Hepatic ketogenesis regulates lipid homeostasis via ACSL1-mediated fatty acid partitioning

Liver-derived ketone bodies play a crucial role in fasting energy homeostasis by fueling the brain and peripheral tissues. Ketogenesis also acts as a conduit to remove excess acetyl-CoA generated from fatty acid oxidation and protects against diet-induced hepatic steatosis. Surprisingly, no study has examined the role of ketogenesis in fasting-associated hepatocellular lipid metabolism. Ketogenesis is driven by the rate-limiting mitochondrial enzyme 3-hydroxymethylglutaryl CoA synthase (HMGCS2) abundantly expressed in the liver. Here, we show that ketogenic insufficiency via disruption of hepatic HMGCS2 exacerbates liver steatosis in fasted chow and high-fat-fed mice. We found that the hepatic steatosis is driven by increased fatty acid partitioning to the endoplasmic reticulum (ER) for re-esterification via acyl-CoA synthetase long-chain family member 1 (ACSL1). Mechanistically, acetyl-CoA accumulation from impaired hepatic ketogenesis is responsible for the elevated translocation of ACSL1 to the ER. Moreover, we show increased ER-localized ACSL1 and re-esterification of lipids in human NASH displaying impaired hepatic ketogenesis. Finally, we show that L-carnitine, which buffers excess acetyl-CoA, decreases the ER-associated ACSL1 and alleviates hepatic steatosis. Thus, ketogenesis via controlling hepatocellular acetyl-CoA homeostasis regulates lipid partitioning and protects against hepatic steatosis.


Introduction
Non-alcoholic fatty liver disease (NAFLD) is one of the most heterogeneous forms of liver disease, with a prevalence of ~ 24% in the United States and rapidly rising worldwide. Hepatic steatosis or fatty liver is the rst hit in NAFLD pathogenesis, wherein lipotoxicity from the accumulated lipids induces oxidative stress, insulin resistance, in ammation, and brosis, leading to the development of non-alcoholic steatohepatitis (NASH). About 20% of NAFLD patients progress to end-stage liver diseases, such as cirrhosis and hepatocellular carcinoma, increasing overall mortality. Despite the increasing incidence and mortality, no FDA-approved drugs are available due to the incomplete understanding of NAFLD pathogenesis 1, 2 3,4 . Moreover, the links between lipid metabolism and NALFD pathogenesis in response to energy demands are not well de ned. Hepatic lipid accumulation is determined by a mismatch in free fatty acid uptake, de novo lipogenesis, fatty acid oxidation, and very low-density lipoprotein (VLDL) secretion 5,6 . In this process, lipid partitioning plays an essential role in adapting to systemic energy needs 7 . For instance, free fatty acids entering the hepatocytes are partitioned toward mitochondrial fatty acid oxidation (FAO) at fasting. In contrast, fatty acids are targeted to the endoplasmic reticulum (ER) for esteri cation and storage as lipid droplets under nutrient-rich conditions 7 . An imbalance in lipid partitioning induces lipid accumulation leading to hepatic steatosis and injury, the prerequisite step for the development and progression of NAFLD and its progressive form NASH 7,8 . Despite the signi cance of fatty acid partitioning in hepatic lipid homeostasis, the mechanisms that govern them are under-studied.
Fasting is the appropriate model to dissect the mechanism of lipid partitioning as the latter dictates FAO in a controlled manner. Fasting is characterized by adipose tissue lipolysis, increased delivery of fatty acids to the liver, and enhanced FAO 9,10 . Hepatic FAO generates acetyl-CoA that is preferably converted into ketone bodies via the mitochondrial 3-hydroxymethulglutaryl-CoA synthase 2 (HMGCS2), the ratelimiting ketogenic enzyme highly expressed in the hepatocytes 11 . Ketone bodies secreted from the liver fuel the peripheral organs such as the brain and muscle 12,13 . Recent studies show that neonates with ketogenic insu ciency develop fatty liver, which could be rescued by early weaning 14,15 . Similarly, feeding a high-fat diet to mice with ketogenic insu ciency induces hepatic steatosis and liver injury 13,15,16 . These data indicate that ketogenesis is a conduit to remove excess dietary lipids, thereby protecting against fatty liver. During fasting, hepatocytes are bombarded with a huge in ux of adipose tissue-derived fatty acids 17 . Although ketogenesis is known to be induced during fasting 17 , its signi cance in regulating hepatocellular lipid homeostasis is not well understood.
We use conditional liver-speci c HMGCS2 knockout mice to report that ketogenesis insu ciency exacerbates fatty liver in chow and high-fat feeding mice at fasted state. Impaired ketogenesis did not affect fatty acid uptake, de novo lipogenesis, and β-oxidation but increased fatty acid storage as lipid droplets. Incoming fatty acids are esteri ed into acyl-CoA by the enzyme Acyl-CoA synthetase long-chain family member 1 (ACSL1) localized on mitochondria and ER 18, 19 . Mitochondrial ACSL1 promotes βoxidation, whereas ER-associated ACSL1 favors lipid storage via re-esteri cation 20 . The mechanisms regulating the ER localization of ACSL1 and its implication in hepatic lipid metabolism are not well understood. We show that acetyl-CoA accumulation from ketogenic insu ciency induces ER translocation of ACSL1, resulting in fatty acid re-esteri cation and steatosis. We report that a similar mechanism exists in human NASH livers, wherein a reduction in HMGCS2 is associated with increased ER-associated ACSL1. Excessive acetyl-CoA is buffered by L-carnitine, and this buffering capacity is diminished in NASH due to acquired carnitine de ciency 21,22 . We show that hepatic L-carnitine levels were signi cantly lower in ketogenic insu cient mice, and restoring L-carnitine attenuated fastinginduced acetyl-CoA accumulation, ER-ACSL1, and hepatic steatosis. We also demonstrate that the lipidlowering effect of L-carnitine is associated with a reduction in the ER-associated ACSL1 in primary hepatocytes derived from NASH patients. Overall, our study de nes the crucial role of hepatic ketogenesis in lipid homeostasis by regulating the partitioning of fatty acids via the acetyl-CoA-ER-ACSL1 axis.

Animal studies
We generated a Hmgcs2 F/F founder line on a C57BL6 background by CRISPR/Cas-9 mediated genome engineering, wherein exon 2 was oxed. Mice with conditional knockout of HMGCS2 (Hmgcs2 ΔLiv mice) were generated by crossing Hmgcs2 F/F mice with mice carrying tamoxifen-inducible Cre recombinase under the control of Albumin promoter (Alb-Cre/ERT2). Littermates that do not express Cre recombinase were used as the control. Tamoxifen (Cayman Chemical, Ann Arbor, MI) was dissolved at 10mg/ml in corn oil (Sigma-Aldrich, St. Louis, MO) and administered intraperitoneally at a dose of 100mg/kg body weight for two consecutive days. All the experiments were carried out at least two weeks after tamoxifen injection. All the animals were maintained on the chow diet provided by the DLAR. For high fat studies, mice were fed with an HFD (60% kcal from fat, D12492; Research Diets) diet for 4 weeks. Mice euthanized at the fed state or after overnight fast (6.00 PM-10.00 AM). For the L-carnitine experiment, 6-8week-old Hmgcs2 ΔLiv mice were provided with drinking water containing L-Carnitine (@10mg/ml TCI, Portland, OR; C0049) 9 hours before fasting. All the mice were maintained at a 12-hour light/dark cycle with free access to food and water. All experiments were performed using age-and sex-matched littermates, unless indicated in the gure legends. All animal experiments were approved by the Animal Care and Use Committee at the University of Pittsburgh.

Mouse primary hepatocyte isolation
Mouse primary hepatocytes were isolated from 6-8-week-old Hmgcs2 ΔLiv male mice. Brie y, mice were anesthetized using iso urane, and the inferior vena cava was cannulated and infused with 10 ml of perfusion buffer (HBSS with no Ca 2+ , no Mg 2+ , no phenol red, supplemented with 0.5 mM of EDTA and 25 mM of HEPES, pH 7.4) with a ow rate of 3ml/min. Then, 10 ml of digestion buffer (HBSS with no Ca 2+ , no Mg 2+ , no phenol red, supplemented with 25 mM of HEPES, pH 7.4) with collagenase type 1 (15mg/50ml) was infused using a peristaltic pump. The liver was excised, minced, and ltered using a 70 µm cell strainer (Falcon) and centrifuged at 50 x g for 2 min to pellet hepatocytes. Dead cells were removed by centrifugation at 50 x g for 10 min in 90% Percoll solution (Sigma-Aldrich) in pre-chilled 10X PBS. Cells were resuspended in DMEM (Gibco) supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin. The cells were counted and plated onto the 12-well or 10 cm plates at a cell density of 0.2 x10 6 cells/ml. After 4 hours of incubation, the cells were refreshed with Williams E media supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin. For Triascin C treatment, hepatocytes were incubated with 200µM bovine serum albumin-conjugated palmitate (BSA-PA; Cayman Chemical) and treated with 5µM Triascin C (Enzo Life Sciences, Farmingdale, NY) or DMSO vehicle for 16 hours. For the acetate treatment, mouse primary hepatocytes were incubated with 200µM BSA-PA or BSA and treated with 20 mM sodium acetate (Sigma-Aldrich) for 16 hours.
Human NASH subjects and primary hepatocyte isolation A total of 20 human subjects (10 NASH and 10 Normal) were scheduled for a liver biopsy to diagnose NASH, and were recruited by participating physicians at the University of Pittsburgh Medical Center (Pittsburgh, PA, USA). All the patients ful lled the inclusion criteria, such as no positive for viral hepatitis, Wilson disease, or any other possible cause of liver dysfunction. The patients were excluded if the alcohol consumption exceeded 20 g/week. Primary hepatocytes were isolated from explanted human liver segments obtained from patients receiving orthotopic liver transplantation for decompensated liver cirrhosis due to NASH. The specimens were obtained by the Human Synthetic Liver Biology Core at the Pittsburgh Liver Research Center (PLRC) using a protocol approved by the Human Research Review Committee and the Institutional Review Board (IRB STUDY20090069) at the University of Pittsburgh.
Liver tissue specimens were protected from ischemic injury by ushing with ice-cold University of Wisconsin (UW) solution immediately after resection in the operating room, keeping the specimens on ice, and transporting the specimens immediately to the laboratory. Hepatocytes were isolated by a modi ed three-step perfusion technique. Brie y, the livers were ushed under a sterile biosafety hood through the hepatic vessels (re-circulation technique) with pre-warmed calcium-free HBSS (Sigma, H6648-1L) supplemented with 0.5 mM EGTA (Thermo Fisher, 50-255-956) and then with collagenase/protease solution (VitaCyte, 007-1010) until the tissue was fully digested. The digestion time for each preparation was in the range of 45-60 min. The digested liver was removed and immediately cooled with ice-cold Leibovitz's L-15 Medium (Invitrogen, 11415114) supplemented with 10% FBS (Sigma, F4135). The nal cell suspension was centrifuged twice at 65xg for 7 min at 4 o C and the medium was aspirated. The yield and viability of freshly isolated hepatocytes were estimated by trypan blue staining.
Neutral lipid staining using Oil-red-O and BODIPY Cells were washed with 1X PBS and xed with 10% PBS-buffered formalin at room temperature for 30 min, washed with ddH 2 O, and incubated with 60% isopropyl alcohol for 5 min. The cells were incubated in 0.5% Oil red O (Sigma-Aldrich) staining reagent for 30-45 min. Then, the cells were washed with ddH 2 O thrice and counterstained with hematoxylin for 15 sec, and washed several times with ddH 2 O. The images were captured using an EVOS microscope (Olympus, Tokyo, Japan). To quantify the intracellular lipids, the stain was extracted in 250 µl isopropanol incubated at room temperature for 10-15 min, and absorbance was measured at 492 nm. For BODIPY staining, cells were xed with 10% PBS-buffered formalin at room temperature for 20 min, washed with 1X PBS, and stained with 250 µl of BODIPY (1mg/ml stock diluted 1:1000; Cayman Chemical) for 15-20 min. The cells were washed with PBS, and images were captured using an EVOS microscope.

Blood and serum analysis
The tail was snipped, and blood glucose levels were measured using a glucometer (Bayer, Parsippany, NJ). Serum β-hydroxybutyrate levels (Cayman Chemical, Ann Arbor, MI) were measured using a calorimetry kit following manufacturers' instructions. Serum triglycerides and cholesterol levels were measured using colorimetric In nity Triglyceride and Cholesterol Reagent kits (Thermo sher Scienti c, Middletown, VA). Serum non-esteri ed fatty acid was quanti ed using a colorimetric assay (Fuji lm Wako Diagnostics, Lexington, MA).

Liver triglyceride assay
Hepatic triglyceride levels were measured as described previously 23 . Brie y, human and mouse frozen liver tissues (∼50 mg) were homogenized in 3 ml of chloroform: methanol (2:1) mixture and vortexed.
The homogenate was incubated for 60 min at room temperature while rotating on a shaker. Then, the homogenate was acidi ed with 1mol/L H 2 SO 4, and lipid fractions in the lower organic phase were collected after centrifugation at 1300 rpm for 10 min and transferred to clean glass vials. Triglyceride levels were determined using the triglyceride reagent (Thermo Fisher Scienti c) and normalized to liver weight.
Thin layer chromatography (TLC) Equal amount of lipids extracted from human and mouse livers were dried completely under a N-EVAP nitrogen evaporator (Organomation, Berlin MA) and resuspended with 100 µl chloroform: methanol (2:1).
Before running the TLC plate, the running chamber was equilibrated with 200ml of the solvent system containing petroleum ether: ethyl ether: acetic acid mixture (25:5:1), and the TLC plate was baked at 75°C for 30 min. Aliquots of 25 µl resuspended lipid extracts and lipid standard (Nu-Chek-Prep, Elysian, MN), were loaded onto the TLC plates (EMD Millipore) and lipids were separated by the solvent system.. The plate was removed from the chamber when the solvent reached about 1 inch to the top edge of the plate, and was dried for 5-10 min (until all solvents evaporated). The plate was then transferred to an equilibrated iodine tank for about 30-45 min for staining, and images were captured with CanoScan LidE 220 imager (Canon). ImageJ software (NIH, Bethesda) was used to quantify the lipid band density.

Liver histology
Liver tissue was excised and immediately xed in 10% PBS-buffered formalin (Thermo sher Scienti c). H&E staining was performed in 6-micron para n-embedded sections, and the images were captured using an EVOS microscope.
RNA isolation, cDNA synthesis, and qPCR analysis Total RNA was extracted using TRIzol reagent (Life Technologies) as per manual instructions. RNA was quanti ed using a Nanodrop, and 1 µg of RNA was reverse-transcribed using Moloney murine leukemia virus (Mu-MLV) reverse transcriptase (Promega, Madison, WI). mRNA levels were analyzed using SYBR Green PCR Master Mix (ApexBio, Houston, TX) with QuantStudio 3 Station qPCR machine (Applied Biosystems, Foster City, CA). The relative expression of target genes was calculated using a comparative delta threshold cycles (ΔCT) method after normalizing to β-actin. The primer sequences are provided in Table S1.

Western blotting
Cells or human and mouse frozen tissues (∼10mg) were homogenized using radioimmunoprecipitation assay lysis buffer (0.5% NP-40, 0.1% sodium deoxycholate, 150 mmol/L NaCl, 50 mmol/L Tris-Cl, pH 7.5) containing 1 mmol/L phenylmethylsulphonyl uoride, protease inhibitor cocktail (Sigma-Aldrich) and 2 mmol/L sodium orthovanadate 24 . The homogenate was centrifuged at 13,000 x g for 10 min at 4°C to collect the supernatant. The protein concentration was quanti ed using a protein assay kit (Bio-Rad, Hercules, CA), and the sample was resolved on a SDS-PAGE. The membranes were blocked with 3% skim milk and incubated with primary antibodies, overnight at 4°C. Secondary antibodies conjugated with DyLight (Cell Signaling Technology) were added to the membranes (antibody details are in Table S2) and visualized using the Odyssey CLx Imaging System (LI-COR, Lincoln, NE).

Subcellular fractionation for western blot
Mitochondrial and microsomal fractions were isolated as described previously 20 . In brief, cell pellet and liver tissues from mice and humans were minced in mitochondrial isolation buffer (MSHE, 70 mM sucrose, 210 mM mannitol, 5 mM HEPES, 1mM EGTA, pH 7.2 with 2% fatty acid-free BSA) with protease and phosphatase inhibitors. The minced tissue was homogenized by stroking 25-30 times using a Te on glass homogenizer on ice. The homogenate was transferred to the Eppendorf tube and centrifuged at 800 x g for 10 min, and the supernatant was centrifuged at 8000 x g for 10 min to pellet the mitochondria. The mitochondrial pellet was washed by resuspending in MSHE buffer and centrifuged at 8000 x g for 10 min. The microsomal fraction was collected by transferring the supernatant of the mitochondrial pellet to a new tube containing 8 mM of CaCl 2 and incubated on a rotating platform in the cold room for 10 min. The microsomal pellet was collected by centrifuging at 30000 x g for 30 min. The mitochondrial and microsomal pellets were resuspended in RIPA buffer containing protease and phosphatase inhibitor cocktail.

Crude mitochondria isolation for respirometry
A half lobe of the freshly sampled liver was transferred to ice-cold PBS, cut into small pieces, and washed three times with PBS. 200 mg of the liver was weighed and transferred to 2mL of SMET buffer (10 mM Tris HCl (pH 7.5), 220 mM Mannitol, 70 mM Sucrose, 1 mM EDTA, 0.25% BSA) and homogenized using a glass dounce and rotating te on pestle, applying eight strokes at max speed. Homogenates were centrifuged at 800× g for 10 min at 4°C to discard cell debris and nuclei. The supernatant was collected, avoiding the top fat layer, and centrifuged again to eliminate remaining cell debris and nuclei. A fraction of the supernatant was saved for protein assays and citrate synthase activity measurement. The remaining supernatant was then centrifuged at 8000 x g for 10min at 4°C to pellet crude mitochondria.
Pellets were resuspended in 400µL SMET buffer, and 80µL was used for respirometry assays. Remaining pellet suspension was centrifuged 8000 x g for 10min at 4°C and saved for protein quanti cation. Mitochondrial respiratory capacity was assessed using an Oroboros O 2 K High-Resolution Respirometer and MiR05 buffer at 37°C under constant mixing in a sealed, 2-ml chamber. Respirometry assays were performed to determine respiration devoted to ATP synthesis, maximum or uncoupled electron transport chain capacity and non-mitochondrial respiration. Membrane integrity was assessed by the addition of cytochrome c after addition of ADP. Substrate and inhibitor concentrations were as follows: palmitoylcarnitine (40 µM), malate (2 mM); adenosine diphosphate (ADP, 4 mM); cytochrome c (10µM), FCCP (1µL titration) and antimycin A (2.5 µM). The protein concentration of the crude mitochondria was determined by BCA protein assay and used to express respiratory capacities as oxygen consumption per protein mass (pmol s − 1 mg − 1 ). Respiration in the presence of antimycin A was subtracted from all respiratory rates to account for non-mitochondrial respiration.

Acetyl-CoA measurement
The acetyl-CoA levels were measured using a uorometric PicoProbe Acetyl CoA assay kit (Abcam, Waltham, MA) following manufacturers' instructions. Human and mouse liver tissue (∼25mg) was homogenized in 1N perchloric acid (PCA) and centrifuged at 10,000 x g for 10 min at 4°C. The samples were neutralized with 3M KHCO 3 to pH ∼ 6-8 while keeping on ice. The supernatant was collected by centrifuging at 13,000 x g for 15 min at 4°C. The assay was carried out per the manual instructions, and the nal concentration was adjusted based on the sample dilution factor and tissue weight. L-carnitine measurement L-carnitine levels were measured by colorimetric L-carnitine assay kit (Abcam, Waltham, MA) according to manufactures instructions. Human and mouse frozen liver tissues (∼25mg) were homogenized with 250 µl assay buffer and the supernatant was collected by centrifuging at 13,000 x g for 10 min. The samples were deproteinized by adding ice-cold perchloric acid (PCA) to a nal concertation of 1 M and mixed well. After incubating on ice for 5 min, the samples were centrifuged at 13,000 x g for 2 min at 4°C. The supernatant was neutralized to pH 6.5-8 using ice-cold 2 M KOH. Finally, the supernatant was collected for the assay by centrifuging at 13,000 x g for 15 min at 4°C. The assay was carried out as per the manual instructions and the nal concentrations of the L-carnitine were adjusted for the sample dilution factors and tissue weight.

Body composition analysis
Body composition was assessed in conscious mice using EchoMRI (EchoMRI LLC, TX, USA). Percent fat and lean mass were calculated for individual mice by dividing the weight of the tissue determined by EchoMRI by the body weight.

Quanti cation and Statistical data analysis
The data is presented as mean ± SEM. Statistical analyses were performed using Prism software version 9.0.0 (GraphPad Software). Comparisons between the two groups were made by conducting Student ttests. One-way ANOVA (Tukey multiple-comparisons test) was used to compare more than two groups.
Differences were considered statistically signi cant if p < 0.05 ( * ); p < 0.01 ( * * ) or p < 0.001 ( * * * ) or p < 0.0001( * * * * ). The band intensities for western blot and TLC were quanti ed using Image J 1.52q. The statistical methods of each experiment are indicated in the gure legends. In vitro experiments were replicated at least three independent times, except for human NASH hepatocytes.

Results
Hepatic ketogenesis protects against fasting-induced hepatic steatosis. Genetic studies dissecting the role of ketone bodies were performed in mice with constitutive knockout of HMGCS2 (using albumin promoter-driven Cre recombinase) in the liver. However, these mice develop severe hepatic steatosis and mitochondrial dysfunction at the neonatal age 14 . To assess the role of hepatic ketogenesis in metabolic homeostasis without any confounding developmental defects, we generated a mouse model with conditional knockdown of HMGCS2 in the liver (Hmgcs2 ΔLiv ) by crossing our Hmgcs2oxed mice (Hmgcs2 F/F ) with mice expressing tamoxifen-inducible cre under the control of albumin promoter (Fig. 1A). Administration of tamoxifen decreased HMGCS2 protein expression in the liver but not in other organs such as the colon, kidney, and heart ( Fig. 1B and S1A), which are also minimally involved in ketone body production 12 . Note, littermate Hmgcs2 F/F mice administered with tamoxifen were used as control throughout the studies. Ketone bodies play a crucial role in fasting energetics 12 ; therefore, we assessed Hmgcs2 F/F and Hmgcs2 ΔLiv mice under fed and 16-hours fasted conditions. Fasting decreased body weight, fat mass, and lean mass with no difference between the genotypes (Fig. 1C and D). As expected, fasting signi cantly increased the serum ketone body, β-hydroxybutyrate (BHB) levels in Hmgcs2 F/F mice (Fig. 1E), though no difference in hepatic HMGCS2 protein expression was noticed ( Fig. 1F). Disruption of HMGCS2 in the liver signi cantly decreased ketone body levels in Hmgcs2 ΔLiv mice (Fig. 1E), consistent with a recent report that ketone bodies are majorly derived from the liver 25 . Ketone body metabolism in peripheral tissues, particularly muscle, increases circulating glucose levels as ketone body oxidation competes with glucose utilization 26 . However, inhibiting hepatic ketogenesis did not affect the fasting glucose levels under nutrient-deprived conditions (Fig S1B).
Previous studies showed that constitutive deletion of hepatic HMGCS2 results in fatty liver in postnatal mice pups (as early as 4 days after birth) 14,15 . However, anti-sense oligomers (ASO)-mediated knockdown of Hmgcs2 in adult mice did not affect the hepatic lipid content under fed conditions 16 . Similarly, temporal disruption of Hmgcs2 in adult mice did not affect liver-to-body weight ratio or hepatic triglyceride content in Hmgcs2 ΔLiv mice under fed conditions ( Fig. 1G and H). However, fasting increased the liver-to-body weight ratio with paler liver in Hmgcs2 ΔLiv mice ( Fig. 1G and I). In line with that, liver triglyceride (TAG) levels were signi cantly elevated in fasted Hmgcs2 ΔLiv mice (Fig. 1H). Further, H&E analysis revealed microsteatosis in the fasted Hmgcs2 ΔLiv livers (Fig. 1J). Serum triglycerides were similar between the fed and fasted mice (Fig S1C). We found a trend towards an increase in serum cholesterol levels in Hmgcs2 ΔLiv mice; however, fasting did not affect cholesterol levels in both genotypes ( Fig S1D). Together, our data suggest that impaired ketogenesis induces hepatic steatosis under nutrientdeprived conditions.
Ketogenic insu ciency induces hepatic steatosis via ACSL1. During fasting, non-esteri ed fatty acids (NEFA) released from the adipose tissue via lipolysis enter the liver to undergo FAO 27 . An increase in adipose tissue lipolysis could induce hepatic steatosis 27 . Adipose tissue levels of lipolysis-related proteins, such as pHSL and ATGL (Fig S2A), and the circulating NEFA levels ( Fig S2B) were similar between the Hmgcs2 F/F and Hmgcs2 ΔLiv mice. Moreover, the hepatic expression of fatty acid uptake genes such as Cd36, Fatp2, and Fatp5 was not different between the genotypes (Fig. 2A). This suggests that adipose tissue lipolysis or enhanced hepatic lipid uptake mechanisms do not contribute to hepatic steatosis in Hmgcs2 ΔLiv mice. Adipose tissue-derived fatty acids activate hepatic PPARα, the nuclear transcription factor that regulates fatty acid oxidation genes 27 . Disruption of hepatic PPARα signaling induces severe steatosis after acute fasting 28 . Therefore, we assessed whether fasting-induced fatty liver in Hmgcs2 ΔLiv mice is caused by impaired PPARα signaling. As expected, fasting increased the mRNA levels of PPARα and its target genes, such as Cpt1, Acox1, Mcad, Cpt2 and Vcad (Fig. 2B). However, fasted Hmgcs2 ΔLiv mice showed a similar increase in the FAO genes (Fig. 2B), suggesting that steatosis in Hmgcs2 ΔLiv mice is not caused by impaired PPARα signaling. Moreover, mRNA levels of lipogenesisrelated genes such as Srebp1, Fasn, Acc1, and Scd1 showed no difference between the genotypes at fed or fasted conditions (Fig. 2C). Together, our data suggest that fasting-induced hepatic steatosis in Hmgcs2 ΔLiv mice is not driven by altered lipolysis, lipid uptake, PPARα signaling, or lipogenesis.
We found that the livers of fasted Hmgcs2 ΔLiv mice expressed a signi cantly elevated level of fat-speci c protein 27 (Fsp27) and perilipin (Plin2) (Fig. 2D). FSP27 increases lipid storage by inhibiting ATGLmediated lipolysis 29 , while perilipin2 confers resistance of lipid droplets to lipophagy 30 . To address the involvement of lipolytic mechanisms in the liver, we assessed the expression of CGI-58 and ATGL, which showed no difference between the genotypes (Fig. 2E). We also did not nd any difference in the levels of phosphorylated HSL (Fig. 2E). Lipophagy is recently recognized as a regulator of lipid droplet size and number 31 . Our analysis did not show any difference in the expression of autophagy-related proteins in fasted Hmgcs2 ΔLiv livers (Fig S2C), indicating that the lipolytic or lipophagic mechanisms in the liver are not responsible for lipid accumulation in Hmgcs2 ΔLiv mice.
Hepatic lipid content is controlled by a tight interaction between fatty acid synthesis, oxidation, and esteri cation 32 . Since FAO, lipogenesis, or lipolysis-related genes are unaltered, we assessed whether Hmgcs2 ΔLiv has altered lipid re-esteri cation. We did not nd any difference in the mRNA levels of esteri cation-related genes such as Dgat1, Dgat2, Gpat, and lipin1 between the genotypes (Fig S2D). Intriguingly, the mRNA levels of acyl-CoA synthetase long-chain family member 1 (ACSL1) were signi cantly elevated in the livers of fasted Hmgcs2 ΔLiv mice (Fig. 2F). However, no difference in other ACSL isoforms (Acls2, Acls3, and Acsl5) was noted (Fig. 2F). Consistent with the mRNA levels, ACSL1 protein expression was also increased in the livers of fasted Hmgcs2 ΔLiv mice (Fig. 2G). ACSL1 is a ratelimiting enzyme involved in converting fatty acids into acyl-CoA, the rst step required for fatty acid oxidation or esteri cation 33 . To determine whether ACSL1 is responsible for steatosis in Hmgcs2 ΔLiv mice, we incubated the primary hepatocytes from Hmgcs2 ΔLiv mice and Hmgcs2 F/F mice under lipogenic condition (BSA-conjugated palmitate) in the presence or absence of Triascin C, a speci c ACSL1 inhibitor 34 . The lipid accumulation was higher in the primary hepatocytes from Hmgcs2 ΔLiv mice compared to Hmgcs2 F/F (Fig. 2H and S2F). Notably, Triascin C alleviated steatosis in Hmgcs2 ΔLiv primary hepatocytes ( Fig. 2H and S2E), suggesting that ASCL1 drives steatosis in Hmgcs2 ΔLiv mice. ACSL1 translocation to ER is regulated by ketogenesis. The subcellular localization of ACSL1 plays a crucial role in targeting the fatty acids for catabolism or storage in the form of lipid droplets 19 . Acyl-CoA generated by the mitochondria-associated ACSL1 is channeled toward fatty acid oxidation 35 . Whereas ER-associated ACSL1 promotes fatty acid esteri cation to synthesize triglycerides 20 . To understand how ACSL1 induces hepatic steatosis in Hmgcs2 ΔLiv mice, we assessed the subcellular localization of ACSL1.
ACSL1 is highly enriched in the isolated mitochondria of fasted livers 20 . Our data showed no difference in the mitochondrial ACSL1 protein levels between the fasted Hmgcs2 F/F and Hmgcs2 ΔLiv mice (Fig. 2I). Interestingly, we found signi cantly elevated ER-associated ACSL1 protein levels in the livers of fasted-Hmgcs2 ΔLiv mice (Fig. 2J). Because ER-associated ASCL1 regulates triglyceride storage via fatty acid reesteri cation 20 , we performed thin-layer chromatography (TLC) to assess whether re-esteri cation is enhanced in Hmgcs2 ΔLiv mice. Our data showed a marked increase in the triglyceride fraction from the fasted Hmgcs2 ΔLiv mice, even after normalizing with phospholipids ( Fig. 2K-M). However, no difference in phospholipids, cholesterol esters, and ceramides factions was noted (Fig. 2K and L). Together, our data indicate that ketogenesis controls the ER-ACSL1 mediated re-esteri cation of fatty acids under fasting conditions. Ketogenesis insu ciency does not affect mitochondrial function under acute fasting. Previous studies showed that ketone bodies are potent inducers of mitochondrial biogenesis by upregulating the expression of the transcription factor, PGC1α 36,37 . Because mitochondrial homeostasis including biogenesis and function is key in regulating lipid metabolism 38 , we assessed whether impaired ketogenesis affected mitochondrial homeostasis in Hmgcs2 ΔLiv mice. We found no difference in the mRNA expression of Pgc1α, Nrf2, and various other mitochondria-associated genes such as Atp6, Cox1, and Nd4 in Hmgcs2 ΔLiv mice (Fig. 3A). Moreover, the expression of OXPHOS proteins such as NDUFB8, SDHB, UQCRC2, and cytochrome C oxidase was similar between the genotypes (Fig. 3B), suggesting that the temporal impairment in ketogenesis does not affect mitochondrial content. A recent study showed that TBK1 acts as a chaperone regulating the mitochondrial translocation and activity of ACSL1 under fasting conditions 20 . We found no difference between the fasted Hmgcs2 F/F and Hmgcs2 ΔLiv mice in the expression of TBK1 or its phosphorylated form (Fig S3A), indicating that the mitochondrial TBK1-ASCL1 axis may not be involved in the fasting-mediated hepatic steatosis in Hmgcs2 ΔLiv mice. Ketogenic insu ciency in neonatal pups increases the acetylation of mitochondrial proteins leading to mitochondrial dysfunction 14 . When we analyzed the isolated mitochondria, pan-protein acetylation was found to be increased in the livers of fasted Hmgcs2 ΔLiv mice (Fig. 3C). We then assessed mitochondrial respiratory capacity in the presence of palmitoylcarnitine using puri ed liver mitochondria from Hmgcs2 F/F and Hmgcs2 ΔLiv mice. There were no differences in mitochondrial respiration devoted to oxidative phosphorylation or maximum electron transport chain activity in the uncoupled state ( Fig. 3D-E). Consistent with the data above, we also found no difference in liver citrate synthase activity (Fig. 3F), which serves as a surrogate of mitochondrial content 39 . Together, these data suggest that ketogenic insu ciency did not result in steatosis due to impaired mitochondrial fatty acid oxidation, nor did it result in compensatory increases in mitochondrial content.
Ketogenic insu ciency exacerbates diet-induced hepatic steatosis via re-esteri cation. To understand whether hepatic ketogenesis insu ciency promotes diet-induced hepatic steatosis in response to fasting, 8-week-old HMGCS2 conditional knockout were treated with tamoxifen to disrupt HMGCS2 and then fed with 60% high-fat diet (HFD) for 4-weeks and then fasted for 16 hours before euthanasia. We found no difference in the body weight between Hmgcs2 F/F and Hmgcs2 ΔLiv mice on an HFD (Fig. 4A). As expected, ketone body levels were signi cantly lower in Hmgcs2 ΔLiv mice (Fig. 4B). No difference in serum triglyceride or cholesterol was noted ( Fig. 4C and S4A). As in chow diet mice, the liver-to-body weight ratio was signi cantly higher in HFD-fed Hmgcs2 ΔLiv mice (Fig. 4D). H&E analysis showed microsteatosis in Hmgcs2 F/F mice, while Hmgcs2 ΔLiv mice showed macrosteatosis (Fig. 4E). Liver triglyceride levels were signi cantly increased in the HFD-fed Hmgcs2 ΔLiv mice (Fig. 4F). We assessed whether enhanced fatty acid re-esteri cation contributed to macrosteatosis in HFD-fed Hmgcs2 ΔLiv mice. Similar to chow-fed mice, TLC analysis showed a signi cant increase in triglyceride fraction in HFD-fed Hmgcs2 ΔLiv mice ( Fig. 4G-I). But no difference in other lipid moieties was observed (Fig. 4G and H). We then assessed whether ER localization of ACSL1 is augmented by HFD feeding in Hmgcs2 ΔLiv mice. We found a signi cant increase in the ER-associated ACSL1 in HFD-fed Hmgcs2 ΔLiv mice (Fig. 4J), but no change in mitochondria-associated ACSL1 was observed (Fig. 4J). Similar to chow-fed mice, acetylated mitochondrial proteins were elevated in the mitochondria of HFD-fed Hmgcs2 ΔLiv mice (Fig. 4K). Collectively, our data show that impaired ketogenesis under nutrient-rich conditions increases ASCL1mediated esteri cation of fatty acids. Impaired ketogenesis in human NASH is associated with increased ER-associated ACSL1 and fatty acid re-esteri cation. NASH is the most common liver disease, where excess and sustained lipid accumulation (steatosis) triggers a myriad of pathological changes culminating in liver in ammation, brosis, and cancer 3 . Hepatic steatosis is majorly driven by a mismatch in the delivery (excessive lipolysis and dietary intake) and handling of lipids (fatty acid synthesis, oxidation, storage, and secretion) by the hepatocytes 5 . When the ux of fatty acids is acute, a compensatory increase in FAO and ketogenesis protects the liver from lipotoxicity 40 . However, chronic steatotic conditions decrease FAO and ketogenesis resulting in the accumulation of lipid droplets. Indeed, studies show that ketone body levels decline progressively in NASH patients 13,41,42 . Based on our results, we investigated whether hepatic ketogenesis is associated with elevated fatty acid esteri cation in NASH patients. To this end, we con rmed higher triglyceride levels in deidenti ed liver tissues from NASH patients undergoing surgical resection compared to control human (normal) subjects (Fig. 5A). Our analysis showed no difference in serum triglyceride levels but BHB levels trended lower in NASH patients (Fig S5A and B). Further, the liver from NASH patients showed a signi cant decrease in the protein levels of HMGCS2 and 3-hydroxybutyrate dehydrogenase (BDH1), the latter catalyzes the interconversion of β-hydroxybutyrate and acetoacetate ( Fig. 5B and S5C). A correlation analysis shows that triglyceride levels increase as the hepatic expression of HMGCS2 decreases (Fig. 5C). Since ketogenesis is impaired in NASH, we asked whether ASCL1 translocation is modulated in the NASH livers. We found signi cantly elevated levels of ER-associated ACSL1 in NASH subjects (Fig. 5D). Moreover, we observed a negative correlation between the ER-ACLS1 and HMGCS2 protein expression (Fig. 5E); however, no difference in the mitochondria-associated ACLS1 was noticed (Fig S5D). TLC analysis showed increased levels of esteri ed triglyceride compared to other lipid fractions (Fig. 5F and S5E), suggesting that impaired ketogenesis is associated with increased ER-ASCL1 mediated re-esteri cation of lipids in NASH patients. Similar to mouse models, impaired ketogenesis is also associated with elevated levels of acetylated mitochondrial proteins in human NASH livers ( Fig S5F). Collectively, our data show that impaired ketogenesis induces lipid accumulation in NASH patients in part by promoting the ASCL1-mediated re-esteri cation of fatty acids.
Hepatic acetyl-CoA-L-carnitine homeostasis regulates lipid partitioning via ACSL1. We then sought to understand how ketogenesis regulates the ER translocation of ACSL1. Ketogenesis is the major route of eliminating acetyl-CoA generated by FAO, though some amount of acetyl-CoA would be metabolized via the TCA cycle, or exported out of mitochondria as citrate 12,43 . We found no difference in the acetyl-CoA levels under fed conditions but fasting increased the acetyl-CoA levels by 2-fold in the livers of Hmgcs2 ΔLiv mice (Fig. 6A). Based on the coincidental increase in acetyl-CoA and ER-associated ACSL1 in fasted Hmgcs2 ΔLiv mice, we posited that acetyl-CoA might regulate the subcellular localization of ACSL1. To test our hypothesis, we incubated mouse primary hepatocytes under lipogenic conditions with or without acetate, which increases intracellular acetyl-CoA levels. Western blot analysis showed that acetate increases ER-associated ACSL1 both in the presence and absence of palmitic acid (Fig. 6B and  S6A). Further, acetate increased lipid accumulation in primary hepatocytes (Fig. 6C and S6B). This data suggests that elevated acetyl-CoA induces ER translocation of ACSL1 and exacerbates steatosis under lipogenic conditions. We then interrogated ways to decrease acetyl-CoA levels as a resort to alleviate hepatic steatosis under conditions of impaired ketogenesis. L-carnitine (LC) acts as a buffer for mitochondrial acetyl-CoA by forming acetyl-carnitine, i.e., excess mitochondrial acetyl-CoA exists in the form of acetyl-carnitine 44 . LC is also required for the mitochondrial import of acyl-CoA for FAO. Thus, excess mitochondrial acetyl-CoA reduces mitochondrial FAO by decreasing free LC levels 21,45 . Our analysis showed signi cantly reduced free LC levels in the livers of fasted Hmgcs2 ΔLiv mice (Fig. 6D). We hypothesized that increasing LC would restore lipid homeostasis in Hmgcs2 ΔLiv mice by decreasing hepatic acetyl-CoA levels. To this end, we supplemented the drinking water with LC @ 10mg/ml and fasted the Hmgcs2 ΔLiv mice for 16 hours. No change in body weight was observed by LC supplementation (Fig S6C). LC supplementation reduced acetyl-CoA levels in the fasted Hmgcs2 ΔLiv mice (Fig. 6E). Notably, the livers of LC-treated Hmgcs2 ΔLiv mice were reddish compared to untreated littermates (Fig. 6F). The liver-to-body weight ratio was signi cantly lower and the liver triglyceride levels were signi cantly reduced in LC-treated fasted Hmgcs2 ΔLiv mice ( Fig. 6G and H). H&E analysis con rmed the reduction in hepatic steatosis in LC-treated fasted Hmgcs2 ΔLiv mice (Fig. 6I). We further demonstrate that LC supplementation decreased the ER localization of ACSL1 (Fig. 6J), resulting in reduced esteri cation of triglyceride in the fasted Hmgcs2 ΔLiv mice ( Fig. 6K and L, and S6D). We further noticed that LC decreased the levels of acetylated mitochondrial proteins in fasted Hmgcs2 ΔLiv mice (Fig. 6M). Together, our data indicate that LC recuperates acetyl-CoA homeostasis and alleviates ER-ACSL1-mediated hepatic steatosis in mice with impaired ketogenesis.
L-carnitine alleviates steatosis in human NASH partly by reducing acetyl-CoA and ER-associated ASCL1. We then evaluated the relationship between acetyl-CoA and ACSL1-mediated esteri cation in the livers of NASH patients. As shown in previous studies 42,46 , liver acetyl-CoA levels were elevated in the NASH subjects (Fig. 7A). Correlative analysis showed that ER-associated ACSL1 was in proportion with acetyl-CoA levels in human NASH livers (Fig. 7B). Similar to the mouse models, free L-carnitine (LC) levels were signi cantly reduced in NASH livers and were positively correlated with hepatic HMGCS2 expression ( Fig. 7C and S7A). Previous studies have demonstrated that LC improves steatosis in NASH patients by increasing mitochondrial function and FAO 47,48 . We tested whether LC affects ACSL1-mediated fatty acid esteri cation using primary hepatocytes from NASH patients. BODIPY staining showed reduced lipid accumulation with LC treatment in primary human NASH hepatocytes cultured under lipogenic conditions (Fig. 7D). We demonstrate that LC decreases the ER-associated ACSL1 (Fig. 7E) and fatty acid esteri cation in human NASH primary hepatocytes (Fig S7B). In contrast to mouse livers, LC increased ASCL1 levels in the mitochondria, and the levels of acetylated mitochondrial proteins were decreased ( Fig  S7C and D). Collectively, we demonstrate that LC alleviates steatosis by reducing acetyl-CoA levels and ER translocation of ACSL1 in human NASH.

Discussion
Perturbation in hepatic fatty acid metabolism is the key risk factor driving several metabolic diseases, including obesity, diabetes, and NAFLD. An increase in free fatty acid in ux from peripheral tissue, augmented lipid biosynthesis, and impaired lipid disposal in the liver lead to triglyceride accumulation triggering hepatic steatosis. Lipid disposal pathways, in particular β-oxidation play an essential role in eliminating excessive hepatic lipids under physiological and pathological conditions 40,49,50 . Indeed, under fasting conditions, the liver disposes up to 250g of lipid per day through β-oxidation and ketogenesis 51 . However, it remains unclear whether hepatic ketogenesis is just a regulator of peripheral energy homeostasis or has a role in regulating hepatic lipid metabolism. We show that the mouse with temporal disruption of hepatic ketogenesis develops severe steatosis at fasting. While there is no change in fatty acid uptake, de novo lipogenesis, and β-oxidation pathways, ketogenesis is found to regulate fatty acid partitioning via ER-associated ACSL1. We show that the accumulation of acetyl-CoA in response to ketogenic insu ciency induces ER translocation of ACSL1 and fatty acid esteri cation. Thus, incoming fatty acids are mistargeted for storage but not FAO, as elevated acetyl-CoA (from ketogenic insu ciency) is misinterpreted as excessive FAO. We demonstrate that L-carnitine supplementation in ketogenic insu ciency mice counteracts the acetyl-CoA-mediated increase in ER-associated ACSL1 and triglyceride esteri cation. We also report elevated acetyl-CoA, ER-associated ACSL1, and enhanced triglyceride esteri cation in NASH patients eliciting impaired ketogenesis. Thus, we de ne a novel role of ketogenesis in regulating hepatic lipid metabolism.
Recent studies demonstrated a causal role of ketogenesis in fatty liver pathogenesis using neonates and adult mice with ketogenic insu ciency (via disruption of hepatic HMGCS2) 14,15 . These mice develop fatty liver when fed lipid-rich mothers' milk and a high-fat diet 15 . We show that acute fasting is su cient to induce steatosis in ketogenesis insu cient mice. Despite the availability of food ad-lib, mice tend to eat at night, i.e. they undergo physiological fasting at day time. No difference in liver triglyceride levels in ad-lib-fed ketogenesis insu cient mice indicates that a small increase in fatty acids from physiological fasting is well-tolerated. Refeeding induces the secretion of hepatic triglycerides as VLDL, which are taken up by peripheral tissues, including adipose tissue, heart, kidney, and vascular tissue 50 . Previous studies showed that ketogenesis regulates the composition of hepatic lipids 16,52 ; therefore, investigating the impact of ketogenic insu ciency on peripheral lipid homeostasis may shed light on the mechanistic underpinnings of metabolic diseases, including cardiovascular disease.
Fasting-mediated adipose tissue lipolysis and a drop-in insulin levels coordinate to induce hepatic ketogenesis. Thus, inhibition of adipose tissue ATGL decreases the circulating ketone body levels 27 . Under these circumstances, ketone body levels are tightly linked to fatty acid oxidation, evident from PPARα knockout mice eliciting diminished fasting-associated ketogenesis despite no difference in circulating fatty acids 28 . Lipid accumulation in ketogenesis insu ciency mice was not due to impairment in hepatic PPARα signaling or lipogenic pathways. Our data show increased fatty acid storage evident from microsteatosis and increased expression of lipid droplet-associated proteins such as Plin2 and FSP27. This led us to speculate that ketogenetic insu ciency promotes lipid accumulation via the esteri cation of fatty acids.
Free fatty acids entering the hepatocytes are immediately esteri ed into acyl CoA by ACSL1. ACLS1 in the outer mitochondria membrane directs acyl-CoA for β-oxidation 19,20 , whereas ER-localized ACSL1 generates lipid substrates for triglyceride synthesis via re-esteri cation 49 . Nonetheless, the mechanisms regulating the ER translocation of ACSL1 are unclear. We demonstrate that ketogenesis is tightly linked to ER localization of ACSL1 in fasted mice. The signi cance of ER-ACSL1 in triglyceride synthesis via reesteri cation is also established in human NASH livers, where the translocation of ACSL1 to ER-localized is dramatically increased and is strongly associated with lower expression of HMGCS2. Thus, our data suggest that impairment in hepatic ketogenesis increases ER localization of ACSL1 in both human and mouse NASH models.
Acetyl-CoA acts as a metabolic node regulating glucose, lipid, and amino acid metabolism 53 . Acetyl-CoA levels increase during fasting and high-fat feeding due to elevated FAO 54 . The conduit of acetyl-CoA as ketone bodies is necessary to maintain cellular acetyl-CoA levels. When FAO exceeds ketogenesis, the increase in acetyl-CoA pauses mitochondrial FAO by acetylating the mitochondrial proteins 55,56 . Until acetyl-CoA homeostasis recuperates, the fatty acids are channeled toward ER for storage as triglycerides.
Despite an increase in the acetylation of mitochondrial proteins, mitochondrial function was similar between the genotypes suggesting that mitochondrial dysfunction is not the primary cause of lipid accumulation in acute fasted ketogenic insu cient mice. Excess acetyl-CoA shuttles to the cytosol as citrate, where cytosolic ATP citrate lyase (ACLY) breaks down citrate into acetyl-CoA and oxaloacetate 53 .
Since ACLY expression is upregulated in the fasted ketogenic insu cient mice, we interrogated whether acetyl-CoA spillover into the cytosol acts as a signal for the ER translocation of ACSL and lipid esteri cation. Acetyl-CoA levels could be increased by treating with the short-chain fatty acid acetate, which is metabolized by the enzyme acetyl-coenzyme A synthetase to generate cytosolic acetyl-coA 57 .
Our data that acetate increases ER-translocation of ACSL1 and lipid-laden hepatocytes support the hypothesis that cytosolic acetyl-CoA promotes fatty liver by inducing the esteri cation of lipids via ACSL1.
Mitochondrial membranes are impermeable to fatty acyl-CoA; therefore, fatty acyl-CoA is esteri ed with Lcarnitine to form acyl-carnitine by the mitochondrial enzyme CPT1. In the mitochondrial matrix, CPTII releases the free carnitine from acyl-carnitine forming acyl-CoA. The free L-carnitine is then exported back to the cytosol to participate in mitochondrial fatty acid transport. L-carnitine also buffers acetyl-CoA by forming acetyl-carnitine, which is excreted in the urine 58, 59 . Studies show that carnitine levels decrease in the latter state of liver disease 22,60 , indicating that carnitine de ciency may not trigger hepatic steatosis but could exacerbate disease progression. However, patients with primary carnitine de ciency due to a lack of organic cation/carnitine transporter develop hepatomegaly and hepatic steatosis, which could be alleviated by L-carnitine supplementation 61-63 . Moreover, studies demonstrate that L-carnitine is ratelimiting for ketogenesis from endogenous fatty acids. For instance, ketosis with carnitine supplementation improves glucose homeostasis, insulin sensitivity, and lipid pro le 64,65 . Thus, carnitine levels are a key determinant of lipid metabolism, ketogenesis, and hepatic steatosis. Here, we demonstrate that carnitine supplementation attenuates hepatic steatosis under ketogenic insu ciency by alleviating ER translocation of ACSL1. Carnitine also improves mitochondrial function by decreasing the levels of acetylated mitochondrial proteins 44 . No export mechanism exists to excrete cellular acetyl-CoA. We posit that carnitine, via forming acetyl-carnitine, acts as a conduit to remove excess acetyl-coA.
Several studies have demonstrated the bene cial anti-in ammatory role of acetyl-carnitine in metabolic diseases 66,67 . Further investigation is required to determine whether carnitine sequestration as acetylcarnitine mediates the anti-in ammatory response in NASH patients exhibiting impaired ketogenesis.
While it is true that ketogenesis is a sel ess function of the liver in providing nutrients to other organs 12 , our data show that the conduit of acetyl-CoA via ketogenesis is crucial to regulate mitochondrial acetylome, ER homeostasis, and lipid partitioning in the liver. Dysregulated acetyl-CoA induces hepatic steatosis via ACSL1-mediated TG synthesis in the ER 19,20,68 . This mechanism particularly plays a crucial role in conditions of increased hepatic fatty acid in ow, such as fasting, overnutrition, and fatty liver disease. Although pharmacological inhibition of ACSL1 ameliorates steatosis 18,20 , the risk of liver injury is enhanced by the accumulating free fatty acids 69 . We provide empirical evidence that L-carnitine inhibits ER translocation of ACSL1 and hepatic steatosis by buffering acetyl-CoA. However, L-carnitine should be recommended with utmost care as carnitine metabolism in the intestine generates TMAO, a metabolite with atherogenic and carcinogenic properties 70,71 . Our study provides a rationale for the therapeutic use of L-carnitine in a subset of NAFLD patients exhibiting lower L-carnitine and ketone bodies. As carnitine is excreted as acetyl-carnitine 72 , therapeutic interventions with L-carnitine could be monitored non-invasively in NASH patients by measuring the urinary acetyl-carnitine levels. Hmgcs2 F/F and Hmgcs2 ΔLiv mice. All the data is presented as mean ± SEM. p < 0.05 ( * ) or p < 0.001 ( * * * ) or p < 0.0001 ( * * * * ) analyzed by One-way ANOVA (Tukey multiple-comparisons test).

Figure 2
Oxphos proteins in the liver of Hmgcs2 ΔLiv mice. (C) WB analysis of acetyl-lysine (AcK) in the hepatic mitochondrial fractions from 16 h fasted Hmgcs2 ΔLiv mice. (D and E) Mitochondrial respiration devoted to oxidative phosphorylation or maximum electron transport chain activity in the uncoupled state. (F) Liver citrate synthase enzymatic activity. All the data is presented as mean ± SEM. Data was analyzed by One-way ANOVA (Tukey multiple-comparisons test) and Two-tailed Student t-test  in the mitochondrial fraction of fasted Hmgcs2 ΔLiv mice treated with or without LC. All the data is presented as mean ± SEM. p < 0.05 ( * ); or p < 0.01 ( * * ) or p < 0.001( * * * ) analyzed by Two-tailed Student t test. WB analysis of ACSL1 in the microsomal fraction of primary human hepatocytes from NASH subjects pre-treated with 1mM L-carnitine for 6 h and loaded with 200mm BSA-PA for 16 h. (F) Schematic diagram. All the data is presented as mean ± SEM. Correlative analysis was performed by using a nonparametric Spearman's test. p < 0.05 ( * ) or p < 0.01 ( * * ) analyzed by the Two-tailed Student t-test.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.