Metabolome alterations in pediatric metabolically unhealthy obesity are primarily linked to abnormal glucose homeostasis

Abstract

The mechanisms underlying metabolically healthy obesity (MHO) in pediatrics remain poorly understood. This study aims to evaluate the plasma metabolome of MHO versus metabolically unhealthy obesity (MUO) in Asian children and adolescents, to identify key metabolic drivers that undergird the MUO phenotype. MHO and MUO were defined by the absence or presence of metabolic syndrome criteria, respectively. We conducted untargeted metabolomics analysis on plasma samples from children and adolescents without obesity (n = 24), with MHO (n = 65) and with MUO (n = 222). Multivariate data analyses identified key metabolites differentiating the groups. Logistic regression assessed metabolite associations with metabolic conditions, while Spearman’s correlation evaluated their links to cardiometabolic parameters. Metabolites such as plasma fatty acids, amino acids and 1,5-anhydroglucitol differentiated MHO from MUO, correlating significantly with parameters of glucose homeostasis. Plasma branched-chain amino acids and 3-hydroxyisobutyric acid were elevated while 1,5-anhydroglucitol was reduced in pediatrics with obesity and abnormal glucose tolerance compared to those with obesity and normal glucose tolerance. Our study revealed distinct metabolome alterations between MHO and MUO in Asian children and adolescents. Notably, we identified that these metabolomic differences between MHO and MUO are primarily linked to abnormal glucose homeostasis, highlighting potential metabolic targets for improving health outcomes in pediatric obesity.

Introduction

The global prevalence of childhood obesity has increased significantly in the past four decades from less than 1% in 1975 to 6–8% in 2016 ¹. This worrying trend is also observed in Singapore where the rate of overweight and obesity in children has increased from 13% in 2017 to 16% in 2021 ². Individuals with obesity often develop metabolic syndrome (MetS) which is a composite of pathologies including hypertension, dyslipidaemia (abnormal high density lipoprotein (HDL)), hypertriglyceridemia and insulin resistance (IR) ³. The presence of MetS in individuals with obesity is termed as metabolically unhealthy obesity (MUO). Conversely, a subgroup of individuals with obesity who do not have MetS are classified as having metabolically healthy obesity (MHO) ⁴. Presently, there is no universally standardized definition of MHO ⁵. In Singapore, the prevalence of MHO among children and adolescents with obesity ranges from 22.4 to 63.5% depending on the stringency of MHO definition used ⁶.

The mechanisms underlying the MHO phenotype are unclear. Individuals with MHO were found to be younger ⁷ and had lower visceral fat accumulation and ectopic fat storage in the liver and skeletal muscle than individuals with MUO ⁷. Furthermore, individuals with MHO displayed a more favorable inflammatory profile compared to individuals with MUO ⁸. Prospective studies posit that the MHO phenotype is a transient and dynamic state, with 30–50% of MHO progressing to MUO phenotype within 5–10 years ^9,10.

Human metabolomic studies performed in well-characterized clinical cohorts are crucial to identify underlying biochemical perturbations in the development of MUO phenotype. In recent years, a large proportion of studies investigating the underlying biochemical pathways altered between MHO and MUO have been conducted mainly in adults ¹¹, with amino acid (AA) and fatty acid (FA) metabolism found to be most significantly altered ¹¹. An extensive metabolomic profiling revealed that Norwegian adults with MHO had lower plasma levels of branched-chain amino acids (BCAA), and a more favorable FA profile composed of higher concentrations of monounsaturated fatty acids (MUFA) and lower concentrations of long chain saturated FA compared to their MUO counterparts ¹². This is consistent with another untargeted metabolomic analysis conducted by Badoud et al. that found significant perturbations in BCAA and aromatic AA metabolism between adults without obesity, adults with MHO and adults with MUO ¹³.

While several metabolomic studies conducted among children and adolescents with obesity have primarily focused on IR and type 2 diabetes mellitus (T2DM) as key outcomes ^{14, 15–16}, the broader MHO and MUO phenotypes remain underexplored. Among the few studies that have examined the MHO and MUO phenotype in children, lower levels of AA such as lysine, histidine and glutamine have been observed in Caucasian children and Korean adolescents with MUO ^17,18. In contrast, higher levels of BCAA (leucine, isoleucine and valine) as well as asparagine were observed in Caucasian and Chinese adolescents with MUO ^19,20. Additionally, altered lipid metabolism, characterized by increased plasma acylcarnitines, glycerophospholipids ¹⁸ and decreased free FA (palmitic and stearic acid) ²⁰ have been reported in children with MUO.

Given the well-documented population differences in metabolic profiles and disease susceptibility ²¹, there is a critical need for more studies focusing on children and Asian populations. Furthermore, as the MUO phenotype is highly heterogeneous, comprising a composite of individual MetS criteria, it is essential to not only examine the broader MHO/MUO spectrum but also investigate the unique biochemical pathways associated with each metabolic abnormality. Such studies could elucidate the underlying metabolic drivers of the MUO phenotype and provide a basis for tailored early interventions in children to prevent the progression from MHO to MUO, ultimately reducing the long-term health burden of obesity.

The objective of our study is to conduct global untargeted metabolomic profiling to uncover underlying plasma metabolome perturbations in a multiethnic Asian cohort of children and adolescents without obesity, with MHO and with MUO. Furthermore, we aim to identify key plasma metabolite markers associated with the various components of metabolic health in obesity.

Results

Clinical metabolic parameters of children and adolescents without obesity, children and adolescents with MHO, and children and adolescents with MUO

Children and adolescents with obesity had significantly higher adiposity measures (BMI z-score and body fat percentage), fasting insulin, fasting glucose, homeostatic model assessment for insulin resistance (HOMA-IR), systolic blood pressure (SBP), triglycerides (TG) and significantly lower HDL cholesterol than children and adolescents without obesity (Supplementary Table 1).

Apart from significant differences in adiposity measures, children and adolescents with MHO exhibited significantly higher fasting insulin (19.6 mU/L (11.9–26.0) vs 5.80 mU/L (5.20–9.70)) and HOMA-IR (4.08 (2.58–5.59) vs 1.19 (1.00–2.02)) compared to children and adolescents without obesity. Conversely, children and adolescents with MUO had significantly higher fasting insulin (25.7 mU/L (17.4–36.0) vs 5.80 mU/L (5.20–9.70)), HOMA-IR (5.65 (3.58–7.67) vs 1.19 (1.00–2.02)), SBP (124 mmHg (116–131) vs 111 mmHg (103–119)), TG (1.29 mmol/L (0.99–1.90) vs 0.85 mmol/L (0.57–1.22)) and lower HDL cholesterol (1.05 mmol/L (0.95–1.19) vs 1.39 mmol/L (1.16–1.61)) than children and adolescents without obesity (Table 1).

Table 1. Clinical characteristics of children and adolescents without obesity, children and adolescents with MHO, and children and adolescents with MUO.

	Without obesity (n = 24)	MHO (n = 65)	MUO (n = 222)	Without obesity vs MHO	Without obesity vs MUO	MHO vs MUO
				p
Sex (male) (%)	62.5	69.2	66.7	0.548	0.682	0.698
Race (C/M/I) (%)	41.7/54.1/4.20	46.2/44.6/9.20	52.3/41.4/6.30	0.610	0.482	0.576
Age (years)#	14.7 (11.3–16.7)	13.1 (10.1–15.4)	15.1 (13.2–16.5)	0.060	1.000	< 0.001*
Waist to hip ratio	–	0.96 (0.91–1.01)	0.96 (0.90–1.00)	–	–	0.471
Body fat %#	21.1 (15.9–26.8)	47.3 (40.0–59.4)	47.6 (39.5–56.3)	< 0.001*	< 0.001*	1.000
BMI z-score#	0.29 (− 0.24–0.62)	2.40 (2.11–2.68)	2.40 (2.11–2.64)	< 0.001*	< 0.001*	1.000
Fasting insulin (mU/L)#	5.80 (5.20–9.70)	19.6 (11.9–26.0)	25.7 (17.4–36.0)	< 0.001*	< 0.001*	< 0.001*
Fasting glucose (mmol/L)	4.70 (4.50–4.80)	4.70 (4.50–4.85)	4.70 (4.53–5.10)	0.453	0.058	0.092
2-h OGTT (mmol/L)	–	5.80 (5.30–6.40)	6.30 (5.30–7.40)	–	–	0.010*
C-peptide (pmol/L)	–	797 (672–800)	1047 (800–1279)	–	–	< 0.001*
HOMA-IR#	1.19 (1.00–2.02)	4.08 (2.58–5.59)	5.65 (3.58–7.67)	< 0.001*	< 0.001*	< 0.001*
SBP (mmHg)#	111 (103–119)	111 (104–118)	124 (116–131)	1.000	< 0.001*	< 0.001*
DBP (mmHg)#	62 (56–68)	61 (57–67)	67 (61–71)	1.000	0.097	< 0.001*
TG (mmol/L)#	0.85 (0.57–1.22)	1.01 (0.79–1.27)	1.29 (0.99–1.90)	1.000	< 0.001*	< 0.001*
Total cholesterol (mmol/L)	4.31 (4.06–4.90)	4.66 (3.96–5.09)	4.53 (4.00–5.14)	0.294	0.344	0.823
LDL cholesterol (mmol/L)	2.51 (2.17–2.82)	2.92 (2.38–3.26)	2.83 (2.34–3.32)	0.058	0.065	0.628
HDL cholesterol (mmol/L)#	1.39 (1.16–1.61)	1.22 (1.15–1.34)	1.05 (0.95–1.19)	0.268	< 0.001*	< 0.001*

Data are provided in median (IQR) or %. Mann–Whitney U test and Chi square test were used for univariate analysis of continuous and categorical data respectively. Significance was set at a p < 0.05. Bolded values with an asterisk indicate metabolites that are significantly different between groups. BMI: body mass index, C/M/I: Chinese/Malay/Indian, DBP: diastolic blood pressure, HDL: high density lipoprotein, HOMA-IR: homeostatic model assessment for insulin resistance, LDL: low density lipoprotein, OGTT: oral glucose tolerance test, SBP: systolic blood pressure, TG: Triglycerides. # represent variables which were analyzed pairwise between groups using post hoc Dunn’s test with Bonferroni correction for multiple comparisons.

Children and adolescents with MHO were significantly younger (13.1 years (10.1–15.4) vs 15.1 years (13.2–16.5)) and exhibited significantly lower fasting insulin (19.6 mU/L (11.9–26.0) vs 25.7 mU/L (17.4–36.0)), 2-h oral glucose tolerance test (OGTT) (5.80 mmol/L (5.30–6.40) vs 6.30 mmol/L (5.30–7.40)), C-peptide (797 pmol/L (672–800) vs 1047 pmol/L (800–1279)), HOMA-IR (4.08 (2.58–5.59) vs 5.65 (3.58–7.67)), SBP (111 mmHg (104–118) vs 124 mmHg (116–131)), diastolic blood pressure (DBP) (61 mmHg (57–67) vs 67 mmHg (61–71)), TG (1.01 mmol/L (0.79–1.27) vs 1.29 mmol/L (0.99–1.90)), and higher HDL cholesterol (1.22 mmol/L (1.15–1.34) vs 1.05 mmol/L (0.95–1.19)) than their MUO counterparts (Table 1).

Proportion of children and adolescents with MUO for each metabolic syndrome criteria

Among the children and adolescents with MUO, 47.8% were found to have two or more MetS criteria (Supplementary Fig. 1A). According to each MetS criteria, 48.6% had elevated blood pressure, 32.0% had hypertriglyceridemia, 55.9% had dyslipidaemia (abnormal HDL) and 26.1% had abnormal glucose tolerance (AGT) (Supplementary Fig. 1B).

Plasma metabolome of children and adolescents without obesity, children and adolescents with MHO and children and adolescents with MUO

Untargeted gas chromatography-time of flight mass spectrometry (GC-TOF/MS) metabolomic profiling of plasma metabolites in children and adolescents without obesity, those with MHO and those with MUO detected 60 putative metabolites, primarily consisting of FA, AA and their downstream metabolites, as detailed in Supplementary Table 2. Supervised partial least square discriminant analysis (PLS-DA) of plasma metabolome showed a discrimination between children and adolescent without obesity and those with obesity (R²X: 0.36, R²Y: 0.36, Q²: 0.29) (Fig. 1A). The corresponding variable importance in projection (VIP) plot revealed a metabolite signature rich in AA (BCAA, phenylalanine, lysine, serine, 2-methylalanine, proline, glycine and alanine), saturated FA (palmitic and stearic acid), unsaturated FA (linoleic and petroselinic acid), citric acid, cholesterol and 1,5-anhydroglucitol (Fig. 1B). Pairwise analysis revealed that children and adolescents without obesity had significantly higher concentrations of 1,5-anhydroglucitol, citric acid, leucine, isoleucine, stearic acid, glycine, serine, pyroglutamic acid and cholesterol and lower concentrations of 2-methylalanine than those with obesity (Supplementary Table 3).

[See PDF for image]

Fig. 1

Multivariate data analysis of plasma metabolites distinguishing between children and adolescents without obesity, those with MHO, and those with MUO. The PLS-DA plot depicts plasma metabolites that discriminate between (A) children and adolescents without obesity and those with obesity, (C) children and adolescents without obesity, those with MHO, and those with MUO, and (E) children and adolescents with MHO versus those with MUO. The Variable Importance in Projection (VIP) plots highlight important metabolites (VIP score > 1) that distinguish between (B) children and adolescents without obesity and those with obesity, (D) children and adolescents without obesity, those with MHO, and those with MUO, and (F) children and adolescents with MHO versus those with MUO. MHO: metabolically healthy obesity, MUO: metabolically unhealthy obesity. MHO: metabolically healthy obesity, MUO: metabolically unhealthy obesity.

Plasma metabolome also distinguished between children and adolescents without obesity, those with MHO and those with MUO (R²X: 0.37, R²Y: 0.16, Q²: 0.11) (Fig. 1C). The corresponding VIP plot revealed metabolites similar to those identified in the distinction between children and adolescents without obesity and those with obesity (Fig. 1D). Children and adolescents without obesity had significantly higher concentrations of citric acid, serine and lower concentrations of phenylalanine compared to those with MHO (Supplementary Table 4). In contrast, children and adolescents without obesity were found to have significantly higher concentrations of 1,5-anhydroglucitol, citric acid, stearic acid, glycine, serine and cholesterol and lower concentrations of leucine and isoleucine compared to those with MUO (Supplementary Table 4).

Plasma metabolites associated with MUO phenotype

Plasma metabolome was found to discriminate between metabolic health between children and adolescents with MHO and those with MUO (R²X: 0.37, R²Y: 0.15, Q²: 0.052) (Fig. 1E). The discriminatory metabolites with VIP score > 1 consisted of AA (BCAA, aspartic acid, ornithine, pyroglutamic acid, proline, methionine and lysine), saturated FA (palmitic and stearic acid), unsaturated FA (palmitoleic, linoleic and petroselinic acid), 1,5-anhydroglucitol and 2-hydroxybutyric acid (Fig. 1F). Children and adolescents with MHO had significantly higher concentrations of 1,5-anhydroglucitol, petroselinic acid, palmitoleic acid, 2-hydroxybutyric acid, aspartic acid and ornithine than those with MUO (Supplementary Table 5).

Correspondingly, lower concentrations of 1,5-anhydroglucitol, palmitoleic acid, petroselinic acid, 2-hydroxybutyric acid, aspartic acid and ornithine were found to be significantly associated with the MUO phenotype after adjusting for race, sex, age and BMI z-score (Supplementary Fig. 2).

Plasma metabolites associated with MetS criteria and cardiometabolic parameters among children and adolescents with MUO

We found that only higher concentrations of pyroglutamic acid (OR: 1.22, 95% CI: 1.00–1.49; p = 0.048) was marginally associated with dyslipidaemia (abnormal HDL). Lower concentrations of 1,5-anhydroglucitol (OR: 0.19, 95% CI: 0.06–0.54; p = 0.002) and higher concentrations of BCAA, including valine (OR: 1.34, 95% CI: 1.05–1.72; p = 0.02), leucine and isoleucine (OR: 1.41, 95% CI: 1.10–1.80; p = 0.006), 2-hydroxybutyric acid (OR: 3.74, 95% CI: 1.24–11.3; p = 0.02) and lysine (OR: 4.10, 95% CI: 1.31–12.9; p = 0.016) were significantly associated with AGT (Table 2). There were no metabolites significantly associated with elevated blood pressure and hypertriglyceridemia.

Table 2. Association of plasma metabolites with individual MetS criteria among children and adolescents with MUO.

	Elevated blood pressure	Hypertriglyceridemia	Dyslipidaemia (Abnormal HDL)	AGT
	Odds Ratio (95% CI)
Monosaccharide
1,5-Anhydroglucitol	0.79 (0.36–1.73)	1.25 (0.54–2.90)	1.74 (0.78–3.88)	0.19 (0.06–0.54)*
Unsaturated FA
Petroselinic acid	1.12 (0.67–1.87)	0.72 (0.41–1.26)	0.87 (0.52–1.48)	1.32 (0.75–2.32)
Linoleic acid	1.18 (0.36–3.84)	0.73 (0.20–2.65)	0.65 (0.19–2.18)	2.23 (0.63–7.98)
Palmitoleic acid	1.19 (0.63–2.25)	0.59 (0.28–1.27)	1.11 (0.57–2.13)	0.74 (0.34–1.61)
Saturated FA
Palmitic acid	1.02 (0.81–1.29)	0.95 (0.74–1.22)	0.92 (0.72–1.17)	1.23 (0.96–1.57)
Stearic acid	0.96 (0.60–1.53)	1.00 (0.61–1.64)	0.82 (0.50–1.35)	1.54 (0.94–2.54)
Ketone body
2-Hydroxybutyric acid	0.87 (0.32–2.43)	0.57 (0.19–1.77)	0.69 (0.24–1.94)	3.74 (1.24–11.3)*
BCAA
Leucine and isoleucine	0.96 (0.77–1.19)	1.17 (0.93–1.47)	0.95 (0.76–1.19)	1.41 (1.10–1.80)*
Valine	0.99 (0.80–1.23)	1.18 (0.94–1.48)	0.84 (0.68–1.05)	1.34 (1.05–1.72)*
Glucogenic AA
Aspartic acid	1.98 (0.61–6.43)	2.00 (0.59–6.86)	1.37 (0.41–4.52)	0.45 (0.11–1.74)
Proline	1.01 (0.81–1.25)	1.13 (0.90–1.43)	0.98 (0.79–1.22)	1.18 (0.93–1.51)
Methionine	1.03 (0.55–1.92)	1.21 (0.63–2.32)	0.98 (0.52–1.84)	1.93 (0.98–3.81)
Ketogenic AA
Lysine	0.96 (0.36–2.55)	1.22 (0.43–3.42)	0.90 (0.33–2.45)	4.10 (1.31–12.9)*
Other AA
Ornithine	2.04 (0.28–14.7)	1.20 (0.15–9.70)	1.13 (0.15–8.37)	2.20 (0.24–20.1)
Pyroglutamic acid	0.99 (0.82–1.20)	0.97 (0.79–1.19)	1.22 (1.00–1.49)*	0.91 (0.73–1.14)

Data are presented as odds ratios (95% CI). Statistical analyses were performed using logistic regression after adjusting for BMI z-score, sex, age and race. Significance was set at a p < 0.05. Bolded values with an asterisk indicate metabolites that are significantly associated with individual MetS criteria. AA: amino acids, AGT: abnormal glucose tolerance, BCAA: branched-chain amino acids, FA: fatty acids, HDL: high density lipoprotein.

Table 3. Correlation of plasma metabolites with cardiometabolic parameters among children and adolescents with MUO.

[See PDF for image]

Data are presented as correlation coefficients. Statistical analyses were performed using Partial Correlation after adjusting for BMI z-score, sex, age and race (except for the outcome of BMI-z score which was adjusted for sex, age and race only). Significance was set at a p < 0.05. Bolded values with an asterisk indicate metabolites that are significantly associated with individual cardiometabolic parameters. Red and green shadings represent negative and positive correlations respectively while the increased intensity of the shadings represent stronger correlations of the metabolites with cardiometabolic parameters. AA: amino acids, BCAA: branched-chain amino acids, BMI: body mass index, DBP: diastolic blood pressure, FA: fatty acids, HDL: high density lipoprotein, HOMA-IR: homeostatic model assessment for insulin resistance, OGTT: oral glucose tolerance test, SBP: systolic blood pressure, TG: Triglycerides.

Aspartic acid (r = 0.171, p = 0.014) and ornithine (r = 0.143, p = 0.041) were significantly positively correlated with SBP (Table 3). 2-Hydroxybutyric acid (r = 0.156, p = 0.026), leucine and isoleucine (r = 0.197, p = 0.005), valine (r = 0.148, p = 0.035), proline (r = 0.142, p = 0.043), methionine (r = 0.178, p = 0.011), and ornithine (r = 0.198, p = 0.004) were significantly positively correlated with fasting glucose while 1,5-anhydroglucitol (r = − 0.199, p = 0.004) was significantly inversely correlated with 2-h OGTT (Table 3). 1,5-anhydroglucitol (r = − 0.147, p = 0.035), petroselinic acid (r = − 0.224, p = 0.001) and palmitoleic acid (r = − 0.213, p = 0.002) were significantly inversely correlated with HOMA-IR while leucine and isoleucine (r = 0.200, p = 0.004), valine (r = 0.201, p = 0.004), proline (r = 0.170, p = 0.015), methionine (r = 0.205, p = 0.003), lysine (r = 0.236, p = 0.001) and ornithine (r = 0.388, p < 0.001) were significantly positively correlated with HOMA-IR (Table 3).

After stratifying the children and adolescents with MUO by sex, we observed a consistent overall pattern across both sexes, with metabolomic perturbations primarily associated with abnormal glucose homeostasis (Supplementary Tables 6 to 9). However, notable sex-specific differences emerged in individual metabolite associations. In males with MUO, lower plasma 1,5-anhydroglucitol was significantly associated with AGT (Supplementary Table 6), and discriminatory metabolites were more strongly linked to fasting glucose (Supplementary Table 7). In contrast, in females with MUO, higher plasma BCAA, methionine, and lysine were significantly associated with AGT (Supplementary Table 8), and discriminatory metabolites were primarily associated with HOMA-IR (Supplementary Table 9).

Plasma metabolome of children and adolescents with obesity and normal glucose tolerance (NGT) and children and adolescents with obesity and AGT

Plasma metabolome of children and adolescents with obesity and NGT could be differentiated from those with obesity and AGT (R²X: 0.38, R²Y: 0.22, Q²: 0.054) (Fig. 2A). The corresponding VIP plot identified 1,5-anhydroglucitol, glyceric acid, palmitoleic acid, leucine, isoleucine, valine, 3-hydroxyisobutyric acid (3-HIB), 2-methylalanine, lysine, methionine, and aspartic acid as key metabolites differentiating between these groups (Fig. 2B).

[See PDF for image]

Fig. 2

Multivariate data analysis of plasma metabolites that discriminate between children and adolescents with obesity and NGT and those with obesity and AGT. (A) PLS-DA plot illustrates the plasma metabolites that discriminate between children and adolescents with obesity and NGT and those with obesity and AGT. (B) Variable Importance in Projection (VIP) plot highlights key metabolites with VIP scores greater than 1 that distinguish between children and adolescents with obesity and normal glucose tolerance (NGT) and those with obesity and abnormal glucose tolerance (AGT). AGT: abnormal glucose tolerance, NGT: normal glucose tolerance.

Children and adolescents with obesity and AGT had significantly higher concentrations of leucine and isoleucine (p = 0.01) and 3-HIB (p = 0.003) and significantly lower concentrations of 1,5-anhydroglucitol (p < 0.001), glyceric acid (p = 0.015) and aspartic acid (p = 0.027) than those with obesity and NGT (Fig. 3 and Supplementary Table 10).

[See PDF for image]

Fig. 3

Fold change (FC) of important plasma metabolite levels (VIP score > 1) in the AGT group relative to the NGT group (with FC set to 1 for NGT). Middle, lower and upper limits of the box plot represent median and IQR (25th and 75th percentile) respectively. Red boxes denote significantly elevated metabolites, while blue boxes indicate significantly reduced metabolites in the AGT group compared to the NGT group. *: p < 0.05; **: p ≤ 0.01; ***: p < 0.001. 1,5-AHG: 1,5-Anhydroglucitol; 3-HIB: 3-Hydroxyisobutyric acid; AA: Amino acid; BCAA: Branched-chain amino acids; FA: Fatty acids.

Discussion

In this cross-sectional untargeted metabolomics study performed in a clinically well-characterized Asian pediatric cohort, we observed distinct alterations in the plasma metabolome between MHO and MUO in Asian children and adolescents. We identified, for the first time, that perturbations in the plasma metabolome of children and adolescents with MUO phenotype are predominantly associated with abnormal glucose homeostasis.

Children and adolescents with obesity exhibited significantly altered plasma concentrations of various AA (glycine, serine and pyroglutamic acid), lipids, 1,5-anhydroglucitol and citric acid compared to their counterparts without obesity. The decreased levels of AA in our cohort of children and adolescents with obesity align with findings from metabolomic studies in other ethnic populations of children with obesity ^14,22,23. Serine, glycine and pyroglutamic acid serve as precursors of glutathione, and their reduced levels may reflect lowered plasma glutathione levels previously observed in children with obesity ²⁴. Glycine, in particular, has been inversely associated with pro-inflammatory marker, IL-6 ²⁵, and lowered glycine levels have been proposed as a biomarker of cardiovascular disease in a small cohort of children with obesity ²⁵. Similarly, reduced levels of citric acid which have been reported in other studies of children with obesity ^17,22 were found in our children and adolescents with obesity. Citric acid is a key intermediate in the tricarboxylic acid cycle that is responsible for ATP synthesis ²⁶. It is produced from the condensation reaction between oxaloacetate and acetyl-CoA, of which the latter is generated through mitochondrial metabolism of glucose, FA and AA ²⁶. The reduced levels of citric acid may signify dysfunctional mitochondrial metabolic processes that are characteristic of obesity ²⁷.

Our findings also demonstrated that plasma metabolome of children and adolescents without obesity was more similar to those with MHO than their MUO counterparts. This finding aligns with other studies that have identified more pronounced biochemical perturbations between children with MUO and children without obesity, as compared to between children with MHO and those without obesity ²⁸. These results reinforce the current paradigm that MHO represents an intermediate and potentially transient state between individuals without obesity and the MUO phenotype ^10,29.

We observed lower concentrations of MUFA such as petroselinic acid and palmitoleic acid in association with MUO phenotype. Petroselinic acid is a ligand of peroxisome proliferator-activated receptors α (PPAR-α) which upregulates FA and glucose uptake and reduces inflammation ³⁰. Higher concentrations of palmitoleic acid have been associated with a lower risk of T2DM, IR and cardiometabolic risk factors ³¹. The association of specific AAs with MUO remains debated. Mihalik et al. and Bervoets et al. reported lower concentrations of BCAA, aromatic AA, histidine, glutamine, lysine, methionine, serine, glycine and citrulline in children and adolescents with MUO ^14,17. In contrast, Mangge et al. and Tong et al. found higher plasma concentrations of BCAA, aromatic AA, arginine, ornithine, citrulline and histidine in juveniles with MUO ^19,20. In our study, lower levels of aspartic acid and ornithine were associated with MUO. Both aspartic acid and ornithine are crucial substrates within the urea cycle which is responsible for conversion of toxic ammonia, produced from AA catabolism, into urea in the mitochondria ³². Lower levels of both AA could be suggestive of dysfunctional mitochondrial catabolic process in MUO leading to AA accumulation. Correspondingly, a lifestyle intervention study conducted for adolescents with MUO found that increases in AA within the urea cycle were associated with improvements with weight and insulin sensitivity ³³.

Unlike other studies that define MUO as obesity in the presence of two or more MetS criteria ^{17, 18, 19–20}, our stringent definition of MUO ⁶ allowed for a more precise classification and study of specific metabolic conditions within the obese phenotype. Further analysis of the relationship between significantly perturbed metabolites and cardiometabolic parameters revealed that a greater number of metabolites were correlated with abnormal glucose homeostasis compared to other cardiometabolic parameters. This emphasizes glucose dysregulation as a major feature underlying the development of MUO ^15,16,34. We also showed that the monosaccharide, 1,5-anhydroglucitol, was reduced in children and adolescents with obesity and AGT. 1,5-Anhydroglucitol competes with glucose for reabsorption from the kidneys ³⁵ and lowered plasma levels are a validated marker of post prandial hyperglycaemia ³⁵. The lower concentration of plasma 1,5-anhydroglucitol aligns with elevated plasma glucose levels in the AGT group. Additionally, elevated BCAA and 3-HIB (a catabolic intermediate of valine) were observed in our children and adolescents with obesity and AGT. Increased plasma BCAA have been previously reported in children with obesity and IR ¹⁹ and are predictive of incident T2DM up to 12 years before diagnosis ³⁶. Elevated BCAA are a result of impaired metabolism in adipose tissues and are postulated to perpetrate mitochondrial dysfunction and subsequent IR by saturating catabolic pathways in the liver and skeletal muscle ³⁷. While elevated 3-HIB had been previously linked to both obesity and IR in adults ³⁸, our study is the first to replicate this finding in a pediatric cohort, showing elevated 3-HIB in children and adolescents with obesity and AGT. 3-HIB mediates IR through facilitating increased FA uptake in skeletal muscle ³⁹ to form toxic lipid intermediates that impair subsequent glucose metabolism and disposal.

We acknowledge several limitations in our study. First, as our study is cross-sectional in nature, causality cannot be inferred from the observed associations. Longitudinal studies are needed to establish whether the identified metabolites predict the transition from MHO (obesity with NGT) to MUO (obesity with AGT), or if they are consequential biomarkers of the MUO phenotype (obesity with AGT). Additionally, mechanistic and interventional studies are warranted to clarify the functional and therapeutic potential of these metabolites. Second, as our study population comprised predominantly Asian ethnicities (Chinese, Malay, Indian), the generalizability of our findings to children of other ethnic backgrounds (e.g. Caucasians, Africans, Hispanics) may be limited. Although similar metabolomic alterations have been reported in non-Asian paediatric cohorts with MUO ^17,19, larger studies in ethnically diverse populations are needed to determine whether these metabolites reflect universal or population-specific signatures of MUO. Third, we acknowledge that the uneven distribution of participants across the lean (n = 24), MHO (n = 65) and MUO (n = 222) groups may introduce bias or skew the analysis. However, our study was adequately powered based on prior sample size calculations (as described in the Methods section) and using the smallest group (n = 24) provided sufficient statistical power to detect significant differences across groups. Increasing the sample size of the MHO or MUO groups would therefore not alter the calculated power. Importantly, all participants with obesity were included without preselection into MHO or MUO groups, minimizing recruitment bias. As such, the group proportions reflect the natural distribution of MHO and MUO phenotypes in our cohort based on the applied classification criteria. Fourth, our study did not account for the influence of factors such as diet and lifestyle habits that may impact plasma metabolite profiles. Fifth, untargeted metabolomics utilizing GC-TOF/MS is less sensitive than targeted liquid chromatography tandem mass spectrometry (LC–MS/MS) for metabolomic profiling. Nevertheless, the untargeted approach by GC-TOF/MS allowed broad metabolite coverage, including AA, FA, intermediates, and key sugars (1,5-anhydroglucitol). To minimize false positives and account for sensitivity limitations, we applied strict quality control (QC) criteria: only metabolites detected in ≥ 70% of QC samples and with a similarity index > 700 (out of 1000) was included. This ensured downstream analyses focused on reliably detected endogenous metabolites.

Conclusion

Our study revealed distinct alterations in plasma metabolome between MHO and MUO in Asian children and adolescents. Notably, we identified, for the first time, that plasma metabolome perturbations in the MUO phenotype are predominantly linked to impaired glucose homeostasis. These findings suggest that metabolic pathways involved in glucose regulation may serve as potentially important therapeutic targets for improving health outcomes in the obese phenotype.

Methods

Study population

Children and adolescents with obesity (n = 287) of Chinese, Malay and Indian race enrolled under the OBesity in Singapore Children (OBiSC) study were included in this study. These participants were recruited from National University Hospital and Health Promotion Board, Singapore. To be eligible for the OBiSC study, children and adolescents aged 5–20 years old had: 1) obesity before age of 10 years, 2) BMI for age ≥ 97th percentile, 3) no syndromic causes of obesity. Children and adolescents without obesity (n = 24) with a BMI for age < 85th percentile were included in this study as controls, and they were family members (siblings) of the participants with obesity. The minimum sample size per group for this study was estimated based on a prior study comparing the plasma metabolome of young individuals (aged 8–18 years) without obesity, with MHO and with MUO ¹⁷. At a significance level of α = 0.05, a minimum of 10 and 13 participants per group were required to detect significant differences with 80% and 90% power, respectively. As our study employed an untargeted metabolomics approach, which is inherently exploratory in nature, we opted to include the full cohort of recruited participants to maximize the discovery potential of metabolites associated with the MUO phenotype. The study was performed in accordance with the Declaration of Helsinki and ethics approval was obtained from Domain Specific Review Board of National Healthcare Group, Singapore (Reference number: 2015/00314). Written informed consent was obtained from all study participants and their parents and/or legal guardians. The study is registered under clinicaltrials.gov (NCT02418377).

Classification of children and adolescents into MHO and MUO groups

There is currently a lack of consensus on the definition of MHO ⁴⁰. While most studies have defined MHO as central obesity with one or no MetS criteria ^17,19, our group previously demonstrated that applying different MHO definitions based on the presence or absence of MetS component yielded varying MHO prevalence within the same cohort of children and adolescents with obesity ⁶. Notably, using the less stringent definition (central obesity with ≤ 1 criteria) can result in classifying individuals with dyslipidaemia, hyperglycaemia, hypertension or type 2 diabetes as MHO, despite these being metabolic disorders. Therefore, we adopted a more stringent definition, classifying MHO as central obesity with no metabolic abnormalities, to ensure a metabolically distinct group. The criteria for our stringent metabolic health definition are aligned with the International Diabetes Federation guidelines for classification of MetS in youths ⁴¹. MHO was defined as obesity with none of the following sub-conditions: (1) hypertriglyceridemia: fasting TG ≥ 1.7 mmol/L or on hyperlipidaemia medication, (2) dyslipidaemia (abnormal HDL): HDL cholesterol < 1.03 mmol/L for children under the age of 16, HDL cholesterol < 1.03 mmol/L for male ≥ 16 years old and HDL cholesterol < 1.29 mmol/L for female ≥ 16 years old, (3) AGT: fasting glucose ≥ 5.6 mmol/L or 2-h OGTT ≥ 7.8 mmol/L or on diabetic medication, (4) elevated blood pressure: blood pressure ≥ 90th percentile based on age, sex, and height or on hypertensive medication.

Anthropometric and biochemical measurements

The standard anthropometric parameters including weight, height, waist circumference and hip circumference were measured. BMI was calculated as weight in kilograms (kg) divided by the square of height in meters (m), and BMI z-score was adjusted for child’s age and sex based on local growth chart ⁴². Body fat percentage was assessed by bioelectrical impedance analysis (BIA) using Tanita body composition analyzer (Model BC-418). Blood pressure was also measured (Carescape V100 Dinamap). The participants fasted for 8–10 h prior to the blood test. Fasting blood samples were collected and analyzed for fasting glucose, insulin, and lipid levels. Following the fasting blood draw, participants underwent an OGTT, during which they consumed a 75 g glucose solution within 5 min. Blood samples were subsequently collected 120 min after completing the glucose drink. No additional food or beverages were permitted during the 120-min waiting period. HOMA-IR was calculated as previously described ⁴³.

Global untargeted metabolomics utilizing GC-TOF/MS

Plasma was extracted from fasting blood of 287 children and adolescents with obesity and 24 children and adolescents without obesity and stored at − 80 °C until use.

Briefly, 50 μL aliquots of plasma samples were first thawed to room temperature (24 °C) before 200 μL of methanol (TEDIA®) containing 0.004 mg/mL myristic-d₂₇ acid (IS) was added to each sample for protein precipitation. Protein precipitated samples were vortexed briefly at room temperature and centrifuged for 5 min at 2000 g at 24 °C. 150 μL of the supernatant was transferred to a 15 mL glass tube and dried under nitrogen gas using a TurboVap. The dried samples were resuspended in 100 μL of toluene, vortexed vigorously for 10 s, and dried again under nitrogen. A two-step derivatization method was employed for the chemical derivatization of metabolites. Samples were first incubated with 30 μL of 2% MOX reagent (methoxyamine chloride in pyridine; Thermofisher Scientific) for 1.5 h at 60 °C. Following this, 30 μL of MSTFA, 1% TMCS (2,2,2-trifluoro-N-methyl-N-(trimethylsilyl)-acetamide, 1% chlorotrimethylsilane; Thermofisher Scientific) was added for another 1 h incubation at 60 °C. A total of 50 μL of derivatized samples were transferred to silanized glass vials for GC-TOF/MS analysis.

Untargeted GC-TOF/MS analysis was performed using an Agilent 7890A gas chromatograph (Agilent Technologies) coupled to a LECO Pegasus (4D) time-of-flight (TOF) mass analyzer operating in GC–MS mode (LECO Corp) under previously described chromatographic methods with modifications ⁴⁴. Consistent sample injection was performed using a CTC PAL autosampler (CTC Analytics AG). Injection temperature and volume were set at 250 °C and 1 μL with a spilt ratio of 1:2, respectively. A DB-5 29.75 m × 0.25 mm, 0.25 μm film thickness (Agilent Technologies) capillary column was used with a constant helium carrier gas flow of 1.5 mL/min. A two-step temperature gradient was adopted, where the temperature was initially held at 70 °C for 2 min before increasing to 200 °C at 20 °C/min and held for another 2 min. Following which, the temperature was increased to 310 °C at 10 °C/min and held for 4 min. The transfer line and ion source temperatures were set at 300 °C and 250 °C, respectively. A 400 s acquisition delay of mass spectra was pre-set post-injection before collecting ionized compounds with a mass-to-charge (m/z) ratio of 45–600 at 20 Hz using ionization energy of 70 eV and a detector voltage of 1650 V.

All chromatograms obtained from GC-TOF/MS analyses were subjected to baseline correction, de-convolution, noise reduction, smoothing, library matching and area calculation using ChromaTOF software (version 4.41, Leco Corporation). De-convolution and peak finding were performed using 1) signal-to-noise threshold of 100, 2) minimum peak width of 3 s and 3) minimum of three apexing masses. Peak area was normalized using total ion chromatogram. Instrument performance was assessed using the peak area of myristic-_d27 acid, the internal standard. The “calibration” function in ChromaTOF was utilized for peak alignment as previously described ⁴⁴.

A reference table of putative metabolites initially constructed using the data obtained from the chromatograms of QC samples. The QC samples comprised pooled plasma from children and adolescents without obesity, those with MHO, and those with MUO from our cohort. A metabolite was only included in the reference table if 1) it was detected in at least 70% of the QC samples and 2) its similarity index exceeded the threshold of 700. These stringent criteria ensured that the reference table only included endogenous putative metabolites that were present in adequately high abundance in plasma. Chromatographic peaks derived from glucose and sugars (except 1,5-anhydroglucitol) were saturated in most samples and were excluded from the reference table.

Statistical analyses

Statistical analyses were performed using IBM SPSS Statistics (version 26.0, IBM Corp.) with level of significance set at 2-sided significance level set at p < 0.05. As data were non-normally distributed, descriptive statistics for numerical variables were presented as median with interquartile range (IQR: 25th quartile-75th quartile), while categorial variables were presented as proportions (%). For continuous data, significant differences between groups were assessed by Kruskal–Wallis test for three-group comparisons and Mann Whitney-U test for two-group comparisons. When the Kruskal–Wallis test indicated a significant difference (p < 0.05) between groups, post hoc Dunn’s test with Bonferroni correction was performed to identify significant pairwise difference between groups (p < 0.05). Categorical data between groups were analyzed by Chi Square test. Binary logistic regression with adjustment for age, sex, race and BMI z-score was utilized to examine metabolites that were associated with MUO phenotype. Partial correlation test adjusted for age, sex, race and BMI z-score (except for the outcome of BMI z-score which was adjusted for age, sex and race only) was used to correlate metabolite levels with cardiometabolic parameters such as TG, HDL, SBP, DBP, BMI z-score, waist to hip ratio, fasting glucose, 2-h OGTT and HOMA-IR. Covariates were selected based on their known or potential influence on MUO risk ^45,46. Supervised PLS-DA was conducted using SIMCA-P + software (version 12.0.1.0, Umetrics, www.umetrics.com) to determine the discriminatory effect of plasma metabolites on different groups. Plasma metabolites with VIP score > 1 were considered important in discriminating between groups.

Author contributions

D.Z.W.N. was responsible for data analysis, results interpretation and writing of the manuscript. Y.H.S., D.Y.Q.K. and A.A.S. contributed to acquisition of data. S.R.P. contributed to data analysis. B.W.L. and Y.S.L. were involved in the review of the manuscript and providing of technical advice pertaining to analyses of data. E.C.Y.C. and D.S.Q.O. were responsible for conceptualisation of study design, supervision of project, review of data analyses and writing and review of manuscript drafts. D.S.Q.O. also contributed to funding acquisition. All authors had final approval of the submitted and published versions.

Funding

This research is supported by the National University Health System Seed Fund 2019 (Grant: A-0002738-01-00).

Data availability

The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

References

1. Abarca-Gómez, L et al. Worldwide trends in body-mass index, underweight, overweight, and obesity from 1975 to 2016: A pooled analysis of 2416 population-based measurement studies in 128 9 million children, adolescents, and adults. The lancet; 2017; 390, 10113 pp. 2627-2642. [DOI: https://dx.doi.org/10.1016/S0140-6736(17)32129-3]

2. Ministry of Health (MOH), S., Annual prevalence of obesity for children aged below 18 over the past five years, their profile and assessed effectiveness of preventative measures. 14 Febuary 2022: https://www.moh.gov.sg/news-highlights/details/annual-prevalence-of-obesity-for-children-aged-below-18-over-past-five-years-their-profile-and-assessed-effectiveness-of-preventive-measures/#:~:text=14th%20Feb%202022&text=From%202017%20to%202021%2C%20the,in%20the%20last%20two%20years.

3. Han, TS; Lean, ME. A clinical perspective of obesity, metabolic syndrome and cardiovascular disease. JRSM Cardiovasc. Dis.; 2016; 5, 2048004016633371. [PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4780070][DOI: https://dx.doi.org/10.1177/2048004016633371]

4. Karelis, A; Brochu, M; Rabasa-Lhoret, R. Can we identify metabolically healthy but obese individuals (MHO)?. Diabetes Metab.; 2004; 30, 6 pp. 569-572.1:STN:280:DC%2BD2M%2FktlagsA%3D%3D [DOI: https://dx.doi.org/10.1016/S1262-3636(07)70156-8]

5. Liu, C et al. The prevalence of metabolically healthy and unhealthy obesity according to different criteria. Obes. Facts; 2019; 12, 1 pp. 78-90.1:CAS:528:DC%2BC1MXisV2ksLfM [PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6465689][DOI: https://dx.doi.org/10.1159/000495852]

6. Ooi, DSQ et al. Prevalence and predictors of metabolically healthy obesity in severely obese Asian children. Pediatric Res.; 2022; 92, pp. 1374-1380.1:CAS:528:DC%2BB38Xjt1els70%3D [DOI: https://dx.doi.org/10.1038/s41390-022-01941-z]

7. Primeau, V et al. Characterizing the profile of obese patients who are metabolically healthy. Int. J. Obes.; 2011; 35, 7 pp. 971-981.1:STN:280:DC%2BC3MnntVOquw%3D%3D [DOI: https://dx.doi.org/10.1038/ijo.2010.216]

8. Phillips, CM; Perry, IJ. Does inflammation determine metabolic health status in obese and nonobese adults?. J. Clin. Endocrinol. Metab.; 2013; 98, 10 pp. E1610-E1619.1:CAS:528:DC%2BC3sXhs1elsLnJ [DOI: https://dx.doi.org/10.1210/jc.2013-2038]

9. Lin, L et al. Transition of metabolic phenotypes and risk of subclinical atherosclerosis according to BMI: A prospective study. Diabetologia; 2020; 63, 7 pp. 1312-1323. [DOI: https://dx.doi.org/10.1007/s00125-020-05116-5]

10. Appleton, SL et al. Diabetes and cardiovascular disease outcomes in the metabolically healthy obese phenotype: A cohort study. Diabetes Care; 2013; 36, 8 pp. 2388-2394. [PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3714523][DOI: https://dx.doi.org/10.2337/dc12-1971]

11. Cheng, D et al. Metabolomic signature between metabolically healthy overweight/obese and metabolically unhealthy overweight/obese: A systematic review. Diabetes Metabol. Syndrome Obesity Targets Therapy; 2021; 14, 991. [DOI: https://dx.doi.org/10.2147/DMSO.S294894]

12. Telle-Hansen, VH et al. A comprehensive metabolic profiling of the metabolically healthy obesity phenotype. Lipids Health Dis.; 2020; 19, 1 pp. 1-12. [DOI: https://dx.doi.org/10.1186/s12944-020-01273-z]

13. Badoud, F et al. Serum and adipose tissue amino acid homeostasis in the metabolically healthy obese. J. Proteome Res.; 2014; 13, 7 pp. 3455-3466.1:CAS:528:DC%2BC2cXpslemsrY%3D [DOI: https://dx.doi.org/10.1021/pr500416v]

14. Mihalik, SJ et al. Metabolomic profiling of fatty acid and amino acid metabolism in youth with obesity and type 2 diabetes: Evidence for enhanced mitochondrial oxidation. Diabetes Care; 2012; 35, 3 pp. 605-611.1:CAS:528:DC%2BC38XltFGksL8%3D [PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3322714][DOI: https://dx.doi.org/10.2337/DC11-1577]

15. Suzuki, Y et al. Associations among amino acid, lipid, and glucose metabolic profiles in childhood obesity. BMC Pediatr.; 2019; 19, 1 pp. 1-11. [DOI: https://dx.doi.org/10.1186/s12887-019-1647-8]

16. Newbern, D et al. Sex differences in biomarkers associated with insulin resistance in obese adolescents: Metabolomic profiling and principal components analysis. J. Clin. Endocrinol. Metab.; 2014; 99, 12 pp. 4730-4739.1:CAS:528:DC%2BC2cXitFalsbbP [PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4328030][DOI: https://dx.doi.org/10.1210/jc.2014-2080]

17. Bervoets, L et al. Identification of metabolic phenotypes in childhood obesity by 1H NMR metabolomics of blood plasma. Future Sci. OA; 2018; 4, 6 FSO310.1:CAS:528:DC%2BC1cXhtlWrtLnM [PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6060399][DOI: https://dx.doi.org/10.4155/fsoa-2017-0146]

18. Jeong, S et al. Identification of biomarkers related to metabolically unhealthy obesity in Korean obese adolescents: A cross-sectional study. Children; 2023; 10, 2 322. [PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9955165][DOI: https://dx.doi.org/10.3390/children10020322]

19. Mangge, H et al. Branched-chain amino acids are associated with cardiometabolic risk profiles found already in lean, overweight and obese young. J. Nutr. Biochem.; 2016; 32, pp. 123-127.1:CAS:528:DC%2BC28XlvVehsrs%3D [DOI: https://dx.doi.org/10.1016/j.jnutbio.2016.02.007]

20. Tong, L et al. Metabolome profiling and pathway analysis in metabolically healthy and unhealthy obesity among Chinese adolescents Aged 11–18 years. Metabolites; 2023; 13, 5 641.1:CAS:528:DC%2BB3sXhtFemtL%2FP [PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10221088][DOI: https://dx.doi.org/10.3390/metabo13050641]

21. Holmes, E; Wilson, ID; Nicholson, JK. Metabolic phenotyping in health and disease. Cell; 2008; 134, 5 pp. 714-717.1:CAS:528:DC%2BD1cXhtFCqs7jE [DOI: https://dx.doi.org/10.1016/j.cell.2008.08.026]

22. Butte, NF et al. Global metabolomic profiling targeting childhood obesity in the Hispanic population. Am. J. Clin. Nutr.; 2015; 102, 2 pp. 256-267.1:CAS:528:DC%2BC2MXhtlOlu7jF [PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4515872][DOI: https://dx.doi.org/10.3945/ajcn.115.111872]

23. Short, KR et al. Effect of obesity and exercise training on plasma amino acids and amino metabolites in American Indian adolescents. J. Clin. Endocrinol. Metab.; 2019; 104, 8 pp. 3249-3261. [PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6584131][DOI: https://dx.doi.org/10.1210/jc.2018-02698]

24. Codoñer-Franch, P et al. Is obesity associated with oxidative stress in children?. Int. J. Pediatr. Obes.; 2010; 5, 1 pp. 56-63. [DOI: https://dx.doi.org/10.3109/17477160903055945]

25. do Prado, WL et al. Preliminary evidence of glycine as a biomarker of cardiovascular disease risk in children with obesity. Int. J. Obesity; 2023; 47, 10 pp. 1023-1026. [DOI: https://dx.doi.org/10.1038/s41366-023-01354-w]

26. Martínez-Reyes, I; Chandel, NS. Mitochondrial TCA cycle metabolites control physiology and disease. Nat. Commun.; 2020; 11, 1 102.2020NatCo.11.102M [PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6941980][DOI: https://dx.doi.org/10.1038/s41467-019-13668-3]

27. de Mello, AH et al. Mitochondrial dysfunction in obesity. Life Sci.; 2018; 192, pp. 26-32. [DOI: https://dx.doi.org/10.1016/j.lfs.2017.11.019]

28. Bagheri, M et al. Obesity-related metabolomic profiles and discrimination of metabolically unhealthy obesity. J. Proteome Res.; 2018; 17, 4 pp. 1452-1462.1:CAS:528:DC%2BC1cXjsVyltro%3D [DOI: https://dx.doi.org/10.1021/acs.jproteome.7b00802]

29. Lind, L et al. Metabolic profiling of obesity with and without the metabolic syndrome: A multisample evaluation. J. Clin. Endocrinol. Metab.; 2022; 107, 5 pp. 1337-1345. [DOI: https://dx.doi.org/10.1210/clinem/dgab922]

30. Koves, TR et al. Peroxisome proliferator-activated receptor-γ co-activator 1α-mediated metabolic remodeling of skeletal myocytes mimics exercise training and reverses lipid-induced mitochondrial inefficiency. J. Biol. Chem.; 2005; 280, 39 pp. 33588-33598.1:CAS:528:DC%2BD2MXhtVensb7M [DOI: https://dx.doi.org/10.1074/jbc.M507621200]

31. Frigolet, ME; Gutiérrez-Aguilar, R. The role of the novel lipokine palmitoleic acid in health and disease. Adv. Nutr.; 2017; 8, 1 pp. 173S-181S.1:CAS:528:DC%2BC1cXitlWksrfI [PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5227969][DOI: https://dx.doi.org/10.3945/an.115.011130]

32. Barmore, W., F. Azad, and W.L. Stone, Physiology, urea cycle. (2018).

33. Jachthuber Trub, C et al. Impact of lifestyle Intervention on branched-chain amino acid catabolism and insulin sensitivity in adolescents with obesity. Endocrinol. Diabetes Metabol.; 2021; 4, 3 e00250.1:CAS:528:DC%2BB3MXhs1amtbfO [DOI: https://dx.doi.org/10.1002/edm2.250]

34. Tricò, D et al. Elevated α-hydroxybutyrate and branched-chain amino acid levels predict deterioration of glycemic control in adolescents. J. Clin. Endocrinol. Metab.; 2017; 102, 7 pp. 2473-2481. [PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5505187][DOI: https://dx.doi.org/10.1210/jc.2017-00475]

35. Dungan, KM. 1, 5-anhydroglucitol (GlycoMark™) as a marker of short-term glycemic control and glycemic excursions. Expert Rev. Mol. Diagn.; 2008; 8, 1 pp. 9-19.1:CAS:528:DC%2BD2sXhsVCgtr7L [DOI: https://dx.doi.org/10.1586/14737159.8.1.9]

36. Wang, TJ et al. Metabolite profiles and the risk of developing diabetes. Nat. Med.; 2011; 17, 4 pp. 448-453. [PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3126616][DOI: https://dx.doi.org/10.1038/nm.2307]

37. Newgard, CB. Interplay between lipids and branched-chain amino acids in development of insulin resistance. Cell Metab.; 2012; 15, 5 pp. 606-614.1:CAS:528:DC%2BC38XmsVCis7k%3D [PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3695706][DOI: https://dx.doi.org/10.1016/j.cmet.2012.01.024]

38. Mardinoglu, A et al. Elevated plasma levels of 3-hydroxyisobutyric acid are associated with incident type 2 diabetes. EBioMedicine; 2018; 27, pp. 151-155. [DOI: https://dx.doi.org/10.1016/j.ebiom.2017.12.008]

39. Jang, C et al. A branched-chain amino acid metabolite drives vascular fatty acid transport and causes insulin resistance. Nat. Med.; 2016; 22, 4 pp. 421-426.1:CAS:528:DC%2BC28XjslGrtLc%3D [PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4949205][DOI: https://dx.doi.org/10.1038/nm.4057]

40. Tsatsoulis, A; Paschou, SA. Metabolically healthy obesity: Criteria, epidemiology, controversies, and consequences. Curr. Obes. Rep.; 2020; 9, 2 pp. 109-120. [DOI: https://dx.doi.org/10.1007/s13679-020-00375-0]

41. Zimmet, P et al. The metabolic syndrome in children and adolescents—an IDF consensus report. Pediatr Diabetes; 2007; 8, 5 pp. 299-306. [DOI: https://dx.doi.org/10.1111/j.1399-5448.2007.00271.x]

42. Health Promotion Board, Singapore. Growth Charts for Children in Singapore <https://www.cfps.org.sg/assets/1-Circular-for-GPs/6-CDS-Guidance-Annex-D-Child-Health-Booklet-2014.pdf>. 2000.

43. Wallace, TM; Levy, JC; Matthews, DR. Use and abuse of HOMA modeling. Diabetes Care; 2004; 27, 6 pp. 1487-1495. [DOI: https://dx.doi.org/10.2337/diacare.27.6.1487]

44. Chan, ECY; Pasikanti, KK; Nicholson, JK. Global urinary metabolic profiling procedures using gas chromatography–mass spectrometry. Nat. Protoc.; 2011; 6, 10 pp. 1483-1499.1:CAS:528:DC%2BC3MXht1CrtrjL [DOI: https://dx.doi.org/10.1038/nprot.2011.375]

45. Iacobini, C et al. Metabolically healthy versus metabolically unhealthy obesity. Metabolism; 2019; 92, pp. 51-60.1:CAS:528:DC%2BC1cXitlWnsb7O [DOI: https://dx.doi.org/10.1016/j.metabol.2018.11.009]

46. Wang, J-S et al. Trends in the prevalence of metabolically healthy obesity among US adults, 1999–2018. JAMA Netw. Open; 2023; 6, 3 pp. e232145-e232145. [PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9999245][DOI: https://dx.doi.org/10.1001/jamanetworkopen.2023.2145]