Assessment of metabolic phenotypic variability in children's urine using 1H NMR spectroscopy

Abstract

The application of metabolic phenotyping in clinical and epidemiological studies is limited by a poor understanding of inter-individual, intra-individual and temporal variability in metabolic phenotypes. Using ¹H NMR spectroscopy we characterised short-term variability in urinary metabolites measured from 20 children aged 8-9 years old. Daily spot morning, night-time and pooled (50:50 morning and night-time) urine samples across six days (18 samples per child) were analysed, and 44 metabolites quantified. Intraclass correlation coefficients (ICC) and mixed effect models were applied to assess the reproducibility and biological variance of metabolic phenotypes. Excellent analytical reproducibility and precision was demonstrated for the ¹H NMR spectroscopic platform (median CV 7.2%). Pooled samples captured the best inter-individual variability with an ICC of 0.40 (median). Trimethylamine, N-acetyl neuraminic acid, 3-hydroxyisobutyrate, 3-hydroxybutyrate/3-aminoisobutyrate, tyrosine, valine and 3-hydroxyisovalerate exhibited the highest stability with over 50% of variance specific to the child. The pooled sample was shown to capture the most inter-individual variance in the metabolic phenotype, which is of importance for molecular epidemiology study design. A substantial proportion of the variation in the urinary metabolome of children is specific to the individual, underlining the potential of such data to inform clinical and exposome studies conducted early in life.

Figure 1

A typical ¹H NMR spectrum of urine from an 8 year old male child collected in the morning (A), night-time (B) and pooled (C; 50:50 morning and night-time) with identified metabolites. Abbreviations: 2-HB, 2-hydroxybutyrate; 3-HB, 3-hydroxybutyrate; 3-HIV, 3-hydroxyvalerate, 3IS, 3-indoxylsulfate; 4-DEA, 4-deoxyerythronic acid; 4-DTA, 4-deoxythreonic acid; NAG, N-acetyl glycoprotein fragments; NAN, N-acetylneuraminic acid; TMAO, trimethylamine-N-oxide.

Figure 2. Short term variability over six days for 44 metabolites in morning, night-time and pooled urine samples based on intra-class correlation coefficients (ICCs) measured in 20 children by ¹H NMR spectroscopy.

Each child was sampled twice daily over a period of one week (morning and night-time).

Figure 3. Distribution of urinary trimethylamine-N-oxide TMAO (top panel) and trimethylamine (bottom panel) based on ¹H NMR spectra of urine samples obtained from 20 children across morning, night-time and pooled samples (50:50 morning and night-time samples).

Metabolite integrals were log transformed. TMAO, Trimethylamine-N-oxide. A.U. arbitrary units.

Figure 4. Decomposition of variance for each annotated metabolite resonance in ¹H NMR spectra.

The plot displays estimates for the proportion of biological variance explained by child characteristic (black) and diurnal (yellow) components. The remainder of the variance is attributed to day-to-day and technical variability and unknown sources (residual, brown). Metabolites are ordered by estimated child specific variance. *Multiple overlapping resonances. Integral assigned to the most likely, abundant metabolite.