Hair metabolomics: identification of fetal compromise provides proof of concept for biomarker discovery

Abstract

Analysis of the human metabolome has yielded valuable insights into health, disease and toxicity. However, the metabolic profile of complex biological fluids such as blood is highly dynamic and this has limited the discovery of robust biomarkers. Hair grows relatively slowly, and both endogenous compounds and environmental exposures are incorporated from blood into hair during growth, which reflects the average chemical composition over several months. We used hair samples to study the metabolite profiles of women with pregnancies complicated by fetal growth restriction (FGR) and healthy matched controls. We report the use of GC-MS metabolite profiling of hair samples for biomarker discovery. Unsupervised statistical analysis showed complete discrimination of FGR from controls based on hair composition alone. A predictive model combining 5 metabolites produced an area under the receiver-operating curve of 0.998. This is the first study of the metabolome of human hair and demonstrates that this biological material contains robust biomarkers, which may lead to the development of a sensitive diagnostic tool for FGR, and perhaps more importantly, to stable biomarkers for a range of other diseases.

Keywords: GC-MS.; biomarker; fetal growth restriction; hair; metabolite profiling.

Figure 1

The ratio of identified metabolite levels in maternal hair samples between Fetal Growth Restriction (FGR) case and control groups. Red circles () indicate metabolite abundance in hair from women identified as FGR cases relative to controls (blue triangles ), plotted using a log₂ scale. Only the metabolites generating statistically significant scores (p <0.01) are presented.

Figure 2

A. Principal Component Analysis (PCA), PC1 vs. PC2. Score plot shows separation between Fetal Growth Restriction (FGR) Cases (red) and Controls (green) based on metabolites abundance significantly different (p<0.01) between the groups. B. Multivariate ROC curve, based on metabolites which were statistically different between Control and FGR Case groups (p<0.01). Var. designated as number of discriminating metabolites (lactate, levulinate, 2-methyloctadecanate, tyrosine, margarate; significance to the model decreasing as listed) included in the AUC (Area Under the Curve) calculations.