Impact of lifestyle Intervention on branched‐chain amino acid catabolism and insulin sensitivity in adolescents with obesity

Introduction Insulin resistance in adolescents with obesity associates with a sex‐dependent metabolic ‘signature’ comprising branched‐chain amino acids (BCAAs), glutamate and C3/C5 acylcarnitines (C3/C5), implicating altered flux through BCAA catabolic pathways. Here, we investigated the effects of lifestyle intervention on BCAA catabolism and insulin sensitivity. We hypothesized (1) weight reduction and improved insulin sensitivity associate with enhanced BCAA catabolism; (2) baseline BCAAs and their metabolic by‐products predict changes in weight and insulin sensitivity during lifestyle intervention. Methods A 33 adolescents with obesity were studied before and after 6 months of lifestyle intervention. Principal component analysis and multiple linear regression models were used to correlate changes in metabolic factors with changes in weight and insulin sensitivity assessed by HOMA‐IR, adiponectin and ratio of triglyceride (TG) to HDL. Baseline metabolic factors were used as explanatory variables in prediction models. Results Weight reduction was associated with reductions in BCAA, glutamate, and C3/C5 (p = .002) and increases in urea cycle AA (p = .029), suggesting an increase in BCAA catabolism. Increases in urea cycle AA during weight reduction were associated with increases in adiponectin, a marker of insulin sensitivity. Markers of insulin resistance (high BCAA, glutamate, and C3/C5 and low urea cycle AA) at baseline predicted increases in metrics of insulin sensitivity (decreased TG/HDL and increased adiponectin) during lifestyle intervention. Conclusions Weight reduction in adolescents is associated with increases in BCAA catabolism and improvements in insulin sensitivity. Our study underscores the therapeutic potential of manipulating BCAA catabolism to treat obesity‐associated insulin resistance in adolescents and prevent progression to T2D.

Here, we investigated the effects of lifestyle intervention on BCAA catabolism and insulin sensitivity. We hypothesized (1) weight reduction and improved insulin sensitivity associate with enhanced BCAA catabolism; (2) baseline BCAAs and their metabolic by-products predict changes in weight and insulin sensitivity during lifestyle intervention.
Methods: A 33 adolescents with obesity were studied before and after 6 months of lifestyle intervention. Principal component analysis and multiple linear regression models were used to correlate changes in metabolic factors with changes in weight and insulin sensitivity assessed by HOMA-IR, adiponectin and ratio of triglyceride (TG) to HDL. Baseline metabolic factors were used as explanatory variables in prediction models.

| INTRODUC TI ON
Obesity and insulin resistance (IR) are the major risk factors for paediatric (and adult) type 2 diabetes (T2D) 1,2 ; however, only half of youth with obesity are insulin resistant and even fewer (2%-8%) progress to T2D. 3,4 Thus, obesity is neither sufficient nor sensitive for predicting who will become insulin resistant and metabolically unhealthy, 5 and who will develop T2D. Additional factors, such as disproportionate body fat distribution and/or increased liver fat content and visceral fat mass, are essential for progression to overt glucose intolerance. 6 Using state-of-the-art targeted metabolomic profiling, principal components analysis (PCA) and regression analysis, we previously showed that IR in adolescents with obesity is associated with a sexdependent metabolic 'signature' comprising the branched-chain amino acids (BCAAs), glutamate and C3/C5 acylcarnitines, implicating an altered flux through the BCAA catabolic pathway. 7 We called this metabolic signature 'BCAA-related factor', which in our PCA was Factor 2. Elevated levels of BCAA and the other components of PCA Factor 2 associate with insulin resistance in obese adults as well as adolescents. [7][8][9][10] As in adults, it is not clear if high BCAA levels in adolescents are a cause or a consequence of obesity and IR or both, [11][12][13] or if levels can be modified by dietary/exercise intervention. [13][14][15][16] However, recent work in rat and mouse models of obesity and IR demonstrates that enhanced catabolism of BCAA mediated by pharmacologic activation of the branched-chain α-keto acid dehydrogenase (BCKDH) improves insulin sensitivity while lowering circulating BCAAs and branch chain ketoacids (BCKAs). 17,18 Building on our novel findings, we investigated here the relationship between BCAA catabolism and insulin sensitivity during lifestyle intervention. We hypothesized that: (1) weight reduction and improved insulin sensitivity during intervention are associated with enhanced BCAA catabolism, evident by decreases in BCAAs and their metabolic by-products (PCA Factor 2); and (2) baseline BCAAs and their metabolic by-products predict subsequent changes in weight and insulin sensitivity. To test these hypotheses, we used targeted metabolomic profiling, PCA and multiple linear regression models to assess the correlations between changes in metabolic factors and changes in weight and insulin sensitivity as assessed by homeostasis model assessment index of insulin resistance (HOMA-IR), adiponectin and the ratio of triglyceride (TG) to HDL. In prediction models, metabolic factors at baseline were used as explanatory variables. We also stratified the data by sex to see whether there are differences that are not detected when data from both sexes are pooled. 19,20 2 | RE S E ARCH DE S I G N AND ME THODS

| The Duke Children's Healthy Lifestyles program
The Duke Children's Healthy Lifestyles programme (HLP) provides comprehensive clinical care for children and adolescents with overweight and obesity and represents the current standard of clinical care for paediatric obesity treatment. Treatment in the HLP uses motivational interviewing to modify dietary and activity behaviours in order to reduce the severity of overweight or obesity and obesityrelated comorbidities. 21 At the initial visit to the HLP (1 h), the medical provider meets with the family to review a comprehensive lifestyle, birth, medical, family and social history; conduct a physical examination focussed on obesity-related conditions; and screen for psychological and disordered eating concerns. The provider discusses the meaning of the patient's BMI (weight-for-length if the patient is <2 years old) as a function of age and sex and assesses the patient's risk for disease in the context of the described behaviours, the laboratory studies and the family history. In addition, the medical provider provides medical management of obesity-associated comorbidities, as necessary. Patients also meet with a registered dietician and paediatric physical therapist to complete a nutrition and fitness assessment, and recommendations are tailored to the individual family needs. Nutritional guidance follows standard recommendations as provided by the American Academy of Pediatrics. 22 Families are encouraged to follow-up with the Healthy Lifestyles team monthly for 1 year. The medical provider provides ongoing management of obesity-associated comorbidities and lifestyle modification (30 min), and the patient meets with a registered dietician for medical nutritional therapy (30 min). The focus of these visits is patient-centred goal setting around evidence-based lifestyle habits known to influence obesity such as reducing sugar-sweetened beverages and screen time, and increasing fruit and vegetable consumption and physical activity. Families in Healthy Lifestyles are also invited to activity sessions and cooking classes at a community predicted increases in metrics of insulin sensitivity (decreased TG/HDL and increased adiponectin) during lifestyle intervention.

Conclusions:
Weight reduction in adolescents is associated with increases in BCAA catabolism and improvements in insulin sensitivity. Our study underscores the therapeutic potential of manipulating BCAA catabolism to treat obesity-associated insulin resistance in adolescents and prevent progression to T2D.

K E Y W O R D S
BCAA, childhood obesity, insulin resistance recreation centre. These sessions are adapted for children with obesity and aim to provide hands-on experience preparing healthy foods and at least 45 min of moderate to vigorous physical activity.
Sessions are offered 6 days per week, and families are encouraged to attend at least 2 sessions per week.

| Patient cohort
Participants were identified prior to enrolment in Duke Children's Healthy Lifestyles Program and followed prospectively for

| Blood samples
Baseline and 6-month follow-up blood samples were obtained after an 8-to 12-h overnight fast. Plasma was stored at −80°C until analysed.

| Anthropometric measurements
Body weight and height were measured by standard methods. Blood pressure was measured twice; the average was used in statistical analyses. Age, sex and height-specific normal values for children are available at https://www.nhlbi.nih.gov/files/ docs/bp_child_pocket. pdf. Body fat percentage was estimated using a Tanita BC-148 segmental body composition analyzer. BMI, BMI percentile (BMI %), BMI z-score and the per cent BMI exceeding the 95 th %ile were calculated using the SAS programme (https://www.cdc.gov/nccdp hp/ dnpao/ growt hchar ts/resou rces/sas.htm). Our cohort consisted of subjects with extreme BMI values, in whom BMI percentile and BMI z-score provide unreliable estimates of the degree of overweight and the response to intervention. 23,24 Consequently, we used 'BMI per cent exceeding the 95 th percentile' for age and sex to track the weight changes over time.

| Plasma acylcarnitines and amino acids
A 45 Acylcarnitines (0.01-40 µmol/L, <15%) and 15 amino acids (5-1000 µmol/L, <15%) were analysed by tandem mass spectrometry (MS/MS) using a Waters TQD instrument. Amino acids and acylcarnitines were analysed by flow injection electrospray ionization tandem mass spectrometry and quantified by isotope or pseudo-isotope dilution using methods described previously. 26,27 Briefly, plasma samples were spiked with a cocktail of heavyisotope internal standards (Cambridge Isotope Laboratories; CDN Isotopes, Canada) and deproteinated with methanol. The methanol supernatants were dried and esterified with either acidified methanol or butanol for acylcarnitine or amino acid analysis, respectively. Mass spectra for acylcarnitine and amino acid esters were obtained using precursor ion and neutral loss scanning methods, respectively. The data were acquired using a Waters TQ (triple quadrupole) detector equipped with AcquityTM UPLC system and a data system controlled by MassLynx 4.1 operating system (Waters). Ion ratios of analyte to respective internal standard computed from centroided spectra were converted to concentrations using calibrators constructed from authentic aliphatic acylcarnitines and amino acids (Sigma; Larodan, Sweden) and dialysed Foetal Bovine Serum (Sigma).
Assays were run in a 96-well-plate format, with a calibration curve and a set of two QC samples at the beginning and end of each plate. Use of two independent QC samples enabled monitoring of intra-and interday precision of the assay. Over the course of several recent years, intra-and interday CV of acylcarnitine measurements in QC plasma were <15%, and <10% for abundant analytes, such as plasma amino acids and acetylcarnitine.

| Statistical analysis
Principal components analysis was used to reduce the large number of correlated metabolites into clusters of fewer components not correlated with each other. The metabolic factors and metabolites comprising these factors are described in our previous study. 7 Minimum sample size was calculated to detect correlations of 0.5 or greater between the 'BCAA-related factor' (PCA Factor 2) and insulin sensitivity. In linear regression models using 4 explanatory variables with one testing variable, a sample size of 32 provides a correlation of 0.5 with power of 0.8 and p < .05. Thus, our sample size of 33 provided adequate statistical power.
Metabolites measured at 6-month follow-up were scored using the PCA results obtained in the original study to construct the same factors. These metabolic factors served as explanatory variables.
HOMA-IR, adiponectin and TG /HDL ratio were natural log transformed to approximate normality. Multiple linear regression models were used to analyse the associations between changes in metabolic factors, changes in surrogate measures of IR and changes in weight. All models were adjusted for age, sex and BMI z. Models for change in insulin sensitivity were also adjusted for change in BMI% exceeding the 95 th percentile. Likewise, the model for change in BMI% exceeding the 95 th percentile was also adjusted for change in HOMA-IR. Statistically significant changes in factors were determined using stepwise linear regression analysis. To investigate if baseline factors predict subsequent changes in insulin sensitivity in response to lifestyle intervention, metabolic factors at baseline were used as explanatory variables in linear regression models.
Anthropometric values and metabolites related to BCAA catabolism across time were compared using paired t tests. Data were stratified by sex for analysing the effects separately for females and males.
Unpaired t tests were used to compare anthropometric values and metabolites related to BCAA catabolism among females and males at baseline and at 6-month follow-up. Baseline anthropometric values and metabolites among participants with and without follow-up were compared using unpaired t test to address selection bias (Table S1).
For all analyses, p < .05 was considered statistically significant; analyses were performed using SAS version 9.4 (SAS Institute Inc.).  surrogate measures of insulin sensitivity at follow-up were comparable to those at baseline (Table 1).

| Comparisons of anthropometric values and metabolites by sex
At baseline, males and females were comparable in age, weight and BMI-related metrics. At baseline and follow-up, males had higher levels of BCAAs (p = .0121) and Glutamate/Glutamine (p = .0186), consistent with our previous findings. 7 C3 (p = .0160) and C5 (p = .0246) acylcarnitine levels were also higher among males at 6 months.

| Association models
Associations between insulin sensitivity and BCAA-related factor (PCA Factor 2) at 6-month follow-up Significant metabolic factors identified in our original study served as explanatory variables in the follow-up regression analysis. Using the 6-month follow-up data, BCAA-related factor (PCA Factor 2) was again significantly associated with HOMA-IR (p = .0050) and the ratio of TG/HDL (p = .0344, Table 2). This is consistent with the results of our original study.  These findings implicate the urea cycle amino acids as novel markers of insulin sensitivity.

Associations between changes in weight and insulin sensitivity and changes in PCA factors
To determine whether associations between BCAA, urea cycle amino acids and insulin sensitivity are mediated by changes in weight, we adjusted models for change in insulin sensitivity for change in 'BMI% exceeding the 95 th percentile' (Table 3B) (Table 3B). This finding suggests that changes in TG/HDL are mediated by, or associated with, changes in BCAA and catabolic by-products independent of change in 'BMI% exceeding the 95th percentile'.

| Prediction models
Did baseline metabolic factors predict subsequent changes in weight or markers of insulin sensitivity? Table 4 provides the selected prediction models for the subsequent High circulating levels of BCAA in rodents and human adults with obesity are thought to reflect a decrease in BCAA catabolism in liver and adipose tissue [31][32][33] and/or an increase in BCAA production by an altered microbiome. 13 Using metabolomic profiling in plasma samples, we showed that IR in adolescents with obesity is associated with a similar metabolic signature comprising BCAAs, glutamate and C3/C5 acylcarnitines implicating an altered flux through the BCAA catabolic pathway. Interestingly, components of the metabolome were associated differentially with IR in teenage boys and girls: fasting BCAAs were higher in boys and correlated most strongly with HOMA-IR and adiponectin; in contrast, BCAAs in girls correlated most strongly with the TG/HDL ratio. 7 As in adults, it is not clear if higher BCAAs in adolescents are a cause or a consequence (or both) of obesity and/or IR or if levels can be modified by dietary/exercise intervention. [11][12][13][14][15][16] In this study, we investigated the changes in the metabolome of youth with obesity 6 months after enrolment into a lifestyle modification programme.
We assessed associations between changes in BCAAs and their  insulin sensitivity. 41 The TG/HDL ratio reflects the balance between TG intake (in the form of dietary chylomicrons), TG clearance by peripheral tissues, and TG synthesis and export from the liver. 42 Transfer of triglyceride from VLDL to HDL particles increases HDL clearance and thereby reduces plasma HDL. HOMA-IR is a measure of hepatic insulin sensitivity as it reflects fasting insulin and glucose levels. 25 Thus, the surrogate measures of IR reflect distinct, but overlapping, components of insulin sensitivity regulated at the level of the liver, adipose tissue and skeletal muscle. Given the differential regulation of BCAA catabolism in these tissues in obese states, [31][32][33] it may not be surprising that correlations between BCAA-related factor (PCA Factor 2) and urea cycle amino acids (PCA Factor 7) and HOMA-IR, adiponectin and TG/HDL varied in response to lifestyle intervention.
We do not yet know why those with highest baseline BCAAs and lowest urea cycle amino acids had the greatest increases in insulin sensitivity, as assessed by TG/HDL and adiponectin. We speculate that lifestyle intervention may have had its greatest impact in  44 In contrast, a smaller study of 80 youths found that BCAAs were negatively associated with HOMA-IR at baseline but not at follow-up. Other investigators found that increases in tyrosine preceded changes in BCAAs in obese children. 45  2-year follow-up. 46 Consistent with our findings, in a cohort of 396 Finnish girls, leucine and isoleucine levels in childhood were associated with and predicted in early adulthood the ratio of TG/HDL but not HOMA-IR or fat mass. 47 Wahl and colleagues found that lower serum concentrations of phosphatidylcholines and smaller waist circumference predicted weight loss. 48 Our study has several limitations. The sample size was small, in part because high attrition rates are common in studies of adolescents with obesity and IR. 21 Nevertheless, our findings provide novel insights into the effects of lifestyle intervention on branched-chain amino acid catabolism and insulin sensitivity in adolescents with obesity. BCAA levels in adolescents with obesity are sex-dependent, and reduction in 'BMI% exceeding the 95 th percentile' during lifestyle intervention is associated modestly with increases in BCAA catabolism: BCAA levels decline and the urea cycle amino acids rise in parallel with reduction in 'BMI% above the 95 th percentile' and an increase in insulin sensitivity. Assessment of BCAA catabolism may provide a useful metric for assessing the response to intervention in adolescents with obesity, and modulation of BCAA catabolism might provide new approaches for treating IR and preventing progression to T2D.

ACK N OWLED G M ENTS
We would like to thank our patients and their families, Emory We would like to thank Huaxia Cui for her technical assistance with immunoassays and conventional metabolites measurements. We also would like to thank Dr. Christopher Newgard for his help with interpretation of the data.

E TH I C A L A PPROVA L
The Institutional Review Board at Duke University approved the research protocol.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.