Low Cytochrome Oxidase 4I1 Links Mitochondrial Dysfunction to Obesity and Type 2 Diabetes in Humans and Mice

ORIGINAL ARTICLE Low cytochrome oxidase 4I1 links mitochondrial dysfunction to obesity and type 2 diabetes in humans and mice

B Van der Schueren1, R Vangoitsenhoven1, B Geeraert2, D De Keyzer2, M Hulsmans2, M Lannoo1, HJ Huber2, C Mathieu1 and P Holvoet2

OBJECTIVES: Cytochrome oxidase (COX) dysfunction is associated with mitochondrial oxidative stress. We determined the association between COX expression, obesity and type 2 diabetes. SUBJECTS/METHODS: COX4I1 and COX10 genes were measured in monocytes of 24 lean controls, 31 glucose-tolerant and 67 diabetic obese patients, and 17 morbidly obese patients before and after bariatric surgery. We investigated the effect of caloric restriction and peroxisome proliferator-activated receptor (PPAR) agonist treatment on Cox in obese diabetic mice, and that of diet-induced insulin resistance in Streptozotocin-treated mice. RESULTS: Low COX4I1 was associated with type 2 diabetes in obese patients, adjusting for age, gender, smoking, interleukin-6 and high-sensitivity C-reactive protein, all related to metabolic syndrome (MetS; odds ratio: 6.1, 95% conﬁdence interval: 2.3–16). In contrast, COX10 was low in glucose-tolerant and diabetic obese patients. In morbidly obese patients, COX4I1 was lower in visceral adipose tissue collected at bariatric surgery. In their monocytes, COX4I1 decreased after bariatric surgery, and low COX4I1 at 4 months was associated with MetS at 7 years. In leptin-deﬁcient obese diabetic mice, Cox4i1 was low in white visceral adipose tissue (n = 13; Po0.001) compared with age-matched lean mice (n = 10). PPARγ-agonist treatment (n = 13), but not caloric restriction (n = 11), increased Cox4i1 (Po0.001). Increase in Cox4i1 depended on the increase in glucose transporter 4 (Glut4) expression and insulin sensitivity, independent of the increase in blood adiponectin. In streptozotocin-treated mice (three groups of seven mice, diet-induced insulin resistance decreased Cox4i1 and Glut4 (Po0.001 for both). CONCLUSION: COX4I1 depression is related to insulin resistance and type 2 diabetes in obesity. In peripheral blood monocytes, it may be a diagnostically useful biomarker. International Journal of Obesity (2015) 39, 1254–1263; doi:10.1038/ijo.2015.58

INTRODUCTION cytochrome oxidase (COX) complex, the terminal node and rate- Obesity is an important risk factor for metabolic syndrome (MetS), limiting step in the mitochondrial electron transport chain, is 17,18 type 2 diabetes and cardiovascular disease.1–3 Low-grade inflam- associated with mitochondrial oxidative stress, acondition 9,10,19,20 mation has been shown to underlie the development of obesity- associated with obesity, MetS and type 2 diabetes. COX4I1 4–6 is suggested to be the most important regulatory subunit related complications. Recent studies have supported the 21,22 contention that increased oxidative stress in adipose tissue is an of COX as it is required for the allosteric feedback inhibition of 7–11 the enzyme by its indirect product ATP. COX10 as this subunit is early instigator of MetS. Oxidative damage of adipose tissues is 23 associated with impaired adipocyte maturation and production of required for COX biogenesis. Hence, the aims of this study were to fl investigate the association of mitochondrial oxidative stress with pro-in ammatory adipocytokines by dysfunctional adipocytes, 24 emphasizing an important role of adiponectin.12,13 Both factors obesity, MetS and type 2 diabetes and to test the potential of increase infiltration of activated monocytes into the adipose COX4I1 and COX10 in peripheral blood monocytes as biomarkers for tissues where they differentiate into macrophages that produce a detrimental metabolic evolution in patients with obesity. To this inflammatory chemokines.14 The existence of specific markers for end, we determined their expressions in adipose tissue and in type 2 diabetes in relation to oxidative stress in adipose tissues is monocytes in representative animal models and patient populations. yet to be determined. Moreover, if these markers would be available with a simple peripheral blood test, they would be a SUBJECTS AND METHODS useful clinical tool to predict the risk of type 2 diabetes. Obesity, MetS, type 2 diabetes and cardiovascular diseases were Subjects found to be associated with systemic oxidative stress, mainly The first patient group consisted of 98 successive obese individuals: 31 evidenced by high levels of oxidized low-density lipoprotein normal glucose-tolerant obese (NGTO) patients and 67 obese patients with (LDL).8–10,15 It has been proposed that mitochondrial decline type 2 diabetes, according to the American Diabetes Association. Inclusion was blinded. The samples were collected at the Division of Endocrinology resulting in mitochondrial oxidative stress contributes to the between 27 September and 20 December 2007. Obese subjects were development of age-related metabolic and cardiovascular compared with 24 lean control persons. All participants were without 16 diseases. However, the relation of mitochondrial oxidative stress symptoms of clinical atherosclerotic cardiovascular disease. This study and those diseases remains to be elucidated. Dysfunction of the complies with the Declaration of Helsinki and the Medical Ethics

1Laboratory for Clinical and Experimental Medicine and Endocrinology, KU Leuven, Leuven, Belgium and 2Division of Atherosclerosis and Metabolism, Department of Cardiovascular Sciences, KU Leuven, Leuven, Belgium. Correspondence: Professor P Holvoet, Division of Atherosclerosis and Metabolism, Department of Cardiovascular Sciences, KU Leuven, Herestraat 49, O&N1, PB 705, B-3000 Leuven, Belgium. E-mail: [email protected] Received 22 January 2015; revised 19 March 2015; accepted 4 April 2015; accepted article preview online 14 April 2015; advance online publication, 19 May 2015 Cytochrome oxidase in obesity and diabetes B Van der Schueren et al 1255 Committee of the KU Leuven approved the study protocol. All human RNA isolation and quantitative real-time PCR analysis participants gave written informed consent. Extraction and purification of RNA and complementary DNA generation The second patient cohort comprised 17 obese individuals. These 17 was performed as described before.29,36,37 qPCR was performed on a 7500 morbidly obese subjects were referred to our hospital for bariatric surgery. Fast Real-Time PCR system using Fast SYBRGreen master mix (Applied After multidisciplinary deliberation, the selected patients received a Biosystems, Ghent, Belgium). Oligonucleotides (Invitrogen, Ghent, Belgium) laparoscopic Roux-en-Y gastric bypass. A 30-ml fully divided gastric pouch used as forward and reverse primers were designed using the ‘Primer was created and the jejunum, 30 cm distal of the ligament of Treitz, was Express’ software (Applied Biosystems; Supplementary Table 1). RNA anastomosed to it with a circular stapler of 25 mm. To restore intestinal expression levels were calculated with the delta-delta-quantification cycle transit, a fully stapled entero–entero anastomose was constructed 120 cm 38 method (ΔΔCq). β-actin for mouse experiments, and HPRT1, SDHA, TBP distal on the alimentary limb. In this way, the food passage was derived and YWHAZ for patient samples were selected as most stable house- away from almost the whole stomach, the duodenum and the proximal 39 25–27 keeping genes using GeNorm. jejunum. Monocytes were isolated from venous blood taken before, In order to test the reproducibility and variance of expression of RNA and at 4 months and 7 years after bariatric surgery. Visceral adipose tissue markers, RNA expression was measured in monocytes of 13 healthy biopsies were collected at the time of surgery. The samples were collected individuals collected at weeks 0, 1 and 2. P-values, determined by repeated between 29 March 2005 and 28 February 2013. measures test, were 0.95 for COX1, 0.96 for COX4I1, 0.92 for glutathione peroxidase (GPX1) and 0.88 for IRAK3. Receiver operating characteristic Plasma analysis (ROC) curves of week 0 and week 1 and week 2 were also compared. Mean Blood samples were centrifuged to prepare plasma samples. Total, high- areas under the curves (AUC) were 0.56 for COX10, 0.54 for COX4I1, 0.53 for density lipoprotein (HDL) cholesterol, triglycerides, adiponectin, oxidized GPX1 and 0.54 for IRAK3, whereby an ROC of around 0.5 indicates no LDL (Ox-LDL), high-sensitivity C-reactive protein (hs-CRP) and interleukin-6 evidence that the data are different according to time of collection. Mean (IL-6) were measured as before.28 LDL cholesterol levels were calculated within-group variability was 14% for COX10, 12% for COX4I1 and 17% for using the Friedewald formula. Plasma glucose was measured using GPX1 and IRAK3. the glucose oxidase method (Johnson & Johnson, Zaventem, Belgium), and insulin with an immunoassay (Biosource Technologies, Fleunes, Statistics 29,30 Belgium). Insulin resistance was calculated by a homeostasis model Two groups were compared using an unpaired Mann–Whitney U-test. More assessment of insulin resistance (HOMA-IR) = fasting plasma insulin − 1 31 than two groups were compared using nonparametric one-way analysis of (mU l ) x fasting blood glucose (mM)/22.5. All laboratory assessments variance (Kruskal–Wallis) followed by comparison by the Dunn’smultiple were performed without the knowledge of clinical data. Blood pressure comparison test or by repeated measures analysis of variance followed by the was taken three times with the participant in a seated position after 5 min Bonferroni’s multiple comparison test. It was tested whether variance in the of quiet rest, and the average of the last two measurements was used. groups that were being statistically compared was similar. Odds ratios were MetS was defined according to the joint interim statement of 2009.1 determined by Fisher exact test. All analyses were performed using GraphPad Prism 5.0 for Windows (GraphPad Software, San Diego, CA, USA), except for Monocyte isolation from human blood ROC analysis, where MedCalc statistical software (Medcalc Software bvba, o Peripheral blood mononuclear cells (PBMCs) were prepared from the Ostend, Belgium) for biomedical research was used. A P-value 0.05 was fi anticoagulated blood using gradient separation on Histopaque-1077. considered statistically signi cant.Weusedtheon-linePower&SampleSize Monocytes were isolated from PBMCs (1 × 107 cells) using anti-CD14 Calculator of Statistical Solutions to determine if the sample size for each coated beads, and an LS column that was put in a MidiMACS Separator study assured an adequate power (at least 0.80) to detect statistical (Miltenyi Biotec Ltd., Surrey, UK).32,33 significance (at least Po0.05).

Mouse studies RESULTS Animal experiments were approved by the Institutional Animal Care and Relation of RNA expressions to type 2 diabetes in patients with Research Advisory Committee of the KU Leuven (P087/2007). Breeding and obesity genotyping of mice deficient in LDL receptor (Jackson Laboratory, Bar Harbor, ME, USA) and leptin ob gene (Jackson Laboratory; DKO), We determined the relation between RNA expressions and backcrossed on the C57BL/6J background (KU Leuven), were performed diabetes in a cohort of 31 NGTO patients and 67 obese patients as previously described.29,30 Food intake of male DKO mice was ≈5.7 g with type 2 diabetes. Compared with 24 lean controls, age- per day; this is almost double of the mean food intake of lean mice (≈3g matched obese persons with type 2 diabetes more often had per day). We studied placebo DKO (n = 13), DKO mice in which the weight MetS. Characteristics for all three populations are given in Table 1. ≈ loss was induced by caloric restriction (( 3 g per day; n = 11) and DKO mice NGTOs had higher HOMA-IR, higher diastolic blood pressure and treated with rosiglitazone (n = 13). Randomization was blinded. Mice were included in the study at 12 weeks and were followed during 12 weeks. were more often treated with diuretics. Furthermore, they had They were compared with age- and gender-matched lean C57BL/6J mice higher blood levels of leptin, insulin, IL-6, hs-CRP and ox-LDL, but (n = 10). Rosiglitazone (Sigma Aldrich, Diegem, Belgium, 10 mg kg − 1 lower levels of adiponectin and HDL cholesterol. Compared with per day) was added to standard chow containing 4% fat (Ssniff, Soest, lean controls, obese patients with type 2 diabetes were older, and Germany), placebo-treated mice received the grinded chow only. more often had MetS and used more medication. They had higher Rosiglitazone did not affect the food intake. Food and water were systolic blood pressure and diastolic blood pressure, higher available ad libitum. At baseline (12 weeks), characteristics of group DKO triglycerides, leptin, IL-6, hs-CRP and more elevated glycemia mice were identical. and higher HOMA-IR values. Levels of adiponectin and HDL Because DKO mice are leptin deficient, we also wanted to study mice with impaired insulin signaling that overexpress leptin as in obese persons. cholesterol were lower. LDL cholesterol was also lower, most likely In brief, male C57BL/6 mice received a single subcutaneous injection of owing to more frequent statin use. When compared with obese Streptozotocin (STZ; Sigma Aldrich) 2 days after birth (performed at Stelic patients without type 2 diabetes, obese patients with diabetes (Tokyo, Japan) under the supervision of Paul Holvoet. In this experiment, were older, and more often had MetS, higher HOMA-IR and mice were fed a high-fat diet (CLEA, Tokyo, Japan) ad libitum after 4 weeks elevated systolic blood pressure. Adiponectin levels were not of age. The composition was as follows: 24.5% casein, 5% egg white different, whereas leptin levels were lower. Patients with diabetes powder, 0.4% L-Cystine, 15.9% beef tallow powder (80%), 20% safflower also had lower ox-LDL, most probably due to more frequent use of (high oleic acid), 5.5% cellulose; 8.3% maltodextrin and 6.9% lactose; 6.8% 34,35 statins. Compared with lean controls, COX10, GPX1 and IRAK3 sucrose. Mice were studied at 4, 8 and 12 weeks after STZ injection were low in monocytes of obese patients without and with type 2 (seven mice per group) and were compared with age-matched C57BL/6 mice (six, six and nine mice per group). All mice were killed by Nembutal diabetes; superoxide dismutase (SOD2) was high. Importantly, overdose.29,30 No animals were excluded. COX10 was lower in obese patients with diabetes compared with The investigator was blinded to the group allocation during the lean controls and NGTO (Table 1). ROC curve analysis confirmed experiment and when assessing the outcome. that COX4I1 was related to type 2 diabetes; its area under the

Table 1. Characteristics and gene expressions in monocytes of lean controls and obese patients without and with type 2 diabetes

Lean controls (n = 24) Obese without T2D (n = 31) Obese with T2D (n = 67)

Characteristics Age (years) 40 ± 13 42 ± 14 54 ± 8.7***/††† Gender (n, % male) 9 (38) 4 (13) 4 (6)*** Smoker (n, %) 1 (4) 0 (0) 6 (9) Ex-smoker (n, %) 1 (4) 0 (0) 0 (0) Obesity (n, %) 0 (0) 31 (100)*** 67 (100)*** Type 2 diabetes (n, %) 0 (0) 0 (0) 67 (100)***/††† MetS (n, %) 1 (4) 19 (61)*** 64 (96)***/††† Statin use (n, %) 1 (4) 7 (23) 48 (72)***/††† Antihypertensive drug use (n, %) 6 (25) 14 (45) 53 (79)***/††† ACEI or ARB (n, %) 2 (8) 4 (13) 32 (48)***/††† CACB (n, %) 1 (4) 4 (13) 16 (24)* Betablocker (n, %) 4 (17) 9 (29) 31 (46)* Diuretics (n, %) 0 (0) 9 (29)** 28 (42)*** Oral antidiabetic drug use (n, %) 0 (0) 0 (0) 54 (81)***/††† Metformin (n, %) 0 (0) 0 (0) 52 (78)***/††† Insulin therapy (n, %) 0 (0) 0 (0) 43 (64)***/††† BMI (kg m − 2)23± 2.7 41 ± 7.7*** 38 ± 5.0***/† Leptin (ng ml − 1) 11.1 ± 8.5 72 ± 40*** 46 ± 30***/†† ADN (μgml− 1) 10.9 ± 6.3 5.0 ± 2.9*** 4.5 ± 3.8*** Glucose (mg dl − 1)92± 14 92 ± 11 136 ± 48***/††† Insulin (mU l − 1) 10.1 ± 6.2 14.4 ± 7.7* 34.4 ± 40.2***/††† HOMA-IR 1.3 ± 0.8 1.8 ± 1.0* 4.4 ± 4.5***/††† HOMA %b 114.6 ± 69.1 136.2 ± 45.9 156.7 ± 172.2 TG (mg dl − 1) 102 ± 48 130 ± 65 153 ± 68*** LDL-C (mg dl − 1) 112 ± 34 103 ± 31 75 ± 28***/††† HDL-C (mg dl − 1)63± 17 51 ± 11** 46 ± 12***/† SBP (mm Hg) 129 ± 15 131 ± 16 139 ± 16*/† DBP (mm Hg) 74 ± 7.4 83 ± 12** 82 ± 9.8*** IL-6 (pg ml − 1) 2.4 ± 1.1 5.3 ± 2.7*** 5.5 ± 2.9*** Hs-CRP (mg l − 1) 1.7 ± 3.5 8.3 ± 11.2** 6.7 ± 7.5*** Ox-LDL (IU l − 1)37± 15 60 ± 22** 48 ± 18††

Gene expressions COX10 1.05 ± 0.16 0.82 ± 0.18*** 0.76 ± 0.17*** COX4I1 1.05 ± 0.15 1.03 ± 0.24 0.85 ± 0.18***/††† GPX1 0.98 ± 0.40 0.56 ± 0.18*** 0.58 ± 0.21*** IRAK3 1.03 ± 0.21 0.75 ± 0.24*** 0.80 ± 0.18*** SOD2 1.05 ± 0.24 2.58 ± 1.38*** 2.53 ± 1.38*** Abbreviations: ACEI, ACE inhibitor; ADN, adiponectin; ARB, angiotensin receptor blocker; BMI, body mass index; CACB, calcium channel blocker; C, cholesterol; COX10, cytochrome c oxidase assembly homolog 10; COX4I1, cytochrome c oxidase subunit IV isoform 1; DBP, diastolic blood pressure; GPX1, glutathione peroxidase 1; HDL, high-density lipoprotein; HOMA-IR, homeostasis model assessment of insulin resistance; IRAK3, interleukin-1 receptor-associated kinase 3; hs-CRP, high-sensitivity C-reactive protein; IL, interleukin; LDL, low-density lipoprotein; MetS, metabolic syndrome; ox-LDL, oxidized LDL; PPAR, peroxisome proliferator-activated receptor; SBP, systolic blood pressure; SOD2: superoxide dismutase, mitochondrial; T2D, type 2 diabetes mellitus; TG, triglycerides. Data shown are mean ± s.d. *Po0.05, **Po0.01 and ***Po0.001 obese persons with and without type 2 diabetes compared with lean controls; †Po0.05, ††Po0.01 and †††Po0.001 obese persons with type 2 diabetes compared with obese persons without type 2 diabetes.

curve (AUC) was higher than 0.70 and its sensitivity was above surgery. Characteristics are shown in Table 2. As expected, obese 70%, similar to that of glucose. More importantly, this test showed patients before surgery more often had type 2 diabetes and MetS, an additive value of COX4I1 and glucose (odds ratio increased and had higher leptin, glucose, HOMA-IR, diastolic blood pressure, from 6.6 and 9.5 to 45). The sensitivity and specificity of the IL-6, hs-CRP and Ox-LDL. They had lower blood adiponectin. They combination of COX4I1 and glucose was above 80%. In addition, were more often treated with oral antidiabetic drugs: seven multiple logistic regression analysis including all genes that were patients used metformin, two patients used sulfonylurea, whereas differentially expressed in monocytes of obese persons showed no patients were on peroxisome proliferator-activated receptor o that COX4I1 predicted type 2 diabetes, while adjusting for age, (PPAR) agonists. COX4I1 (0.79 ± 0.13 vs 1.18 ± 0.30; P 0.05), o gender, smoking, IL-6 and hs-CRP, which were all related to MetS COX10 (0.88 ± 0.16 vs 1.12 ± 0.23; P 0.05, GPX1 (0.76 ± 0.15 vs o (odds ratio: 6.1, 95% confidence interval: 2.3–16). In addition, 1.21 ± 0.23; P 0.01) and SOD3 (0.66 ± 0.051 vs 0.93 ± 0.094; o fi IRAK3 was associated with diabetes (adjusted odds ratio: 4.7, 95% P 0.05) expressions were signi cantly lower in the visceral confidence interval: 1.3–17). Together COX4I1 and IRAK3 predicted adipose tissue of obese patients when compared with that of the lean controls (n = 6). Lower values were associated with impaired allocation of 83% of controls without type 2 diabetes and 89% of adipose tissue differentiation evidenced by lower expressions of cases with type 2 diabetes correctly (overall prediction 87%). ADIPOQ (0.68 ± 0.21 vs 1.43 ± 0.50; Po0.05), GLUT4 (0.37 ± 0.19 vs 1.32 ± 0.53; Po0.01), PPARα (0.65 ± 0.12 vs 1.09 ± 0.37; Po0.01) Expressions in adipose tissue and monocytes of obese persons and PPARβ/δ (0.83 ± 0.12 vs 0.99 ± 0.10; Po0.01), and a trend of and response to bariatric surgery lower PPARγ (0.79 ± 0.21 vs 1.13 ± 0.56; P = 0.06). We compared RNA expressions in visceral adipose tissue of an We measured the expressions in monocytes of these independent group of 17 obese subjects scheduled for bariatric 17 patients with obesity. Before bariatric surgery, we found low

Table 2. Characteristics and gene expressions in monocytes before and after bariatric surgery in patients with obesity

Lean controls Obese patients (n = 17) (n = 24) Before bariatric 4 Months after bariatric 7 Years after bariatric ANOVA surgery surgery surgery

Characteristics Age (years) 40 ± 13 40 ± 14 40 ± 14 48 ± 14†††/‡‡‡ Po0.001 Gender (n, % male) 9 (38) 5 (29) 5 (29) 5 (29) NA Smoker (n, %) 1 (4) 3 (18) 3 (18) 4 (24) NA Ex-smoker (n, %) 1 (4) 4 (24) 4 (24) 5 (29) NA Obesity (n, %) 0 (0) 17 (100)*** 16 (94)*** 13 (76)*** NA Type 2 diabetes (n, %) 0 (0) 9 (53)*** 5 (29)** 5 (29)** NA MetS (n, %) 1 (4) 12 (71)*** 6 (35)* 9 (53)*** NA Statin use (%) 1 (4) 6 (35)* 6 (35)* 6 (35)* NA Antihypertensive drug use (n, %) 6 (25) 6 (35) 3 (18) 4 (24) NA ACEI or ARB (n, %) 2 (8) 3 (18) 2 (12) 1 (6) NA CACB (n, %) 1 (4) 1 (6) 0 (0) 0 (0) NA Betablocker (n, %) 4 (17) 4 (24) 1 (6) 3 (18) NA Diuretics (n, %) 0 (0) 4 (24)* 2 (12) 4 (24)* NA Oral antidiabetic drug use (n, %) 0 (0) 8 (47)*** 7 (41)*** 4 (24)* NA Metformin (n, %) 0 (0) 7 (41)*** 7 (41)*** 4 (24)* NA Insulin therapy (n, %) 0 (0) 1 (6) 0 (0) 2 (12) NA BMI (kg m − 2)23± 2.7 45 ± 7*** 37 ± 6***/††† 33 ± 4**/††† Po0.001 Leptin (ng ml − 1) 11.1 ± 8.5 70 ± 35*** 27 ± 20††† 33 ± 22*/††† Po0.001 ADN (μgml− 1) 10.9 ± 6.3 4.1 ± 4.0*** 7.8 ± 6.9†† 12.4 ± 7.8†††/‡‡‡ Po0.001 Glucose (mg dl − 1)92± 14 123 ± 44** 95 ± 18††† 99 ± 20†† Po0.001 Insulin (mU l − 1) 10.1 ± 6.2 17.8 ± 9.5 8.3 ± 3.9††† 8.7 ± 5.9††† Po0.001 HOMA-IR 1.3 ± 0.8 2.4 ± 1.2* 1.1 ± 0.5††† 1.2 ± 0.8††† Po0.001 HOMA % b 114.6 ± 69.1 119.3 ± 71.6 97.1 ± 41.9 84.4 ± 30.4† Po0.05 TG (mg dl − 1) 102 ± 48 148 ± 67 110 ± 48† 99 ± 49†† Po0.01 LDL-C (mg dl − 1) 112 ± 34 94 ± 32 88 ± 27 88 ± 20 NS HDL-C (mg dl − 1)63± 17 50 ± 12 47 ± 10* 64 ± 20†††/‡‡‡ Po0.001 SBP (mm Hg) 129 ± 15 141 ± 17 118 ± 4††† 131 ± 20 Po0.001 DBP (mm Hg) 74 ± 7.4 87 ± 12* 59 ± 8***/††† 79 ± 8‡‡‡ Po0.001 IL-6 (pg ml − 1) 2.4 ± 1.1 5.1 ± 3.2*** 3.8 ± 2.4 4.3 ± 3.4 NS Hs-CRP (mg l − 1) 1.7 ± 3.5 5.7 ± 6.0** 7.3 ± 15.6 1.5 ± 1.5 NS Ox-LDL (IU l − 1)54± 20 76 ± 21** 73 ± 20* 49 ± 10†††/‡‡‡ Po0.001

Gene expressions COX10 1.05 ± 0.16 0.91 ± 0.19 0.94 ± 0.19* 0.78 ± 0.14***/† Po0.05 COX4I1 1.05 ± 0.15 1.09 ± 0.26 1.13 ± 0.24 0.71 ± 0.12***/†††/‡‡‡ Po0.001 GPX1 0.98 ± 0.40 0.59 ± 0.26*** 0.57 ± 0.37** 0.87 ± 0.28†††/‡‡‡ Po0.001 IRAK3 1.03 ± 0.21 0.63 ± 0.32*** 0.76 ± 0.23*** 1.57 ± 0.38***/†††/‡‡‡ Po0.001 SOD2 1.05 ± 0.24 2.07 ± 1.03*** 1.99 ± 1.03** 1.04 ± 0.21†††/‡‡‡ Po0.001 Abbreviations: ACEI, ACE inhibitor; ADN, adiponectin; ANOVA, analysis of variance; ARB, angiotensin receptor blocker; BMI, body mass index; CACB, calcium channel blocker; C, cholesterol; COX10, cytochrome c oxidase assembly homolog 10; COX4I1, cytochrome c oxidase subunit IV isoform 1; DBP, diastolic blood pressure; GPX1, glutathione peroxidase 1; HDL, high-density lipoprotein; HOMA-IR, homeostasis model assessment of insulin resistance; IRAK3, interleukin-1 receptor-associated kinase 3; hs-CRP, high-sensitivity C-reactive protein; IL, interleukin; LDL, low-density lipoprotein; MetS, metabolic syndrome; NA, not applicable; NS, not signiﬁcant; ox-LDL, oxidized LDL; PPAR, peroxisome proliferator-activated receptor; SBP, systolic blood pressure; SOD2: superoxide dismutase, mitochondrial; TG, triglycerides. Data shown are mean ± s.d. *Po0.05, **Po0.01 and ***Po0.001 compared with lean controls (Mann–Whitney test); †Po0.05, ††Po0.01 and †††Po0.001 compared with before bariatric surgery and ‡Po0.05, ‡‡Po0.01 and ‡‡‡Po0.001 compared with 4 months after bariatric surgery (repeated measures ANOVA followed by the Bonferroni's multiple comparison test).

IRAK3 and high SOD2, as before.28 We also found a decrease of In contrast, there was an increased trend in the number of patients GPX1 in patients with obesity (Table 2), whereas COX4I1 and with MetS (from 6 to 9) 7 years after the surgery. Mean levels of COX10 were not different when compared with lean controls. leptin, glucose, insulin, triglycerides and ox-LDL, and HOMA-IR and Four months after surgery, 16 out of 17 patients still had a body HOMA %b were lower, whereas adiponectin and HDL cholesterol mass index 430 kg m − 2, although significant weight loss had was higher than before surgery. At 7 years, adiponectin and HDL occurred and a decreased trend in the number of patients with cholesterol were higher and ox-LDL was lower than at 4 months. MetS (from 12 to 6) and type 2 diabetes (from 9 to 5) was Interestingly, the levels of glucose, insulin, adiponectin, triglycerides, observed. Mean levels of leptin, glucose, insulin, triglycerides and hs-CRP, ox-LDL and HOMA-IR were similar to those in lean controls, systolic blood pressure and diastolic blood pressure were lower, even though body mass index and leptin were still higher than in whereas the level of adiponectin was higher. There were no lean controls. The use of antihypertensive and oral antidiabetic significant changes in drug treatment. At 7 years, 13 out of 17 drugs was not statistically different (Table 2). patients were still obese, despite the persistence of significantly At 4 months, GPX1 and IRAK3 were still lower and SOD2 was still lower body weight. The prevalence of type 2 diabetes remained higher than in lean controls. At 7 years, IRAK3 increased to a level unchanged when compared with that of patients at 4 months that was even higher than that in lean controls. This was post surgery although HbA1c tended to decrease from 7.1 to 5.9% associated with an increase in GPX1 and a decrease in SOD2 that (54.1 to 41.0 mmol mol − 1), not reaching statistical significance. were both similar to the levels observed in lean controls. COX10

Figure 1. Characteristics of lean C57BL/6J and obese DKO mice. Scatter dot plots (with mean) of weight, glucose, insulin resistance (HOMA-IR) and blood levels of adiponectin, triglycerides and total cholesterol in lean C57BL/6J control mice (white circles, n = 10), placebo DKO (black diamonds, n = 13), diet-restricted DKO mice (black squares, n = 11), and DKO mice treated with rosiglitazone (black downward triangles, n = 13). **Po0.01 and ***Po0.001 compared with lean C57BL/6J mice; †Po0.05, ††Po0.01 and †††Po0.001 compared with placebo DKO mice.

was lower at 4 months, and COX10 and COX4I1 were lower at 7 years hyperglycemia and insulin resistance in leptin-deficient and compared with lean controls and compared with the expression leptin-sufficient mice. levels 4 months after surgery (Table 2). COX4i1 (1.02 ± 0.24 vs Obese DKO mice with diabetes had higher weight, hyperglycemia 1.26 ± 0.15; Po0.05) and GPX1 (0.40 ± 0.34 vs 0.76 ± 0.34; Po0.05) and insulin resistance (evidenced by higher HOMA-IR) and lower waslowerat4monthsinpatientswhohadMetSat7years plasma levels of adiponectin compared with lean C57BL/6J control compared with those without. Other expressions of investigated mice. Triglyceride and cholesterol levels were higher (Figure 1). RNA markers were not different. Gene expressions before surgery and at expressions of Cox4i1 and Cox10 were lower in white visceral adipose 4 months and 7 years after surgery were not different between tissues of DKO compared with those in lean mice (Figure 2). Impaired patients with or without type 2 diabetes. In obese patients who had adipogenesis was evidenced by lower expression of glucose transporter 4 (Glut4) and Ppar-α (Figure 2). Pparβ/δ (0.56 ± 0.15 vs diabetes, however, COX10 decreased between 4 months and 7 years γ (−0.12 ± 0.18), whereas in patients who did not have type 2 diabetes 0.99 ± 0.24) and Ppar (0.22 ± 0.07 vs 0.91 ± 0.17) were also lower (Po0.001 for both). In addition, RNA expressions of Gpx1 and Sod3 COX10 increased (0.17 ± 0.16; P = 0.01 for difference). were lower (Figure 2). Both caloric restriction and rosiglitazone treatment reduced Mouse studies weight in DKO mice, but the weight of rosiglitazone-treated mice The selection of two mouse models allowed us to study the was still higher than that of lean controls. Caloric restriction did expressions of Cox genes in relation to impaired adipogenesis, not significantly lower glucose and HOMA-IR, but lowered

Figure 2. Gene expressions in white visceral adipose tissues of lean and obese mice. Scatter dot plots (with mean) of RNA expressions of Cox4i1, Cox10, Glut4, Pparα, Gpx1 and Sod3 in adipose tissue samples of lean C57BL/6J control mice (white circles, n = 10), placebo DKO (black diamonds, n = 13), diet-restricted DKO mice (black squares, n = 11) and DKO mice treated with rosiglitazone (black downward triangles, n = 13). *Po0.05 and ***Po0.001 compared with lean C57BL/6J mice; †Po0.05 and †††Po0.001 compared with placebo DKO mice. adiponectin. Rosiglitazone treatment reduced glucose, HOMA-IR compared with those of control mice. They were not different at and adiponectin. None of the interventions lowered triglycerides 8 weeks, but lower at 12 weeks. They were lower in STZ mice at 8 or cholesterol (Figure 1). Caloric restriction did not increase Cox4i1 and 12 weeks than at 4 weeks (Figure 3). Triglyceride levels were and Cox10. In contrast, rosiglitazone treatment increased both higher in STZ mice at 12 weeks, compared with those of control (Figure 2). Caloric restriction and rosiglitazone treatment increased and STZ mice at 4 weeks. Total cholesterol levels were higher in Glut4 and PPARα, with a higher increase with rosiglitazone than STZ mice at 4, 8 and 12 weeks, compared with those of control caloric restriction (Po0.05; Figure 2). In addition, caloric restriction and STZ mice at 4 weeks (Figure 3). Plasma leptin levels in control and rosiglitazone treatment increased Pparβ/δ, (0.95 ± 0.19 and and STZ-treated mice were similar at 4 weeks (1.6 ± 0.6 vs 0.75 ± 0.13; Po0.001 for both) and Pparγ (0.43 ± 0.07 and 1.7 ± 1.1 ng ml − 1), but were higher in STZ mice at 8 weeks 0.69 ± 0.13; Po0.01 and Po0.001). Both treatments also (4.2 ± 1.6 vs 1.4 ± 0.5 ng ml − 1) and at 12 weeks (6.6 ± 5.2 ng ml − 1 increased Gpx1 and Sod3 (Figure 2). Cox4i1 was strongly associated vs 1.3 ± 0.7 ng ml − 1; Po0.01). There were no age-dependent with Glut4 (R2 = 0.48, Po0.001; slope: 0.96 ± 0.15). changes in weight, glucose, HOMA-IR, adiponectin, triglycerides Weight of STZ (streptozotocin and high-fat diet-fed) mice was and total cholesterol in control mice. lower, compared with that of control mice, at 4 weeks and at Cox4i1 was lower in STZ mice with diabetes at 8 and 12 weeks 8 weeks, but not at 12 weeks. STZ mice had hyperglycemia from compared with controls; Cox10 was lower at 12 weeks (Figure 4). 4 weeks on and insulin resistance from 8 weeks on (Figure 3). Glut4 was lower at 8 and 12 weeks; Pparα was lower at 12 weeks. Plasma adiponectin levels of STZ mice were higher at 4 weeks At all time points, Gpx1 and Sod3 were similar in STZ and in control

Figure 3. Characteristics of lean C57BL/6J and STZ mice. Scatter dot plots (with mean) of weight, glucose, HOMA-IR and plasma levels of adiponectin, triglycerides and total cholesterol in lean control mice at 4 weeks (white circles, n = 6), 8 weeks (black circles, n = 6) and 12 weeks (white squares, n = 9) and STZ mice at 4 weeks (black squares, n = 7), 8 weeks (white upward triangles, n = 7) and 12 weeks (black upward triangles, n = 7). **Po0.01 and ***Po0.001 compared with lean control C57BL/6 J mice; †Po0.05, ††Po0.01 and †††Po0.001 compared with STZ mice at 4 weeks.

mice (Figure 4). Pparβ/δ (0.47 ± 0.27 vs 1.02 ± 0.21; Po0.001) was rather associated with impaired insulin signaling than with lower at 12 weeks; Pparγ was not different between STZ mice and impaired adipose tissue differentiation alone. control. Again, there were no age-dependent changes in RNA Previously we reported on the association between oxidative expressions in adipose tissues of control mice. Cox4i1 was strongly stress, obesity and insulin resistance.10,20 Our current data support associated with Glut4 (R2 = 0.62, Po0.001; slope: 0.85 ± 0.18). a possible mechanistic link between attenuated insulin signaling and oxidative stress in the adipose tissue that may be governed by the decreased expression of the insulin-dependent glucose transporter GLUT4.40 Indeed, we found low Glut4 expression in DISCUSSION adipose tissues of leptin-deficient obese mice with diabetes We set out to investigate the association between obesity and after caloric restriction to be associated with low Cox4i1 despite type 2 diabetes and low COX4I1 RNA expression. In humans, presence of elevated adiponectin. In turn, an elevation of Glut4 COX4I1 was lower in monocytes of obese patients with type 2 expression in rosiglitazone-treated obese mice increased Cox4i1. diabetes compared with NGTO and to lean controls. In both This suggests that an improvement of adipose tissue differentia- leptin-deficient and leptin-sufficient mice with diabetes, Cox4i1 tion and glucose uptake is required for restoration of COX expression in visceral adipose tissues was low, indicating that complex components, particularly Cox4i1. In STZ mice, a downregulation of Cox4i1 is leptin independent. PPARγ agonist decreased expression of Cox4i1 was already observed at 8 weeks treatment that led to an expected increase in adiponectin and when Glut4 was low, whereas adiponectin expression was still insulin sensitivity was associated with an increase of Cox4i1 normal. Interestingly, the increase in Pparα in leptin-deficient mice expression. In contrast, caloric restriction, which increased with diabetes after caloric restriction or rosiglitazone treatment adiponectin, but left insulin sensitivity unaffected, did not increase correlated with the expression of its target Glut4. Conversely, the Cox4i1. Cox4i1 expression was already low in STZ mice at 8 weeks decrease in Pparα in STZ mice correlated with the decrease of when insulin sensitivity was low, but when adiponectin was still Glut4. Our combined observations in leptin-deficient obese mice normal. These data support the hypothesis that low Cox4i1 is and STZ mice with diabetes showed also a correlation of Cox4i1

Figure 4. Gene expressions in white visceral adipose tissues of lean C57BL/6J and STZ mice. Scatter dot plots (with mean) of RNA expressions of Cox4i1, Cox10, Glut4, Pparα, Gpx1 and Sod3 in white visceral adipose tissues in control mice at 4 weeks (white circles, n = 6), 8 weeks (black circles, n = 6) and 12 weeks (white squares, n = 9) and STZ mice at 4 weeks (black squares, n = 7), 8 weeks (white upward triangles, n = 7) and 12 weeks (black upward triangles, n = 7). *Po0.05, **Po0.01 and ***Po0.001 compared with age-matched lean control C57BL/6J mice; †Po0.05 and ††Po0.01 compared with STZ mice at 4 weeks. with Pparβ/δ, but not with Pparγ, in agreement with earlier of insulin sensitivity.44 In addition, reactive oxygen species observations in Pparβ/δ over-expressing mice.41 Overall, our production was associated with mitochondrial alterations in the mouse data link low COX4I1 expression to Pparα/β/δ-dependent muscle of hyperglycemic streptozotocin-treated mice, and GLUT4 expression and insulin signaling. The relation between low normalization of glycemia or antioxidant treatment decreased PPAR, low GLUT4, high HOMA-IR and low COX4I1 was also reactive oxygen species production and restored mitochondrial observed in white visceral adipose tissues of obese patients. integrity.45 Thus, our data are in agreement with above literature However, as the number of adipose tissue samples was too low to data showing a close relation between insulin resistance and discriminate between obese patients with diabetes and NGTO oxidative stress. patients, we studied this relation in monocytes, which are more In mouse and human samples, we observed a decrease of clinically accessible. Low COX4I1 in monocytes discriminated several antioxidative systems, such as COX, GPX1 and SOD between obese with diabetes and NGTO patients. Remarkably, systems. Interestingly, low Cox4i1 expression was observed when considering low IRAK3 in addition to low COX4I1 in our adjusted Gpx1 and Sod3 were low, as in DKO mice, or high as in STZ-treated multivariate analysis model improved prediction of type 2 mice, respectively. In aggregate, our data demonstrate that the diabetes in obese patients. This suggests a further synergistic, antioxidant COX complex can be impaired specifically and IRAK3-dependent path to insulin resistance. To this end, independently of other antioxidative systems such as Gpx, which we previously demonstrated that IRAK3 is required for the detoxifies reactive oxygen species,46 and Sod3, which detoxifies PPAR-dependent antioxidative and anti-inflammatory action of superoxide.47 In humans, COX4I1 was lower in monocytes of obese adiponectin.42 patients with type 2 diabetes compared with NGTO and lean Previously a diet high in fat increased the H(2)O(2)-emitting controls, whereas GPX1 was already lower in NGTO. SOD2 potential of mitochondria, shifted the cellular redox machinery to expression was higher in NGTO and patients with diabetes. a more oxidized state, and decreased the redox-buffering capacity Overall, our human data confirm that low GPX and high SOD2 in in the absence of any change in mitochondrial respiratory association with obesity and impaired adipogenesis (low function.43 Interestingly, attenuating mitochondrial H(2)O(2) emis- adiponectin) do not necessarily lead to an impairment of COX. sion completely preserved insulin sensitivity. These findings As the study was based on cross-sectional data, we cannot specifically linked intracellular metabolic balance to the control make conclusions about possible causal relationship. Second, as

© 2015 Macmillan Publishers Limited International Journal of Obesity (2015) 1254 – 1263 Cytochrome oxidase in obesity and diabetes B Van der Schueren et al 1262 we have not gathered data on other interventions than Roux-en-Y 7 Holvoet P, Mertens A, Verhamme P, Bogaerts K, Beyens G, Verhaeghe R et al. bypass, it remains elusive whether other interventions such as Circulating oxidized LDL is a useful marker for identifying patients with coronary sleeve gastrectomy, gastric banding or weight loss induced by artery disease. Arterioscler Thromb Vasc Biol 2001; 21: 844–848. pharmacotherapy would yield similar associations and thus 8 Holvoet P, Harris TB, Tracy RP, Verhamme P, Newman AB, Rubin SM et al. provide similar predictive value for long-term consequences of Association of high coronary heart disease risk status with circulating oxidized LDL in the well-functioning elderly: findings from the Health, Aging, and Body obesity. It has to be said, however, that Roux-en-Y gastric bypass is Composition study. Arterioscler Thromb Vasc Biol 2003; 23: 1444–1448. the preferred and most powerful therapeutic option for 9 Holvoet P, Kritchevsky SB, Tracy RP, Mertens A, Rubin SM, Butler J et al. obese patients with type 2 diabetes. Although we only followed The metabolic syndrome, circulating oxidized LDL, and risk of myocardial 17 patients after Roux-en-Y bypass, this small group still allowed infarction in well-functioning elderly people in the health, aging, and body appreciating significant beneficial effects of gastric bypass on composition cohort. Diabetes 2004; 53: 1068–1073. metabolic parameters, as well as statistically significant changes in 10 Holvoet P, Lee DH, Steffes M, Gross M, Jacobs Jr DR . Association between cir- gene expressions in monocytes from the same patients, culating oxidized low-density lipoprotein and incidence of the metabolic syn- 299 – which were comparable to those observed between NGTO and drome. JAMA 2008; : 2287 2293. T2D patients. 11 Furukawa S, Fujita T, Shimabukuro M, Iwaki M, Yamada Y, Nakajima Y et al. Increased oxidative stress in obesity and its impact on metabolic syndrome. In conclusion, we found that the expression of COX4I1,an J Clin Invest 2004; 114:1752–1761. essential element for ATP feedback inhibition of COX, is down- 12 Antonopoulos AS, Margaritis M, Coutinho P, Shirodaria C, Psarros C, Herdman L regulated in type 2 diabetes. COX4I1 decrease can be independent et al. Adiponectin as a link between type 2 diabetes mellitus and vascular NADPH- from the canonical radical detoxification machinery (Gpx1 and oxidase activity in the human arterial wall: the regulatory role of perivascular Sod3), independent of the IRAK3/adiponectin system and adipose tissue. Diabetes 2014; e-pub ahead of print 31 December 2014. independent of COX biogenesis. COX4I1 decrease is rather linked 13 Liu Y, Palanivel R, Rai E, Park M, Gabor TV, Scheid MP et al. Adiponectin stimulates to decreased glucose uptake and insulin signaling as evidenced by autophagy and reduces oxidative stress to enhance insulin sensitivity during 64 – decrease of GLUT4 expression, which may partly depend on PPAR high-fat diet feeding in mice. Diabetes 2015; :36 48. expression. COX4I1 downregulation results in reduced protection 14 Reilly MP, Lehrke M, Wolfe ML, Rohatgi A, Lazar MA, Rader DJ. Resistin is an inflammatory marker of atherosclerosis in humans. Circulation 2005; 111: against mitochondrial oxidative stress in obesity, and may be 932–939. related to metabolic dysregulation, ultimately leading to MetS and 15 Holvoet P, Vanhaecke J, Janssens S, Van de Werf F, Collen D. Oxidized LDL and type 2 diabetes. As this dysregulation was not only found in malondialdehyde-modified LDL in patients with acute coronary syndromes and visceral adipose tissue, but was also detected in circulating stable coronary artery disease. Circulation 1998; 98:1487–1494. monocytes, monitoring COX4I1 expression offers practical tool for 16 Wallace DC. A mitochondrial paradigm of metabolic and degenerative diseases, patient diagnosis and monitoring. Finally, the COX complex might aging, and cancer: a dawn for evolutionary medicine. Annu Rev Genet 2005; 39: offer new targets for pharmacotherapy in obesity to prevent or 359–407. treat MetS and type 2 diabetes. 17 Poyton RO, McEwen JE. Crosstalk between nuclear and mitochondrial genomes. Annu Rev Biochem 1996; 65:563–607. 18 Villani G, Attardi G. In vivo control of respiration by cytochrome c oxidase in wild- CONFLICT OF INTEREST type and mitochondrial DNA mutation-carrying human cells. Proc Natl Acad Sci USA 1997; 94:1166–1171. fl The authors declare no con ict of interest. 19 Njajou OT, Kanaya AM, Holvoet P, Connelly S, Strotmeyer ES, Harris TB et al. Association between oxidized LDL, obesity and type 2 diabetes in a population- based cohort, the Health, Aging and Body Composition Study. Diabetes Metab Res ACKNOWLEDGEMENTS Rev 2009; 25:733–739. We thank Roxane Menten for her excellent technical support. This work was funded 20 Park K, Gross M, Lee DH, Holvoet P, Himes JH, Shikany JM et al. Oxidative by the Bijzonder Onderzoeksfonds of the KU Leuven (PF/10/014; Center of stress and insulin resistance: the coronary artery risk development in young Excellence), by the Interdisciplinair Ontwikkelingsfonds—Kennisplatform (KP/12/009) adults study. Diabetes Care 2009; 32: 1302–1307. and by the Fonds voor Wetenschappelijk Onderzoek-Vlaanderen (G0846.11 and 21 Arnold S. Cytochrome c oxidase and its role in neurodegeneration and Vascular Biology Network). MH is a postdoctoral fellow of the Fonds voor neuroprotection. Adv Exp Med Biol 2012; 748: 305–339. Wetenschappelijk Onderzoek-Vlaanderen. 22 Arnold S. The power of life--cytochrome c oxidase takes center stage in metabolic control, cell signalling and survival. Mitochondrion 2012; 12:46–56. 23 Williams SL, Valnot I, Rustin P, Taanman JW. Cytochrome c oxidase subassemblies REFERENCES in fibroblast cultures from patients carrying mutations in COX10, SCO1, or SURF1. 279 – 1 Alberti KG, Eckel RH, Grundy SM, Zimmet PZ, Cleeman JI, Donato KA et al. J Biol Chem 2004; : 7462 7469. Harmonizing the metabolic syndrome: a joint interim statement of the Interna- 24 Luo M, Guan X, Luczak ED, Lang D, Kutschke W, Gao Z et al. Diabetes increases 123 tional Diabetes Federation Task Force on Epidemiology and Prevention; National mortality after myocardial infarction by oxidizing CaMKII. J Clin Invest 2013; : – Heart, Lung, and Blood Institute; American Heart Association; World Heart 1262 1274. Federation; International Atherosclerosis Society; and international association for 25 Cummings DE, Overduin J, Foster-Schubert KE. Gastric bypass for obesity: the Study of Obesity. Circulation 2009; 120: 1640–1645. mechanisms of weight loss and diabetes resolution. J Clin Endocrinol Metab 2004; 2 Grundy SM. Obesity, metabolic syndrome, and cardiovascular disease. 89: 2608–2615. J Clin Endocrinol Metab 2004; 89: 2595–2600. 26 Klein S, Burke LE, Bray GA, Blair S, Allison DB, Pi-Sunyer X et al. Clinical implications 3 Grundy SM, Brewer Jr HB, Cleeman JI, Smith Jr SC, Lenfant C. Definition of of obesity with specific focus on cardiovascular disease: a statement for metabolic syndrome: report of the National Heart, Lung, and Blood Institute/ professionals from the American Heart Association Council on Nutrition, Physical American Heart Association conference on scientific issues related to definition. Activity, and Metabolism: endorsed by the American College of Cardiology Arterioscler Thromb Vasc Biol 2004; 24: e13–e18. Foundation. Circulation 2004; 110:2952–2967. 4 Muntner P, He J, Chen J, Fonseca V, Whelton PK. Prevalence of non-traditional 27 Sjostrom L, Lindroos AK, Peltonen M, Torgerson J, Bouchard C, Carlsson B et al. cardiovascular disease risk factors among persons with impaired fasting glucose, Lifestyle, diabetes, and cardiovascular risk factors 10 years after bariatric surgery. impaired glucose tolerance, diabetes, and the metabolic syndrome: analysis of N Engl J Med 2004; 351:2683–2693. the Third National Health and Nutrition Examination Survey (NHANES III). Ann 28 Hulsmans M, Geeraert B, De Keyzer D, Mertens A, Lannoo M, Vanaudenaerde B Epidemiol 2004; 14: 686–695. et al. Interleukin-1 receptor-associated kinase-3 is a key inhibitor of inflammation 5 Ridker PM, Rifai N, Stampfer MJ, Hennekens CH. Plasma concentration of in obesity and metabolic syndrome. PLoS One 2012; 7: e30414. interleukin-6 and the risk of future myocardial infarction among apparently 29 Mertens A, Verhamme P, Bielicki JK, Phillips MC, Quarck R, Verreth W et al. healthy men. Circulation 2000; 101: 1767–1772. Increased low-density lipoprotein oxidation and impaired high-density lipopro- 6 Ridker PM, Buring JE, Cook NR, Rifai N. C-reactive protein, the metabolic tein antioxidant defense are associated with increased macrophage homing and syndrome, and risk of incident cardiovascular events: an 8-year follow-up of 14 atherosclerosis in dyslipidemic obese mice: LCAT gene transfer decreases ather- 719 initially healthy American women. Circulation 2003; 107: 391–397. osclerosis. Circulation 2003; 107: 1640–1646.

International Journal of Obesity (2015) 1254 – 1263 © 2015 Macmillan Publishers Limited Cytochrome oxidase in obesity and diabetes B Van der Schueren et al 1263 30 Verreth W, De KD, Pelat M, Verhamme P, Ganame J, Bielicki JK et al. Weight-loss- 39 Vandesompele J, De Preter K, Pattyn F, Poppe B, Van Roy N, De Paepe A et al. associated induction of peroxisome proliferator-activated receptor-alpha and Accurate normalization of real-time quantitative RT-PCR data by geometric peroxisome proliferator-activated receptor-gamma correlate with reduced averaging of multiple internal control genes. Genome Biol 2002; 3. atherosclerosis and improved cardiovascular function in obese insulin- RESEARCH0034. resistant mice. Circulation 2004; 110: 3259–3269. 40 Wang CH, Wang CC, Huang HC, Wei YH. Mitochondrial dysfunction leads to 31 Ridderstrale M, Carlsson E, Klannemark M, Cederberg A, Kosters C, Tornqvist H impairment of insulin sensitivity and adiponectin secretion in adipocytes. FEBS J et al. FOXC2 mRNA expression and a 5' untranslated region polymorphism of the 2013; 280: 1039–1050. gene are associated with insulin resistance. Diabetes 2002; 51: 3554–3560. 41 Burkart EM, Sambandam N, Han X, Gross RW, Courtois M, Gierasch CM et al. 32 Pickl WF, Majdic O, Kohl P, Stockl J, Riedl E, Scheinecker C et al. Molecular and Nuclear receptors PPARbeta/delta and PPARalpha direct distinct metabolic functional characteristics of dendritic cells generated from highly purified CD14+ regulatory programs in the mouse heart. J Clin Invest 2007; 117:3930–3939. peripheral blood monocytes. J Immunol 1996; 157: 3850–3859. 42 Hulsmans M, Geeraert B, Arnould T, Tsatsanis C, Holvoet P. PPAR agonist-induced 33 Salio M, Cerundolo V, Lanzavecchia A. Dendritic cell maturation is induced by reduction of Mcp1 in atherosclerotic plaques of obese, insulin-resistant mycoplasma infection but not by necrotic cells. Eur J Immunol 2000; 30:705–708. mice depends on adiponectin-induced Irak3 expression. PLoS One 2013; 8: 34 Cynis H, Kehlen A, Haegele M, Hoffmann T, Heiser U, Fujii M et al. Inhibition of e62253. glutaminyl cyclases alleviates CCL2-mediated inflammation of non-alcoholic fatty 43 Rindler PM, Crewe CL, Fernandes J, Kinter M, Szweda LI. Redox regulation of liver disease in mice. Int J Exp Pathol 2013; 94: 217–225. insulin sensitivity due to enhanced fatty acid utilization in the mitochondria. 35 Fujii M, Shibazaki Y, Wakamatsu K, Honda Y, Kawauchi Y, Suzuki K et al. A murine Am J Physiol Heart Circ Physiol 2013; 305: H634–H643. model for non-alcoholic steatohepatitis showing evidence of association between 44 Anderson EJ, Lustig ME, Boyle KE, Woodlief TL, Kane DA, Lin CT et al. diabetes and hepatocellular carcinoma. Med Mol Morphol 2013; 46: 141–152. Mitochondrial H2O2 emission and cellular redox state link excess fat 36 Geeraert B, De Keyzer D, Davey PC, Crombe F, Benhabiles N, Holvoet P. Oxidized intake to insulin resistance in both rodents and humans. J Clin Invest 2009; 119: low-density lipoprotein-induced expression of ABCA1 in blood monocytes pre- 573–581. cedes coronary atherosclerosis and is associated with plaque complexity in 45 Bonnard C, Durand A, Peyrol S, Chanseaume E, Chauvin MA, Morio B et al. hypercholesterolemic pigs. J Thromb Haemost 2007; 5:2529–2536. Mitochondrial dysfunction results from oxidative stress in the skeletal muscle of 37 Hulsmans M, De Keyzer D, Holvoet P. MicroRNAs regulating oxidative stress and diet-induced insulin-resistant mice. J Clin Invest 2008; 118:789–800. inflammation in relation to obesity and atherosclerosis. FASEB J 2011; 25: 46 Kadenbach B, Ramzan R, Vogt S. Degenerative diseases, oxidative stress and 2515–2527. cytochrome c oxidase function. Trends Mol Med 2009; 15:139–147. 38 Livak KJ, Schmittgen TD. Analysis of relative gene expression data using real-time 47 Fridovich I. Superoxide dismutases. An adaptation to a paramagnetic gas. J Biol quantitative PCR and the 2(-Delta Delta C(T)) Method. Methods 2001; 25:402–408. Chem 1989; 264: 7761–7764.

Supplementary Information accompanies this paper on International Journal of Obesity website (http://www.nature.com/ijo)