32048
High Resolution Mass Spectrometry for Biomarker and Risk Factor Discovery
Hypothesis-based epidemiologic study of candidate prenatal risk factors for autism spectrum disorder (ASD) commonly utilize exposure or exposure-response biomarkers obtained from biospecimens collected in the pre- or peri-natal window. When these targeted analyses utilize liquid chromatography-high-resolution mass spectrometry (LC-HRMS) on modern instruments, the picture of the molecular contents of the sample can often be expanded by semi-quantitative description of the wider chemical space (metabolomics/exposomics), which may facilitate the development of hypothesis-agnostic, discovery type analyses. Correlation of this wider metabolome/exposome with neurodevelopmental outcomes may help understand the role of environment and gene/environment interactions in risk of ASD, especially across cohorts.
Objectives:
We previously conducted targeted analyses for four exposure or exposure-response biomarkers using LC-HRMS in two different cohorts across 4 sample types and with 5 classes of targeted analytes using hybrid targeted/untargeted metabolomics. Our objective here is to mine the semi-quantitative data from these experiments to determine if we characterized the prenatal metabolome/exposome with overlap between different methods, sample types, and cohorts. Such overlap would increase the power of single or limited risk factor studies, since each targeted study could then serve as a replication of other studies that quantified the same molecules.
Methods:
In the EARLI cohort, we quantified the sex steroid and phthalate metabolite content by two separate methods in meconium samples (n=193) as well as metabolites of prostaglandin E2 from maternal urine (n= 547 across multiple gestational visits). In addition, in maternal serum (n=1002 across multiple gestational visits) and newborn blood spots (n=400) we quantified polyunsaturated fatty acids (PUFAs) from a case-control study built from the state of California registry and biobank. All experiments were conducted on a Q Exactive Plus high resolution mass spectrometer coupled to an Ultimate 3000 HPLC in the same laboratory by blinded analysts. After targeted analysis was conducted, data was re-interrogated with untargeted metabolomics pipelines.
Results:
Targeted assays with known performance characteristics provided quantitative abundance on select analytes. Simultaneously, semi-quantitative relative abundance was collected on around 10,000-14,000 features per experiment, with number of features depending on the biosample and method of analysis. Putative identification was conducted by database searching. Confirmatory identification for select analytes was conducted by LC-MS/HRMS, and elution time matching with an authentic standard. Some targeted analytes, including PUFAs, we detected by other analytical methods in other sample types. Other molecules, including unconjugated testosterone, were only detected in targeted analysis. This allowed us to compare findings across some, but not all, studies.
Conclusions:
Mixed targeted/untargeted analytical methods can provide data for both hypothesis testing and hypothesis generating from limited biospecimens from epidemiological studies of risk factors for ASD. Untargeted data can potentially, but not necessarily, be compared across experiments.