Monitoring the Response of the Human Urinary Metabolome to Brief

Sep 30, 2015 - (3-6) The corollaries of exercise range from acute effects on biological parameters, which are readily observed from as shortly as seco...
1 downloads 14 Views 4MB Size
Article pubs.acs.org/jpr

Monitoring the Response of the Human Urinary Metabolome to Brief Maximal Exercise by a Combination of RP-UPLC-MS and 1H NMR Spectroscopy Alexandros Pechlivanis,†,‡,∥ Konstantinos G. Papaioannou,§ George Tsalis,§ Ploutarchos Saraslanidis,§ Vassilis Mougios,§,⊥ and Georgios A. Theodoridis*,‡ †

Biomolecular Medicine, Division of Computational and Systems Medicine, Department of Surgery and Cancer, Faculty of Medicine, Imperial College London, SW7 2AZ London, United Kingdom ‡ School of Chemistry, Aristotle University of Thessaloniki, 54124 Thessaloniki, Greece § School of Physical Education and Sport Science at Thessaloniki, Aristotle University of Thessaloniki, 54124 Thessaloniki, Greece S Supporting Information *

ABSTRACT: The delineation of exercise biochemistry by utilizing metabolic fingerprinting has become an established strategy. We present a combined RP-UPLC-MS and 1H NMR strategy, supplemented by photometric assays, to monitor the response of the human urinary metabolome to short maximal exercise. Seventeen male volunteers performed two identical sprint sessions on separate days, consisting of three 80 m maximal runs. Using univariate and multivariate analyses, we followed the fluctuation of 37 metabolites at 1, 1.5, and 2 h postexercise. 2-Hydroxyisovalerate, 2-hydroxybutyrate, 2-oxoisocaproate, 3-methyl-2-oxovalerate, 3-hydroxyisobutyrate, 2-oxoisovalerate, 3-hydroxybutyrate, 2-hydroxyisobutyrate, alanine, pyruvate, and fumarate increased 1 h postexercise and then returned toward baseline. Lactate and acetate were higher than baseline at 1 and 1.5 h. Hypoxanthine and inosine remained above baseline throughout the postexercise period. Urate decreased at 1 h and increased at 1.5 h before returning to baseline. Valine, isoleucine, succinate, citrate, trimethylamine, trimethylamine N-oxide, tyrosine, and formate decreased at 1 h and/or 1.5 h postexercise and then returned to baseline. Creatinine gradually decreased over the sampling period. Glycine, 4-aminohippurate, and hippurate remained below baseline throughout the postexercise period. Our findings show that even one-half minute of maximal exercise elicited major perturbations in human metabolism, several of which persisted for at least 2 h. KEYWORDS: physical exercise, metabonomics, UPLC-MS, NMR spectroscopy, sprint running, urinary metabolite fingerprinting

1. INTRODUCTION

aforementioned metabonomic techniques with classical biochemical assays has been demonstrated in recent reports,12,13 as was done in the present work. Further to our previous reports on the short- and long-term effects of brief maximal exercise on the human urinary14 and serum metabolome,15 herein we present the monitoring of the urinary metabolic fingerprint for 2 h after exercise through the combination of RP-UPLC-MS and 1H NMR spectroscopy. By delineating the postexercise kinetics of 37 urinary metabolites, mainly involved in the pathways of ATP, carbohydrate, lipid, and amino acid degradation, we present evidence for differences in the duration and magnitude of exercise-induced perturbations in each pathway.

Metabolic fingerprinting has proven a significant addition to the battery of classical tools for the understanding of mechanisms regulating the response of biological systems to exogenous interventions.1,2 Physical exercise is an intervention that can affect many biochemical pathways in both beneficial and injurious ways.3−6 The corollaries of exercise range from acute effects on biological parameters, which are readily observed from as shortly as seconds to as late as days after its conclusion,7 to chronic effects, which persist for much longer.8 The overwhelming complexity of the network of metabolic pathways affected by physical exercise warrants the simultaneous monitoring of multiple metabolites on a prolonged time scale. This challenge can be met with the use of multiplatform metabolic fingerprinting. Metabolic fingerprinting with the use of NMR technology can rely on the advantages of exceptional robustness, high identification power, reproducibility, and repeatability.9 The lack of sensitivity of the technique can be met with the utilization of hyphenated MS technologies, which, in turn, excel in sensitivity and applicability but lag behind in reproducibility, robustness, and identification power.10,11 The complementarity of the © XXXX American Chemical Society

2. EXPERIMENTAL SECTION 2.1. Experimental Intervention and Sample Collection

Seventeen young, physically active, healthy male volunteers (physical education students; age, 19 ± 1 years; body mass, Received: May 27, 2015

A

DOI: 10.1021/acs.jproteome.5b00470 J. Proteome Res. XXXX, XXX, XXX−XXX

Article

Journal of Proteome Research 72 ± 9 kg; height, 1.82 ± 0.07 m; BMI 21.6 ± 1.8 kg/m2; all mean ± SD) provided written informed consent to participate in the study. The study was approved by the institutional ethics committee, and all procedures were in accordance with the Helsinki declaration of 1975, as revised in 1996. The volunteers underwent two identical intermittent sprint sessions (test and retest) consisting of three 80 m maximal runs with intervals of 10 min between the first and second runs (to activate the energy production mechanisms) and 10 s between the second and third runs (to maximize lactate production). The average time of each run was 10.33 s. Sessions were performed at least 3 days apart. The main objective of the study was to test the reliability (reproducibility) of the urine lactate concentration. In order to obtain a broader view of changes in the metabolic fingerprint as a result of brief maximal exercise, two state-of-theart techniques, LC-MS and 1H NMR spectroscopy, were selected for the holistic analysis of the samples. The tests were performed in the morning, 2−3 h after consumption of a light breakfast. The volunteers kept a detailed record of their diet the day before the test and were asked to repeat the diet the day before the retest. Urine samples were collected right before exercise as well as 1, 1.5, and 2 h after exercise. In each instance, the volunteers emptied their bladders. During the sampling period, they were allowed only water ad libitum. After thorough mixing of each urine sample, a few milliliters were stored at −80 °C until analyzed. The study design is graphically presented in Figure 1.

that aliquots of the urine samples collected 1 h postexercise were diluted 1:100, whereas aliquots of all other samples were diluted 1:11 for the assay. Additionally, urate was determined by use of a reagent kit from Spinreact (Santa Coloma, Spain) in order to complement the picture of purine metabolism obtained from the RP-UPLC-MS and 1H NMR analyses, since urate could not be safely identified through those techniques. 2.4. RP-UPLC-MS Analysis

The samples were analyzed using an Acquity UPLC system (Waters, Elstree, U.K.), connected with the LCT Premier Time of Flight (ToF) mass spectrometer (Waters Micromass), using an electrospray ionization (ESI) source. The LockSpray system was used for the constant determination and correction of the instrument’s mass accuracy through a leu-enkephalin solution (0.2 mg/L) at a flow rate of 20 μL/min. The analysis was performed at both positive (ESI+) and negative electrospray ionization (ESI−). The samples were kept at 4 °C pending analysis. For the chromatographic separation, we used the reversedphase column, Acquity UPLC HSS T3 C18 (1.8 μm, 2.1 mm × 100 mm), at a temperature of 40 °C. The solvents used were 0.1% (by vol.) formate in water and 0.1% (by volume) formate in acetonitrile. The samples were analyzed at a flow rate of 0.5 mL/min using a 12 min linear gradient program with the following percentages of formate in water: 99.9 (0.0−1.0 min), 99.9−85 (1.0−3.0 min), 85−50 (3.0−6.0 min), 50−5 (6.0− 9.0 min), 5 (9.0−10.0 min), 5−99.9 (10.0−10.1 min), and 99.9 (10.1−12.0 min). The injection volume was 5 μL. After each injection, the syringe and injection valve were washed with 200 μL of “strong solvent” (H2O/MeOH, 10:90 by volume), followed by 600 μL of “weak solvent” (H2O/MeOH, 80:20 by volume). The mass spectrometer conditions were as follows: source temperature, 120 °C; desolvation temperature, 350 °C; cone gas flow, 25 L/h; desolvation gas flow, 900 L/h; capillary voltage, 3200 V for ESI+ and 2400 V for ESI−; cone voltage, 35 V. Nitrogen was used as desolvation gas. Data were recorded for the m/z region of 50−1000 in V-mode with a scan time of 0.20 s and dwell time of 0.01 s between scans. Data were collected in centroid mode. The analyte-to-reference ratio was 99:1. The instrument was calibrated prior to the analyses using 0.5 mmol/L sodium formate (resolution ∼ 8000, accuracy ∼ 10 ppm for both types of ionization). The samples were analyzed in random order and QC samples were inserted every ten samples during the analysis to monitor the stability of the analytical platform. QC samples were also used to condition the column at the start of each assay (10 samples for ESI+ and 8 samples for ESI−). In order to obtain fragmentation information for structural identification, we subjected QC samples to exact mass measurement at low and high collision energy (MSE) without the selection of precursor ions in the quadrupole of a XEVO G2 QToF mass spectrometer under the same chromatographic conditions as above.18

Figure 1. Experimental design of each exercise session (test and retest). Boxes represent 80 m sprint runs and arrows represent urine sampling points. Time is not in scale for clarity.

2.2. Sample Processing

Sample preparation for RP-UPLC-MS analysis was performed by adapting the protocol from Want and co-workers.16 Urine samples that were to be analyzed in a single instrument run were thawed just prior to processing. Then 150 μL of each sample was diluted with 150 μL of water and, after vortexing for 30 s, the mixture was centrifuged at 9660g for 10 min. Next, 200 μL of each supernatant was transferred to a 96-well plate, while 50 μL was mixed with equal volumes from the other supernatants in a glass vial for the preparation of a quality control (QC) sample. The samples were kept at −20 °C until the day of analysis, when they were thawed and centrifuged at 9660g for 10 min before being placed in the autosampler of the liquid chromatograph. Sample preparation for 1H NMR analysis was performed by adapting the protocol from Beckonert and co-workers.17 An amount of 500 μL of each urine sample was centrifuged at 9660g for 10 min. Then 400 μL of the supernatant was mixed with 200 μL of 0.2 mol/L phosphate buffer, pH 7.4, in D2O, containing 0.684 mmol/L sodium 3-trimethylsilyl-(2,2,3,3-2H4)-1propionate (TSP) and 0.2 mg/mL NaN3. The mixture was vortexed and transferred into 5 mm wide NMR tubes.

2.5. 1H NMR Spectroscopy Analysis

All 1H NMR experiments were performed on a Bruker (Karlsruhe, Germany) Avance III 600-MHz spectrometer, using an inverse detection probe (5 mm) with z-gradients, at 300 K. One-dimensional (1D) 1H NMR spectra were obtained using a standard pulse sequence, 1D NOESY (noesypr1d), included in the spectrometer’s library, with presaturation during relaxation and mixing time for water suppression. For each 1D

2.3. Photometric Assays

For validation purposes, lactate was determined through an enzymic photometric assay as described,14 with the exception B

DOI: 10.1021/acs.jproteome.5b00470 J. Proteome Res. XXXX, XXX, XXX−XXX

Article

Journal of Proteome Research

each compound were normalized to the median of the preexercise values. The significance of changes with exercise was examined by Friedman’s ANOVA. Significant main effects were followed up with Wilcoxon’s matched-pairs signed-rank test with Bonferroni correction for the six possible pairwise comparisons of four time points. Effect sizes were calculated as the Z of the Wilcoxon’s test, divided by the square root of observations, and were considered small if they were up to 0.3, medium if they were above 0.3 and up to 0.5, and large if they exceeded 0.5.22 The level of statistical significance was set at α = 0.05. Data were analyzed in SPSS (version 22, SPSS, Chicago, IL).

NOESY spectrum, a total of 128 scans were collected in 65 536 data points over a spectral width of 12 019.23 Hz, using a relaxation delay of 2 s, an acquisition time of 4.54 s, and a mixing time of 0.1 s. An exponential weighting factor corresponding to a line broadening of 0.3 Hz was applied to all acquired free induction decays prior to Fourier transformation and phase correction. 2.6. RP-UPLC-MS Data Treatment and Analysis

The three-dimensional RP-UPLC-MS raw data were preprocessed using the R statistics language and the freely available XCMS19 software in order to convert them into two-dimensional, time-aligned detected features. Specifically, for the detection of the features, the “centWave”20 algorithm was used with a peak width window of 3−20 s. Ten ppm was chosen as the maximum allowable m/z variation between two successive scans, and the signal-to-noise ratio was set at 10. Peak integration was performed on the raw data. Grouping of the detected peaks was done according to peak density, using a standard deviation of 30 s for the smoothing algorithm for the first grouping of peaks. Retention time correction was performed using a polynomial smoothing function that utilized the “well-behaved” peaks as “anchors.” The last grouping was done using a standard deviation of 4 s. The “minFrac” filter was utilized to filter out features that were not present in at least 50% of the samples of at least one group of similar samples (i.e., pre-exercise, 1 h postexercise etc.). Normalization of the data was done using the median fold change algorithm.21 From the table thus produced, we filtered out the features for which the coefficient of variation in the QC samples was ≥30% and which were not present in at least 90% of the QC samples. The resulting feature table was subjected to multivariate analysis using the SIMCA 13 software (Umetrics, Umea, Sweden) for principal component analysis (PCA). Pairwise comparisons were performed using orthogonal partial leastsquares discriminant analysis (OPLS-DA).

3. RESULTS 3.1. RP-UPLC-MS Data

Initial PCA modeling of the data set after Pareto scaling, including the study samples and the QC samples, was used to assess the stability of the analytical system during the analysis. The stability of the analytical system was also audited through the examination/scrutiny of raw chromatographic data as extracted ion chromatograms of representative mass traces throughout the chromatographic run. When we compared samples taken at the same time point at test and retest through OPLS-DA after Pareto scaling of the extracted features, no classification through a valid model could be achieved. This led us to conclude that metabolic fingerprints were highly reproducible, thus enabling us to pool the data from the test and retest sessions at each time point and, hence, increase the number of observations and statistical power of the analyses. Since the stability of the analytical system was satisfactory, we could safely assume that the observed variation among samples was mainly biological, not technical. Supporting Information Figure S1 shows the PCA scores plots for ESI+ and ESI−, both in two and three dimensions. The QC samples can be seen at the center of a multiparametric space in a “tight formation,” confirming the stability of the analytical system. An interesting feature of the scores plots was the relative clustering of samples collected at different time points, although the samples of 2 h postexercise tended to disperse away from the samples of the other time points. This variation may be attributed to the fact that the final sampling had been preceded by over 2 h of variable (ad libitum) hydration, resulting in samples that were varyingly and considerably diluted. Comparisons between the samples of each time point were performed by constructing pairwise OPLS-DA cross-validated models. The scores plots for ESI+, along with the respective permutation plots, are presented in Figure 2. A summary of the models’ characteristics is presented in Supporting Information Table S2. In Supporting Information Figure S2, we present the S-plot of the OPLS-DA model comparing the baseline samples with those obtained 1 h after exercise. Based on that and on similar plots of the models comparing baseline with 1.5 and 2 h after exercise (not shown), we could identify the features that were significant for the discrimination between pre- and postexercise. With the aid of MSE experiments, we identified hypoxantine and inosine as the main discriminators. Hypoxanthine increased 1 h after exercise and remained high up until 2 h. Inosine, on the other hand, increased 1 h after exercise but returned to baseline from 1.5 h onward. In Supporting Information Figure S3, the mass spectra and identified features of the two compounds are presented. The same workflow as above was followed for the ESI− data. Pairwise comparisons between the samples collected at each time

2.7. 1H NMR Data Treatment and Analysis

All obtained spectra were preprocessed using an in-house Matlab script, through which the full range of the spectra (from −1 to 10 ppm at a resolution of 0.00033 ppm) is inserted in the software, followed by automatic phase correction, baseline correction, and chemical shift scaling on the basis of the internal standard, TSP. After removal of the region corresponding to residual water (4.55−5.18 ppm), data were aligned and then normalized by using the median fold change algorithm, as in the case of the RP-UPLC-MS data. The spectral data were subsequently subjected to univariate scaling and, finally, analyzed by PCA and OPLS-DA using the SIMCA 13 software. In addition to multivariate analysis, the 1H NMR data were subjected to univariate statistical analysis. To this end, the aggregate spectral signals of each identified compound that could be safely measured in each pair of test and retest samples were averaged, and the resulting data sets were examined for normality of distribution by the Shapiro-Wilk test. Because the distribution of a large portion (40%) of the variables did not follow a Gaussian distribution, we subjected the data to nonparametric descriptive and inferential statistical analysis. Nevertheless, Supporting Information Table S1 contains the parametric descriptive statistics (that is, mean and SD) of the variables for additional information. For nonparametric statistics, the lactate level was correlated to the lactate concentration (as determined photometrically) through Spearman’s ρ correlation analysis. Then the levels of C

DOI: 10.1021/acs.jproteome.5b00470 J. Proteome Res. XXXX, XXX, XXX−XXX

Article

Journal of Proteome Research

Figure 2. Scores plot for the OPLS-DA model on the basis of RP-UPLC-MS ESI+ data, comparing, in pairwise fashion, the baseline samples (black circles), the samples obtained 1 h after exercise (red circles), the samples obtained 1.5 h after exercise (blue circles), and the samples obtained 2 h after exercise (green circles). Insets depict the permutation plots for each OPLS-DA model. No valid model could be constructed for the comparison of 1.5 and 2 h.

point were performed by constructing pairwise OPLS-DA crossvalidated models. The scores plots for ESI-, along with the respective permutation plots, are presented in Supporting Information Figure S4. Comparing the samples obtained 1 h after exercise with those obtained 1.5 and 2 h after exercise gave a similar image as in ESI+, whereas, again, no model could discriminate the samples obtained 1.5 h after exercise from those obtained 2 h after exercise. Using MSE analysis in ESI- mode, we confirmed the contribution of inosine to the discrimination among time points. Through the use of databases such as HMDB and MassBank, along with an in-house database and MSE experiments a number of biomolecules could be identified, but these were not found to contribute to the group differentiation. For example, eight acylcarntines (namely, carnitine, acetylcarnitine, propionylcarnitine, butyrylcarnitine, isobutyrylcarnitine, valerylcarnitine, isovalerylcarnitine, and octanoylcarnitine) could

be identified by retention time and mass spectral match in comparison to a laboratory based database and injection of reference standards. No other discriminating feature of the chromatograms, obtained in either the ESI+ or ESI− mode, could be identified. 3.2. 1H NMR Spectroscopy Data

3.2.1. Multivariate Analysis. In line with the RP-UPLC-MS data analysis, the reproducibility of the metabolic fingerprints obtained through 1H NMR spectroscopy between test and retest at all time points of sampling was examined by OPLS-DA. No classification could be achieved for any binary comparisons through significant models, leading us to pool the data from test and retest, as we did with the RP-UPLC-MS data. The 1H NMR spectra of urine exhibited a plethora of metabolite signals, as shown in Supporting Information Figure S5, which depicts the median total spectra before and 1 h after exercise. Identification of metabolites was performed according D

DOI: 10.1021/acs.jproteome.5b00470 J. Proteome Res. XXXX, XXX, XXX−XXX

E

2-hydroxyisovalerate 2-hydroxybutyrate 2-oxoisocaproate 3-methyl-2-oxovalerate valine isoleucine 3-hydroxyisobutyrate 2-oxoisovalerate 3-hydroxybutyrate lactate

2-hydroxyisobutyrate alanine acetate acetoacetate oxaloacetate pyruvate succinate citrate dimethylamine trimethylamine 2-oxoglutarate creatinine malonate carnitine trimethylamine n-oxide taurine glycine allantoin ureac xanthosinec fumarate trans-aconitate 4-aminohippurate tyrosine hippurate tryptophanc hypoxanthine

inosine

formate

M1 M2 M3 M4 M5 M6 M7 M8 M9 M10

M11 M12 M13 M14 M15 M16 M17 M18 M19 M20 M21 M22 M23 M24 M25

M38

M39

M26 M27 M28 M29 M30 M31 M32 M33 M34 M35 M36 M37

metabolite

metabolite key

(s) (d), 3,78 (q) (s) (s), 3.44 (s) (s) (s) (s) (d), 2.71 (d) (s) (s) (t), 3.02 (t) (s), 4.06 (s) (s) (s) (s)

8.24 (s), 8.35 (s), 6.1 (d), 4.28 (dd), 4.4 (dd), 3.91 (dd), 3.83 (dd) 8.46 (s)

3.27 (t), 3.43 (t) 3.57 (s) 5.39 (s) 5.8 (br) 5.85 (d), 7.8 (s) 6.52 (s) 3.44 (s), 6.59 (s) 6.86 (d), 7.7 (d) 6.9 (m), 7.2 (m) 3.86 (d), 7.5 (t), 7.6 (tt), 7.8 (dd) 7.74 (d), 7.5 (d), 7.3 (s), 7.28 (m) 8.19 (s), 8.21 (s)

1.36 1.48 1.93 2.24 2.34 2.38 2.41 2.57 2.73 2.87 2.45 3.05 3.11 3.23 3.27

0.84 (d), 0.97 (d) 0.90 (t), 1.65/1.74 (m), 4.01 (dd) 0.94 (d), 2.1 (m), 2.61 (d) 0.89 (t),1.1 (d) 0.97 (d), 1.04 (d) 0.94 (t), 1.01 (d) 1.07 (d), 2.48 (m), 3.68 (m) 1.11 (d), 3.02 (m) 1.2 (d), 2.3/2.4 (m), 4.16 (m) 1.33 (d), 4.12 (q)

chemical shift (ppm)

19.36 (15.64−25.74)a, d 8.52 (3.72−17.34)a 0.39 (0.36−0.57)a, c, d

1.00 (0.74−1.60)b, c, d 1.00 (0.87−1.30)b, c, d 1.00 (0.70−1.12)b

1.61 1.10 0.63 0.82 0.78

(1.38−1.91)a, c (0.91−1.16) (0.43−0.74)a (0.70−0.93)a (0.67−1.05)a 1.00 1.00 1.00 1.00 1.00

(0.73−1.08)b (0.93−1.14) (0.65−1.13)b, d (0.87−1.19)b (0.76−1.88)b, c, d

1.44 (1.22−1.86) 2.09 (1.76−2.86)a, c, d 1.83 (1.67−2.21)a, c, d 2.15 (1.98−2.65)a, c, d 0.90 (0.89−0.97)a 0.99 (0.94−1.07)c 1.98 (1.48−2.42)a, c, d 1.10 (1.02−1.17)a, c 1.39 (1.27−1.73)a, c, d 179.02 (140.61− 249.50)a, c, d 2.00 (1.79−2.76)a, c, d 1.49 (1.22−1.73)a, c, d 1.71 (1.51−2.37)a, c, d 1.05 (1.00−1.10) 1.12 (0.89−1.27) 8.27 (6.66−13.90)a, c, d 0.69 (0.62−0.81)a, c 0.68 (0.53−0.79)a, c, d 0.91 (0.87−1.00) 0.79 (0.73−1.33)a 0.90 (0.81−1.00) 0.90 (0.81−0.97) 1.09 (0.85−1.44) 0.96 (0.59−1.41) 0.80 (0.67−0.93)a, c

a, c, d

1 h postexercise

0.76 (0.64−1.24) 0.67 (0.54−0.73)a, c 0.80 (0.67−0.94)

(0.60−1.14)b (0.82−1.14)b, d (0.93−1.16)b, c (0.93−1.05) (0.91−1.48) (0.94−1.04)b (0.89−1.12)b (0.82−1.27)b (0.85−1.14) (0.77−1.78)b (0.94−1.13) (0.86−1.13)d (0.81−1.22) (0.69−1.59) (0.91−1.25)b, d

(0.86−1.15) (0.78−1.15)b (0.94−1.02)b (0.88−1.22)b (0.90−1.13)b, c (0.91−1.10)c (0.87−1.14)b (0.89−1.07)b (0.88−1.15)b (0.92−1.04)b, c

b

1.00 (0.80−1.51) 1.00 (0.87−1.31)b, c, d 1.00 (0.74−1.26)

1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00

1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00

pre-exercise

(0.98−1.39)b (0.82−1.11)b (1.17−1.52)a, b, d (0.95−1.07) (0.82−1.42) (0.94−1.76)b (0.75−1.02)b (0.89−1.46)b (0.89−1.20) d (0.77−1.56) (0.87−1.00) (0.85−1.18) d (0.87−1.34) (0.71−2.01) (0.78−1.14)b

(0.77−1.19) (0.86−1.43)b (0.92−1.37)b (0.98−1.70)b (0.84−1.04)a (0.83−0.97)a, b (0.92−1.53)b (0.89−1.10)b (0.96−1.13)b (2.40−32.05)a, b, d

(0.81−1.23)b, d (0.97−1.31) (0.49−1.06) (0.68−1.20) (0.58−1.01)a

0.73 (0.57−1.06)b, d

2.38 (1.54−9.97)a

21.49 (16.03−30.56)a, d

0.90 1.14 0.80 0.92 0.76

1.26 (0.59−1.59) 0.79 (0.69−0.98)a, b 0.97 (0.73−1.17)

1.08 0.93 1.29 1.02 0.91 1.11 0.90 1.16 1.05 1.07 0.90 0.90 1.04 1.04 0.94

1.02 1.10 1.08 1.12 0.92 0.90 1.15 0.97 1.01 4.71

b

1.5 h postexercise

Table 1. Changes in Human Urine Metabolite Levels, Identified by 1H NMR Spectroscopy, after Maximal Intermittent Exercisea

(0.76−1.16)b (0.72−0.88)a, b (0.98−1.27)b, c (0.97−1.26) (0.74−1.53) (0.94−1.10)b (0.72−0.96) (0.69−1.33)b (0.74−1.02) c (0.73−1.27) (0.78−1.00) (0.72−0.92)a, c (0.80−1.31) (0.61−1.45) (0.63−1.04)a

(0.74−1.11) (0.75−1.22)b (0.83−1.13)b (0.79−1.19)b (0.90−1.13) (0.83−1.12) (0.81−1.19)b (0.68−1.13) (0.89−1.16)b (0.87−2.61)b, c

(0.90−2.39)c (1.04−1.68) (0.27−0.98)a (0.59−1.20) (0.51−0.94)a

1.17 (0.77−1.45)b, c

12.29 (6.54− 17.68)a, b, c 3.16 (1.62−4.77)a

1.31 1.20 0.80 0.91 0.76

0.87 (0.61−1.45) 0.75 (0.60−0.92)a 0.89 (0.66−1.30)

0.94 0.82 1.05 1.05 1.00 1.02 0.84 0.98 0.82 0.97 0.89 0.82 0.95 0.83 0.86

0.95 0.90 0.94 0.97 0.95 0.96 1.06 0.91 1.00 1.12

b

2 h postexercise