ad libitum access to food and water. After at least 3 d of adaptation to the respective diet, the mice were housed at a thermone￾utral temperature (30 °C), 22 °C, 16 °C or 6 °C for 7 d in a humidity-controlled climate chamber before their organs and fecal samples were harvested. For measurements of body temperature, we implemented a stepwise reduction in ambient temperature with 24-h steps. We activated FXR in vivo by oral admin￾istration of the agonist Phenex 20606 (Phenex Pharmaceuticals AG; 40 mg/ml in DMSO, dissolved in olive oil to a final concentration of 2 mg/ml) on days 8, 9 and 10. Control mice received an oral gavage of the vehicle. For β3-adrener￾gic receptor stimulation we supplemented the cholesterol-enriched HFD with CL316243 (0.001% wt/wt; Tocris). To block intestinal cholesterol absorption, we treated mice with an ezetimibe-supplemented diet (10 mg ezetimibe per 167 mg HFD; Ezetrol, MSD) for the entire experimental period. For depletion of intesti￾nal microbiota we added bacitracin, neomycin and streptomycin (1 g/liter each; Sigma-Aldrich) to the drinking water. We anesthetized mice after a 4-h fasting period and transcardially withdrew blood with syringes that were pre-treated with EDTA (Sigma-Aldrich) to isolate plasma. Subsequently, we perfused the animals with PBS and harvested the tissues, which were immediately conserved either in TRIzol reagent (Invitrogen) or stored at −80 °C for further processing. Intestinal and fecal samples were immediately stored on dry ice. Microbiome analysis. We performed sequencing of the 16S-rRNA-encoding gene of fecal samples as described48. We performed microbiome analysis using R. To minimize the effect of compositional outliers, multivariate outliers were detected and removed using the ‘mvoutlier’ package (https://cran.r-project.org/ web/packages/mvoutlier/index.html) based on a core microbiota, including all genera with a mean relative abundance of greater than 0.5%. Alpha diversity measures, including observed number, Chao1 estimator and Shannon index, were calculated using the ‘diversity()’ and ‘estimateR()’ functions in the ‘vegan’ package for R based on genus level and operational taxonomic unit (OTU) level abundance tables. We used pairwise Wilcoxon rank-sum tests to determine significant differences between treatment groups. The Bray–Curtis dissimilarity was calculated on the OTU and genus level abundance tables using the ‘vegdist()’ function from the ‘vegan’ package. Permutationals MANOVA was used to detect differences in beta diversity between the treatment groups using the ‘adonis()’ function and 10,000 permutations. Identification of differentially abundant gen￾era- and species-level OTUs was done based on a core microbiota, including those taxa with a mean relative abundance >1% in at least one of the treatment groups. The ‘mvabund’ package49 was used for fitting generalized linear models with a negative binomial distribution to the single taxa. We performed ANOVA with 10,000 permutations to assess which taxa deviated significantly between groups. We adjusted P values for multiple-testing by applying the Benjamini– Hochberg correction50. In box plots, center lines represent medians, limits are first and third quartiles (inter-quartile range; IQR), whiskers represent the lowest and highest values within a 1.5 × IQR range ± the first/third quartile, and points are values outside the 1.5 × IQR range ± the first/third quartile. Plasma parameters. We quantified plasma cholesterol and triglyceride lev￾els using commercial kits (Roche) that were adapted to 96-well microtiter plates. For lipoprotein profiling, we used 200 µl plasma, which was separated by fast-performance liquid chromatography (FPLC) on a Superose 6 10/300 GL column (GE Healthcare) with a flow rate of 0.5 ml/min. Forty fractions (0.5 ml each) were collected. Triglyceride and cholesterol concentrations were measured in each fraction. Cholesterol uptake studies. To analyze cholesterol uptake, we housed mice at thermoneutrality or 6 °C and fed them the cholesterol-enriched HFD described above. 4 h before organ and blood collection, we orally administered a 200-µl lipid emulsion (10 mg human triglyceride-rich lipoprotein-derived lipid per ml saline) enriched with a tracer dose of 14C-cholesterol (185 kBq/mouse). For measurement of radioactivity by β-scintillation counting, we solubilized organs in Solvable (PerkinElmer, 100 µl per 10 mg organ). Cholesterol uptake was calculated as a percentage of the injected dose. Bile acid and hydroxysterol measurements. We performed targeted bile acid and hydroxysterol analysis by HPLC coupled to electrospray ioniza￾tion tandem mass spectrometry as recently described51. In brief, sample preparation comprised a simple methanol liquid–liquid extraction of tissues and biofluids. Quantitative measurement of bile acids was performed using a HPLC-ESI-QqQ system in multiple reaction monitoring (MRM) mode (HPLC: 1200 Infinity Quaternary LC System (Agilent); column: Accucore Polar Premium (2.6 µm, 150 mm × 2.1 mm inner diameter, Thermo Fisher Scienctific, Inc.); QqQ: API 4000 Q trap (ABSCIEX)). Peak identifica￾tion and quantification were performed by comparing retention times, as well as MRM transitions and peak areas, respectively, to corresponding standard chromatograms. Gene expression analysis. We isolated and purified total RNA from mouse organs by using the NucleoSpin RNA II kit (Macherey & Nagel) and prepared cDNA using SuperScript III Reverse Transcriptase (Invitrogen). We used exist￾ing human liver cDNA samples that were published previously52. Quantitative real-time PCR reactions for the indicated genes were conducted on a 7900HT sequence detection system (Applied Biosystems) using TaqManAssay-on￾Demand primer sets (Applied Biosystems, Cyp7a1: Mm00484150_m1, Cyp8b1: Mm00501637_s1, Cyp7b1: Mm00484157_m1, Cyp27a1: Mm00470430_m1, Baat: Mm00476075_m1, Abcb11: Mm00445168_m1, Nr0b2: Mm00442278_m1, Slc10a2: Mm00488258_m1, Ucp1: Mm00494069_m1, Dio2: Mm00515664_m1, Elovl3: Mm00468164_m1, Ppargc1a (Pgc1a): Mm00447183_m1, Prdm16: Mm00712556_m1, Slc10a1: Mm00441421_m1, Col1a1: Mm00801666_g1, Cxcl10: Mm00445235_m1, Emr1: Mm00802530_m1, Tgfb1: Mm00441724_m1, Tnf: Mm00443258_m1, Tbp: Mm00446973_m1, CYP7A1: Hs00167982_m1, CYP7B1: Hs00191385_m1, CYP8B1: Hs00244754_s1 and TAF1: Hs00270322_ m1). Cycle thresholds (Ct values) were normalized to those of the housekeeping genes TATA-box binding protein (Tbp) or TATA-box binding protein associated factor 1 (TAF1). Study subjects. The study design, subjects and sample collection have been described in detail52. In brief, liver samples were obtained from subjects who were undergoing bariatric surgery (obese) or abdominal surgery for early color￾ectal cancer or for benign diseases such as sigma-diverticulosis (non-obese) at the Department of General and Visceral Surgery, University of Ulm, Germany. Individuals undergoing bariatric surgery had not been subjected to a weight￾loss regimen before surgery, but they were commended to adhere to a regular meal pattern to minimize the confounding effects of negative weight balance. We selected gender-matched subgroups including subjects who were non￾obese without diabetes (non-obese), subjects who were obese without diabetes (obese) and subjects who were obese and had type 2 diabetes (obese + T2D) from the total cohort. Permission was given under the number 112/2003 from the local ethics committee in Ulm, Germany, and informed consent in writing was obtained from each subject. Plasma activity of alanine aminotransferase and aspartate aminotransferase. For assessment of liver damage, we determined plasma activities of alanine ami￾notransferase (ALT) and aspartate aminotransferase (AST) using the COBAS Mira System (Roche)