Skeletal muscle alkaline Pi pool is decreased in overweight-to-obese sedentary subjects and relates to mitochondrial capacity and phosphodiester content

Defects in skeletal muscle energy metabolism are indicative of systemic disorders such as obesity or type 2 diabetes. Phosphorus magnetic resonance spectroscopy (31P-MRS), in particularly dynamic 31P-MRS, provides a powerful tool for the non-invasive investigation of muscular oxidative metabolism. The increase in spectral and temporal resolution of 31P-MRS at ultra high fields (i.e., 7T) uncovers new potential for previously implemented techniques, e.g., saturation transfer (ST) or highly resolved static spectra. In this study, we aimed to investigate the differences in muscle metabolism between overweight-to-obese sedentary (Ob/Sed) and lean active (L/Ac) individuals through dynamic, static, and ST 31P-MRS at 7T. In addition, as the dynamic 31P-MRS requires a complex setup and patient exercise, our aim was to identify an alternative technique that might provide a biomarker of oxidative metabolism. The Ob/Sed group exhibited lower mitochondrial capacity, and, in addition, static 31P-MRS also revealed differences in the Pi-to-ATP exchange flux, the alkaline Pi-pool, and glycero-phosphocholine concentrations between the groups. In addition to these differences, we have identified correlations between dynamically measured oxidative flux and static concentrations of the alkaline Pi-pool and glycero-phosphocholine, suggesting the possibility of using high spectral resolution 31P-MRS data, acquired at rest, as a marker of oxidative metabolism.


Results
Between groups comparison. In addition to a significantly higher BMI and lower VO 2max , the Ob/Sed individuals also differed from the L/Ac volunteers in the metabolic parameters derived from 31 P-MRS. The concentration of the main muscular PDE (i.e., glycero-phosphocholine [GPC]), as well as the total [PDE], were significantly higher, while the concentration of the alkaline Pi-pool ([Pi 2 ]) and its ratio to the main Pi concentration ([Pi 1 ]), i.e., (Pi 2 /Pi), were significantly lower in the Ob/Sed group compared to the L/Ac group. In addition, the group of Ob/Sed subjects had significantly lower mitochondrial capacity (Q max ) and Pi-to-ATP exchange flux (F ATP ) values compared to the L/Ac group. Detailed information about the measured physiological and muscle energy metabolism parameters are listed in Table 1. In Fig. 1 are depicted representative 31 P-MR spectra acquired at rest and during the exercise-recovery experiment and Fig. 2 depicts the comparison between the groups.
Multivariate stepwise regression analysis of Q max including physiological and metabolic parameters derived from 31 P-MRS data acquired at rest, identified [Pi 2 ] (r 2 = 0.46, adjusted r 2 = 0.44, p = 0.0001) as the strongest and F ATP (r 2 = 0.54, adjusted r 2 = 0.50, p = 0.00001) as the second-strongest independent predictor of Q max . Detailed results are given in Table 2.

Discussion
In this study, we compared parameters of skeletal muscle metabolism, measured by static and dynamic 31 P-MRS methods, between a group of overweight-to-obese sedentary subjects, who are prone to diabetes, and a group of lean active individuals. We have found that the combination of increased BMI and sedentary lifestyle leads to significant differences in the alkaline Pi pool in skeletal muscle, as well as in other metabolic 31 P-MRS parameters, such as the concentration of PDE, the Pi-to-ATP metabolic flux, and mitochondrial capacity. In addition, significant correlations were found between the concentration of PDE, the alkaline Pi 2 /Pi ratio, and the resting Pi-to-ATP exchange rate and flux, measured by 31 P-MRS techniques at rest, and the maximal mitochondrial oxidative flux, measured by an exercise-recovery experiment. Dynamic 31 P-MRS provides a parameter closely related to training status, i.e., the mitochondrial capacity (Q max ) of the muscle tissue 15 . This was also demonstrated in our study, as the Q max of the overweight-to-obese sedentary subjects was significantly lower when compared to active, lean individuals. The correlation between Q max and BMI found in this study can be explained by the decreased physical activity in more obese individuals, as our regression analysis showed a primary connection of Q max with other parameters of muscle metabolism and not with BMI. This is in good agreement with a recent in vitro study, which found no differences in mitochondrial Scientific RepoRts | 6:20087 | DOI: 10.1038/srep20087 respiratory capacity and mitochondrial content in myocellular tissue samples between lean and obese subjects with similar training status 29 . Similarly, a different in vivo study did not find any changes in mitochondrial capacity in humans after weight reduction stimulated by diet-only; however, if combined with increased physical activity, an improvement in aerobic capacity was observed 30 .
Significantly higher myocellular PDE levels were found in our Ob/Sed subjects when compared to the L/Ac group. This was further supported with the positive correlation found between [PDE] and BMI. This is in good agreement with the finding of a previous report by Szendroedi et al. 26 in subjects with a comparable physical activity index. A correlation of [PDE] and age was also reported, in the current study and in 26,31 ; however, our regression analysis showed that physical activity and BMI, rather than age, primarily predict the PDE levels (data not shown). The increased spectral resolution of the 7 T MR system, used in our study, reveals that the measured [PDE] is mainly attributable to [GPC], with only a small contribution from glycero-phosphoethanolamine ([GPE]), and that, in fact, it is the [GPC] that is responsible for the differences between the two groups. This separation in PDE signals was not visible in the previous study performed at 3 T 26 .
The ratio of alkaline Pi 2 to cytosolic Pi (Pi 2 /Pi) was lower in the Ob/Sed group in comparison to the L/Ac group. Recently, van Oorschot et al. reported a dependence of Pi 2 /Pi, measured in the vastus lateralis, on the training status, when they compared highly trained runners with normally active individuals 28 . The potential influence of BMI was, however, not considered in the aforementioned study. The results of our study suggest such a dependence of the Pi 2 /Pi ratio in skeletal muscle on BMI, as a linear correlation between BMI and Pi 2 /Pi was found. Nevertheless, the results of our regression analysis identified only [GPC] and Q max as the primary predictors of the Pi 2 /Pi (data not shown). The differences in Pi 2 /Pi found between the groups can be directly attributed to the changes in [Pi 2 ], which was also significantly higher in the L/Ac group in comparison to the Ob/Sed group.
Although the mean F ATP values of our Ob/Sed group (F ATP = 0.25 ± 0.06 mM.s −1 ) are still above the decreased values reported previously in patients with type 2 diabetes (F ATP = 0.21 ± 0.05 mM.s −1 ) 25 the physical inactivity together with the overweight of our volunteers caused a significant reduction in the myocellular Pi-to-ATP metabolic flux, when compared to L/Ac individuals (F ATP = 0.31 ± 0.04 mM.s −1 ).

Interrelations between metabolic parameters measured by dynamic and static 31 P-MRS.
We report several correlations between the parameters of static 31 P-MR spectra, exchange rates and metabolite fluxes measured by ST at rest, and oxidative metabolism markers measured by exercise-recovery experiments. The alkaline Pi 2 resonance is suspected to represent mitochondrial density in the muscle tissue and depends on the amount of regular physical activity 28 . Its relation to training status was also confirmed in our study, as the [Pi 2 ], as well as Pi 2 /Pi, positively correlated with the maximal oxidative flux (Q max ), determined during the dynamic experiment, and also with the k ATP and F ATP , defining the Pi-to-ATP exchange rate and metabolic flux. Linear correlations between Q max and metabolic parameters measured by ST experiments at rest (i.e., k ATP and F ATP ) reported in our previous study on overweight-to-obese subjects 22 , were also found in this study combining the two different population groups. In addition, the multivariate regression analysis identified [Pi 2 ] and F ATP as independent predictors of Q max , suggesting the potential use of highlyspectrally resolved static 31 P-MRS at 7 T and ST as alternative techniques to dynamic exercise-recovery experiments.
The identified correlation between [PDE] and measured Pi-to-ATP metabolite flux (F ATP ) is in good agreement with a recent report by Szendroedi et al. 26 . Significant correlations were found also between [PDE] and mitochondrial capacity (Q max ), as well as other 31 [32][33][34][35][36][37] . In particular, Farber et al., studying a model of membrane defect of Alzheimer's disease, reported that an inhibition of oxidative phosphorylation causes accumulation of GPC through accelerated PC turnover 34 . Impaired oxidative metabolism and elevated PDE levels have been also reported in patients with spinal cord injury 35 and congenital lipodystrophy 36 . Muscular PDE content was also related to glucometabolic control in type II diabetes 26 . Furthermore, excessive amounts of PDE have been reported in fibromyalgia 33 , Duchenne muscular dystrophy 32 , or Becker muscular dystrophy 37 , connecting abnormal membrane metabolism with muscle dysfunction. Nonetheless, further investigations of this relation are still necessary.   The findings of this study support our previous report on correlations between dynamic and ST parameters in this Ob/Sed group 22 and provide additional information through analysis of Pi 2 /Pi and GPC, and moreover, by comparison to a lean active group of individuals. As to the technical limitations of our study, we should note that although care was taken to perfectly reposition the subject in the second MR system, when applicable, some small mislocalizations could not be fully excluded. The effect of individual anatomy must be also considered, as the localization through the sensitivity of the surface coil used in this study might cover different portions of the quadriceps muscles between subjects. Localization techniques, recently proposed for dynamic examinations of the lower leg muscles, e.g., frequency selective 31 P-MRI 38-40 , semi-LASER for single voxel localization 41 or depth-resolved surface coil MRS 42 , could be used in future studies to measure muscle-specific metabolism. However, the muscles of the quadriceps covered by the sensitivity volume of the used surface coil are all active   during knee-extension 43 , and, therefore, the inter-subject variability of the covered muscle volumes should have had only a minor effect on our results.
In conclusion, overweight-to-obese sedentary pre-diabetics exhibit increased concentrations of glycero-phosphocholine, a lower amount of alkaline Pi, a slower Pi-to-ATP exchange rate, and decreased mitochondrial capacity compared to lean active individuals. Associations found between the parameters of myocellular metabolism measured at rest and during exercise suggest that highly spectrally resolved static 31 P-MRS and saturation transfer measurements at rest could provide markers of muscle mitochondrial metabolism.

Methods
Fifteen young, overweight-to-obese, sedentary individuals (10/5 male/female; age 34.6 ± 7.1 years) with a body mass index (BMI) ≥ 27.0 kg.m −2 , a sedentary lifestyle without regular physical activity, no pharmacotherapy, and no medical history of type 2 diabetes were recruited for this study and classified as the overweight-to-obese/sedentary (Ob/Sed) group. Thirteen of these volunteers had already participated in our previous study on the interrelations between mitochondrial capacity and Pi-to-ATP exchange rates in this particular type of population 22 . Fifteen young, lean, physically active participants (10/5 male/female; age 29.3 ± 5.5 years) were recruited for the current study as the control lean/active (L/Ac) group.
Written, informed consent was obtained from each participant in the study after an explanation of the purpose, nature and potential risks of the study. The examination protocol was approved by the appropriate institutional ethical boards of the Medical University of Vienna and of the University Hospital Bratislava, Comenius University Bratislava, and the study was carried out in accordance with the approved guidelines.
Physiological tests. Within a week before the MR examination, the participants underwent a physical examination and physiological testing. BMI was measured by an analog weight scale and standard measuring tape. Bioelectric impedance, measured using an Omron BF511 (Omron Healthcare, Matsusaka, Japan), was used to evaluate total adiposity (%Fat) and to estimate the lean body mass (LBM). The maximal aerobic capacity (i.e., whole-body oxygen uptake [VO 2max ]) was measured during an incremental exercise test performed on a Lode Corival cycle ergometer (Lode, Groningen, The Nederlands). Continuous measurement of the gas exchange rate was obtained with the Ergostik (Geratherm Respiratory, Bad Kissingen, Germany), and the maximal oxygen consumption rate was expressed relative to LBM. The ergometry was performed at least three days prior to the MR examinations. The activity level was evaluated based on two working days and a weekend of accelerometer recordings and expressed as the number of steps per 24 hours. 31 P-MRS. Each participant underwent the entire MR examination protocol in one day, starting two hours after a standardized meal. The dynamic 31 P-MRS exercise-recovery experiment was performed on either a 7 T MR system (Magnetom, Siemens Healthcare, Erlangen, Germany) or a 3 T MR system (TIM Trio) from the same manufacturer, due to initial compatibility problems of our ergometer (Quadspect, Ergospect, Innsbruck, Austria) with the 7 T. Dual-tuned ( 31 P-1 H) circular surface coils (10 cm diameter, Rapid Biomedical, Rimpar, Germany), with similar sensitivity volumes 22 were used on both MR systems. The use of two MR systems, equipped with the same ergometer and surface coils with a similar sensitivity volume, has recently been shown to have no effect on the metabolic data derived from dynamic 31 P-MRS 44 .
Static 31 P-MRS experiments were performed exclusively at 7 T, as the increased spectral resolution is necessary for separation of the Pi 2 , as well as the GPE and GPC signals 27 , and the increase in signal-to-noise ratio allows significant reduction in measurement time of the ST experiment, compared to 3 T 45 . The subjects were investigated while lying inside the MR scanner with the surface coil fixed to the quadriceps femoris muscle (Fig. 1a) and the coil positions were marked to allow precise repositioning in the other MR system, if applicable. When the dynamic measurements were performed at 3 T (i.e., in case of first 10 Ob/Sed subjects), the order of examinations was randomized to allow simultaneous examinations of two subjects; otherwise the measurements at rest were always performed prior to the exercise-recovery experiment.
For the assessment of intramyocellular metabolite concentrations and the Pi 2 /Pi ratio, a pulse-acquire 31 P-MR spectrum (acquisition delay = 0.4 ms; repetition time = 15 s; bandwidth = 5 kHz; 16 averages in 4 minutes) was acquired at rest (Fig. 1b) and corrected for longitudinal relaxation times, as measured for 31 P muscle metabolites at 7 T 27,46 . The γ -ATP signal was used as an internal concentration reference, assuming a stable ATP concentration of 8.2 mM in the skeletal muscle 12 .
The exchange rate between ATP and Pi (i.e., ATP synthesis) was investigated using an ST experiment applying continuous irradiation, and the apparent longitudinal relaxation time (T 1 app ) was determined via an inversion recovery experiment, as described previously 45 . The total measurement time of the ST experiment was under 9 minutes.
The exercise-recovery protocol involved six minutes of repeated knee extensions against an air pressure, set to 30% of the maximal voluntary contraction force, once every repetition time (i.e., 2 s), followed by six minutes of recovery 15 . The volunteers were instructed by an audio signal to time the contraction-relaxation periods, so that the spectra were acquired always in the relaxed state of the muscle.
Analyses and calculations. Due to patient noncompliance (n Ob/Sed = 1) and technical problems (n Ob/Sed = 1), 28 complete datasets and one incomplete dataset (dynamic 31 P-MRS only), were available for analyses.
All acquired 31 P-MR spectra were analyzed using jMRUI software with the AMARES time domain fitting algorithm 47 . The resonance lines of PCr, two Pi signals, and two PDEs-glycero-phosphocholine (GPC) and glycero-phosphoethanolamine (GPE)-were fitted as single Lorentzians, whereas γ -and α -ATP were fitted as doublets and β -ATP as a triplet. The line width of the Pi 2 peak was constrained with respect to the line width of the main Pi peak, and the expected frequency difference between Pi 2 and Pi was set to ~0.4 ppm, to ensure a good Scientific RepoRts | 6:20087 | DOI: 10.1038/srep20087 fit for the Pi 2 peak 27,28 . The shift in resonance position between PCr and Pi signals in parts per million (δ ) was used to calculate intramyocellular pH 48 , according to the Henderson-Hasselbalch equation: pH = 6.75 + log((δ − 3.27)/ (5.63− δ )). The free cytosolic ADP concentration ([ADP]) was calculated according to the method described by Kemp et al. 49 , assuming that 15% of total creatine [Cr] was not phosphorylated in the resting state 50 .
The chemical exchange rate constant (k ATP ) was calculated from the fractional reduction of Pi magnetization upon selective saturation of γ -ATP (Fig. 1c) 45 . The resting unidirectional forward exchange flux was then calculated as F ATP = k ATP × [Pi].
To calculate the time constant of PCr resynthesis (τ PCr ), the PCr signal changes during the recovery period of the dynamic experiment (Fig. 1d) were fitted to a monoexponential function using MATLAB (MathWorks, Nattick, MA, USA). The initial PCr recovery rate (V PCr ), which roughly represents ATP turnover at the end of exercise, was determined and used to calculate the maximal rate of oxidative phosphorylation (Q max ) according to the ADP-based model of Michaelis and Menten 49 .
Data are presented as means ± standard deviations and compared between the groups by an unpaired Student t-test. The relationships between metabolic parameters were analyzed by linear regression analysis, using Pearson's correlation coefficient, to estimate the strength of the relationship. Multivariate stepwise regression analysis for the dependent variable Q max was performed using the independent variables (i.e., BMI, age,