Studying autism using untargeted metabolomics in newborn screening samples

BackgroundThe etiopathology of autism spectrum disorder (ASD) is unclear. Main risk factors include both genetic and non-genetic factors, especially prenatal and perinatal events. The Danish Neonatal Screening Biobank in connection with registry data provides unique opportunities to study early signs of disease. Therefore, we aimed to study the metabolomic profiles of dried blood spot (DBS) of newborns later diagnosed with ASD. MethodsFrom the iPsych cohort, we randomly selected 37 subjects born in 2005 and diagnosed with ASD in 2012 (cases) together with 37 matched controls and submitted their biobanked DBS to an LC-MS/MS-based untargeted metabolomics protocol. Raw data were preprocessed using MZmine 2.41.2 and metabolites were subsequently putatively annotated using mzCloud, GNPS feature-based molecular networking and other metabolome mining tools (MolNetEnhancer). Statistical analyses and data visualization included principal coordinates analyses, PERMANOVAs, t-tests, and fold-change analyses. Results4360 mass spectral features were detected, of which 150 could be putatively annotated at a high confidence level. Chemical structure information at a broad level could be retrieved for a total of 1009 metabolites, covering 31 chemical classes including bile acids, various lipids, nucleotides, amino acids, acylcarnitines and steroids. Although the untargeted analysis revealed no clear distinction between cases and controls, 18 compounds repeatedly reported in the ASD literature could be detected in our study and three mass spectral features were found differentially abundant in cases and controls before FDR correction. In addition, our results pinpointed important other factors influencing chemical profiles of newborn DBS samples such as gestational age, age at sampling and month of birth. LimitationsInherent to pilot studies, our sample size was insufficient to reveal metabolic markers of ASD. Nevertheless, we were able to establish an efficient metabolomic data acquisition and analysis pipeline and flag main confounders to be considered in future studies. ConclusionsIn this first untargeted DBS metabolomic study, newborns later diagnosed with ASD did not show a significantly different metabolic profile when compared to controls. Nevertheless, our method covered many metabolites associated with ASD in previous studies, suggesting that biochemical markers of ASD are present at birth and may be monitored during newborn screening.