Effect of Obesity and Exercise Training on Plasma Amino Acids and Amino Metabolites in American Indian Adolescents

Abstract

Context

Amino acids (AAs) and their metabolites are altered with obesity and may be predictive of future diabetes in adults, but there are fewer studies on AAs, as well as conflicting findings on how they vary with obesity, in adolescents.

Objective

To determine whether plasma AAs vary with body composition and insulin sensitivity and are altered in response to exercise training.

Design

Cross-sectional, and an exercise intervention.

Setting

Tribal wellness center.

Participants

American Indian boys and girls, 11 to 17 years of age with obesity (Ob, n = 58) or normal weight (NW, n = 36).

Intervention

The Ob group completed 16 weeks of aerobic exercise training.

Main Outcome Measure

A panel of 42 plasma AAs.

Results

Compared with the NW group, the Ob group had lower aerobic fitness and insulin sensitivity (interactive homeostasis model assessment 2), 17 AAs that were higher, and 7 AAs that were lower. Branched-chain AAs (+10% to 16%), aromatic AAs (+15% to 32%), and glutamate were among the higher AAs; all were positively correlated with body fat and negatively correlated with insulin sensitivity. The lysine metabolite 2-aminoadipic acid (2-AAA) and the valine metabolite β-aminoisobutyric acid (BAIBA) were 47% higher and 29% lower, respectively, in the Ob group, and were positively (2-AAA) and negatively (BAIBA) correlated with insulin sensitivity. Exercise training increased aerobic fitness by 10%, but body composition, insulin sensitivity, and AAs were not significantly changed.

Conclusions

Several plasma AAs are altered in American Indian adolescents with obesity and are associated with insulin sensitivity, but they were not altered with this exercise intervention.

In addition to their essential role as the building blocks of proteins, amino acids (AAs) are involved in many metabolic processes, and they may serve as effectors of, and biomarkers for, cardiometabolic diseases. The concentrations of specific AAs in circulation, particularly branched-chain AAs (BCAAs) and the aromatic AAs, have been found to be elevated in adults who are obese and/or insulin resistant, compared with their healthier peers (1–3). In retrospective studies of adults, elevated plasma BCAAs at baseline predicted future onset of type 2 diabetes (T2D) during 7 to 12 years of follow-up (4, 5). Other AAs, and some AA metabolites, also vary with metabolic health status. Plasma glutamate was higher, and serine and glycine were lower, in people with obesity, insulin resistance, or nonalcoholic fatty liver disease (6–9). Among the AA metabolites, the lysine-derived 2-aminoadipic acid (2-AAA) has promise as a biomarker of T2D risk. Adults with T2D had elevated 2-AAA in blood samples collected up to 12 years before they developed the disease (10). Likewise, a metabolite of valine and thymine, β-aminoisobutyric acid (BAIBA), was reported to be reduced in adults with obesity, negatively correlated with insulin resistance in animals and humans, and increased in circulation at the end of an endurance exercise program completed by previously sedentary adults (11, 12).

Most of the studies that examined the role of AAs as correlates or mediators of insulin resistance and T2D have been conducted with adults. The rise in obesity and cardiometabolic diseases in childhood supports the need to examine the causes and consequences of these diseases in young people, to determine whether they differ from adults, and to develop effective prevention and treatment strategies (13, 14). There are fewer studies that examined whether circulating AAs are altered in children with obesity, or relate to insulin sensitivity, and those findings have been conflicting. For example, two studies of children reported results that were opposite of what had been previously demonstrated in adults; that is, BCAAs and the aromatic AA, phenylalanine, were lower in insulin-resistant adolescents with obesity compared with their normal weight peers (15, 16). In other reports, however, children with obesity and normal weight children did not have differing concentrations of BCAA (17) or phenylalanine and tyrosine (17, 18). In some studies children with obesity had higher BCAAs (18–20), phenylalanine (19, 20), or tyrosine (20). Among those investigations, there was no consistent relationship between specific AAs and body mass index (BMI), body fat, or insulin sensitivity (15–20). Importantly, there are no reports on novel AA metabolites that could be potential predictors of T2D, such as 2-AAA or BAIBA, in children. Determining whether AAs are altered with obesity and insulin resistance, or change in response to lifestyle interventions, could help determine the signatures of metabolic disease processes in children, which could guide efforts to improve the health of young people.

The need to prevent and treat cardiometabolic diseases is especially important for American Indians. American Indian adolescents have higher-than-average rates of obesity and the highest incidence of T2D compared with other racial and ethnic groups in the United States (14, 21). None of the prior metabolomic studies related to obesity or insulin resistance included AA values for American Indian adolescents, so it is unclear whether circulating AAs vary with metabolic health in this understudied population. Thus, we performed a targeted analysis of plasma AAs in American Indian boys and girls to test the hypothesis that BCAAs and aromatic AAs would be elevated with obesity and positively correlated with insulin sensitivity. We also determined whether other amino metabolites, including 2-AAA and BAIBA, were altered with obesity, and whether the concentrations of AAs change in response to a 16-week exercise training program. To our knowledge, the effect of exercise training on plasma AAs in adolescents with obesity has not been reported.

Materials and Methods

Design

The results presented herein are secondary analyses from a randomized trial that was designed to test the effects of financial incentives on the frequency and duration of exercise performed by American Indian adolescents who were overweight/obese (22). For the current analyses, we used a subset of that cohort who were obese and enrolled into the exercise program (Ob group, N = 58) and a group of normal weight peers (NW group, N = 36) who completed the same baseline testing but did not perform exercise training. The study was approved by the Institutional Review Boards of the Choctaw Nation of Oklahoma and the University of Oklahoma Health Sciences Center, respectively. The trial was registered at ClinicalTrials.gov (NCT01848353) on 25 April 2013.

Participants

All participants were American Indian boys and girls, 11.0 to 17.7 years old, living in the Choctaw Nation Health Service Area of southeast Oklahoma. All participants had reached sexual maturation [Tanner stage ≥2 for breasts (girls) or genitalia (boys) (23, 24)], as determined by a pediatrician. As previously described (22), the exercise promotion trial had 77 participants (43 girls, 34 boys) who were (i) overweight or obese [defined as BMI ≥85th percentile for age- and sex-specific norms (25)], (ii) 11.0 to 20.9 years old, and (iii) not engaged in sports programs, and attaining (via self-report) <30 minutes of structured moderate-to-vigorous intensity exercise on ≤3 days per week during the 3 months before enrollment. From that group, we excluded participants who were >18.0 years of age and/or had a BMI that was not within the obese range (≥95th percentile). There was no upper limit of BMI. We also excluded participants who were missing adequate samples for the current analyses or had other missing data. The NW group was defined as having a BMI between the 25th and 84th percentile for age and sex (25). There were no criteria for physical activity for the NW group. Participants from either group were excluded from enrollment when they had confirmed diabetes or other potentially confounding health conditions, including pregnancy, polycystic ovary syndrome, active cardiovascular, endocrine or liver disease, kidney or other organ dysfunction, chronic debilitating disease or impaired physical mobility, anemia, symptoms of undiagnosed illness, tobacco use (regular use within the past 6 months), or substance abuse. Use of medications to treat cardiometabolic conditions, promote weight loss, or that were known to influence glucose, lipid, or protein metabolism were also causes for exclusion. Preliminary eligibility was assessed during an initial phone call or in-person discussion with the study participants’ parents or guardians. Participants and their parents or guardians provided their informed, written consent and assent to enroll in the study in accordance with Institutional Review Board guidelines. After enrollment, a qualified medical provider conducted a medical history and examination to confirm eligibility for the study.

Exercise training program

Upon completion of baseline tests and procedures, described below, the Ob group entered into an exercise program that was 48 weeks in duration and subdivided into three consecutive 16-week phases. Each phase was designed to test how different incentive structures would affect exercise frequency and/or duration. A detailed description of the program and the main behavioral outcomes have been published elsewhere (22). For the current analysis, we determined the effects of exercise training only during phase 1 (weeks 1 to 16). Because there was no effect of the financial incentives on exercise frequency or duration during phase 1, data were analyzed as a pooled set for the current report, without regard for the type of incentive the participant received. Forty-two participants (20 girls, 22 boys) completed testing at the end of phase 1 and had plasma available for the current analyses.

The Ob group members were instructed to exercise at their local wellness center operated by Choctaw Nation in Southeast Oklahoma (22). Participants could exercise on schedules of their own choosing and could perform any type of aerobic or resistive exercise that was available at the wellness center. Participants were encouraged to complete at least three exercise sessions per week, with a minimum of 20 minutes per session of moderate-to-vigorous physical activity (MVPA). During exercise, participants wore a chest strap heart rate monitor (Spirit System; Interactive Health Technologies, Austin, TX). The monitor and accompanying software recorded the duration and intensity of each exercise session, calculated the time spent in MVPA, and provided feedback during and after exercise.

Clinical and physiological tests

The following tests and measurements were performed at baseline in all participants, and at the end of the exercise program for the Ob group (22). Trained clinical staff measured height, body mass, waist circumference, and blood pressure. BMI was calculated using body mass and height (kg/m²), and BMI percentile was calculated using growth charts from the Centers for Disease Control and Prevention (25). Total body fat, fat-free mass, and trunk fat were measured using bioelectrical impedance (Model BC-418; Tanita Corporation, Arlington Heights, IL). A bicycle ergometer test with increasing workloads was used to measure peak work output and peak rate of oxygen consumption (VO₂ peak). Continuous measurements of expired gases were performed with a facemask and metabolic measurement system (TrueOne 2400; ParvoMedics, Sandy, UT). Daily ambulatory activity was measured with accelerometers worn on the waist (Fitbit Zip; Fitbit, San Francisco, CA) for 7 days.

Blood analysis

Venous blood was collected in the morning following an overnight fast. Glycated hemoglobin (HbA1c) was measured on whole blood at the time of collection (Siemens DCA Vantage analyzer; Siemens Healthineers, Tarrytown, NY). After centrifugation, aliquots of plasma and serum were stored at −80°C until analysis. Plasma glucose was measured by the glucose oxidase method (2300STAT Plus; Yellow Springs Instruments, Yellow Springs, OH). Intra-assay and interassay coefficients of variation for glucose were 2% and 5%, respectively. Serum insulin was measured using a chemiluminescent ELISA (no. 80-INSHU-CH10; ALPCO, Salem, NH). The intra-assay and interassay coefficients of variation for insulin at 30 pmol/L were 10% and 16%, respectively, and 3% and 8%, respectively, at 450 pmol/L. Insulin sensitivity and β cell function were calculated using glucose and insulin concentrations with the revised integrated homeostatic model of assessment 2 (iHOMA2) (26). A targeted approach was used to measure the plasma concentration of 42 individual amino acids and their metabolites using ultra-performance liquid chromatography–mass spectrometry at the Mayo Clinic Metabolomics Core laboratory, as previously described (27, 28). Samples were thawed on ice and spiked with ¹³C-labeled internal standards, and 10-point standard curves were used to calculate the concentration of each AA and AA metabolite (28). A quality control sample and one of the concentration standards were injected throughout each sequence and were used to calculate the intra-assay and interassay coefficients of variation, which averaged 2% and 9%, respectively, across analytes.

Data analyses

Summary values are presented as mean ± SD or median (interquartile range). Between-group differences at baseline and within-group changes from the beginning to the end of the exercise program were evaluated with t tests and 95% CIs, or Wilcoxon rank-sum tests, as appropriate. A χ² test was used to compare the proportions of girls and boys between the NW and Ob groups. The cross-sectional comparison of AAs between the NW and Ob groups and the change in AAs within the Ob group from baseline to the end of exercise were analyzed using MetaboAnalyst software (www.metaboanalyst.ca), which includes the Benjamini–Hochberg procedure (29), to calculate false discovery rates as a correction for multiple testing. Most of the AA data, and several of the physiological characteristics, including body composition and insulin sensitivity, had skewed distributions and were, therefore, log transformed before analysis. Bivariable correlations were calculated to determine strength of association between amino metabolites, age, Tanner stage, total body fat, trunk fat, and insulin sensitivity, using Pearson or Spearman coefficients. Stepwise multivariable modeling was used to determine whether there were sets of two or more amino metabolites that explained a greater proportion of the variance in insulin sensitivity than either body fat or individual amino metabolites alone. P values less than an α of 0.05 were considered significant for all tests.

Results

Baseline comparisons between the NW and Ob groups

Age and HbA1c were not different between the NW and Ob groups, but all other measures of body size and composition, aerobic fitness, physical activity, blood pressure, glucose, insulin, and insulin sensitivity differed between groups (Table 1). None of the participants met the criteria for T2D according to guidelines from the American Diabetes Association (30) upon enrollment; fasting glucose was <7.0 mmol/L (range, 4.4 to 5.9 mmol/L) and HbA1c was <6.5% (range, 4.5% to 6.1%).

Table 1.

Physiological Characteristics of the Normal Weight and Obese Groups at Baseline

	Normal Weight	Obese	P Value (95% CI for Difference)
Girls/boys, n	16/20	28/30	0.717
Age, y	14.3 ± 1.7	13.9 ± 1.7	0.269 (−1.1, 0.3)
BMI, kg/m2	20.4 ± 1.9	34.7 ± 6.5	<0.001 (12.5, 16.5)
BMI, percentile	60 ± 18	98 ± 2	<0.001 (32, 44)
Waist circumference, cm	71.1 ± 6.8	107.7 ± 14.4	<0.001 (32.0, 41.2)
Body fat, %	22 ± 7	43 ± 7	<0.001 (18, 24)
Fat-free mass, kg	42.6 ± 8.1	53.5 ± 13.2	<0.001 (6.5, 15.4)
Fat mass, kg	12.4 ± 4.2	41.9 ± 16.4	<0.001 (24.9, 34.0)
Trunk fat, kg	5.2 ± 2.3	17.2 ± 6.8	<0.001 (10.0, 14.0)
Peak workload, W	147 ± 35	115 ± 28	<0.001 (−46, −19)
VO2 peak, mL/kg FFM/min	54.6 ± 10.2	33.2 ± 8.2	<0.001 (−25.6, −17.2)
Steps per day	9198 ± 3109	6159 ± 2908	<0.001 (−4358, −1718)
Systolic BP, mm Hg	111 ± 9	123 ± 11	<0.001 (8, 17)
Diastolic BP, mm Hg	67 ± 9	72 ± 10	0.013 (1, 9)
Glucose, mmol/L	4.9 ± 0.3	5.2 ± 0.3	<0.001 (0.1, 0.4)
Insulin, pmol/L	36 ± 15	135 ± 128	<0.001 (65, 133)
iHOMA2, %S	175 ± 84	68 ± 44	<0.001 (−138, −78)
iHOMA2, %B	75 ± 22	159 ± 91	<0.001 (59, 108)
HbA1c, % [mmol/mol]	5.4 ± 0.3 [35.3 ± 3.3]	5.3 ± 0.4 [34.6 ± 3.2]	0.325 (−0.2, 0.1) [−2.1, 0.7]

All values, except the number of participants, are shown as mean ± SD. The P value for between-group proportions of girls and boys is for a χ² test. For all other variables, P values and 95% CIs are for between-group comparisons with t tests.

Abbreviations: BP, blood pressure; FFM, fat-free mass; %B, β cell function; %S, insulin sensitivity.

Out of the 42 amino metabolites in the panel, three were undetected (anserine, carnosine, and cystathionine 2) and were excluded from further analyses. Of the remaining analytes, the concentrations of 17 were higher in the Ob group vs the NW group, 7 were lower in the Ob group, and 15 were not different between groups (Table 2). The concentrations of all three BCAAs were 10% to 16% higher in the Ob group than in the NW group (Fig. 1). The sum of the three BCAAs was positively correlated with log-transformed trunk fat and negatively correlated with log-transformed insulin sensitivity, respectively (Fig. 1). Similarly, the aromatic AAs phenylalanine and tyrosine were 15% to 32% higher in the Ob group (Fig. 2). Both phenylalanine and tyrosine were positively correlated with log-transformed trunk fat and negatively correlated with log-transformed insulin sensitivity, respectively.

Table 2.

Plasma Concentration of AAs and Metabolites at Baseline

	Normal Weight	Obese	P Value	Correlation With: Body Fat iHOMA2 (%S)
Higher in Ob group
Glutamate	14.6 (12.0, 20.7)	85.9 (63.1, 126.5)	4.96 × 10−11	0.742	−0.610
Phosphoethanolamine	0.36 (0.29, 0.45)	1.12 (0.86, 1.76)	3.3 × 10−10	0.616	−0.570
Aspartate	0.99 (0.81, 1.53)	4.42 (3.06, 5.90)	1.85 × 10−9	0.556	−0.568
Cystathionine 1	0.00 (0.00, 0.10)	0.17 (0.14, 0.25)	7.04 × 10−8	0.469	−0.492
Tyrosine	58.1 (54.0, 65.1)	78.8 (66.5, 88.0)	9.07 × 10−8	0.562	−0.590
Alloisoleucine	1.17 (0.01, 1.64)	1.82 (1.47, 2.16)	8.29 × 10−6	0.401	−0.341
Phenylalanine	51.0 (48.1, 56.5)	59.0 (54.4, 63.6)	1.35 × 10−5	0.493	−0.467
Leucine	106.3 (94.7, 114.1)	122.0 (107.7, 131.6)	1.86 × 10−4	0.409	−0.360
Alanine	323.0 (279.1, 354.9)	371.7 (330.8, 422.0)	2.61 × 10−4	0.309	−0.382
Valine	215.7 (183.5, 235.1)	239.8 (214.0, 260.1)	4.04 × 10−4	0.385	−0.286
β-Alanine	1.87 (1.15, 2.81)	2.73 (2.20, 3.26)	9.55 × 10−4	0.300	−0.225
Ornithine	40.3 (36.2, 47.7)	46.9 (41.0, 52.5)	0.0035	0.397	−0.302
2-AAA	0.45 (0.33, 0.59)	0.60 (0.44, 0.87)	0.0040	0.276	−0.443
Proline	170.3 (129.1, 200.6)	187.3 (165.0, 225.8)	0.0045	0.430	−0.340
Histidine	99.4 (88.8, 113.8)	110.5 (97.7, 139.3)	0.0052	0.355	−0.218
Lysine	158.2 (142.3, 180.2)	179.8 (160.4, 196.2)	0.0067	0.400	−0.334
Isoleucine	58.9 (54.6, 64.0)	65.1 (57.0, 73.8)	0.0195	0.260	−0.283
Not different between groups
Tryptophan	53.2 (47.9, 56.9)	55.2 (49.9, 62.0)	0.0526	0.109	−0.201
Arginine	69.1 (62, 75.7)	78.4 (67.5, 87.4)	0.0526	0.125	−0.190
Methionine	21.1 (18.8, 22.9)	22.1 (20.0, 24.6)	0.1177	0.179	−0.183
Taurine	39.6 (31.0, 44.2)	42.4 (34.6, 50.9)	0.1263	0.042	−0.140
2-Amino-N-butyric acid	15.7 (13.1, 17)	16.5 (13.3, 22.8)	0.1456	0.178	−0.027
3-Methylhistidine	3.23 (2.61, 4.01)	3.76 (2.62, 4.50)	0.3002	0.062	−0.036
Hydroxylysine 1	0.03 (0.02, 0.05)	0.03 (0.02, 0.05)	0.4140	0.022	−0.188
1-Methylhistidine	3.30 (1.09, 6.23)	2.73 (1.65, 9.03)	0.4943	0.060	−0.018
Sarcosine	15.8 (5.4, 24.9)	15.7 (13.5, 18.3)	0.6881	0.030	−0.044
Serine	82.8 (78.2, 104.6)	89.2 (78.5, 101.1)	0.8277	0.104	−0.037
Ethanolamine	6.96 (5.94, 7.54)	6.69 (6.05, 7.46)	0.9256	−0.043	0.007
Hydroxylysine 2	0.86 (0.54, 1.10)	0.75 (0.58, 1.09)	0.8277	−0.189	−0.131
Threonine	130.0 (108.3, 151.5)	123.9 (106.3, 138.9)	0.4140	0.007	0.046
Hydroxyproline	24.3 (17.9, 26.8)	22.0 (17.0, 24.9)	0.3893	−0.244	−0.006
Citrulline	33.3 (26.1, 40.5)	30.3 (26.8, 35.3)	0.1510	−0.286	0.273
Lower in Ob group
Glycine	257.3 (208.5, 303.3)	215.5 (189.3, 239.9)	0.0063	−0.290	0.297
Glutamine	524.2 (494.8, 578.9)	487.7 (395.3, 559.7)	0.0053	−0.335	0.322
BAIBA	0.91 (0.63, 1.22)	0.64 (0.41, 0.90)	0.0040	−0.260	0.348
Cysteine	18.7 (13.6, 22.4)	2.0 (0.8, 15.9)	2.08 × 10−4	−0.302	0.307
Asparagine	42.5 (40.6, 51.9)	36.2 (30.3, 41.3)	9.38 × 10−5	−0.388	0.467
Homocysteine	0.59 (0.42, 0.74)	0.33 (0.27, 0.41)	2.41 × 10−6	−0.490	0.278
GABA	0.11 (0.09, 0.13)	0.08 (0.05, 0.09)	7.31 × 10−7	−0.543	0.352

All concentrations are µmol/L. Values are shown as median (interquartile range) for 36 normal weight participants and 58 participants with obesity, respectively. Between-group comparisons were made with a Wilcoxon rank-sum tests. False discovery rate–adjusted P values are shown, with significant values in bold. Correlation coefficients are listed for total body fat and insulin sensitivity [iHOMA2 (%S)], performed on log-transformed values. Significant correlations are in bold.

Figure 1.

BCAAs in the NW and Ob groups. (A–C) Group values for valine, leucine, and isoleucine, respectively. Boxes show the median (center line) and interquartile range (IQR) for each group. Whiskers show the range of values within 1.5-fold of the IQR. Filled circles show the mean, and open circles are outlier values outside of 1.5-fold of the IQR. Between-group comparisons are shown as the percentage difference and false discovery rate–adjusted P value for a Wilcoxon rank-sum test. (D and E) Correlations between trunk fat and insulin sensitivity [iHOMA2 (%S)], respectively, and the sum of the three BCAAs. Triangles indicate normal weight; circles indicate obesity; filled symbols represent girls; open symbols represent boys. All results are for 36 normal weight participants and 58 participants with obesity at baseline, before the Ob group started the exercise program.

Figure 2.

Phenylalanine (Phe) and tyrosine (Tyr) in NW and Ob groups. (A and B) Group values for Phe and Tyr, respectively. Boxes show the median (center line) and interquartile range (IQR) for each group. Whiskers show the range of values within 1.5-fold of the IQR. Filled circles show the mean, and open circles are outlier values outside of 1.5-fold of the IQR. Between-group comparisons are shown as percentage difference and false discovery rate–adjusted P value for a Wilcoxon rank-sum test. (C–F) Correlations between trunk fat and insulin sensitivity [iHOMA2 (%S)], respectively, and Phe and Tyr. Triangles indicate normal weight; circles indicate obesity; filled symbols represent girls; open symbols represent boys. All results are for 36 normal weight participants and 58 participants with obesity at baseline, before the Ob group started the exercise program.

The amino metabolite with the largest between-group difference was glutamate, which was threefold higher in the Ob group than in the NW group (Fig. 3). In contrast, glutamine and γ-amino-N-butyric acid (GABA), which are both formed from glutamate, were lower in the Ob group (Fig. 3). Glutamate was the amino metabolite with the strongest bivariable correlation with trunk fat and insulin sensitivity (Fig. 3).

Figure 3.

Glutamate, glutamine, and GABA in NW and Ob groups. (A–C) Group values for glutamate, glutamine, and GABA, respectively. Boxes show the median (center line) and interquartile range (IQR) for each group. Whiskers show the range of values within 1.5-fold of the IQR. Closed circles show the mean, and open circles are outlier values outside of 1.5-fold of the IQR. Between-group comparisons are shown as percentage difference and false discovery rate–adjusted P value for Wilcoxon a rank-sum test. (D and E) Correlations between trunk fat and insulin sensitivity [iHOMA2 (%S)], respectively, and glutamate. Triangles indicate normal weight; circles indicate obesity; filled symbols represent girls; open symbols represent boys. All results are for 36 normal weight participants and 58 participants with obesity at baseline, before the Ob group started the exercise program.

Lysine (Table 2) and its metabolite, 2-AAA (Fig. 4), were higher in the Ob group. 2-AAA was negatively correlated with insulin sensitivity (Fig. 4) and positively correlated with log-transformed trunk fat (r = 0.28, P < 0.01). The valine metabolite, BAIBA, was lower in the Ob group than in the NW group (Fig. 4). BAIBA was positively correlated with insulin sensitivity (Fig. 4) and negatively correlated with log-transformed trunk fat (r = −0.23, P = 0.029).

Figure 4.

Amino acid metabolites, 2-AAA, and BAIBA in NW and Ob groups. (A) Group values for 2-AAA. (B) Correlation between insulin sensitivity and 2-AAA. (C) Group values for BAIBA. (D) Correlation between insulin sensitivity and BAIBA. For group comparisons in (A) and (C), boxes show the median (center line) and interquartile range (IQR) for each group. Whiskers show the range of values within 1.5-fold of the IQR. Closed circles show the mean, and open circles are outlier values outside of 1.5-fold of the IQR. Between-group comparisons are shown as percentage difference and false discovery rate–adjusted P value for a Wilcoxon rank-sum test. For scatterplots in (B) and (D), triangles indicate normal weight, circles indicate obesity, filled symbols represent girls, and open symbols represent boys. All results are for 36 normal weight participants and 58 participants with obesity at baseline, before the Ob group started the exercise program.

We determined which set of AAs was most closely associated with insulin sensitivity using bivariable correlations. Age and Tanner stage were not significantly correlated with any AA. The only AAs that differed between sexes were hydroxyproline and ethanolamine (28% and 9% higher in girls, respectively). For multivariable modeling, we initially included all AAs and measures of adiposity (BMI, waist circumference, total body fat, and trunk fat). The best stepwise model for explaining the variance in insulin sensitivity had three variables: total body fat, 2-AAA, and BAIBA (R² = 0.595, adjusted R² = 0.581). Total body fat was the first variable entered (R² = 0.502, adjusted R² = 0.496), followed by 2-AAA and BAIBA. When total body fat was removed from the list of input variables, trunk fat replaced it as the first variable into the model (R² = 0.435, adjusted R² = 0.428), followed by tyrosine, asparagine, 2-AAA, and valine (final R² = 0.606, adjusted R² = 0.582). When only amino metabolites were made available for the model (no body composition variables), the first variable to enter was glutamate (R² = 0.370, adjusted R² = 0.364), followed by tyrosine, citrulline, asparagine, and cysteine (final R² = 0.591, adjusted R² = 0.567).

The Ob group was comprised of 11 participants with class 1 obesity (≥95th percentile for age and sex up to 119% of the BMI at the 95th percentile), 28 with class 2 obesity (BMI corresponding to 120% to 139% of the 95th percentile, or 35.0 kg/m², whichever is less), and 19 with class 3 obesity (BMI ≥140% of the 95th percentile, or 40.0 kg/m², whichever is less) (31). We explored whether there were differences in the physiological or biochemical variables among those subgroups. The class 1 obesity group was younger (12.9 ± 1.2 years) than the class 2 (14.0 ± 1.8 years, P = 0.31) and class 3 (14.3 ± 1.9 years, P = 0.018) subgroups, but Tanner stage did not differ among them. Fat mass was progressively greater for classes 1, 2, and 3, respectively (23.7 ± 7, 38.3 ± 8.7, and 57.3 ± 15.3 kg, P < 0.01 for all between-group comparisons). HbA1c (5.3 ± 0.2, 5.3 ± 0.2, and 5.4 ± 0.4 for classes 1, 2, and 3, respectively), and fasting glucose (5.1 ± 0.4, 5.1 ± 0.3, and 5.3 ± 0.3 mmol/L for classes 1, 2, and 3, respectively) did not differ among obesity classes, but fasting insulin was lower in the class 1 subgroup (68 ± 24 pmol/L) than either the class 2 (142 ± 153 pmol/L, P = 0.030 vs class 1) or class 3 (165 ± 112 pmol/L, P < 0.001 vs class 1) subgroups. Insulin sensitivity was higher in the class 1 subgroup (92 ± 41% insulin sensitivity) than in either the class 2 (72 ± 48% insulin sensitivity, P = 0.032 vs class 1) or class 3 (48 ± 28% insulin sensitivity, P < 0.001 vs class 1) subgroups. Despite the differences in body fat, insulin concentration, and insulin sensitivity, none of the AA concentrations differed significantly among obesity classes.

Changes in response to the exercise program

Participants in the exercise program completed an average of 32 out of 48 planned exercise sessions during 16 weeks (range, 7 to 71 sessions), with an average duration of 34 ± 8 minutes of MVPA per session. The cumulative exercise time was 18.3 ± 8.6 hours (range, 11.4 to 22.6 hours). After 16 weeks, absolute BMI (but not BMI percentile) and total body fat increased from baseline, but trunk fat was not significantly changed (Table 3). There was a small but significant improvement in exercise capacity, shown by increased peak power and oxygen consumption during the bicycle test. However, fat-free mass, physical activity, blood pressure, glucose, insulin, insulin sensitivity, and HbA1c were not significantly changed (Table 3). None of the AAs shown in Figs. 1–4 was significantly changed in response to exercise (Fig. 5). Likewise, none of the other AAs was significantly altered following the training program; all of the false discovery rate–adjusted P values for the change from baseline to the end of phase 1 exercise training were >0.9.

Table 3.

Physiological Characteristics of the Exercise Intervention Group Before and After Their Participation in the 16-wk Exercise Promotion Program

	Baseline Value	Change	P Value
BMI, kg/m2	35.3 ± 6.7	0.8 ± 1.6	0.002
BMI, percentile	98 ± 3	0 ± 1	0.070
Body fat, %	44 ± 8	1 ± 3	0.032
Fat-free mass, kg	51.1 ± 10.3	0.5 ± 3.3	0.311
Fat mass, kg	42.0 ± 17.5	0.1 ± 11.7	0.002
Trunk fat, kg	16.8 ± 6.5	0.6 ± 1.8	0.066
Peak workload, W	114 ± 30	6 ± 19	0.043
VO2 peak, mL/kg FFM/min	33.2 ± 7.1	3.1 ± 6.3	0.009
Steps per day	6380 ± 3323	−478 ± 2448	0.224
Systolic BP, mm Hg	124 ± 11	−2 ± 12	0.290
Diastolic BP, mm Hg	72 ± 10	−1 ± 13	0.596
Glucose, mmol/L	5.2 ± 0.4	0.2 ± 0.8	0.152
Insulin, pmol/L	145 ± 142	−11 ± 143	0.623
iHOMA2, %S	66 ± 44	−9 ± 33	0.099
iHOMA2, %B	164 ± 95	−13 ± 87	0.355
HbA1c, % (mmol/mol)	5.3 ± 0.2 (34.2 ± 2.7)	0.1 ± 0.2 (0.7 ± 2.5)	0.090

Values are shown as mean ± SD for 42 participants. P values are shown for within-group comparisons (paired t tests), with significant changes from baseline to follow-up shown in bold.

Abbreviations: BP, blood pressure; FFM, fat-free mass; %B, β cell function; %S, insulin sensitivity.

Figure 5.

No change in AAs or amino metabolites in response to exercise training performed by the Ob group. (A–J) For each of the selected AAs/metabolites, the figure on the left shows the preexercise and postexercise values, and the right panel shows the change in values from preexercise to postexercise. Boxes show the median (center line) and interquartile range (IQR). Whiskers show the range of values within 1.5-fold of the IQR. Closed circles show the mean, and open circles are outlier values outside of 1.5-fold of the IQR. There were no statistically significant changes from preexercise to postexercise. All results are for 42 participants.

Because there was a broad range of exercise sessions performed, we explored whether any of the outcome variables was related to the frequency or volume of exercise completed. Bivariable correlations revealed that exercise frequency and total hours of exercise performed were positively correlated with the change in peak workload (r = 0.36, P < 0.02) and VO₂ peak (r = 0.41, P < 0.01) measured during the aerobic fitness test. However, exercise behavior was not significantly correlated with changes in insulin sensitivity or plasma AAs. The overall lack of change in AAs was not attributable to different responses between boys and girls. The number of exercise sessions (girls, 29 ± 11; boys, 34 ± 16; P = 0.220) and duration of exercise sessions (girls, 36 ± 8 minutes of MVPA; boys, 33 ± 7; P = 0.315) did not differ between sexes. Likewise, there were no differences between girls and boys for the change in concentration of any AAs from baseline to the end of the exercise program.

We examined whether changes in AA concentrations or insulin sensitivity may be related to timing of the blood collection relative to the last exercise session. Postexercise blood samples in the whole group were collected at a median (interquartile range) of 8 (4, 18) days from the last exercise session. The wide range in the timing of the postexercise sample is attributable to participants who stopped exercising at the wellness center before 16 weeks but returned for follow-up testing as scheduled. Within the group of exercisers, timing of blood collection was not significantly correlated with changes in AA concentrations (r_s = −0.29 to 0.20, P > 0.05) or insulin sensitivity (r_s = 0.13, P > 0.05). A subgroup of 20 participants completed at least 40 exercise sessions and had their postexercise blood collection between 2 and 8 days after their last exercise session. Within this subgroup with the highest compliance, none of the AAs changed from preexercise to postexercise training, nor were there associations between timing of blood collection and changes in AA concentrations (r_s = −0.32 to 0.25, P > 0.05) or insulin sensitivity (r_s = 0.13, P > 0.05).

Discussion

We found that the plasma concentration of several plasma AAs and their metabolites are altered in obese American Indian adolescents and are associated with insulin sensitivity, but are not altered with this exercise intervention. Despite having fasting glucose and HbA1c within the normal range, the Ob group had several known risk factors for future development of T2D, including elevated body fat and lower insulin sensitivity, physical activity, and aerobic fitness compared with the NW group. The profile of elevated BCAAs, aromatic AAs, glutamine, and 2-AAA, and lower BIABA, among other AA differences, fit with the growing body of literature demonstrating that alterations in several AAs signal a metabolic disruption in people with insulin resistance and potential future onset of T2D (4).

Plasma BCAA and aromatic AA concentrations were higher in the Ob group than in the NW group, and they were positively correlated with adiposity and negatively correlated with insulin sensitivity. Although these changes are primarily understood to be the consequence, rather than a cause, of obesity and insulin resistance, the elevation of BCAAs and aromatic AAs may exacerbate hyperinsulinemia by acting as insulin secretagogues (32). Our interest in measuring BCAA and aromatic AA was prompted by reports that they are predictive of incident diabetes in adults (4, 5). Tyrosine was also reported to be positively correlated with insulin resistance in children with obesity, before and after a weight loss program (33). There are conflicting findings on whether BCAAs and aromatic AAs are altered in children with obesity, and no prior studies on this subject included American Indian adolescents, a population with high rates of obesity and T2D (14, 21). In pediatric studies, BCAAs were found to be either higher (18–20), lower (15, 16), or not altered (17) in the presence of obesity. Another investigation that included only children who were obese found that BCAAs were higher in boys than girls, but they were unrelated to insulin resistance in either sex (34). Similarly, phenylalanine was either higher (19, 20) or lower (15, 16), and tyrosine was either higher (20) or not different (15, 17) in groups of children with obesity and insulin resistance compared with children who were normal weight and insulin sensitive. Some of those studies included a wide age range, with a mix of prepubertal and pubertal children, which may contribute to inconsistent findings among studies (17, 19, 20, 33). We chose to include only adolescents who had reached Tanner developmental stage ≥2 to avoid the potential confounding effect of the increased insulin resistance that occurs at the onset of puberty (35).

Glutamate concentration was higher and glutamine and GABA were lower in the Ob group. Elevated glutamate may be another predictive measure of future development of cardiometabolic disorders, as it was elevated prior to the development of T2D and coronary artery disease in adults (6) and was associated with insulin resistance and liver disease (7). There are only a few studies that have reported whether circulating glutamate, glutamine, and/or GABA are altered in groups with obesity vs normal weight groups of children or adolescents. Previously, glutamate was reported as increased with obesity (19) or positively correlated with BMI and insulin resistance (18), whereas glutamine was lower in children with obesity (17). However, neither glutamine or GABA was significantly correlated with BMI or insulin resistance in the study by McCormack et al. (18). Recent evidence suggests that altered populations of gut microbiota that metabolize glutamate and glutamine may contribute to the changes in the plasma concentrations of these AAs (36, 37). In a mouse model of diet-induced obesity, changes in intestinal glutamate metabolism were linked to reduced plasma GABA, reduced insulin sensitivity in the brain, and behavioral changes consistent with anxiety or depression (38). We did not assess mood or cognitive function, so we cannot determine whether the changes in glutamine and GABA are related to differences in brain function.

A novel finding was that 2-AAA was higher and BAIBA was lower in the Ob group, and those metabolites were negatively and positively correlated with insulin sensitivity, respectively. To our knowledge, values for 2-AAA and BAIBA have not previously been reported in children or adolescents. The importance of 2-AAA and BAIBA in the current study was supported by their inclusion, with total body fat, in the multivariable models to explain insulin sensitivity. In adults, 2-AAA was significantly correlated with BMI, and insulin resistance and was predictive of incident diabetes (10, 39). Similar to other AAs, 2-AAA appears to be an insulin secretagogue, which may play a role in maintaining normal glycemia in the face of insulin resistance prior to the onset of diabetes (10). Importantly, the concentration of 2-AAA is modifiable through changes in insulin action, as shown by the reduction in 2-AAA following 3 months of insulin sensitizer therapy in insulin-resistant adults with obesity (27). Because insulin inhibits protein breakdown in multiple organs (40), the increased circulating 2-AAA in people with obesity may be related to impaired insulin regulation of protein turnover. BAIBA was of interest because it was shown in animals to be reduced with obesity, but increased following exercise training (11). That study provided preliminary evidence that BAIBA is produced by skeletal muscle in response to exercise, can be secreted into the circulation, and acts on the liver and adipose to improve the metabolic profile (11). There are few data on the metabolic role of BAIBA in humans, and no previous results in children. In adults, BAIBA was inversely correlated with adiposity and insulin resistance (11, 12) and increased 17% following a 20-week aerobic training program (11).

We expected that exercise training would partially correct some of the AA concentrations that were altered in the Ob group, even in the absence of a change in body mass. Although aerobic fitness increased in response to training, the lack of significant changes in AAs and insulin sensitivity demonstrated that the volume and/or type of exercise performed was insufficient for improvement in those metabolic pathways in this cohort of adolescents. Participants were given the flexibility to exercise on their own schedule, but compliance with the exercise target was lower than expected, with an average of only two sessions per week. Barriers to compliance included lack of transportation and competing time demands (22). In previous studies with children who were overweight/obese, insulin sensitivity was increased following 12 to 13 weeks of aerobic training, without a change in body mass (41–43). An exercise-induced improvement in insulin sensitivity could be expected to result in changes in plasma AAs in previously untrained people with obesity, reflecting the effects of insulin action on AA metabolism in skeletal muscle, liver, or other organs. However, the effect of an exercise-only program on circulating AAs in children or adolescents has, to our knowledge, not previously been reported. Thus, it is not yet clear how much exercise is required to shift the AA profile of obese adolescents, and whether weight loss or other physiological changes are required. In two prior studies of children with obesity, a lifestyle intervention program that included moderate-intensity exercise and nutritional and behavioral components resulted in either small or no changes in plasma AAs (44, 45). The first of those studies had a 3-month intervention and the main outcomes were reductions in whole-body rates of leucine appearance and nonoxidative disposal (measures of protein breakdown and synthesis, respectively); plasma leucine concentration and body mass were unchanged (44). In comparison, there is some evidence that AAs may change in obese children in response to a 12-month lifestyle intervention (45). In a subset of children from that study who lost an average of 6% fat mass, there were small increases in plasma glutamine and methionine, although no other AA concentrations were altered. Thus, circulating AAs that are altered with obesity in children or adolescents may be slow to change in response to exercise or weight loss. That possibility is supported by a recent investigation of adults who deliberately increased and then decreased their body mass (46). In that study, phenylalanine and BCAAs were among a group of metabolites that increased during weight gain but did not return to baseline during weight loss, demonstrating persistent effects of weight gain. Likewise, two metabolomic studies of adults with obesity reported that 6 months of aerobic exercise resulted in no changes in plasma AAs (47, 48), even with a 5% reduction in body mass in one of the groups (48). Collectively, the available data from children and adults suggest that large, sustained changes in exercise behavior, body composition, and/or insulin sensitivity may be required to normalize the alterations in plasma AAs that accompany obesity and insulin resistance.

A limitation of the investigation was that diet was not controlled prior to the study visits and dietary history was not assessed. Differences among participants in diet, particularly protein intake, could have added variability to the AA outcomes or influenced the results in ways that were not measured. Another potential limitation is that insulin sensitivity was measured only in the fasting state (quantified using iHOMA2). We initially planned to have each participant complete an oral glucose tolerance test, which could have provided additional insights about insulin action on glucose and AAs in the postprandial period. However, that procedure was removed at the outset of participant enrollment to reduce the time demand on participants and clinical staff, as previously described (22). The decision to modify the protocol was part of an ongoing collaborative partnership between the university-based investigators and the clinical, research, and administrative leaders of the Choctaw Nation of Oklahoma (49). Most of the prior reports of AA profiles in children with obesity have also used fasting blood samples to measure insulin resistance (18–20, 33, 34), although there are some that used either an oral glucose tolerance test (33) or a hyperinsulinemic-euglycemic clamp (15, 16) to acquire a dynamic measure of insulin action. The latter approaches could provide additional insights into the effect of insulin action on AA metabolism and AA concentrations and should be used when feasible. In the multivariable models, we found that several AAs made significant contributions to explaining variance in insulin sensitivity after accounting for body fat. There are other circulating compounds that have also been shown to vary with, and may contribute to, the variance in insulin sensitivity in children, such as adiponectin (50), that were not measured in this study. Finally, our study design does not allow for determination of the cause for the differences in AAs between the NW and Ob groups. The concentration of each AA in circulation is the cumulative result of uptake, release, and metabolism of AAs in many organs and tissues. Thus, even though the concentrations of several AAs were correlated with body fat, it is unclear how much of the variation in any given AA can be attributed to AA metabolism within adipose depots, or the indirect actions of adipose-derived compounds such as leptin, adiponectin, or inflammatory molecules on other tissues.

Conclusion

Several plasma AAs are altered in American Indian adolescents with obesity and are correlated with insulin sensitivity, but they are not altered with this exercise intervention. Alternative approaches to modifying exercise and diet are needed to produce an effective, sustainable reduction of risk for cardiometabolic diseases in adolescents with obesity.

Acknowledgments

The authors extend their appreciation to the many people of Choctaw Nation who helped with the development and implementation of this project. We thank J. Neil Henderson for his role as Principal Investigator and Director of the American Indian Diabetes Prevention Center. We are also grateful for the excellent technical assistance from the staff at the Mayo Clinic Metabolomics Resource Core.

Financial Support: This work was primarily supported by Grant P20 MD000528 from the National Center on Minority Health and Health Disparities. The funding agency had no input into the study design, data collection, data analysis, data interpretation, or preparation of reports and manuscripts. This publication was also made possible by the Mayo Clinic Metabolomics Resource Core through Grant U24DK100469 from the National Institute of Diabetes and Digestive and Kidney Diseases and originates from the National Institutes of Health Director’s Common Fund. Additional support for equipment, personnel, and other resources was provided by Choctaw Nation of Oklahoma, the Presbyterian Health Foundation, the Children’s Health Foundation, and the Children’s Hospital Foundation Metabolic Research Program.

Clinical Trial Information: ClinicalTrials.gov no. NCT01848353 (registered 25 April 2013).

Disclosure Summary: The authors have nothing to disclose.

Abbreviations:

Abbreviations:
 
• 2-AAA

  2-aminoadipic acid

 
• AA

  amino acid

 
• BAIBA

  β-aminoisobutyric acid

 
• BCAA

  branched-chain AA

 
• BMI

  body mass index

 
• GABA

  γ-amino-N-butyric acid

 
• HbA1c

  glycated hemoglobin

 
• iHOMA2

  interactive homeostasis model assessment 2

 
• MVPA

  moderate-to-vigorous physical activity

 
• NW group

  normal weight group

 
• Ob group

  group with obesity

 
• T2D

  type 2 diabetes

 
• VO₂ peak

  peak rate of oxygen consumption