EXERCISE MODALITY AND METABOLIC STATUS DRIVE DISTINCT LIPIDOMIC RESPONSES AFTER ACUTE AEROBIC, RESISTANCE AND COMBINED TRAINING
Description
Lipid metabolism is significantly altered in obesity and type 2 diabetes (T2D) and is modulated by exercise modality, intensity, and duration. While high-intensity aerobic training (AT) and resistance training (RT) have been shown to modify metabolic profiles, the acute lipidomic response to moderate-intensity exercise, particularly in individuals with metabolic disorders, remains unclear. This study used high-resolution direct-infusion mass spectrometry (DIMS) lipidomics to investigate the acute lipidomic responses to AT, RT, and combined training (CT) in individuals with normal weight (NW), obesity (OB), and T2D. A randomised crossover design was employed, with 57 individuals (18 NW, 22 OB, 17 T2D) undergoing three supervised training sessions (AT, RT, CT). Blood samples were collected pre-, post-, 30 min and 60 min post-exercise and lipidomic profiles were assessed using untargeted direct infusion–high-resolution mass spectrometry (DI-HRMS). NW individuals showed the most dynamic responses, particularly in glycerophospholipids (GP) and sphingolipids (SP), reflecting greater insulin sensitivity. OB and T2D groups showed delayed lipid recovery, with prolonged elevated levels of triacylglycerols (TAGs), diacylglycerols (DAGs), and ceramides (Cers), indicative of insulin resistance (IR), and metabolic inflexibility. RT and CT were more effective than AT in reducing pro-inflammatory lipids, such as DAG, ganglioside GM3 and Cer. Hierarchical clustering analysis (HCA) further revealed distinct lipid regulation patterns and lipid class-specific cluster enrichment, modulated by both exercise and metabolic phenotype. These findings highlight importance of personalised exercise prescriptions, particularly those incorporating resistance components, have the potential to optimise metabolic responses and enhance lipid metabolism in individuals with IR.
Files
Steps to reproduce
All samples were randomized for data acquisition and directly infused in a HESI LTQ Orbitrap-MS Discovery (Thermo Scientific, Bremen, Germany) with 30 000 FWHM of mass resolution. Direct injection of the sample was carried out in positive mode, and acquisitions were performed in quintuplicate within the mass range of m/z 200 to 1200. The flow rate utilized was 10 μL/min, whilst the temperature was set at 280°C. The spray voltage was set at 5 kV, and the sheath gas was maintained at 10 arbitrary units. For spectra acquisition and visualisation, the software employed was XCalibur 3.0.63 (Thermo Scientific, Bremen, Germany). Fragmentation of selected markers after statistical analysis was achieved by an MS/MS (MS2) experiment directly injected into the Thermo Q Exactive mass spectrometer (Thermo Scientific, Bremen, Germany). An inclusion list of markers was inserted, followed by the preparation of a sample pool for each group after the addition of 10 μL of sample and 990 μL of methanol. The pool was then centrifuged at 1088 g for 5 min at 5°C. Then 250 μL of the supernatant and 250 μL of water containing 0.1% (v/v) formic acid were transferred to an Eppendorf tube. A full scan of each spectrum was collected, followed by two experiments: the first performed with low collision energy at 20, 30, and 40 eV, and the second with high collision energy at 40, 50, and 60 eV. The flow rate was set to 15 μL/min, the temperature to 320°C, the spray voltage to 4 kV, and the sheath gas to 12 arbitrary units. A resolution of 17,500 in positive mode was used to acquire spectra in the mass range of m/z 200 to 1,200. Spectra were acquired and processed using XCalibur 3.0.63 (Thermo Scientific, Bremen, Germany). The DIMS data was processed through the Galaxy-M workflow in R. Once the table of peak intensities and mass-to-charge ratio (m/z) values was obtained, the data underwent a filtering process prior to statistical analyses. Initially, variables with a value of zero in more than 20% of samples were removed, as well as those with a coefficient of variation (CV) greater than 20% in technical replicates. The missing data were then imputed using a constant value of half the minimum (HM). The data were subsequently transformed using the logarithm to the base 10 (log10) function to decrease the variation in peak intensities among samples and then scaled using the Pareto-scaling method to reduce data variability between samples.
Institutions
- University of Aberdeen