Metabolomic profiling of intensity-dependent responses to acute exercise in healthy humans.

Skip to main content An official website of the United States government Here's how you know Here's how you know Official websites use .gov A .gov website belongs to an official government organization in the United States. Secure .gov websites use HTTPS A lock ( Lock Locked padlock icon ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites. Search Log in Dashboard Publications Account settings Log out Search… Search NCBI Primary site navigation Search Logged in as: Dashboard Publications Account settings Log in Search PMC Full-Text Archive Search in PMC Journal List User Guide PERMALINK Copy As a library, NLM provides access to scientific literature. Inclusion in an NLM database does not imply endorsement of, or agreement with, the contents by NLM or the National Institutes of Health. Learn more: PMC Disclaimer | PMC Copyright Notice J Transl Med . 2026 Mar 5;24:502. doi: 10.1186/s12967-026-07937-1 Search in PMC Search in PubMed View in NLM Catalog Add to search Metabolomic profiling of intensity-dependent responses to acute exercise in healthy humans Jianxiu Liu Jianxiu Liu 1 Division of Sports Science and Physical Education, Tsinghua University, Beijing, 100084 China 2 IDG/McGovern Institute for Brain Research, Tsinghua University, Beijing, 100084 China Find articles by Jianxiu Liu 1, 2, # , Junxin Zhang Junxin Zhang 3 Sports Coaching College, Beijing Sport University, Beijing, 100084 China Find articles by Junxin Zhang 3, # , Xingtian Li Xingtian Li 1 Division of Sports Science and Physical Education, Tsinghua University, Beijing, 100084 China Find articles by Xingtian Li 1, # , BaiLe Wu BaiLe Wu 4 Department of Exercise Physiology, School of Sports Science, Beijing Sport University, Beijing, 100084 China Find articles by BaiLe Wu 4 , Alimjan Ablitip Alimjan Ablitip 1 Division of Sports Science and Physical Education, Tsinghua University, Beijing, 100084 China Find articles by Alimjan Ablitip 1 , Dizhi Wang Dizhi Wang 1 Division of Sports Science and Physical Education, Tsinghua University, Beijing, 100084 China Find articles by Dizhi Wang 1 , Limei Ke Limei Ke 5 School of Biomedical Engineering, Tsinghua University, Beijing, China Find articles by Limei Ke 5 , Qian Di Qian Di 6 Vanke School of Public Health, Tsinghua University, Beijing, China 7 Institute for Healthy China, Tsinghua University, Beijing, China Find articles by Qian Di 6, 7, ✉ , Ruidong Liu Ruidong Liu 3 Sports Coaching College, Beijing Sport University, Beijing, 100084 China Find articles by Ruidong Liu 3, ✉ , Xindong Ma Xindong Ma 1 Division of Sports Science and Physical Education, Tsinghua University, Beijing, 100084 China 2 IDG/McGovern Institute for Brain Research, Tsinghua University, Beijing, 100084 China Find articles by Xindong Ma 1, 2, ✉ Author information Article notes Copyright and License information 1 Division of Sports Science and Physical Education, Tsinghua University, Beijing, 100084 China 2 IDG/McGovern Institute for Brain Research, Tsinghua University, Beijing, 100084 China 3 Sports Coaching College, Beijing Sport University, Beijing, 100084 China 4 Department of Exercise Physiology, School of Sports Science, Beijing Sport University, Beijing, 100084 China 5 School of Biomedical Engineering, Tsinghua University, Beijing, China 6 Vanke School of Public Health, Tsinghua University, Beijing, China 7 Institute for Healthy China, Tsinghua University, Beijing, China ✉ Corresponding author. # Contributed equally. Received 2025 Sep 26; Accepted 2026 Feb 10; Collection date 2026. © The Author(s) 2026 Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/ . PMC Copyright notice PMCID: PMC13072661  PMID: 41787548 Introduction Exercise confers numerous health benefits, yet the intensity-specific impact of acute exercise on systemic metabolism remains incompletely characterized. This study aimed to characterize serum metabolomic responses to different treadmill-based exercise intensities using liquid chromatography–mass spectrometry (LC–MS) and to assess associations between metabolite signatures and exercise capacity indicators. Methods We conducted a randomized controlled trial (ClinicalTrials.gov Identifier: NCT04830059 ). Fifty-eight healthy, inactive university students were randomized to high-intensity interval training (HIIT, n = 20), moderate-intensity continuous training (MICT, n = 19), or light-intensity training (LIT, n = 19). Serum samples were collected at rest and immediately post-exercise. An untargeted LC–MS metabolomics approach was used to quantify alterations in 480 serum metabolites across the three protocols. Results Acute HIIT, MICT, and LIT significantly altered 204, 198, and 60 serum metabolites, respectively, with a shared core of 37 metabolites. Metabolic perturbations were clearly intensity-dependent: HIIT elicited the most profound shifts, characterized by rapid accumulation of glycolytic, TCA-cycle, and antioxidant-associated metabolites. MICT induced broad activation of glycolysis, the TCA cycle, and fatty acid oxidation, whereas LIT caused milder changes primarily in glycolysis and bile acid signaling. Across all intensities, strong positive correlations between lactate and TCA-cycle intermediates underscored the tight coupling between glycolytic flux and mitochondrial respiration. Baseline VO 2 max and HRmax showed distinct metabolite association profiles, with VO 2 max positively related to amino acid–derived and redox-related metabolites, whereas HRmax was linked to organic acid, bile acid, and dipeptide species. Conclusions This parallel-group metabolomics trial demonstrates that a single bout of treadmill exercise elicits intensity-dependent, yet partly overlapping, metabolic programs spanning central carbon, amino acid, lipid, and bile acid pathways. By integrating lactate-centered correlations with VO 2max - and HRmax-related metabolite signatures, our findings outline a molecular framework linking acute exercise intensities metabolism to cardiorespiratory fitness and provide a basis for refining intensity-specific exercise prescriptions. The study was registered on ClinicalTrials.gov (Identifier: NCT04830059 ). Supplementary Information The online version contains supplementary material available at 10.1186/s12967-026-07937-1. Keywords: Exercise intensity, Metabolomics, Lactate, VO 2 max Highlights This randomized controlled trial maps distinct serum metabolomic signatures for high-intensity interval training (HIIT), moderate-intensity continuous training (MICT), and light-intensity training (LIT). A core metabolic scaffold involving central carbon pathways is shared across all three intensities. HIIT drives rapid energy flux, MICT promotes fat oxidation, and LIT primarily modulates bile acid signalling. Lactate acts as a central hub connecting glycolysis, mitochondrial oxidation, and lipid remodelling. VO 2 max correlates with antioxidant and amino acid–related metabolites, whereas HRmax links to organic acids, bile acids, and dipeptides. Supplementary Information The online version contains supplementary material available at 10.1186/s12967-026-07937-1. Introduction Regular physical activity improves metabolic health and enhances insulin sensitivity [ 1 ], while also supporting immune function [ 2 ]. It is an effective non-pharmacological strategy for the prevention and treatment of cardiovascular and metabolic disorders, as well as other chronic diseases [ 3 ]. However, the mechanisms through which exercise exerts these systemic benefits remain only partially understood. Emerging work suggests that a substantial proportion of these effects is mediated by complex, coordinated metabolic responses during and after exercise, involving multiple interconnected biochemical pathways and metabolites [ 4 ]. Given the vast complexity of the human metabolome—reflected in pathway resources cataloguing > 110,000 metabolic and signalling pathways [ 5 ]—approaches that quantify only a limited panel of metabolites may not fully capture coordinated, system-wide metabolic adaptations [ 6 ]. Recent advances in high-throughput metabolite profiling have made it possible to capture a much more comprehensive snapshot of the organism’ s metabolic status [ 7 ]. Metabolomics, defined as the systematic characterization of small molecules (metabolites) in biological samples, usually with a molecular weight below 1500 Da [ 8 , 9 ], offers a powerful approach to interrogate these responses. Because metabolites represent the integrated end products of genomic, transcriptomic, proteomic, and environmental interactions [ 9 ], metabolic profiling in the context of exercise can provide deeper insight into how physical activity promotes health and contributes to disease prevention. Exercise intensity is a key determinant of both acute metabolic responses and long-term training adaptations. High-intensity interval training (HIIT), moderate-intensity continuous training (MICT), and light-intensity training (LIT) are widely used in practice and span a broad spectrum of relative intensities. HIIT typically involves brief bouts of vigorous exercise interspersed with periods of rest or low-intensity activity, whereas MICT is characterized by sustained, moderate effort [ 4 ]. LIT, although less metabolically demanding, contributes to health by helping to maintain energy balance and supporting cardiovascular function [ 10 ]. At the metabolic level, HIIT is often associated with greater engagement of glycolytic pathways, accompanied by pronounced lactate accumulation and marked perturbations in central carbon and amino-acid metabolism [ 11 ]. By contrast, MICT promotes more sustained aerobic metabolism and higher rates of fatty acid oxidation [ 12 ], while LIT can still induce favourable changes in substrate utilization and cardiometabolic risk markers, particularly when performed regularly over time [ 13 ]. These intensity-dependent differences suggest that distinct metabolic programmes may be recruited across light-, moderate-, and high-intensity domains. However, the specific metabolite signatures and pathway-level networks that characterize each intensity domain—as well as their relationships to homeostatic regulation and exercise capacity—remain incompletely defined. Metabolomics-based studies have begun to map the primarily acute, and to a lesser extent chronic, metabolic responses to exercise in humans, revealing broad alterations in amino acid, lipid, carbohydrate, and bile acid pathways [ 6 , 14 ]. Systematic reviews highlight both the promise of exercise metabolomics for characterizing the metabolic correlates of training adaptations and performance, and important limitations of the existing literature, including small sample sizes, heterogeneous protocols, and limited coverage of exercise intensities and fitness phenotypes [ 6 , 9 ]. In particular, relatively few randomized controlled trials have systematically compared metabolomic responses across clearly defined intensity domains within a unified experimental framework, while concurrently relating these metabolic changes to cardiorespiratory fitness. In the present randomized controlled trial, we used untargeted serum metabolomics to characterize acute metabolic responses to three treadmill exercise intensity domains (HIIT, MICT, and LIT). We aimed (i) to identify shared versus intensity-specific metabolite signatures and pathway-level networks, and (ii) to relate baseline profiles and exercise-induced changes to key indicators of exercise capacity (blood lactate, VO₂max, and HRmax). By integrating intensity-stratified metabolomics with physiological phenotyping within a unified experimental framework, we sought to address gaps highlighted by recent reviews regarding the limited availability of well-controlled studies spanning different exercise intensities/workloads and the under-explored links between exercise-induced metabolomic remodeling and cardiorespiratory fitness-related traits [ 6 , 14 ]. Methods Study design and population This randomized controlled trial was registered at ClinicalTrials.gov ( NCT04830059 ). The present manuscript reports a sub-analysis focusing on exercise intensity-dependent acute metabolomic responses. Participants were recruited from the university campus through poster advertisements and online promotions. The inclusion criteria were as follows: (1) healthy, inactive university students aged 20–30 years with a normal BMI; (2) no history of chronic diseases or mental illness; (3) no regular physical activity habits, as determined by the International Physical Activity Questionnaire (IPAQ); and (4) no physical restrictions, as assessed by the Physical Activity Readiness Questionnaire (PAR-Q). A priori power analysis (G*Power v3.1) for a repeated-measures ANOVA (within–between interaction) indicated that a minimum of 42 participants (14 per group) was required to detect a medium effect size (f = 0.25) with 80% power (α = 0.05, r = 0.5, ε = 1) [ 15 , 16 ]. To account for potential attrition and ensure robust power for secondary metabolomics analyses, we targeted a larger enrollment. Ultimately, 60 eligible participants were randomized to the HIIT, MICT, or LIT groups. Following data quality control, two participants were excluded as outliers based on Principal Component Analysis (PCA), resulting in a final analytical sample of 58 participants (HIIT n = 20, MICT n = 19, LIT n = 19; 27 males, 31 females). This final sample size exceeds the a priori requirement and is comparable to recent exercise metabolomics trials that have successfully detected significant metabolic alterations with similar cohort sizes [ 17 ]. Participant characteristics are detailed in Table 1 . Table 1. Basic characterization of participants Group HIIT MICT LIT p N(male) 20 (7) 19 (10) 19 (10) Age (years) 24.5 ± 2.1 25.8 ± 1.4 24.9 ± 2.1 0.1 Height (cm) 168.4 ± 6.5 171.1 ± 7.4 168.6 ± 6.6 0.4 Weight (kg) 60.6 ± 10 66.6 ± 14.6 62.7 ± 10.8 0.3 BMI (kg/m 2 ) 21.2 ± 2.6 22.5 ± 3.5 22 ± 3.1 0.5 VO 2 max[mL/(kg·min)] 43.6 ± 7.6 42.5 ± 6.2 44.6 ± 6.4 0.6 HRmax 188.4 ± 8.4 184.2 ± 9.5 191.6 ± 11.4 0.07 Open in a new tab HIIT: high-intensity interval training, LIT: light-intensity training, MICT: moderate-intensity continuous training, VO 2 max: maximum oxygen uptake, HRmax: maximum heart rate. Values are group mean ± SD Participants were simply randomly assigned by computer-generated random numbers to ensure equal probability of assignment. Randomization was performed by independent personnel who were not involved in subsequent measurements or data analysis to minimize possible bias. Throughout the study, participants were identified only by unique identifiers. Group assignments were stored in opaque, sealed envelopes and opened only after baseline assessments were completed, ensuring allocation concealment and reducing selection bias. Outcome assessors and data analysts remained blinded to group allocation throughout the trial. To minimize potential contamination, participants from different intervention arms attended the fitness center on separate days and were not allowed to observe the exercise protocols of other groups. After researchers calculated individual intervention protocols based on VO₂max, participants then underwent treadmill exercise; however, they remained unaware of their specific group assignments and the exercise protocols of others. All experimental sessions started at 7:30 a.m. and participants had to fast overnight to minimize external variables. The study received ethical approval from the Ethics Committee of Tsinghua University (IRB 20190091) and was conducted in strict accordance with the Declaration of Helsinki. Informed consent was obtained from all participants, confirming their voluntary participation in the study. Process and interventions Study procedure Participants arrived at the Tsinghua University Physical Training and Rehabilitation Research Center after a 12-hour fast. Participants remained fasted before and after exercise to minimize external influences. Baseline blood samples (baseline) were collected from 7:30 to 9:00 a.m. After warming up, participants underwent an acute exercise intervention, with continuous heart rate monitoring during exercise. All measurements were completed in the morning to avoid time-of-day interference. The post-exercise blood samples (post-exercise) were collected immediately after exercise. To maintain sample stability and prevent metabolic degradation, all the blood samples were placed on ice for 30 min and then centrifuged at 4000 rpm for 10 min at 4 °C. Serum was collected and stored at − 80 °C until analysis. Although plasma samples require anticoagulants to prevent clotting, this study specifically utilized serum to align with our predefined metabolomics workflow and ensure consistency with previous serum-based metabolite profiling studies. The experimental process of the RCT study is illustrated in Fig. 1 A. Fig. 1. Open in a new tab Research design, molecular responses to exercise, and individual variability. Study design. ( A ) Participants were randomly assigned to one of three acute exercise interventions (HIIT, MICT, or LIT). Serum samples were collected immediately before and after the exercise bout for metabolomic analysis. ( B1 - B5 ) Principal component analysis (PCA) of serum metabolite profiles. ( B1 ) Baseline PCA of all participants, with HA, MA, and LA denoting baseline samples from the HIIT, MICT, and LIT groups, respectively. ( B2 ) Post-exercise PCA of all participants, with HB, MB, and LB denoting post-exercise samples from the HIIT, MICT, and LIT groups, respectively. ( B3 – B5 ) Within-group PCA plots showing pre- versus post-exercise separation for the HIIT ( B3 ), MICT ( B4 ), and LIT ( B5 ) groups. ( C ) Overall inter-individual variability of metabolite levels at baseline and in response to exercise. Boxplots display the distributions of the coefficient of variation (CV%) of baseline concentrations (left) and exercise-induced fold changes (right) for the HIIT, MICT, and LIT groups. ( D1 - D3 ) Metabolite-wise inter-individual variability under different exercise intensities. Each point represents a single metabolite, plotting the coefficient of variation (CV) of baseline levels (x-axis, %) against the CV of exercise-induced fold changes (y-axis, %) for the HIIT ( D1 ), MICT ( D2 ), and LIT ( D3 ) groups The measurement of VO 2 max and intervention protocol VO 2 max was measured one week before the intervention to determine exercise intensity for each participant. VO 2 max is a method for measuring the capacity to uptake and utilize oxygen during intense whole-body exercise, providing a more accurate method for determining individualized intensity levels in HIIT and MICT [ 18 ]. The Bruce treadmill protocol, as described by Hanson et al., was used to measure VO 2 max, with oxygen consumption and ventilation monitored using the Cosmed Fitmate metabolic system (Cosmed, Rome, Italy) [ 19 ]. Participants ran on a treadmill with gradually increasing speed and incline until termination criteria were met, including symptoms like dyspnea, cyanosis, dizziness, nausea, chest pain, extreme fatigue, abnormal blood pressure, arrhythmia, RPE ≥ 19, respiratory exchange ratio ≥ 1.1, inability to continue running, or inability to maintain the required speed for 10 s. Heart rate, expiratory, and inspiratory volumes were recorded during the test [ 20 ]. Two professional trainers and research assistants ensured participant safety during the VO 2 max test. Individualized HIIT and MICT intervention protocols were designed based on standard methods from previous studies [ 21 – 23 ]. For the HIIT group, participants ran at their maximum VO 2 speed for 1 min, followed by running at 50% of their VO 2 max speed for 1 min, alternating for 20 min and 10 cycles. The MICT group ran at a speed equivalent to 70%-75% of their individual VO 2 max for 20 min. Heart rate monitors (Polar H10, Finland) were used to detect heart rate and monitor running speed. Both HIIT and MICT sessions included a 10-minute warm-up and cool-down period, totaling 30 min of intervention. The LIT group performed 30 min of light-intensity stretching activities. Liquid chromatography-mass spectrometry for untargeted metabolomics Untargeted metabolomics analysis was conducted at the Metabolomics and Lipidomics Center of Tsinghua University, following a previously established and published methodology [ 24 ]. Metabolite extracts underwent analysis using HILIC and RPLC separations, conducted in both positive and negative ionization modes. Data collection was performed using a Q Exactive HFX mass spectrometer for HILIC and a Q Exactive HF mass spectrometer for RPLC, both from Thermo Fisher Scientific, CA. Both instruments featured HESI-II probes and operated in full MS-ddms2 scan mode for each sample. HILIC separations utilized a BEH Amide column (2.1 × 100 mm, 1.7 μm, Waters), with mobile phases containing 10 mM ammonium acetate in 95% acetonitrile/5% water (A) and 10 mM ammonium formate in 50% acetonitrile/50% water (B). RPLC separations employed a BEH C18 column (2.1 × 100 mm, 1.7 μm, Waters) with mobile phases consisting of 5 mM ammonium bicarbonate in water (A) and pure acetonitrile (B). To maintain data integrity, the system was equilibrated by injecting 10 samples before running the RPLC and HILIC sequences. Additionally, a pooled QC sample was injected every 20 injections to monitor signal stability, and for each sample the mass accuracy, retention time, and peak shape of the spiked-in internal standards were verified to ensure data integrity and analytical consistency. Data were acquired in positive and negative ion modes, covering mass ranges of m/z 80-1200 and m/z 70-1050, respectively, using data-dependent MSMS acquisition. Full scans and fragment spectra were recorded at resolutions of 60,000 and 15,000, respectively. The mass spectrometry parameters included a spray voltage of 2.8 kV in negative ion mode and 3.2 kV in positive ion mode, a capillary temperature of 320 °C, a heater temperature of 300 °C, a sheath gas flow rate of 35 units, and an auxiliary gas flow rate of 10 units. Metabolites were identified using Tracefinder 3.2 (Thermo, CA), referencing an internal library containing MS/MS spectra of over 1500 endogenous metabolites. Mass tolerance for precursor and fragment ions was set at 10 ppm and 15 ppm, respectively. Metabolite identification was based on matching fragment ions with library spectra, with two levels of identification confidence: one confirmed by MS/MS data and the other based on accurate mass assignment of precursor ions. Relative quantification was performed by measuring chromatographic peak areas, allowing for retention time shifts of up to 0.25 min for peak alignment. These raw peak areas, validated for stability by the QC samples and internal standards, were then used for statistical comparisons between experimental groups to determine significant metabolic changes. Statistical analysis Statistical analyses were performed using R version 4.3.1 on data obtained from serum samples subjected to three different intensities of acute interventions (HIIT, MICT, LIT). To ensure standardization and mitigate the impact of dimensional differences, all metabolite data were log2-transformed and Z-score standardized. To strictly account for within-subject correlations inherent in the repeated-measures design, differences in metabolite levels pre- and post-intervention were assessed using linear mixed models implemented via the limma R package. This approach allows for a robust estimation of intervention effects across multiple groups. To control for multiple hypothesis testing, raw P-values were adjusted using the Benjamini-Hochberg (BH) procedure, with statistical significance defined as a False Discovery Rate (FDR) of < 0.05. Principal component analysis Principal component analysis (PCA) was performed to explore differences in metabolite levels within and between the three groups at baseline and post-intervention. For visualization, the “fviz_pca_ind ”function in R was used to plot principal component analysis for individual samples, with ellipses representing group clustering. The first two principal components (PC1 and PC2) were used to describe variance, with differences between groups highlighted by color coding. PCA plots were created for comparing the post-intervention data for all the three groups to the baseline data, as well as the baseline and post-intervention data for each group and for the entire dataset. Inter-individual Variability To avoid potential bias, the coefficient of variation (CV) for each metabolite was calculated using the non-imputed datasets. The coefficient of variation was calculated as CV = (standard deviation/mean) * 100. Metabolomics datasets were normalized according to experimental conditions. Inter-individual variability in response to exercise was assessed by calculating the median CV for each metabolite at all time points after exercise relative to baseline. Box plots show the first quartile (lower edge of box), median (center line), and third quartile (upper edge of box). The upper spoke line represents the third quartile + 1.5 x (interquartile spacing) and the lower spoke line represents the first quartile − 1.5 x (interquartile spacing). Pathway enrichment analysis We performed KEGG pathway enrichment analysis of significant metabolites using the MetaboAnalyst platform. P values were corrected by the BH method in order to control for false-positive rates due to multiple hypothesis testing. We defined pathways with FDR less than 0.05 as significantly enriched pathways [ 25 ]. To assess the overall trend of these pathways, we determined whether each pathway was up- or down-regulated as a whole by calculating the median fold change value of significant molecules in the pathway. Partial least squares discriminant analysis Partial Least Squares Discriminant Analysis (PLS-DA) was performed using the MetaboAnalyst platform to identify metabolites that contributed significantly to between-group differences. The analysis assessed both within-group differences (baseline vs. post-test) and between-group differences (post-test comparisons among the three groups). The importance of each metabolite in the PLS-DA model was ranked by calculating the predictor variable importance (VIP) score. Hierarchical clustering analysis After log2 transformation and Z-score normalization, hierarchical clustering was performed using the hclust function in R (v4.3.1). The Euclidean distance matrix was calculated from the normalized data to quantify the similarity of the paired samples. Clustering was then performed using the Ward.D2 linkage method, which minimizes variance within clusters to produce tightly clustered and balanced groups. Correlation network analysis Pairwise Spearman rank correlations were calculated using the R package “Hmisc” (v4.1-1), and weighted undirected networks were plotted using “igraph” (v1.2.1). Correlations with Bonferroni-adjusted P-values below 0.05 were included and displayed using the Fruchterman-Reingold method. Only the main network, represented by the largest connected component, was plotted. Nodes were color-coded according to the metabolite group or correlation direction, and their sizes represent both the ‘betweenness centrality’ calculated by the ‘betweenness’ function in ‘igraph’ and the fold change response to exercise. Correlation analysis between L-lactic acid and metabolites Pairwise Pearson correlation analysis was used in order to investigate the relationship between L-Lactic Acid and other metabolites. The analysis included baseline and post-exercise measurements for three different groups. Correlation coefficients (r) and associated P-values were calculated using the Pearson correlation method. P values were adjusted using BH, and metabolites with an FDR of less than 0.05 were considered to be significantly correlated with L-Lactic Acid. Multivariate linear regression to identify metabolites associated with VO 2 max, HRmax, and L-lactic acid changes Multiple linear regression analyses were used to investigate the relationship between baseline metabolite levels and changes in metabolite levels (post test - baseline) with VO 2 max, HRmax, and L-Lactic Acid changes (post-test - baseline) across the three groups. The multivariate linear regression model were adjusted for sex, age, and BMI. P-values were calculated for regression coefficients related to VO 2 max and HRmax and adjusted for P-values using the BH method, with metabolites having an FDR of less than 0.05 considered significant. Results Effects of acute exercise patterns on metabolomic changes Table 1 summarizes the basic characteristics and exercise parameters of the participants. There were no significant differences in age, height, body mass index (BMI), VO 2 max and HRmax between the intervention groups in the baseline, with a mean age of 25 ± 2 years and a mean BMI of 22 ± 3 kg/m². A total of 480 metabolites were identified and quantified in these three sets of serum samples (Table S1 ). Based on these metabolite datasets, a principal component analysis (PCA) was performed (Fig. 1 B1–B5). The results of the PCA revealed that the metabolic profiles of the three exercise groups were highly overlapping at baseline (Fig. 1 B1). Following the intervention, visible shifts in metabolic trajectories were observed; however, the global PCA of post-exercise data showed that the profiles of the HIIT, MICT, and LIT groups still largely overlapped without distinct separation in this unsupervised space (Fig. 1 B2). Within-group comparisons showed clearer pre- to post-exercise shifts. A separation of pre- and post-exercise metabolic profiles was evident in the HIIT group (Fig. 1 B3). Similarly, the MICT group exhibited a clear shift from baseline to post-exercise (Fig. 1 B4). In contrast, changes in the LIT group were less pronounced, with a greater degree of overlap between pre- and post-exercise confidence ellipses (Fig. 1 B5). Further descriptive statistics analyzed the variability of overall metabolism across exercise-intensity groups (Fig. 1 C). At baseline, the MICT group exhibited the highest median coefficient of variation (CV%), followed by the HIIT and LIT groups (42.20%, 37.54%, and 36.28%, respectively; Fig. 1 C, left side). For exercise-induced metabolic changes (fold change), the median CV% remained highest in the MICT group, followed by the HIIT and LIT groups (44.21%, 42.18%, and 38.38%, respectively; Fig. 1 C, right side). Metabolite-wise variability analysis (Fig. 1 D1–D3) showed distinct patterns of heterogeneity. In the HIIT group, medium- and long-chain acylcarnitines (e.g., L-octanoylcarnitine, myristoylcarnitine, hydroxyhexanoylcarnitine) had among the highest fold-change CVs. In the MICT and LIT groups, several lysophospholipids (e.g., LysoPCs and LysoPEs), as well as sphingosine and phytosphingosine, also exhibited high CVs. In addition, small molecules such as quinic acid, caffeine, and 2-methylbutyrylglycine frequently displayed elevated baseline CVs across groups. Metabolic responses to acute exercise Exercise interventions of different intensities were associated with significant changes in metabolite levels, with 204, 198, and 60 metabolites significantly altered in the HIIT, MICT, and LIT groups, respectively (Fig. S1 A). Of these, 63 metabolites were unique to HIIT, 60 were unique to MICT, and 10 were unique to LIT (Fig. 2 A and B; Table S2 ). In addition, 37 metabolites were consistently up- or down-regulated across all three exercise interventions (Fig. 2 C), including 28 upregulated and 9 downregulated metabolites. The upregulated metabolites included pyruvic acid, lactic acid, malic acid, and fumaric acid, which are annotated to energy- and amino-acid–related pathways. The downregulated metabolites included LPE(18:1), LPE(18:2), and 2-methylbutyric acid, which are annotated to lipid metabolism and branched-chain amino acid metabolism. The HIIT group had 169 upregulated and 35 downregulated metabolites, compared with 148 upregulated and 50 downregulated metabolites in the MICT group and 37 upregulated and 23 downregulated metabolites in the LIT group (Fig. 2 D). Fig. 2. Open in a new tab Metabolite changes, classification, and enrichment analysis under different exercise modes. ( A ) Venn diagram of significantly altered metabolites after acute exercise. The diagram shows the distribution of significant metabolites across the three exercise groups (HIIT, MICT, LIT), including those unique to each group and those shared between groups. (Unit: Number). ( B ) Circular plots showing the classification of significant metabolites that are unique to each exercise group or shared between groups. Numbers indicate the number of significant metabolites in each group or combination, and colors represent different metabolite classes. ( C ) Heatmap of metabolites that are significantly altered across all three exercise interventions. Colors represent the log2 fold change (post vs. baseline), with red indicating increased levels and blue indicating decreased levels. ( D ) Classification and quantification of significant metabolites by biochemical class. Bar plots summarize the numbers of significantly up- and down-regulated metabolites in each category for the HIIT, MICT, and LIT groups. Red bars represent up-regulated metabolites, and blue bars represent down-regulated metabolite. ( E ) Serum metabolomic pathway enrichment analysis. The dot plot depicts significantly enriched metabolic pathways (y-axis) for each exercise group (x-axis: HIIT, MICT, LIT). Dot color represents the median log2 fold change of significant metabolites within each pathway (red, net upregulation; blue, net downregulation), and dot size is proportional to the enrichment significance (− log10 p-value). ( F ) VIP analysis for baseline versus post-exercise comparisons. Each panel shows VIP scores of metabolites for one exercise group (HIIT, MICT, LIT), derived from models comparing baseline and post-exercise samples. Red points indicate metabolites with higher post-exercise levels, and blue points indicate metabolites with lower post-exercise levels. ( G ) VIP analysis for pairwise post-exercise comparisons between exercise groups. Each panel shows VIP scores of metabolites based on post-exercise data only (HIIT vs. MICT, HIIT vs. LIT, MICT vs. LIT). Each point represents a metabolite; red points indicate higher levels in the first group (or lower levels in the second), and blue points indicate lower levels in the first group (or higher levels in the second) Beyond these overall counts, the three exercise intensities exhibited distinct acute metabolite signatures (Fig. S1 B and 2 E; Table S3 and S4 ). Immediately after HIIT, levels of pyruvic acid, lactate, and TCA cycle intermediates (e.g., succinate, fumarate, malic acid, oxoglutaric acid) were markedly increased (Fig. S1 B). Ascorbic acid, beta-alanine, and pantothenic acid (vitamin B5) also showed significant elevations, whereas bile acid metabolites (e.g., glycodeoxycholic acid, glycoursodeoxycholic acid) and certain lipids (e.g., LPE(18:1)) were significantly decreased. MICT similarly increased succinate, pyruvic acid, lactate, and fumarate, and also elevated 1-pyrroline-2-carboxylic acid and hypoxanthine. In contrast, several lysophospholipids, especially lysophosphatidylinositols (LPIs; e.g., LPI(17:0)-H, LPI(16:0)-H, LPI(18:0)) and metabolites such as 5-methoxytryptophan, showed significant decreases (Fig. S1 A). In the LIT group, changes were mainly characterized by elevations in dipeptides (e.g., Leu–Phe, Ile–Phe), lactate, and, to a lesser extent, TCA cycle intermediates such as citrate, fumarate, and malic acid. LIT also significantly reduced citrulline, 4-trimethylammoniobutanal, and several bile acids (e.g., glycochenodeoxycholic acid), together with lysophosphatidylethanolamines (e.g., LPE(20:4), LPE(18:2)) (Fig. S1 B). Metabolites with high Variable Importance in Projection (VIP) values were considered among the key drivers of the model’s discriminatory power. In the HIIT group, energy substrates (succinate, pyruvic acid, lactate) and the antioxidant ascorbic acid exhibited the highest VIP values (Fig. 2 F). In the MICT group, high VIP values were observed for succinate, pyruvic acid, fumarate, and the cofactor pantothenic acid. For LIT, metabolites with high VIP values included lactate, pyruvic acid, and several dipeptides such as Leu–Phe and Ile–Phe. Comparative VIP analysis (Fig. 2 G) showed distinct sets of discriminative metabolites for each pairwise comparison. In the comparison between HIIT and MICT, metabolites with the largest VIP values included TCA cycle intermediates such as malic acid and fumarate. In the comparison between MICT and LIT, succinate, pyruvic acid, the buffering agent β-alanine, and palmitoleic acid were among the main discriminative metabolites. Metabolite correlations with exercise capacity indicators (Lactate, VO 2 max, and HRmax) across exercise intensities A total of 195 metabolites were identified that were significantly correlated with L-lactic acid (FDR < 0.05) (Fig. 3 A; Table S5 ). Among these, L-lactic acid exhibited the strongest positive correlations with pyruvic acid, fumarate, malic acid, succinate, and pantothenic acid. In contrast, significant negative correlations were observed between L-lactic acid and LPI(17:0)-H, 2-methylbutyric acid, L-threonine, LPE(18:2), and LPE(18:1). In addition, correlation analysis revealed distinct clustering patterns among metabolites associated with L-lactic acid (Fig. 3 B). Specifically, metabolites annotated to the TCA cycle, fatty-acid oxidation (e.g., acylcarnitines), and alanine, aspartate, and glutamate metabolism formed a densely connected region with predominantly positive correlations. By contrast, membrane-lipid metabolites such as LPEs and LPIs tended to show negative correlations with this central cluster (Fig. 3 B). Multiple linear regression analyses revealed significant associations between changes in metabolites and changes in lactate within each of the three exercise groups (Table S6 ). In the HIIT group, increases in metabolites related to glycolysis (pyruvic acid), alanine metabolism (L-alanine and β-alanine), and aspartate metabolism (aspartic acid) were positively associated with lactate differences (FDR < 0.05). Associations with alanine metabolism (L-alanine) were also observed in both MICT and LIT groups. In the MICT group, lactate changes were additionally correlated with the glycolytic metabolite pyruvic acid, whereas in the LIT group positive associations were observed with TCA cycle intermediates such as malic acid and fumarate (Table S6 ). Fig. 3. Open in a new tab Correlation and regression analyses of metabolites with L-Lactic acid, and VO 2 max under different exercise modalities. ( A ) L-Lactic Acid Correlation Analysis. Pearson correlation analysis identified the top five metabolites most positively and most negatively correlated with L-Lactic Acid. Yellow represents baseline measurements for the three groups (HIIT, MICT, LIT), red represents post-HIIT measurements, blue represents post-MICT measurements, and green represents post-LIT measurements. ( B ) Spearman Correlation Network Analysis of Metabolites Significantly Associated with L-Lactic Acid. Node size represents betweenness centrality, and edges are color-coded based on the direction of the correlation (positive or negative). ( C ) Regression Analysis of Metabolite Changes with VO 2 max. The figure displays the regression coefficients between metabolite changes and VO 2 max across the three groups. The analysis controlled for age, sex, and BMI. Red indicates metabolites with positive regression coefficients, while blue indicates those with negative regression coefficients We conducted multiple linear regression analyses relating baseline metabolites and exercise-induced changes to VO 2 max and HRmax. In the baseline models, VO 2 max was significantly associated with 26 metabolites and HRmax with 22 metabolites (Fig. S1 C and S1 D). VO 2 max showed positive associations with metabolites such as N-acetylornithine, ergothioneine, and kynurenic acid, and negative correlations with uracil, inosine, and cholesterol sulfate. HRmax was positively associated with metabolites such as 4-hydroxybenzoate and L-glutamic acid, and negatively correlated with 1-(β-D-ribofuranosyl)-1,4-dihydronicotinamide, cholic acid, and dipeptides (e.g., Ile–Ile). Subsequent analysis examined the associations between exercise-induced metabolite changes and physiological parameters across the three training groups (Fig. 3 C and S1 E). Within the HIIT group, VO 2 max showed negative associations with metabolites such as trimethylamine N-oxide and kynurenic acid. In contrast, the MICT group showed positive correlations between VO 2 max and α-tocopherol, glutathione, and maltol, alongside positive associations between HRmax and branched-chain amino acid–related metabolites (e.g., 2-keto valeric acid) and purine metabolism intermediates such as hypoxanthine. For the LIT group, VO 2 max was negatively correlated with metabolites including N-acetyl-L-phenylalanine and quinic acid, whereas HRmax was positively correlated with sugar derivatives (e.g., D-glucuronic acid and gluconic acid) and amino acids such as histidine and L-ornithine. Hierarchical clustering and correlation network analysis of metabolite regulation As metabolites showed distinct pre- to post-exercise responses across intensities, correlation-based hierarchical clustering was performed separately for the HIIT, MICT, and LIT groups using metabolites that changed significantly after exercise (Fig. 4 A–C). Within each training modality, significantly regulated metabolites segregated into two primary clusters: one predominantly composed of metabolites related to central carbon and amino acid metabolism, and the other enriched in lipid- and steroid-related metabolites. Fig. 4. Open in a new tab Hierarchical clustering and metabolite network analysis. ( A - C ) Correlation networks and hierarchical clustering of significantly altered metabolites in the HIIT ( A ), MICT ( B ), and LIT ( C ) groups. Nodes represent metabolites, with node size reflecting the fold change relative to baseline. Edges indicate significant pairwise Spearman correlations. The top 5 hub metabolites with the highest node degree (number of direct connections) are highlighted in each network In the HIIT group (Fig. 4 A), Cluster 1 was defined by metabolites involved in arginine and proline metabolism, arginine biosynthesis, the citrate cycle (TCA cycle), butanoate metabolism, and alanine, aspartate, and glutamate metabolism. Within this cluster, lactate, fumarate, malic acid, pantothenic acid, and the dipeptide Glu–Glu showed the highest node degrees. Cluster 2 was dominated by metabolites assigned to glycerophospholipid metabolism, ether lipid metabolism, and steroid hormone biosynthesis, with highly connected metabolites such as propionylcarnitine, 3-hydroxyisovalerylcarnitine, PC(38:5), PC(36:4), and PC(36:3) forming a dense lipid-related subnetwork. A comparable clustering topology was evident in the MICT group (Fig. 4 B). Cluster 1 similarly encompassed metabolites from arginine and proline metabolism, arginine biosynthesis, the TCA cycle, butanoate metabolism, and alanine, aspartate, and glutamate metabolism. Long-chain fatty acids including palmitic acid, α-linolenic acid (18:3), linoleic acid (18:2), oleic acid, and FA(22:5)-H exhibited the highest node degrees in this cluster. Cluster 2 was characterized by metabolites annotated to arachidonic acid metabolism, steroidogenesis, and oxidation of branched-chain fatty acids, with PC(40:7), propionylcarnitine, myristoylcarnitine, tetradecanoylcarnitine, and 3-hydroxyisovalerylcarnitine acting as the main hubs. In the LIT group (Fig. 4 C), Cluster 1 comprised metabolites linked to pyruvate metabolism, arginine biosynthesis, the TCA cycle, butanoate metabolism, and alanine, aspartate, and glutamate metabolism. Malic acid, pyruvic acid, lactate, fumarate, and succinate showed the highest node degrees in this cluster. Cluster 2 was distinguished by metabolites related to steroid hormone biosynthesis and primary bile acid biosynthesis, with bile acids such as glycochenodeoxycholic acid, glycoursodeoxycholic acid, GUDCA, and glycodeoxycholic acid forming the main hubs. Differential activation of metabolic pathways Six metabolic pathways were significantly enriched on the basis of metabolites exhibiting pre- to post-exercise changes (Fig. 2 E), and the key metabolites within these pathways are mapped in Fig. 5 . TCA cycle intermediates were broadly increased after exercise across all intensities: fumarate, malic acid, succinate, and 2-oxoglutarate were elevated in all three groups (HIIT, MICT, and LIT), whereas citrate increases were confined to the HIIT and LIT groups. In the arginine and proline metabolism pathway, guanidinoacetate, creatine, and 1-pyrroline-2-carboxylate were significantly elevated in the HIIT and MICT groups, with no significant changes observed in the LIT group. For alanine, aspartate, and glutamate metabolism, L-alanine and pyruvic acid were significantly increased across all three exercise groups. In contrast, N-acetyl-L-aspartate, succinate semialdehyde, and L-glutamate increased significantly only in the HIIT and MICT groups, and L-aspartate was significantly elevated exclusively in the HIIT group, whereas glucosamine 6-phosphate was significantly reduced only in the LIT group. In the arginine biosynthesis pathway, N-acetyl-L-glutamate was significantly upregulated in both the HIIT and MICT groups, while downstream intermediates showed decreases: L-ornithine was reduced in the HIIT and MICT groups, and citrulline was significantly reduced in the MICT and LIT groups. Additionally, in the butanoate metabolism pathway, 3-hydroxybutyric acid and 2-hydroxyglutarate were significantly increased in the HIIT and MICT groups, with no significant changes in the LIT group. Within histidine metabolism, urocanic acid increased significantly in the HIIT and LIT groups, whereas anserine was significantly decreased only in the HIIT group. Fig. 5. Open in a new tab Key metabolic pathways. Integrated metabolic map of pathway alterations across exercise intensities. The schematic depicts the TCA cycle (blue) and interconnected pathways: arginine biosynthesis (red); arginine and proline (yellow); alanine, aspartate, and glutamate (green); butanoate (purple); and histidine (brown) metabolism. Embedded bar graphs show metabolite abundance in HIIT, MICT, and LIT groups (mean ± SEM). Statistical significance was determined using linear mixed models (* P < 0.05, ** P < 0.01, *** P < 0.001) Discussion In this randomized controlled trial, we investigated serum metabolomic signatures elicited by a single acute exercise bout after participants were assigned to one of three intensity protocols (HIIT, MICT, or LIT). First, we identified a shared core of 37 metabolites that changed concordantly across all three intensities, with pathway annotation implicating convergent regulation of central energy and lipid-related processes. Specifically, upregulated changes were primarily centered on glycolysis and TCA cycle, accompanied by increases in multiple acylcarnitines and amino acid–related metabolites; conversely, downregulated changes were enriched in membrane lipid and bile acid–related molecules (e.g., LPE(18:1), LPE(18:2), and several conjugated bile acids) and included selected metabolites linked to branched-chain amino acid catabolism (e.g., 2-methylbutyric acid). Second, although all three exercise intensities affected these pathways, the response was clearly intensity dependent, with LIT producing smaller perturbations than HIIT and MICT. Importantly, despite partially overlapping pathway-level changes, the responses to HIIT and MICT may be driven by distinct physiological mechanisms. HIIT was characterized by rapid energy mobilization, evidenced by elevated pyruvate, lactate, and several TCA-cycle intermediates, together with significant changes in amino acid metabolism–related and antioxidant-associated metabolites (e.g., ascorbic acid). In contrast, MICT exhibited a pattern more consistent with sustained energy provision, characterized by increases in glycolytic and TCA-cycle intermediates along with alterations in markers of fatty acid mobilization/oxidation (e.g., acylcarnitines). Although lower in intensity, LIT may contribute to maintaining metabolic stability through modulation of glycolysis, the TCA cycle, and neurotransmitter-related metabolism. Finally, we examined associations between metabolites and indices of exercise capacity. Lactate was positively correlated with TCA-cycle metabolites but inversely correlated with fatty acid oxidation and membrane lipid metabolism. In addition, VO₂max was positively associated with metabolites such as N-acetylornithine, ergothioneine, and kynurenic acid, and negatively associated with uracil, inosine, and cholesterol sulfate. The results in this study indicated that all three exercise modalities showed consistent post-exercise increases in lactate and pyruvate together with coordinated shifts in TCA-cycle intermediates, a pattern broadly consistent with acute mobilization of central carbon metabolism under increased energetic demand [ 12 , 26 ]. Aligning with prior evidence that exercise engages both lipid and amino-acid metabolic pathways to support fuel use and metabolic homeostasis [ 27 , 28 ], we found that metabolites related to whole-body substrate handling, including fatty acid and oxidation related markers and amino acid related metabolites, also changed across intensities. Furthermore, multiple lysophosphatidylethanolamine (LPE) species, including LPE(18:1), LPE(18:2), and LPE(18:3), were consistently decreased across HIIT, MICT, and LIT. Consistent with exercise-metabolomics syntheses, decreases in lysophospholipids (including LPCs and LPEs) and concurrent shifts in bile-acid related species have been repeatedly reported after acute exercise [ 6 ]. Prior studies indicate that acute exercise elicits a shift in substrate utilization, accompanied by lipid mobilization and regulation of mitochondrial fatty acid oxidation (β-oxidation) [ 29 , 30 ]. Consequently, the observed reduction in LPE levels may reflect perturbations in lipid turnover and phospholipid metabolism during the immediate post-exercise period. Collectively, these shared pathway-level changes highlight a common metabolic backbone of the acute exercise response across intensity domains. Although all three exercise interventions significantly altered common metabolic pathways, the metabolic profiles induced by different exercise modalities exhibited distinct intensity-dependence. These differences in exercise intensity are fundamentally reflected in energy metabolism processes and substrate selection, leading to divergent metabolic signatures. The metabolic profile of HIIT is characterized by rapid energy mobilization, evidenced by significant elevations in lactate and pyruvate. These changes suggest marked engagement of glycolysis-related processes and are accompanied by synchronous increases in TCA-cycle intermediates (e.g., fumarate and malate). Notably, HIIT was concurrently associated with significant alterations in amino acid–metabolism–related metabolites (e.g., L-alanine and β-alanine). Integrating previous findings, under the high metabolic demands corresponding to high-intensity loads, amino acid–related metabolic processes may provide anaplerotic inputs to replenish TCA-cycle intermediates, thereby supporting oxidative metabolism [ 31 , 32 ]. Furthermore, the elevation of ascorbate levels following HIIT suggests an acute response in redox/antioxidant-related metabolism; however, its causal relationship with oxidative-stress burden requires further verification combined with oxidative-stress markers and antioxidant enzyme activities [ 33 ]. In contrast to the heavy reliance on carbohydrate metabolism seen in HIIT, MICT more favorably supports a balanced energy supply from both fatty acids and carbohydrates. MICT emphasizes the synergistic utilization of multiple energy sources (e.g., glycolysis, TCA cycle, and fatty acid metabolism), accompanied by a relatively balanced amino acid metabolism. Consistent with this, Van Loon et al. [ 12 ] found that lipid oxidation peaks at moderate exercise intensities (approximately 55% Wmax) and declines at higher intensities, supporting the role of MICT in optimizing lipid energy metabolism. Repeated exposure to such metabolic conditions may drive long-term adaptations, thereby enhancing lipid oxidation efficiency. Previous studies have indicated that MICT can effectively augment lipid oxidation capacity [ 34 , 35 ] and yield significant metabolic benefits through its sustained impact on lipid oxidative pathways [ 36 ]. Our results are consistent with existing research, further reinforcing the classic metabolic adaptation model of MICT. Furthermore, LIT primarily maintained energy supply through modest shifts in glycolysis- and TCA-cycle–related metabolites. Beyond these energy-related changes, LIT exhibited a signaling-oriented metabolomic profile characterized by coordinated alterations in lipids, steroids, and taurine derivatives. Notably, several identified molecules have been discussed in prior literature in neuroactive or neuroendocrine contexts. For example, N-acyl taurines have been linked to TRP-family ion channel modulation [ 32 , 37 ], and oleamide has been discussed in relation to arousal/sleep regulation [ 32 , 38 ]. Steroid-related changes, including progesterone-associated species, have also been reported within neuroendocrine/neurosteroid signalling frameworks [ 39 ]. However, because we measured circulating rather than central metabolite levels, we interpret these observations cautiously. Rather than inferring direct central modulation, we propose that LIT induces a peripheral metabolic signature compatible with neuroendocrine and neurotransmission-related signalling frameworks described in prior literature. These findings are hypothesis-generating and warrant future targeted and mechanistic studies. Beyond these intensity-specific patterns, our correlation and network analyses highlight lactate as a central integrator of the acute metabolic response. Across all three exercise modalities, lactate showed strong positive associations with pyruvate, multiple TCA-cycle intermediates (e.g., fumarate, malate, succinate, 2-oxoglutarate), cofactor-related metabolites such as pantothenic acid, and fatty acid oxidation related markers including selected acylcarnitines. Pantothenic acid plays an active role in fatty-acid metabolism and facilitates acetyl-CoA entry into the TCA cycle. Thus, its positive coupling with lactate and TCA intermediates may reflect increased cofactor/oxidative demand to sustain high oxidative throughput during the immediate post-exercise period [ 32 ]. In our network, lactate emerged as a high-degree hub within the central carbon and amino acid module. Consistent with the paradigm shift from viewing lactate as a “dead-end waste product of glycolysis” to the lactate-shuttle, this architecture is compatible with lactate acting as an integrative metabolite rather than merely a glycolytic byproduct [ 40 ]. This interpretation is consistent with the lactate-shuttle framework [ 41 ], in which lactate produced by glycolytic fibres or other producer cells is oxidized by more oxidative cell populations or mitochondrial compartments as an important oxidative substrate and signalling molecule linking cytosolic glycolysis to oxidative phosphorylation and gluconeogenesis [ 42 ]. By contrast, metabolites related to membrane lipid metabolism, particularly LPEs and LPIs, clustered on the opposite side of the lactate-TCA-acylcarnitine module and exhibited predominantly negative correlations with lactate. Similar decreases in lysophospholipids (including LPCs and LPEs) after exercise have been summarized in prior metabolomics reviews [ 6 ]. Rather than simply indicating suppression of lipid metabolism, this inverse pattern might reflect dynamic membrane remodelling to support transporter activity under conditions of high metabolic demand and substrate flux [ 43 ]. We also found that higher VO₂max was associated with amino acid derived and redox-related metabolites (e.g., N-acetylornithine, ergothioneine, and kynurenic acid), whereas HRmax showed a distinct pattern involving purine, bile acid, and dipeptide species. This separation suggests that individuals with greater aerobic fitness are characterized by a metabolomic profile compatible with more efficient oxidative metabolism and antioxidant defence [ 44 , 45 ], in line with recent work identifying ergothioneine as a diet-derived mitochondrial antioxidant that enhances exercise performance [ 46 ], as well as N-acyl amino acids as candidate mediators of cardiorespiratory fitness [ 47 ]. Building on these baseline associations, the correlations between exercise-induced metabolite changes and physiological adaptation indicators further delineated putative molecular features of intensity-specific responses. Notably, improvements in the HIIT condition were inversely associated with markers of metabolic stress, including kynurenic acid and trimethylamine N-oxide (TMAO). Prior evidence suggests that accumulation of TMAO can promote oxidative stress and impair mitochondrial function [ 48 ], and that excessive activation of the kynurenine pathway is closely linked to stress-related metabolic responses [ 49 ]. Collectively, these observations suggest that under high-intensity loading, the accumulation of certain metabolic byproducts—or potential saturation of their clearance pathways—may constrain the magnitude of physiological improvement. In contrast, MICT was associated with coordinated changes in long-chain acylcarnitines and antioxidant-related metabolites (e.g., α-tocopherol and glutathione). Given that shifts in acylcarnitine profiles are sensitive indicators of skeletal-muscle mitochondrial fatty-acid oxidation flux and fuel selection [ 50 ], and that the integrity of antioxidant defenses is essential for maintaining exercise redox homeostasis [ 33 ], these findings highlight that MICT may be more conducive to sustaining mitochondrial oxidative flux. This study presents several limitations. First, the participants were healthy, inactive young individuals, which restricts the generalizability of our findings to older adults or populations with medical conditions. Second, we focused exclusively on serum metabolomic changes in response to acute exercise. Integrating metabolomics with other omics layers, such as targeted metabolomics, proteomics and transcriptomics, would enable a more comprehensive characterization of the molecular response to different exercise intensities and of the interactions among distinct biological pathways. Third, although the present trial used a randomized parallel-group design, the analyses of metabolite-phenotype relationships are correlative and based on an acute response to exercise within a discovery-oriented untargeted LC-MS framework. Future investigations utilizing targeted metabolomics and multi-cohort designs are required to formally verify these patterns. Despite these limitations, the study also has several strengths. First, we conducted a randomized controlled trial to compare acute serum metabolomic responses across clearly defined light-, moderate-, and high-intensity exercise domains. These domains reflect common real-world exercise prescriptions for healthy adults and together provide a comprehensive characterization of intensity-dependent metabolic remodeling. This design also responds to recent systematic reviews in exercise and sports metabolomics that have called for well-controlled study designs to improve interpretability and comparability across studies [ 9 ]. Moreover, we discovered that LIT is not simply a “scaled-down” version of MICT and HIIT, it was characterized by the unique modulation of bile acids (e.g., glycochenodeoxycholic acid) and dipeptide. This finding redefines that LIT may exert health benefits through endocrine signaling rather than caloric expenditure. Second, our correlation network analysis identified L-lactic acid as a central connectivity hub showing strong positive correlations with TCA cycle intermediates (fumarate, succinate) and acylcarnitines (fatty acid oxidation markers) across all intensities. This provides empirical metabolomics validation for the “dual-fuel” model. Third, we related both baseline and exercise-induced changes in metabolic profiles to indicators of exercise capacity, including maximal oxygen uptake, lactate responses, and maximal heart rate. This integrative analysis helps to link intensity-specific metabolic programs to cardiorespiratory fitness and to the balance between glycolytic flux and mitochondrial energy pathways, such as fatty acid oxidation and amino acid metabolism, under different exercise conditions. Collectively, these insights lay a foundation for future work and may contribute to the refinement of more precise, intensity-tailored exercise strategies. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 (75.2MB, tif) Supplementary Material 2 (27.6KB, xlsx) Supplementary Material 3 (62.1KB, docx) Acknowledgements We thank the participants, project staff, and the supports from Tsinghua University Physical Fitness and Rehabilitation Center. We also thank the staff members at Metabolomics and Lipidomics Center at Tsinghua - National Protein Science Facility (Beijing) for providing facility support. Abbreviations HIIT High-intensity interval training MICT Moderate-intensity continuous training LIT Light-intensity training RCT Randomized controlled trial FDR False Discovery Rate BH Benjamini-Hochberg PCA Principal component analysis CV Coefficient of variation PLS-DA Partial Least Squares Discriminant Analysis VIP Variable importance in projection VO 2 max Maximal oxygen uptake PC Phosphatidylcholine LPE Lysophosphatidylethanolamine TCA cycle Citric acid cycle HRmax Maximum heart rate Author contributions Jianxiu Liu: Conceptualization, Data curation, Formal analysis, Project administration, Visualization, Writing – original draft, Writing review & editing. Junxing Zhang: Data curation, Formal analysis, Project administration, Writing – review & editing. Xintian Li: Data curation, Project administration, Writing – review & editing. Baile Wu: Data curation, Formal analysis, Visualization, Writing – review & editing. Alimjan ablitip, Dizhi Wang, & Limei Ke: Writing – review & editing. Qian Di: Conceptualization, Writing – review & editing, Funding acquisition, Supervision Ruidong Liu: Conceptualization, Writing – review & editing, Funding acquisition, Supervision. Xindong Ma: Conceptualization, Funding acquisition, Supervision. Funding We acknowledge the research support from the National Natural Science Foundation of China (No. 42277419), the National Key Research and Development Program of China (No. 2024YFC3607002), the National Social Science Fund of China (No. 25CTY033), the Research Plan for Higher Education Science of the Chinese Higher Education Association (No. 23TY0205), and the Research Fund of Vanke School of Public Health, Tsinghua University. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Data and code availability The study did not generate any unique code. Materials availability The study did not generate new unique reagents. Resource availability Lead contact. Requests for further information, resources, and reagents should be directed to and will be fulfilled by the lead contact, Jianxiu Liu ([email protected]). Declarations Ethics approval and consent to participate All participants provided written informed consent following a comprehensive explanation of the study’s procedures. This study received approval from the Ethics Committee of the School of Medicine at Tsinghua University (IRB 20190091). Data sharing This data is available on Metabolomics Workbench, Study ID 6897. Competing interests All other authors declare they have no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Jianxiu Liu, Junxin Zhang and Xingtian Li contributed equally to this work. Contributor Information Qian Di, Email: [email protected]. Ruidong Liu, Email: [email protected]. Xindong Ma, Email: [email protected], Email: [email protected]. References 1. Thyfault JP, Bergouignan A. Exercise and metabolic health: beyond skeletal muscle. Diabetologia. 2020;63(8):1464–74. 10.1007/s00125-020-05177-6. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 2. Nieman DC, Lila MA, Gillitt ND. Immunometabolism: a multi-omics approach to interpreting the influence of exercise and diet on the immune system. Annual Rev food Sci technology. 2019 Mar;25:10:341–63. 10.1146/annurev-food-032818-121316. [ DOI ] [ PubMed ] [ Google Scholar ] 3. McGee SL, Hargreaves M. Exercise adaptations: molecular mechanisms and potential targets for therapeutic benefit. Nat reviews Endocrinol. 2020;16(9):495–505. 10.1038/s41574-020-0377-1. [ DOI ] [ PubMed ] [ Google Scholar ] 4. Gibala MJ, Little JP, MacDonald MJ, Hawley JA. Physiological adaptations to low-volume, high-intensity interval training in health and disease. J Physiol. 2012;590(5):1077–84. 10.1113/jphysiol.2011.224725. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 5. Wishart DS, Li C, Marcu A, Badran H, Pon A, Budinski Z, et al. PathBank: a comprehensive pathway database for model organisms. Nucleic Acids Res. 2020;48(D1):D470–8. 10.1093/nar/gkz861. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 6. Sakaguchi CA, Nieman DC, Signini EF, Abreu RM, Catai AM. Metabolomics-based studies assessing exercise-induced alterations of the human metabolome: a systematic review. Metabolites. 2019 Aug;9(8). 10.3390/metabo9080164. [ DOI ] [ PMC free article ] [ PubMed ] 7. Babu M, Snyder MJM, Proteomics C. Multi-omics profiling for health. 2023;22(6). 10.1016/j.mcpro.2023.100561. [ DOI ] [ PMC free article ] [ PubMed ] 8. Fiehn O. Metabolomics–the link between genotypes and phenotypes. Plant Mol Biol. 2002;48(1–2):155–71. 10.1023/A:1013713905833. [ PubMed ] [ Google Scholar ] 9. Khoramipour K, Sandbakk Ø, Keshteli AH, Gaeini AA, Wishart DS, Chamari K. Metabolomics in exercise and sports: a systematic review. Sports Medicine (Auckland, NZ). 2022 Mar;52(3):547–83. 10.1007/s40279-021-01582-y. [ DOI ] [ PubMed ] 10. Ross R, Janssen I, Tremblay MSJJS, Science H. Public health importance of light intensity physical activity. 2024;13(5):674. 10.1016/j.jshs.2024.01.010. [ DOI ] [ PMC free article ] [ PubMed ] 11. Torma F, Gombos Z, Jokai M, Takeda M, Mimura T, Radak ZJSM, et al. High intensity interval training and molecular adaptive response of skeletal muscle. 2019;1(1):24–32. 10.1016/j.smhs.2019.08.003. [ DOI ] [ PMC free article ] [ PubMed ] 12. van Loon LJ, Greenhaff PL, Constantin-Teodosiu D, Saris WH, Wagenmakers AJ. The effects of increasing exercise intensity on muscle fuel utilisation in humans. J Physiol. 2001;536(Pt 1):295–304. 10.1111/j.1469-7793.2001.00295.x. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 13. Chastin SFM, De Craemer M, De Cocker K, Powell L, Van Cauwenberg J, Dall P, et al. How does light-intensity physical activity associate with adult cardiometabolic health and mortality? Systematic review with meta-analysis of experimental and observational studies. Br J Sports Med. 2019;53(6):370–6. 10.1136/bjsports-2017-097563. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 14. Schranner D, Kastenmüller G, Schönfelder M, Römisch-Margl W, Wackerhage H. Metabolite concentration changes in humans after a bout of exercise: a systematic review of exercise metabolomics studies. Sports Med - Open. 2020/02/10. 2020;6(1):11. 10.1186/s40798-020-0238-4. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 15. Cohen J. Statistical power analysis for the behavioral sciences. Routledge; 2013. 16. Faul F, Erdfelder E, Lang A-G, Buchner, AJBrm. G* Power 3: A flexible statistical power analysis program for the social. Behav biomedical Sci. 2007;39(2):175–91. [ DOI ] [ PubMed ] [ Google Scholar ] 17. Morville T, Sahl RE, Moritz T, Helge JW, Clemmensen C. Plasma metabolome profiling of resistance exercise and endurance exercise in humans. Cell Rep Dec. 2020;29(13):108554. 10.1016/j.celrep.2020.108554. [ DOI ] [ PubMed ] [ Google Scholar ] 18. Snell PG, Stray-Gundersen J, Levine BD, Hawkins MN, Raven PB. Maximal oxygen uptake as a parametric measure of cardiorespiratory capacity. Med Sci Sports Exerc. 2007;39(1):103–7. 10.1249/01.mss.0000241641.75101.64. [ DOI ] [ PubMed ] [ Google Scholar ] 19. Hanson NJ, Scheadler CM, Lee TL, Neuenfeldt NC, Michael TJ, Miller MG. Modality determines VO2max achieved in self-paced exercise tests: validation with the Bruce protocol. Eur J Appl Physiol. 2016;116(7):1313–9. 10.1007/s00421-016-3384-0. [ DOI ] [ PubMed ] [ Google Scholar ] 20. Nieman D, Austin M, Benezra L, Pearce S, McInnis T, Unick J, et al. Validation of Cosmed’s FitMate™ in measuring oxygen consumption and estimating resting metabolic rate. Res Sports Med. 2006;14(2):89–96. 10.1080/15438620600651512. [ DOI ] [ PubMed ] [ Google Scholar ] 21. Liu J, You Y, Liu R, Shen L, Wang D, Li X, et al. The joint effect and hemodynamic mechanism of PA and PM2.5 exposure on cognitive function: a randomized controlled trial study. J Hazard Mater. 2023;460. 10.1016/j.jhazmat.2023.132415. N.PAG-N.PAG. [ DOI ] [ PubMed ] 22. Poon ET-C, Siu PM-F, Wongpipit W, Gibala M, Wong SH-S. Alternating high-intensity interval training and continuous training is efficacious in improving cardiometabolic health in obese middle-aged men. J Exerc Sci Fit. 2022;20(1):40–7. 10.1016/j.jesf.2021.11.003. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 23. Racil G, Ben Ounis O, Hammouda O, Kallel A, Zouhal H, Chamari K, et al. Effects of high vs. moderate exercise intensity during interval training on lipids and adiponectin levels in obese young females. Eur J Appl Physiol. 2013;113(10):2531–40. 10.1007/s00421-013-2689-5. [ DOI ] [ PubMed ] [ Google Scholar ] 24. Tang H, Wang X, Xu L, Ran X, Li X, Chen L, et al. Establishment of local searching methods for orbitrap-based high throughput metabolomics analysis. 2016;156:163–71. 10.1016/j.talanta. [ DOI ] [ PubMed ] 25. Contrepois K, Wu S, Moneghetti KJ, Hornburg D, Ahadi S, Tsai M-S, et al. Mol choreography acute Exerc. 2020;181(5):1112–30. e16. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 26. Liu X, Cooper DE, Cluntun AA, Warmoes MO, Zhao S, Reid MA, et al. Acetate production from glucose and coupling to mitochondrial metabolism in mammals. Cell. 2018;175(2):502. 10.1016/j.cell.2018.08.040. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 27. Egan B, Zierath JR. Exercise metabolism and the molecular regulation of skeletal muscle adaptation. Cell Metabol. 2013;17(2):162–84. 10.1016/j.cmet.2012.12.012. [ DOI ] [ PubMed ] [ Google Scholar ] 28. Gleeson M. Dosing and efficacy of glutamine supplementation in human exercise and sport training. J Nutr. 2008;138(10):S2045–9. 10.1093/jn/138.10.2045S. [ DOI ] [ PubMed ] [ Google Scholar ] 29. Romijn JA, Coyle EF, Sidossis LS, Gastaldelli A, Horowitz JF, Endert E, et al. Regulation of endogenous fat and carbohydrate metabolism in relation to exercise intensity and duration. Am J Physiol. 1993;265(3 Pt 1):E380–91. 10.1152/ajpendo.1993.265.3.E380. [ DOI ] [ PubMed ] [ Google Scholar ] 30. Hargreaves M, Spriet LL. Skeletal muscle energy metabolism during exercise. Nat Metab Sep. 2020;2(9):817–28. 10.1038/s42255-020-0251-4. [ DOI ] [ PubMed ] [ Google Scholar ] 31. Gibala MJ, MacLean DA, Graham TE, Saltin B. Tricarboxylic acid cycle intermediate pool size and estimated cycle flux in human muscle during exercise. Am J Physiol. 1998 Aug;275(2):E235–42. 10.1152/ajpendo.1998.275.2.E235. [ DOI ] [ PubMed ] 32. Lewis GD, Farrell L, Wood MJ, Martinovic M, Arany Z, Rowe GC, et al. Metabolic signatures of exercise in human plasma. Sci Transl Med. May 2010;26(33):33ra7. 10.1126/scitranslmed.3001006. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 33. Powers SK, Deminice R, Ozdemir M, Yoshihara T, Bomkamp MP, Hyatt H. Exercise-induced oxidative stress: Friend or foe? J Sport Health Sci Sep. 2020;9(5):415–25. 10.1016/j.jshs.2020.04.001. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 34. Stander Z, Luies L, Mienie LJ, Keane KM, Howatson G, Clifford T, et al. The altered human serum metabolome induced by a marathon. Metabolomics: Official J Metabolomic Soc. 2018;14(11):150. 10.1007/s11306-018-1447-4. [ DOI ] [ PubMed ] [ Google Scholar ] 35. Tabone M, Bressa C, García-Merino JA, Moreno-Pérez D, Van EC, Castelli FA, et al. The effect of acute moderate-intensity exercise on the serum and fecal metabolomes and the gut microbiota of cross-country endurance athletes. Sci Rep. 2021;11(1):1–12. 10.1038/s41598-021-82947-1. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 36. Zhang H, Liu J, Cui M, Chai H, Chen L, Zhang T, et al. Moderate-intensity continuous training has time-specific effects on the lipid metabolism of adolescents. J translational Intern Med. 2023;11(1):57–69. 10.2478/jtim-2022-0050. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 37. Saghatelian A, McKinney MK, Bandell M, Patapoutian A, Cravatt BF. A FAAH-regulated class of N-acyl taurines that activates TRP ion channels. Biochemistry. 2006 Aug 1;45(30):9007-15. 10.1021/bi0608008. [ DOI ] [ PubMed ] 38. Cravatt BF, Prospero-Garcia O, Siuzdak G, Gilula NB, Henriksen SJ, Boger DL, et al. Chemical characterization of a family of brain lipids that induce sleep. Science Jun. 1995;9(5216):1506–9. 10.1126/science.7770779. [ DOI ] [ PubMed ] [ Google Scholar ] 39. Wang M. Neurosteroids and GABA-A receptor function. Front Endocrinol (Lausanne). 2011;2:44. 10.3389/fendo.2011.00044. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 40. Gladden LB. Lactate metabolism: a new paradigm for the third millennium. J Physiol. 2004;558(1):5–30. 10.1113/jphysiol.2003.058701. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 41. Brooks GA. The science and translation of lactate shuttle theory. Cell Metabol. 2018;27(4):757–85. 10.1016/j.cmet.2018.03.008. [ DOI ] [ PubMed ] [ Google Scholar ] 42. Brooks GA. Lactate as a fulcrum of metabolism. Redox Biol. 2020;35:101454. 10.1016/j.redox.2020.101454. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 43. Newsom SA, Brozinick JT, Kiseljak-Vassiliades K, Strauss AN, Bacon SD, Kerege AA, et al. Skeletal muscle phosphatidylcholine and phosphatidylethanolamine are related to insulin sensitivity and respond to acute exercise in humans. J Appl Physiol (1985) Jun. 2016;1(11):1355–63. 10.1152/japplphysiol.00664.2015. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] 44. Peake JM, Tan SJ, Markworth JF, Broadbent JA, Skinner TL, Cameron-Smith D. Metabolic and hormonal responses to isoenergetic high-intensity interval exercise and continuous moderate-intensity exercise. Am J Physiol Endocrinol metabolism. 2014;307(7):E539–52. 10.1152/ajpendo.00276.2014. [ DOI ] [ PubMed ] [ Google Scholar ] 45. Wu L, Wang J, Cao X, Tian Y, Li JJSR. Effect of acute high-intensity exercise on myocardium metabolic profiles in rat and human study via metabolomics approach. 2022;12(1):6791. 10.1038/s41598-022-10976-5. [ DOI ] [ PMC free article ] [ PubMed ] 46. Sprenger HG, Mittenbühler MJ, Sun Y, Van Vranken JG, Schindler S, Jayaraj A, et al. Ergothioneine boosts mitochondrial respiration and exercise performance via direct activation of MPST. bioRxiv. 2024 Apr 10. 10.1101/2024.04.10.588849. 47. Robbins JM, Benson M, Verkerke ARP, Tiwari G, Deng S, Rao P, et al. N-Palmitoyl glutamine is a candidate mediator of cardiorespiratory fitness. Circulation. 2025 Nov 12. 10.1161/circulationaha.125.074187. [ DOI ] [ PMC free article ] [ PubMed ] 48. Caradonna E, Abate F, Schiano E, Paparella F, Ferrara F, Vanoli E, et al. Trimethylamine-N-Oxide (TMAO) as a rising-star metabolite: implications for human health. Metabolites. 2025 Mar;24(4). 10.3390/metabo15040220. [ DOI ] [ PMC free article ] [ PubMed ] 49. Agudelo LZ, Femenía T, Orhan F, Porsmyr-Palmertz M, Goiny M, Martinez-Redondo V, et al. Skeletal muscle PGC-1α1 modulates kynurenine metabolism and mediates resilience to stress-induced depression. Cell Sep. 2014;25(1):33–45. 10.1016/j.cell.2014.07.051. [ DOI ] [ PubMed ] [ Google Scholar ] 50. Overmyer KA, Evans CR, Qi NR, Minogue CE, Carson JJ, Chermside-Scabbo CJ, et al. Maximal oxidative capacity during exercise is associated with skeletal muscle fuel selection and dynamic changes in mitochondrial protein acetylation. Cell Metab Mar. 2015;3(3):468–78. 10.1016/j.cmet.2015.02.007. [ DOI ] [ PMC free article ] [ PubMed ] [ Google Scholar ] Associated Data This section collects any data citations, data availability statements, or supplementary materials included in this article. Supplementary Materials Supplementary Material 1 (75.2MB, tif) Supplementary Material 2 (27.6KB, xlsx) Supplementary Material 3 (62.1KB, docx) Data Availability Statement The study did not generate any unique code. The study did not generate new unique reagents. Lead contact. Requests for further information, resources, and reagents should be directed to and will be fulfilled by the lead contact, Jianxiu Liu ([email protected]). Articles from Journal of Translational Medicine are provided here courtesy of BMC ACTIONS View on publisher site PDF (8.4 MB) Cite Collections Permalink PERMALINK Copy RESOURCES Similar articles Cited by other articles Links to NCBI Databases Cite Copy Download .nbib .nbib Format: AMA APA MLA NLM Add to Collections Create a new collection Add to an existing collection Name your collection * Choose a collection Unable to load your collection due to an error Please try again Add Cancel Follow NCBI NCBI on X (formerly known as Twitter) NCBI on Facebook NCBI on LinkedIn NCBI on GitHub NCBI RSS feed Connect with NLM NLM on X (formerly known as Twitter) NLM on Facebook NLM on YouTube National Library of Medicine 8600 Rockville Pike Bethesda, MD 20894 Web Policies FOIA HHS Vulnerability Disclosure Help Accessibility Careers NLM NIH HHS USA.gov Back to Top