Ethical approvals

This research complies with all relevant ethical regulations. Animal studies carried out in Leeds were regulated under the Animals (Scientific Procedures) Act 1986 with procedures carried out in accordance with U.K. Home Office protocols under a U.K. Home Office Project License (PP8169223) by a U.K. Home Office Personal License Holder. Animal studies carried out at the University of Sau Paulo were approved and followed the institutional guidelines for animal care and use for research (CEUA ICB USP #2144240425). Acute human exercise studies carried out at the University of Newcastle received approval from the NHS HRA North East Tyne & Wear South Research Ethics and Newcastle University Ethics Committees (16/NE/0192, ISRCTN63739203). Human exercise training studies carried out at Wageningen University received ethical approval from the Medical Ethical Committee of Wageningen University.

Human primary skeletal myocyte culture

Adult human skeletal myoblasts (Cell applications Inc. Cat no. 150-05a) were grown in human skeletal muscle cell growth medium (Cell applications Inc. Cat no.151-500) at 37 °C within a humidified atmosphere at 5% CO2. Subculture of human skeletal myoblasts occurred once 70% confluency was reached. Experiments were limited to the 5th passage. Human skeletal myoblasts were seeded at 9500 cells per cm2. Once confluent, myoblasts were cultured for 6–8 days in skeletal muscle differentiation media (Cell applications Inc. Cat no.151D-250) to induce myoblast differentiation to myotubes. Myoblasts were treated with either human skeletal muscle cell growth medium (Cell applications Inc.) containing BAIBA (10 μM; Sigma Aldrich, 217794) immediately after plating to assess proliferation or human skeletal muscle cell differentiation medium (Cell applications Inc.) containing BAIBA (10 µM) once confluent to assess effects of BAIBA on differentiation or human skeletal muscle cell differentiation medium (Cell applications Inc.) containing L-BAIBA (10 µM) or D-BAIBA (10 µM) (Adipogen). Media was changed every day during proliferation and every 2 days during differentiation.

The effect of BAIBA on proliferation was assessed with a non-destructive method to evaluate confluence71. Photographic images of HskMCs within wells were captured by photographing five areas under a standard inverted phase contrast light microscope using a digital camera with a camera lens adaptor. Images were analysed for confluence using ImageJ (1.54 f) freeware resulting in a measure of confluence known as an Area Fraction (AF). Five photographs were taken of the cells, four from the corner and one from the centre of the well. The photographs were taken consistently in the same position of the well at 24 h, 48 h and 72 h post seeding/differentiation.

Informed consent was obtained from donors. The cells were approved and complied with ethics according to:

Collection, generation, research purpose, and sale: Cell Applications, Inc. 5820 Oberlin Dr. Suite 101, San Diego, CA 92121.

Use in compliance with Human Tissue Act (UK) by Leeds Institute of Cardiovascular and Metabolic Medicine, University of Leeds, Leeds, LS2 9JT UK in 2022.

Fusion index assay

Primary HSkMCs (Cell Applications, S150a-05a) were seeded and differentiated on collagen-coated coverslips. Myotubes were treated with either human skeletal muscle cell differentiation medium (Cell applications Inc.) containing BAIBA (10 μM; 217794, Sigma-Aldrich) or L-BAIBA (10 μM; Adipogen). Fusion was assessed at days 3 and 6 after the induction of differentiation. The cells were washed with PBS and fixed with 4% PFA/PBS for 10 minutes, followed by 7 minutes permeabilization in 0.3% TRITON, 20 minutes blocking with BSA (2%) and were incubated with primary antibodies overnight at 4 °C (monoclonal anti-myogenin antibody, clone F5D, Invitrogen catalogue no. 14-5643-82; 1:150 dilution) as a marker of differentiation, anti-myosin heavy chain antibody monoclonal MF20 Myosin heavy chain, sarcomere (MHC) (clone MF-20, Developmental Studies Hybridoma Bank catalogue no. MF20, 1;100 dilution) to visualize myotubes. The cells were then incubated with a secondary antibody (polyclonal goat anti-Mouse IgG1 Alexa Fluor 555; Thermofisher catalogue number A-21127; dilution 1:150). Slides were coverslipped in mounting medium and stained 3 min with Hoechst (20 µg/mL, Sigma) to visualize nuclei.

To establish the differentiation potential and the fusion index of the cultures, at least 1,000 nuclei from MF20-positive cells were counted from several random fields. The percentage of differentiated cells was calculated as: (nuclei within MF20-stained myocytes/total number of nuclei) × 100; or the fusion index calculated as: (MF20-stained myocytes containing ≥2 nuclei/total number of nuclei) × 100. The number of myogenin-positive nuclei was manually counted. Images were taken using a x20 objective lens using a fluorescence microscope slide scanner (Axioscan Z1, ZEISS). Images were recorded using ZEN software (Version 3.40, ZEISS).

Seahorse bioanalyzer cellular mitochondrial respiration analysis

HSkMCs were plated at a density of 10,000 cells per well on a collagen-coated 96-well Seahorse cell culture plate and grown for 72 h in human skeletal muscle cell growth medium (Cell Applications Inc.) to assess proliferation. After 72 h, the medium was replaced with human skeletal muscle cell differentiation medium (Cell Applications Inc.) supplemented with BAIBA (10 µM; Sigma-Aldrich, cat no. 217794) to evaluate differentiation. As a negative control, the four corner wells of the plate were left cell-free and filled only with Seahorse medium (XF DMEM Medium, pH 7.4, cat no. 103575) containing 5 mM HEPES. Twelve h prior to the assay, Seahorse sensor cartridges were hydrated with Seahorse calibrant solution according to the manufacturer’s protocol and incubated in a 37 °C, CO2-free incubator. On the day of the assay, cells were washed and incubated with Seahorse medium. The sensor cartridge was then fitted onto the cell culture plate, which was subsequently placed in a 37 °C, CO2-free incubator for one hour to equilibrate. The mitochondrial stress test was performed using the Seahorse XFe96 Analyzer (Agilent) operated with Wave Controller (v. 2.6.1 Agilent). During the assay, the following inhibitors were sequentially injected according to the standard Mito Stress Test protocol: oligomycin (1 mM), FCCP (1 mM), and rotenone/antimycin A (0.5 mM). Cells were then fixed with 10% PFA and stained with Hoechst. Nuclei counts were measured using a Synergy H1 plate reader to assess fluorescence intensity and to normalize the seahorse data to cell number.

siRNA-mediated PPARδ and MRGPRD knockdown in vitro

FlexiTube GeneSolution GS5467 siRNA against PPARδ (SI05383420), FlexiTube GeneSolution GS116512 siRNA against MRGPRD (SI00164444), AllStars negative control siRNA, and HiPerFect Transfection Reagent were purchased from Qiagen. HSkMC (Cell applications Inc. Cat no. 150-05a) transfection was performed according to the manufacturer’s instructions (75 ng siRNA, 3 μL transfection reagent per well, 60 nmol/L final siRNA concentration) on days 2 and 4 of differentiation.

Gene expression analysis

Total RNA extraction from myocytes and skeletal muscle; cDNA conversion; and quantitative RT-PCR followed published protocols19,42. All data were normalized to RPLP0 or TBP rRNA (mouse skeletal muscle; mouse RPLP0 primer PPM03561B-200, mouse skeletal muscle; mouse Tbp primer PPM03560F-200, human primary myocytes, human RPLP0 primer PPH21138F-200, human primary myocytes, human TBP primer PH01091G-200, Qiagen) and quantitative measures were obtained using the ΔΔCT method. Data were analyzed using StepOne™ Software (version 2.1 Applied Biosystems). Primers are given in Supplementary Table 3.

Animal experimentation

Six-week-old male and female C57BL6/J mice (Charles River, UK) were weight-matched and assigned to groups for treatment. Mice were treated with 100 mg/kg/day BAIBA in their drinking water for 6 weeks and fed standard chow (Special Diet Services; 801151; 3.592 kcal/g). For obesity studies six-week-old male C57BL6/J mice (Charles River, UK) were weight-matched and assigned to groups for treatment. Mice were either placed on 60% fat-diet (Bio Serv F3282; kcal 5.49 kcal/g) or standard chow (Special Diet Services; 801151; 3.592 kcal/g) for 8 weeks. After 8 weeks of the study a randomly assigned half of the 60% fat-diet group and half the standard chow fed group then began treatment with 100 mg/kg/day BAIBA for a further 14 weeks. Feeding was ad libitum.

Abat knockdown studies: 6 week old male C57BL6/J mice (Charles River, UK) were weight-matched and assigned to groups for treatment. Mice were either assigned to the unexercised group or to the exercised group which underwent 6 weeks of treadmill training as described below. The exercised and unexercised groups were further subdivided into groups receiving intramuscular injections into both hindlimb gastrocnemius skeletal muscles of either scrambled control Antisense LNA GapmeR (GeneGlobe nb: LG00000002, Qiagen) or Antisense LNA GapmeR against Abat (Abat-207_1 GeneGlobe nb: LG00838129 Sequence: CACGATAGACAAATAGA, Qiagen) at a concentration of 5 mg/Kg/injection twice a week for six weeks. GapmeRs enter cells without the need for transfection reagents and are active in vivo without the need of formulation72,73,74. Mice were fed standard chow (Special Diet Services; 801151; 3.592 kcal/g) ad libitum.

Pparδ knockdown studies: 6 week old male C57BL6/J mice (Charles River, UK) were weight-matched and assigned to groups for treatment. Mice were either assigned to a control group or to a group receiving 100 mg/Kg/day BAIBA in drinking water. The BAIBA-treated and control groups were further subdivided into groups receiving intramuscular injections into the hindlimb gastrocnemius skeletal muscle of either Accell non-targeting control custom siRNA ((Sense: 5’ U.G.G.U.U.U.A.C.A.U.G.U.C.G.A.C.U.A.A.U.U 3’; Antisense: 5’ 5’-P.U.U.A.G.U.C.G.A.C.A.U.G.U.A.A.A.C.C.A.U.U 3’) based on D-001950-01, Dharmacon™ Custom siRNA, Horizon Discovery or Accell siRNA14 against Pparδ ((Sense: 5’ C.U.C.C.A.A.A.U.C.U.G.A.A.A.U.G.U.A.U.U.U 3’; Antisense: 5’ 5’-P.A.U.A.C.A.U.U.U.C.A.G.A.U.U.U.G.G.A.G.U.U 3’) based on A-042751-14, Dharmacon™ Custom siRNA, Horizon Discovery) twice a week for six weeks. siRNA was delivered at 100 µg per intramuscular injection per mouse. Mice were fed standard chow (Special Diet Services; 801151; 3.592 kcal/g) ad libitum.

D- and L-BAIBA enantiomer studies: Six-week-old male C57BL6/J mice (Charles River, UK) were weight-matched and assigned to groups for treatment. Mice were treated with either 100 mg/kg/day D-BAIBA or L-BAIBA (Adipogen) in their drinking water for 6 weeks and fed standard chow (Special Diet Services; 801151; 3.592 kcal/g) ad libitum.

All animals were housed in conventional cages at room temperature with humidity maintained at 40–60% and a 12-h light/dark photoperiod. All studies were regulated under the Animals (Scientific Procedures) Act 1986 and complied with national and local ethical regulations for animal research. All procedures were carried out in accordance with U.K. Home Office protocols under a U.K. Home Office Project License (PP8169223) by a U.K. Home Office Personal License Holder.

Muscle injury model: Three-month-old male C57BL/6 J mice were obtained from the animal facility at the Institute of Biomedical Sciences, University of São Paulo, Brazil following institutional ethical approval. Mice were housed with ad libitum access to standard chow (Special Diet Services; 801151; 3.592 kcal/g) and water, maintained on a 12 h light/dark cycle at 22 ± 2 °C. Animals were randomized into two experimental groups (n = 7) either receiving 100 mg/kg/day BAIBA (Sigma-Aldrich) in drinking water or standard drinking water. BAIBA administration was initiated 7 days prior to injury (pre-loading phase) and continued for 14 days following cardiotoxin injury. On the day of injury, mice were anesthetized with an intraperitoneal injection of ketamine (80 mg/kg) and xylazine (10 mg/kg). Cardiotoxin (CTX; Naja mossambica) was prepared in sterile phosphate-buffered saline (PBS) and injected into the tibialis anterior (TA) muscle. A total volume of 50 µL of 10 µM CTX was administered into the mid-belly of each TA muscle using a 29-gauge insulin syringe. In all animals, the left hindlimb was subjected to cardiotoxin injury, while the right hindlimb remained uninjured. This experimental design resulted in four experimental conditions: Control (standard drinking water, right hindlimb), Injured (standard drinking water, left hindlimb), BAIBA-control (BAIBA-supplemented drinking water, right hindlimb) and BAIBA-Injured (BAIBA-supplemented drinking water, left hindlimb). Fourteen days after injury, mice were euthanized by overdose of ketamine/xylazine (ketamine 300 mg/kg and xylazine 30 mg/kg, intraperitoneally).

Reporting of animal experiments follows the ARRIVE guidelines.

Freewheel running

Aerobic exercise training was carried out using a 6-week free wheel running protocol in male C57BL/6j mice (Charles River, UK) as previously published19. Mice were singly housed and given access to free wheel running (Tecniplast, Cat no.1284L0106) for 6 weeks. Controls were age-matched littermates unexercised and singly housed in matched cages.

Treadmill training

Seven week old mice were acclimated to the treadmill apparatus (5-lane treadmill, Harvard Apparatus, Panlab) for 10 min treadmill running per day at 10 cm/s at 0% grade in the 5 days leading up to the training programme. For 6 weeks, mice ran at a speed of 25 cm/s at 10% grade during a 45-minute session per day, 4 days per week. Each session was preceded and followed by a 5-minute warm-up and cool-down at a speed of 10 cm/s respectively. After the 6 week exercise training programme a maximal running capacity test was performed, as detailed below, to identify the individual Vmax, defined as the maximum speed each mouse was capable of running voluntarily. Treadmill training was performed in groups of five mice and was voluntary. No animals refused to run.

Maximal running test

Maximal running capacity test was performed on a motor treadmill (5-lane treadmill, Harvard Apparatus, Panlab) using a graded exercise protocol modified from previously published papers75,76. Briefly, mice underwent an adaptation period where the mice were placed on a stationary treadmill for 10 min per day for 5 consecutive days. On test day, mice were placed on the treadmill and exercise intensity started at 10 cm/sec and was increased for 5 cm/sec every 3 min at 0% grade until mice were able to run no longer. The time elapsed and distance run was recorded.

Indirect calorimetry and monitored wheel running

All experiments were performed according to previously published protocols42. CLAMS (Columbus Instruments) was used to monitor oxygen consumption, carbon dioxide production, food intake, and voluntary wheel running using Oxymax software (version 5.37.05, Columbus Instruments). The CLAMS was calibrated before each experiment. Animals were subjected to a 3-day acclimation period in a training cage with running wheel to habituate to the environment of the metabolic cages. Animals were maintained in normal bedding at 22 °C throughout the monitoring period. Ten-minute interval measurements for each animal were obtained for oxygen and carbon dioxide with ad libitum access to food and water on a controlled 12-h light/dark cycle. Ambulatory, locomotor and wheel running activity were constantly monitored. Cages contained one mass sensor to monitor food intake. Data was collected for a 48 h period after the 3-day acclimation. Data were analyzed using CaIR (version 1.3) (https://calrapp.org/)77.

Intraperitoneal glucose tolerance tests

Intraperitoneal glucose tolerance tests were performed as described42. Mice were fasted for 8 h with free access to water prior to baseline glucose measurements. Administration of glucose (Sigma Aldrich) was performed by intraperitoneal injection (glucose 1.5 mg/g of body weight; glucose solution 150 mg/ml). Blood was obtained from the tail vein immediately prior to glucose injection and then at 30, 60, 90, and 120 min post injection. Glucose levels were measured using a Bayer Contour Glucose Meter (Bayer Healthcare).

Magnetic resonance imaging

Anaesthesia was induced using 5% isoflurane in 100% oxygen and then maintained using 1.5–3% isoflurane at 2 l/min oxygen flow. Animals were positioned prone on a dedicated mouse cradle. Body temperature was maintained with a custom resistive blanket placed on the back of the animal. Cardiac and respiratory signals were continuously monitored (BIOPAC Systems, Inc., Goleta, USA). Mice were imaged on a 7 T preclinical MRI scanner with a 660 mT/m shielded gradient system using either a quadrature-driven transmit/receive volume coil with inner diameter of 72 mm (Bruker BioSpin MRI GmbH, Ettlingen, Germany) or 4-element volume-array (Neos BioTec, Pamplona, Spain) operated with ParaVision software (version 6.0.1; Bruker BioSpin MRI GmbH, Ettlingen, Germany). A 2D cardiac-triggered and respiratory-gated 3-point Dixon spoiled gradient-echo sequence was used: TR = 5.65 ms, TE = 2.42/2.75/3.09 ms, matrix = 256 × 128, field-of-view = 80 × 30 mm, number of slices = 28 in sagittal orientation, slice thickness = 1 mm, number of signal averages = 8/ 1 (volume coil / volume array), total scan time ~30 min. The data were analysed in MATLAB (2022; MathWorks, Natick, USA) using the hierarchical iterative decomposition of water and fat with echo asymmetry and least squares estimation (IDEAL) method78. The proton density fat fraction (PDFF, the amount of lipid signal over total signal) was used to segment adipose tissue depots. Visceral adipose depots were segmented separately using Osirix Lite v11.0.2 (Bernex, Switzerland) 2D threshold region-growing algorithm tool with segmentation parameters set to a lower threshold of 80% proton density fat fraction (PDFF) i.e., a minimum of 80% of the tissue volume consisted of lipid.

Tissue collection

Mice were killed by cervical dislocation. Blood was immediately removed via cardiac puncture and placed into EDTA-coated eppendorfs before being centrifuged at 2000 x g for 10 min. Serum was collected, frozen in liquid nitrogen, and stored at -80 °C until use. Soleus, gastrocnemius, extensor digitorum longus, and tibialis anterior muscle and inguinal and gonadal adipose tissue were removed, weighed and either used for high-resolution respirometry, contractile analysis, histology or flash-frozen in liquid nitrogen.

For the muscle injury model TA muscles were excised, trimmed, transversely sectioned, immersed in isopentane, cooled in liquid nitrogen, and stored at −80 °C.

Immunofluorescent muscle fiber-typing

Immediately after sacrifice, the soleus muscle was embedded in OCT embedding matrix (VWR, 361603E, UK) and flash frozen in liquid nitrogen-cooled isopentane at −150 °C. Samples were stored at −80 °C until further use. Samples were sliced into 8 µm cryosections with a CM1860 Cryostat, (Leica).

Immunohistochemical staining of Myosin Heavy Chain (MHC) isoforms was performed using mouse monoclonal antibodies BA-D5 (monoclonal anti-MHC-I/ β-MHC/Myh7 specific, IgG2b, clone/catalogue no. BA-D5, 1:100 dilution), SC-71 (monoclonal anti-MHC-IIa specific, IgG1, clone/catalogue no. SC-71 1:100 dilution), BF-F3 (monoclonal anti-MHC-IIb specific, IgM, clone/catalogue no. BF-F3, 1:100 dilution), and 6H1 (monoclonal anti-MHC-IIx specific, IgM, clone/catalogue no. 6H1, 1:100 dilution) obtained from Developmental Studies Hybridoma Bank (DSHB, University of Iowa). Laminin staining was performed to label basement membranes using a rabbit polyclonal anti-laminin antibody (Sigma-Aldrich catalogue no. L9393, 1:50 dilution).

Muscle sections were washed twice (5 min each) in PBS, saturated in Blocking Buffer (FBS 4%, BSA 4%, in PBS) for 1 h, washed once in PBS, incubated with 1:100 Fab Fragment Goat Anti-Mouse IgG (Jackson ImmunoResearch Catalogue no. 115-007-003) for 20 min, then washed once in PBS. Primary antibodies were incubated overnight at 4 °C with agitation. After washing 3 times in 0.1% Tween20 in PBS, secondary antibodies were used to selectively bind to each primary antibody: polyclonal goat anti-Mouse IgG2b cross-adsorbed secondary antibody, Alexa Fluor® 647 (Invitrogen catalogue no. A-21242), polyclonal goat anti-Mouse IgG1 cross-adsorbed secondary antibody, Alexa Fluor® 350 (Invitrogen catalogue no. A21120), polyclonal goat anti-mouse IgG1 cross-adsorbed secondary antibody, Alexa Fluor® 488 (Invitrogen catalogue no. A-21121), polyclonal goat anti-mouse IgM conjugated with Alexa Fluor® 555 (Invitrogen – A-21426), polyclonal goat anti-Rabbit IgG (H + L) Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor™ 488 (Invitrogen catalogue no. A-11034) all diluted at 1:400 in PBS. After three washes in 0.1% Tween20 in PBS, muscle sections were mounted using mounting media.

Images were taken using a x20 objective lens using a fluorescence microscope slide scanner (Axioscan Z1, ZEISS). Images were recorded using ZEN software (Version 3.40, ZEISS). Fiber type distribution and number were measured manually using ImageJ software.

Histology and morphometric analysis of muscle injury model

Frozen TA muscles were sectioned at 8 µm thickness using a cryostat at −25 °C (Leica CM1850, Germany). Sections were mounted on glass slides and stained with hematoxylin and eosin (H&E). Hematoxylin was used to stain nuclei, followed by eosin counterstaining to visualize cytoplasmic and extracellular components. Images were acquired using a Zeiss Axio Imager 2 microscope with Zeiss Zen 2 Pro Software (version 2.0) (Zeiss, Germany) under brightfield illumination. Centralized nuclei were quantified using ImageJ software (National Institutes of Health, Bethesda, MD, USA). Multiple non- overlapping fields from the mid-belly region of each muscle were analyzed. For centralized nuclei data, 600 fibers were analyzed per group.

Muscle force frequency and fatigue contraction protocols

Following cervical dislocation of the animal, the right soleus was immediately dissected and placed in a Krebs–Henseleit solution (117 mM NaCl, 4.7 mM KCl, 1.2 mM MgSO4, 1.2 mM KH2PO4, 24.8 mM NaHCO3, 2.5 mM CaCl2, 11.1 mM glucose). Silk sutures (4.0 – Fine Science Tools GmbH, Heidelberg, Germany) were attached to tendons at either end of the soleus and were suspended vertically in a buffer-filled organ bath between a hook and a length-controlled lever system (Aurora Scientific, Aurora, Canada). In vitro field stimulation using platinum electrodes was provided via a high-power bipolar stimulator (Aurora Scientific, Aurora, Canada) outputting supramaximal current (700 mA) operated with Dynamic Muscle Control software (version 5; Aurora Scientific, Aurora, Canada). After optimal contractile length (L0) was determined, the muscle was thermo-equilibrated in a Krebs–Henseleit solution for 15 minutes at 35 °C79.

The force-frequency relationship was then determined across stimulation frequencies of 1-150 Hz (1 s train duration; 0.25 ms pulse width, each separated by 1 minute). Following a 5 min recovery period in which muscle length was measured using digital calipers, fatigue resistance (expressed as a relative % to initial force) was further assessed across 150 repeated tetanic contractions (40 Hz, 1 s train duration, interspersed by 1 s). At the end of each experiment, Force (N) was normalized to muscle cross-sectional area (CSA; cm2) after dividing muscle mass (g) by the product of L0 (cm) and estimated muscle density (1.06 g/cm3) to allow specific force in N/cm2 to be calculated79.

High-resolution respirometry

Left soleus was immediately removed, weighed, and placed into relaxing and biopsy preservation solution for high-resolution respirometry (BIOPS: 2.77 mM CaK2EGTA, 7.23 mM K2EGTA, 5.77 mM Na2ATP, 6.56 mM MgCl2·6H2O, 20 mM taurine, 15 mM Na2phosphocreatine, 20 mM imidazole, 0.5 mM Dithiothreitol (DTT), and 50 mM MES hydrate). Under a dissecting microscope (MEIJI-LABAX Co LTD, 13D46, Tokyo, Japan) individual muscle fibers from the soleus muscle were gently separated along their longitudinal axis to create a thin sheet of muscle fibers one layer thick. Separation of muscle fibers was conducted in ice-cold BIOPS on a bed of ice and was done in less than 5 minutes to ensure sample dissection consistency. Tissue was then incubated with saponin (50 μg·mL-1) to permeabilize the tissue and washed with MiR05 (MiR05: 110 mM sucrose, 60 mM K-lactobionate, 20 mM HEPES, 20 mM taurine, 10 mM KH2PO4, 3 mM MgCl2, 0.5 mM EGTA, 1% (w/v) fatty acid-free BSA, pH 7.1). Samples were added to the Oxygraph-2K (Oroboros Instruments, Innsbrück, Austria) which contained 2 mL MiR05, under constant stirring (750 rpm) at 37 °C.

Substrates and inhibitors were added to the chamber and steady rates of respiration recorded. Complex I was assessed with glutamate (Glu; 10 mM) and malate (Mal; 1 mM), and pyruvate (Pyr; 5 mM). ADP (2.5 mM) was titrated to provide a measure of maximal complex I-supported oxidative phosphorylation (OXPHOS). Complex I & II-supported OXPHOS was assessed with the addition of succinate (Succ; 10 mM). Carbonyl cyanide m-chlorophenyl hydrazine (CCCP; 5 μM dissolved in DMSO) yielded an uncoupled state as a measure of maximal ETC capacity. Antimycin A rates were subtracted as a correcting factor from all respiratory measurements (AA; 12.5 μM dissolved in 95 % ethanol) to account for non-mitochondrial residual respiration. Complex IV activity was assessed as a proxy for mitochondrial content. Ascorbate (Asc; 2 mM) and TMPD (N, N, N’, N’-Tetramethyl-p-phenylenediamine dihydrochloride; 0.5 mM) were added to the chambers to assess complex IV activity and Sodium Azide (AZ; 20 mM) was added to inhibit all mitochondrial respiration. Data were processed using DatLab (version 6.1, Oroboros Instruments).

Analysis of mitochondrial function in soleus muscle of HFD fed mice used a modified methodology to assess fatty acid β-oxidation as previously described43. OXPHOS activity was monitored using a Substrate- Uncoupler- Inhibitor-Titration (SUIT) protocol using a carnitine-conjugated long chain fatty acid palmitoyl-L-carnitine substrate (0.04 mM). All data were processed using DatLab (version 6.1, Oroboros Instruments).

Citrate synthase assay

Citrate synthase activity was assayed according to published protocols42. Muscle tissue samples were homogenized in 100 mM K2HPO4/KH2PO4, 5 mM EDTA, 0.1-mM fructose-2,6-bisphosphate, 0.1% Triton X-100, and 1 mM dithiothreitol, pH 7.2. Citrate synthase activity was measured at 412 nm to detect the transfer of sulfhydryl groups to 5,5′-dithiobis(2-nitrobenzoic acid) (DTNB). Reaction buffer composition was 100 mM Tris · HCl, 0.2 mM acetyl CoA, 0.1 mM DTNB, and 1 mM oxaloacetate (omitted for control), pH 8.0. The reaction rates were linear for ≥4 min. Assays were performed in duplicate, and means were analyzed. Specific activities were expressed in international units (μmol substrate transformed to product/min) normalized to tissue weight.

Immunoblotting

Protein lysate extracts were analysed by SDS gel electrophoresis, loading equal amounts of protein (30 μg per sample). All samples were resolved using a 4–12% Bis-Tris gradient gel (Invitrogen, Paisley UK), The proteins were transferred to 0.2 µm nitrocellulose membranes (7 × 8.5 cm) (#1704270, Bio-Rad). Total protein was visualised using Ponceau and washed off in Tris-buffered saline with 0.1% Tween 20 detergent (TBST). Membranes were blocked with skimmed milk powder for 1 h at room temperature and incubated with primary antibodies overnight at 4 °C; Total OXPHOS Rodent WB Antibody Cocktail 1:1000 (ab110413; Abcam). A cocktail of 5 mouse antibodies, one each against CI subunit NDUFB8 (ab110242; Clone 220E9DH10C12), CII-30kDa (ab14714; Clone 21A11AE7), CIII-Core protein 2 (ab14745; clone 13G12AF12BB11), CIV subunit I (ab14705, Clone 1D6E1A8) and CV alpha subunit (ab14748, Clone 15H4C4). Blots were developed with horseradish peroxidase-linked secondary antibody: m-IgG Fc BP-HRP (recombinant protein; sc-525409, Santa Cruz), 1:1000 dilution, using enhanced chemiluminescence and were detected with SuperSignal™ West Atto Ultimate Sensitivity Substrate (A38554, Thermo Fisher). Images were recorded using iBrightTM FL1500 imaging system (Invitrogen). Quantitative densitometry was performed using ImageJ 1.54 f software.

Human acute endurance exercise study

Participants were aged 18–65 years, non-smokers, and free from chronic disease (except type 1 diabetes). Participants were excluded if they had diabetes-related complications, other chronic conditions, history of smoking, or BP > 140/90 mmHg at study visits. Sixty participants were enrolled in total and both male and female participants were included. Sex was established by self-report. Participant demographics are shown in Supplementary Table 1. Participants provided written informed consent prior to enrolment following approval from the NHS HRA North East Tyne & Wear South Research Ethics and Newcastle University Ethics Committees (16/NE/0192, ISRCTN63739203).

Study visit 1: Participants’ height, weight (seca 220 stadiometer/seca 889 scale; seca, Hamburg, Germany) and medical history were recorded. Participants were screened for cardiac anomalies using a modified 12-lead resting/exercising ECG. Peak oxygen uptake (VO2PEAK) and peak heart rate were defined using a maximal-graded walking treadmill (Valiant 2 CPET; Lode, Groningen, the Netherlands) test80.

Study visit 2: Participants arrived at the clinical research facility at 08:30 A.M. following an overnight fast, had abstained from exercise for 48 h. Participants were cannulated and resting (baseline) blood samples (10 ml) were drawn. Participants walked on an incline for 45 minutes at 60% VO2PEAK (Valiant 2 CPET; Lode, Groningen, the Netherlands) operated with Lode Ergometry Manager software (version 9; Lode, Groningen, the Netherlands). Participants’ treadmill velocity and gradient were calculated using VO2, velocity, and gradient data from the preliminary VO2PEAK test. Breath-by-breath respiratory parameters (Metalyzer 3B-R3, Cortex) were recorded, with gradient adjusted at 10 and 30 minutes if VO2 was >10% different than target VO2. On exercise completion blood was immediately drawn from the cannula. Blood was processed for serum which was stored at – 80 °C.

Human aerobic exercise training study

The study received ethical approval from the Medical Ethical Committee of Wageningen University, in accordance with the Declaration of Helsinki. For the present analysis a random sub-set of the serum from thirty-three participants were used. A detailed description of subject participation, experimental design, endurance training program, whole-body physiological outcome measures can be found in previous publications47. Demographics of the patients used in this study can be found in Supplementary Table 2.

Healthy male volunteers gave full written informed consent to participate in a 10-week aerobic exercise training program. Sex was established by self-reporting. All subjects were physically active, performing sports on a non-competitive basis between 1 and 4 h/week. The total study duration was 10 weeks, and included 28 endurance training sessions. Training sessions involved a 10-min warmup on a cycle ergometer, followed by the endurance training session of 60-min continuous cycling. All training sessions were conducted under the supervision of a researcher using indoor, mechanically braked spinning bikes (Body Bike Smart, Body Bike International). Heart rate (HR) for each session was determined (Polar Electro), HR and rate of perceived exertion were assessed at start and every 5 min. After the endurance training sessions participants performed a 10-min cooling down period on the same cycle ergometer. Exercise intensity was determined using published approaches81. The exercise intensity is “vigorous”82. Fasting blood was collected in EDTA-coated evacuated tubes (BD Biosciences) by venepuncture at baseline and study end. Fasting blood samples were collected after an overnight fast 3–4 days before the first training session (baseline), and following the 28th training session (Exercise trained).

LC-MS analysis of total BAIBA

LC-MS analysis of plasma and serum BAIBA was conducted following published methods19,83. Plasma samples were prepared for LC-MS analyses via protein precipitation with the addition of nine volumes of 74.9:24.9:0.2 vol/vol/vol acetonitrile/methanol/formic acid. The samples were centrifuged (10 min, 15,000 g, 4 °C) An Acquity UPLC system (Waters, USA) equipped with an Atlantis HILIC silica column (3 µm, 2.1 × 150 mm) held at 30 ◦C was used for the analysis of BAIBA and operated with MassLynx software (version 4.1; Waters, USA). The binary solvent system was solvent A comprising LC − MS-grade water, 0.1% formic acid, 10 mM ammonium formate and solvent B comprising 0.1% formic acid in acetonitrile.

Elution Gradient: 95% B held for 0.5 min at a flow rate of 250 μl/min, decreasing to 40% B over 10 mins at 250 μl/min with a hold of 40%B for 5 min at 250 μl/min. Then a return to 95% B for 2 min at 250 μl/min followed by an increase to 400 μl/min for 12.5 min. Flow was then returned to 250 μl/min over 1 minute and held for 0.5 min. Total runtime was 32 minutes.

The Acquity UPLC system was coupled to a Xevo TQ-XS mass spectrometer (Waters). Positive electrospray ionisation mode, a cone gas flow rate of 50 ml/h, and a desolvation temperature of 650 °C were used. BAIBA analyses were performed using multiple reaction monitoring (MRM) with a parent ion (m/z) in Q1 of 104.1, a fragment ion (m/z) in Q2 of 86 and a collision energy of 17. Samples were analysed in technical triplicate. For data related to Supplementary Fig. 1b, 23 independent samples were analysed. For data related to Supplementary Fig. 3a, 8 independent samples were analysed. For data related to Supplementary Fig. 5d, 40 independent samples were analysed. For data related to Supplementary Fig. 6a, 16 independent samples were analysed.

Data were processed and peak integration performed using Waters TargetLynx Version 4.1 (Waters, USA).

LC-MS analysis of D-BAIBA and L-BAIBA enantiomers

LC-MS analysis of D-BAIBA and L-BAIBA was based on published methods84. An internal standard spiking solution of 10 μM D,L-BAIBA-d3 (CDN Isotopes) in 0.1% formic acid in LC-MS-grade methanol was prepared. Serum or plasma (10 µl) was protein-precipitated with 40 µl internal standard spiking solution, vortex mixed and centrifuged (15 min; 13,000 g; 4 °C). The supernatant fraction was transferred to LC vials. Supernatant (10 µl) was injected on to an Acquity UPLC system (Waters, USA) equipped with a chiral SPP-TeicoShell, 150 × 4.6 mm, 2.7 µm column (AZYP LLC., Arlington, TX) with a Max-RP 50 × 2.0 mm guard column (Phenomenex, Torrance, CA) and held at room temperature. The binary solvent system used for the analysis of the BAIBA enantiomers was solvent A comprising LC − MS-grade methanol and solvent B comprising 0.005% formic acid and 2.5 mM ammonium formate in LC-MS-grade water. The mobile phase was set at a flow rate of 0.6 mL/min. The column mobile phase was held at 25% solvent B for 0–10 min before increasing to 98% solvent B over 10–10.1 min. The mobile phase was then held at 98% B from 10.1–17 min before returning to 25% B 17.1-25 min. The Acquity UPLC system was coupled to a Xevo TQ-XS mass spectrometer (Waters) and operated with MassLynx software (version 4.1; Waters, USA. Analyses were performed using multiple reaction monitoring (L-BAIBA and D-BAIBA 104.1 m/z → 86 m/z, cone voltage = 12 V, collision energy = 12 eV) (Supplementary Fig. 11). Positive electrospray ionisation was used. A cone gas flow rate of 50 ml/h and a desolvation temperature of 650 °C were used. Samples were analysed in technical triplicate. For data related to Fig. 7a, b, 24 independent samples were analysed. For data related to Fig. 8a–d and Supplementary Fig. 9a–f, 60 independent samples were analysed. For data related to Fig. 8e–g, 66 independent samples were analysed. For data related to Fig. 9c and Supplementary Fig 10c, 40 independent samples were analysed. Data were processed and peak integration performed using Waters TargetLynx Version 4.1 (Waters, USA).

RNAseq analysis

All RNA samples were quality checked on the Tapestation 4200 operated with TapeStation Analysis Software (version 3; Agilent), using the RNA tapes (Agilent), as well as the Quant-iT RNA kit (Invitrogen). 200 ng of total RNA was taken forward, and libraries prepared using the Illumina Stranded mRNA library preparation kit (Illumina) using their standard operating procedure. 12 cycles were used for the enrichment PCR. Libraries were quality checked on the Qubit BR dsDNA kit (Invitrogen) and Tapestation 4200 with D1000 tapes (Agilent). Any libraries that still had visible adapter dimer had an additional bead clean up to remove adapter dimer. Libraries were pooled at 23 ng each, and sequenced on the Illumina NextSeq2000 using the P2 100 cycle SBS kit in a single end sequencing run. Data was converted using Illumina’s BCL_Convert off instrument.

Raw FASTQ files were trimmed using TrimGalore to remove low-quality reads and overrepresented sequences. A reference mouse genome was retrieved from the Ensembl genome database (GRCm39). STAR aligner was then used to align the trimmed sequences to the reference human genome. Once aligned, raw read counts were calculated using FeatureCounts in R. DEGs were identified using DESeq2 package (v1.40). Gene ontology enrichment was performed using ClusterProfileR (v4.14.4). Gene Set Enrichment Analysis was performed using Enrichr (2021)28 and the MSigDB Molecular Signatures Database29. For all RNA-seq analyses, adjusted p value < 0.05 was set as a threshold for statistical significance. The R version used for the analysis was v4.3.0. The raw data of the RNAseq data was deposited in the EBI ArrayExpress repository (https://www.ebi.ac.uk/biostudies/arrayexpress) can be accessed via https://www.ebi.ac.uk/arrayexpress/experiments/E-MTAB-14746 (human skeletal muscle cell datasets) and https://www.ebi.ac.uk/arrayexpress/experiments/E-MTAB-14747 (murine soleus datasets).

Statistical analysis

Sample sizes were calculated using power calculations. Animals/cell culture wells were randomly assigned to experimental groups. Group variance was analyzed with an F test. Statistical significance was assessed using one-way or two-way ANOVA (Tukey’s, Holm–Sidak’s, Fisher’s and Dunnett’s post hoc test), or two-tailed unpaired Student’s t test, as detailed. In each case, n  ≥  3 and represents independent biological repeats. The significance level was set to p  ≤  0.05 or adjusted p  ≤  0.05 (FDR; RNAseq data). Univariate analysis was conducted using GraphPad Prism (version 10) software. Differential expression analysis statistics were conducted in DESeq2 (v1.4) (Bioconductor). Indirect calorimetry data were analyzed, and p values were calculated using ANCOVA/Generalized Linear Model with body mass as a covariate in CaIR (version 1.3, https://calrapp.org/)77.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.