Work Description

Date: 26 April, 2020

Dataset Title: Mitochondrial nutrient utilization underlying the association between metabolites and insulin resistance in adolescents [Supplemental materials]

Dataset Creators:
LaBarre, Jennifer L.
Peterson, Karen E.
Kachman, Maureen T.
Perng, Wei.
Tang, Lu.
Hao, Wei.
Zhou, Ling.
Karnovsky, Alla.
Cantoral, Alejandra.
Téllez-Rojo, Martha María.
Song, Peter XK.
Burant, Charles F.

Dataset Contact:
Burant, Charles F. burantc@med.umich.edu

Research Overview:
Objectives. Untargeted metabolomics was used to identify metabolites associated with metabolic health in adolescents. Path analysis classified how habitual dietary intake influences BMI z-score (BMIz) and insulin resistance (IR) through changes in the metabolome.

Study Population. Participants were enrolled in the Early Life Exposure in Mexico to ENvironmental Toxicants (ELEMENT) project which was started in 1994 and consists of three sequentially-enrolled birth cohorts from Mexico City Maternity Hospitals. A subset of these children, age 8-14 years, were contacted through their primary caregiver to provide urine samples, serum samples, anthropometry and complete an interview-based questionnaire (n=250). Subjects for this analysis have baseline and follow data on anthropometry, metabolic biomarkers and adequate serum volume for metabolomics analyses (n=206).

Methodology:
Metabolomic Profiling. Fasting serum samples were analyzed using an untargeted metabolomics platform on an Agilent 1200 HPLC/6530 quadrupole Time-of-Flight mass spectrometry (MS) system (Agilent Technologies, Inc., Santa Clara, CA USA) using the Waters Acquity HSS T3 1.8 μm column (Waters Corporation, Milford, MA). The eluent was analyzed in both positive and negative ion mode electrospray ionization. Chromatographic peaks that represent features were detected using a modified version of existing commercial software (Agilent MassHunter Qualitative Analysis). Data normalization used “pooled” reference samples that were analyzed repeatedly throughout each batch. The incorporation of Binner, a new method to visualize closely eluting features, allowed for the removal of redundant features. Missing peak intensities were imputed by K-nearest neighbor (K = 5) in features with at least 70% detection across samples. Features with less than 70% detection across samples were removed. Annotated metabolites were identified via comparing their MS/MS spectra to authentic standards, purchased internal or external standards ran on the same instrument. All lipids will be mentioned with the nomenclature as X:Y, where X is the length of the carbon chain and Y is the number of double bonds. The final metabolomics dataset contained 550 annotated metabolites and 2722 unannotated metabolites.

Anthropometry and Metabolic Biomarkers. Anthropometry was obtained using standardized procedures by trained personnel following Lohman Methodology. Weight and height were used to calculate BMI (kg/m2) and participants were classified as obese (BMI z-score>2) and normal weight (-2 2) vs. normal weight (-20.2 (Pearson’s correlation) and a minimum of 5 metabolites within, resulting in 34 clusters and 110 singleton metabolites without clusters. A cluster score, ‘f1’, is created by a weighted linear sum of metabolites with weights of factor loadings of metabolites within the cluster. Step 5. In the path analysis, we identified beta coefficients describing the relationship between (1) energy-adjusted fat and carbohydrate intake (mean 0, std. deviation 1) with metabolome clusters (β1) and (2) metabolome clusters with BMIz and standardized HOMA-CP (β2), adjusting for sex, age, and puberty onset. Our analysis is composed of two parts: (1) a structural part linking the metabolome to BMIz or HOMA-CP through a linear model such as BMIz = β0 + f1 * β1 + puberty onset * β2 + age * β3 + sex * β4 + e1 and (2) a confirmatory factor analysis model via another linear model such as f1 = γ0 + Fat * γ1 + puberty onset * γ2 + age * γ3 + sex* γ4 + e2. All parameters, including loadings, β values and γ parameters are estimated via a maximum likelihood estimation method using the R package “lavaan”. Step 6. Our hypothesis is that a person’s intrinsic metabolism, reflected in the metabolome, influences the consequence of macronutrient intake on BMIz and IR. To assess this relationship, we used path analysis, a multiple regression technique used to examine causal relationships between variables. The path effect of an independent variable on a dependent variable is expressed as products of regression parameters from the two models. To obtain confidence intervals for the path effect, Sobel testing was used, which is known as being a conservative but widely used approach for assessing the significance of a path.

Instrument and/or Software specifications:
Statistical analyses were performed using SAS 9.4 and R, Version 3.6.2. Heatmaps were created using an in-house software, CoolMap. Figures were created using GraphPad Prism version 7.4 (La Jolla, California).

Files contained here:
Online_SuppMaterial_Abbreviations.pdf
Online_SuppMaterial_Figure_S1_S5.pdf
Online_SuppMaterial_TableS1.xlsx
Online_SuppMaterial_TableS2.xlsx
Online_SuppMaterial_TableS3.xlsx
Online_SuppMaterial_TableS4.xlsx
Online_SuppMaterial_TableS5.xlsx
Online_SuppMaterial_TableS6.xlsx
Online_SuppMaterial_TableS7.xlsx
Online_SuppMaterial_TableS8.xlsx

Use and Access:
This data set is made available under a Creative Commons Attribution-Noncommercial license (CC-BY-NC 4.0)

To Cite Data:
LaBarre, J., Peterson, K., Kachman, M., Perng, W., Tang, L., Hao, W., Zhou, L., Karnovsky, A., Cantoral, A., Téllez-Rojo, M., Song, P., Burant, C. Mitochondrial nutrient utilization underlying the association between metabolites and insulin resistance in adolescents [Data set]. University of Michigan - Deep Blue. https://doi.org/10.7302/00cs-x605