Received: 22 March 2020 | Revised: 12 June 2020 | Accepted: 19 June 2020 DOI: 10.1096/fj.202000678R RESEARCH ARTICLE Liver CPT1A gene therapy reduces diet-induced hepatic steatosis in mice and highlights potential lipid biomarkers for human NAFLD Minéia Weber1 | Paula Mera1,2 | Josefina Casas3,4 | Javier Salvador2,5 | Amaia Rodríguez2,6 | Sergio Alonso7 | David Sebastián8,9 | M. Carmen Soler-Vázquez1 | Carla Montironi10,11 | Sandra Recalde1 | Raquel Fucho1 | María Calderón-Domínguez1 | Joan Francesc Mir1 | Ramon Bartrons12 | Joan Carles Escola-Gil13,9 | David Sánchez- Infantes14,2 | Antonio Zorzano8,9 | Vicenta Llorente-Cortes15,16,17 | Núria Casals18,2 | Víctor Valentí6,19 | Gema Frühbeck2,6 | Laura Herrero1,2 | Dolors Serra1,2 1 Department of Biochemistry and Physiology, School of Pharmacy and Food Sciences, Institut de Biomedicina de la Universitat de Barcelona (IBUB), Universitat de Barcelona, Barcelona, Spain 2 Centro de Investigación Biomédica en Red de Fisiopatología de la Obesidad y la Nutrición (CIBEROBN), Instituto de Salud Carlos III, Madrid, Spain 3 Research Unit on BioActive Molecules, Department of Biological Chemistry, Institute of Advanced Chemistry of Catalonia (IQAC)/CSIC, Barcelona, Spain 4 Centro de Investigación Biomédica en Red de Enfermedades Hepáticas y Digestivas (CIBEREHD), Instituto de Salud Carlos III, Madrid, Spain 5 Department of Endocrinology & Nutrition, Clínica Universidad de Navarra, Pamplona, Spain 6 Metabolic Research Laboratory, Clínica Universidad de Navarra, IdiSNA, Pamplona, Spain 7 Cancer Genetics and Epigenetics Group, Program of Predictive and Personalized Medicine of Cancer, Germans Trias i Pujol Research Institute (IGTP- PMPPC), Campus Can Ruti, Barcelona, Spain 8 Institute for Research in Biomedicine (IRB Barcelona), Barcelona Institute of Science and Technology, Departament de Bioquímica i Biomedicina Molecular, Facultat de Biologia, Universitat de Barcelona, Barcelona, Spain 9 Centro de Investigación Biomédica en Red de Diabetes y Enfermedades Metabólicas Asociadas (CIBERDEM), Instituto de Salud Carlos III, Madrid, Spain 10 Pathology Department, Hospital Clinic de Barcelona, Barcelona, Spain 11 Liver Cancer Translational Research Laboratory, Liver Unit, IDIBAPS-Hospital Clínic, Universitat de Barcelona, Barcelona, Spain 12 Departament de Ciències Fisiològiques, Facultat de Medicina i Ciències de la Salut, Universitat de Barcelona, Spain 13 IIB Sant Pau, Departament de Bioquímica i Biologia Molecular, Universitat Autònoma de Barcelona, Barcelona, Spain 14 Germans Trias i Pujol Research Institute (IGTP-PMPPC), Campus Can Ruti, Barcelona, Spain 15 Institute of Biomedical Research of Barcelona (IIBB), Spanish National Research Council (CSIC), Biomedical Research Institute Sant Pau (IIB Sant Pau), Barcelona, Spain 16 CIBERCV, Institute of Health Carlos III, Madrid, Spain 17 Cardiovascular Research Center, CSIC-ICCC, Barcelona, Spain 18 Basic Sciences Department, Faculty of Medicine and Health Sciences, Universitat Internacional de Catalunya, Sant Cugat del Vallès, Spain Abbreviations: AAV, adeno-associated viruses; ACC1, acetyl-CoA carboxylase 1; ASP, acid-soluble products; BHB-CoA, β-hydroxybutyryl-CoA; CPT1A, carnitine palmitoyltransferase 1A; DAG, diacylglycerides; DGAT2, diacylglycerol O-acyltransferase homolog 2; FA, fatty acids; FFA, free fatty acid; FAO, fatty acid oxidation; GFP, green fluorescent protein; G6Pase, glucose-6-phosphatase; GTT, glucose tolerance test; HFD, high-fat diet; HMGCS2, hydroxymethylglutaryl- CoA synthase 2; IR, insulin resistance; LPE, lysophosphatidylethanolamine; MCD, malonyl-CoA decarboxylase; MTP, microsomal triacylglycerol transfer protein; NAFLD, nonalcoholic fatty liver disease; NAS, NAFLD activity score; qRT-PCR, real-time quantitative PCR; PC, phosphatidylcholine; ROS, reactive oxygen species; RYGB, Roux-en-Y gastric bypass; TAG, triacylglyceride; UCP2, uncoupling protein 2; VLDL, very-low-density lipoprotein. Minéia Weber and Paula Mera contributed equally to this study. This is an open access article under the terms of the Creative Commons Attribution NonCommercial License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited and is not used for commercial purposes. © 2020 The Authors. The FASEB Journal published by Wiley Periodicals LLC on behalf of Federation of American Societies for Experimental Biology 11816 |  wileyonlinelibrary.com/journal/fsb2 The FASEB Journal. 2020;34:11816–11837. | 15306860, 2020, 9, Downloaded from https://faseb.onlinelibrary.wiley.com/doi/10.1096/fj.202000678R by Readcube (Labtiva Inc.), Wiley Online Library on [15/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License WEBER et al.     11817 19 Department of Surgery, Clínica Star; L'Oréal-UNESCO; Ciência sem Fronteiras-CNPq, Grant/Award Number: 237976/2012-9 Universidad de Navarra, Pamplona, Spain Abstract Correspondence The prevalence of nonalcoholic fatty liver disease (NAFLD) has increased drastically Dolors Serra, Department of Biochemistry and Physiology, School of Pharmacy and due to the global obesity pandemic but at present there are no approved therapies. Food Sciences, University of Barcelona, Here, we aimed to revert high-fat diet (HFD)-induced obesity and NAFLD in mice by Av. Joan XXIII, 27-31, E-08028 Barcelona, enhancing liver fatty acid oxidation (FAO). Moreover, we searched for potential new Spain. Email: dserra@ub.edu lipid biomarkers for monitoring liver steatosis in humans. We used adeno-associated virus (AAV) to deliver a permanently active mutant form of human carnitine pal- Funding information Ministry of Spain (MINECO), Grant/ mitoyltransferase 1A (hCPT1AM), the key enzyme in FAO, in the liver of a mouse Award Number: SAF2014-52223-C2-1-R, model of HFD-induced obesity and NAFLD. Expression of hCPT1AM enhanced he- and SAF2017-83813-C3-1-R; European patic FAO and autophagy, reduced liver steatosis, and improved glucose homeostasis. Regional Development Fund [ERDF; Juan de la Cierva—Incorporación, Grant/ Lipidomic analysis in mice and humans before and after therapeutic interventions, Award Number: IJCI-2016-28313; such as hepatic AAV9-hCPT1AM administration and RYGB surgery, respectively, Centro de Investigación Biomédica en led to the identification of specific triacylglyceride (TAG) specie (C50:1) as a poten- Red de Fisiopatología de la Obesidad y la Nutrición (CIBEROBN), Grant/Award tial biomarker to monitor NAFFLD disease. To sum up, here we show for the first Number: CB06/03/0001; Government time that liver hCPT1AM gene therapy in a mouse model of established obesity, of Catalonia, Grant/Award Number: diabetes, and NAFLD can reduce HFD-induced derangements. Moreover, our study 2014SGR465; Instituto de Salud Carlos III, Grant/Award Number: FISPI18/01584, highlights TAG (C50:1) as a potential noninvasive biomarker that might be useful to FIS-PI17/00412, FIS PI18/01484, monitor NAFLD in mice and humans. FIS17/01455, CP15/00106, CM19/00039 ; European Regional Development Fund KEYWORDS (ERDF); Fundació La Marató de TV3, Grant/Award Number: 201627-30, 201627- AAV, CPT1A, fatty-acid oxidation, gene therapy, hepatic steatosis, lipid biomarker, obesity 31; European Foundation for the Study of Diabetes (EFSD); Lilly; Janssen-Rising 1 | IN T RO D U C T ION This enzyme is physiologically inhibited by malonyl-CoA, a glucose-derived metabolite that is an intermediate in the Despite enormous efforts by health-care providers and de novo synthesis of fatty acids (FA). Increased glucose the research community, obesity rates continue to rise. metabolism raises levels of malonyl-CoA, which inhibits Unfortunately, safe, long-term efficient treatments are cur- CPT1 and blocks FAO, resulting in lipid partitioning and rently unavailable. Of great concern is the parallel increase in storage. A significant number of studies in rodents have the prevalence of obesity-associated pathological conditions tried to indirectly promote FAO to reduce systemic obesity such as insulin resistance, type 2 diabetes (T2D), nonalco- and prevent the development of obesity-associated meta- holic fatty liver disease (NAFLD), cardiovascular disease, bolic disorders. These studies targeted the regulation of en- and cancer. Hence, it is vital to develop novel therapeutic zymes and transcription factors involved in FA metabolism, strategies to combat this epidemic. such as acetyl-CoA carboxylase (ACC),2 malonyl-CoA One of the main causes of obesity is the modern lifestyle, decarboxylase (MCD),3 uncoupling protein 1 (UCP1),4 which is characterized by excessive calorie intake and in- AMP-activated protein kinase (AMPK),5 peroxisome pro- sufficient physical activity. The liver plays an essential role liferator-activated receptor (PPAR),6 and long-chain acyl- in regulating whole-body energy balance and lipid/glucose CoA dehydrogenase (VLCAD).7 Studies that increased homeostasis. In conditions associated with prolonged excess short-term CPT1A activity in liver demonstrated a decrease energy intake, the liver can store significant quantities of lip- in hepatic triglyceride (TAG) content,8 but did not report ids, which eventually leads to steatosis. Liver steatosis can any improvement in insulin sensitivity. Furthermore, we cause hepatic inflammation and progress to steatohepatitis and others have shown that a long-term increase in CPT1A (NASH),1 which predisposes to hepatic injury and cancer. and a mutated isoform that is insensitive to malonyl-CoA The design of new therapeutic strategies to enhance fat burn- (CPT1AM)9 prevented steatosis and the development of ing in liver may alleviate these obesity-related disorders. obesity in mice fed a high-fat diet (HFD).10,11 Still, these Mitochondrial long-chain fatty acid β-oxidation (FAO) studies could not revert the HFD-induced derangements in is regulated by carnitine palmitoyltransferase 1 (CPT1). an established model of obesity. |    15306860, 2020, 9, Downloaded from https://faseb.onlinelibrary.wiley.com/doi/10.1096/fj.202000678R by Readcube (Labtiva Inc.), Wiley Online Library on [15/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 11818 WEBER et al. The aim of the present study was to determine whether measured by enzymatic tests (Hitachi Modular P800, Roche, AAV-mediated long-term activation of hepatic FAO could Basel, Switzerland). Total cholesterol, HDL-cholesterol, be a successful strategy to revert an already established obe- LDL-cholesterol, and triacylglycerol concentrations (Roche) sity with associated hyperglycemia and NAFLD in mice. We were calculated as previously described.13 High sensitiv- focused on three novel strategies: (1) use of the AAV9 se- ity C-reactive protein (CRP) was measured using the Tina- rotype, which is the most effective AAV in gene therapy at quant CRP (Latex) ultrasensitive assay (Roche). Insulin was targeting liver; (2) use of a mouse model with established measured by an enzyme-amplified chemiluminescence assay obesity, type 2 diabetes, and NAFLD; and (3) use of the (IMMULITE; Diagnostics Products Corp., Los Angeles, CA, human isoform of CPT1AM (hCPT1AM), with the prospect USA). Intra- and inter-assay coefficients of variation were of future clinical therapeutic applications. Our results show 4.2% and 5.7%, respectively. Insulin resistance and sensitiv- that enhanced hepatic FAO through hCPT1AM-mediated ity were calculated using the homeostasis model assessment gene therapy reduces diet-induced weight gain, and hepatic (HOMA) and a quantitative insulin sensitivity check index steatosis and improves hepatic insulin signaling in obese (QUICKI), respectively. Leptin was quantified by a double- mice. Mechanistically, hCPT1AM-expressing mice showed antibody RIA method (Linco Research, Inc, St. Charles, MO, enhanced autophagy, ketogenesis, and oxidative phosphory- USA).14,15 Intra- and inter-assay coefficients of variation lation in the liver. Importantly, hCPT1AM expression also were 6.7% and 7.8%, respectively. All reported investiga- changes the hepatic lipidomic profile. One of the altered lipid tions were carried out in accordance with the principles of the species, C50:1 triacylglyceride, was concomitantly lower in Declaration of Helsinki, as revised in 2013, and approved by the serum of human patients with obesity and NAFLD after the Hospital's Ethical Committee for Research (061/2014). bariatric surgery. Thus, we identified a potential biomarker to Written informed consent was obtained from all participants. monitor the progression of hepatic steatosis in humans. 2.2 | Adeno-associated vectors 2 | M AT E R IA L S A N D ME T HODS AAV vectors, serotype 9, were used to drive the expression 2.1 | Human study cohort and biological of GFP (AAV9-GFP) or hCPT1AM (AAV9-hCPT1AM) in analysis mouse liver. Plasmid vectors carried: (a) the human albumin enhancer element and the human α1-antitrypsin (EalbAATp) Serum samples were obtained from obese patients liver-specific promoter described by Kramer et al,16 (b) the (BMI ≥ 30 kg/m2 and body fat percentage [BF] ≥35% for cDNA sequence of GFP or hCPT1AM, (c) The woodchuck women and ≥ 25% for men) before and 6 months after Roux- posttranscriptional regulatory element (WPRE, acc #AY468- en-Y gastric bypass (RYGB) (n = 15) and from healthy lean 486),17 (d) The bovine growth hormone polyadenosine volunteers (n = 15) at the Clínica Universidad de Navarra. transcription termination signal [bGH-poly(A)] (bases 2326- Inclusion criteria included a complete diagnosis: physical ex- 2533 GenBank acc #M57764). The expression cassette was amination, laboratory investigation, ultrasound echography, flanked by two-inverted terminal repeats (ITRs) derived from and a liver biopsy consistent with the diagnosis of nonalco- AAV2. The vector preparations used titers of 1.34 × 1013 holic fatty liver disease (NAFLD), according to Brunt crite- and 1.16 × 1013 genome copies (gc)/mL for AAV9-GFP and ria.12 Exclusion criteria were: (i) excess alcohol consumption AAV9-hCPT1AM, respectively. (≥20 g for women and ≥30 g for men); (ii) the presence of hepatitis B virus surface antigen or hepatitis C virus anti- bodies in the absence of a history of vaccination; (iii) use 2.3 | Animals of drugs linked to NAFLD, including amiodarone, val- proate, tamoxifen, methotrexate, corticosteroids, or antiret- The animal study was designed as follows: 40 seven-week- rovirals; (iv) evidence of other specific liver diseases, such old male C57BL/6J mice were purchased from Janvier (SC- as autoimmune liver disease, hemochromatosis, Wilson's C57J-M) and housed in our animal facility for 1 week to disease, or α-1-antitrypsin deficiency. Patients with type 2 acclimatize. Mice were kept at standard laboratory condi- diabetes (T2D) were not on any antidiabetic treatment be- tions (12 hours light/dark cycle, 20-22ºC). Animals were fore or during the study. Obese patients with T2D did not randomly assigned to one of the following groups: NCD- have a long history of diabetes (less than 2-3 years or even GFP, HFD-GFP, NCD-hCPT1AM, and HFD-hCPT1AM. de novo diagnosis as evidenced from their anamnesis and Five animals from each group were housed in the same cage. biochemical determinations). Glucose, uric acid, ALT, AST, Next, mice were fed with either NCD (manufactured by alkaline phosphatase, and γ-glutamyltransferase (γ-GT) were TestDiet [catalogue number #58Y2], 10% kcal fat) of HFD | 15306860, 2020, 9, Downloaded from https://faseb.onlinelibrary.wiley.com/doi/10.1096/fj.202000678R by Readcube (Labtiva Inc.), Wiley Online Library on [15/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License WEBER et al.     11819 (manufactured by TestDiet [catalogue number #58Y1], 60% 2.5 | Isolation of primary mouse hepatocytes kcal fat) for the entire duration of the experiment. This diet causes steatosis but not severe liver injury (fibrosis). Body The collagenase method was used to isolate primary mouse weight was measured every week and blood glucose once hepatocytes.21 Livers were perfused with Hank's balanced a month. After 9 weeks of diet, most of the mice already salt solution (HBSS, pH 7.4, 37°C, gassed with 95% O2 and showed an obese phenotype characterized by a significant 5% CO2: 5.4 mM KCl, 0.44 mM KH2PO4, 138 mM NaCl, increase in circulating levels of glucose and insulin and 4.17 mM NaHCO3, 0.338 mM Na2HPO4, 5.56 mM glucose, higher body weight than the control NCD-fed group. AAV9- 50 mM HEPES, and 0.5 mM EGTA) through the portal hCPT1AM and AAV9-GFP vectors were administered by vein for 5 minutes at a rate of 5 mL/min. Then, livers were tail-vein injection with a single dose (1.5 × 1012 gc/kg). To perfused with HBSS (without EGTA) containing 5 mM of do that, mice were anesthetized in the morning (~10 AM) CaCl2 and 0.25% (w/v) collagenase IV (Sigma) for 15 min- using isoflurane, and then, placed in a restrainer for the tail- utes approximately. The liver was then removed and gently vein injection. After injection, mice were returned to their teased apart in HBSS, and the cell suspension was washed home cage and room. Eleven weeks after virus injection, three times in HBSS. Cell viability as assessed by the trypan mice were anesthetized using isoflurane, and then, killed blue exclusion test was always higher than 80%. Hepatocytes by cervical dislocation in fasting conditions (6 hours). The were seeded at a density of 3.5 × 106 cells in 0.1% (w/v) gela- total duration of the HFD feeding paradigm was 20 weeks tin-treated 25-cm2 flasks in DMEM medium (Gibco #11966), (9 weeks before AAV9 vector administration and 11 weeks supplemented with 10% of FBS, 10 mM of glucose, 10 μg/ after that). The Animal Experimentation Ethics Committee mL of streptomycin, 100 units/mL of penicillin, 100 nM of of the University of Barcelona (CEEA-UB) approved all the dexamethasone (Sigma), and 100 nM of insulin (Sigma). procedures with mice. Permits 9611 and 9609 were obtained from the Government of Catalonia, according to European Directive 2010/63/EU. 2.6 | Fatty acid oxidation Fatty acid oxidation (FAO) to CO2 and acid-soluble products 2.4 | Construction of site- (ASPs), essentially consisting of acyl-carnitine, Krebs-cycle directed hCPT1AM intermediates, and acetyl-CoA, was measured in isolated pri- mary mouse hepatocytes cultured in 25-cm2 flasks. Following hCPT1A wild-type (wt) cDNA (GenBank Access No. cell attachment after isolation (approximately 5 hours), the HGNC: 2328) was obtained by hCPT1A cDNA am- medium was replaced for 16 hours by fresh FAO medium plification from the HepG2 cell line. The primers used (DMEM medium containing 0.1% FA-free BSA, 1 mM car- were: hCPT1AHindIII.for (5′TCGATAAGCTTATAAA nitine, 20 mM glucose, 10 nM insulin, 10 μg/mL strepto- ATGGCAGCTCACCAAGC3′) and hCPT1AEcoR1.rev mycin and 100 units/mL penicillin). Then, the medium was (5′CATCCGGAATTCCTTTTTGGAATTAGA replaced by fresh FAO medium containing 0.3 mM of oleate ACTG3′). The amplification was cloned in pYES plasmid and 1 μCi/ml [1-14C]oleic acid bound to 1% (w/v) BSA and to obtain the construct pYEShCPT1Awt. Plasmid pYESh- hepatocytes were incubated for 3 hours at 37°C in a CO2-free CPTIAM593S was constructed using the Quick-Change incubator. Oxidation to CO2 and ASPs was measured as de- PCR-based mutagenesis procedure (Stratagene) with the scribed previously.22 YEShCPT1Awt plasmid as template. Appropriate substitu- tions and the absence of unwanted mutations were confirmed by sequencing the inserts in both directions using Applied 2.7 | Isolation of mitochondria from Biosystems 373 automated DNA sequences. To check liver and assay of CPT1 activity hCPT1AM insensitivity to malonyl-CoA, we expressed the constructs containing hCPT1A wt and mutant (hCPT1AM) Mitochondrial-enriched fractions from cultured hepatocytes in yeast cells, and then, prepared cell extracts as described in were obtained by differential centrifugation. Briefly, hepato- Ref. (9) for the assay of CPT1 activity. S cerevisiae was cho- cytes were scraped and resuspended in 2 mL of solution A sen as an expression system for hCPT1A wt and hCPT1AM (70 mM sucrose, 220 mM mannitol, 2 mM EDTA, 5 mM because it does not have endogenous CPT1 activity. The HEPES, and protease inhibitor cocktail from Roche, pH cDNA of the hCPT1AM gene used in our study contains the 7.4). Then, cells were homogenized using a glass Wheaton- NH2-terminal sequence that direct the protein to the mito- Dounce homogenizer (20 strokes with both the loose and chondria and two more sequences that codify for two α-helix the tight pestle) and centrifuged at 2000 g for 3 minutes at fragments that insert this protein into the mitochondrial outer 4°C. Next, the supernatant was centrifuged at 16 000 g for membrane.18–20 30 minutes at 4°C. Finally, the mitochondria-enriched pellet |    15306860, 2020, 9, Downloaded from https://faseb.onlinelibrary.wiley.com/doi/10.1096/fj.202000678R by Readcube (Labtiva Inc.), Wiley Online Library on [15/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 11820 WEBER et al. was resuspended in 50-100 µL of solution A. CPT1 activity signaling #5114S (1:1000 in 5% Milk PBS-Tween); PLIN2, was measured in 10 µg of protein from those mitochondria- Abcam #Ab37516 (1:200 in 5% Milk PBS-Tween); PLIN5, enriched fractions, using the radiometric method as described Abcam #Ab192749 (1:2000 in 5% Milk PBS-Tween); LDLr, previously.23 Concisely, the substrates L-[methyl-3H]carni- Abcam #Ab52818 (1:1500 in 5% Milk PBS-Tween); LRP1, tine and palmitoyl-CoA were added to the protein suspension Abcam #Ab92544 (1:7000 in 5% Milk PBS-Tween); VLDLr, and enzyme activity was assayed for 4 minutes at 30°C in a Santa Cruz #sc-18824 (1:1500 in 5% Milk PBS-Tween); total volume of 200 μL. Next, butanol was used to isolate the TIM44, BD Bioscience #612582 (1:5000 in 5% Milk PBS- fraction containing the [3H]palmitoyl-carnitine and radioac- Tween); pAkt (Ser473), Cell signaling #4051S (1:1000 in 5% tivity was measured by liquid scintillation counting. Milk PBS-Tween); Akt, Cell signaling #9272S (1:1000 in 5% Milk PBS-Tween); UCP2, Cell signaling #89326 (1:1000 in 5% Milk PBS-Tween); HMGS2, produced in our laboratory 2.8 | Serum biochemical analyses (1:6000 in 5% Milk PBS-Tween); β-actin, Sigma Aldrich #A3854 (1:50.000 in PBS-Tween); Total OXPHOS Rodent Blood samples were obtained from the tails of awake mice WB Antibody Cocktail, Abcam #ab110413 (1:5000 in 5% after an overnight fast. Blood glucose concentrations were Milk PBS-Tween); Lipolysis Activation Antibody Sampler measured using a Glucometer Elite (Bayer). Serum insulin Kit, Cell signaling #8334 (for all antibodies 1:1000 in 5% was measured by ELISA (Millipore). Plasma triacylglyc- Milk PBS-Tween); ATGL, Cell signaling #2138 (1:1000 in erides, β-hydroxybutyrate, cholesterol, HDL cholesterol, 5% Milk PBS-Tween). AST, and ALT were determined enzymatically using com- mercial kits adapted to a COBAS 6000 autoanalyzer (Roche Diagnostics, Rotkreuz, Switzerland).24 2.12 | Measurements of liver TAG content To measure TAG content in the liver, 100 mg of frozen tis- 2.9 | Glucose tolerance test sue were homogenized in 500 μL of PBS. Then, chloroform- extracted lipids were dried under an N2 stream and dissolved For the GTT, glucose (2.0 g per kg body weight) was dis- in isopropanol. TAG were quantified in the extracted lipids solved in PBS and administered by intraperitoneal injection using the Cromatest Triglycerides MR kit (Linear Chemicals after an overnight fast. Then, blood glucose concentrations SL). were measured using a Glucometer Elite (Bayer) before, 15, 30, 60, 90, and 120 minutes after administration. 2.13 | ATP extraction and measurement 2.10 | Analysis of insulin/AKT signaling Frozen liver samples were powdered and homogenized in pathway in hepatocytes 1.0 mL of 40% ice-cold HCLO4 and incubated for 10 min- utes. The precipitated proteins were removed by cen- Analyses of insulin/AKT signaling were performed in iso- trifugation (10 000 g for 10 minutes). Next, 300 μL of the lated primary mouse hepatocytes. Sixteen hours after isola- supernatant was neutralized by 5 M KHCO3. Then, 10 μL tion, cells were incubated with 100 nM of insulin (Humulin was pipetted into the wells of a white nonphosphorescent 100 UI/mL, Lilly) for 5 minutes. After that, protein extracts microplate, placed in a luminometer (Turner System), and from cells were used to measure phosphorylation of AKT processed by addition of 90 μL of ATP monitoring reagent using Western blot, as described below. (Molecular Probe, Life Technologies A22066). ATP concen- trations were calculated from a calibration curve constructed at the same time by means of standard ATP dissolved in the 2.11 | Western blot appropriate solution for each experiment. Protein determination by Western blot was performed using standard methods. CPT1A detection was undertaken in mi- 2.14 | Measurement of oxygen consumption tochondria-enriched fractions. The antibodies used are the in isolated mitochondria following: Rabbit CPT1A-specific polyclonal antibody pro- duced in our laboratory against amino acids 317-430 of the Respiration of isolated liver mitochondria (500 µg protein) rat CPTIA,10 Sci Crunch Cat# CPT1A, RRID:AB_2636894 was measured at 37°C by high-resolution respirometry (1:6000 in 5% Milk PBS-Tween); LC3B, Cell signaling with the Oxygraph-2k (OROBOROS INSTRUMENTS, #2775 (1:1000 in 5% Milk PBS-Tween); SQSTM/p62, Cell Innsbruck, Austria). Mitochondria were obtained in buffer | 15306860, 2020, 9, Downloaded from https://faseb.onlinelibrary.wiley.com/doi/10.1096/fj.202000678R by Readcube (Labtiva Inc.), Wiley Online Library on [15/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License WEBER et al.     11821 II (sucrose 0.5 M, KCl 50 mM, EDTA 5 mM, sodium py- PCR amplifications, the LightCycler 480 SYBR Green I rophosphate 1 mM, MgCl2 5 mM, pH 7.4, and protease Master kit (Roche) was used and measured with the Roche inhibitors) and 500 µL of mitochondria was brought to a LightCycler 480 Real-Time PCR System. The sequences of final volume of 2.1 mL with respiration medium (EGTA the oligonucleotides used are: 0.5 mM, MgCl2.6H2O 3 mM, taurine 20 mM, KH2P04 10 mM, HEPES 20 mM, BSA 1 g/l, K-lactobionate Abca1 For: 5′-TTCTTGCGATACACTCTGGTGC Rev: 60 mM, sucrose 110 mM, pH 7.1) and added into the oxy- 5′-CGGGATTGAATGTTCTTGTCGT; graph chamber. All respiration measurements were made Atgl For: 5′-TGACCATCTGCCTTCCAGA Rev: in triplicate. Resting respiration (state 4, absence of ade- 5′-TGTGGTGGCGCAAGAACA; nylates) was assessed by the addition of 10 mM glutamate Atg7 For: 5′-CCGGTGGCTTCCTACTGTTA Rev: and 2 mM malate as the complex I substrate supply. Then, 5′-AAGGCAGCGTTGATGACC; state 3 respiration was assessed by the addition of 2.5 mM β-actin For: 5′-ATGCTCCCCGGGCTGTAT Rev: ADP. RCR was calculated as the quotient state 3/state 4. 5′-CATGGAGTCCTTCTGACCCAT; The integrity of the outer mitochondrial membrane was es- Hsl For: 5′-GCGCTGGAGGAGTGTTTTT Rev: tablished by the addition of 10 μM cytochrome c and no 5′-CGCTCTCCAGTTGAACCAAG; stimulation of respiration was observed. The addition of Mgl For: 5′-TGAGAAAGGCTTTAAGAACTGGG Rev: 10 mM succinate provided state 3 respiration with parallel 5′-GACCACCTGTGTGATGTGGG; electron input to complexes I and II. The concentrations of Mtp For: 5′-GTGGAGGAATCCTGATGGTGA Rev: substrates and inhibitors used were based on prior experi- 5′-TGATCTTGGTGTACTTTTGCCC; ments conducted for optimization of the titration protocols. hcpt1a For: 5′-CAATCGGACTCTGGAAACG Rev: 5′-CCGCTGACCACGTTCTTC. 2.15 | Measurements of liver temperature For Abcg8c and Abcg5 gene expression analyses, cDNA was subjected to real-time PCR amplification using TaqMan The thermal images from mice liver were made using a FLIR Master Mix (Applied Biosystems) and specific mouse T420 infrared camera (FLIR T420 Systems AB, Sweden) TaqMan probes (Applied Biosystems) (Mm00446241_m1 with thermal sensitivity of 0.05°C and resolution set at [Abcg5] and Mm00445970_m1 [Abcg8]). 320 × 240 pixels. Mice were anesthetized and their abdomi- nal cavity was opened. The first image was taken of a piece of aluminum foil to obtain a measurement of the reflected tem- 2.17 | Oil Red O staining in hepatocytes perature for each set of images. To take the images, the cam- era was placed 30 cm from the midpoint of the liver. Thermal TAG content in isolated primary mouse hepatocytes was data were extracted from the pictures using a region of inter- measured using Oil Red O staining (Sigma, O0625), fol- est (ROI)-based approach. The ROIs were manually drawn in lowing standard procedures. Oil Red O stock solution was the images of the liver region using FLIR Research IR Max prepared in isopropanol (3 mg/mL). It was then filtered after software version 4.40.6.24 for Windows (FLIR Systems Inc, overnight agitation and stored protected from light. The Oil North Billerica, MA, USA). The reflected temperature was Red O working solution was prepared by diluting the stock obtained by placing a rounded ROI on the aluminum foil 3:2 in distilled water. Sixteen hours after hepatocyte isolation, image and retaining the mean value (°C) for adjustments. cells were washed twice with PBS and fixed using formalin Images were not normalized by liver size or weight, since the 4% of formaldehyde for 30 minutes at room temperature. thermographic camera measures the minimum and maximal After fixation, hepatocytes were washed again with PBS, and temperatures for a given ROI. Thus, these measurements do then, permeabilized for 30 minutes with 60% of isopropanol. not vary according to liver size or weight.25 Next, cells were incubated with Oil Red O working solution for 15 minutes. Finally, cells were washed with water and ex- amined under the microscope. Lipid-packed vesicles appear 2.16 | RNA extraction and quantitative RT- as bright red after Oil Red O staining. PCR RNA extraction and quantitative RT-PCR were performed 2.18 | Autophagy flux assay using standard procedures. Briefly, total RNA was ex- tracted from frozen, pulverized tissue using the RNeasy Autophagy/LC3II flux was performed in isolated primary kit (QIAGEN). Then, cDNA was synthesized with the mouse hepatocytes. Sixteen hours after isolation cells were in- Transcriptor First Strand cDNA Synthesis kit (Roche). For cubated with chloroquine (CQ) (20 μM) for 6 hours. After that, |    15306860, 2020, 9, Downloaded from https://faseb.onlinelibrary.wiley.com/doi/10.1096/fj.202000678R by Readcube (Labtiva Inc.), Wiley Online Library on [15/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 11822 WEBER et al. protein extracts from cells were used to measure LC3II pro- the addition of 0.5 mL of methanol. Samples were evaporated tein levels using western blot, as described above. Autophagy again and resuspended in 150 µL of methanol. The tubes flux was calculated by subtracting the densitometry values of were centrifuged at 13 000 g for 3 minutes, and 130 µL of LC3II in CQ untreated from CQ-treated samples. the supernatants was transferred to ultra-performance liquid chromatography (UPLC) vials for injection and analysis. 2.19 | Determination of acylcarnitine content 2.20.2 | Sphingolipids To determine the acylcarnitines in liver and serum sam- A total of 750 µL of a methanol-chloroform (2:1, ples, 140 μL of serum and tissue homogenate was used for vol/vol) solution containing internal standards acylcarnitine extraction. First, 10 μL of Deuterated palmi- (N-dodecanoylsphingosine, N-dodecanoylglucosylsphingosine, toylcarnitine [20 μM] was added to get 200 pmol in 150 N-dodecanoylsphingosylphosphorylcholine, and C17- μL of final resuspension. Next, 750 μL of Folch reagent sphinganine, 0.2 nmol each, from Avanti Polar Lipids) were (chloroform:metanol 2:1), 100 μL of dH2O and 50 μL 0.1 M added to 0.05 mL of serum. Samples were extracted at 48°C KOH were added to the samples. Phases were separated overnight and cooled. Then, 75 µL of 1 M KOH in methanol after vortexing and centrifugation at 2000 rcf. Following, was added, and the mixture was incubated for 2 hours at 37°C. the top aqueous phase was carefully removed and the or- Following addition of 75 µL of 1 M acetic acid, samples were ganic phase containing the acylcarnitines was washed with evaporated to dryness, and stored at −20°C until analysis. Before 200 μL of methanol:dH2O:chloroform (48:47:3). After wash- analysis, 150 µL of methanol were add to the samples, centri- ing, the organic solvent was dried using a SpeedVac vacuum fuged at 13 000 g for 5 minutes and 130 µL of the supernatant concentrator. The day of the analyses, samples were resus- were transferred to a new vial and injected. pended with 150 μL of methanol. Acylcarnitines were ana- LC-HRMS analysis was performed using an Acquity ultra lyzed using an Acquity UPLC-TOF system (Waters) with an high-performance liquid chromatography (UHPLC) system BEH C8 column (1.7 mm particle size, 100 mm, 62.1 mm, (Waters, USA) connected to a Time of Flight (LCT Premier Waters). The two mobile phases were 1 mM of ammonium XE) Detector. Full scan spectra from 50 to 1800 Da were formate in methanol (phase A) and 2 mM of ammonium for- acquired, and individual spectra were summed to produce mate in H2O (phase B), both phases with 0.05 mM of for- data points each of 0.2 seconds. Mass accuracy at a resolving mic acid. The following gradient was programed: 0 minute, power of 10 000 and reproducibility were maintained using an 65% A; 10 minutes, 90% A; 15 minutes, 99% A; 17 minutes, independent reference spray via the LockSpray interference. 99% A; 20 minutes, 65% A, and a flow rate of 0.3 mL min−1. Lipid extracts were injected onto an Acquity UHPLC BEH Quantification was carried out using the extracted ion chro- C8 column (1.7 µm particle size, 100 mm × 2.1 mm, Waters, matogram of each compound, using 50-mDa windows. The Ireland) at a flow rate of 0.3 mL/min and a column tempera- linear dynamic range was determined by injecting standard ture of 30°C. The mobile phases were methanol with 2 mM of mixtures. Positive identification of compounds was based ammonium formate and 0.2% of formic acid (A)/water with on the accurate mass measurement with an error, 5 ppm and 2 mM of ammonium formate and 0.2% of formic acid (B). A their LC retention time compared to that of a standard (62%). linear gradient was programed as follows: 0.0 minutes: 20% B; 3 minutes: 10% B; 6 minutes: 10% B; 15 minutes: 1% B; 18 minutes: 1% B; 20 minutes: 20% B; 22 minutes: 20% B. 2.20 | Lipidomic analyses Sphingolipids (Cer, ceramide; SM, sphingomyelin; GlucCer, glucosylceramide; HexCer: hexosylceramides), glycerophospho- 2.20.1 | Phospholipids and neutral lipids lipids (PC, phosphatidylcholine; LPC: lysophosphatidylcholine; pPC: plasmanylcholine; LPEA: lysophosphatidylethanolamine; A total of 750 µL of a methanol-chloroform (1:2, vol/vol) and PC O-: ether-linked phosphatidylcholine), neutral lipids solution containing internal standards (16:0 D31_18:1 phos- (DAG: diacylglycerol; TAG: triacylglycerol; CE: cholesterol phocholine, 16:0 D31_18:1 phosphoethanolamine, 16:0 ester). All Lipid species are annotated using the “lipid subclass” D31-18:1 phosphoserine, 17:0 lyso-phosphocholine, 17:1 and “C followed by the total fatty acyl chain length:total number lyso-phosphoethanolamine, 17:1 lyso-phosphoserine, 17:0 of unsaturated bonds” and (eg, PC (C32:2)). D5_17:0 diacylglycerol, 17:0/17:0/17:0 triacylglycerol and C17:0 cholesteryl ester, 0.2 nmol each, from Avanti Polar Lipids) were added to 0.020 mL of serum. Samples were 2.21 | Statistical analysis vortexed and sonicated until they appeared dispersed. They were extracted at 48°C overnight. The samples were then Data are presented as mean ± SEM. For the statistical anal- evaporated and transferred to 1.5 mL Eppendorf tubes after yses, we used analyses of variance considering two factors | 15306860, 2020, 9, Downloaded from https://faseb.onlinelibrary.wiley.com/doi/10.1096/fj.202000678R by Readcube (Labtiva Inc.), Wiley Online Library on [15/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License WEBER et al.     11823 (two-way ANOVA: factor one was diet [NCD or HFD] as a control (AAV9-hCPT1AM and AAV9-GFP, respectively). and factor two was the virus injected [AAV-GFP or AAV- The genome of each virus contained the target gene under the hCPT1AM]), followed by Tukey's multiple comparison test. control of the liver-specific promoter EalbAATp, which con- For the lipidomic studies of mice samples, we used two-way strained the expression of hCPT1AM to this tissue (Figure S1A). ANOVA analysis, correcting for multiple comparisons by Eight-week old C57Bl/6J mice were fed either NCD or controlling the false discovery rate (FDR) (two-stage step-up HFD for 9 weeks. After that time, mice showed an obese phe- method of Benjamini, Kieger, and Yekutieli). Hierarchical notype characterized by a significant increase in circulating clustering was performed using R.26 For these analyses, pre- levels of glucose and insulin and higher body weight than processed lipidomics data were log-transformed and nor- the control NCD-fed group (Figure S1B-C). NCD- and HFD- malized between samples using limma package.27 To avoid fed mice were then administered 1.5 × 1012 gc/kg of AAV9- mathematical errors, before log-transformation zeros were GFP or AAV9-hCPT1AM by tail-vein injection (Figure 1A). substituted by 1/10th of the minimum value (other than zero) Mice were studied throughout the treatment and killed in the data matrices. Hierarchical unsupervised clustering at 11 weeks after AAV administration. At that time point, was performed using heatmap 3 package,28 applying com- most hepatocytes expressed GFP in both NCD- or HFD-fed plete linkage on the euclidean distances matrices. For the mice treated with AAV9-GFP, which indicated that the in- analysis of human lipidomic measurements, principal com- fection efficiency was similar in both groups (Figure S2A). ponent analyses (PCA) were performed using the function As expected, mRNA expression of hCPT1A was observed prcomp from R base after scaling the log-transformed data to in NCD-hCPT1AM and HFD-hCPT1AM mice but not in account for the large numerical differences among the mean their respective GFP controls (Figure S2B). In addition, values of the different lipidic molecules. UMAP analyses CPT1A protein levels were 2- and 1.7-fold higher in NCD- were performed on log-transformed data using R umap pack- hCPT1AM and HFD-hCPT1AM mice, respectively, than in age29 that implements the Uniform Manifold Approximation their GFP controls (Figure S2C). Note that the CPT1A anti- and Projection method.30 The lipid content change (ΔLC) body recognizes both endogenous CPT1A and the expressed was calculated for every lipid in obese patients as the dif- hCPT1AM protein. Concomitantly, CPT1 activity was 1.62- ference in the log-transformed content of that particular fold higher in NCD-hCPT1AM mice and 1.42-fold higher in lipid after surgery vs before surgery. Similarly, changes in HFD-hCPT1AM mice than in their respective GFP controls NAFLD-related parameters were calculated in obese patients (Figure 1B). To discard changes in mitochondrial content due as the difference in those parameters after surgery. Partial to hCPT1AM expression, we analyzed mitochondrial protein correlations between ΔLC and changes in NAFLD-related, TIM44 (translocase of the mitochondrial inner membrane) correcting for age, gender, and changes in BMI were calcu- levels in the liver of all mice. No differences were seen in lated using the package ppcor31 and visualized using corrplot mitochondrial content between groups (Figure S2D). CPT1A package.32 Essentially, the partial correlation analyses evalu- protein levels in other tissues with strong AAV9 tropism, ated the lineal regression equation “ΔNAFLD.variable ~age such as heart and skeletal muscle (hamstring), did not change +gender + ΔBMI + Δlipid.content,” where ΔNAFLD. after administration of AAV9-hCPT1AM (Figure S2E). variable was the change in the studied variable (ie, glucose, Consistent with increased CPT1 protein and activity, the AST, ALT, GGT, or ADT/ALT) after surgery, ΔBMI was levels of long-chain acylcarnitines, which are direct prod- the change in BMI after surgery and Δlipid.content was the ucts of CPT1A activity, were increased in the liver of hCP- change of the studied lipid after surgery. When required, T1AM-expressing mice compared to GFP-expressing controls P values were adjusted for multiple testing using the false (Figure 1C). Conversely, long-chain acylcarnitine concen- discovery method (FDR)33 implemented in the function trations in serum samples were similar among all groups p.adjust. When very strong correlation among tests was ex- (Figure 1C). Moreover, FAO was 32% higher in cultured pri- pected (for instance, in the partial correlation analyses), no mary hepatocytes from HFD-hCPT1AM mice than in HFD- further P value adjustment was applied to minimize type-II GFP controls (Figure 1D). The same was true in hepatocytes errors. Differences were considered significant at P < .05. from NCD-hCPT1AM mice, albeit to a lesser extent. 3 | R E S U LTS 3.2 | hCPT1AM expression in mouse liver reduces hepatic steatosis, hepatic 3.1 | AAV9-mediated expression of insulin resistance, hyperglycemia, and hCPT1AM in mouse liver increases FAO hyperinsulinemia We used AAVs to express a human mutated isoform of CPT1A HFD-hCPT1AM mice weighed significantly less than (hCPT1AM) or green fluorescent protein (GFP), which was used HFD-GFP control mice, whereas no differences were seen |    15306860, 2020, 9, Downloaded from https://faseb.onlinelibrary.wiley.com/doi/10.1096/fj.202000678R by Readcube (Labtiva Inc.), Wiley Online Library on [15/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 11824 WEBER et al. F I G U R E 1 Expression of hCPT1AM increases FAO in hepatocytes and improves metabolic parameters and decreases hepatic steatosis in HFD-fed obese mice. A, Experiment design and time course. B, CPT1A activity in isolated mitochondria from NCD- and HFD-fed mice expressing either GFP or hCPT1AM. C, Long-chain acylcarnitines were measured in liver and serum samples from GFP- and hCPT1AM-expressing mice as an indirect indicator of CPT1A activity. D, Total oleate oxidation in isolated hepatocytes from NCD- and HFD-fed mice expressing either GFP or hCPT1AM. E, Body weight (*P < .05, **P < .01, HFD-GFP vs HFD-hCPT1AM) of NCD- and HFD-fed mice after treatment with AAV9-GFP or AAV9-hCPT1AM. Fasting blood (F) glucose and (G) insulin levels NCD- and HFD-fed mice after treatment with AAV9-GFP or AAV9- hCPT1AM. H, Glucose tolerance test of NCD- and HFD-fed mice after treatment with AAV9-GFP or AAV9-hCPT1AM. I, Insulin signaling in isolated hepatocytes from NCD- and HFD-fed mice expressing either GFP or hCPT1AM, as indicated by Western blotting of insulin-induced Akt phosphorylation (pAkt, Ser473). Quantification of pAkt normalized by total Akt (N = 4). AU =arbitrary units. N = 5-6 (unless stated otherwise), *P < .05, **P < .01, ***P < .001 in NCD-hCPT1AM mice (Figure 1E and Table 1). The dif- and the obesity-induced diabetic phenotype observed in ferences in body weight were not attributable to changes in HFD-fed mice. food consumption, since daily rates of food intake were equal Liver weight and TAG content of HFD-hCPT1AM mice in GFP- and hCPT1AM-expressing mice fed either NCD or were lower than in HFD-GFP control mice (Figure 2A, HFD (Figure S3A). Fasting blood glucose and insulin levels Table 1 and Figure S3B). In addition, histopathological anal- were lower in HFD-hCPT1AM mice than in control HFD- ysis of liver samples confirmed that HFD-hCPT1AM mice GFP mice, and similar to levels observed in NCD-treated showed a reduction in the NAFLD activity score (NAS)34 mice (Figure 1F-G). Glucose tolerance (measured via in- (Table S1). Furthermore, hepatocellular ballooning was not traperitoneal GTT) was improved in HFD-hCPT1AM mice observed in any of the HFD-hCPT1AM mice (Table S1). compared to HFD-GFP control mice (Figure 1H). In addi- Of note, hCPT1AM expression did not significantly affect tion, pAkt levels were higher in cultured primary hepatocytes liver histological parameters in NCD-fed mice (Figure 2A from HFD-hCPT1AM mice than in those from HFD-GFP and Table S1). Finally, hepatic injury was lower in HFD- mice, as shown in Figure 1I. Overall, these results indicate hCPT1AM mice than in HFD-GFP controls, as indicated by that hCPT1AM expression improves hepatic insulin signaling lower serum levels of AST and ALT enzymes (Table 1). In | 15306860, 2020, 9, Downloaded from https://faseb.onlinelibrary.wiley.com/doi/10.1096/fj.202000678R by Readcube (Labtiva Inc.), Wiley Online Library on [15/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License WEBER et al.     11825 TABLE 1 Analyses of body weight, serum enzymes and metabolites, liver weight and liver triglyceride (TAG) content in mice NCD HFD AAV-GFP AAV-hCPT1AM AAV-GFP AAV-hCPT1AM # Body weight (g) 32.3 ± 0.86 32.7 ± 1.41 50.7 ± 0.99 45 ± 1.02* Total CHOL (mg/dL) 105.0 ± 7.58 107.8 ± 9.06 205.2 ± 6.52# 180.2 ± 7.36* TAG (mg/dL) 39.8 ± 2.97 42.2 ± 7.14 42.8 ± 1.35 38.8 ± 1.74 LDL CHOL (mg/dL) 29.8 ± 4.68 33.7 ± 4.32 117.2 ± 6.44# 89.3 ± 6.28* HDL CHOL (md/dL) 67.1 ± 2.91 65.6 ± 4.55 79.4 ± 0.96 82.6 ± 1.69 BHB (mM) 1.2 ± 0.06 1.1 ± 0.08 1.5 ± 0.09# 1.9 ± 0.10* ALT (U/L) 8.3 ± 1.66 4.2 ± 1.72 109.2 ± 22.50# 34 ± 5.75* AST (U/L) 35.2 ± 5.31 27.3 ± 2.22 110.4 ± 18.33# 67.2 ± 8.41* # Liver weight (g) 1.1 ± 0.04 1.1 ± 0.08 2.4 ± 0.12 1.5 ± 0.10* Liver TAG (mg/g liver) 2.2 ± 0.12 1.2 ± 0.06 48 ± 0.67# 29.4 ± 0.50* Note: Measures were performed in 28-week old mice, 11 weeks after injection of the AAV9s. Data are means ± SEM. n = 5-6. Abbreviations: ALT, alanine aminotransferase; AST, aspartate aminotransferase; BHB, β-hydroxybutyric acid; CHOL, total cholesterol; HDL, high-density lipoprotein cholesterol; HFD, high-fat diet; LDL, low-density lipoprotein cholesterol; NCD, normal chow diet. *P < .05 versus HFD-GFP; # P < .01 versus NCD-GFP. summary, these results indicate that hepatic hCPT1AM ex- oxidative phosphorylation, both uncoupled and coupled to pression decreases HFD-induced hepatic steatosis and injury. ATP synthesis. Moreover, HFD reduced the mitochondrial respiratory control ratio (RCR) in GFP-expressing mice but not in hCP- 3.3 | Increased liver FAO activates T1AM-expressing mice. This indicates higher respiratory mitochondrial energy metabolism coupling efficiency in liver mitochondria from HFD-fed hCPT1AM-expressing mice (Figures 2E and S3C). The Increased FAO due to hCPT1AM expression could produce protein levels of the OXPHOS complexes CV-ATP5A, CIV- a surplus of acetyl-CoA that can be metabolized by two MTCO1, CII-SDHB, and CI-NDUFB8 were similar in HFD- pathways: (1) conversion to ketone bodies and (2) entry to hCPT1AM and HFD-GFP control mice (Figure S3D). No the tricarboxylic acid (TCA) cycle to complete oxidation. alteration in the mRNA levels of the genes involved in mito- Oxidation of acetyl-CoA can be coupled to ATP synthesis chondrial dynamics was observed in any of the experimental or not, in which case uncoupling protein 2 (UCP2) allows groups (Figure S4A). the dissipation of energy as heat. First, we analyzed liver protein levels of mitochondrial hydroxymethylglutaryl- CoA synthase 2 (HMGCS2), the rate-limiting enzyme of 3.4 | hCPT1AM expression activates hepatic ketogenesis. HMGCS2 protein levels were higher autophagy in liver in hCPT1AM-expressing mice than in GFP control mice in both NCD- and HFD-fed groups (Figure 2B). In addition, Because the expression of hCPT1AM in the liver of HFD- serum levels of ketone bodies such as β-hydroxybutyric acid fed mice decreases the accumulation of lipid droplets (LD) (BHB) were higher in HFD-hCPT1AM mice than in HFD- (Table 1, Figures 2A and S3B), we next examined the two GFP control mice (Table 1). These results suggest that excess main catabolic pathways of LDs: lipolysis and autophagy. substrates in HFD-hCPT1AM mice promote the production First, we measured the levels of perilipin 2 and 5 (PLIN2 of ketone bodies. Second, we observed that temperature and PLIN5), which are key components of the LDs and con- and UCP2 protein levels were higher in the liver of HFD- sidered regulators of lipolysis.35 HFD-induced increase in hCPT1AM mice than in the HFD-GFP control group, which PLIN2 and PLIN5 protein levels was blunted in hCPT1AM- indicates an increase in energy dissipation (Figure 2B-C). expressing mice (Figure 3A). Next, we analyzed the mRNA Finally, liver ATP levels were higher in HFD-hCPT1AM levels of the genes Atgl, Hsl, and Mgl, which encode enzymes mice than in control HFD-GFP mice (Figure 2D), whereas involved in the lipolysis pathway. In all cases, values were no differences were observed in NCD-fed mice. Altogether, higher in HFD-hCPT1AM than in HFD-GFP control mice these results indicate that hCPT1AM expression increments (Figure S4B). However, no changes were observed when FAO in liver, resulting in an increase in ketogenesis and we analyzed the protein concentrations of the lipases ATGL |    15306860, 2020, 9, Downloaded from https://faseb.onlinelibrary.wiley.com/doi/10.1096/fj.202000678R by Readcube (Labtiva Inc.), Wiley Online Library on [15/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 11826 WEBER et al. F I G U R E 2 Expression of hCPT1AM decreases hepatic steatosis in HFD-fed obese mice. A, Liver macroscopic appearance (left panel), and histology (hematoxylin and eosin staining, scale bar is 50 μm) (right panel) in NCD- and HFD-fed mice after treatment with AAV9-GFP or AAV9-hCPT1AM. B, Protein levels of HMGCS2, the rate-limiting enzyme of hepatic ketogenesis, and UCP2 in liver of NCD- and HFD-fed mice expressing either GFP or hCPT1AM, as indicated by Western blotting. Quantification is normalized by β-actin (N = 3). C, Liver temperature as indicated by infrared thermography and D, liver ATP levels in NCD- and HFD-fed mice after treatment with AAV9-GFP or AAV9-hCPT1AM. E, Mitochondrial respiratory capacity in isolated mitochondria from NCD- and HFD-fed mice expressing either GFP or hCPT1AM. AU =arbitrary units. N = 5-6 (unless stated otherwise), *P < .05, **P < .01, ***P < .001 and the phosphorylated active form of HSL (Figure S4C). hCPT1AM expression in HFD-fed mice increases the expres- Altogether, these results suggest that the decrease in LD ob- sion of the machinery of autophagosome formation. served in HFD-hCPT1AM mice is independent of the LD- associated lipases activity. Recent studies have revealed a role for autophagy in LD 3.5 | hCPT1AM expression improves breakdown.36,37 Therefore, next we analyzed the expression cholesterol levels in HFD-fed mice of autophagy-related protein 7 (Atg7) and the protein levels of the LC3II/LC3I ratio and p62, autophagy biomarkers. ATG7 HFD-hCPT1AM mice showed considerably lower total and is a key protein in LC3II formation during the initial steps of LDL-cholesterol in serum than obese HFD-GFP control mice the autophagy process. We observed that Atg7 mRNA levels (Table 1). To determine which mechanisms might be involved were higher in HFD-hCPT1AM than in HFD-GFP control in the observed low serum cholesterol levels we analyzed: (1) mice, which indicates an increase in the machinery to activate mRNA levels for the key enzyme HMG-CoA reductase and autophagy (Figure 3B). Moreover, the LC3II/LC3I ratio was the transcription factor SREBP-2 (both involved in de novo higher and p62 protein levels were lower in HFD-hCPT1AM cholesterol synthesis), (2) protein levels of liver lipoprotein re- mice than in the HFD-GFP group (Figure 3C). Accumulation ceptors, and (3) mRNA levels for transporters involved in liver of LC3II in the presence of Chloroquine (CQ) indicates its cholesterol export. Our results showed no changes in HMG- flux through lysosomes. Of note, quantification of LC3II flux CoA reductase and SREBP-2 mRNA levels due to hCPT1AM revealed an induction of autophagy in cultured hepatocytes expression (Figure S4D). We next measured protein levels of from hCPT1AM-expressing mice when compared to GFP liver lipoprotein receptors involved in the import and export of controls (Figure 3D). Altogether, these results suggest that liver cholesterol. Specifically, we analyzed LDLr (low-density lipoprotein receptor), VLDLr (very-low-density lipoprotein | 15306860, 2020, 9, Downloaded from https://faseb.onlinelibrary.wiley.com/doi/10.1096/fj.202000678R by Readcube (Labtiva Inc.), Wiley Online Library on [15/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License WEBER et al.     11827 receptor), and LRP1 (low-density lipoprotein receptor-related than in NCD-GFP mice and did not change after the expres- protein). HFD-induced increase in LDLr protein levels was sion of hCPT1AM (Figure 3F). The mRNA levels for ABCA1 blunted in hCPT1AM-expressing mice (Figure 3E). No signifi- and ABCG5/G8 transporters, which are involved in choles- cant changes were observed in VLDLr and LRP1 protein levels. terol export, were lower in obese HFD-GFP mice. However, The mRNA levels for MTP (microsomal transporter protein), a hCPT1AM expression was sufficient to revert the decrease protein involved in ApoB formation, were lower in HFD-GFP caused by HFD (Figure 3F). Altogether, these results suggest |    15306860, 2020, 9, Downloaded from https://faseb.onlinelibrary.wiley.com/doi/10.1096/fj.202000678R by Readcube (Labtiva Inc.), Wiley Online Library on [15/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 11828 WEBER et al. F I G U R E 3 Expression of hCPT1AM in liver increases hepatic autophagy and modulates cholesterol metabolism in HFD-fed mice. A, Levels of PLIN2 and PLIN5 in liver of NCD- and HFD-fed mice expressing either GFP or hCPT1AM, as indicated by Western blotting. Quantification is normalized by β-actin. B, mRNA levels of Atg7 in the liver of NCD- and HFD-fed mice after treatment with AAV9-GFP or AAV9-hCPT1AM. C, Levels of LC3I/II and p62 in the liver of NCD- and HFD-fed mice expressing either GFP or hCPT1AM, as indicated by Western blotting. Quantification is normalized by β-actin. (N = 3). D, Quantification of net autophagy flux in isolated hepatocytes from NCD- and HFD-fed mice expressing either GFP or hCPT1AM, as indicated by Western blotting of CQ-induced LC3II accumulation. Quantification is normalized by β-actin (N = 2). E, Levels of LDLr, VLDLr, and LRP1 in the liver of NCD- and HFD-fed mice expressing either GFP or hCPT1AM, as indicated by Western blotting. Quantification is normalized by β-actin (N = 3). F, mRNA levels of Mtp, Abca1, Abcg5, and Abcg8 in the liver of NCD- and HFD-fed mice after treatment with AAV9-GFP or AAV9-hCPT1AM. AU =arbitrary units. CQ = chloroquine. N = 5-6 (unless stated otherwise), *P < .05, **P < .01, **P < .001 that some aspects of liver cholesterol metabolism are improved protein (all but one in the cluster in green) (Figure 4A). As in HFD-hCPT1AM mice, which could explain the reduction in expected, not all the lipid species that changed in the liver of total serum cholesterol levels observed in these mice. HFD-hCPT1AM mice were also changed in serum. As shown in Figure 4B, a subset of 24 lipids exhibited statistically sig- nificant differences between HFD-fed and NCD-fed mice 3.6 | Comparative liver and serum lipidomic (two-way ANOVA, P < .01 adjusted for multiple comparison analysis of AAV9-hCPT1AM-treated mice using the FDR method). Expression of hCPT1AM in HFD-fed shows a specific lipidomic signature to monitor mice reverted the levels of 9 of these lipid species to what is hepatic steatosis seen in NCD-fed mice (shown in red in Figure 4B). Among them, we found that ceramide Cer (C20:0) was lower in the Reliable noninvasive methods to monitor NAFLD remain serum of HFD-hCPT1AM mice than in that of HFD-GFP limited. Here, we aimed to uncover serum lipid biomarkers control mice (Figure 4B). Several species of glycerophospho- for noninvasive clinical detection and monitoring of hepatic lipids were also reduced in HFD-hCPT1AM serum, including steatosis. We performed a comparative lipidomic study in phosphatidylcholines (PC (C34:1, C36:3, C36:4, and C38:6)), liver and serum samples from NCD- and HFD-fed mice ex- plasmalogen (pPC (C32:0)), and lysophosphatidylethanola- pressing either GFP or hCPT1AM. We analyzed 66 species mine (LPEA (C18:0)). Finally, only a few neutral lipids (DAG of sphingolipids, 16 species of glycerophospholipids, and 94 (C32:5) and TAG (C50:1)) showed significantly lower levels species of neutral lipids by electrospray ionization mass spec- in the serum of HFD-hCPT1AM mice than in that of HFD- trometry (ESI-MS). In the liver, a subset of 42 lipids exhib- GFP control mice. As before, the clustering analysis was per- ited statistically significant differences among NCD-GFP and formed using complete linkage on the euclidean distances of HFD-GFP groups (two-way ANOVA P < .01 adjusted for the log-transformed values of the selected lipids. In this case, multiple comparisons using FDR) (Figure 4A). Noteworthy, the analysis yielded three main clusters, and a fourth clus- expression of hCPT1AM in HFD-fed mice restored the levels ter containing a single animal (HFD-GFP 2), most likely an of 30 of these lipid species to what is seen in NCD-fed mice outlier. In general, the differences among groups were not as (shown in red in Figure 4A). These include ceramides (Cer pronounced as those found when studying liver lipids, and the (C16:0, C18:0, C20:0, and C22:0)), dihydroceramides (dhCre separation between HFD-hCPT1AM and HFD-GFP animals (C22:0)) and phospholipids, specifically lysophosphatidyl- was not as definite (Figure 4B). Altogether, this compara- cholines (LPC (C16:0 and C18:0)). However, the most sig- tive lipidomic analysis revealed that some lipid species (Cer nificant changes were observed in the levels of neutral lipids, (C20:0) and TAG (C50:1)) increase in the liver and serum of such as DAG and TAG. While the livers of HFD-GFP mice HFD-fed mice, yet, the expression of hCPT1AM in the liver show a vast increase in the levels of DAG and TAG, specifi- is sufficient to partially or totally revert this increase both in cally those with a long-chain fatty acyl group, the expression the tissue and in serum (Figure 4C-D). Therefore, we hypoth- of hCPT1AM in HFD-fed mice was sufficient to reverse the esized that these lipid species might be potential circulating accumulation of these lipid species (Figure 4A). The cluster- biomarkers to monitor the amelioration of NAFLD during a ing analysis was performed using complete linkage method on therapeutic intervention. the euclidean distances of the log-transformed values of the 42 liver lipid biomarkers that showed statistically significant differences based on the type of diet. Accordingly, the clas- 3.7 | Serum lipidomic switch after bariatric sification provided by the clustering analysis mainly reflected surgery in obese humans the difference in diet, albeit it also clearly showed differences between the HFD-fed animals expressing hCPT1AM (all of Lipidomic analyses in mice highlighted interesting potential them in the cluster in red) vs those expressing the control GFP biomarkers for monitoring human NAFLD. To determine the | 15306860, 2020, 9, Downloaded from https://faseb.onlinelibrary.wiley.com/doi/10.1096/fj.202000678R by Readcube (Labtiva Inc.), Wiley Online Library on [15/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License WEBER et al.     11829 F I G U R E 4 Lipidomic analyses of NCD- and HFD-fed mice after treatment with AAV9-GFP or AAV9-hCPT1AM. Lipidomic analyses of NCD- and HFD-fed mice after treatment with AAV9-GFP or AAV9-hCPT1AM. Unsupervised clustering of 18 mice (in columns) based on the quantification of (A) 42 lipids (in rows) in liver samples and (B) 24 lipids (in rows) in serum samples that exhibited statistically significant differences among NCD-GFP and HFD-GFP groups (two-way ANOVA P < .01 adjusted for multiple comparisons using FDR). The unsupervised clustering was performed using complete linkage on the Euclidean distance among samples. The heatmap colors represent the values scaled by the standard deviation of every lipid value across all samples and centered on the average of the NCD-GFP mice. Blue cells indicate values below the average value of the NCD-GFP mice, while red cells indicate values above that average, in z-scores. The lipid species highlighted in red exhibited statistically significant differences among HFD-GFP and HFD-hCPT1AM groups. Concentration of Cer (C20:0) and TAG (C50:1) (expressed in pmol equivalent) in liver (C) and serum (D) samples from NCD- and HFD-fed mice expressing either GFP or hCPT1AM (n = 4-6, **P < .01, ***P < .001, ****P < .0001). Cer: ceramide; SM: sphingomyelin; GlucCer: glucosylceramide; PC: phosphatidylcholine; PE: phosphatidylethanolamine; PS: phosphatidylserine; DAG: diacylglycerol; TAG: triacylglycerol. Lipid species are annotated using the “lipid subclass” and “C followed by the total fatty acyl chain length:total number of unsaturated bonds” and (eg, PC (C32:2)). If the sphingoid base residue was dihydrosphingosine, the name contains a “dh” prefix. Plasmalogens and lysophospholipids are annotated as described above, except that a “p” or “L” prefix is included, respectively clinical relevance of our findings, we performed a compara- with NAFLD by liver biopsy (Table 3) and did not receive tive serum lipidomic analysis of 15 lean control subjects and any antidiabetic treatment before surgery. The only patient 15 obese patients, before and 6 months after Roux-en-Y gas- on statin medication did not require lipid-lowering therapy tric bypass (RYGB) (Table 2). Obese patients were diagnosed 6 months after surgery. After RYGB, obese patients showed |    15306860, 2020, 9, Downloaded from https://faseb.onlinelibrary.wiley.com/doi/10.1096/fj.202000678R by Readcube (Labtiva Inc.), Wiley Online Library on [15/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 11830 WEBER et al. TABLE 2 Clinical characteristics of the subjects included in the study P (lean vs obese P (obese pre- vs Lean Obese Pre-surgery Obese Post-surgery pre-surgery) post-surgery) N 15 15 15 – – Sex (male/female) 4/11 4/11 4/11 .659 .659 Age (years) 42 ± 3 45 ± 3 46 ± 3 .470 <.0001 Height (m) 1.66 ± 0.02 1.64 ± 0.02 1.64 ± 0.02 .606 .334 Weight (kg) 63 ± 3 117 ± 6 79 ± 5 <.00001 <.00001 BMI (kg/m2) 22.7 ± 0.8 43.0 ± 1.2 29.1 ± 1.0 <.00001 <.00001 Body fat (%) 28.7 ± 2.4 50.8 ± 1.3 36.1 ± 2.0 <.00001 <.00001 Waist circumference 80 ± 3 120 ± 4 95 ± 3 <.00001 <.00001 (cm) Glucose (mg/dL) 84 ± 2 104 ± 3 90 ± 8 <.00001 <.00001 - Glucose 2-h OGTT 95 ± 4 157 ± 8 <.00001 – (mg/dL) Insulin (μU/mL) 5.8 ± 0.9 19.2 ± 2.0 6.9 ± 1.0 <.00001 <.00001 - Insulin 2-h OGTT 44.9 ± 7.0 163.8 ± 24.4 .002 – (μU/mL) HOMA 1.2 ± 0.2 5.0 ± 0.6 1.6 ± 0.2 <.00001 <.00001 QUICKI 0.39 ± 0.01 0.31 ± 0.01 0.37 ± 0.01 <.00001 <.00001 Triacylglycerols 78 ± 8 155 ± 17 83 ± 9 <.00001 .002 (mg/dL) Total cholesterol 192 ± 6 222 ± 9 165 ± 10 .009 .623 (mg/dL) LDL-cholesterol 109 ± 6 136 ± 8 102 ± 6 .007 .041 (mg/dL) HDL-cholesterol 68 ± 4 51 ± 5 54 ± 3 .014 .050 (mg/dL) CRP (mg/L) 2.0 ± 0.6 10.2 ± 2.0 2.9 ± 0.9 .003 .012 Uric acid (mg/dL) 4.3 ± 0.2 6.2 ± 0.4 5.0 ± 0.4 <.00001 .032 Leptin (ng/mL) 11.0 ± 2.2 48.6 ± 6.3 12.7 ± 3.1 <.00001 <.00001 AST (IU/L) 13 ± 1 18 ± 2 15 ± 1 .047 .057 ALT (IU/L) 9±1 29 ± 5 22 ± 5 .001 <.00001 Alkaline 85 ± 4 100 ± 8 77 ± 10 .097 .132 phosphatase (IU/L) γ-GT (IU/L) 12 ± 2 32 ± 9 17 ± 6 .042 <.00001 TSH (μU/mL) 1.7 ± 0.3 3.2 ± 1.3 – .396 – Daily alcohol intake 2.1 ± 1.1 0.9 ± 0.7 – .405 – (g) Antihypertensive 1 (6.7%) 6 (40.0%) 1 (6.7%) .060 .019 therapy, n (%) Antidiabetic therapy, 0 (0%) 0 (0%) 0 (0%) 1.000 1.000 n (%) Lipid-lowering 0 (0%) 1 (6.7%) 0 (0%) .536 .334 therapy, n (%) Note: Bold values denote statistically significant P values. Differences between lean subjects and patients with obesity before surgery were analyzed by Student's t test or χ2 test, where appropriate. The impact of weight loss in patients with obesity before and 6 months after Roux-en-Y gastric bypass (RYGB) was analyzed by paired two-tailed Student's t test. n = 15. Abbreviations: ALT, alanine aminotransferase; AST, aspartate aminotransferase; BMI, body mass index; CRP, high-sensitivity C-reactive protein; HOMA, homeostasis model assessment; IGT, impaired glucose tolerance; NG, normoglycemia; QUICKI, quantitative insulin sensitivity check index; T2D, type 2 diabetes; TSH, thyroid-stimulating hormone; γ-GT, γ-glutamyltransferase. | 15306860, 2020, 9, Downloaded from https://faseb.onlinelibrary.wiley.com/doi/10.1096/fj.202000678R by Readcube (Labtiva Inc.), Wiley Online Library on [15/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License WEBER et al.     11831 T A B L E 3 Pathology analysis of liver biopsies obtained from (P = 2.2 × 10-10, Chi-squared test). Notably Cluster 3, com- patients with morbid obesity and NAFLD prising most post-surgery individuals, was closer to Cluster 1 Obese NAFLD (mostly composed of lean control subjects) than to Cluster 2 Pathological characteristics (%) (mostly obese pre-surgery individuals). These three clusters Steatosis were not associated with gender (P = .55, Chi-squared test) <5% 27 or age (P = .71, ANOVA). We also performed two complementary dimension reduc- 5%-33% 53 tion analyses, a classical principal component analysis (PCA) 34%-66% 13 and a more sophisticated Uniform Manifold Approximation >66% 7 and Projection (UMAP), which is a nonlinear dimension Lobular inflammation reduction tool. In line with the results shown in Figure 4A, No foci 33 the dimension reduction analyses showed that the serum <2 foci, per 200X field 13 lipidomics values of the obese post-surgery patients were 2-4 foci, per 200X field 47 closer to those of lean controls than to the patients before >4 foci, per 200X field 7 surgery (Figure 5B). In the PCA, there was a very large dif- Portal inflammation ference in the first principal component mean value of obese pre-surgery patients with lean controls (difference = 10, P- None to minimal 27 adjusted = 4.7 × 10−7) and obese post-surgery patients (dif- Greater than minimal 73 ference = 9.2, P-adjusted = 4.6 × 10−6), but not between Fibrosis stage lean and obese post-surgery patients (difference = 0.8, P- None 60 adjusted = .87). Similarly, the first dimension of the UMAP Perisinusoidal or periportal/portal 13 (UMAP 1) revealed a significant difference between obese Perisinusoidal and portal/periportal 20 pre-surgery individuals and lean controls (difference = 2.1, Bridging fibrosis 7 P-adjusted = 1.7 × 10-6), as well as with obese post-surgery Cirrhosis 0 individuals (difference = 3.1, P-adjusted = 1.5 × 10-9), but Hepatocyte ballooning not between lean and obese post-surgery individuals (differ- None 60 ence = 0.9, P-adjusted = 0.044). Next, we evaluated the correlation between changes in Few balloon cells 33 serum lipid profiles and NAFLD variables, adjusting for Many prominent balloon cells 7% changes in BMI. First, we calculated the serum lipid content Mallory's hyaline change (ΔLC = content after surgery-content before surgery) None to rare 100 and its correlation with BMI change (ΔBMI) in obese pa- Many 0 tients subjected to RYGB surgery. Fifty-five of the lipids ana- Glycogenated nuclei lyzed exhibited statistically significant changes after surgery None to rare 100 (P < .05, paired t test corrected for multiple comparisons), Many 0 and only eight showed a significant correlation with changes Anatomopathological diagnosis in BMI (P < .05, correlation test corrected for multiple com- parisons) (Figure 5C). Notably, there was no overlap between NAFLD 60 the set of lipids with significant changes after surgery and NASH 40 those that best correlated with changes in BMI. In general, lipids with larger ΔLC showed weak correlation with ΔBMI lower weight and BMI (from 43.0 ± 1.2 to 29.1 ± 1 kg/m2), and, conversely, lipids whose ΔLC strongly correlated with decreased leptin levels and overall improved clinical bio- ΔBMI had very modest ΔLC. This observation suggests that chemistry (Table 2). We analyzed 46 species of sphingolip- the most significant changes in serum lipid content do not ids, 40 species of glycerophospholipids, and 50 species of directly reflect the loss of BMI. neutral lipids (MAG, DAG, TAG, and cholesterol esters) by When we compared the lipidomic analysis in mice serum ESI-MS in serum from lean control subjects and obese pa- and liver and human serum before and after therapeutic in- tients before and after RYGB. terventions such as AAV9-hCPT1AM administration and A subset of 26 lipids exhibited statistically significant RYGB surgery, respectively, we noticed that TAG (C50:1) differences among groups (ANOVA, P < .01 adjusted for was increased in obese/NAFLD mice and humans but re- multiple comparison using the FDR method) (Figure 5A). covered after hCPT1AM expression in mice and RYGB Unsupervised clustering analyses yielded three clus- surgery in humans (Figures 4D and 5D). As mentioned ters with a very strong association with the sample type above, changes in TAG (C50:1) after bariatric surgery did |    15306860, 2020, 9, Downloaded from https://faseb.onlinelibrary.wiley.com/doi/10.1096/fj.202000678R by Readcube (Labtiva Inc.), Wiley Online Library on [15/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 11832 WEBER et al. not correlate with changes in BMI. Therefore, the putative 4 | DISCUSSION correlation of TAG and NAFLD score is likely to be inde- pendent. Unfortunately, some parameters for calculating the Here, we describe a partial reversion of an already estab- NAFLD score were not available. Nonetheless, we evaluated lished obese and diabetic phenotype through an enhanced the association between ΔLC and NAFLD-related parame- FAO therapeutic intervention. AAVs have become a focus ters, such as glucose levels, the AST/ALT ratio, GGT levels, of attention in the field, especially in clinical-stage experi- ALT levels, and AST levels (Figure 5E), correcting for age, mental therapeutic strategies, and are considered one of the gender, and changes in BMI. As shown in Figure 5E, changes safest strategies for gene therapy.38 Two key factors of the in TAG (C50:1) after bariatric surgery show a partial correla- present study are the use of the AAV9 serotype which ef- tion with changes in NAFLD-related parameters. Altogether, ficiently and specifically targets hepatocytes and the use of these results highlight TAG (C50:1) as a potential circulating the human isoform of CPT1AM (hCPT1AM) with the pros- biomarker to monitor the amelioration of NAFLD during a pect of future clinical therapeutic applications. Thus, here we therapeutic intervention. describe that AAV9-mediated hCPT1AM-expressing mice | 15306860, 2020, 9, Downloaded from https://faseb.onlinelibrary.wiley.com/doi/10.1096/fj.202000678R by Readcube (Labtiva Inc.), Wiley Online Library on [15/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License WEBER et al.     11833 F I G U R E 5 Lipidomic analyses in humans. A, Unsupervised clustering of 42 donors (in columns) based on the quantification of 26 lipids (in rows) that exhibited statistically significant differences among groups (ANOVA P < .01 adjusted for multiple comparisons using FDR). The unsupervised clustering was performed using complete linkage on the Euclidean distance among samples. The heatmap colors represents the values scaled by the standard deviation of every lipid values across all samples and centered on the average of the lean individuals. Blue cells indicate values below the average value of the lean individuals, while red cells indicate values above that average, in z-scores. B, (Left) Principal component analysis (PCA) of 42 donors based on the quantification of 141 lipids. To account for the large numerical differences among lipidic molecules, values were scaled to have unit variance before the PCA analysis. In blue, lean individuals (L); in yellow, obese patients pre-surgery (O); in orange the same patients post-surgery (OP). The vertical lines indicate the mean value on the first principal component (PC1, explaining around 27% of the total variance) for the three groups, using the same color code. (Right) Uniform manifold approximation and projection (UMAP) of the same data. Color code as in the figure on the left. C, (Left) Changes in serum lipid content (ΔLC, in x-axis) and its correlation with changes in BMI (y-axis). Every dot represents a lipid. ΔLC is the average difference in lipid content between pre- and post-surgery, measured in 12 patients with complete data. To account for large variability in the average value of different lipids, the t-Statistic values rather than the mean difference values are shown in the graph. Negative ΔLC t-Statistic values indicate lipids whose average value decreased after surgery. Conversely, positive values indicate lipids that increased after surgery. Dot sizes are proportional to the FDR-adjusted P value. In red, lipids with FDR-adjusted P value < .05. (Right) Same plot than on the left, but dot sizes are proportional to the FDR-adjusted P value of the correlation test. In red, those with FDR-adjusted P value < .05. D, Concentration of TAG (C50:1) (expressed in pmol equivalent) in sera samples from lean controls and obese patients before and after bariatric surgery (N = 15, ****P < .0001). E, Partial correlations of changes in lipid content with changes in NAFLD- related parameters, after selecting lipids that exhibited at least one of the five partial correlations with a P value < .05 (dot size is proportional to the magnitude of the correlation. In red, negative partial correlations. In blue positive partial correlations). F, Schematic representation of how AAV9-hCPT1AM reverses obesity-induced hepatic steatosis and the lipidomic strategy followed to identify TAG (C50:1) as a possible circulating biomarker for NAFLD. Cer: ceramide; SM: sphingomyelin; HexCer: hexosylceramides; PC, phosphatidylcholine; LPC: lysophosphatidylcholine; PE: phosphatidylethanolamine; LPEA: lysophosphatidylethanolamine; PC O-: ether-linked phosphatidylcholine; DAG: diacylglycerol; TAG: triacylglycerol; CE: cholesterol ester. Lipid species are annotated using the “lipid subclass” and “C followed by the total fatty acyl chain length:total number of unsaturated bonds” and (eg, PC (C32:2)) under HFD showed increased liver FAO and, importantly, of UCP2, might attenuate mitochondrial ROS production and reduced NAFLD and improved glucose tolerance and hepatic protect against ROS-related cellular damage.44-48 We hypoth- insulin signaling compared to HFD-fed control mice. In addi- esized that an increase in mitochondrial FAs import in HFD- tion, hepatic hCPT1AM-expression in HFD-fed mice is suf- hCPT1AM mice might trigger a mild uncoupling intended to ficient to reduce potentially toxic lipid intermediates in the mitigate ROS production. liver, such as DAG and ceramides, that can activate protein We examined the molecular mechanisms underlying kinase C (PKCƸ) resulting in an impairment in the insulin the beneficial effects of hCPT1AM expression. Autophagy signaling pathway.39–42 These reduced ceramides and DAGs and lipolysis are two pathways implicated in the mobiliza- levels could explain the enhanced hepatic insulin signaling tion of lipid droplets (LDs) that are compromised after the observed in HFD-hCPT1AM mice. Notably, hCPT1AM ex- onset of T2D and in genetic and diet-induced models of obe- pression in NCD-fed mice did not significantly affect the he- sity.36,49–52 In HFD-hCPT1AM mice, increased FAO results patic and metabolic endpoints that were examined. in a remarkable reduction of hepatic LDs. Consistently, levels The expression of hCPT1AM does not have any notice- of the structural proteins PLIN2 and PLIN5 were also low in able deleterious effect in the liver, yet, it reverses hepatic HFD-hCPT1AM mice. However, protein analyses indicated injury as indicated by the lower serum levels of AST and that expression of hCPT1AM does not result in the activation ALT enzymes observed in HFD-hCPT1AM mice. However, of LD-associated lipases. Instead, the autophagy pathway ap- it has been suggested that excessive FAO in muscle might pears to be activated in the liver of HFD-hCPT1AM mice. lead to mitochondria overload and cell toxicity.43 Here, we Importantly, recent studies have revealed a role for autophagy show that in HFD-hCPT1AM mice, three mechanisms could in hepatic LD breakdown.37 Here, we propose that this mech- prevent possible oxidative stress caused by the accumulation anism could contribute to the reduction of hepatic lipid stores of FAO metabolites in the mitochondria. First, the surplus in HFD-hCPT1AM mice. of acetyl-CoA generated from FAO activates the ketogenic HFD-hCPT1AM mice showed lower levels of serum pathway. Second, part of the acetyl-CoA enters the TCA LDL-cholesterol than HFD-GFP controls. Given that one of cycle for complete oxidation and ATP synthesis. Third, part the main determinants of VLDL synthesis is hepatic TAG of the energy produced is dissipated as heat. Nonetheless, the content, which is significantly reduced in HFD-hCPT1AM parallel increase in mitochondrial uncoupled and coupled mice, we speculate that secretion of VLDL might be reduced respiration observed after hCPT1AM expression might seem in HFD hCPT1AM-expressing mice. Consequently, the de- paradoxical. One possible explanation could lie in previous rived LDL levels are lower. Moreover, hCPT1AM expression reports showing that mild uncoupling, caused by activation in the liver restored some aspects of cholesterol metabolism, |    15306860, 2020, 9, Downloaded from https://faseb.onlinelibrary.wiley.com/doi/10.1096/fj.202000678R by Readcube (Labtiva Inc.), Wiley Online Library on [15/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 11834 WEBER et al. such as protein levels of lipoprotein transporters and mRNA our study did not investigate the role of CPT1A in human levels of cholesterol transporters. NAFLD or in the observed changes in serum lipidomic pro- There is great interest in finding new circulating biomark- files after RYGB. Nonetheless, previous reports have shown ers for NAFLD/NASH diagnosis and monitoring of treatment that liver CPT1A expression is remarkably decreased in bi- effectiveness. While some studies have analyzed specific opsy samples from NAFLD patients.60,61 This observation lipid profiles in models of NAFLD, obesity, and T2D,53–55 indicates that decreased β-oxidation in human NAFLD could only a few studies have focused on changes of lipid species contribute to fatty acid accumulation in hepatocytes. This in liver and serum under interventions that improve hepatic suggests that our preclinical observations in mice could be steatosis and obesity. Here, we performed a comparative lip- relevant for future human research. Overall, the strength of idomic analyses in mice and humans that aimed to look for our lipidomic study is based on the comparison of mice and potential noninvasive markers to monitor the progression human samples of hepatic steatosis, before and after thera- of NAFLD in pathological conditions (ie, obesity) before peutic intervention. This strategy allows us to look for spe- and after therapeutic interventions intended to reduce body cific, robust serum biomarkers that reflect the progression of weight and hepatic steatosis (ie, AAV9-mediated hCPT1AM NAFLD and the effectiveness of a therapeutic intervention. expression in mice and Roux-en-Y gastric bypass in humans). In the future, the hypothesized candidate biomarker should In mice, our results show that TAG with a longer, unsatu- be validated in a larger cohort of participants before it is used rated chain of FAs (TAG (C50:1/2/3), TAG (C52:3/4/5), and for clinical purposes. TAG (C54:1/4/5/6)) increased in the liver of HFD-fed mice. In summary, here we report an efficient gene therapy However, the expression of hCPT1AM decreases the concen- approach to greatly reduce liver steatosis and body weight tration of these lipid species back to what is seen in NCD- and to improve glucose tolerance and hepatic insulin signal- fed mice. Interestingly, one of these species, TAG (C50:1), ing in a mouse model of HFD-induced obesity (Figure 5F). follows the same pattern in the serum of HFD-hCPT1AM, Moreover, we shed light on the molecular mechanisms in- which suggests that this TAG could be a circulating bio- volved in the beneficial effects of increasing FAO in hepato- marker of hepatic steatosis. These results are in line with a cytes. Finally, based on a comparative lipidomic analysis of previous study showing that treatment of HFD-fed mice with mouse and human samples, we hypothesize a candidate cir- compound K-enriched with ginsenosides, which decreases culating biomarker to monitor obesity and hepatic steatosis hepatic steatosis in treated mice, also reduced serum levels and the effectiveness of therapeutic interventions (Figure 5F). of TAG species with a total number of carbons above 50.56 The future directions of this work should include long-term Remarkably, in the serum of patients with obesity and experiments (1 year or more) in mice, rats, and higher animal NAFLD, similar TAG species (TAG (C50:1/2/3) and TAG species (monkeys) to ratify the effectiveness of the gene ther- (C52:3/4)) are reduced 6 months after bariatric surgery. apy and identify potential long-term side effects or cytotoxic- These results are in line with previous studies reporting a ity. Moreover, studies in larger human cohorts would confirm reduction in TAG (C50:1), among other TAGs, in plasma/ the prospective clinical applicability of the suggested cir- serum of obese patients after RYGB interventions.57,58 culating biomarker. Altogether, our results point toward the Conceivably the mechanisms behind TAG changes might be human isoform of CPT1A as a potential therapeutic target different in humans and in hCPT1AM-expressing mice. In against NAFLD and obesity-induced disorders. HFD-hCPT1AM mice the increase in hepatic FAO mobilizes stored lipid droplets and reduces the number of substrates ACKNOWLEDGMENTS for hepatic TAG synthesis. In humans, we can speculate that We thank Gloria González-Aseguinolaza for kindly sup- decreased intestinal fat absorption and altered hormone lev- plying the EalbAATp promoter. Financial support state- els might reduce hepatic lipid accumulation. In our study, ment: This study was supported by the Ministry of Spain changes in many of these lipid content in the serum of obese (MINECO) (SAF2014-52223-C2-1-R to DS, SAF2017- patients after bariatric surgery correlate with changes in 83813-C3-1-R to DS and LH, SAF2017-82813-C3-3R to NAFLD-related parameters and do not directly reflect the NC co-financed by the European Regional Development loss of BMI. Yet, the decrease in circulating TAG (C50:1) Fund [ERDF], a doctoral fellowship to JFM, and a Juan after RYGB could be mediated by other parameters beyond de la Cierva—Incorporación research fellowship [IJCI- hepatic steatosis. Nonetheless, our results in mice and human 2016-28313] to PM and DS), the Centro de Investigación samples support speculation on its possible role as a noninva- Biomédica en Red de Fisiopatología de la Obesidad y sive biomarker for human NAFLD diagnosis and prognosis. la Nutrición (CIBEROBN) (CB06/03/0001 to DS), the Additionally, we cannot reject the possibility that other cir- Government of Catalonia (2014SGR465 to DS), Instituto de culating lipid species that changed in our human intervention Salud Carlos III (FISPI18/01584 to VL-C, FIS-PI17/00412 study, as well as in other published paper 57–59 could be cir- to RB, FIS PI18/01484 to SA and FIS17/01455 to culating biomarkers for NAFLD. It is important to note that DS-I, Miguel Servet Fund CP15/00106 to DS-I and Rio | 15306860, 2020, 9, Downloaded from https://faseb.onlinelibrary.wiley.com/doi/10.1096/fj.202000678R by Readcube (Labtiva Inc.), Wiley Online Library on [15/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License WEBER et al.     11835 Hortega grant (CM19/00039) to CM) and co-financed 6. Ferré P. The biology of peroxisome proliferator-activated recep- by the European Regional Development Fund (ERDF), tors: relationship with lipid metabolism and insulin sensitivity. Diabetes. 2004;53(Suppl 1):S43-50. Centro de Investigación Biomédica en Red de Diabetes 7. Zhang D, Christianson J, Liu Z-X, et al. Resistance to high-fat y Enfermedades Metabólicas Asociadas (CIBERDEM) to diet-induced obesity and insulin resistance in mice with very JC-EG, the Fundació La Marató de TV3 (201627-30 to DS long-chain acyl-CoA dehydrogenase deficiency. Cell Metab. and 201627-31 to NC), the European Foundation for the 2010;11:402-411. Study of Diabetes (EFSD)/Lilly and EFSD/Janssen-Rising 8. Stefanovic-Racic M, Perdomo G, Mantell BS, Sipula IJ, Brown Star and L'Oréal-UNESCO “For Women in Science” re- NF, O’Doherty RM. A moderate increase in carnitine palmi- search fellowships to LH, a doctoral fellowship from the toyltransferase 1a activity is sufficient to substantially reduce hepatic triglyceride levels. Am J Physiol Endocrinol Metab. Ciência sem Fronteiras-CNPq (237976/2012-9) to MW. 2008;294:E969-E977. Figure 4F was created with BioRender.com. 9. Morillas M, Gómez-Puertas P, Bentebibel A, et al. Identification of conserved amino acid residues in rat liver carnitine palmito- CONFLICT OF INTEREST yltransferase I critical for malonyl-CoA inhibition. Mutation of The material reported in this manuscript is original and has methionine 593 abolishes malonyl-CoA inhibition. J Biol Chem. not been previously published or submitted for publication 2003;278:9058-9063. elsewhere. All authors approve the manuscript and declare 10. Orellana-Gavaldà JM, Herrero L, Malandrino MI, et al. Molecular no conflict of interest. therapy for obesity and diabetes based on a long-term increase in hepatic fatty-acid oxidation. Hepatology. 2011;53:821-832. 11. Monsénégo J, Mansouri A, Akkaoui M, et al. Enhancing liver mi- AUTHOR CONTRIBUTIONS tochondrial fatty acid oxidation capacity in obese mice improves M. Weber, P. Mera, R. Bartrons, A. Zorzano, V. Llorente- insulin sensitivity independently of hepatic steatosis. J Hepatol. Cortes, G. Frühbeck, N. Casals, L. Herrero, and D. Serra 2012;56:632-639. designed research and conceptualized the study. M. Weber, 12. Brunt EM. Nonalcoholic steatohepatitis. Semin Liver Dis. P. Mera, F. Casas, D. Sebastián, M. C. Soler-Vázquez, S. 2004;24:3-20. Recalde, R. Fucho, M. Calderon-Dominguez, J. F. Mir, and 13. Ezquerro S, Mocha F, Frühbeck G, et al. Ghrelin reduces TNF- J. C. Escola-Gil performed research. M. Weber, P. Mera, α-induced human hepatocyte apoptosis, autophagy and pyropto- sis: role in obesity-associated NAFLD. J Clin Endocrinol Metab. S. Alonso, and D. Sánchez-Infantes prepared images and 2018;104(1):21-37. https://doi.org/10.1210/jc.2018-01171 graphical data. M. Weber, P. Mera, F. Casas, A. Rodríguez, 14. Matthews DR, Hosker JP, Rudenski AS, Naylor BA, Treacher DF, S. Alonso, J. C. Escola-Gil, and D. Sánchez-Infantes ana- Turner RC. Homeostasis model assessment: insulin resistance and lyzed data. P. Mera, L. Herrero, and D. Serra wrote the β-cell function from fasting plasma glucose and insulin concentra- paper. J. Salvador, A. Rodríguez, G. Frühbeck, and V. tions in man. Diabetologia. 1985;28(7):412-419. Valentí contributed clinical samples. Supervision, pro- 15. Katz A, Nambi SS, Mather K, et al. Quantitative insulin sensitivity ject administration, and funding acquisition: L. Herrero, check index: a simple, accurate method for assessing insulin sensi- tivity in humans. J Clin Endocrinol Metab. 2000;85(7):2402-2410. D. Serra. P. Mera, L. Herrero, and D. Serra had primary https://doi.org/10.1210/jcem.85.7.6661 responsibility for final content. All authors read and ap- 16. Kramer MG, Barajas M, Razquin N, et al. In vitro and in vivo proved the final manuscript. comparative study of chimeric liver-specific promoters. Mol Ther. 2003;7:375-385. R E F E R E NC E S 17. Donello JE, Loeb JE, Hope TJ. Woodchuck hepatitis virus con- 1. Younossi ZM, Koenig AB, Abdelatif D, Fazel Y, Henry L, Wymer tains a tripartite posttranscriptional regulatory element. J Virol. M. Global epidemiology of nonalcoholic fatty liver disease-Me- 1998;72:5085-5092. ta-analytic assessment of prevalence, incidence, and outcomes. 18. Fraser F, Corstorphine CG, Zammit VA. Topology of carnitine pal- Hepatology. 2016;64:73-84. mitoyltransferase I in the mitochondrial outer membrane. Biochem 2. Abu-Elheiga L, Oh W, Kordari P, Wakil SJ. Acetyl-CoA carbox- J. 1997;323:711-718. ylase 2 mutant mice are protected against obesity and diabetes 19. Lee K, Kerner J, Hoppel CL. Mitochondrial carnitine palmito- induced by high-fat/high-carbohydrate diets. Proc Natl Acad Sci yltransferase 1a (CPT1a) is part of an outer membrane fatty acid USA. 2003;100:10207-10212. transfer complex. J Biol Chem. 2011;286:25655-25662. 3. An J, Muoio DM, Shiota M, et al. Hepatic expression of malo- 20. Morillas M, Gómez-Puertas P, Roca R, et al. Structural model of nyl-CoA decarboxylase reverses muscle, liver and whole-animal the catalytic core of carnitine palmitoyltransferase I and carnitine insulin resistance. Nat Med. 2004;10:268-274. octanoyltransferase (COT): Mutation of CPT I histidine 473 and 4. Ishigaki Y, Katagiri H, Yamada T, et al. Dissipating excess energy alanine 381 and COT alanine 238 impairs the catalytic activity. J stored in the liver is a potential treatment strategy for diabetes asso- Biol Chem. 2001;276:45001-45008. ciated with obesity. Diabetes. 2005;54:322-332. 21. Dentin R, Pégorier J-P, Benhamed F, et al. Hepatic glucokinase 5. Woods A, Williams JR, Muckett PJ, et al. Liver-specific activa- is required for the synergistic action of ChREBP and SREBP-1c tion of AMPK prevents steatosis on a high-fructose diet. Cell Rep. on glycolytic and lipogenic gene expression. J Biol Chem. 2017;18:3043-3051. 2004;279:20314-20326. |    15306860, 2020, 9, Downloaded from https://faseb.onlinelibrary.wiley.com/doi/10.1096/fj.202000678R by Readcube (Labtiva Inc.), Wiley Online Library on [15/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 11836 WEBER et al. 22. Veerkamp JH, van Moerkerk TB, Glatz JF, Zuurveld JG, Jacobs disease and insulin resistance: revisiting the role of protein Kinase AE, Wagenmakers AJ. 14CO2 production is no adequate mea- C ε. Hepatology. 2019;70(6):2217-2220. https://doi.org/10.1002/ sure of [14C]fatty acid oxidation. Biochem Med Metab Biol. hep.30829 1986;35:248-259. 41. Turinsky J, O’Sullivan DM, Bayly BP. 1,2-Diacylglycerol and ce- 23. Morillas M, Clotet J, Rubí B, et al. Identification of the two his- ramide levels in insulin-resistant tissues of the rat in vivo. J Biol tidine residues responsible for the inhibition by malonyl-CoA in Chem. 1990;265:16880-16885. peroxisomal carnitine octanoyltransferase from rat liver. FEBS. 42. Ussher JR, Koves TR, Cadete VJJ, et al. Inhibition of De Novo 2000;Lett:466. ceramide synthesis reverses diet-induced insulin resistance 24. Escolà-Gil JC, Llaverias G, Julve J, Jauhiainen M, Méndez- and enhances whole-body oxygen consumption. Diabetes. González J, Blanco-Vaca F. The cholesterol content of western 2010;59(10):2453-2464. https://doi.org/10.2337/db09-1293 diets plays a major role in the paradoxical increase in high-density 43. Koves TR, Ussher JR, Noland RC, et al. Mitochondrial overload lipoprotein cholesterol and upregulates the macrophage reverse and incomplete fatty acid oxidation contribute to skeletal mus- cholesterol transport pathway. Arterioscler Thromb Vasc Biol. cle insulin resistance. Cell Metab. 2008;7(1):45-56. https://doi. 2011;31:2493-2499. org/10.1016/j.cmet.2007.10.013 25. Martínez-Sánchez N, Seoane-Collazo P, Contreras C, et al. 44. Skulachev VP. Role of uncoupled and non-coupled oxidations in Hypothalamic AMPK-ER stress-JNK1 axis mediates the cen- maintenance of safely low levels of oxygen and its one-electron tral actions of thyroid hormones on energy balance. Cell Metab. reductants. Q Rev Biophys. 1996;29:169-202. 2017;26:212-229.e12. 45. Nègre-Salvayre A, Hirtz C, Carrera G, et al. A role for uncoupling 26. R: A language and environment for statistical computing. protein-2 as a regulator of mitochondrial hydrogen peroxide gener- 27. Ritchie ME, Phipson B, Wu D, et al. Limma powers differential ation. FASEB J. 1997;11:809-815. expression analyses for RNA-sequencing and microarray studies. 46. Echtay KS, Roussel D, St-Plerre J, et al. Superoxide activates mito- Nucleic Acids Res. 2015;43:e47. chondrial uncoupling proteins. Nature. 2002;415:96-99. 28. Shilin Zhao A, Yin L, Guo Y, Sheng Q, Shyr Maintainer Shilin 47. Cortez-Pinto H, Zhi Lin H, Qi Yang S, Odwin Da Costa S, Diehl Zhao Y. Package “heatmap3” Type Package Title An Improved AM. Lipids up-regulate uncoupling protein 2 expression in rat he- Heatmap Package; 2019. patocytes. Gastroenterology. 1999;116:1184-1193. 29. McInnes, L., John Healy, J., Melville, J. (2018)UMAP: Uniform 48. Casteilla L, Rigoulet M, Pénicaud L. Mitochondrial ROS me- Manifold Approximation and Projection for Dimension Reduction. tabolism: modulation by uncoupling proteins. IUBMB Life. arXiv e-prints: 1802.03426 2001;52:181-188. 30. McInnes L, Healy J, Saul N, Großberger L. UMAP: uniform 49. Yang L, Li P, Fu S, Calay ES, Hotamisligil GS. Defective hepatic manifold approximation and projection. J Open Source Software. autophagy in obesity promotes ER stress and causes insulin resis- 2018;3(29):861. https://doi.org/10.21105/​joss.00861 tance. Cell Metab. 2010;11:467-478. 31. Kim S. ppcor: an R package for a fast calculation to semi-par- 50. Martinez-Lopez N, Singh R. Autophagy and Lipid Droplets in the tial correlation coefficients. Commun Stat Appl Methods. Liver. Annu Re Nutr. 2015;35:215-237. 2015;22:665-674. 51. Ezquerro S, Mocha F, Frühbeck G, et al. Ghrelin reduces TNF- 32. Wei T, Simko V. R package “corrplot”: visualization of a cor- α-induced human hepatocyte apoptosis, autophagy and pyropto- relation matrix (Version 0.84); 2017. https://github.com/taiyu​n/ sis: role in obesity-associated NAFLD. J Clin Endocrinol Metab. corrplot 2018;104:21-37. 33. Benjamini Y, Hochberg Y. Benjamini-1995.pdf. J R Stat Soc B. 52. Alexaki A, Gupta SD, Majumder S, et al. Autophagy regulates 1995;57:289-300. sphingolipid levels in the liver. J Lipid Res. 2014;55:2521-2531. 34. Kleiner DE, Brunt EM, Van Natta M, et al. Design and validation 53. Puri P, Baillie RA, Wiest MM, et al. A lipidomic analysis of nonal- of a histological scoring system for nonalcoholic fatty liver disease. coholic fatty liver disease. Hepatology. 2007;46:1081-1090. Hepatology. 2005;41:1313-1321. 54. Sanyal AJ, Pacana T. A lipidomic readout of disease progression 35. Kimmel AR, Sztalryd C. The Perilipins: major cytosolic lipid in a diet-induced mouse model of nonalcoholic fatty liver disease. droplet-associated proteins and their roles in cellular lipid stor- Trans Am ClinClimatol Assoc. 2015;126:271-288. age, mobilization, and systemic homeostasis. Annu Rev Nutr. 55. Kahle M, Schäfer A, Seelig A, et al. High fat diet-induced modi- 2016;36:471-509. fications in membrane lipid and mitochondrial-membrane protein 36. Singh R, Cuervo AM. Lipophagy: connecting autophagy and lipid signatures precede the development of hepatic insulin resistance in metabolism. Int J Cell Biol. 2012;2012:1-12. mice. Mol Metab. 2014;4:39-50. 37. Singh R, Kaushik S, Wang Y, et al. Autophagy regulates lipid me- 56. Shon JC, Shin H-S, Seo YK, Yoon Y-R, Shin H, Liu K-H. Direct tabolism. Nature. 2009;458:1131-1135. infusion MS-based lipid profiling reveals the pharmacological ef- 38. Naso MF, Tomkowicz B, Perry WL, Strohl WR. Adeno- fects of compound K-reinforced ginsenosides in high-fat diet in- Associated Virus (AAV) as a vector for gene therapy. BioDrugs. duced obese mice. J Agric Food Chem. 2015;63:2919-2929. 2017;31:317-334. 57. Graessler J, Bornstein TD, Goel D, et al. Lipidomic profiling be- 39. Gassaway BM, Petersen MC, Surovtseva YV, et al. PKCε con- fore and after Roux-en-Y gastric bypass in obese patients with dia- tributes to lipid-induced insulin resistance through cross talk with betes. Pharmacogenomics J. 2014;14:201-207. p70S6K and through previously unknown regulators of insulin sig- 58. Velagapudi V, Welbourn R, Arora T, et al. Roux-en-Y gastric by- naling. Proc Natl Acad Sci. 2018;115:E8996-E9005. pass surgery induces early plasma metabolomic and lipidomic al- 40. Samuel VT, Petersen MC, Gassaway BM, Vatner DF, Rinehart J, terations in humans associated with diabetes remission. PLoS One. Shulman GI. Considering the links between nonalcoholic fatty liver 2015;10:e0126401. | 15306860, 2020, 9, Downloaded from https://faseb.onlinelibrary.wiley.com/doi/10.1096/fj.202000678R by Readcube (Labtiva Inc.), Wiley Online Library on [15/11/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License WEBER et al.     11837 59. Kayser BD, Lhomme M, Dao MC, et al. Serum lipidom- SUPPORTING INFORMATION ics reveals early differential effects of gastric bypass com- Additional Supporting Information may be found online in pared with banding on phospholipids and sphingolipids the Supporting Information section. independent of differences in weight loss. Int J Obes. 2017;41: 917-925. 60. Kohjima M, Enjoji M, Higuchi N, et al. Re-evaluation of fatty acid metabolism-related gene expression in nonalcoholic fatty liver dis- How to cite this article: Weber M, Mera P, Casas F, ease. Int J Mol Med. 2007;20(3):351-358. https://doi.org/10.3892/ et al. Liver CPT1A gene therapy reduces diet-induced ijmm.20.3.351 hepatic steatosis in mice and highlights potential lipid 61. Nakamuta M, Kohjima M, Morizono S, et al. Evaluation of fatty biomarkers for human NAFLD. The FASEB Journal. acid metabolism-related gene expression in nonalcoholic fatty liver 2020;34:11816–11837. https://doi.org/10.1096/ disease. Int J Mol Med. 2005;16:631-635. fj.20200​0678R