Abstract
Background
Adiponectin links obesity with insulin resistance, which causes various metabolic abnormalities including dyslipidaemia. Apolipoprotein E (apoE) phenotypes also affect lipoprotein profiles. We aimed to determine whether low adiponectin concentrations are associated with insulin resistance and downstream metabolic abnormalities in obese children.
Methods
We measured fasting concentrations of lipids, apoE, glucose, insulin and adiponectin, as well as anthropometric parameters, in 191 obese children aged 6–15 years. ApoE phenotypes were determined by isoelectric focusing. Boys (n = 79) and girls (n = 39) with apoE3/3 were classified into tertiles according to their adiponectin concentrations. Metabolic parameters, were compared among these three groups in boys and girls separately.
Results
The low adiponectin groups had higher median homeostasis model assessment of insulin resistance (HOMA-IR) than the middle and high adiponectin groups in both boys [5.3 (low) versus 3.1 (middle; P < 0.05) and 3.5 (high; P < 0.05)] and girls [5.0 (low) versus 4.4 (middle) and 3.0 (high; P < 0.05)]. However, only boys who were in the low adiponectin group exhibited significantly higher concentrations of blood pressure, triglycerides, LDL-cholesterol, and remnant-like particle-cholesterol, and lower concentrations of HDL-cholesterol compared with the middle or high adiponectin groups.
Conclusion
Low adiponectin concentration is associated with insulin resistance in obese children. Furthermore, decreased adiponectin with E3/3 exhibited more prominent downstream metabolic abnormalities in obese boys than in obese girls.
Introduction
The prevalence of childhood obesity in Japan has been steadily increasing because of the introduction of Westernized life styles, including a high-calorie, high-fat diet and low physical activity. 1 Similar to obese adults, obese children also suffer from multiple health problems closely related to insulin resistance. 2–4 In typical cases, they have hypertension, 3,4 dyslipidaemia (hypertriglyceridaemia and/or hypo-α-lipoproteinaemia), 2,3 and impaired glucose tolerance. 3 Studies have shown that these metabolic abnormalities inevitably accumulate in obese subjects because of imbalances in adipocytokines (bioactive substances secreted from adipocytes). 5
Adiponectin is an anti-atherogenic adipocytokine, and serum concentrations have been found to be reduced in obese subjects. 6,7 In human serum, adiponectin exists as three types of multimers with differing molecular weights. Adiponectin multimers with high molecular weights bind to adiponectin receptors more efficiently than those with middle or low molecular weights. 8 Although subjects with low adiponectin concentrations have more metabolic abnormalities than those with high adiponectin concentrations, clinical manifestations vary considerably among subjects with similar adiponectin concentrations. 9–12 Such individual variations may result from the effects of other confounding factors such as apolipoprotein E (apoE). ApoE phenotypes significantly affect serum concentrations of LDL-cholesterol (LDL-C), triglycerides (TG) and remnant-like particle cholesterol (RLP-C). 13–17 Furthermore, apoE phenotypes are thought to be associated with insulin resistance in obese women. 18
The aim of this study was to determine whether low adiponectin concentrations are associated with insulin resistance and downstream metabolic abnormalities in obese children. To eliminate the confounding effects of differing apoE phenotypes, we selected only E3/3 children to be included in the study and compared metabolic abnormalities among subgroups classified by adiponectin concentrations.
Methods
Subjects
For this study, we enrolled 191 obese (132 boys and 59 girls) and 91 non-obese children (50 boys and 41 girls) in the age range of 6–15 years. The obese children were selected from public schools in Niigata Prefecture, Japan. A percentage of the standard weight (POW) was calculated according to sex, age and body height, as reported by the Ministry of Education, Sciences and Culture of Japan in 1990. 19 We defined obesity as POW equal to or more than 20%.
We measured several anthropometric parameters including height, weight and systolic and diastolic blood pressures (SBP and DBP) in all children. Fasting (at least 12 hours) and non-fasting venous blood was collected from obese and normal children, respectively, for measurement of lipoprotein and adiponectin later. Serum was separated with low-speed centrifugation, and aliquots were frozen at −80°C until used for later analysis. Informed consent for all children was given from the parents or guardians before blood sampling. This study was approved by the ethics committee of Niigata University Graduate School of Medical and Dental Sciences.
Determination of Apolipoprotein E phenotypes
ApoE phenotypes were determined by isoelectric focusing (IEF) followed by immunoblotting. 20 Briefly, a serum sample (100 μL) was mixed with the pretreatment buffer (0.005 mol/L dithiothreitol, 0.5% Tween 20) at 4°C for 15 minutes. IEF was carried out at 8°C and 13 W for 2000 Vh using a 5% polyacrylamide gel containing 5% ampholyte (pH 4.5–5.4 and pH 5–8, 1:2) and 3 mol/L urea. The separated proteins were transferred overnight onto the nitrocellulose membrane by simple diffusion at 4°C.
Nitrocellulose membranes were reacted with polyclonal rabbit antibodies against human apoE, and then horseradish-conjugated goat antibodies against rabbit immunoglobulin. After visualizing apoE bands by phosphate-buffered saline containing diaminobenzidine (0.072 g/L) and 0.1% hydrogen peroxide (v/v), we determined apoE phenotypes from the multiple banding patterns. 21
In some cases with unclear results, we desialylated apoE bands prior to IEF. 22 Serum samples (10 μL) were incubated with 0.1 mol/L acetate buffer (pH 5.0; 20 μL) containing 2 unit/mL neuraminidase (10 μL) at 37°C for 4 hours. The mixture was delipidated with acetone/ethanol (1:1, v/v) at −80°C for 4 hours, acetone/ethanol solution at −80°C for 2 hours, and then diethyl ether at −80°C for 1 hour. The delipidated protein was dried under a stream of nitrogen. The desialylated samples were incubated with 100 μL sample buffer (0.01 mol/L Tris–HCl [pH 8.6] containing 0.01 mol/L dithiothreitol and 5.2 mol/L urea) for 1 hour at room temperature. IEF was carried out in the same manner as described above except that we used a polyacrylamide gel that contained 5.2 mol/L urea.
Measurements of lipoproteins and adiponectin
Total cholesterol (TC) and TG concentrations were measured enzymatically using an automatic analyser (Hitachi-7450, Tokyo, Japan). LDL-C and HDL-C concentrations were determined by homogeneous assays, and apoE concentration was measured by a turbidometric immunoassay. RLP-C concentrations were determined by detergent-based homogenous assay (RemL-C; Kyowa Medex, Tokyo, Japan). 23 The values obtained using these kits were highly correlated with those measured by the original methods of Nakajima et al. 24 using an immunoaffinity gel. Serum total adiponectin concentrations were measured by immunoassay using a commercial kit (Human Adiponectin ELISA Kit; Otsuka, Osaka, Japan). As we could not obtain fasting samples from non-obese children, TG, RLP-C and adiponectin were measured only in the obese children.
Insulin resistance
Fasting plasma glucose (FPG) concentration was measured using an enzymatic method. Serum insulin concentrations were measured with a commercial enzyme-linked immunoassay kit (LS ‘Eiken’ Insulin; Eiken Chemical Co., Ltd., Tokyo, Japan). Insulin resistance was evaluated with the homeostasis model assessment of insulin resistance (HOMA-IR) calculated by the following equation:
Statistical analysis
In this study, we compared variables according to sex. Data were reported as the mean (±SD) or median (interquartile range). Data distributions were checked for normality using a one-sample Kolmogorov-Smirnov test. Markedly skewed data were logarithmically transformed prior to statistical analysis. Continuous variables between groups were compared by either unpaired Student's t-test or non-parametric Mann-Whitney U test. Multiple stepwise regression analysis was carried out to define the independent risk factors for insulin resistance using Statcel (OMS, Tokorozawa, Japan), an add-in software for Microsoft Excel (Microsoft Japan, Tokyo, Japan). Differences were considered significant at P values <0.05 (two-sided tests).
Results
Apolipoprotein E phenotype and ϵ allele frequencies
Obese and non-obese children had similar relative frequencies of apoE phenotypes and ϵ alleles. In both groups, about two-thirds of the children were E3/3. The incidences of other major phenotypes (E4/3 and E3/2) were similar between the obese and non-obese children. As reflected by these data, the frequencies of ϵ2, ϵ3 and ϵ4 were similar between the two groups. One apoE5 carrier was found in the non-obese group, while four apoE7 carriers were detected in the obese group. Because of the small number of subjects for these two phenotypes, their prevalence did not reach statistical significance (Table 1).
Comparisons of apoE phenotypes and ϵ allele frequencies between obese and non-obese children
ApoE phenotypes were determined by isoelectric focusing followed by immunoblotting. The apoE5 and apoE7 bands were confirmed using serum pretreated with neuraminidase as described in the Subjects and Methods section
In the following comparison, we focused our analyses on subjects with the three most common phenotypes (i.e. E2/3, E3/3 and E3/4) as the other phenotypes accounted for <1–3% of the study subjects. Two apoE3/3 children were excluded from the analysis because they were not fasting at the time of blood sampling.
Effects of apolipoprotein E phenotypes on lipoprotein profiles and insulin resistance
The effects of apoE phenotypes on lipoprotein profiles were more evident in obese children than in non-obese children. However, apoE phenotypes did not affect SBP or DBP concentrations in either group. In non-obese children, apoE concentration was the only metabolic parameter that differed significantly among the three most common apoE phenotypes. In boys, apoE concentration was highest in the E3/2 carriers, lowest in the E4/3 carriers and intermediate in the E3/3 carriers (Table 2). In girls, apoE concentration was significantly higher in E3/2 carriers than in E4/3 and E3/3 carriers.
Effects of apoE phenotypes on anthropometric and lipoprotein profiles in non-obese children
POW, a percentage of the standard weight for age, height and sex; SBP, systolic blood pressure; DBP, diastolic blood pressure; TC, total cholesterol; LDL-C, low-density lipoprotein-cholesterol; HDL-C, high-density lipoprotein-cholesterol; apoE, apolipoprotein E. Data are presented as the mean (SD) except for apoE (median (25th, 75th percentiles])
*P < 0.05 vs. E3/2 children
† P < 0.05 vs. E3/3 children
In contrast, in obese children, TC and LDL-C concentrations were highest in the E4/3 carriers, lowest in the E3/2 carriers, and intermediate in the E3/3 carriers (Table 3). TG and RLP-C concentrations did not differ among the three phenotypes. As in the non-obese children, apoE concentration was highest in the E3/2 carriers independent of sex.
Effects of apoE phenotypes on anthropometric and lipoprotein profiles, and HOMA-IR values in obese children
POW, a percentage of the standard weight for age, height and sex; SBP, systolic blood pressure; DBP, diastolic blood pressure; TC, total cholesterol; LDL-C, low-density lipoprotein-cholesterol; HDL-C, high-density lipoprotein-cholesterol; TG, triglycerides; apoE, apolipoprotein E; RLP-C, remnant-like particle cholesterol; FPG, fasting blood plasma; HOMA-IR, homeostasis model assessment of insulin resistance. Data are presented as mean (SD) or median (25th, 75th percentiles)
*P < 0.05 vs. E3/3 children
† P < 0.05 vs. E3/2 children
Note that E4/3 obese girls, but not boys, displayed abnormalities in parameters related to insulin resistance. Fasting insulin was greater in E4/3 carriers than in E3/3 (P < 0.05) and E3/2 carriers (P = 0.07). Furthermore, HOMA-IR values in obese girls tended to be higher in E4/3 carriers than in E3/3 (P = 0.05) and E3/2 carriers (P = 0.07).
Effects of adiponectin concentration on metabolic parameters in obese children
First, we examined the effects of adiponectin on metabolic parameters in obese children with various apoE phenotypes. Obese boys and girls were divided into tertiles according to their adiponectin concentrations. The median HOMA-IR value was higher in the low adiponectin group than either the middle or high adiponectin group in the obese boys (5.59 [4.07, 7.35] (low) versus 3.68 [2.64, 4.85] (middle), P < 0.001; versus 3.12 [2.09, 4.93] (high), P < 0.001) and girls [5.01 (3.73, 7.90) (low) versus 4.72 (3.28, 8.67) (middle), not significant; versus 2.98 (2.29, 4.03) (high), P < 0.01]. However, we failed to find any significant differences among the subgroups in both obese boys and girls for any other variables (data not shown).
Second, we performed the same analysis in obese children who had the same apoE phenotype. In both obese boys and girls, children with the same apoE phenotype were divided into tertiles according to their adiponectin concentrations. However, we did not examine obese girls with E4/3 and E3/2 because of the small number of individual subgroups. In the obese children with E3/3, fasting insulin concentrations were highest in the low adiponectin groups in both obese boys and girls, although FPG concentrations were not different among groups (Table 4). The low adiponectin groups had almost a two-fold higher HOMA-IR than either the middle or high adiponectin groups (Figure 1, a). The median values of both adiponectin and HOMA-IR were very similar between the corresponding groups of obese boys and girls (Figure 1, a).
Homeostasis model assessment of insulin resistance (HOMA-IR) and remnant-like particle cholesterol (RLP-C) in obese E3/3 children according to sex and serum adiponectin concentration. Obese boys and girls with E3/3 were classified into tertiles according to their fasting adiponectin concentrations. HOMA-IR (a) was calculated from fasting plasma glucose and insulin concentrations. Fasting RLP-C concentrations (b) were measured by the detergent-based method as described in the Subjects and Methods. Upper and lower edges of the boxes indicate the 75th and 25th percentiles, respectively. The lines and closed circles in the boxes denote the median and mean values of the data. Upper and lower ends of the vertical lines indicate the 90th and 10th percentiles, respectively
Effects of adiponectin concentrations on anthropometric and lipoprotein profiles, and homeostasis model assessment of insulin resistance values in obese children with apoE3/3
POW, a percentage of the standard weight for age, height and sex; SBP, systolic blood pressure; DBP, diastolic blood pressure; TC, total cholesterol; TG, triglycerides; apoE, apolipoprotein E; FPG, fasting plasma glucose. Data are presented as the mean (SD) or median (25th, 75th percentiles). Subjects are divided into tertiles (low, middle and high groups) according to adiponectin concentrations; cut-off values of adiponectin concentrations were 5.6 and 7.6 μg/mL in boys and 5.0 and 6.5 μg/mL in girls
*P < 0.05 vs. middle adiponectin groups
† P < 0.05 vs. high adiponectin groups
In the obese boys with E4/3 and E3/2, fasting insulin concentrations were higher in the low adiponectin groups (28.9 ± 18.0 mU/L; 30.2 ± 17.8 mU/L) than in the middle (16.9 ± 4.8 mU/L; P = 0.07; 26.7 ± 8.8 mU/L) and high adiponectin groups (18.0 ± 15.1 mU/L; 14.2 ± 11.4 mU/L, P = 0.1). The mean FPG concentrations were very similar among groups. As a result, the mean values of HOMA-IR tended to be higher in the low adiponectin groups (6.7 ± 4.1; 6.8 ± 3.9) than in the middle (4.0 ± 1.0, P = 0.08; 6.2 ± 2.0) and high adiponectin groups (4.0 ± 3.2; 3.2 ± 2.6, P = 0.1).
Unexpectedly, the effects of adiponectin on anthropometric and lipoprotein parameters were more prominent in obese boys than in obese girls. In obese boys with E3/3, the low adiponectin group had slightly but significantly higher SBP and DBP concentrations compared with the middle and high adiponectin groups (Table 4). However, no significant difference was observed in BP among the three groups of obese girls. In obese boys, the low adiponectin group had the highest LDL-C (Figure 2a) and the lowest HDL-C concentrations (Figure 2b) of the three groups. The respective median values of TG and RLP-C were 43 and 69% higher in the low adiponectin group compared with those in the middle adiponectin group. In addition, the respective median values of TG and RLP-C were 56 and 96% higher in the low adiponectin group compared with those in the high adiponectin group (Table 4; Figure 1b). However, the median apoE concentration was very similar among the three groups (Table 4). In contrast, in obese girls with E3/3, a lower adiponectin concentration showed no significant association with any lipoprotein parameter (Table 4, Figures 1 and 2).
LDL-C and HDL-C concentrations in obese E3/3 children according to sex and serum adiponectin concentration. Obese boys and girls with E3/3 were classified into tertiles according to their fasting adiponectin concentrations. L, M, and H represent the low, middle, and high adiponectin groups, respectively. Fasting LDL-C (a) and HDL-C (b) concentrations were measured by homogenous assays. Data are presented as the mean ± SD
Multiple stepwise regression analysis was performed in obese boys and girls separately, using HOMA-IR as a dependent variable, and using age, adiponectin, apoE phenotype, POW, SBP and TG as independent variables. On account of collinearity, insulin, glucose, HDL-C, apoE, and RLP-C were excluded from the independent variables. These models yielded sufficient correlations in obese boys (R = 0.660, P < 0.00001) and obese girls (R = 0.815, P < 0.00001). In the obese children, adiponectin concentration was a significant determinant of HOMA-IR (Table 5). Age, TG and POW were also selected as independent variables for HOMA-IR.
Multiple stepwise regression analysis for homeostasis model assessment of insulin resistance
Multiple stepwise regression analysis was performed in obese boys and girls separately.
The F value to enter was set at 2.0 at each step. Sex was excluded from independent variables in this model
POW, a percentage of the standard weight for age, height and sex; TG, triglyceride; SBP, systolic blood pressure; apoE, apolipoprotein E
Discussion
Our findings demonstrate that low adiponectin concentrations are associated with insulin resistance and downstream metabolic abnormalities in obese children. This association seems to be more evident in obese boys than girls. We found that the low adiponectin groups had greater HOMA-IR values than the middle and high adiponectin groups in obese children, independent of sex (Figure 1a; Table 5). Obese boys with E3/3 in the low adiponectin group had high blood pressure, LDL-C, TG and RLP-C, as well as low HDL-C; however, the corresponding obese girls demonstrated fewer clinical manifestations (Table 4; Figures 1 and 2).
Many studies have shown that the apoE phenotype is an important regulator of lipoprotein profiles. 13–18 Three common apoE alleles, ϵ2, ϵ3 and ϵ4, result in six major apoE phenotypes (E2/2, E3/2, E4/2, E3/3, E4/3 and E4/4). In addition, many apoE variants, including apoE5 and E7, have been identified. 25–28 Among them, E3/3 is the most frequent phenotype reported in every country. The prevalence of E3/3 in Japan is reported to be between 63.6 and 74.2%, 29–31 which is in accordance with our data (Table 1). E4 is associated with high LDL-C and low apoE concentrations, whereas E2 is associated with low LDL-C and high apoE concentrations. 13,15,17,18,29,32 E5 and E7 were first discovered in Japan by Yamamura et al., 25,26 who found that both apoE variants were frequently associated with coronary artery disease (CAD) and hypercholesterolaemia. In our studies, one E5 carrier and four E7 carriers were detected in non-obese and obese children, respectively, and they tended to have hypercholesterolaemia.
Note that apoE phenotypes affected the lipoprotein profiles of obese children, but not non-obese children (Tables 2 and 3). Such inconsistencies are most likely attributable to differences in dietary fat and carbohydrate intake between the two groups. In young Finns aged 9–24 years, increases in LDL-C owing to E4 were 1.9- and 1.4-fold greater in those individuals on a high saturated fatty acid (SAFA)-cholesterol diet (high SAFA/Chol group) compared to those consuming low and middle SAFA-cholesterol diets (low and middle SAFA/Chol groups). 33 In their study, the LDL-C-lowering effect of E2 was 2.4- and 1.6-fold greater in the high SAFA/Chol group compared with the low and middle SAFA/Chol groups, respectively. In the CAD patients, the E2-associated increases in TG were evident only in those on a high sucrose diet. 34 Furthermore, the analyses using polyacrylamide gel electrophoresis and ultracentrifugation revealed that most E7 heterozygotes exhibited elevated remnant lipoproteins irrespective of the presence or absence of hyperlipidaemia. This abnormality was markedly ameliorated by a low-calorie, low-fat and low-cholesterol diet. 27
Collectively, it appears that the effects of adiponectin on lipoprotein profiles are likely to be masked in groups of obese children who consist of various apoE carriers. In general, serum adiponectin concentration is closely associated with visceral adiposity, 6,7 hypercholesterolaemia, 9–12 hypertriglyceridaemia, 6,9,10,12,35 hypo-α-lipoproteinaemia, 6,9,10,12 hypertension, 11,36 hyperglycaemia and insulin resistance. 6,7,9–11 In obese children, however, studies sometimes found only limited clinical manifestations associated with the low adiponectin concentration. In Japanese obese boys, for example, adiponectin concentration was inversely related to both LDL-C and fasting insulin concentrations; however, no correlations with SBP, DBP, TG and HDL-C were observed. 37 In American obese children and adolescents, adiponectin concentrations were positively associated with the whole body insulin sensitivity index and HDL-C, but no associations were found with LDL-C, TG, SBP, DBP and HOMA-IR. 38 When we analysed all obese children with various apoE phenotypes together, the low adiponectin group was significantly correlated with high HOMA-IR, but not with other metabolic abnormalities. However, when the same analysis was performed in E3/3 children, low adiponectin concentrations were significantly associated with several metabolic parameters including TG, LDL-C, HDL-C, RLP-C, SBP, DBP and HOMA-IR (Table 4; Figures 1 and 2). Therefore, when apoE phenotypes are taken into account, the serum adiponectin concentration appears to have a much greater impact on these parameters than previously reported.
Our findings suggest that some sex differences exist in the effects of adiponectin on metabolic parameters. We found that obese boys in the low adiponectin group had a greater number of abnormalities in lipoprotein profiles and blood pressure than the corresponding obese girls (Table 4; Figures 1 and 2). As both groups had similar values of serum adiponectin and HOMA-IR, the female subjects may have been protected against atherosclerosis by a low susceptibility to metabolic abnormalities induced by insulin resistance. In most cohort studies, menopause decreases HDL-C, but increases LDL-C and TG concentrations. 39 Hormone replacement therapy reduces LDL-C and BP, but increases HDL-C. 40,41 In the obese girls, these favourable effects of oestrogen may obscure the metabolic abnormalities caused by low adiponectin concentration. Further studies are needed to elucidate the specific underlying mechanisms.
We conclude that a low adiponectin concentration is associated with insulin resistance and downstream metabolic abnormalities in obese children, notably when having E3/3. This association seems to be more evident in obese boys than in obese girls. We speculate that the decreased susceptibility of girls to the effects of low adiponectin concentrations may partially account for the low frequency of observed metabolic syndrome and CAD in adult females.