Insulin Resistance Is Associated With Reduced Mitochondrial Oxidative Capacity Measured by ³¹P-Magnetic Resonance Spectroscopy in Participants Without Diabetes From the Baltimore Longitudinal Study of Aging

Muscle biopsy specimens showed that individuals with type 2 diabetes have lower muscle mitochondrial content and oxidative capacity, mitochondrial proteins, and expression of oxidative genes than control subjects (1), but findings are inconsistent across studies (2,3), especially when accounting for obesity. Not all of the studies using ³¹P-magnetic resonance spectroscopy (³¹P-MRS) confirmed that individuals with type 2 diabetes have impaired mitochondrial function (4). Moreover, similar mitochondrial capacity was found in individuals with early-stage type 2 diabetes and prediabetes compared with normoglycemic control subjects (5,6). Thus, whether insulin resistance (IR) and prediabetes are linked to impaired mitochondrial capacity remains unclear. Inconsistent findings among studies may be due to small sample size, use of different diagnostic criteria, and inadequate adjustment for confounders (Supplementary Table 1).

Using data on ³¹P-MRS collected in 248 participants without diabetes in the Baltimore Longitudinal Study of Aging (BLSA), we tested whether impaired mitochondrial capacity is associated with prediabetes, IR, and the duration and severity of hyperglycemia exposure.

BLSA is a study of human aging established in 1958 by the National Institute on Aging Intramural Research Program (7). BLSA continuously enrolls healthy volunteers ≥20 years of age, followed at intervals of 1–4 years, with more frequent follow-up visits for older persons. The study protocol is reviewed by the National Institute of Environmental Health Sciences Institutional Review Board, and participants consent to participate after receiving an extensive description of the study.

We studied 248 BLSA participants without diabetes, who were 51–97 years old (mean age 73.5 ± 9.6 years, 46.4% men) with available data from ³¹P-MRS, oral glucose tolerance test (OGTT) results, and confounders. Information on race (Caucasian vs other), smoking (current smoker, former smoker, or no smoking), and physical activity (questionnaire on different activities transformed in metabolic equivalents of resting oxygen consumption) (8) were self-reported. Body weight in kilograms, and height and waist circumference in centimeters were assessed. BMI was calculated as weight/height (kg/m²). Total body DEXA was performed using the Prodigy Scanner (General Electric) to obtain measures of total body fat mass (TBFM) and trunk fat mass (FM) in kilograms. TBFM was also normalized by body weight and expressed as the percentage of body fatness (i.e., TBFM divided by body weight and multiplied by 100). Serum interleukin-6 (IL-6) and CRP were measured with ELISA (R&D Systems, Minneapolis, MN; and Alpco, Salem, NH, respectively). Plasma leptin level was measured using ELISA (LINCO Research, St. Charles, MO), whereas plasma adiponectin level was measured by radioimmunoassay (LINCO Research).

³¹P-MRS measurements of phosphorus-containing metabolites were obtained from the vastus lateralis muscle of the left thigh using a 3-T Achieva MRI scanner (Philips, Best, the Netherlands) and a 10-cm ³¹P-tuned, flat surface coil (PulseTeq, Surrey, U.K.). Participants performed rapid and intense ballistic knee extensions (9) to maximally recruit quadriceps and rapidly deplete phosphocreatine (PCr) with minimal acidification, which were practiced at least 15 min before starting the actual examination. Exercise duration was controlled to achieve a 33–67% reduction in PCr peak height and never exceeded 42 s, resulting in a postexercise recovery period of 5.8–6.3 min, which was sufficient for reliable fitting of the recovery coefficient. MRI signals were acquired before, during, and after this exercise protocol, with a repetition time of 1.5 s using a pulse-acquire sequence adiabatic radiofrequency excitation over a period of 7.5 min, starting 1 min before exercise to define baseline values, with a time resolution of 6 s (10). Data were analyzed by jMRUI (version 5.0), and metabolites were quantified using a nonlinear least squares algorithm (AMARES), based on the integrated area of the spectral line (11,12). The postexercise PCr recovery time constant (τ_PCr) was calculated by fitting the time-dependent recovery of PCr to the monoexponential recovery function, as follows:

where PCr₀ represents the amount of PCr immediately after in-magnet exercise, and ΔPCr is the difference in PCr between baseline and postexercise conditions.

Muscle pH was monitored during the protocol based on the chemical shift difference between inorganic phosphate (Pi) and PCr. Subjects showing pH values of <6.8 were excluded from further analysis (13).

Participants were classified as being “normal,” having prediabetes, or having diabetes, according to the American Diabetes Association criteria using fasting plasma glucose (FPG) and/or 2-h post-OGTT glucose (2hG) levels (14). For OGTT, fasting participants consumed 75 g of an oral solution of glucose, and blood samples were collected after fasting and then every 20 min for 2 h. Type 2 diabetes was defined as history of type 2 diabetes and taking hypoglycemic medications; or FPG ≥126 mg/dL and 2hG ≥200 mg/dL at the same visit; or FPG ≥126 mg/dL or 2hG ≥200 mg/dL at consecutive visits. Participants with diabetes were excluded from the current study. Prediabetes was defined as follows: FPG ≥100 mg/dL but <126 mg/dL or 2hG ≥ 140 mg/dL but <200 mg/dL at the same visit; or FPG ≥100 mg/dL but <126 mg/dL or 2hG ≥140 mg/dL but <200 mg/dL during two consecutive visits. IR was estimated by HOMA-IR, calculated as FPG (mg/dL) × fasting insulin (mU/L) divided by 405 (15). Insulin sensitivity was assessed by the Matsuda index (10,000/square root of [(fasting glucose × fasting insulin) × (mean glucose × mean insulin during OGTT)]) (16) and the Reduced Matsuda Index (10,000/square root of [(fasting glucose × fasting insulin) × (glucose at 120 min × insulin at 120 min)]) (17). Duration of hyperglycemia exposure was assessed as the number of years from the onset of prediabetes. Severity of hyperglycemia exposure was estimated for participants with longitudinal data (N = 205) as the average OGTT 2-h glucose level from the time of the first OGTT to mitochondrial capacity assessment.

Variables were summarized as the mean ± SD, median (interquartile range), or number (percentage). The relationships between τ_PCr (in seconds) and prediabetes, HOMA index, Matsuda index, and reduced Matsuda index were explored using Spearman correlations. Logistic and linear regression models were fitted to test the associations of mitochondrial oxidative capacity with prediabetes, IR/insulin sensitivity, and the severity and duration of hyperglycemia exposure. Covariates included age, sex and percentage of PCr depletion (model I); race, education, smoking status, physical activity level, body weight, body height, and DEXA-measured trunk FM (model II); and levels of plasma IL-6, CRP, leptin, and adiponectin (model III [or the fully adjusted model]). Alternative anthropometric and body composition measures were also tested as confounders (model IV–VII). All analyses were performed by the SAS statistical package, version 9.3 (SAS Institute Inc., Cary, NC) and R version 3.1.2 (R Foundation for Statistical Computing, Vienna, Austria).

The prevalence of prediabetes was 40.3%. The median time from the onset of diabetes was 14 years. The main characteristics of the population according to prediabetes status are presented in Table 1. τ_PCr was significantly higher in subjects with prediabetes compared with normal subjects and according to HOMA tertiles, and was significantly lower according to tertiles of the Matsuda index and reduced Matsuda index (Fig. 1). Independent of all confounders, τ_PCr significantly and positively correlated with prediabetes and HOMA, whereas it correlated negatively with the Matsuda index (Supplementary Table 2). Participants with higher τ_PCr values were more likely to have prediabetes and to be in a higher tertile of IR (odd ratios >1) and were less likely to be a higher tertile of insulin sensitivity (odd ratios <1) than those with lower τ_PCr values (Table 2). Comparing tertiles of τ_PCr, both the highest and the medium tertiles were significantly associated with prediabetes and impaired insulin sensitivity compared with the lowest tertiles (Supplementary Table 3).

Relationships of τ_PCr with the Matsuda index and the reduced Matsuda index did not depart from linearity (Supplementary Fig. 1), whereas for HOMA-IR a cubic spline provided the best fit (Supplementary Table 4). Using OGTT-based surrogate indices developed by Abdul-Ghani et al. (18), we found that τ_PCr was not associated with liver IR, but was inversely and significantly correlated with muscle insulin sensitivity. Adjusting for physical activity and/or trunk FM, the association was attenuated and no longer significant (Supplementary Table 5).

Adjusting for confounders, and greater severity and longer duration of hyperglycemia exposure, even when both were included as predictors in the same model, were independently associated with greater mitochondrial impairment (Table 3).

Using data from 248 BLSA participants with diabetes, we examined the relationship between IR and ³¹P-MRS–measured mitochondrial oxidative capacity. Independent of confounders, a larger τ_PCr value was significantly associated with a higher likelihood of having prediabetes, more severe HOMA-IR, and lower insulin sensitivity (Matsuda index). The severity and duration of hyperglycemia exposure were independently associated with larger τ_PCr values.

Time course studies in rodents suggest that mitochondrial impairment follows diet-induced diabetes (3). However, human studies showed mitochondrial impairment in insulin-resistant offspring of patients with types 2 diabetes, suggesting that impaired oxidative capacity may drive IR (19). Weight loss has no effect on mitochondrial capacity (20), but Coen et al. (21) recently found that exercise superimposed on bariatric surgery enhances mitochondrial respiration and improves insulin sensitivity in patients with severe obesity. Similarly, exercise in overweight-to-obese older adults improves insulin sensitivity and increases muscle oxidative capacity (1), perhaps by enhancing mitochondrial biogenesis. Of note, prolonged exercise training can, at least in part, reverse mitochondrial impairment associated with a long-standing diabetic state (6).

Our study suggests that mitochondrial impairment occurs before IR progression to type 2 diabetes. Consistently, in a small group of older adults who did not have diabetes but had IR, a significant reduction in ³¹P-nuclear magnetic resonance saturation transfer (ST)–measured muscle phosphorylation activity was found compared with young control subjects (22). Our study further supports these findings in a larger and better characterized sample population, accounting for the potential confounding variables of body composition and physical inactivity, and demonstrating the independent association of both the severity and duration of hyperglycemia with reduced mitochondrial capacity. Noteworthy, ST is a technique that allows measurement of the ATP synthesis rate at rest and minimizes the variability of exercise. The postexercise PCr synthesis used in our study expresses the capacity to respond to an energetic challenge and may be more sensitive to even minor metabolic derangements than ST (1).

The mechanism by which mitochondrial impairments could lead to IR is unclear. In impaired mitochondria, the flux of electrons in the electron transport chain is impaired, and spare electrons are more likely to leak in the mitochondrial matrix and produce radical oxygen species. The resulting oxidative stress may further damage mitochondria and affect IR directly or indirectly by triggering a local inflammatory response (4). It also is possible that hyperglycemia plays a key role in mitochondrial impairment. Patients with well-controlled type 2 diabetes have diminished muscle ATP synthesis despite normal glucose transport/phosphorylation (23). In addition, ³¹P-nuclear magnetic resonance ST–measured mitochondrial activity correlates positively with insulin sensitivity and negatively with insulin secretion, suggesting that reduced insulin signaling might limit mitochondrial oxidative rates (24). Also, mitochondrial activity negatively correlates with hepatic lipid content, supporting a link between altered muscle and liver energy metabolism in promoting IR (24). However, short-term elevation of plasma lipid levels does not affect ATP synthase in human skeletal muscle, suggesting that lipid-induced IR is unlikely to be due to a direct effect of lipid species on mitochondrial function (25).

A major limitation of our study is the cross-sectional design. Hyperinsulinemic-euglycemic clamp would have been more accurate and robust than surrogate indices to evaluate insulin sensitivity, but is time consuming and difficult to perform in large epidemiological studies (18), like the BLSA. Moreover, the severity of hyperglycemia exposure was estimated from OGTT 2-h glucose levels rather than from fasting glucose levels, but the fasting glucose level predominantly reflects liver IR, rather than muscle IR. Also, the severity of hyperglycemia exposure was based on different lengths of follow-up across participants, but we accounted for that in the analysis. Furthermore, we could not investigate race differences in mitochondrial function since BLSA participants are mostly Caucasian. Objective measurements of physical activity by accelerometers would have been more accurate than self-reports, but the accelerometer was recently introduced in the BLSA and was not available for the current sample. Finally, investigating the relationship between mitochondrial function and intramyocellular lipid content would be interesting, but these data are still unavailable in the BLSA.

Despite limitations, our study suggests that reduced mitochondrial oxidative capacity is an integral part of the IR in adult and older persons, and should be considered a possible therapeutic target to reduce the consequences of this condition and prevent progression to type 2 diabetes.

Funding. This research was supported entirely by the Intramural Research Program of the National Institutes of Health, National Institute on Aging. Data for these analyses were obtained from the Baltimore Longitudinal Study of Aging, a study performed by the National Institute on Aging (03-AG-0325).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. E.F. contributed to the analysis and interpretation of the data, statistical analysis, and drafting of the manuscript. C.W.C. contributed to the acquisition, analysis, and interpretation of the data and to critical revision of the manuscript for important intellectual content. R.G.S. and D.A.R. contributed to the acquisition of the data and critical revision of the manuscript for important intellectual content. K.W.F., D.C., and A.C.Z. contributed to the acquisition of the data. Z.A.M., M.G.-F., M.Z., S.A.S., R.R.K., and J.M.E. contributed to the critical revision of the manuscript for important intellectual content. L.F. contributed to the study concept and design; study supervision; acquisition, analysis, and interpretation of the data; statistical analysis; drafting of the manuscript; and critical revision of the manuscript for important intellectual content. E.F. and L.F. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.