2-Aminoadipic acid (2-AAA) as a potential biomarker for insulin resistance in childhood obesity

Insulin resistance is an important clinical feature of metabolic syndrome, which includes obesity and type 2 diabetes. Increased adipose energy storage in obesity promote insulin resistance and other metabolic adverse effects. To identify a new link between adipocyte and insulin resistance, we performed targeted metabolite profiling of differentiated adipocytes and studied the association between adipogenic metabolites and insulin resistance. We found a correlation between 2-aminoadipic acid (2-AAA) and adipogenic differentiation. Also, circulatory 2-AAA was positively associated with obesity-related factors (fat mass, fat percent, waist circumference, BMI, BMI z-score, triglycerides, insulin, and HOMA-IR) at baseline and after 2 years in the children cohort study. Of these factors, increased BMI z-score and HOMA-IR were the primary independent factors associated with higher 2-AAA levels, and the baseline 2-AAA level was an indicator of the BMI z-score after 2 years. To validate the relationship between 2-AAA and obesity-related factors, we analyzed changes in 2-AAA levels following obesity intervention programs in two independent studies. In both studies, changes in 2-AAA levels during the intervention period were positively correlated with changes in the BMI z-score and HOMA-IR after adjusting for confounders. Moreover, the 2-AAA levels were increased in cell and mouse models of obesity-related insulin resistance. Excess 2-AAA levels led to impaired insulin signaling in insulin-sensitive cells (liver, skeletal muscle and adipose cells) and caused abnormal gluconeogenesis. Our results demonstrate that 2-AAA is associated with adipogenesis and insulin resistance. In this regard, 2-AAA could be a potential biomarker of obesity and obesity-related metabolic disorders.

Among the metabolites evaluated, 2-aminoadipic acid (2-AAA) is produced by lysine degradation. Lysine residues are deaminated by metal-catalyzed oxidation to form allysine, which is oxidized to generate 2-AAA 11 . In previous studies, the levels of 2-AAA and other lysine pathway metabolites were significantly higher in diabetic mice. The level of allysine, a precursor of 2-AAA, was also increased in streptozotocin-induced diabetic rats 12,13 . In a metabolite profiling analysis of two cohorts with long follow-up periods, 2-AAA was associated with insulin resistance and markers predictive of an increased future risk of type 2 diabetes 14 .
As discussed above, altered levels of 2-AAA are important in the development of diabetes; however, whether they are also associated with obesity-related insulin resistance remains unclear. Therefore, we hypothesized that increased levels of 2-AAA are involved in adipogenesis and may be associated with insulin resistance. We evaluated whether the levels of 2-AAA were correlated with adipogenic differentiation at the cellular level and obesity-related insulin resistance in cross-sectional and follow-up cohort/intervention studies. Additionally, we sought to confirm the association between the levels of 2-AAA and insulin resistance using cell and mouse models. Next, we assessed whether accumulation of 2-AAA leads to insulin resistance, which could occur through impaired insulin signaling and abnormal gluconeogenesis.

Results
Association of 2-AAA levels with adipogenesis. Excess fat has been linked to many insulin resistance-related states, including obesity and type 2 diabetes 4,15 . Therefore, we investigated whether various metabolites were altered following the differentiation from preadipocytes to adipocytes. Cell lysates were extracted from subcutaneous preadipocytes and adipocytes (from a nondiabetic 39-year-old Caucasian woman with a body mass index (BMI) of 38 kg/m 2 ), and the metabolite levels were measured. Following differentiation, metabolites with significantly decreased and increased levels were screened (Supplementary Table 1). Surprisingly, 2-AAA was detected in differentiated adipocytes but not in preadipocytes (Table 1). Thus, we found that alteration of 2-AAA levels was associated with adipogenesis.
Association of 2-AAA levels with obesity and obesity-related factors in a cross-sectional study. To evaluate the association of 2-AAA levels with obesity and obesity-related factors, we first examined 2-AAA levels according to obesity status (Supplementary Table 2, Fig. 1) in a cross-sectional study (KoCAS-1). The anthropometric measurements, lipid profiles, glycemic index, and 2-AAA levels were higher in the The associations between plasma 2-AAA levels and obesity-related factors are shown in Table 2. Circulatory 2-AAA was positively associated with adiposity indices (fat mass, fat percent, waist circumference, BMI, and BMI z-score; all p ≤ 0.0336), lipid parameters (TG level; p < 0.0001), and glycemic parameters (glucose and insulin levels, HOMA-IR; all p ≤ 0.004) but was negatively correlated with the HDL-C level (p < 0.0001). According to the multiple linear stepwise regression results, the BMI z-score and HOMA-IR value were independent factors for high plasma 2-AAA levels ( Table 2).
Replication of the results in the 2-year follow-up study. We next performed replication studies in independent subjects in the 2-year follow-up study (KoCAS-2; Supplementary Table 3). As in KoCAS-1, the baseline levels of 2-AAA were positively associated with baseline adiposity indices (fat mass, fat percent, waist circumference, BMI, and BMI z-score; all p ≤ 0.0001), lipid parameters (TG level; p = 0.0016), and glycemic parameters (insulin level and HOMA-IR; all p ≤ 0.0004) but were negatively correlated with the HDL-C level (p = 0.0105; Supplementary Table 4). The association between the baseline levels of 2-AAA and these obesity-related factors remained significant after 2 years (Table 2).

Simple regression
Stepwise regression b In the stepwise analyses, the baseline 2-AAA level was associated independently with the baseline BMI z-score, glucose level, and HOMA-IR (Supplementary Table 4). In particular, the baseline 2-AAA level was a major indicator of the BMI z-score after 2 years (Table 2).
The independent obesity intervention study also showed lower levels of 2-AAA in the BMI z-score loss group compared with the BMI z-score gain group following a 6-month intervention (p = 0.0148; Supplementary Table 6 Table 7 Alteration of 2-AAA levels by insulin resistance. Chronic consumption of a high-fat diet is closely related to obesity and insulin resistance 16,17 . Whole proteins were extracted from the adipose tissues of C57BL6 mice fed a high-fat diet and from age-matched mice fed a standard diet, and the levels of 2-AAA were assessed. As expected, the high-fat diet-fed mice exhibited higher body weight and glucose tolerance test (GTT) scores than the standard diet-fed mice (data not shown). Interestingly, 2-AAA levels were elevated in the adipose tissue of mice fed a high-fat diet compared with mice fed a standard diet (Fig. 4A). Palmitic acid (PA) mediates abnormal gluconeogenesis by inducing endoplasmic reticulum (ER) stress and insulin resistance 18,19 . We confirmed that ER stress markers (peIF2α, CHOP, and GRP78) and gluconeogenesis-related factors (PEPCK, PGC1α, and G6Pase) were increased in PA-treated SK-Hep I human liver cells (Fig. 4C, Supplementary Fig. 1). Additionally, excess PA resulted in elevated TG levels compared with the control (Fig. 4D). Remarkably, the levels of 2-AAA were higher in both lysates and supernatants from SK-Hep I cells treated with PA (Fig. 4B). These results suggest that increased levels of 2-AAA are associated with insulin resistance, which induces abnormal gluconeogenesis. Impaired insulin signaling by increased levels of 2-AAA. In this study, increased 2-AAA levels were correlated with insulin resistance and obesity ( Table 2, Figs 1-3). Additionally, we observed that 2-AAA levels were related to adipogenesis and increased in animal and cell models of induced insulin resistance (Table 1, Fig. 4). Therefore, excessive 2-AAA is a risk factor for insulin resistance. To explore the effect of 2-AAA on insulin signaling, SK-Hep I human liver cells, C2C12 mouse myotubes, and human subcutaneous adipocytes were treated with 2-AAA. Insulin signaling-related events, such as phosphorylation of AKT and phosphorylation of IR, were significantly reduced in 2-AAA-treated SK-Hep I cells (Fig. 5A, Supplementary Fig. 2A). At the same time, gluconeogenesis was significantly increased as a result of increased expression of PEPCK, G6Pase, and PGC1α (Fig. 5B, Supplementary Fig. 2B). As expected, AKT phosphorylation was decreased in 2-AAA-treated C2C12 mouse myotubes (Fig. 5C, Supplementary Fig. 2C). Treatment with 2-AAA caused a reduction in phosphorylation of AKT and a slight decrease in phosphorylation of IR in human subcutaneous adipocytes (Fig. 5D,  Supplementary Fig. 2D). Overall, these results indicate that excess 2-AAA impairs insulin signaling. An increase in 2-AAA levels may contribute to the future development of diabetes due to abnormal gluconeogenesis.

Discussion
Adipose tissue is an important organ involved in the regulation of insulin sensitivity and development of diabetes through its fat storage capacity and thermogenic regulation. However, when adipose tissue is no longer able to store fat, excess fat accumulates in ectopic lipid depots, such as the liver, intra-abdominal/visceral sites, and skeletal muscles. Increased ectopic lipid accumulation leads to dyslipidemia and insulin resistance [20][21][22] .
Insulin resistance is caused by impaired insulin signaling, and β cells promote abnormal production of insulin to maintain glucose homeostasis. When appropriate insulin secretion is not achieved in the pancreas, hyperglycemia and insulin resistance occur. Insulin resistance is an important clinical feature of metabolic syndrome, which includes obesity and type 2 diabetes [3][4][5]23 . Little is known about the metabolic mechanism involved in adipogenesis and insulin resistance. In this study, we differentiated human subcutaneous preadipocytes into adipocytes and performed metabolite profiling. Surprisingly, 2-AAA was detected in adipocytes but not in preadipocytes (Table 1). 2-AAA is produced primarily by oxidative stress and has been reported to be strongly linked to an increased risk of diabetes 11,14 . Thus, we generated a mouse model using C57BL6 mice in which insulin resistance was induced via the accumulation of excess fat; interestingly, 2-AAA levels were found to be elevated in the adipose tissue of these mice (Fig. 4A). We showed that excess palmitate plays an important role in the abnormal increase in gluconeogenesis by inducing the ER stress response and impairing insulin signaling 19 . As expected, the levels of 2-AAA were elevated in PA-treated SK-Hep I cells (Fig. 4B). Therefore, our results showed that 2-AAA is strongly correlated with insulin resistance.
The incidence of childhood obesity continues to increase globally and is a major public health challenge because it can promote many disorders in adulthood. It can also cause many acute health problems, including type 2 diabetes, early puberty, hypertension, musculoskeletal disorders, and psychological issues, leading to a considerable amount of suffering. Therefore, it is important to establish means for early prediction and preventive strategies for obesity because of the close correlation between childhood and adult obesity 1,2,24 . We identified a positive correlation of the plasma levels of 2-AAA with BMI and HOMA-IR in KoCAS. 2-AAA levels were significantly associated with obesity-related factors, including lipid parameters and HOMA-IR, at baseline and the 2-year follow-up. These results were validated in the obesity intervention study (Table 2, Fig. 3), in which changes  www.nature.com/scientificreports www.nature.com/scientificreports/ intake have also been found to be correlated with obesity and insulin resistance, although some lipids exhibit differences according to race [28][29][30] . Previous research from our laboratory showed that ADMA is a potential biomarker for obesity-related insulin resistance 9 . Thus, the discovery of new biomarkers for disease can improve prediction and prevention in high-risk groups and establish new biological pathways associated with the onset of disease. Therefore, the identification of 2-AAA as a risk factor for childhood obesity and insulin resistance is noteworthy. Overall, our results demonstrate that 2-AAA is associated with adipogenesis and that altered levels of 2-AAA impair insulin signaling. Thus, 2-AAA could be a potential biomarker for insulin resistance-related obesity and may contribute to the early prevention of metabolic disorders, including obesity and type 2 diabetes.

Materials and Methods
Study population. The data were obtained from the Korean Children-Adolescents Study (KoCAS), conducted by the Korean National Institute of Health. The KoCAS-1 subjects were 449 adolescents aged 12-16 years from Seoul and Gyeonggi provinces for whom clinical biomarker data were collected between 2011 and 2012. The KoCAS-2 subjects were 200 independent individuals aged 9-11 years in 2008-2009, from whom health and metabolite data were obtained once more after 2 years. Obesity was defined as a body mass index (BMI) > 25 kg/m 2 or being in the 95 th percentile for age and sex according to the 2007 Korean growth standard for children and adolescents. www.nature.com/scientificreports www.nature.com/scientificreports/ For the short-term obesity intervention study, 88 morbidly obese adolescents aged 12-16 years with BMIs ≥ 99 th percentile (or ≥30 kg/m 2 ) were recruited from Seoul and Gyeonggi provinces in 2012 as part of KoCAS. The intervention program included minimal exercise and nutrition education, in which the subjects participated three times over the course of 10 weeks. Data were also obtained from the Intervention for Childhood and Adolescent Obesity via Activity and Nutrition (ICAAN) study performed in 2016. The ICAAN study is an intensive multidisciplinary intervention program that includes exercise, nutrition education, and behavioral modification for obese children in Korea. In this study, we used data from 67 children aged 9-13 years with BMIs ≥ 90 th percentile who had completed the 6-month intervention program.
We obtained written informed consent prior to study participation from the parents of all subjects. All study protocols were approved by the institutional review board of Seoul-Paik Hospital, Inje University for the KoCAS (IIT-2009 Clinical variables. Professionally trained personnel performed the anthropometric examinations using a standardized protocol. Height was measured with an automatic stadiometer (DS-102; Dong Sahn Jenix Co., Ltd., Seoul, South Korea), and body weight and composition were determined using a bioimpedance analyzer (BC-418; Tanita Corp., Tokyo, Japan) for KoCAS. In the ICAAN study, height was measured with a stadiometer (DS-103; Dong Sahn Jenix Co., Ltd.), and body weight and composition were measured using dual-energy X-ray absorptiometry (Lunar Prodigy Advance; GE Medical Systems Lunar, Madison, WI, USA). BMI was calculated as body weight in kilograms divided by squared height in meters (kg/m 2 ) and converted into percentiles and z-scores based on the age-and sex-specific BMIs of the 2007 Korean national growth charts 31 . Waist circumference was measured at the midpoint between the lower border of the ribcage and iliac crest using a nonelastic tape measure.
Laboratory tests were performed only for individuals whose parents had agreed to the blood tests and provided written informed consent in advance. The concentrations of serum total cholesterol (TC), high-density lipoprotein-cholesterol (HDL-C), triglycerides (TG), and glucose were measured using enzymatic assays and an autoanalyzer (model 7600II; Hitachi, Tokyo, Japan). Fasting serum insulin was measured using a Roche E170 analyzer (Roche Diagnostics, Mannheim, Germany). The insulin resistance index was calculated using the homeostasis model assessment of insulin resistance (HOMA-IR) 32 .
Information on physical activity was collected using a self-reported questionnaire. Adolescents who met the national physical activity guidelines (moderate-intensity activity for ≥150 min/week on ≥5 days/week, or vigorous-intensity activity for ≥60 min/week on ≥3 days/week) were assigned to the active group. Pubertal development, determined using the method of Marshall and Tanner, was divided into three stages according to genitalia for boys and breasts for girls. We defined Tanner stage 1 as prepubertal, stages 2 and 3 as pubertal, and stages 4 and 5 as postpubertal development.
Statistical analysis. All statistical tests were conducted using the SAS statistical software package (ver. 9.4; SAS Institute, Inc., Cary, NC, USA), and values are expressed as the means ± standard deviation (SD) for continuous variables and as percentages for categorical variables. Variables with nonnormal distributions (TG, insulin, HOMA-IR, 2-AAA) were log-transformed before the analysis. A paired t-test was used to comparatively assess altered metabolite levels between preadipocytes and adipocytes isolated from the same individual's adipose tissues. The differences between the normal-weight and obese groups were tested using a general linear model adjusted for age, sex, physical activity, and pubertal stage. Logistic regression analysis was used to estimate odds ratios (ORs) for the presence of overweight, obesity, and severe obesity. Univariate linear regression and multivariate linear stepwise regression analyses were performed to assess associations between obesity-related clinical markers and plasma 2-AAA levels. The statistical significance of changes from baseline to after intervention program completion was assessed using a paired t-test. Analysis of covariance (ANCOVA) was used to compare differences among the BMI z-score groups following completion of the intervention programs after adjustment for baseline values, age, sex, physical activity, and pubertal stage. The association between changes in the 2-AAA level and changes in obesity-related factors during the obesity intervention period was assessed using partial correlation coefficients (r) controlling for age, sex, baseline BMI, physical activity, and pubertal stage. P-values < 0.05 were considered to indicate statistical significance.
Chemicals and animals. Palmitate (PA), insulin, and DL-2-AAA were obtained from Sigma-Aldrich (St. Louis, MO, USA). To prepare PA-bovine serum albumin (BSA) solutions, PA was dissolved in ethanol and mixed with fatty acid-free BSA (2% w/v in water; Bovogen Biologicals, Keilor East, Australia) at 37 °C for 2 h with shaking. TG levels were determined using a TG quantification colorimetric/fluorometric kit (BioVision, Milpitas, CA, USA).
Male C57BL/6 mice were obtained from Japan Shizuoka Laboratory Animal Center (SLC). The mice were housed individually in plastic cages with bedding and allowed free access to a control chow diet and tap water. All mice were used for experiments at 5 weeks of age following an acclimation period of 1 week. The animals were