Hepatic ketogenesis is not required for exercise training to mitigate diet-induced liver steatosis in male mice

Abstract

Accelerated hepatic fatty acid oxidation during acute exercise has been proposed as a contributor to the antisteatotic effects of exercise training. Ketogenesis, which produces acetoacetate (AcAc) and β-hydroxybutyrate (βOHB) from fatty acids, is stimulated by exercise and supports fat oxidation. This study tested the hypothesis that hepatic ketogenesis is necessary for exercise training to lower liver lipids. Liver-specific 3-hydroxymethylglutaryl-CoA synthase 2 knockout (HMGCS2 KO) mice and wild-type (WT) littermates underwent sedentary, acute exercise (treadmill running), and exercise training (6-wk treadmill running regime) protocols. Liver ketone bodies and lipids were determined via mass spectrometry. Stable isotope infusions in conscious, unrestrained mice defined mitochondrial oxidative fluxes during rest and treadmill running. In untrained mice, hepatic HMGCS2 deletion lowered liver AcAc and βOHB and impaired their increase during acute exercise. Liver triacylglycerides (TAGs) were comparable between genotypes at rest (ad libitum fed and short-fasted conditions). In contrast, liver TAGs were higher in HMGCS2 KO compared with WT mice following acute, nonexhaustive exercise. Acute exercise stimulated TCA cycle flux in both genotypes; however, liver TCA cycle flux was higher in KO mice during rest and acute exercise. This suggests that enhanced lipid oxidation via the TCA cycle may be sufficient for TAG homeostasis in HMGCS2 KO mice at rest, but not during acute exercise. Exercise training decreased liver TAGs similarly in WT and KO mice when assessed under short-fasted conditions. In conclusion, hepatic ketogenesis supports liver lipid homeostasis during acute exercise, but is not required for exercise training to mitigate diet-induced fatty liver.

NEW & NOTEWORTHY Exercise training has been proposed to mitigate liver steatosis partly through enhanced hepatic fat oxidation. During acute exercise, the oxidation of fatty acids to ketone bodies is stimulated. This study tested the hypothesis that hepatic ketogenesis was required for exercise training to reduce liver fat in mice. The results show that hepatic ketogenesis supports lipid homeostasis during acute exercise, but is not necessary for exercise training to mitigate diet-induced fatty liver.

INTRODUCTION

Regular exercise is a first-line intervention for metabolic dysfunction-associated steatotic liver disease (MASLD) (1). The effectiveness of aerobic exercise to mitigate liver steatosis has been linked to its ability to enhance hepatic lipid oxidation both acutely and chronically (2–4). During acute aerobic exercise, hepatic fat oxidation (fatty acid catabolism to acetyl-CoA via β-oxidation and its complete oxidation to CO₂ in the TCA cycle) is functionally coupled to glucose production. More specifically, liver glucose production is increased to match the heightened glucose demands of muscular work (5, 6). The enhanced liver glucose production is accomplished partly by a rise in gluconeogenesis, an energetically costly process that contributes to the greater hydrolysis of ATP to ADP and AMP during exercise (5–11). To blunt this energy discharge, fatty acid oxidation to CO₂ via TCA cycle flux and oxidative phosphorylation is stimulated (2, 9, 10). Thus, exercise accelerates fat oxidation so that the potential energy in fatty acids is transformed to ATP to support gluconeogenesis and, subsequently, working muscle (12). Of note, this acceleration of fatty acid oxidation to support gluconeogenesis and, subsequently, muscle glucose demands becomes increasingly more important as aerobic exercise duration is prolonged and glycogen levels in the liver and muscle decline (6, 12, 13). We previously tested whether the exercise-mediated elevation in fat oxidation via the TCA cycle was required for exercise training to lower liver lipids. Loss of hepatic phosphoenolpyruvate carboxykinase 1 (PCK1) prevented the rise in gluconeogenic and TCA cycle flux during acute exercise in mice (10). However, the lowering of liver triacylglycerides (TAGs) by training was not impaired (10).

In addition to the TCA cycle, acute exercise engages other hepatic lipid catabolism pathways responsible for transforming the potential energy in fatty acids to alternative forms of energy currency that are then released from the liver to support muscular work. This includes promoting the production of acetoacetate (AcAc) and β-hydroxybutyrate (βOHB) from fatty acids via ketogenesis (14). Ketogenesis is a mitochondrial pathway initiated by 3-hydroxymethylglutaryl-CoA synthase 2 (HMGCS2), which synthesizes HMG-CoA from β-oxidation-derived acetoacetyl-CoA and acetyl-CoA (15). Next, HMG-CoA lyase (HMGCL) cleaves HMG-CoA to produce AcAc (15). AcAc can be released into the circulation; however, most is reduced by βOHB dehydrogenase 1 (BDH1) to βOHB before hepatic output (14, 15). Exercise provokes ketogenesis by increasing fatty acid delivery to the liver, elevating hepatic extraction of fatty acids from the circulation, and enhancing the intrahepatic conversion of fatty acids to ketone bodies (6, 16). Although these ketogenic responses to exercise have received significant investigation, it remains unclear whether the stimulation of ketogenesis during acute exercise contributes to the efficacy of exercise training to combat liver steatosis.

The experiments presented herein tested the hypothesis that hepatic ketogenesis is necessary for exercise training to lower liver lipids. Liver-specific HMGCS2 knockout (KO) mice and wild-type (WT) littermates were fed a Gubra Amylin-NASH (GAN) diet to promote fatty liver. WT and liver-specific HMGCS2 KO mice were untrained or exercise-trained via treadmill running. Targeted metabolomics was completed to quantify liver ketone bodies, shotgun lipidomics was performed to determine liver lipids, indirect calorimetry assessed whole body energy metabolism, and stable isotope infusions in conscious unrestrained mice defined glucose and mitochondrial oxidative fluxes at rest and during acute exercise. The results of this study show that hepatic ketogenesis promotes liver lipid homeostasis during acute exercise, but is not necessary for exercise training to mitigate diet-induced fatty liver.

MATERIALS AND METHODS

Mouse Models and Husbandry

The University of Minnesota Institutional Animal Care and Use Committee approved all experimental procedures. Mice on a C57BL/6NJ background with a hepatocyte-specific KO of HMGCS2 were generated by breeding mice expressing Cre recombinase under control of the albumin promoter (B6.Cg-Tg(Alb-cre)21Mgn/J) with HMGCS2 floxed mice as previously described (17). Floxed HMGCS2 littermates of the liver-specific HMGCS2 KO mice that did not express Cre recombinase were used as WT controls. All mice were fed Teklad Global 18% Protein Rodent Diet (18% kcal from fat, 58% kcal from carbohydrate, and 24% kcal from protein; 2918, Inotiv, Indianapolis, IN) upon weaning at 3 wk of age. At 6 wk of age, mice were provided a Gubra Amylin-NASH (GAN) diet [40% kcal from fat (mostly palm oil), 20% kcal from fructose, and 2% cholesterol; D09100310; Research Diets Inc., New Brunswick, NJ]. WT and liver HMGCS2 KO mice remained on the GAN diet for 6, 12, or 32 wk until they were euthanized at 12, 18, or 38 wk of age. One cohort of WT and liver HMGCS2 KO mice received a ketogenic diet (90.5% kcal from fat and 9.5% kcal from protein; TD.160153, Inotiv, Indianapolis, IN, n = 5 per genotype) for 72 h prior to being euthanized at 12 wk of age. Additional studies used hepatocyte-specific BDH1 KO mice on a C57BL/6NJ background that were generated by crossing B6.Cg-Tg(Alb-cre)21Mgn/J mice with BDH1-floxed mice as previously described (18). Cre-negative, floxed BDH1 littermates were used as WT controls. These mice were fed the Teklad Global 18% Protein Rodent Diet (2918, Inotiv, Indianapolis, IN) from weaning until euthanized at 12 wk of age (n = 14 per genotype). Deletion of liver HMGCS2 and BDH1 was confirmed by both PCR genotyping and immunoblotting. One cohort of 16-wk-old C57BL/6J mice (Strain No. 000664, The Jackson Laboratory, Bar Harbor, ME, n = 5) fed the Teklad Global 18% Protein Rodent Diet (2918, Inotiv, Indianapolis, IN) was studied.

Mice were housed with cellulose bedding (Cellu-nest, Shepherd Specialty Papers, Watertown, TN) in temperature-controlled (∼22°C) and humidity-controlled (30%–70%) conditions maintained on a 14:10 h light:dark cycle. Both food and water were provided to mice ad libitum. Male mice were used for all experiments and were euthanized via cervical dislocation under one of five conditions: ad libitum fed (n = 8 per genotype), 6.5-h fasted (n = 8 per genotype in mice fed a GAN diet for 6 wk; n = 5 per genotype in mice fed a GAN diet for 32 wk), 18-h fasted (n = 7 or 8 per genotype), following a 60-min treadmill run at 45% of maximal running speed (n = 8 per genotype), and following a treadmill run to exhaustion at 22.5 m·min⁻¹ (n = 7–12 per genotype). Each condition was conducted sequentially as an independent experiment to assess the effect of genotype within a specific physiological state, precluding direct comparisons between the different conditions. Liver and skeletal muscles were rapidly excised, freeze-clamped in liquid nitrogen, and stored at −80°C. For 6.5-h fasting conditions, food was withdrawn during the first hour of the light cycle. For 18-h fasting conditions, food was removed during the last hour of the light cycle.

Exercise Stress Test

The exercise stress test was performed to determine maximal running speed as previously described (9, 19). The maximal running speed obtained is a marker of cardiopulmonary fitness, given that it positively correlates with maximum whole body oxygen uptake (19). Mice completed one (for acute exercise experiments) or two (for exercise training experiments) exercise stress tests. The stress test for acute exercise experiments was performed 48 h before the acute exercise bout. For exercise training, the first stress test was completed 24 h before the start of a 6-wk exercise training protocol. The second was performed 24 h following the completion of the 6-wk training protocol. Mice were acclimated to running on an enclosed single-lane treadmill (Columbus Instruments, Columbus, OH) by performing two 10-min exercise bouts at 10 m·min⁻¹ (0% incline) 24 and 48 h before the exercise stress test. For the exercise stress test, mice were placed in the enclosed single-lane treadmill. Following a 10-min sedentary period, mice initiated running at 10 m·min⁻¹ (0% incline). The treadmill belt speed was increased by 4 m·min⁻¹ every 3 min until exhaustion. Exhaustion was defined as the point in which the mouse remained on the shock grid at the back of the treadmill for greater than 5 s. A workload index was determined to provide an assessment of the impact of differences in body weight on maximal running speed outcomes. It was calculated as the product of body weight (kg) and distance traveled during the stress test (m).

Exercise Training Protocol

The exercise training protocol was completed as previously outlined (10). In brief, mice began receiving the GAN diet at 6 wk of age, and the exercise training was initiated when mice were 12 wk of age. Trained mice completed a 60-min treadmill running bout, 5 days a week for 6 wk (n = 8 or 9 per genotype). The treadmill running speed was 45% of the mouse’s initial maximal running speed during the first 2 wk of the training protocol and increased by 5% of the initial maximal running speed every 2 wk. Untrained mice (n = 8 or 9 per genotype) were placed in an enclosed container on top of the treadmill. Mice were euthanized 72 h following the final exercise training bout and 48 h after the final exercise stress test mice.

Exercise to Exhaustion (Endurance) Test

Twelve-week-old, untrained mice were acclimated to running on an enclosed single-lane treadmill (Columbus Instruments, Columbus, OH) as outlined for the exercise stress tests. The endurance test was completed 48 h after the final acclimation bout. The treadmill belt speed started at 10 m·min⁻¹ (0% incline) and increased by ∼4.2 m·min⁻¹ every 3 min until 22.5 m·min⁻¹ was reached. The treadmill speed remained at 22.5 m·min⁻¹ until exhaustion. Of note, exercise stress tests were not performed in the mice that completed the endurance test. Based on the stress test of other cohorts, 22.5 m·min⁻¹ is ∼70%–75% of maximal running speed.

Body Composition

Body composition was determined in both untrained and trained mice in the exercise training cohort using an EchoMRI-100 Body Composition Analyzer (EchoMRI LLC, Houston, TX).

Indirect Calorimetry

Indirect calorimetry was performed using the Promethion Core System (Sable Systems International, North Las Vegas, NV). Mice were individually housed for 6 days on ALPHA-dri PLUS bedding (Shepherd Specialty Papers, Watertown, TN) in temperature (∼22°C)- and humidity (30%–70%)-controlled conditions maintained on a 12:12-h light:dark cycle. The first 3 days were the acclimation period followed by 3 days of data collection to determine food intake, ambulatory activity, voluntary running wheel characteristics, oxygen consumption (V̇o₂), CO₂ production (V̇co₂), respiratory exchange ratio (RER; V̇co₂/V̇o₂), and energy expenditure. Energy expenditure was calculated from V̇o₂ and V̇co₂ using the Weir equation (20). Mice (n = 7 or 8 per genotype) used in the indirect calorimetry experiments were 12 wk of age, had been receiving a GAN diet (D09100310; Research Diets Inc., New Brunswick, NJ) ad libitum for 6 wk, and were untrained with no prior exposure to voluntary running wheels. Access to running wheels only occurred during the indirect calorimetry experiments and were included to provide insight into the impact of ketogenic insufficiency on voluntary exercise characteristics. These data were obtained to complement and facilitate interpretation of results from the maximal running speed and endurance tests conducted via forced treadmill running.

Surgical Procedures

An initial cohort of 15-wk-old, male C57BL/6J mice underwent surgery to implant a carotid artery catheter for measurements of circulating βOHB. This initial cohort of mice was studied to confirm the running intensity during acute exercise and exercise training protocols was sufficient to increase ketone body concentrations. Seventeen-week-old, untrained WT and liver HMGCS2 KO mice had catheters implanted in the jugular vein and carotid artery for isotope infusion and sampling protocols as previously described (10, 21). The exteriorized ends of the implanted catheters were flushed with 200 U·mL⁻¹ heparinized saline and sealed with stainless-steel plugs. Following surgery, mice were housed individually and provided ∼9 or 10 days of postoperative recovery before stable isotope infusion and/or arterial sampling experiments. On postoperative days 5–8, mice underwent acclimation and stress test exercise bouts.

Stable Isotope Infusions

Stable isotope infusions were performed at rest and during acute exercise in ∼18-wk-old mice (n = 7 per genotype) as previously described (10). During the first hour of the light cycle, both food and water were withdrawn for the remainder of the experiment (7 h). Two hours into the fast, mice were moved to an enclosed single-lane treadmill, and the exteriorized vascular catheters were connected to infusion syringes. Mice were given a 1-h acclimation period before obtaining an 80 μL arterial blood sample to determine natural isotopic enrichment of plasma glucose. Immediately after this sample acquisition, a stable isotope infusion protocol was started to facilitate the quantification of endogenous glucose production and associated oxidative fluxes as previously described (10, 21, 22). In brief, a ²H₂O (99.9%) bolus containing [6,6-²H₂]glucose (99%) was intravenously infused for 25 min to enrich body H₂O and provide a [6,6-²H₂]glucose prime (440 μmol·kg⁻¹). An independent, continuous infusion of [6,6-²H₂]glucose (4.4 μmol·kg⁻¹·min⁻¹) was initiated after the ²H₂O bolus and [6,6-²H₂]glucose prime. A primed (1.1 mmol·kg⁻¹), continuous (0.055 mmol·kg⁻¹·min⁻¹) intravenous infusion of [U-¹³C]propionate (99%, sodium salt) was started 2 h after the ²H₂O bolus and [6,6-²H₂]glucose prime. Four arterial blood samples (100 µL) were obtained 90–120 min following the [U-¹³C]propionate bolus (time = 0–30 min of treadmill running) to determine arterial glucose, insulin, glucagon, and nonesterified fatty acids (NEFAs). These samples were also used to complete ²H/¹³C metabolic flux analysis. The sample acquired at 90 min following the [U-¹³C]propionate bolus (time = 0 min) was obtained while mice were sedentary on a stationary treadmill. Samples taken 100–120 min following the [U-¹³C]propionate bolus (time = 10–30 min of exercise) were acquired while mice were completing an acute treadmill running bout at 45% of their maximal running speed. Plasma samples were stored at −80°C. Donor red blood cells were resuspended in 4.5% ²H₂O-10 U·mL⁻¹ heparinized saline (∼0.4–0.5 vol/vol) and intravenously infused throughout the experiment to maintain hematocrit between ∼0.4 and 0.5 vol/vol. Mice were removed from the treadmill and euthanized by cervical dislocation immediately after the final sample was taken. Liver and skeletal muscle were rapidly excised, freeze-clamped in liquid nitrogen, and stored at −80°C.

Glucose Derivatization and Gas Chromatography-Mass Spectrometry Analysis

Approximately 40 µL of plasma was acquired before starting the stable isotope infusions and at the 0, 10, 20, and 30 min time points of the treadmill running bout to prepare di-O-isopropylidene propionate, aldonitrile pentapropionate, and methyloxime pentapropionate derivatives of glucose (9). Gas chromatography-mass spectrometry (GC-MS) analysis was performed, and uncorrected mass isotopomer distributions (MIDs) for six fragment ions were determined as previously described (10, 21). In brief, GC-MS protocols were performed using a Thermo Scientific ISQ 7000 GC-MS system coupled to a Thermo Scientific TRACE 1310 gas chromatograph (Waltham, MA) with an HP-5 ms capillary column (19091S-433, Agilent Technologies Inc., Santa Clara, CA). The GC-MS protocol for methyloxime derivatives was run using a column temperature of 80°C for 1 min, which increased to 280°C at a rate of 20°C·min⁻¹ and held for 4 min. The column temperature was then ramped up to 325°C at 40°C·min⁻¹. For aldonitrile and di-O-isopropylidene derivatives, the column temperature was 80°C for 1 min, ramped up to 280°C at 10°C·min⁻¹, and held for 4 min until being increased to 325°C at 40°C·min⁻¹. The MS was run in scan mode for methyloxime (m/z 140–260), aldonitrile (m/z 100–500), and di-O-isopropylidene derivatives (m/z 301–314). A custom MATLAB function was used to integrate derivative peaks and determine MIDs. MIDs were obtained for six glucose fragment ions. The following fragment ion ranges were used for determining MIDs: methyloxime, m/z 145–149; aldonitrile, m/z 173–177, 259–264, 284–288, 370–379; and di-O-isopropylidene, m/z 301–314. All sample derivatives were run in duplicate.

²H/¹³C Metabolic Flux Analysis

The in vivo metabolic flux analysis used in this study has been detailed previously (22, 23). In brief, a reaction network was generated using Isotopomer Network Compartmental Analysis (INCA) software (24). This reaction network defined both carbon and hydrogen transitions for endogenous glucose production and associated oxidative metabolism reactions. The flux through each network reaction was determined relative to citrate synthase flux (V_CS) by minimizing the sum of squared residuals between experimentally determined and simulated MIDs of the six fragment ions previously described (23, 25). Flux estimations were repeated 50 times from random initial values. Goodness of fit was assessed by a chi-square test (P = 0.05). Body weights of mice and the infusion rate of [6,6-²H₂]glucose were used to determine absolute values.

Circulating Metabolite and Hormone Analyses

Blood glucose was measured via Contour blood glucose meters (Ascensia Diabetes Care, Parsippany, NJ). Plasma NEFAs were quantified with the Wako HR series NEFA-HR(2) assay (FUJIFILM Medical Systems USA, Lexington, MA). A Precision Xtra Blood Glucose & Ketone meter (Abbott Diabetes Care, Alameda, CA) was used to measure blood β-ketone bodies in the C57BL/6J mouse experiments. Plasma insulin was determined using the Mercodia Mouse Insulin ELISA (10-1247-01, Winston Salem, NC). Plasma glucagon was quantified via the Mercodia Glucagon ELISA (10–1281-01, Winston Salem, NC). Circulating glucose, β-ketone bodies, insulin, and glucagon were determined via single measurements. Plasma NEFAs were measured in duplicate.

Liver and Skeletal Muscle Enzyme and Metabolite Analysis

For citrate synthase activity, gastrocnemius (∼10–20 mg) was homogenized in a potassium phosphate buffer (100 mM KPO₄, 5 mM EDTA, and 5 mM EGTA at pH 7.4), freeze-thawed three times, and assayed via a spectrophotometric assay in triplicate at 25°C as previously described (26). For glycogen measurements, 30–50 mg gastrocnemius and liver samples were homogenized in 0.03 N HCl, incubated in an amyloglucosidase-sodium acetate (0.2 M NaOAc) buffer, and the resulting free glucose was determined in duplicate by an enzymatic assay as previously described (27). For TAG measurements, 30–50 mg of gastrocnemius, superficial vastus lateralis, and liver were homogenized in a 3 M KOH-ethanol solution, heated at 70°C for 1 h, and incubated overnight at room temperature before neutralization with 2 M Tris-HCl (pH 7.5). TAGs were determined in triplicate with the Triglycerides—Liquid Reagent Set (Pointe Scientific, Inc., Lincoln Park, MI). Shotgun lipidomics was performed via a single measurement for liver samples (∼30–50 mg) from one cohort of untrained and trained mice to quantify liver TAGs, diacylglycerides (DAGs), cholesterol esters, phosphatidylethanolamine, phosphatidylcholine, and phosphatidylglycerol as previously described (28).

Liver AcAc, βOHB, and total ketone bodies (TKBs; defined as the sum of AcAc and βOHB) were quantified via single measurement by ultra-high-performance liquid chromatography tandem mass spectrometry (UHPLC-MS/MS) as previously detailed (29). In brief, liver samples (∼40 mg) were homogenized in an acetonitrile (ACN):methanol (MeOH):water (2:2:1 vol/vol/vol) solution with [U-¹³C₄]AcAc and D-[3,4,4,4-²H₂] βOHB internal standards. Liver homogenates underwent three cycles of vortexing (10 s), flash freezing in liquid nitrogen (30 s), and sonication at 25°C (5 min). Samples were then centrifuged for 10 min at 15,000 g and 4°C, and the supernatants were transferred to LC-MS vials. Analysis of samples was performed using a Thermo Fisher Scientific Vanquish LC system and a Cortecs T3 column (186008500, Waters Corporation, Milford, MA) coupled to a Thermo Fisher Scientific QExactive Plus hybrid quadrupole-orbitrap mass spectrometer with a heated electrospray ionization source.

Immunoblotting

Liver homogenates were prepared as previously described (10, 21, 22). Liver proteins (15 µg) were separated via gel electrophoresis on a NuPAGE 4%–12% Bis-Tris gel (Invitrogen, Carlsbad, CA) and transferred to a PVDF membrane. The antibodies used for immunoblotting are provided in Supplemental Table S1. Following incubation with primary and secondary antibodies, PVDF membranes were treated with a chemiluminescent substrate (Thermo Fisher Scientific, Waltham, MA), and images were obtained using a ChemiDoc Imaging system and Image Lab software (Bio-Rad, Hercules, CA). A BLOT-FastStain (G-Bioscience, St. Louis, MO) assay was completed to determine total protein, which was used as the loading control. Densitometry was completed using ImageJ software.

Statistical Analyses

GraphPad Prism software (GraphPad Software LLC., San Diego, CA) was used to perform statistical analyses. Given that acute conditions were completed as independent, sequential experiments, comparisons between genotypes were performed using individual Student’s t tests for each condition. A one-way repeated measures ANOVA was used for analysis of C57BL/6J mice experiments. A two-way repeated measures ANOVA was used for comparisons in the ²H/¹³C metabolic flux analysis experiments. Two-way and three-way ANOVAs were used for comparisons in exercise training experiments. If a significant interaction was detected for two- and three-way ANOVAs, a Sidak post hoc test was performed. Statistical differences were considered significant if P < 0.05. All data are reported as means ± SE.

RESULTS

Liver TAGs Are Higher in Liver HMGCS2 KO Mice Compared With WT Mice After a 60-Min Exercise Bout

Initial studies fed mice a GAN diet for 6 wk to model the early stages of MASLD. To determine the impact of hepatic HMGCS2 deletion on liver ketone bodies and TAGs, we quantified these metabolites in WT and liver HMGCS2 KO mice across four distinct experimental cohorts: ad libitum fed, 6.5-h fasted, 60-min exercise (45% of maximal running speed), and exercise to exhaustion (Fig. 1, A–C). Within the ad libitum-fed cohort, loss of hepatic HMGCS2 lowered liver AcAc, βOHB, and TKBs (Fig. 1D). However, liver TAGs were similar between genotypes (Fig. 1D). This pattern persisted following a 6.5-h fast, where HMGCS2 KO mice showed lower liver ketone bodies but maintained liver TAG levels comparable with WT mice (Fig. 1E). Following 60 min of treadmill running, liver HMGCS2 KO mice again exhibited lower liver AcAc, βOHB, and TKBs compared with WT controls (Fig. 1F). Notably, the 60-min exercise bout showed a genotype effect for liver TAGs. Specifically, liver TAGs were higher in liver HMGCS2 KO mice compared with WT littermates after the 60-min treadmill run (Fig. 1F). In the exhaustive exercise cohort, although ketone bodies remained lower in liver HMGCS2 KO mice (Fig. 1G), the genotype difference in liver TAGs was no longer observed, with levels being similar between WT and KO mice at the point of exhaustion (Fig. 1G).

Figure 1.

Liver triacylglycerides (TAGs) are higher in liver 3-hydroxymethylglutaryl-CoA synthase 2 knockout (HMGCS2 KO) mice compared with wild-type (WT) mice after a 60-min exercise bout. A: a schematic representation of the experimental design. Male, liver-specific HMGCS2 knockout (KO) mice and wild-type (WT) littermates were fed a Gubra Amylin-NASH (GAN) diet for 6 wk. B: a representative immunoblot of liver HMGCS2 in WT and liver-specific HMGCS2 KO mice. C: a schematic representation of the ketogenic pathway with select metabolites and enzymes. D: liver acetoacetate (AcAc; nmol·g⁻¹), β-hydroxybutyrate (βOHB; nmol·g⁻¹), total ketone bodies (TKBs; nmol·g⁻¹), and liver triacylglycerides (TAGs) in ad libitum-fed mice (n = 7 or 8 per genotype). E: liver AcAc (nmol·g⁻¹), βOHB (nmol·g⁻¹), TKB (nmol·g⁻¹), and liver TAGs in 6.5-h fasted liver HMGCS2 KO mice (n = 7 or 8 per genotype). F: liver AcAc (nmol·g⁻¹), βOHB (nmol·g⁻¹), TKB (nmol·g⁻¹), and liver TAGs in liver HMGCS2 KO mice exercised on a treadmill for 60 min at 45% of their maximal running speed (n = 6–8 per genotype). G: liver AcAc (nmol·g⁻¹), βOHB (nmol·g⁻¹), TKB (nmol·g⁻¹), and liver TAGs in liver HMGCS2 KO mice exercised on a treadmill at 22.5 m·min⁻¹ until exhaustion (n = 7–12 per genotype). Data are presented as means ± SE. Statistical differences were determined by Student’s t tests and accepted as significant if P < 0.05 for all data in D–G. *P < 0.05, **P < 0.01, and ****P < 0.0001 between specified groups. Figure created with a licensed version of BioRender.com.

The use of internal standards in our LC-MS measurements of ketone bodies allowed for a cross-condition comparison (Supplemental Fig. S1). A two-way ANOVA revealed that the 60-min exercise bout increased liver βOHB and TKBs in WT mice compared with all other conditions, including the exhaustive exercise state (Supplemental Fig. S1). These data suggest that nonexhaustive exercise triggers a ketogenic response in WT mice, which was prevented in the absence of hepatic HMGCS2. Furthermore, these findings suggest that HMGCS2-dependent ketogenesis supports liver lipid homeostasis during acute, nonexhaustive exercise.

Liver TAGs Are Higher in Liver HMGCS2 KO Mice Compared With WT Mice Under Conditions Characterized by Acutely Elevated Lipid Availability

Prior, independent work has determined that exercise is accompanied by increased fatty acid delivery to the liver (6), which could place greater reliance on hepatic ketogenesis to maintain liver lipid homeostasis. We therefore investigated whether hepatic HMGCS2 is required for liver lipid homeostasis during conditions characterized by acute (18-h fast and 72-h ketogenic diet) and chronic (32-wk GAN diet) elevations in liver lipid availability.

During fasting, a decline in insulin promotes adipose tissue lipolysis and the delivery of fatty acids to the liver (30). To further evaluate the role of ketogenesis in regulating liver lipid homeostasis during fasting, WT and liver HMGCS2 KO mice were fed a GAN diet for 6 wk and then fasted for 18 h. Body weight was comparable between genotypes (Fig. 2A). Liver AcAc, βOHB, and TKBs were decreased in liver HMGCS2 KO mice compared with WT mice (Fig. 2B). Moreover, mice lacking hepatic HMGCS2 exhibited elevated liver TAGs relative to WT littermates (Fig. 2C). These data indicate that ketogenic insufficiency promotes higher TAGs during extended fasting.

Figure 2.

Liver triacylglycerides (TAGs) are higher in liver 3-hydroxymethylglutaryl-CoA synthase 2 knockout (HMGCS2 KO) mice compared with wild-type (WT) mice under conditions characterized by acutely elevated lipid availability. A: body weights of 12-wk-old, male, liver-specific HMGCS2 knockout (KO) mice and wild-type (WT) littermates fed a GAN diet for 6 wk (n = 7 or 8 per genotype). B: liver acetoacetate (AcAc; nmol·g⁻¹), β-hydroxybutyrate (βOHB; nmol·g⁻¹), and total ketone bodies (TKBs; nmol·g⁻¹) in 18-h fasted WT and liver-specific HMGCS2 KO mice following 6 wk on a GAN diet (n = 6 or 7 per genotype). C: liver triacylglycerides (TAGs; mg·g⁻¹) in 18-h fasted WT and liver HMGCS2 KO mice following 6 wk on a GAN diet (n = 7 or 8 per genotype). D: body weights of 12-wk-old, male, liver HMGCS2 KO mice and WT littermates fed a ketogenic diet for 72 h (n = 5 per genotype). E: liver AcAc (nmol·g⁻¹), βOHB (nmol·g⁻¹), and TKBs (nmol·g⁻¹) in 6.5-h fasted WT and liver HMGCS2 KO mice following 72 h on a ketogenic diet (n = 5 per genotype). F: liver TAGs (mg·g⁻¹) in 6.5-h fasted WT and liver HMGCS2 KO mice following 72 h on a ketogenic diet (n = 5 per genotype). G: body weights of 38-wk-old, male, liver HMGCS2 KO mice and WT littermates fed a GAN diet for 32 wk (n = 5 per genotype). H: liver AcAc (nmol·g⁻¹), βOHB (nmol·g⁻¹), and TKBs (nmol·g⁻¹) in 6.5-h fasted WT and liver HMGCS2 KO mice following 32 wk on a GAN diet (n = 5 per genotype). I: liver TAGs (mg·g⁻¹) in 6.5-h fasted WT and liver HMGCS2 KO mice following 32 wk on a GAN diet (n = 5 per genotype). Data are presented as means ± SE. Statistical differences were determined by Student’s t tests and accepted as significant if P < 0.05 for data in A–C and E–I. **P < 0.01, ***P < 0.001, and ****P < 0.0001. Statistical difference (P < 0.05) for data in D was determined by a two-way repeated measures ANOVA followed by Sidak’s post hoc tests. Significant main and interaction effects are presented in D. #P < 0.05, ##P < 0.01, and ###P < 0.001 compared with day 0 within genotypes. ND, not determined because concentrations are below quantifiable limits. Figure created with a licensed version of BioRender.com.

To further test the necessity of hepatic HMGCS2 during acute elevations in lipid availability, we fed WT and liver HMGCS2 KO mice a ketogenic diet (90.5% kcal from fat) for 72 h to acutely increase liver lipid availability from an exogenous source. There was a main effect of diet (P < 0.0001) indicating that body weight decreased in response to the ketogenic diet (Fig. 2D). An interaction between diet and genotype was also observed (P < 0.0001; Fig. 2D). More specifically, liver HMGCS2 KO mice showed a progressive decline in body weight over the 72-h ketogenic diet feeding, whereas the decrease observed in WT mice plateaued by the 72-h time point (Fig. 2D). Notably, there were no statistical differences in body weight between genotypes at any individual time point. As expected, liver AcAc, βOHB, and TKBs were lower in KO mice compared with WT controls (Fig. 2E). Liver TAGs were ∼7.75-fold higher in liver HMGCS2 KO mice relative to WT mice (Fig. 2F).

To investigate the impact of hepatic HMGCS2 deficiency under conditions of long-term elevations in liver lipid availability from exogenous sources, we fed mice a GAN diet for 32 wk. Body weight was similar between liver HMGCS2 KO mice and WT controls (Fig. 2G). Loss of hepatic HMGCS2 reduced liver AcAc, βOHB, and TKBs (Fig. 2H). Interestingly, in contrast to the 18-h fast and 72-h ketogenic diet cohorts, liver TAGs were similar between genotypes following the long-term GAN diet feeding (Fig. 2I).

Together, these data are consistent with our findings from the acute exercise experiments (Fig. 1). Specifically, under conditions characterized by short-term or acute elevations in liver lipid availability, hepatic HMGCS2 deletion increases liver TAGs relative to WT controls.

Loss of Hepatic BDH1 Does Not Promote Higher Liver TAGs After Acute Exercise

Twelve-week-old, chow-fed, male mice lacking liver BDH1 were studied to define the importance of disposing of fatty acids to AcAc on liver lipid homeostasis during exercise (Fig. 3A). Body weight, fat mass, and lean mass were similar between genotypes (Fig. 3B). Loss of hepatic BDH1 did not impact maximal running speed (Fig. 3C). Following a 60-min treadmill run, liver AcAc was higher, βOHB was decreased, and TKBs were unchanged in liver BDH1 KO compared with WT mice (Fig. 3D). Notably, liver TAGs were similar between liver BDH1 KO mice and WT littermates (Fig. 3E). To further support our findings from the acute exercise experiments, WT and liver BDH1 KO mice were fasted 18 h. Loss of hepatic BDH1 increased AcAc, decreased βOHB, and lowered TKBs in 18-h fasted mice (Fig. 3F). Liver TAGs were comparable between WT and liver BDH1 KO mice (Fig. 3G). These results suggest that the ability to divert acetyl-CoA and/or dispose of fatty acids toward AcAc is important for maintaining liver lipid homeostasis during acute exercise.

Figure 3.

Liver triacylglycerides (TAGs) are not higher in liver βOHB dehydrogenase 1 (BDH1) knockout (KO) mice compared to wild-type (WT) mice under conditions characterized by acutely elevated lipid availability. A: a schematic representation of the ketogenic pathway with select metabolites and enzymes and a representative immunoblot of liver BDH1 in liver-specific BDH1 knockout (KO) mice and wild-type (WT) littermates. B: body weight (g), fat mass (g), and lean mass (g) in 12-wk-old, chow-fed (∼18% kcal from fat), male WT and liver BDH1 KO mice (n = 6 or 7 per genotype). C: maximal running speed in (m·min⁻¹) in WT and liver BDH1 KO mice (n = 6 or 7 per genotype). D: liver acetoacetate (AcAc; nmol·g⁻¹), β-hydroxybutyrate (βOHB; nmol·g⁻¹), and total ketone bodies (TKBs; nmol·g⁻¹) in WT and liver-specific BDH1 KO mice following a 60-min treadmill run at 45% of maximal running speed (n = 6 or 7 per genotype). E: liver TAGs (mg·g⁻¹) in WT and liver BDH1 KO mice following a 60-min treadmill run at 45% of maximal running speed (n = 6 or 7 per genotype). F: liver AcAc (nmol·g⁻¹), βOHB (nmol·g⁻¹), and TKBs (nmol·g⁻¹) in 18-h fasted WT and liver BDH1 KO mice (n = 7 or 8 per genotype). G: liver TAGs (mg·g⁻¹) in 18-h fasted WT and liver BDH1 KO mice (n = 7 or 8 per genotype). Data are presented as means ± SE. Statistical differences were determined by Student’s t tests and accepted as significant if P < 0.05 for all data in B–G. ***P < 0.001 and ****P < 0.0001. Figure created with a licensed version of BioRender.com.

TCA Cycle Flux Is Higher in Hepatic HMGCS2 KO Mice at Rest and During Exercise

In addition to ketogenesis, acetyl-CoA and fatty acids are also oxidized by the TCA cycle to support the energetic demands of gluconeogenesis. As such, TCA cycle fluxes were quantified in male mice at rest and during a 30-min exercise bout (Fig. 4A) to further test the impact of ketogenic insufficiency on lipid homeostasis. Plasma insulin decreased in response to exercise, however, there were no differences between WT and liver HMGCS2 KO mice (Fig. 4B). Plasma glucagon increased comparably in both genotypes during the 30-min treadmill run (Fig. 4C). Analysis of plasma NEFAs revealed a significant genotype effect, wherein levels were higher in liver HMGCS2 KO mice compared with WT littermates (Fig. 4D). Arterial glucose was similar between genotypes, though a significant main effect of exercise was observed, reflecting an increase in circulating glucose levels during exercise in both genotypes (Fig. 4E).

Figure 4.

Glucose and oxidative fluxes at rest and during acute exercise in mice lacking liver 3-hydroxymethylglutaryl-CoA synthase 2 knockout (HMGCS2). A: 18-wk-old, male, liver-specific HMGCS2 knockout mice and wild-type (WT) littermates fed a Gubra Amylin-NASH (GAN) diet for 12 wk underwent ²H- and ¹³C-isotope infusions to quantify glucose and oxidative fluxes at rest and during an acute exercise bout. B: plasma insulin (pmol·L⁻¹) before and at the conclusion of a 30-min treadmill run (n = 6 or 7 per genotype). C: plasma glucagon (pmol·L⁻¹) before and at the conclusion of a 30-min treadmill run (n = 7 per genotype). D: plasma nonesterified fatty acids (NEFAs; mmol·L⁻¹) before and at the conclusion of a 30-min treadmill run (n = 7 per genotype). E: blood glucose (mmol·L⁻¹) before and during a 30-minute treadmill run (n = 6 or 7 per genotype). Model-estimated, absolute nutrient fluxes (µmol·kg⁻¹·min⁻¹) in WT and liver-specific HMGCS2 KO mice (n = 6 or 7 per genotype) before and during a 30-min of treadmill run for endogenous glucose production (V_EGP) (F), glycogenolysis (V_PYGL) (G), total gluconeogenesis (GNG; V_Aldo) (H), GNG from glycerol (V_GK) (I), GNG from phosphoenolpyruvate (PEP; V_Enol) (J), tricarboxylic acid cycle cataplerosis (V_PCK) (K), anaplerosis from pyruvate (V_PC) (L), anaplerosis from propionyl-CoA (V_PCC) (M), flux from oxaloacetate and acetyl-CoA to citrate (V_CS) (N), and flux from succinyl-CoA to oxaloacetate (V_SDH) (O). Data are presented as means ± SE. Statistical differences (P < 0.05) were determined by a two-way repeated measures ANOVA, and significant main effects are specified in B–O. Figure created with a licensed version of BioRender.com.

Endogenous glucose production (V_EGP), glycogenolysis (V_PYGL), and total gluconeogenesis (V_Aldo) were similar at rest and increased during exercise in both genotypes (Fig. 4, F–H). The rise in total gluconeogenesis during exercise in WT and liver HMGCS2 KO mice was due to elevated rates of gluconeogenesis from glycerol (V_GK), gluconeogenesis from phosphoenolpyruvate (V_Enol), and cataplerosis from the TCA cycle (Fig. 4, I–K). Anaplerosis was also higher in response to exercise in both WT and liver HMGCS2 KO mice (Fig. 4, L and M). This included flux of pyruvate to oxaloacetate (V_PC; Fig. 4L) and flux from propionyl-CoA to succinyl-CoA (V_PCC; Fig. 4M). Exercise increased TCA cycle fluxes (V_CS and V_SDH) in both liver HMGCS2 KO mice and WT littermates (Fig. 4, N and O). Importantly, there was a main effect of genotype observed for TCA cycle fluxes, indicating that loss of HMGCS2 increases these oxidative fluxes (Fig. 4, N and O).

Loss of Hepatic HMGCS2 Increases Voluntary Wheel Running

Next, we performed indirect calorimetry to assess the extent to which dysregulated liver oxidative metabolism in HMGCS2 KO mice impacted systemic energy metabolism. Body weight was not different between genotypes (Fig. 5A). RER was characterized by a significant interaction between genotype and light/dark cycle (P < 0.036; Fig. 5, B and C). Specifically, only HMGCS2 KO mice exhibited an increase in RER during the dark cycle relative to the light cycle (Fig. 5, B and C). Energy expenditure (Fig. 5, D and E) and food intake (Fig. 5, F and G) were similar between genotypes across the light, dark, and 24-h cycles. However, a significant main effect of the light/dark cycle was observed as energy expenditure and food intake increased during the dark cycle (Fig. 5, D and F). Analysis of ambulatory activity revealed a main effect of the light/dark cycle, with increased movement during the dark phase (Fig. 5H). In addition, a genotype effect was observed, wherein HMGCS2 KO mice exhibited greater ambulatory movement off the voluntary running wheel compared with WT littermates (Fig. 5, H and I). Although voluntary running wheel distance was comparable between genotypes during the light cycle (Fig. 5J), HMGCS2 KO mice displayed a higher voluntary running wheel distance during the dark cycle and 24-h cycles (Fig. 5, J and K). This increased wheel running distance in mice lacking hepatic HMGCS2 was associated with a greater amount of time spent running at wheel speeds above 25 m·min⁻¹ in the dark and 24-h cycles (Fig. 5, L–N). To further test whether the differences in wheel running characteristics between genotypes were voluntary or physiological in nature, mice underwent a forced treadmill run to exhaustion. Time to exhaustion running at 22.5 m·min⁻¹ was similar between WT and liver HMGCS2 KO mice (Fig. 5O). Collectively, our results suggest that the increased voluntary wheel running distance and speeds in liver HMGCS2 KO mice are not due to heightened exercise endurance.

Figure 5.

Voluntary wheel running (VWR) is increased in mice lacking liver 3-hydroxymethylglutaryl-CoA synthase 2 knockout (HMGCS2). Twelve-week-old, male, liver-specific HMGCS2 knockout mice and wild-type (WT) littermates were fed a Gubra Amylin-NASH (GAN) diet for 6 wk before indirect calorimetry and voluntary wheel running experiments. A: body weight (g; n = 7 or 8 per genotype). B: respiratory exchange ratio (RER; unitless; n = 7 or 8 per genotype) for 12-h light and dark cycles. C: RER (unitless; n = 7 or 8 per genotype) for 24 h. D: energy expenditure (kcal·h⁻¹; n = 7 or 8 per genotype) for 12-h light and dark cycles. E: energy expenditure (kcal·h⁻¹; n = 7 or 8 per genotype) for 24 h. F: food intake (g; n = 6–8 per genotype) for 12-h light and dark cycles. G: food intake (g; n = 6–8 per genotype) for 24 h. H: all meters by mice while not on the voluntary running wheel (n = 7 or 8 per genotype) during 12-h light and dark cycles. I: all meters by mice while not on the voluntary running wheel (n = 7 or 8 per genotype) for 24 h. J: running distance (m) while on the voluntary running wheel (n = 7 or 8 per genotype) during 12-h light and dark cycles. K: running distance (m) while on the voluntary running wheel (n = 7 or 8 per genotype) for 24 h. L: time spent (s) at different voluntary wheel running speeds during the light cycle (n = 7 or 8 per genotype). M: time spent (s) at different voluntary wheel running speeds during the dark cycle (n = 7 or 8 per genotype). N: time spent (seconds) at different voluntary wheel running speeds during 24-h period (n = 7 or 8 per genotype). O: time to exhaustion (min) while running on a treadmill at 22.5 m·min⁻¹ (n = 7–11 per genotype). Data are presented as means ± SE. Statistical differences were determined by Student’s t tests and accepted as significant if P < 0.05 for data in A, C, E, G, I, K, and O. Statistical differences (P < 0.05) for data in B, D, F, H, J, and L–N were determined by a two-way ANOVA followed by Sidak’s post hoc tests. Significant main and interaction effects are presented within each panel. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Figure created with a licensed version of BioRender.com.

Exercise Training Limits Gains in Body Weight and Adiposity in Both WT and KO Mice

Given that ketogenic insufficiency dysregulated liver lipid homeostasis during acute exercise, the impact of exercise training on systemic and tissue-specific lipid metabolism was tested in male WT and liver HMGCS2 KO mice. The exercise training protocol consisted of 60 min of treadmill running, 5 days per week for 6 wk and was initiated when mice had been receiving a GAN diet for 6 wk (Fig. 6A). The treadmill running speed was 45% of the mouse’s initial maximal running speed during the first 2 wk of the training protocol and increased by 5% of the maximal running speed every 2 wk. This initial speed and duration were selected because they stimulate an increase in circulating βOHB in C57BL/6J mice (Fig. 6B). Body weight, adiposity, and lean mass were comparable between WT and KO mice before the exercise training protocol at 12 wk of age (Fig. 6C). Reassessment of these metrics following 6 wk of exercise training showed body weight, fat mass, and lean mass were lower in trained mice compared with their untrained counterparts (Fig. 6D). Six weeks of exercise training prevented a gain in body weight and adiposity in both genotypes (Fig. 6, E and F).

Figure 6.

Anthropometric adaptations to exercise training in mice lacking liver 3-hydroxymethylglutaryl-CoA synthase 2 knockout (HMGCS2). A: a schematic representation of the exercise training protocol. Male, liver-specific HMGCS2 knockout (KO) mice and wild-type (WT) littermates were fed a Gubra Amylin-NASH (GAN) diet starting at 6 wk of age. At 12 wk of age, mice were randomly assigned to sedentary (untrained) or exercise training (trained) groups. B: arterial blood β-hydroxybutyrate (βOHB; mmol·L⁻¹) in C57BL/6J mice at rest and during 60-min treadmill run at 45% of maximal running speed (n = 5 mice per time point). C: body weight (g), fat mass (g), and lean mass (g) before the exercise training protocol (n = 8 or 9 per genotype). D: body weight (g), fat mass (g), and lean mass (g) following the exercise training protocol (n = 8 or 9 per genotype). E: time course of body weight (g) during the 6-wk exercise training protocol (n = 8 or 9 per genotype). F: change in body weight (g) and fat mass (g) during the 6-wk exercise training protocol (n = 8 or 9 per genotype). Data are presented as means ± SE. Statistical differences for data in B were determined by a one-way, repeated measures ANOVA and accepted as significant if P < 0.05. Statistical differences (P < 0.05) for data in C, D, and F were determined by a two-way ANOVA. Statistical differences (P < 0.05) for data in E were determined by a three-way ANOVA. Significant main and/or interaction effects are presented within each panel. Figure created with a licensed version of BioRender.com.

Mitigation of Diet-Induced Fatty Liver by Exercise Training Is Not Compromised in Mice Lacking Hepatic HMGCS2

Experiments also tested the impact of exercise training on liver metabolism in male WT and liver HMGCS2 KO mice fed a GAN diet (Fig. 7). Liver HMGCS2 protein was comparable in untrained and trained WT mice (Fig. 7A). Liver BDH1 protein was similar in untrained WT and KO mice (Fig. 7A). Training decreased liver BDH1 in both genotypes (Fig. 7A). As expected, liver AcAc, βOHB, and TKBs were lower in untrained and trained KO mice compared with their WT controls (Fig. 7B). Trained WT mice had decreased liver AcAc, βOHB, and TKBs relative to untrained WT mice (Fig. 7B). Liver TAGs were similar in untrained WT and KO mice (Fig. 7C). Training decreased liver TAGs in both genotypes (Fig. 7C). In addition to total concentration, we assessed the fatty acyl chains comprising liver TAGs. Training decreased all fatty acids quantified in both WT and liver HMGCS2 KO mice (Supplemental Fig. S2). Liver DAGs were similar in untrained WT and KO mice (Fig. 7D) and were reduced by training in both genotypes (Fig. 7D). Liver cholesterol esters (Fig. 7E), phosphatidylethanolamine (Fig. 7F), phosphatidylcholine (Fig. 7G), and phosphatidylglycerol (Fig. 7H) were comparable between all groups. The key finding related to liver lipid metabolism presented here is that hepatic HMGCS2 is not necessary for exercise training to lower liver TAGs.

Figure 7.

Liver adaptations to exercise training in mice lacking liver 3-hydroxymethylglutaryl-CoA synthase 2 knockout (HMGCS2). Livers were analyzed in ∼18-wk-old, male, untrained and exercise-trained mice with a liver-specific HMGCS2 knockout (KO) and wild-type (WT) littermates fed a Gubra Amylin-NASH (GAN) diet starting at 6 wk of age. A: liver HMGCS2 and βOHB dehydrogenase 1 (BDH1) as determined by immunoblotting and representative immunoblots (arbitrary units; n = 7 or 8 per genotype). B: liver acetoacetate (AcAc; nmol·g⁻¹), β-hydroxybutyrate (βOHB; nmol·g⁻¹), and total ketone bodies (TKBs; nmol·g⁻¹; n = 7–9). C: liver triacylglycerides (TAGs; nmol·mg protein⁻¹; n = 7–9 per genotype). D: liver diacylglycerides (DAGs; nmol·mg protein⁻¹; n = 8 per genotype). E: liver cholesterol esters (nmol·mg protein⁻¹; n = 8 or 9 per genotype). F: liver phosphatidylethanolamine (PE; nmol·mg protein⁻¹; n = 8 or 9 per genotype). G: liver phosphatidylcholine (PC; nmol·mg protein⁻¹; n = 7–9 per genotype). H: liver phosphatidylglycerol (PG; nmol·mg protein⁻¹; n = 8 or 9 per genotype). I: liver glycogen (nmol·mg protein⁻¹; n = 7–9 per genotype). J: liver phospho-glycogen synthase (pGS)-to-total GS ratio, phospho-glycogen phosphorylase (pPYGL)-to-total PYGL ratio, cytosolic phosphoenolpyruvate carboxykinase (PCK1), and glucose-6-phosphatase catalytic subunit (G6PC) as determined by immunoblotting and representative immunoblots (arbitrary units; n = 7 or 8 per genotype). Data are presented as means ± SE. Statistical differences (P < 0.05) for liver HMGCS2 between untrained WT and trained WT mice were assessed by Student’s t test. Statistical differences (P < 0.05) for data in all other panels were determined by a two-way ANOVA followed by Sidak’s post hoc tests. Significant main and/or interaction effects are presented within each panel. **P < 0.01, ***P < 0.001, and ****P < 0.0001. Figure created with a licensed version of BioRender.com.

In addition to lipid characteristics, indices of liver glycogen metabolism were evaluated in untrained and trained mice. Liver glycogen was lower in untrained and trained KO mice compared with their WT counterparts (Fig. 7I). Training elevated liver glycogen in both WT and liver HMGCS2 KO mice (Fig. 7I). Molecular regulators of glycogen metabolism were assessed. The liver phospho-glycogen phosphorylase (pPYGL)-to-PYGL ratio and phospho-glycogen synthase (pGS)-to-GS ratio did not differ between genotypes or in response to training (Fig. 7J). Regulators of glucose-6-phosphate, a glycogen precursor/product, were determined. Phosphoenolpyruvate carboxykinase 1 (PCK1) protein was unaffected (Fig. 7J). A main effect of training effect showed higher liver glucose-6-phosphatase (G6PC) protein (Fig. 7J). These results indicate ketogenic insufficiency lowers liver glycogen, but does not impede the training-induced increase in liver glycogen deposition.

Systemic and Skeletal Muscle Adaptations in Mice Lacking Liver HMGCS2

Given that liver-sourced ketone bodies can support the energetic demands of extrahepatic tissues during exercise, adaptations in exercise performance and metabolism were assessed in GAN diet-fed, male WT and liver HMGCS2 KO mice following 6 wk of exercise training. Analysis of maximal running speed revealed a three-way interaction between time, training group, and genotype (P < 0.043). This interaction was driven specifically by a reduction in maximal speed in the untrained WT mice at the end of the 6-wk period compared with their pretraining time point (Fig. 8A). Notably, no such decline was observed in the liver HMGCS2 KO mice, which maintained consistent maximal running speeds regardless of training status or time point (Fig. 8A). A workload index was determined to account for differences in body weight during the maximal running speed tests. Similar to maximal running speed, a significant three-way interaction between time, training, and genotype (P < 0.032) was observed for the workload index (Fig. 8B). The interaction suggests that the decline in maximal running speed observed in the untrained WT group over the 6-wk period was not due to changes in body weight.

Figure 8.

Systemic and skeletal muscle adaptations to exercise training in mice lacking liver 3-hydroxymethylglutaryl-CoA synthase 2 (HMGCS2). A: maximal running speed (m·min⁻¹) during an exercise stress test in Gubra Amylin-NASH (GAN) diet-fed, male mice with a liver-specific knockout (KO) of HMGCS2 and wild-type (WT) littermates before and following 6 wk of exercise training (n = 8 or 9 per group). B: workload index (kg·m) during an exercise stress test in GAN-fed mice before and following 6 wk of exercise training (n = 8 or 9 per group). Skeletal muscle was analyzed in 18-wk-old untrained and exercise-trained mice. C: gastrocnemius citrate synthase (CS) activity (µmol·g⁻¹·min⁻¹; n = 8 or 9 per genotype). D: gastrocnemius glycogen (mg·g⁻¹; n = 8 or 9 per group). E: gastrocnemius triacylglycerides (TAGs; mg·g⁻¹; n = 7 or 8 per group). F: superficial vastus lateralis (SVL) TAGs (mg·g⁻¹; n = 8 per group). Data are presented as means ± SE. Statistical differences (P < 0.05) for data in A and B were determined by a three-way ANOVA followed by Sidak’s post hoc tests. Statistical differences (P < 0.05) for data in C–F were determined by a two-way ANOVA. Significant main and/or interaction effects are presented within each panel. *P < 0.05. Figure created with a licensed version of BioRender.com.

Citrate synthase (CS) activity in the gastrocnemius was similar between untrained WT and liver HMGCS2 KO mice (Fig. 8C). Training elevated gastrocnemius CS activity in both genotypes (Fig. 8C). Gastrocnemius glycogen was comparable between all groups (Fig. 8D). Gastrocnemius TAGs were not different between untrained WT and liver HMGCS2 KO mice (Fig. 8E); however, trained WT and KO mice had lower gastrocnemius TAGs relative to their untrained counterparts (Fig. 8E). Superficial vastus lateralis (SVL) TAGs were similar between all groups (Fig. 8F).

Together, our data suggest that GAN diet feeding may promote a decline in maximal running speed. However, this decline is prevented by either exercise training or hepatic ketogenic insufficiency. Furthermore, because skeletal muscle characteristics and adaptations to training were similar in both genotypes, the preservation of maximal running speed in liver HMGCS2 KO mice may be independent of changes in skeletal muscle.

DISCUSSION

Acute aerobic exercise stimulates hepatic lipid oxidation pathways that transform the potential energy in fatty acids to alternative forms of energy currency that are then released from the liver to support muscular work. This includes increasing the production of AcAc and βOHB from fatty acids via ketogenesis (14). A gap in knowledge has been whether the stimulation of hepatic ketogenesis by acute exercise is critical for exercise training to mitigate liver steatosis. In this study, we tested the hypothesis that hepatic ketogenesis is necessary for exercise training to lower liver lipids. The key findings of our work in male mice are 1) hepatic HMGCS2-mediated ketogenesis is required to maintain liver lipid homeostasis during acute, nonexhaustive exercise, as its loss results in higher liver TAG levels relative to WT littermates after a 60-min treadmill run and 2) hepatic HMGCS2-dependent ketogenesis is not necessary for exercise training to mitigate diet-induced fatty liver.

Role of Ketogenic Insufficiency in Liver Lipid Homeostasis During Acute Exercise

Exercise stimulates ketogenesis through 1) the mobilization of fatty acids from adipose tissue and delivery to the liver, 2) the hepatic extraction of fatty acids from the circulation, and 3) the intrahepatic conversion of fatty acids to ketone bodies (6). Studies in both humans and experimental models support that the fall in insulin during exercise promotes all three of these processes (6, 31–33). In addition, the exercise-mediated rise in glucagon enhances the conversion of fatty acids to both AcAc and βOHB within the liver (6, 16). Consistent with this prior work, our data show that exercise lowers insulin, increases glucagon, and elevates liver ketone bodies in WT mice. Liver HMGCS2 KO mice exhibited a similar hormonal response to exercise compared with WT mice. However, impeding the conversion of fatty acids to ketone bodies via deletion of hepatic HMGCS2 led to a blunted rise in liver ketone bodies in response to exercise. This was associated with increased liver TAGs in HMGCS2 KO mice compared with WT mice at the end of the acute exercise bout. To further test the role of ketogenic insufficiency in promoting higher liver TAGs compared with controls during exercise, we studied hepatic BDH1 KO mice to permit the generation of AcAc, but not βOHB, from fatty acids. In contrast to hepatic HMGCS2 KO mice, loss of BDH1 in the livers of mice did not result in higher liver TAGs compared with WT mice after an exercise bout. Our results indicate that HMGCS2 is a key node through which ketogenesis supports liver lipid homeostasis during nonexhaustive exercise.

Although the deletion of hepatic HMGCS2 inhibits the production of ketone bodies from fatty acids within the liver, ketogenic insufficiency is unlikely to be solely responsible for the higher liver fat in HMGCS2 KO mice in relation to WT mice after exercise. Here we show that, under sedentary conditions, ad libitum-fed and 6.5-h fasted mice lacking hepatic HMGCS2 had lower liver ketone bodies but similar liver TAG concentrations compared with WT controls. Prior work using adult mice with a whole body and liver-specific HMGCS2 KO or knockdown has also reported that ketogenic insufficiency does not promote liver steatosis in fed and short-fasted (<8 h) states (17, 34–37). These results indicate that the inhibition of hepatic ketogenesis alone is not sufficient to promote liver steatosis. As previously mentioned, hepatic fatty acid delivery and ketone body production are positively correlated during exercise (38). The current study had mice undergo an 18-h fast and short-term ketogenic diet feeding because they lower circulating insulin and acutely increase liver lipid availability (39–43). Under these conditions, liver HMGCS2 KO mice had increased liver TAGs relative to WT mice. Our results are in agreement with independent studies demonstrating that long fasting durations (>8 h) and very high-fat diets (>90.5%) elevate liver TAGs in mice lacking HMGCS2 (17, 35). The interaction between ketogenic insufficiency and lipid availability on liver steatosis can also be gleaned from studies in neonatal mice. Before postnatal day 14, mice with a whole body knockdown or KO of HMGCS2 display fatty liver (37, 44). This may be due to the primary nutrient source for rodents during this period of life being breast milk, which is high in fat content (44). Notably, this liver steatosis phenotype in ketogenesis-insufficient neonatal mice is lost when transitioned to a high-carbohydrate, low-fat diet (37). It is currently unclear why the GAN diet used in our study did not promote fatty liver in sedentary HMGCS2 KO mice. Although the GAN diet increases lipid availability (40% kcal from fat), it also elevates circulating insulin levels in mice (45). Wasserman et al. (31) identified that preventing the fall in insulin during exercise attenuates hepatic extraction of fatty acids and stimulation of ketogenesis. We showed that liver-specific HMGCS2 KO mice had elevated circulating NEFAs. It may be that the higher insulin levels in mice fed a GAN diet limit hepatic extraction of lipids in sedentary, liver HMGCS2 KO mice. Together, our data and the work of others suggest that loss of hepatic HMGCS2 promotes higher liver lipids compared with WT mice after an acute exercise because the inability to convert fatty acids to ketone bodies in the liver is accompanied by a concurrent increase in lipid delivery to the liver and/or hepatic extraction of lipids from the circulation.

Conditional compensation by other metabolic pathways could contribute to our observations that HMGCS2 KO mice have higher liver TAGs compared with WT mice after exercise, but not under ad libitum or short-fasted conditions. Previous studies have determined that liver mitochondrial acetyl-CoA is increased by ketogenic insufficiency (46). In addition to ketogenesis, complete oxidation of β-oxidation-derived acetyl-CoA in the TCA cycle is a prominent means of lipid catabolism in the liver (47). Our data show that TCA cycle flux is higher in liver HMGCS2 KO mice relative to WT littermates at rest and during exercise. Interestingly, we recently observed that liver HMGCS2 KO mice have elevated liver TAGs despite increased TCA cycle flux when fed a ketogenic diet under sedentary conditions (17). These findings suggest that fatty acid oxidation via the TCA cycle increases to compensate for ketogenic insufficiency. However, the TCA cycle may not be sufficient to prevent liver lipid accretion when confronted with conditions that place further demands on TCA cycle flux (i.e., an exercise challenge) and/or significantly increase lipid availability.

Role of Hepatic HMGCS2 in Lipid Adaptations to Exercise Training

Regular exercise reduces liver steatosis in both humans and experimental models (2, 22, 48–55). Given that hepatic ketogenesis supports liver TAG homeostasis during acute exercise, we tested whether loss of hepatic HMGCS2 would attenuate the effectiveness of exercise training to mitigate diet-induced fatty liver. Liver HMGCS2 KO mice and WT littermates remained untrained or were trained via a 6-wk treadmill running protocol. Forced exercise training was performed to better facilitate the consistent implementation of key exercise stimuli, including duration, intensity, and frequency (56). This was important because our studies showed that liver HMGCS2 KO mice spent more time running at higher speeds and for greater distances when provided with voluntary running wheels. The factors promoting increased voluntary wheel running in liver HMGCS2 KO mice and why the higher voluntary exercise is not accompanied by elevated energy expenditure remain to be determined. Nevertheless, in response to forced exercise training, liver TAGs were lower in trained WT compared with the untrained WT mice. Unexpectedly, training was equally effective at lowering liver TAGs in liver HMGCS2 KO mice as it was in WT mice. This outcome indicates that hepatic ketogenesis is not required for training to mitigate diet-induced fatty liver.

The effectiveness of training to combat fatty liver in the presence of ketogenic insufficiency is linked to multiple metabolic mechanisms. As previously discussed, TCA cycle flux is increased in liver HMGCS2 KO mice. A persistent elevation in fatty acid oxidation via the increased TCA cycle flux may aid in lowering liver TAGs in trained KO mice. Similar to liver HMGCS2 KO mice, we previously showed that the inhibition of exercise-stimulated TCA cycle flux through deletion of hepatic PCK1 in mice led to increased liver TAGs relative to WT mice during acute exercise, but did not compromise the lipid-lowering actions of training (10). Although TCA cycle flux was impaired, the liver PCK1 KO mice exhibited the delayed suppression of ketogenesis and/or ketosis upon refeeding following an acute exercise bout (10). Thus, individual mitochondrial oxidative pathways may compensate for impairments in the others over the longer timeline of exercise training. Beyond metabolic compensation effects on fatty acid oxidation to CO₂, aerobic exercise training could promote chronic adaptations in hepatic pathways that lower liver TAGs. For instance, training routinely increases the molecular machinery and functional capacity of liver β-oxidation, TCA cycle flux, and oxidative phosphorylation (2, 3, 57–60). In addition, exercise training elicits markers of reduced hepatic de novo lipogenesis, which could lead to reduced liver lipid accretion (57–60).

Adaptations in extrahepatic metabolism may also contribute to the prevention of fatty liver in trained mice lacking hepatic HMGCS2. We observed that training blunted the gain in body weight and fat mass in both WT and liver HMGCS2 KO mice. Repeated bouts of exercise also enhanced markers of mitochondrial content and/or oxidation in skeletal muscle. Specifically, muscle citrate synthase activity was increased, and TAGs were decreased in response to training. Our results are consistent with a recent publication that exercise training increases skeletal muscle mitochondrial respiration similarly in rats with a hepatic knockdown of HMGCS2 and WT controls (61). Ultimately, lower adiposity and an elevation in skeletal muscle lipid oxidation in trained mice could reduce fatty acid delivery to the liver and, subsequently, limit liver lipid accretion. The decreased liver AcAc and βOHB in trained WT mice are agreeable with the hypothesis that liver lipid supply is diminished in response to training.

Importantly, limitations of the present study warrant consideration. First, this study was conducted exclusively using male mice. Prior work has reported that female mice exhibit a more robust ketosis and/or ketogenic response to various aerobic exercise intensities (62, 63). Given this, the loss of HMGCS2 may impact exercise efficacy for fatty liver treatment more significantly in females. Future studies are required to determine whether the reliance on hepatic HMGCS2-dependent ketogenesis during exercise is conserved across sexes.

A second limitation is that the mice in this study were housed at room temperature (∼22°C), which is below thermoneutrality for mice (∼30°C). Housing below thermoneutrality results in increased energy expenditure at rest and during physical activity (64, 65). In addition, diet-induced obesity and MASLD phenotypes are less pronounced and/or the pathological progression is slowed in mice housed at room temperature compared with thermoneutrality (66, 67). Given these factors, the necessity of hepatic ketogenesis for the ability of exercise to combat fatty liver may be critical under conditions of thermoneutrality and warrants future testing.

In conclusion, hepatic HMGCS2-dependent ketogenesis is important for liver lipid homeostasis during an acute, nonexhaustive exercise bout in male mice. In contrast, hepatic HMGCS2 was not required for training to treat diet-induced fatty liver. Our results suggest that the efficacy of training to lower liver fat in the absence of HMGCS2 is linked to compensatory fat oxidation via the TCA cycle and extrahepatic adaptations that limit liver lipid availability. Furthermore, these studies highlight the multifactorial and integrated mechanisms through which training reduces fatty liver.

DATA AVAILABILITY

Data will be made available upon request.

GRANTS

This research was supported by the NIH Grants DK136772 (to C.C.H.), DK091538 (to P.A.C.), and AG069781 (to P.A.C.). The Functional Lipidomics Core at the Barshop Institute is partially supported by NIH Grants P30 AG013319 (to X.H.) and P30 AG044271 (to X.H.). A.F.O. was supported by the National Institutes of Health Ruth L. Kirschstein National Research Service Award T32 DK007293. E.D.Q. was supported by the National Institutes of Health Ruth L. Kirschstein National Research Service Award T32 HL166142.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

C.M.V.: Data curation, Formal analysis, Investigation, Writing - review & editing; A.F.O.: Data curation, Formal analysis, Investigation, Writing – review & editing; R.E.P.: Data curation, Formal analysis, Investigation, Writing – review & editing; J.L.H.: Investigation, Writing – review & editing; G.S.H.: Investigation, Writing – review & editing; H.W.: Investigation, Writing – review & editing; E.D.Q.: Investigation, Writing – review & editing; P.A.C.: Funding acquisition, Resources, Writing – review & editing; X.H.: Funding acquisition, Resources, Writing – review & editing; C.C.H.: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Project administration, Resources, Supervision, Visualization, Writing – original draft, Writing – review & editing.