Skeletal muscle ex vivo mitochondrial respiration parallels decline in vivo oxidative capacity, cardiorespiratory fitness, and muscle strength: The Baltimore Longitudinal Study of Aging

Summary Mitochondrial function in human skeletal muscle declines with age. Most evidence for this decline comes from studies that assessed mitochondrial function indirectly, and the impact of such deterioration with respect to physical function has not been clearly delineated. We hypothesized that mitochondrial respiration in permeabilized human muscle fibers declines with age and correlates with phosphocreatine postexercise recovery rate (kPCr), muscle performance, and aerobic fitness. Mitochondrial respiration was assessed by high‐resolution respirometry in saponin‐permeabilized fibers from vastus lateralis muscle biopsies of 38 participants from the Baltimore Longitudinal Study of Aging (BLSA; 21 men, age 24–91 years) who also had available measures of peak oxygen consumption (VO 2max) from treadmill tests, gait speed in different tasks, 31P magnetic resonance spectroscopy, isokinetic knee extension, and grip strength. Results indicated a significant reduction in mitochondrial respiration with age (p < .05) that was independent of other potential confounders. Mitochondrial respiratory capacity was also associated with VO 2max, muscle strength, kPCr, and time to complete a 400‐m walk (p < .05). A negative trend toward significance (p = .074) was observed between mitochondrial respiration and BMI. Finally, transcriptional profiling revealed a reduced mRNA expression of mitochondrial gene networks with aging (p < .05). Overall, our findings reinforce the notion that mitochondrial function declines with age and may contribute to age‐associated loss of muscle performance and cardiorespiratory fitness.


| INTRODUCTION
Mitochondrial function declines with age across multiple tissues, especially those with high energy demand such as skeletal muscle, brain, and heart (Gonz alez-Freire et al., 2015;Newgard & Pessin, 2014;Peterson, Johannsen & Ravussin, 2012). This functional decline is associated with substantial reduction in various biochemical properties including respiratory capacity, ATP production, and calcium handling capacity and content, as well as increased apoptosis (Gonz alez-Freire et al., 2015). Part of the age-related decline of mitochondrial function in skeletal muscle is due to a reduction in physical activity and can be rescued by exercise (Broskey et al., 2014;Kent and Fitzgerald, 2016;Larsen et al., 2012;Menshikova et al., 2006). However, physical fitness, which is strongly connected with mitochondrial respiration, declines with aging even in individuals who remain physically active (Distefano et al., 2017;Robinson et al., 2017). Therefore, the mechanisms and the extent by which chronological age "per se" affects the mitochondrial energy production remain unclear.
In previous studies, we assessed muscle mitochondrial function in a relatively large group of participants from the Baltimore Longitudinal Study of Aging (BLSA) using 31 P magnetic resonance spectroscopy (MRS) and showed that muscle bioenergetics, assessed as postexercise phosphocreatine recovery rate (kPCr), declines with aging and correlates with poorer performance in mobility tasks, especially those performed at fast pace or on long distances (Choi et al., 2016). Consistent with this study and using the same in vivo measurements, Zane et al. (2017) demonstrated that muscle strength significantly mediates the relationship of mitochondrial function with walking performance. However, studies that assessed mitochondrial electron transport chain respiration "ex vivo" by performing assays in permeabilized muscle fibers yielded controversial results concerning the effect of aging as well as correlations with physical capacity tests Coen et al., 2013;Distefano et al., 2017;Tyrrell et al., 2015). This discordance in the literature may be related to inherent methodological differences. 31 P MRS assesses mitochondrial capacity to generate ATP following muscle contraction in vivo and integrates all aspects of mitochondrial content and function including oxygen consumption, efficiency, and Ca 2+ handling, which together contribute to produce ATP. 31 P MRS may be sensitive to factors external to mitochondria such as adequate perfusion and oxygen delivery during and after exercise, changes in intracellular pH and other characteristics of the microenvironment. Therefore, it is unclear whether the decline in oxidative capacity observed in our previous studies is a direct consequence of an intrinsic impairment in the mitochondrial electron transport chain (ETC) with aging. Ex vivo respiration assays in permeabilized muscle fibers may address this problem as they specifically assess ETC function and are conducted in highly controlled conditions including saturating levels of oxygen and substrates.
To explore whether aging is associated with a primary intrinsic dysfunction of mitochondrial ETC respiratory capacity, we performed ex vivo mitochondrial respiration in permeabilized fibers from vastus lateralis muscle biopsies obtained from 38 BLSA participants who also had data available on treadmill peak oxygen consumption (VO 2max ), gait speed in multiple mobility tasks, isokinetic knee extension strength, and postexercise 31 P MRS. We hypothesized that ex vivo mitochondrial respiration correlates with in vivo measures of oxidative capacity and is also associated with muscle strength and performance in mobility tasks. Finally, we determined whether alterations in mitochondrial gene expression networks could support this changes in the mitochondrial phenotype observed with aging in skeletal muscle.

| Participant characteristics
The main characteristics of the study population, including demographics, physiological, and mitochondrial respiration parameters, are summarized in Table 1 and Table S1.

| Mitochondrial respiration and its association with aging and body composition
Mitochondrial respiration was carried out with permeabilized muscle fiber bundles from 38 participants using the Oxygraph-2k. The correlations and partial correlations between the different mitochondrial respiration measurements and age are summarized in Table 2. Sex or race did not significantly alter the mitochondrial respiration values, although a slightly higher mitochondrial respiration was observed in men compared to women (data not shown). State 4 respiration, defined as oxygen consumption before the addition of ADP, therefore in the absence of adenylates, and the respiration rate following the addition of the three lowest doses of ADP (31.25-62.5 lM) were not correlated with aging (p > .05). However, a statistically significant association between ADP concentrations ranging from 125 lM to 2 mM and aging was found, which was even more evident after adjustment for sex and BMI ( Figure 1). The measure of respiration that closely correlated with aging was obtained with 0.25 mM ADP (now referred as 5 ADP or Submaximal State 3) rather than the highest ADP concentration used (2 mM; defined as State 3 respiration; r = À.452, p = .004 versus r = À.375, p = .020). The significant difference in respiration between 0.25 and 2 mM ADP occurred independently of age, sex, or BMI (p < .05). These findings suggest that with aging, mitochondria become less able to increase oxidative phosphorylation under maximal ADP stimulation, perhaps because of a lower efficiency of the electron transport chain to keep up with the demand (Figure 2). Results were similar when calculating AUC of the entire titration curve from every participant using the trapezoidal formula (see method section). The AUC represents total mitochondrial oxygen consumption over time, for example, the ability to the mitochondria to produce ATP. A statistically significant negative association existed between mitochondrial oxygen consumption and aging that was independent of sex or BMI (r = À.381, p = .018).
Next, we examined the relationships between mitochondrial respiration and measurements of body composition such as height, weight, BMI, fat mass, and lean body mass. While none of the parameters of mitochondrial respiration were significantly correlated with height, weight, or lean body mass (p > .05), a borderline significant inverse association was observed between respiration at Submaximal State 3 and BMI or fat mass (r = À.292, p = .074; r = À.302, p = .065, respectively) after adjustment for age and sex ( Figure S2).

| Mitochondrial respiration and its association
with mobility, cardiorespiratory measurements, and muscle function Significant positive correlations between mitochondrial respiration and VO 2max , muscle strength, and mitochondrial oxidative capacity were observed, whereas mitochondrial respiration was negatively associated with the time to perform a 400-m walk (p < .05; Table 3; Figure S3). No significant association was found between mitochondrial respiration and gait speed assessed at the 6-m walking tasks (data not shown).   Table S4. A list of all GO Terms in the two pairwise comparisons is presented in Table S5.

| DISCUSSION
In this study, we tested the hypothesis that the intrinsic capacity of the mitochondrial electron transport chain declines with aging. Furthermore, we attempted to link such decline with clinically important F I G U R E 2 Association of the mitochondrial respiration derived from the different concentrations of ADP in function of age. Line graphs showing the p-values (upper panel) and r (bottom panel) of the bivariate association between ADP concentrations and age (model not adjusted, adjusted for sex, and then adjusted for sex and BMI). 1ADP to 8ADP, ADP concentrations of 31.25 lM to 1 mM; State 3: 2 mM; State 4: malate (5 mM), glutamate (10 mM), succinate (10 mM) measurements of muscle and physical function. Indeed, we found that, independent of sex and BMI, the capacity of the muscle mitochondria to produce energy is significantly lower at older ages.
These results are consistent with an age-associated decline of "in vivo" energetic capacity observed previously in BLSA participants (Choi et al., 2016;Zane et al., 2017). Specifically, we found that at incremental concentrations of ADP, the ability of the mitochondria to produce ATP was diminished in older adults (Figure 2), suggesting that with age mitochondria might lose ability to step up energy production when maximal performance is required, but show relatively normal function when the energetic demand is lower. Such a decline of energy efficiency at a high portion of the energy demand spectrum was previously hypothesized and is consistent with the results of this study, although the molecular mechanisms that explain this observation remain unknown (Porter et al., 2015). Further, we did not find any changes with age in the number of mitochondria (Figure S4); therefore, we believe that the reduction in mitochondrial respiration with age could be due to reduced mitochondrial oxidative phosphorylation efficiency rather than a reduction in mitochondrial number. Although controversial, these mechanisms have been proposed to explain the inability of old mitochondria to step up energy production while showing relatively normal function when the energetic demand is lower. These findings have been already observed in small series of muscle biopsies and may explain the inconsistencies across studies (Coen et al., 2013;Distefano et al., 2017;Tyrrell et al., 2015). We note that the observation of lower energetic capacity at the highest levels of demand is more consistent with intrinsic defect in oxidative phosphorylation when challenged above a certain threshold. For example, studies in mice have found that skeletal muscle mitochondria are organized in a highly reticular structure around the myofibrils (Glancy et al., 2015;Patel, Glancy & Balaban, 2016;Porter et al., 2015). Glancy et al. (2014) have hypothesized that this structure is essential to ensure that mitochondrial function is preserved at the center of the muscle fiber in condition of high energetic demand. A loss of integrity and connectivity of the mitochondrial network would lead to the observed age-related decline of energetic efficiency and adaptability to challenging conditions.  Recently, we have shown that (i) mitochondrial function correlates with insulin resistance and its severity in nondiabetic older persons, suggesting that decline in mitochondrial function occur in the early stage of development of diabetes type 2 (Fabbri et al., 2017) and (ii) that part of the decline in mitochondrial function is due to vidual, and therefore, the aging effect could be hidden by the heterogeneity of fitness across age in the studied sample. Thus, we believe that some of the age-related changes in mitochondrial function might be likely the result of changes in behavior, such as reduced physical activity, and might be muscle type dependent, as it is known that aging affect or modify the movement patterns, and therefore, the different muscles may be affected differently. Previous studies have shown that the tibialis anterior is one of the muscles that is more affected by changes in the gait pattern during aging (Kent and Fitzgerald, 2016). We did not find any association between mitochondrial respiration and gait speed, with or without adjustment for covariates.
These findings contrast with those reported by Tyrrell et al. (2015) who showed that mitochondria respiration parameters from isolated skeletal muscle mitochondria were associated with gait speed. Also, Choi et al. (2016) found that kPCr, a measurement in vivo of mitochondria function, is a predictor of gait speed in BLSA participants.
Again, our findings could be due to the heterogeneity of the effect in the analyzed samples. For example, it has been shown that the mitochondrial function is associated with walking performance in higher functioning active older adults, but not lower functioning sedentary adults (Santanasto et al., 2016). Interestingly, we found a significant negative correlation of respiration in State 4, Submaximal Similarly to previous studies, we found a nonsignificant negative trend of mitochondrial respiration with BMI, suggesting that mitochondrial respiration might be negatively affected by adiposity or, alternatively, may contribute to increased adiposity Coen et al. 2015). This trend supports the findings from  (Reznick et al., 2007;Sandri et al., 2004;Su et al., 2015;Zabielski et al. 2016). Notably, the age-related underrepresentation of the spliceosome pathway in the MA_Y and O_Y pairwise comparisons is of significance as this pathway has been recently found to be negatively impacted in the myonuclei proteome of old mice (Cutler et al., 2017). Interestingly, the upregulation of pathways related with muscle structure and remodeling was also found in our study. Sarcopenia is a process whereby agerelated decline in skeletal muscle mass and strength occurs, with the muscle undergoing morphological changes and physiological alterations mainly due to a reduction in physical activity. This, in turn, results in a remodeling of the motor unit and decline in the number of motor neurons that innervate the muscle fibers, particularly the type II fibers (Gonz alez-Freire, de Cabo, Studenski & Ferrucci, 2014). These, together with many other factors such as mitochondrial dysfunction, oxidative stress, inflammation, and lipid infiltration, play a key role in the musculoskeletal impairment that occurs with aging.
A noteworthy feature of this study is that measurements focused on several aspects of mitochondrial function were found to be significantly correlated. In Figure 4, it is quite clear that all the measurements related to oxidative capacity are well-clustered together.
Indeed, there was a positive association between two distinct mito-  Color intensity is proportional to the correlation coefficients (Betadine â ) and ethyl alcohol, and the outside areas covered with sterile drapes. The biopsy site was injected intradermally and subcutaneously with about 2 cc of 2% Lidocaine. Then, a 22 gauge 9 1 1/2 inch needle was introduced progressively just above the fascia (either sensed by the operator or at the depth indicated on the MRI image). This method ensured that the subcutaneous tissue and muscle fascia were anesthetized without distorting the muscle sample to be biopsied. A 6-mm Bergstrom biopsy needle was inserted through the skin in the muscle. Multiple muscle specimens were obtained until at least 250 mg of tissue was collected. A small portion of muscle tissue (~15 mg) was immediately placed in ice-cold respirometry buffer (BIOPS) for mitochondrial respiration analysis (see next section). The rest of the biopsy specimen was snap frozen in liquid nitrogen and subsequently stored at À80°C until used for further analyses.

| High-resolution respirometry
The muscle bundles were carefully dissected in ice-cold BIOPS buffer to eliminate intermuscular adipose and connective tissue using a dissecting microscope and sharp Dumont forceps. The fiber bundles were then permeabilized in freshly prepared saponin solution (5 lg/ ml in BIOPS buffer) for 30 min on wet ice with agitation after which they were transferred to a culture dish plate containing 2 ml buffer Z + EGTA on ice (3 9 10 min). Composition of buffer Z and BIOPS buffer was the same as previously described by Distefano et al. (2017). The wet weight of the fiber bundles was measured for data normalization. The fiber bundles were placed into the chambers of the Oxygraph-2k (O2k, Oroboros Instruments, Innsbruck, Austria) after air calibration. A summary of the protocol is shown in Figure S1.
The light in the chambers was turned off, and blebbistatin (Bleb), a myosin II ATPase inhibitor, capable of blocking spontaneous contraction of fiber bundles and mitigating the effects of contraction in respiratory kinetics (Perry et al., 2011), was added to each chamber.
After signal stabilization (~10-15 min), malate (5 mM), glutamate (10 mM), and succinate (10 mM) were added followed by the addition of 8 ADP titrations in increasing amounts (starting at 31.25 lM) until a final concentration of 2 mM, a concentration that we previously tested is saturating and elicits maximal State 3 respiration. Finally, cytochrome C (10 lM) was added at the end of the ADP titration curve to check the integrity of the outer mitochondrial membrane.
The fact that respiration did not increase by more than 15% indicated the lack of damage of the outer membrane. The oxygen flux levels were collected using DatLab 4 software (OROBOROS Instruments (2•60) = 8.33 (Gnaiger, 2009

| Mobility measurements
Details of the 400-m walk test were described elsewhere, and none of the participants had medical contraindications for this test (Newman et al., 2006;Simonsick, Montgomery, Newman, Bauer & Harris, 2001). Briefly, participants were instructed to walk for ten 40-m laps as quickly as possible receiving feedback and encouragement after completion of each lap. All participants completed the test. Lap time was recorded after the completion of each lap using a stopwatch.
Usual and rapid gait speed were measured on a 6-m course in an uncarpeted corridor. Participants were asked to walk at their usual, comfortable pace and then at the fastest possible pace. Times to complete the 6-m course were used in the analysis.

| Cardiorespiratory fitness
Whole-body aerobic capacity was determined by continuous measurement of oxygen consumption during a modified version of the treadmill Balke protocol (Fleg et al., 2005 Achieva MR scanner (Philips, Best, the Netherlands) using a method previously described (Choi et al., 2016;Zane et al., 2017). Briefly, the participants performed a rapid and intense ballistic knee extension exercise and a series of 31 P-MRS spectra were acquired before, during, and after the exercise using a 10-cm 31 P-tuned, flat surface coil (PulseTeq, Surrey, UK) secured over the vastus lateralis muscle of the left thigh. The duration of exercise was controlled to achieve depletion in phosphocreatine (PCr) signal amplitude to a value within 33% to 67% of the resting amplitude. Spectra were processed using The inverse of s PCr, k PCr is the PCr recovery rate constant determined as 1/s.

| Immunoblotting
Muscle homogenates were processed for Western blot as previously reported (Distefano et al., 2017 A total of 750 ng biotinylated RNA was hybridized to human HT-12v4 BeadChips (Illumina, San Diego, CA, USA). Following posthybridization rinses, arrays were incubated with streptavidin-conjugated Cy3 and scanned using an Illumina iScan scanner at 0.54 micron resolution. Hybridization intensity data were extracted from the scanned images using Illumina GenomeStudio software, V2011.1. Raw data were subjected to Z-normalization, as described elsewhere (Cheadle, Cho-Chung et al., 2003;. Principal component analysis (PCA) was performed on the normalized Z-scores of all of the detectable probes in the samples using DIANE 6.0 software (Cheadle, Vawter, Freed & Becker, 2003). Significant genes were selected by the z-test <0.05, false discovery rate <0.30, as well as z ratio >1.5 in both directions and ANOVA p-value <.05. Parametric analysis of gene set enrichment (PAGE; Broad Institute, M.I.T., Cambridge, MA) was analyzed as previously described (Kim & Volsky, 2005). All raw data are available in the Gene Expression Omnibus database (Accession No. GSE98613).

| Statistical analysis
Distributions of population characteristics were examined through histograms and boxplots. Characteristics of the population are reported as mean AE standard deviation (SD) or percentage. Bivariate correlations (Pearson correlations, partial correlations and linear regressions) were performed to test the relationships between the different mitochondrial respiration measurements with age and the other covariates of the analysis.
Multiple linear regressions were used to evaluate the relationship between the three most significant measurements of mitochondrial function (Model 1: State 4; Model 2: 5ADP; and Model 3: State 3) and the different outcomes: fitness (VO 2 max, Grip Strength, Leg Strength), kPCr, and mobility measurements (time in 400 m) independent of age, sex, and body composition. All linear regression coefficients were standardized so as to be comparable (Table S2). Hierarchical clustering correlation heatmap was used to show the association between all the covariates included in the analysis. Hierarchical clustering is used to classify the different outcomes based on their similarity into groups (clusters). This similarity is computed according to a distance between variables (correlation-based distances). The heatmap represents the Pearson correlation matrix across the different covariates included in the study. All analyses were performed using R version 3.3.1 (R Foundation for Statistical Computing, Vienna, Austria).