sign in sign up
<<
Home , Insulin

Pathophysiology/Complications ORIGINAL ARTICLE

Relationship Between Levels, Sensitivity, and Mitochondrial Function in Men

1 4 NELLY PITTELOUD, MD DEVJIT TRIPATHY, MD, DM nsulin resistance has assumed increas- 2 1 VAMSI K. MOOTHA, MD MARIA YIALAMAS, MD ing importance as a risk factor for car- 1 4 ANDREW A. DWYER, BA LEIF GROOP, MD, PHD 1 diovascular disease coincident with the EGAN ARDIN BA 5 I M H , DARIUSH ELAHI, PHD dramatic increase in the prevalence of 3 1 HANG LEE, PHD RANCES AYES MB, BCH, BAO 4 F J. H , in the Western world. Recent KARL-FREDRIK ERIKSSON, MD studies using genetic analysis (1,2), func- tional imaging (3,4), and animal models of differential aerobic capacity (5) have shed light on the role of mitochondrial function in inducing the metabolic distur- OBJECTIVE — The goal of this study was to examine the relationship between serum bances characteristic of insulin-resistant testosterone levels and insulin sensitivity and mitochondrial function in men. states. Using set enrichment analysis RESEARCH DESIGN AND METHODS — A total of 60 men (mean age 60.5 Ϯ 1.2 to look for changes in sets of com- years) had a detailed hormonal and metabolic evaluation. Insulin sensitivity was measured using piled on the basis of function, Mootha et a hyperinsulinemic-euglycemic clamp. Mitochondrial function was assessed by measuring max- al. (1) showed decreased maximal aerobic O imal aerobic capacity (VO2max) and expression of oxidative genes in skeletal capacity (V 2max) and decreased expres- muscle. sion of mitochondrial genes involved in oxidative phosphorylation (OXPHOS) in RESULTS — A total of 45% of subjects had normal tolerance, 20% had impaired men of Northern European descent with glucose tolerance, and 35% had type 2 . Testosterone levels were positively correlated impaired glucose tolerance (IGT) and with insulin sensitivity (r ϭ 0.4, P Ͻ 0.005). Subjects with hypogonadal testosterone levels (n ϭ . Using quantitative real- 10) had a BMI Ͼ25 kg/m2 and a threefold higher prevalence of the than time PCR, Patti et al. (2) reported de- their eugonadal counterparts (n ϭ 50); this relationship held true after adjusting for age and sex creased OXPHOS in a ϭ –binding globulin but not BMI. Testosterone levels also correlated with VO2max (r Mexican-American population of type 2 Ͻ ϭ Ͻ 0.43, P 0.05) and oxidative phosphorylation gene expression (r 0.57, P 0.0001). diabetic subjects as well as insulin- resistant first-degree relatives of type 2 di- CONCLUSIONS — These data indicate that low serum testosterone levels are associated abetic subjects with normal glucose with an adverse metabolic profile and suggest a novel unifying mechanism for the previously tolerance (NGT). Magnetic resonance independent observations that low testosterone levels and impaired mitochondrial function promote in men. spectroscopy has shown that decreased mitochondrial oxidative and phosphory- Diabetes Care 28:1636–1642, 2005 lation activity underlies the insulin resis- tance seen with aging (3) and in lean offspring of patients with type 2 diabetes (4). Recent data from rats bred to have low aerobic capacity also implicate im- ●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●● paired mitochondrial function in the de- From the 1Reproductive Endocrine Unit of the Department of Medicine, Massachusetts General Hospital, velopment of the cardiovascular risk Boston, Massachusetts; the 2Broad Institute, Harvard University and Massachusetts Institute of Technology, profile characteristic of the metabolic syn- Cambridge, Massachusetts; the 3Department of Biostatistics and General Clinical Research Center, Massa- drome (5). chusetts General Hospital, Boston, Massachusetts; the 4Department of , Wallenberg Labora- Little is known about the interaction 5 tory, University Hospital MAS, Lund University, Malmo, Sweden; and the Department of Surgery, between testosterone levels and insulin University of Massachusetts Medical Center, Worcester, Massachusetts. Address correspondence and reprint requests to Frances J. Hayes, MD, Reproductive Endocrine Unit, sensitivity in men, in contrast to the abun- BHX 511 Massachusetts General Hospital, 55 Fruit St., Boston, MA 02114. E-mail: [email protected]. dant literature on this relationship in Received for publication 11 November 2004 and accepted in revised form 22 March 2005. women (6). Cross-sectional studies dem- L.G. has served on an advisory panel for Bristol-Myers Squibb. ␣ onstrate an inverse relationship between Abbreviations: E2, ; IGT, impaired glucose tolerance; NGT, normal glucose tolerance; PGC-1 , peroxisome proliferator–activated -␥ coactivator; OXPHOS, oxidative phosphorylation; SHBG, sex testosterone and fasting insulin levels in hormone–binding globulin; UQCRB, ubiquinol cytochrome c reductase–binding ; WHR, waist-to- men independent of age, obesity, and hip ratio. body distribution (7–11). A link be- A table elsewhere in this issue shows conventional and Syste`me International (SI) units and conversion tween testosterone deficiency and diabe- factors for many substances. tes has also been suggested with the © 2005 by the American Diabetes Association. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby demonstration that men with type 2 dia- marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. betes have lower testosterone levels than

1636 DIABETES CARE, VOLUME 28, NUMBER 7, JULY 2005 Pitteloud and Associates weight-matched nondiabetic control sub- provided written informed consent be- were 497 Ϯ 15 pmol/l and 1,188 Ϯ 75 Ϫ jects (12,13). In addition, six large pro- fore the initiation of any study proce- pmol/l for insulin doses of 40 mU m 2 Ϫ Ϫ Ϫ spective studies have shown that low dures. min 1 and 80 mU m 2 min 1, respec- testosterone levels predict development tively. All statistical analyses using insulin of type 2 diabetes in men (14–19). Two Anthropometric measures sensitivity as a parameter were corrected studies demonstrate a positive relation- Height and weight were measured by for insulin levels during the clamp. ship between total testosterone levels and standard procedures to calculate BMI as insulin sensitivity in normal (20) and di- weight (kg) divided by height squared Maximal aerobic capacity 2 abetic men (21). In contrast, data on the (m ). Waist-to-hip ratio (WHR) was cal- VO2max was measured in the Swedish sub- relationship between free testosterone culated by measuring waist circumfer- jects using an incremental work- levels and insulin sensitivity are conflict- ence at the level of the umbilicus and hip conducted upright test with a ing, with two studies showing no correla- circumference at the level of the greatest bicycle ergometer (Monark Varberg, Swe- tion (21,22), whereas a third study hip girth. In the 42 Swedish subjects, per- den) combined with continuous analysis demonstrates a weak positive relationship cent body fat was measured using bioel- of expiratory gases and minute ventila- (20). ectrical impedance analysis. tion. Exercise was started at a workload Given that low testosterone levels varying between 30 and 100 W depend- predict development of type 2 diabetes in Oral ing on the history of endurance training men (14–19) and that aging is accompa- A standardized 2-h oral glucose tolerance or exercise habits and increased by 20–50 nied by insulin resistance and a decline in test using 75 g glucose was performed on W every 3 min, until a perceived exhaus- testosterone (23–25), we hy- the basis of which subjects were classified tion or a respiratory quotient of 1.0 was pothesized that testosterone is an impor- as having NGT, IGT, or type 2 diabetes reached. Maximal aerobic capacity was tant modulator of insulin sensitivity and using 1985 criteria from the World defined as the VO2 during the last 30 s of mitochondrial function in men. Thus, the Health Organization (27). exercise and is expressed per lean body aims of this study were to 1) examine the mass calculated by bioelectrical imped- relationship between serum testosterone Insulin sensitivity ance analysis. levels and insulin sensitivity in men Insulin sensitivity was determined by the across a wide spectrum of insulin sensi- hyperinsulinemic-euglycemic clamp Percutaneous muscle biopsy tivity, 2) dissect the role of obesity and technique (28). For 3 days before the Percutaneous muscle biopsies (20–50 alterations in sex hormone–binding glob- study, all subjects consumed a weight- mg) were taken from the vastus lateralis ulin (SHBG) levels in mediating this rela- maintaining containing 300 g carbo- muscle of the Swedish subjects. Biopsies tionship, and 3) determine if there is an hydrate per day. The clamp studies were were performed under local anesthesia association between serum testosterone performed at 7:30 A.M. after a 12-h fast. (1% lidocaine) using a Bergstro¨m needle levels and mitochondrial function re- An intravenous cannula was inserted into after the hyperinsulinemic-euglycemic flected by VO2max and OXPHOS gene ex- an antecubital vein for the infusion of in- clamp. Fiber-type composition and pression in . sulin and glucose. A second was whole-muscle concentration inserted retrogradely into a hand vein for were determined as previously described RESEARCH DESIGN AND blood sampling; the hand was kept (29). Quantification and calculation of METHODS — A total of 60 men aged heated in a warming chamber to arterial- the fibers was performed using the COM- 39–69 years (mean 60.5 Ϯ 1.2) were en- ize venous samples. Blood glucose was FAS image analysis system (Scan Beam, rolled in this study. Subjects with a his- measured every 5 min for 120 min. Forty- Hadsun, Denmark). tory of testicular disorders or who were two clamps were performed with an insu- Ϫ Ϫ taking known to interfere lin infusion rate of 40 mU m 2 min 1 Gene set enrichment analysis with testosterone secretion/action or glu- with use of [3-3H]glucose to correct for Total RNA was isolated from the muscle cose were excluded from the hepatic glucose production. There were biopsies, and the targets were hybridized study. Healthy normal men, obese men, 18 clamps performed with an insulin in- to the Affymetrix HG-U133A chip. Ϫ Ϫ and men with newly diagnosed type 2 di- fusion rate of 80 mU m 2 min 1, a dose Marker analysis was first performed using abetes on no hypoglycemic agents were sufficient to suppress hepatic glucose pro- GeneCluster as previously described (1). enrolled to cover the full spectrum of in- duction. An infusion of 20% glucose was The software package SAM was used to sulin sensitivity. A total of 42 Caucasian initiated and titrated to maintain blood identify genes on the microarrays with men, who were participants in the Malmo glucose at the fasting level. The glucose statistically significant changes in expres- Prevention Study (26), were studied in disposal rate was determined during the sion. The approach of gene set enrich- Sweden. Some of the metabolic character- last 30 min of the 2-h clamp. Under ment analysis was then used to analyze istics of this Swedish cohort have already steady-state conditions of euglycemia, the previously defined sets of functionally re- been reported (1). There were 18 subjects rate of exogenous glucose infusion, cor- lated genes and search for systematic ex- enrolled in the U.S., of whom 16 were rected for glucose space and hepatic glu- pression differences among the genes in Caucasian and 2 were African-American. cose production, is equal to the rate of the set according to glucose tolerance. The study was approved by the Ethics insulin-stimulated glucose disposal, ex- Using this technique, a subset of Ϫ Committee at Lund University and the pressed as the M value (mg kg 1 genes involved in oxidative phosphoryla- Ϫ Research Committee of Massa- min 1), a measure of insulin sensitivity. tion (OXPHOS-CR) was recently identi- chusetts General Hospital. All subjects Mean insulin levels during the clamp fied that is tightly coregulated across

DIABETES CARE, VOLUME 28, NUMBER 7, JULY 2005 1637 Testosterone and mitochondrial function in men

Table 1—Clinical and biochemical characteristics of the study population centrations of testosterone and SHBG and measures of adiposity (BMI, WHR, and Testosterone Testosterone percent body fat), insulin sensitivity (M), All subjects Ͻ9.7 nmol/l Ն9.7 nmol/l P and mitochondrial function (VO2max,OX- PHOS gene expression). Multiple regres- n 60 10 50 — sion analysis was performed to control for Age (years) 60 Ϯ 157Ϯ 3.5 61 Ϯ 1NSpotential confounding variables includ- NGT (n)27126ing age, insulin, SHBG, BMI, WHR, and IGT (n)1239percent body fat. Testosterone levels were Type 2 diabetes (n)21615 analyzed as both a continuous and binary Body composition variable (hypogonadal ϭ testosterone BMI (kg/m2) 27 Ϯ 0.6 32 Ϯ 226Ϯ 0.5 0.0004 Ͻ9.7 nmol/l and eugonadal ϭ testoster- WHR 0.96 Ϯ 0.01 1.02 Ϯ 0.05 0.95 Ϯ 0.01 0.008 one Ն9.7 nmol/l). All analyses used study Body fat (%) 24.0 Ϯ 0.97 30.1 Ϯ 2.8 23 Ϯ 0.97 0.008 site as a covariate to correct for differences Biochemical characteristics between the Swedish and U.S. cohorts. Testosterone (nmol/l) 16 Ϯ 0.8 6.8 Ϯ 0.5 17 Ϯ 0.6 Ͻ0.0001 SHBG (nmol/l) 36 Ϯ 220Ϯ 339Ϯ 2 0.0008 Fasting glucose (mmol/l) 5.9 Ϯ 0.2 7.5 Ϯ 0.6 5.6 Ϯ 0.3 0.017 RESULTS — The characteristics of the Fasting insulin (pmol/l) 66 Ϯ 497Ϯ 360Ϯ 4 0.04 study population are provided in Table 1. Cholesterol (mmol/l) 5.3 Ϯ 0.14 5.9 Ϯ 0.4 5.2 Ϯ 0.1 NS Of the subjects, 45% had NGT, 20% had (mmol/l) 1.4 Ϯ 0.1 2.3 Ϯ 0.5 1.3 Ϯ 0.1 0.01 IGT, and 35% had type 2 diabetes. Body HDL cholesterol (mmol/l) 1.3 Ϯ 0.05 1.0 Ϯ 0.8 1.32 Ϯ 0.05 NS composition varied significantly, with a Free fatty acids (␮mol/l) 509 Ϯ 24 516 Ϯ 68 507 Ϯ 25 NS range of BMI from 18.7 to 46.3 kg/m2, Glucose clamp data WHR from 0.67 to 1.13, and percent Insulin sensitivity (mg 6.5 Ϯ 0.4 3.6 Ϯ 0.6 7.3 Ϯ 3 0.0007 body fat from 12.3 to 29.1. Serum testos- Ϫ Ϫ kg 1 min 1) terone levels covered the spectrum from Data are means Ϯ SE. P values represent differences between subjects with testosterone levels Ͻ9.7 vs. Ն9.7 hypogonadal to eugonadal, ranging from nmol/l. 3 to 31 nmol/l (normal range 9.7–34.6). Testosterone levels were lower in men with IGT and type 2 diabetes than in those many tissues, is highly expressed in sites matography (Esoterix, Calabasas Hills, with NGT (13.5 Ϯ 6 and 13.1 Ϯ 4 vs. Ϯ Ͻ of insulin-mediated glucose disposal, and CA). The E2 assay has a sensitivity of 18 18 6.5 nmol/l, respectively P 0.05). for which expression is decreased by pmol/l and, based on a male serum pool, The glucose clamp studies revealed a ϳ20% in skeletal muscle of men with IGT has an intra-assay CV of 4.9% and an in- spectrum of insulin sensitivity, with insu- and type 2 diabetes (1). Of the 34 genes in terassay CV of 15%. Plasma glucose was lin sensitivity values ranging from 1.4 to Ϫ Ϫ the OXPHOS-CR set, the top-ranking measured with the glucose oxidase 13.3 mg kg 1 min 1. gene (i.e., the gene with the largest ex- method (Beckman Instruments, Fuller- A negative correlation was observed pression difference between normal and ton, CA). Insulin was measured by RIA between insulin sensitivity and indexes of diabetic muscle) was ubiquinol cyto- using 125I-labeled human insulin and hu- obesity, including BMI (r ϭϪ0.58, P Ͻ chrome c reductase–binding protein man insulin antiserum (Linco Research, 0.0001), WHR (r ϭϪ0.6, P Ͻ 0.0001), (UQCRB). In the present study, the data St. Charles, MO). SHBG was measured by and percent body fat (r ϭϪ0.4, P ϭ on OXPHOS-CR gene expression were a chemiluminescent immuno- 0.009). There was an inverse relationship analyzed to determine if there was an as- metric assay (DPC; Immulite, Los Ange- between testosterone and BMI (r ϭϪ0.5, sociation with serum testosterone levels. les, CA), which has an intra-assay CV of P Ͻ 0.0001), testosterone and WHR (r ϭ Ͻ7% and an interassay CV of Ͻ8%. Free Ϫ0.38, P Ͻ 0.005), and testosterone and Biochemical analysis fatty acids were measured by an enzy- percent body fat (r ϭϪ0.5, P Ͻ 0.0001). Fasting blood samples collected before matic colorimetric method (WAKO All subjects with hypogonadal testoster- the start of the glucose clamp at Ϫ30, Chemicals, Richmond, VA). one levels had a BMI Ͼ25 kg/m2 and a Ϫ20, and Ϫ10 min were pooled and used WHR Ͼ0.9. An inverse relationship was to determine serum levels of testosterone, Statistical analysis also observed between SHBG and BMI Ϯ ϭϪ Ͻ estradiol (E2), SHBG, total and HDL cho- Data are presented as means SE. Differ- (r 0.6, P 0.001), SHBG and WHR lesterol, triglycerides, and free fatty acids. ences in the baseline characteristics of (r ϭϪ0.46, P Ͻ 0.001), and SHBG and eugonadal and hypogonadal subjects percent body fat (r ϭϪ0.4, P ϭ 0.008). A Immunoassays were compared with a two-sample t test positive correlation was observed be- Serum testosterone concentrations were for continuous variables; categorical vari- tween testosterone levels and insulin sen- measured using the DPC Coat-A-Count ables were compared using the Fisher’s sitivity (r ϭ 0.4, P Ͻ 0.005; Fig. 1) and RIA , which has an intra- and interassay exact test. A P value Ͻ0.05 was consid- between SHBG and insulin sensitivity Ͻ ϭ Ͻ coefficient of variation (CV) of 10%. E2 ered statistically significant. Pearson’s (r 0.44, P 0.005; Fig. 1). In contrast was measured with an RIA using hexane correlation coefficients were used to as- to testosterone, no significant relation- ethylacetate extraction and LH-20 chro- sess the relationship between serum con- ship was observed between E2 levels and

1638 DIABETES CARE, VOLUME 28, NUMBER 7, JULY 2005 Pitteloud and Associates

Figure 1—Correlation between insulin sensitivity (M) and serum testosterone (T) levels (A) and SHBG levels (B) in 60 men; 27 had NGT ( ), 12 had IGT (‚), and 21 had type 2 diabetes (F). Shaded area represents values for subjects with hypogonadal testosterone levels, i.e., Ͻ9.7 nmol/l. In a subset of 42 men, serum testosterone levels were correlated with maximal aerobic capacity (VO2max)(C) and expression of UQCRB in skeletal muscle (D); 17 men had NGT ( ), 7 had IGT (‚), and 18 had type 2 diabetes (F).

BMI (r ϭ 0.19), WHR (r ϭ 0.1), or insulin (P Ͻ 0.05). The relationship between tes- pogonadal testosterone level had NGT sensitivity (r ϭ 0.14). tosterone and insulin sensitivity was no compared with 26 of 50 eugonadal men. As depicted in Fig. 1, the relationship longer significant after adjusting for BMI Expression of OXPHOS-CR genes in ϭ ϭ between testosterone levels and insulin (P 0.23), percent body fat (P 0.17), skeletal muscle correlated with VO2max sensitivity suggests the presence of two or WHR (0.38) expressed as continuous (r ϭ 0.5, P Ͻ 0.001) as previously re- distinct populations. All subjects with a variables. However, body composition ported (1) and with insulin sensitivity serum testosterone level Ͻ9.7 nmol/l had had less of an impact on this relationship (r ϭ 0.33, P Ͻ 0.05). A stronger relation- a low insulin sensitivity value. In contrast, when analyzed using categorical vari- ship was observed between the top- men with a low insulin sensitivity value ables, i.e., BMI Ն or Ͻ30 kg/m2 (P ϭ ranking OXPHOS-CR gene, UQCRB, and Ն Ͻ ϭ ϭ Ͻ exhibited a full range of serum testoster- 0.09), body fat or 25% (P 0.07), or both VO2max (r 0.57, P 0.0001) and one levels. Subjects with hypogonadal WHR Ͼ or Յ0.9 (P ϭ 0.15). Significant insulin sensitivity (r ϭ 0.38, P Ͻ 0.05). testosterone levels (n ϭ 10) were more differences were also observed in the met- Testosterone levels correlated positively ϭ Ͻ insulin resistant than their eugonadal abolic profile of the hypogonadal and with both VO2max (r 0.43, P 0.05; counterparts (n ϭ 50) (insulin sensitiv- eugonadal men (Table 1). Using the defi- Fig. 1) and UQCRB expression (r ϭ 0.57, Ϫ ity ϭ 3.6 Ϯ 0.6 vs. 7.3 Ϯ 3mg kg 1 nition proposed by the Third National P Ͻ 0.0001; Fig. 1). The relationship be- Ϫ min 1, respectively; P Ͻ 0.0007) (Table Health and Nutrition Examination Survey tween testosterone and expression of the 1). Using regression analysis with testos- (30), 90% of subjects with low testoster- entire OXPHOS-CR gene set was close to terone as the dependent variable, the as- one levels met criteria for the metabolic statistical significance (r ϭ 0.3, P ϭ 0.08). sociation between low testosterone levels syndrome compared with 29% of those BMI correlated negatively with VO2max and insulin resistance held true after con- with normal testosterone levels (P Ͻ (r ϭϪ0.4, P ϭ 0.009). However, there trolling for age, insulin levels, and SHBG 0.001). Only 1 of 10 subjects with a hy- was no relationship between BMI and ex-

DIABETES CARE, VOLUME 28, NUMBER 7, JULY 2005 1639 Testosterone and mitochondrial function in men pression of either UQCRB (r ϭϪ0.18, mass spectroscopy (32–34). A recent poprotein lipase activity in abdominal ad- P ϭ 0.24) or OXPHOS-CR (r ϭϪ0.15, study showed no relationship between ipose tissue, leading to decreased P ϭ 0.36). Similarly, a negative correla- free testosterone levels and insulin sensi- uptake in central fat depots tion was seen between percent body fat tivity in obese and diabetic men (22). (42,43). Thus, low testosterone levels ϭϪ ϭ and VO2max (r 0.49, P 0.001), and However, the commercial double anti- may predispose to visceral obesity, lead- ϭϪ ϭ WHR and VO2max (r 0.43, P body system used to measure free testos- ing to dysregulation of metabo- 0.008). There was a trend for percent terone in this study correlates poorly with lism, which in turn promotes insulin body fat to correlate with UQCRB (r ϭ results obtained by equilibrium dialysis resistance (44). Ϫ0.28, P ϭ 0.07), whereas there was no (32,35), which is considered the gold Although the present study demon- relationship between WHR and UQCRB standard. In addition, there is controversy strates a positive correlation between tes- (r ϭϪ0.23, P ϭ 0.16). The relationship as to whether it is only free testosterone tosterone levels and insulin sensitivity in between testosterone and VO2max re- that is biologically active, given that tes- men, no conclusions can be drawn about mained significant when adjusted for BMI tosterone bound to SHBG can bind to cell causality given the cross-sectional nature (P Ͻ 0.05), percent body fat (P Ͻ 0.05), surface receptors in prostate tissue, lead- of the data. However, in male rats, castra- and SHBG (P Ͻ 0.0001), but not WHR ing to activation of adenylyl cyclase and tion leads to the rapid development of in- (P ϭ 0.07). Similarly, the relationship be- generation of cAMP (36). For these rea- sulin resistance, which is corrected by tween testosterone and UQCRB expres- sons, we chose to use total testosterone physiological testosterone replacement sion was still significant after adjusting for levels as the most robust index of andro- (45). In the human, relatively little is SHBG levels (P Ͻ 0.0001). Finally, no re- genicity and to use multiple regression known about the impact of on lationship was observed between serum analysis to assess the contribution of insulin sensitivity. In men with prostate testosterone levels and muscle glycogen SHBG to any association between testos- cancer, induction of hypogonadism with content or the percentage of type 1, type terone levels and insulin sensitivity. a -releasing hormone ago- 2a, or type 2b muscle fibers. The fact that the correlation between nist results in a 60% increase in fasting testosterone levels and insulin sensitivity insulin levels at 3 months (46,47). Data CONCLUSIONS — These data dem- is no longer significant after controlling on the impact of supplementa- onstrate a positive correlation between se- for BMI could mean that obesity is caus- tion on insulin sensitivity in men are con- rum testosterone levels and insulin ing both low testosterone levels and insu- flicting depending on the population sensitivity in men across the full spectrum lin resistance and that there is no direct studied. Androgen administration to men of glucose tolerance. Moreover, men with relationship between testosterone and in- with central obesity and testosterone lev- hypogonadal testosterone levels are twice sulin sensitivity. However, the fact that els in the low normal range increases in- as insulin resistant as their eugonadal the impact of BMI on the relationship be- sulin sensitivity (48–50). However, a counterparts, and 90% fulfill criteria for tween testosterone and insulin sensitivity recent study of human chorionic gonad- the metabolic syndrome. From a clinical is attenuated significantly when BMI is ex- otropin (hCG) administration to healthy perspective, these data highlight the im- pressed as a categorical variable suggests older men with low normal testosterone portance of performing a comprehensive that this may not be the case. A second levels showed no change in insulin sensi- metabolic evaluation in men with hypo- possibility is that the effect of testosterone tivity (51). Differences in the outcome of gonadism. The demonstration that testos- on insulin sensitivity is mediated through this study may reflect the fact that the terone levels correlate not only with changes in BMI. The relationship between study population was less obese and that insulin sensitivity but also with genetic androgens and BMI is likely bi-direc- use of human chorionic gonadotropin (OXPHOS gene expression) and func- tional. Morbid obesity has negative effects was associated with very high E2 levels, tional (VO2max) markers of mitochondrial on the hypothalamic-pituitary-gonadal which may have negated the beneficial ef- function suggests a novel molecular axis in men (37). In addition, low testos- fects of androgens. In men with type 2 mechanism whereby testosterone might terone levels predispose to central (38) diabetes, one small nonrandomized study modulate insulin sensitivity in men. and visceral adiposity (39). In our analy- showed no beneficial effect of testoster- Previous studies on the relationship sis, all men with hypogonadal testoster- one replacement on glycemic control between androgens and insulin sensitivity one levels have a BMI Ͼ25 kg/m2 and a (52), whereas a larger non–- in men gave conflicting results depending WHR Ͼ0.9. Whereas visceral fat was not controlled study showed a significant re- on whether total or free testosterone levels assessed in the present study, androgens duction in HbA1c (A1C) levels after 3 were used. One explanation for this dis- have been shown to have important ef- months of testosterone therapy (53). crepancy is that SHBG is mediating the fects on visceral fat mediated Recent studies provide insight into link between testosterone and insulin largely by stimulation of . In the role of mitochondrial function in the sensitivity. Proponents of the use of free vitro, testosterone enhances cate- pathogenesis of insulin resistance with testosterone argue that it is the best index cholamine-induced lipolysis by increas- the demonstration of decreased expres- of androgenicity in insulin-resistant men ing the number of ␤3-adrenergic sion of peroxisome proliferator–activated given the low SHBG levels that pertain in receptors on rat precursor cells receptor-␥ coactivator (PGC-1␣) and this setting (31). An alternative explana- (40). Similarly, castration of male rats re- downregulation of OXPHOS genes in tion for the discordant results is that the duces lipolysis, which is reversed by skeletal muscle of insulin-resistant sub- assays used to measure free testosterone physiological testosterone replacement jects (1,2). A model for the pathogenesis have serious methodological limitations, (41). In two small studies of men with of insulin resistance has thus been pro- as highlighted in studies validated by central obesity, testosterone inhibits li- posed whereby, in genetically susceptible

1640 DIABETES CARE, VOLUME 28, NUMBER 7, JULY 2005 Pitteloud and Associates individuals, environmental risk factors abolic syndrome and/or type 2 diabetes in H, Sannerstedt R: Visceral fat accumula- such as physical inactivity contribute to men. tion in men is positively associated with decreased expression of PGC-1␣. Re- insulin, glucose, and C- levels, but duced PGC-1␣ expression, in turn, leads negatively with testosterone levels. Metab- to decreased of metabolic Acknowledgments— This work was sup- olism 39:897–901, 1990 10. Pasquali R, Casimirri F, Cantobelli S, Mel- and mitochondrial genes and thus de- ported by National Institutes of Health Grants K23 DK 02858-04, RO3 DK064276-01, and chionda N, Morselli Labate AM, Fabbri R, creased oxidative phosphorylation, de- M01-RR-01066, National Institutes of Health, Capelli M, Bortoluzzi L: Effect of obesity creased lipid oxidation, intracellular National Center for Research Resources, Gen- and body fat distribution on sex hor- accumulation of triglycerides in skeletal eral Clinical Research Centers Program. mones and insulin in men. Metabolism 40: muscle, and ultimately insulin resistance. V.K.M. was funded by a physician postdoc- 101–104, 1991 In this study, we demonstrate that testos- toral fellowship from Howard Hughes Medical 11. Simon D, Preziosi P, Barrett-Connor E, terone levels correlate positively with Institute. Roger M, Saint-Paul M, Nahoul K, Papoz We thank Virginia Hughes for her review of L: Interrelation between plasma testoster- VO2max and OXPHOS-CR gene expres- sion. From a mechanistic perspective, lit- the manuscript and helpful comments. one and plasma insulin in healthy adult tle is known about how testosterone men: the Telecom Study. Diabetologia 35: 173–177, 1992 might influence insulin action. Induction References 12. Barrett-Connor E: Lower endogenous an- of hypogonadism with a gonadotropin- 1. Mootha VK, Lindgren CM, Eriksson KF, drogen levels and in men releasing hormone (GnRH) in Subramanian A, Sihag S, Lehar J, Puig- with non-insulin-dependent diabetes normal men decreases lipid oxidation and server P, Carlsson E, Ridderstrale M, Lau- mellitus. Ann Intern Med 117:807–811, resting energy expenditure (54). The in- rila E: PGC-1alpha-responsive genes 1992 sulin resistance resulting from castration involved in oxidative phosphorylation are 13. Andersson B, Marin P, Lissner L, Ver- in rats is associated with a decrease in gly- coordinately downregulated in human di- meulen A, Bjorntorp P: Testosterone con- cogen synthase activity (45). Lean off- abetes. Nat Genet 34:267–273, 2003 centrations in women and men with spring of patients with type 2 diabetes (4) 2. Patti ME, Butte AJ, Crunkhorn S, Cusi K, NIDDM. Diabetes Care 17:405–411, have also been shown to have decreased Berria R, Kashyap S, Miyazaki Y, Kohane 1994 I, Costello M, Saccone R: Coordinated re- 14. Tibblin G, Adlerberth A, Lindstedt G, activity, which has duction of genes of oxidative metabolism Bjorntorp P: The pituitary-gonadal axis been attributed to intramyocellular tri- in with insulin resistance and di- and health in elderly men: a study of men glyceride accumulation. Data from the abetes: potential role of PGC1 and NRF1. born in 1913. Diabetes 45:1605–1609, Otsuka Long Evans Fatty (OLETF) rat, a Proc Natl Acad SciUSA100:8466–8471, 1996 genetic model of obesity and type 2 dia- 2003 15. Haffner SM, Shaten J, Stern MP, Smith betes, support a role for androgens in 3. Petersen KF, Befroy D, Dufour S, Dziura J, GD, Kuller L: Low levels of sex hormone- modulating mitochondrial function (55). Ariyan C, Rothman DL, DiPietro L, Cline binding globulin and testosterone predict In this model, expression levels of uncou- GW, Shulman GI: Mitochondrial dys- the development of non-insulin-depen- pling protein 1 (UCP-1) and its upstream function in the elderly: possible role in dent diabetes mellitus in men. Am J Epi- ␣ ␤ insulin resistance Science 300:1140– demiol 143:889–897, 1996 regulators, PGC-1 and the 3 adrenergic ␤ 1142, 2003 16. Stellato RK, Feldman HA, Hamdy O, Hor- receptor ( 3AR), are reduced, leading to 4. Petersen KF, Dufour S, Befroy D, Garcia ton ES, McKinlay JB: Testosterone, sex inefficient energy utilization and obesity R, Shulman GI: Impaired mitochondrial hormone-binding globulin, and the de- (55). Administration of dehydroepi- activity in the insulin-resistant offspring velopment of type 2 diabetes in middle- androsterone (DHEA) for 14 days caused of patients with type 2 diabetes. N Engl aged men. Diabetes Care 23:490–494, ␤ a significant increase in UCP-1, 3AR, J Med 350:664–671, 2004 2000 and PGC-1␣ expression and reversal of 5. Wisloff U, Najjar SM, Ellingsen O, Haram 17. Oh J-Y, Barrett-Connor E, Wedick NM, the adverse metabolic phenotype that PM, Swoap S, Al-Share Q, Fernstrom M, Wingard DL: Endogenous sex characterizes the OLETF rat with a reduc- Rezaei K, Lee SJ, Koch LG, Britton SL: and the development of type 2 diabetes in tion in body weight, glucose, insulin, free Cardiovascular risk factors emerge after older men and women: the Rancho Ber- fatty acids, and levels (55). artificial selection for low aerobic capac- nardo Study. Diabetes Care 25:55–60, ity. Science 307:418–420, 2005 2002 Thus, it is plausible that hypogonad- 6. Dunaif A: Insulin resistance and the poly- 18. Svartberg J, Jenssen T, Sundsfjord J, Jorde ism may cause insulin resistance via dys- cystic syndrome: mechanism and R: The associations of endogenous testos- regulation of and implications for pathogenesis. Endocr Rev terone and sex hormone-binding globulin that low testosterone levels may represent 18:774–800, 1997 with glycosylated hemoglobin levels, in an additional environmental factor con- 7. Phillips GB: Relationship between serum community dwelling men: the Tromso tributing to decreased expression of genes sex hormones and glucose, insulin and Study. Diabetes Metab 30:29–34, 2004 involved in oxidative metabolism. Fur- lipid abnormalities in men with myocar- 19. Laaksonen DE, Niskanen L, Punnonen K, ther studies are required to test this hy- dial infarction. Proc Natl Acad SciUSA Nyysonen K, Tuomainen T-P, Valkonen pothesis by examining gene expression 74:1729–1733, 1977 V-P, Salonen R, Salonen JT: Testosterone profiles in skeletal muscle before and after 8. Lichtenstein MJ, Yarnell JW, Elwood PC, and sex hormone-binding globulin pre- Beswick AD, Sweetnam PM, Marks V, dict the metabolic syndrome and diabetes androgen manipulations. If testosterone Teale D, Riad-Fahmy D: Sex hormones, in middle-aged men. Diabetes Care is shown to modulate OXPHOS gene ex- insulin, lipids, and prevalent ischemic 27:1036–1041, 2004 pression, androgen supplementation may disease. Am J Epidemiol 126:647– 20. Haffner SM, Karhapaa P, Mykkanen L, represent an important therapeutic mo- 657, 1987 Laakso M: Insulin resistance, body fat dis- dality for preventing or treating the met- 9. Seidell JC, Bjorntorp P, Sjostrom L, Kvist tribution, and sex hormones in men. Di-

DIABETES CARE, VOLUME 28, NUMBER 7, JULY 2005 1641 Testosterone and mitochondrial function in men

abetes 43:212–219, 1994 terone in normal women and women with 46. Smith JC, Bennett S, Evans LM, Kynaston 21. Birkeland KI, Hanssen KF, Torjesen PA, androgen deficiency: comparison of HG, Parmar M, Mason MD, Cockcroft JR, Vaaler S: Level of sex hormone-binding methods. J Clin Endocrinol Metab 89:525– Scanlon MF, Davies JS: The effects of in- globulin is positively correlated with in- 533, 2004 duced hypogonadism on arterial stiffness, sulin sensitivity in men with type 2 diabe- 33. Wang C, Catlin DH, Demers LM, body composition, and metabolic param- tes. J Clin Endocrinol Metab 76:275–278, Starcevic B, Swerdloff RS: Measurement eters in males with prostate cancer. J Clin 1993 of total serum testosterone in adult men: Endocrinol Metab 86:4261–4267, 2001 22. Abate N, Haffner SM, Garg A, Peshock comparison of current laboratory meth- 47. Dockery F, Bulpitt CJ, Agarwal S, Donald- RM, Grundy SM: Sex steroid hormones, ods versus liquid chromatography-tan- son M, Rajkumar C: Testosterone sup- upper body obesity, and insulin resis- dem mass spectrometry. J Clin Endocrinol pression in men with prostate cancer tance. J Clin Endocrinol Metab 87:4522– Metab 89:534–543, 2004 leads to an increase in arterial stiffness and 4527, 2002 34. Matsumoto AM, Bremner WJ: Serum tes- . Clin Sci 104:195–201, 23. Morley JE, Kaiser FE, Perry HM 3rd, tosterone assays: accuracy matters (Edito- 2003 Patrick P, Morley PM, Stauber PM, Vellas rial). J Clin Endocrinol Metab 89:520–523, 48. Marin P, Holmang S, Jonsson L, Sjostrom B, Baumgartner RN, Garry PJ: Longitudi- 2004 L, Kvist H, Holm G, Lindstedt G, Bjorn- nal changes in testosterone, luteinizing 35. Vermeulen A, Verdonck L, Kaufman JM: torp P: The effects of testosterone treat- hormone, and follicle-stimulating hor- A critical evaluation of simple methods ment on body composition and mone in healthy older men. Metabolism for the estimation of free testosterone in metabolism in middle-aged obese men. 46:410–413, 1997 serum. J Clin Endocrinol Metab 84:3666– Int J Obes 16:991–997, 1992 24. Harman SM, Metter EJ, Tobin JD, Pearson 3672, 1993 49. Marin P, Holmang S, Gustafsson C, Jon- J, Blackman MR: Longitudinal effects of 36. Rosner W, Hyrb DJ, Khan MS, Nakhla sson L, Kvist H, Elander A, Eldh J, Sjos- aging on serum total and free testosterone AM, Romas, NA: Androgen and trom L, Bjorntorp P: Androgen treatment levels in healthy men: Baltimore Longitu- signaling at the via G-pro- of abdominally obese men. Obes Res dinal Study of Aging. J Clin Endocrinol teins and cyclic adenosine monophos- 1:245–251, 1993 Metab 86:724–731, 2001 phate. Steroids 64:100–106, 1999 50. Simon D, Charles MA, Lahlou N, Nahoul 25. Feldman HA, Longcope C, Derby CA, Jo- 37. Giagulli VA, Kaufman JM, Vermeulen A: K, Oppert JM, Gouault-Heilmann M, hannes CB, Araujo AB, Coviello AD, Pathogenesis of the decreased androgen Lemort N, Thibult N, Joubert E, Balkau B, Bremner WJ, McKinlay JB: Age trends in levels in obese men. J Clin Endocrinol Eschwege E: Androgen therapy improves the level of serum testosterone and other Metab 79:997–1000, 1994 insulin sensitivity and decreases leptin hormones in middle-aged men: longitu- 38. Khaw KT, Barrett-Connor E: Lower en- level in healthy adult men with low dinal results from the Massachusetts male dogenous androgens predict central adi- plasma total testosterone: a 3-month ran- aging study. J Clin Endocrinol Metab 87: posity in men. Ann Epidemiol 2:675–682, domized placebo-controlled trial. Diabe- 589–598, 2002 1992 tes Care 24:2149–2151, 2001 26. Eriksson KF, Lindgarde F: Impaired glu- 39. Tsai EC, Boyko EJ, Leonetti DL, Fujimoto 51. Liu PY, Wishart SM, Celermajer DS, Jime- cose tolerance in a middle-aged male ur- WY: Low serum testosterone level as a nez M, Pierro ID, Conway AJ, Handels- ban population: a new approach for predictor of increased visceral fat in Japa- man DJ: Do reproductive hormones identifying high-risk cases. Diabetologia nese-American men. Int J Obes Relat Metab modify insulin sensitivity and metabolism 33:526–531, 1990 Disord 4:485–491, 2000 in older men? A randomized, placebo- 27. World Health Organization: Diabetes Mel- 40. Xu XF, De Pergola G, Bjorntorp P: The controlled clinical trial of recombinant litus: Report of a WHO Study Group. Ge- effects of androgens on the regulation of human chorionic gonadotropin. Eur J En- neva, World Health Org., 1985 (Tech. lipolysis in adipose precursor cells. Endo- docrinol 148:55–66, 2003 Rep. Ser., no. 727) crinology 126:1229–1234, 1990 52. Corrales JJ, Burgo RM, Garca-Berrocal B, 28. DeFronzo RA, Tobin JD, Andres R: Glu- 41. Xu XF, De Pergola G, Bjorntorp P: Testos- Almeida M, Alberca I, Gonzalez-Buitrago cose clamp technique: a method for quan- terone increases lipolysis and the number JM, Orfao A, Miralles JM: Partial androgen tifying insulin secretion and resistance. of beta-adrenoreceptors in male rat adipo- deficiency in aging type 2 diabetic men Am J Physiol 237:E214–E223, 1979 cytes. Endocrinology 128:379–382, 1991 and its relationship to glycemic control. 29. Schalin-Jantti C, Laurila E, Lofman M, 42. Rebuffe-Scrive M, Marin P, Bjorntorp P: Metabolism 53:666–672, 2004 Groop LC: Determinants of insulin-stim- Effect of testosterone on abdominal adi- 53. Boyanov MA, Boneva Z, Christov VG: ulated skeletal muscle glycogen metabo- pose tissue in men. Int J Obes 15:791–795, Testosterone supplementation in men lism in man. Eur J Clin Invest 25:693–698, 1991 with type 2 diabetes, visceral obesity and 1995 43. Marin P, Oden B, Bjorntorp P: Assimila- partial androgen deficiency. Aging Male 30. Ford ES, Giles WH, Dietz WH: Prevalence tion and mobilization of triglycerides in 6:1–7, 2003 of the metabolic syndrome among US subcutaneous abdominal and femoral ad- 54. Mauras N, Hayes V, Welch S, Rini A, adults: findings from the Third National ipose tissue in vivo in men: effects of an- Helgeson K, Dokler M, Veldhuis JD, Ur- Health and Nutrition Examination Sur- drogens. J Clin Endocrinol Metab 80:239– ban RJ: Testosterone deficiency in young vey. JAMA 287:356–359, 2002 243, 1995 men: marked alterations in whole body 31. Plymate SR, Matej LA, Jones RE, Friedl 44. Boden G, Jadali F, White J, Liang Y, Moz- protein kinetics, strength, and adiposity. KE: Inhibition of sex hormone-binding zoli M, Chen X, Coleman E, Smith C: Ef- J Clin Endocrinol Metab 83:1886–1892, globulin production in the human hepa- fects of fat on insulin-stimulated 1998 toma (Hep G2) cell line by insulin and metabolism in normal men. 55. Ryu JW, Kim MS, Kim CH, Song KH, Park . J Clin Endocrinol Metab 67:460– J Clin Invest 88:960–966, 1991 JY, Lee JD, Kim JB, Lee KU: DHEA admin- 464, 1988 45. Holmang A, Bjorntorp P: The effects of istration increases brown fat uncoupling 32. Miller KK, Rosner W, Lee H, Hier J, testosterone on insulin sensitivity in male protein 1 levels in obese OLETF rats. Bio- Sesmilo G, Schoenfeld D, Neubauer G, rats. Acta Physiol Scand 146:505–510, chem Biophys Res Commun 303:726–731, Klibanski A: Measurement of free testos- 1992 2003

1642 DIABETES CARE, VOLUME 28, NUMBER 7, JULY 2005