Dynamic Metabolic Changes Driven by Exercise Intensity in Acute Swimming

ABSTRACT

Purpose

Utilizing wearable devices and high-throughput omics technologies, the study aimed to establish an intensity-specific human acute swimming exercise model and draw a swimming metabolome atlas to clarify the commonalities and specificities in a panoramic dynamic physiological response under different intensities.

Methods

Forty-two healthy nonathlete young adults were randomly divided into moderate-intensity continuous training (MICT) and high-intensity interval training (HIIT) acute swimming groups. Blood samples were collected at baseline, 0, 15 and 30 min after swimming for metabolomics and targeted lipidomics testing.

Results

1) Baseline results showed high homogeneity among participants. There were no statistical differences in total exercise duration or energy expenditure between groups (P > 0.05), but HIIT showed significantly higher average speeds and maximum heart rates than MICT (P < 0.001). 2) Two hundred nine metabolites (56.5%) following swimming exercise significantly changed (P < 0.05), whereas five metabolites were identified by their significant intensity-dependent characteristics, and Pearson correlation analysis revealed the strongest correlation (r = 0.917) between levels of N-acetylvaline and lactic acid. 3) Combining longitudinal sampling with omics temporal dynamic analysis, acute swimming atlas showed MICT and HIIT significantly upregulated the tricarboxylic acid cycle and presented commonalities and intensity differences in amino acids metabolism related to the emotional regulation pathway and lipid metabolism related to fat hydrolysis and oxidative utilization.

Conclusions

The study established an intensity-specific human acute swimming model. N-Acetylvaline may be a potential novel intensity-specific exercise biomarker. The acute swimming metabolome atlas reveals substrate utilization and recovery characteristics driven by intensities, especially for emotional regulation pathway and lipid metabolism.

Swimming is a highly adherent exercise program with wide-ranging health benefits, resulting from adaptation to cumulative acute exercise-induced stresses (1–4). The biological effects caused by acute exercise are usually determined jointly by the exercise volume and intensity. The exercise intensity represents the depth of the stimulation of the body by exercise and is the core element for establishing exercise prescriptions and in the relationship of exercise health dose-response (5–7). Swimming always incorporates both moderate-intensity continuous training (MICT) and high-intensity interval training (HIIT) (4). Buchheit et al. (8,9) confirmed that, unlike the exercise modalities such as running and cycling, in-water HIIT may require a specific programming approach.

Advancements in high-throughput omics technologies have enabled diverse exercise modalities to break free from traditional measurement constraints limited to measuring only a few parameters (10). Acute exercise metabolomics studies have commenced, generating numerous insights and biomarkers. For instance, Contrepois’ longitudinal multiomics study with incremental treadmill exercise to exhaustion decoded underlying molecular mechanisms and identified potential blood biomarkers related to resting peak oxygen uptake (11). Similarly, Morville’s randomized crossover design involving cycling and resistance exercises distinguished metabolome features across modes via untargeted metabolomics (12). A 2020 review on acute exercise metabolomics affirmed that even within endurance exercise, load intensity affects metabolite concentrations, yet intramode intensity research is scarce (13). These discoveries accentuate the potential of high-throughput metabolomics in uncovering intensity-driven metabolic changes. However, because of limitations in monitoring in-water exercise intensity and sampling, extant swimming studies predominantly concentrate on assessing the long-term effects on specific macroscopic parameters or physiological markers, such as maximal oxygen uptake, body fat percentage, and cardiovascular function or indicators (14,15). Consequently, comprehensive studies on the acute physiological changes in humans after swimming exercise are notably lacking.

The study aimed to conduct a panoramic dynamic physiological response study of acute swimming on 42 healthy, swimming-skilled, nonathlete young adults by using wearable devices and high-throughput omics technologies. The main purpose of the present study was 1) to establish an intensity-specific human acute swimming exercise model; 2) to screen for exercise health biomarkers with trainable potential and intensity dependence; and 3) to draw a temporal map or atlas of swimming metabolism and clarify the commonalities and specificities of substrate utilization and recovery characteristics under different intensities, providing molecular navigation for potential health effects and mechanism studies of swimming exercises.

METHODS

Participants

The inclusion criteria of participants were as follows: age between 18 and 35 years, possessing swimming skills, and capable of independently completing 50m of breaststroke in a swimming pool. Exclusion criteria were as follows: professional sports training experience or concurrent participation in other experimental interventions; cardiovascular, metabolic, respiratory, or musculoskeletal disorders; recent use of medications or dietary supplements within the past month; BMI <18.5 kg·m⁻², and swimming a 50-m distance within 50 s. All participants completed a questionnaire survey before inclusion, covering basic information, the Physical Activity Readiness Questionnaire for Everyone, the International Physical Activity Questionnaire, and a cardiovascular disease risk factors screening. Then they underwent a physical morphology test, with body fat percentage measured using a dual-energy x-ray absorptiometry scanner (GE Lunar iDXA), and swimming ability was assessed with a 50-m sprint swim screening test to ensure the safety and feasibility of exercise testing.

Ultimately, 42 participants were enrolled in the research and randomly categorized into two swimming groups by the lottery method: the MICT group and the HIIT group (Fig. 1A). This study was conducted with approval from the Ethics Committee for Sports Science Research at Beijing Sport University (BSU-IRB) under protocol number 2022188H. Participants voluntarily enrolled and provided informed consent. All procedures were in accordance with the latest version of the Declaration of Helsinki. Descriptive characteristics of participants are included in Table 1.

FIGURE 1.

Study design and preliminary analysis of metabolomics. (A) Schematic of the study design and analysis program. LC-MS/MS, liquid chromatography-tandem mass spectrometry. (B) PCA based on the baseline level of the metabolomic dataset for all measured metabolites. The metabolites in MICT and HIIT were indicated by blue and red circles, respectively, and the first principal component (PC1) and the second principal component (PC2) were used to visualize the data distribution. (C) PCA based on postexercise levels of all measured metabolites. In the PCA of the metabolomics data, the first principal component (PC1) and the second principal component (PC2) were used to visualize the data distribution. (D) PCA was conducted on pre- and postexercise metabolites for MICT. The results were differentiated by color for baseline (MB), 0 min (M0), 15 min (M15), and 30 min (M30) postexercise. In the PCA of the metabolomics data, PC1 and PC2 were used to visualize the data distribution. (E) PCA was conducted on pre- and postexercise metabolites for HIIT. The results were differentiated by color for baseline (HB), 0 min (H0), 15 min (H15), and 30 min (H30) postexercise. In the PCA of the metabolomics data, PC1 and PC2 were used to visualize the data distribution. (F) Venn diagram showing the number of metabolites significantly affected by swimming. (G) Venn diagrams showing the number of upregulated (top) and downregulated (bottom) metabolites after swimming exercise of different intensities.

TABLE 1.

Baseline demographics of participants.

	MICT	HIIT	P
N	21	21	–
Sex (male; female)	10; 11	10; 11	–
Basic information
Age, years	21.0 (20.0, 25.0)	22.0 (20.0, 24.0)	0.732
Height, cm	168.0 (8.7)	169.4 (10.2)	0.489
Weight, kg	60.2 (55.7, 72.3)	66.6 (53.9, 75.8)	0.624
BMI, kg·m−2	22.2 (20.4, 24.1)	22.6 (20.4, 24.4)	0.782
Body fat, %	27.0 (5.7)	26.9 (6.6)	0.734
Waist-to-hip ratio	0.79 (0.06)	0.80 (0.08)	0.494
Resting heart rate, bpm	67.0 (58.5, 71.5)	61.0 (57.0, 68.5)	0.194
METs, MET-min·wk−1	898 (435, 1824)	989 (680, 2290)	0.649
Screening test
Duration of 50-m sprint swimming, s	71.0 (70.0–92.0)	71.0 (61.0–80.0)	0.130
Heart rate of 50-m sprint swimming, bpm	162.2 (14.0)	163.7 (15.8)	0.860
RPE of 50-m sprint swimming (Borg 6 to 20)	15.0 (14.0–17.0)	16.0 (15.0–17.0)	0.287

Data were mean (SD) for normally distributed, and median (P25, P75) for not normally distributed; P values for intergroup comparison were computed using nonparametric tests (Mann–Whitney U test).

Study design

To familiarize participants with the test commands and minimize the impact of environmental stimuli factors such as water temperature and swimming skills, both groups underwent 1 wk of swimming adaptation training, consisting of two sessions with at least 1 d of rest in between. On the third morning following the final adaptation training session, participants performed a single formal swimming test while in a fasted state. During the swimming exercise and testing period, participants maintained their regular diets and refrained from additional moderate- to high-intensity physical activities. Fasting began after 8:00 pm the evening before the formal test, ensuring a fasting period of 10 to 12 h overnight. The swimming exercises were conducted in a 50-m standard indoor swimming pool, where swimming professionals and medical personnel were present.

Swimming exercise protocol

After standardized 10-min warm-up, the MICT group swam for 30 min at a moderate intensity (heart rate range of 70% to 80% HR_peak), with the last 50 m completed just before exiting the pool. The HIIT group engaged in a 30-min swim consisting of sprinting 50 m followed by a 2-min passive recovery period, represented as (50-m sprint × 2-min passive recovery). After completing the final 50-m sprint, they exited the pool. During both MICT and HIIT swimming sessions, the BHT-TEAM (BoHaoTong, China) head-worn wearable telemetry heart rate system was employed for continuous heart rate monitoring and energy expenditure calculation. A professional swimming coach utilized a stopwatch and a tablet for real-time oversight. For HIIT’s single 50-m sprints, the 50-m sprint performance from the screening test served as the benchmarks for sprint quality. The Borg 6 to 20 scale was promptly applied to record the rate of perceived exertion (RPE) after swimming. The BHT-TEAM system calculates energy expenditure using motion acceleration and heart rate signals, capable of recording energy expenditure throughout the exercise, including HIIT intervals (16–18) (Fig. 1A).

Blood collection and sample preparation

Fasting morning elbow venous blood samples were collected at baseline. Additional samples were collected from both groups immediately after exiting the pool (within 2 min) and during the recovery periods at 15 and 30 min postexercise. Blood samples were collected into red-top BD vacutainers without anticoagulants. After standing at room temperature for 30 to 60 min, the samples were centrifuged at 1200 g for 10 min in a refrigerated centrifuge at 4 °C. The separated serum samples were then transferred to labeled 1.5-mL centrifuge tubes and stored at −80 °C until analysis.

Serum precise metabolomics and targeted lipidomics analysis

Untargeted metabolomics (19–21) and targeted lipidomics (22,23) for triglycerides (TGs) were conducted at LipidALL Technologies (Changzhou, China) by liquid chromatography-tandem mass spectrometry (Agilent 1290 II; Agilent Technologies, Germany). Data processing was conducted using MarkerView (Version 1.3, AB Sciex) and PeakView (Version 2.2, AB Sciex). Metabolite identification was accomplished through comparison with standard references in HMDB (https://hmdb.ca) and METLIN (https://metlin.scripps.edu) databases. Quality control (QC) samples were prepared by pooling aliquots from all samples into a mixed solution, injected between every 10 actual samples. The coefficient of variation for major metabolite classes was determined by calculating the median relative standard deviation (RSD) of metabolites. Instrument variability was assessed by calculating the RSD of internal standard metabolites.

Statistical analysis and data visualization

Metabolomics and lipidomics profiling were meticulously processed using R (v4.3.1) and illustrated by Adobe Illustrator 2023. The “preprocessCore” R package (v3.1.7) and “scales” R package (v1.3.0) were employed for quantile normalization and standardization.

Paired t-test was conducted to confirm the number of differential metabolites caused by swimming exercise (Fig. 2). One-way ANOVA was used to conduct difference analysis on the fold change (FC) values of differential metabolites at different time points after each intensity of exercise, and to create a heatmap for atlas production, with significance determined by Bonferroni-adjusted P values <0.05. Variable importance in projection (VIP) analysis and the area under the curve (AUC) were utilized to identify metabolites that exhibited intensity-dependent characteristics and were perturbed obviously by acute exercise stimulation. A higher VIP value indicates a greater contribution of the metabolite to differentiating between interventions; AUC represents perturbation of metabolites at various times before and after exercise under each intensity, providing clues to determine the amplitude of disturbance and the peak value of the most affected metabolites. Fuzzy c-means clustering was applied to determine the commonality and specificity of temporal changes under different swimming intensities (Fig. 3). To evaluate enriched pathways, Fisher exact test with the false discovery rate was used, a threshold of false discovery rate <0.05 was set to be the criterion for significance in the analysis of temporal metabolic pathways. Pairwise correlations among molecules within each cluster and the centrality of metabolites (the number of connections one metabolite has with other metabolites) were calculated to identify central metabolites in the metabolic network (Fig. 4). Integrating multilayered analytical framework, accompanied by heatmaps, the acute swimming metabolic atlas was depicted to clearly show various metabolic pathways of energy turnover driven by intensities (Fig. 5).

FIGURE 2.

Panoramic variations in the metabolome and exercise intensity-dependent metabolites screening. (A) Metabolites changes in response to acute HIIT and MICT in swimming. (B) UPSET plots illustrate the metabolomes associated with responses to training of different intensities. Numbers at the connection points indicate the quantity of differential metabolites specific to each intensity. The total count of regulatory metabolites specific to MICT and HIIT, displayed in the donut charts, is color coded by metabolite classification. (C) The VIP score shows the top 20 metabolites responsible for the separation of the data at each time point. Metabolites that are more in HIIT than MICT are shown in red and vice versa in blue. (D) The top 20 metabolites, distinguished by the highest AUC significance levels when comparing MICT to HIIT, were analyzed using the Mann–Whitney U test. The data are presented as median (quartile). (E) Pearson correlation of serum metabolites with l-lactic acid levels before and after exercise for the two exercise intensities in this cohort; green for baseline, blue for MICT, and red for HIIT.

FIGURE 3.

Time-resolved panoramic variations in the metabolome. (A, C, E, G) Longitudinal clustering identifies clusters for MICT and HIIT through time-series analysis, representing the mean values of metabolites within these clusters over time. Donut charts illustrate the total number of metabolites in each cluster, along with their classification. Meanwhile, Venn plots display the intersecting number of metabolites between MICT and HIIT within each cluster. (B, D, F, H) Top 10 significantly enriched pathways for MICT and HIIT under each cluster.

FIGURE 4.

Metabolic network alterations induced by varying exercise intensities. (A–L) The graphs illustrate the Pearson correlation network of metabolites for each MICT and HIIT cluster. In these networks, the size of a node reflects the centrality of a metabolite, whereas the thickness of a line denotes the correlation level between metabolites. In addition, the bars identify the top 5 metabolites with the highest centrality in each cluster, which are highlighted in the network diagram with fonts colored in blue or red. To form these networks, a correlation coefficient threshold of r > 0.7 and a significance level of P < 0.05 were applied. Only metabolites with a centrality >5 are displayed in the metabolic networks, ensuring that only the most interconnected metabolites are highlighted.

FIGURE 5.

The metabolome atlas of swimming reveals substrate utilization and recovery characteristics driven by intensities. The metabolic diagram summarizes the major metabolic pathways driven by swimming exercise intensity; direct linkages between metabolites are indicated using solid lines, whereas indirect linkages are indicated using dashed lines. The heatmap schematically illustrates exercise intensity-induced log₂FC changes in metabolites from baseline, with red indicating upregulation and blue indicating downregulation. Specific definitions of abbreviations that appear in the atlas but not in the original text are as follows: IMP, inosinc acid, inosinemon phosphate; AMP, adenosine monophosphate; GMP, guanosine monophosphate; UMP, uridine monophosphate; XMP, xanthosine monophosphate. 5-α-DHT glucuronide, 5-alpha-dihydrotestosterone glucuronide; 12-MTA, 12-methyltridecanoic acid; C17 S1P, C17 sphingosine-1-phosphate; EDA, eicosadienoic acid; EPA, eicosapentaenoic acid; GroPlns, glycerophosphoinositol; S1P, sphingosine 1-phosphate.

RESULTS

QC analysis of both acute swimming exercise model establishment and serum metabolome testing

Baseline data revealed no significant differences between groups in terms of age (P = 0.732), body morphology (P > 0.1), resting heart rate (P = 0.194), and levels of physical activity (P = 0.649) (Table 1). No significant total swimming duration (P = 0.053) and energy expenditure (P = 0.204) differences between groups were observed in acute swimming testing results, but the HIIT group exhibited significantly shorter net swimming time, higher maximum heart rate, and average swimming speed compared with the MICT group (P < 0.001) (Table 2).

TABLE 2.

Acute swimming testing data.

	MICT	HIIT	P
Total swimming duration, min	30.0 (30.0, 31.0)	29.0 (28.0, 31.0)	0.053
Net swimming duration, min	30.0 (30.0, 31.0)	12.0 (11.0, 12.5)	<0.001
50-m sprint duration in HIIT actual tesing, s	–	70.5 (8.9)	–
Total swimming distance, m	906.4 (131.5)	492.9 (32.7)	<0.001
Total energy expenditure,a kcal	249.4 (47.7)	270.8 (62.8)	0.204
Average swimming speed, m·s−1	0.50 (0.07)	0.71 (0.09)	<0.001
Average heart rate in swimming,b bpm	138.0 (9.4)	143.9 (11.3)	0.082
Maximum heart rate in swimming, bpm	155.4 (5.6)	182.8 (5.0)	<0.001
RPE of swimming (Borg 6 to 20)	12 (11, 14)	17 (15, 18)	<0.001

Data were mean (SD) for normally distributed, and median (P25, P75) for not normally distributed; P values for intergroup comparison were computed using nonparametric tests (Mann–Whitney U test).

^aThe values of total energy expenditure during swimming exercise were obtained by automatic estimation by the BHT-TEAM (BoHaoTong, China) wearable telemetry heart rate system.

^bThe average heart rate during swimming was the overall average heart rate; that is, the calculation of the average heart rate for the HIIT group included the heart rate during the interval periods.

Serum metabolome testing results showed that the overall variability of major metabolite categories and internal standard metabolites, as well as instrument variability, had a median RSD of less than 20%. The QC samples clustered together while the Spearman correlation coefficients averaged 0.99. These findings suggest that the metabolomics data demonstrate good consistency, with stable signals during the detection process, ensuring high data quality.

Unsupervised principal component analysis (PCA) results showed a high degree of overlap and consistency in baseline metabolites between groups (Fig. 1B), whereas PCA plot after exercise for MICT and HIIT groups showed a clear separation trend (Fig. 1C). Furthermore, the PCA results (Figs. 1D, E) for MICT and HIIT before (B) and after exercise immediately (M0/H0), at 15 min (M15/H15), and at 30 min postexercise showed that the degree of separation from the baseline was notably higher in the HIIT group. From the perspective of the quantity of detected metabolites, a total of 168 human serum samples were subjected to precise metabolomics analysis by liquid chromatography-tandem mass spectrometry, resulting in the detection of 370 metabolites. Summary of paired t-tests revealed significant changes in 209 metabolites (56.5%) following swimming exercise (P < 0.05). Among them, 103 metabolites exhibited significant alterations after MICT, whereas 183 after HIIT. Furthermore, the distinct metabolites of HIIT are 4.1 times those of MICT (Fig. 1F).

By synthesizing the analytical findings, this study revealed high homogeneity in participants' external demographic and internal biological characteristics. The physiological differences observed between groups were primarily attributed to exercise intensity rather than volume or participant demographics. Additionally, HIIT swimming induced a broader range of serum metabolite changes within a shorter net exercise duration, indicating that a valid and reliable intensity-specific swimming exercise model was established with rigorous quality control for precise metabolomic testing.

Panoramic variations in the metabolome driven by exercise intensity

Among 209 metabolites, 115 were significantly upregulated and 98 were significantly downregulated, whereas 4 metabolites (2-aminoheptanoate, 2-methyl-tridecanedioic acid, dodecadienoic acid, and pyroglutamic acid) were upregulated in MICT but downregulated in HIIT, and 2 were amino acid derivatives and the other 2 were fatty acids. HIIT had a higher number of up- and downregulated and distinct metabolites than MICT (Fig. 1G).

For easier metabolite categorization, the study referred to the HMDB database (https://hmdb.ca) and exercise metabolomic research (11–13) for secondary classification, as shown in Figure 2A, B, and the energy category primarily comprised metabolites within the tricarboxylic acid (TCA) cycle. An upregulation trend of metabolites in the energy category at various time points for both intensities was observed, with a greater number in HIIT compared with MICT (Fig. 2A). The upset plot and donut charts revealed that among the metabolites specifically upregulated by MICT, lipid and fatty acids (36.4%) and amino acids (13.6%) represented a high proportion, whereas among the downregulated metabolites, amino acids (25.0%), lipid and fatty acids (16.7%), and cofactor and vitamins (16.7%) constituted a higher proportion. Although for HIIT, purine and nucleotides (22.7%), amino acids (18.2%), lipid and fatty acids (15.9%), and keto acids (11.4%) represented a high proportion in the upregulated metabolites, and lipid and fatty acids (36.4%), amino acids (15.2%), and peptides (12.1%) among the downregulated metabolites (Fig. 2B). These results suggested that the TCA cycle can be significantly upregulated in both intensities. In MICT, lipid and fatty acids and amino acids turnover (biosynthesis and utilization) appeared to dominate, whereas energy utilization in HIIT might involve a broader range of substrates, including ATP, amino acids, fatty acids, and carbohydrates (keto acids being metabolic byproducts of fatty acids in the absence of carbohydrates). Furthermore, HIIT and MICT may have opposing regulatory patterns in lipid and fatty acids metabolism.

Intensity-dependent differential metabolites with significant fluctuations stimulated by swimming exercise

To accurately determine intensity-dependent metabolites between groups, a two-step screening process was conducted: 1) VIP values were calculated for each time point postexercise to identify the top 20 metabolites contributing most to discriminating between the two intensities (Fig. 2C). 2) Mann–Whitney U test was conducted on AUC values of metabolites between intensities to select the top 20 metabolites with the highest significance (Fig. 2D). The screening criteria were metabolites that simultaneously appeared in the results of VIP analysis at any time point and AUC analysis. The results indicated that there were a total of nine metabolites meeting the aforementioned criteria: l-lactic acid, pyruvate, fumaric acid, N-acetylvaline (NAV), N-lactoyl-phenylalanine (Lac-Phe), phenylpyruvic acid, pantothenic acid, dodecadienoic acid, and ketoleucine. Notably, the first five metabolites ranked among the top 20 in VIP values at all three time points, suggesting they may exhibit the highest level of continuous intensity specificity (by VIP) and homeostasis perturbation (by AUC) among all metabolites within 30 min postexercise.

l-Lactic acid is a classic biomarker of exercise intensity with a greater range of fluctuation compared with glucose, which endows it with extensive regulatory effects. By integrating baseline and two swimming modes data, Pearson analysis revealed strong correlations (r > 0.85) between NAV, pyruvate, fumaric acid, Lac-Phe, and lactic acid levels, with the NAV correlation coefficient reaching 0.917. These metrics all demonstrated significant partitioning with increasing intensity (Fig. 2E), suggesting they may serve as potential exercise-induced, intensity-regulated biomarkers with greater dynamic range and significant intensity-dependent fluctuations.

Temporal metabolic changes following acute swimming in response to exercise intensities

Through C-means fuzzy clustering analysis, four distinct metabolic trends were identified: cluster 1 exhibited a decrease followed by recovery (Fig. 3A); cluster 2 showed sustained elevation (Fig. 3C); cluster 3 demonstrated continuous decline (Fig. 3E); cluster 4 displayed an increase followed by recovery (Fig. 3G). Both the total number of metabolites and the number jointly regulated by MICT and HIIT were higher in the nonrecovery clusters (clusters 2 and 3) than in the recovery clusters (clusters 1 and 4), suggesting that 30 min postexercise is insufficient for most metabolites to fully recover to baseline levels, regardless of swimming intensity.

The TCA cycle pathway only appeared in the upregulated clusters (clusters 2 and 4). Combining the related network analysis for cluster 2 (Fig. 4D–F), MICT and HIIT were also found to coregulate steroid and keto acid metabolism pathways. Specifically, MICT formed an upregulated network centered around 3-oxotetradecanoic acid, focusing on unsaturated fatty acid synthesis and metabolism, whereas HIIT formed a network involving pyruvate metabolism, purine metabolism, short- to medium-chain fatty acid metabolism, branched-chain keto acids and short-chain keto acids. Notably, HIIT's network featured Lac-Phe and pyruvate as central metabolites, facilitating rapid energy utilization and oxidative stress regulation. The related network analysis results for the recoverable cluster 4 (Fig. 4J–L) indicated both MICT- and HIIT-regulated glycolysis metabolism; the central metabolic pathways and metabolites for MICT included carbohydrates and amino acids, centered in Lac-Phe, whereas for HIIT, the central metabolic pathways also involved purine metabolism, with the central metabolites being fumaric acid and NAV.

In the downregulated clusters (clusters 1 and 3), cluster 1 represented the change state of downregulation by exercise followed by equilibrium through metabolic turnover, where MICT induced a metabolic network dominated by amino acid metabolism, with a central metabolite of tyrosyl-alanine (Fig. 4A, C), and HIIT induced a metabolic network dominated by the metabolism of long-chain polyunsaturated fatty acids and glutamic acid, with glutathione as the central metabolite (Fig. 4B, C). It is noteworthy that the fatty acid categories regulated by HIIT in cluster 1 were similar to those regulated by MICT in cluster 2, such as arachidonic acid, eicosadienoic acid, FFA C22:5, etc. (Fig. 4B, D). In addition, long-chain fatty acid metabolism appeared in almost all clusters of MICT, whereas in HIIT, it only appeared in the downregulated recovery cluster 1. Purine metabolism was also observed in cluster 1 of MICT (Fig. 3B), whereas in HIIT, it was enriched in upregulated clusters 2 and 4. Notably, the central metabolite of purine metabolism was xanthosine (Figs. 3D, H and 4E, K). In cluster 3, two intensities coregulated the largest number of metabolic pathways (Figs. 3F and 4G, H), including primary bile acid metabolism, hydroxyl fatty acids metabolism, as well as those involved in the metabolism of a wide range of amino acids such as branched-chain amino acid (BCAA) metabolism, phenylalanine metabolism, alanine and aspartate metabolism, taurine, and hypotaurine metabolism. Furthermore, MICT specifically regulated long-chain fatty acids metabolism, cysteine metabolism, and glycine metabolism, with central metabolites including lipids (13(S)-HpOTrE, 27-norcholestanehexol) and amino acids (cysteine-S-sulfate, l-homocysteine, sarcosine). In contrast, HIIT specifically regulated aminoacyl-tRNA biosynthesis and sphingolipid metabolism, with bile acids serving as central metabolites (Fig. 4I).

The metabolome atlas of swimming reveals substrate utilization and recovery characteristics driven by intensities

Building on the previously discussed multilevel analysis framework, Figure 5, supplemented by a heatmap, illustrated the global metabolic effects in healthy individuals after acute swimming exercise, clarifying the various metabolic pathways involved in energy turnover at different swimming intensities.

TCA cycle and carbohydrate metabolism

The TCA cycle showed significant upregulation of key metabolites such as citrate, isocitrate, α-ketoglutarate (oxoglutarate), succinate, fumarate, and l-malate in both MICT and HIIT swims, with higher FCs typically seen in HIIT (Fig. 5). Similarly, glycolysis metabolites, including pyruvate, lactic acid, and alanine, were also affected, linking the Cori cycle and the glucose-alanine cycle. After MICT, lactic acid levels were recoverable and blood glucose levels remained stable. However, after HIIT swimming, blood lactic acid and glucose levels were higher and more volatile, not returning to baseline within 30 min.

Amino acids metabolism

Both intensities of swimming exercise consistently downregulated BCAA metabolism within 30 min, while corresponding keto acids like ketoleucine, α-ketoisovaleric acid, and ketomethylvaleric acid significantly increased. The downregulation of metabolites in the urea cycle maintaining nitrogen balance (citrulline, l-arginine, ornithine, etc.) suggested increased amino acids utilization for energy during exercise. Moreover, Figure 5 illustrated an exercise-induced neurotransmitter regulation pathway influencing emotions and promoting sleep formed by phenylalanine, tyrosine, tryptophan, and l-glutamate. Although these amino acids were downregulated under both intensities, their metabolic products such as gamma-aminobutyric acid, 4-hydroxybutyric acid (24), kynurenic acid (25), as well as dopamine-related metabolites like homovanillic acid sulfate (26) and 4-hydroxyphenylpyruvic acid were upregulated. Furthermore, glutamate, along with cysteine and glycine, formed glutathione (GSH), which decreased under exercise. Conversely, oxidized glutathione (GSSG) was significantly upregulated in HIIT, indicating a decreased GSH/GSSG ratio and increased oxidative stress during HIIT swimming, thus increasing the risk of cellular oxidative damage (27). In addition, l-glutamine, derived from glutamate, showed a downward trend in HIIT, suggesting a lower immunity post-HIIT swimming (28). Besides, it is worth noting that the time-dependent changes in l-homocysteine (HCY) showed opposite trends under the two swimming intensities.

Purine and pyrimidine metabolism

The end product of purine metabolism is uric acid. Figure 5 shows in purine metabolism that MICT downregulated uric acid, whereas HIIT upregulated it, although it was not statistically significant. The end products of pyrimidine metabolism are various coenzymes, which are used for energy utilization and substance metabolism through TCA cycle.

Lipids metabolism

Human blood lipids include TGs, phospholipids and cholesterol. According to the results of serum TGs (Fig. 6), HIIT caused a downward trend in most TGs. For MICT, TGs with relatively shorter chain lengths and lower degrees of unsaturation showed an upregulation trend, whereas TGs with longer chain lengths and higher degrees of unsaturation showed a downward trend. The fatty acids results indicated both intensities of swimming upregulated saturated fatty acids, such as stearic acid, palmitic acid, and arachidic acid. However, for various long-chain polyunsaturated fatty acids, MICT showed an upregulation trend, whereas HIIT displayed a downregulation trend followed by recovery (Fig. 5). Besides, both intensities resulted in a downward trend in serum complex lipids (glycerophospholipids and sphingolipids), and O-phosphoethanolamine was a phospholipid metabolite, significantly upregulated immediately after HIIT (Fig. 2C). Cholesterol metabolism generates bile acids, steroid hormones, and VD3. Both intensities resulted in a nonsignificant upregulation trend in serum 23S,25,26-trihydroxyvitamin D3 and steroid hormones (e.g., pregnanolone sulfate, dehydroepiandrosterone sulfate or DHEAS). Bile acid components like cholic acid and taurine were downregulated, suggesting accelerated turnover of the enterohepatic cycle. HIIT, in particular, showed a more pronounced downward trend in bile acids, likely indicating its stronger capacity for fatty acid utilization.

FIGURE 6.

Changes in AUC of serum TGs after swimming exercise of different intensities (targeted lipidomics results).

DISCUSSION

This study utilized a head-worn heart rate monitoring system and high-throughput omics testing to complete a precisely controlled acute swimming physiological analysis on 42 healthy young adults. A review on aquatic HIIT demonstrated that a physiologically beneficial HIIT swimming protocol involved a total duration of 28.1 ± 6.5 min, with individual bouts ranging from 15 s to 4 min and recovery periods from 10 s to 3 min (29). The HIIT protocol in the study had a total duration of 29 min, with bouts lasting 70.5 (8.9) s (Table 2) and intervals of 2 min, aligning with these parameters, ensuring precise metabolomic testing with reliable results. It is noteworthy that because of the adaptation training, the 50-m breaststroke sprint times in the screening test (Table 1) did not represent participants’ individual best performances for the 50-m breaststroke. Consequently, in the formal swimming tests, the average swimming speed of repeated breaststroke sprints (Table 2) showed slightly better results than those in the screening test (71.0 vs 70.5 s). This phenomenon is related to the skill proficiency of swimming “skill-based events” after adaptation training. More importantly, compared with MICT, HIIT swimming induced more extensive and specific changes in metabolites within a shorter net exercise time, indicating that average HR has at least limitations in reflecting the metabolic responses of the body during HIIT swimming exercises. Only observing changes in average HR may overlook the greater metabolic responses brought about by the HIIT swimming, so we have listed the maximum heart rate results in Table 2.

Based on the intensity-dependent significant homeostasis disturbances caused by swimming, five metabolites were selected: l-lactic acid, pyruvate, fumaric acid, Lac-Phe, and NAV. Notably, they also exhibited preferable centrality in the time-series change metabolic network analysis in Figure 4. In recent years, the lactate shuttle theory has revolutionized the traditional perception of lactate as merely a metabolic byproduct (30,31), suggesting that lactate serves not only as a viable energy substrate comparable to glucose but also as a signaling myokine and exerkine with extensive metabolic and physiological regulatory functions. This theory posits that lactatemia may be a “strain” rather than a “stress” biomarker. At rest, lactate in healthy human adults ranges from 0.8 to 1.5 mmol·L⁻¹ and glucose ranges from usually 4.0 to 5.0 mmol·L⁻¹, but when exercising, lactate can increase to over four times those of glucose during exercise. The greater range of fluctuation induced by exercise endows lactate with extensive regulatory effects (32). Pyruvate and fumaric acid are important intermediates in the TCA cycle and theoretically closely related to lactic acid in energy regulation. Lac-Phe was proven to be a lactate-derived signaling metabolite, synthesized in CNPD2+ cells, tightly linked to lactate metabolism and glycolytic flux, as well as to phenylalanine levels (33). It was found to be significantly upregulated in mice after acute exhaustive treadmill and horse racing exercises in previous studies (34). The latest research revealed that exercise-inducible Lac-Phe synthesis was driven by increased circulating levels of muscle-derived lactate, and the levels were associated with adipose tissue loss during endurance training in humans with obesity (35). The study supplemented the evidence of significant upregulation of Lac-Phe in human studies during MICT and HIIT in water environments.

Similar to Lac-Phe, NAV is also an acylated amino acid, which appears to be derived from valine and acetyl groups. The results of this study demonstrated that NAV was significantly upregulated by both MICT and HIIT swimming protocols, with a more substantial upregulation observed in the HIIT group. The high correlation coefficient (r = 0.917) between NAV and l-lactic acid, as depicted in Figure 2E, further suggests that NAV has the potential to serve as a novel exercise intensity-dependent biomarker. Previously published literature (36–38) has established that when endurance exercises were carried out under low carbohydrate availability, there was a potential increase in protein requirements, accompanied by a rise in serum BCAA metabolites. In this study, endurance exercises were commenced following a 10- to 12-h fasting period. It is hypothesized that the body was in a state of reduced carbohydrate utilization, and this state was likely more pronounced during HIIT swimming. Under such circumstances, the metabolism of BCAAs surged. Concurrently, the incomplete oxidation of fatty acids probably augmented the flux of acetyl-CoA, furnishing additional sources of acetyl groups and consequently fueling the significant production of NAV. This suggests that the dual stimuli of fasting and exercise potentially influenced the substantial upregulation of NAV observed in this study. Although the precise function of NAV remains to be fully elucidated, its strong correlation with vital health outcomes, including all-cause and cardiovascular mortality (39), cholesterol levels (40), diabetes (41), and kidney injury (42), underscores its potential importance in exercise-induced metabolic health.

The timing of sample collection indicated the metabolomic data reflected a delayed response to swimming exercise rather than real-time monitoring, which can, on one hand, intuitively showcase the recovery characteristics of energy substances postexercise and, on the other hand, help infer the utilization of energy substrates during exercise.

In terms of recovery characteristics, 30 min after both intensities of swimming was insufficient to restore most metabolites to baseline, indicating an energy deficit after swimming. The downregulated networks showed a significant decrease in a large number of amino acids and phospholipids. The study confirmed that the balanced ammonia cycle guided by HIIT in cluster 1 may primarily depend on glutamate metabolism (Figs. 3B and 4C), related not only to neurotransmitters involved in emotional regulation (43) but also to oxidative stress and immune recovery regulation (27,44). Phospholipids are crucial for cell membrane integrity; these results suggested a risk of cellular membrane damage after acute exercise, possibly because of enhanced oxidative stress, especially in HIIT. The specific regulation of glutamine and phospholipid metabolites by HIIT suggested that, besides the higher demand for muscle repair and limited protein synthesis after HIIT swimming, there was also a stronger risk of oxidative damage, including damage to cell membranes and the immune system (28), which affected overall recovery. This implied that unlike MICT swimming, where conventional high-quality protein or BCAA supplements after exercise may be enough, HIIT still necessitates needed attention to antioxidant nutrients, high-quality fat sources, and other dietary nutritional strategies.

In terms of the utilization of energy substrates, these intensity-specific changes were based on significant upregulation in the TCA pathway of both intensities, offering a perspective on exercise energy metabolism to analyze metabolites changes. MICT swimming induced an upregulated network of unsaturated fatty acids, with recoverable energy substrates being carbohydrates and amino acids, whereas HIIT swimming activated multiple fast energy utilization pathways, indicating the utilization of ATP, carbohydrates, amino acids, and medium-chain fatty acids during exercise (Fig. 3, Figure 4). For lipid substrate utilization, HIIT and MICT mainly differed in their capabilities of hydrolysis and oxidative utilization. From the perspective of endogenous fat, serum TGs originate from the release of visceral and subcutaneous adipose tissue, whereas serum fatty acids come from the hydrolysis of adipose tissue and serum TGs. Studies have demonstrated that subcutaneous adipose tissue is characterized by a higher proportion of unsaturated fatty acids (45), whereas visceral adipose tissue contains a greater proportion of saturated fatty acids in healthy individuals. In addition, visceral adipocytes are more metabolically active and exhibit higher lipolytic activity compared with subcutaneous adipocytes (46). Results showed that both intensities upregulate serum saturated fatty acids, suggesting that both intensities may mobilize visceral fat and promote its hydrolysis (45–47). MICT swimming promoted the release of TGs from adipose tissue into circulation and enhanced TGs hydrolysis in serum and adipose tissue, thereby upregulating serum fatty acid levels. HIIT swimming promoted serum TGs hydrolysis and the utilization of serum unsaturated fatty acids, resulting in lower serum TGs and unsaturated fatty acid levels. The difference in unsaturated fatty acid metabolism suggested MICT had a higher capacity for the hydrolysis of long-chain unsaturated fats compared with oxidative utilization, indicating a stronger mobilization and hydrolysis capacity of subcutaneous adipose tissue (45,46). Although HIIT had a stronger fatty acid utilization and blood lipid-lowering effect but a weaker hydrolysis capacity for unsaturated fats in adipose tissue, this capacity may gradually increase with prolonged recovery time. This may be related to the high blood glucose fluctuations and elevated lactate levels after HIIT swimming (48). Post-HIIT swimming hyperglycemia can increase insulin secretion, which inhibited hormone-sensitive lipase activity, thereby reducing TGs hydrolysis (48), and an inverse relationship between blood lactate and free fatty acid concentrations was noted (31). As time passes, blood lactate and glucose levels decreased, lifting the inhibitory effect. Thus, both forms of swimming promoted significant changes in fatty acid mobilization and oxidation pathways (49–51). MICT swimming contributed more to increasing lipolysis and fatty acid mobilization in adipose tissues, but fatty acid oxidation represents the actual consumption of fat; HIIT swimming enhanced the utilization of serum TGs and serum fatty acids, potentially offering long-term benefits in improving blood lipids and overall body fat percentage. So this may be the reason why long-term swimming experiments usually show HIIT swimming has a better improvement effect on body fat in obese groups than MICT (15).

Some metabolic changes that are not significant in healthy populations but do exist could provide insights into exercise prescriptions for populations with metabolic disorders or health risks. For instance, the observed fluctuations in blood glucose, elevated blood uric acid, and increased levels of HCY (52) following HIIT suggested potential metabolic risks. Given that healthy populations possess self-recovery and adaptation capabilities (53), individuals with metabolic disorders may struggle to recover from these fluctuations. Therefore, for individuals with glucose disorders, gout, and those at cardiovascular risk, MICT swimming may be more suitable than HIIT. And understanding the recovery pathways of these indices elevated by exercise in healthy individuals may guide research into preventing or treating metabolic disorders in at-risk populations (54,55).

This study also has some limitations. Although there was no statistically significant difference in the baseline information of the two groups of exercising populations, when conditions permit, a randomized crossover experiment that conducts different exercise interventions on the same individual is the best choice for the experimental design. The gender differences in different swimming exercise interventions still need further study. If the microsampling omics technology can be widely developed, conducting baseline sampling after the adaptation training and carrying out sampling observations for a longer period after exercise will be conducive to achieving more rigorous and comprehensive observations. Moreover, for skill-dependent exercises like swimming, apart from the aquatic environment, skill proficiency also influences the intensity response. To precisely capture the physiological responses to exercise intensity, conducting studies using less skill-dependent exercise modalities (e.g., cycling on an ergometer) may be better. In addition, given the unique aquatic environment and skill requirements of swimming, the metabolomic results obtained in this study may not fully generalize to other exercise modalities.

CONCLUSIONS

Based on sophisticated exercise test control and high-quality metabolic results, a reproducible exercise intensity-specific human acute swimming model was established. NAV was identified as a novel intensity-dependent biomarker most closely linked to lactate. The metabolic atlas of acute swimming showed both intensities significantly upregulated TCA cycle and induced amino acid pathways associated with mood regulation and sleep improvement, with HIIT swimming having a more pronounced effect, resulting in a compounded nutritional recovery strategy need for HIIT. MICT swimming promotes TGs hydrolysis more effectively, whereas HIIT swimming enhances fatty acid oxidative utilization, but prolonging exercise recovery time can lift its hydrolysis limitation. In addition, different trends of indicators such as uric acid and L-HCY between intensities in healthy populations were observed, providing molecular evidence for selecting swimming exercise prescriptions for populations at relevant risk of disease.