Challenges and rewards of nontargeted studies of marine dissolved organic matter (DOM) composition using liquid chromatography coupled to tandem mass spectrometry (LC MS/MS): Case studies from coastal, marine environments.

Lihini Aluwihare1, Daniel Petras1, Irina Koester2, Louis-Félix Nothias3, Marcus Ludwig4, Kai Duhrkop4, Mingxun Wang3, Sebastian Böcker4 and Pieter Dorrestein3, (1)University of California San Diego, Scripps Institution of Oceanography, La Jolla, United States, (2)University of California San Diego, Scripps Institution of Oceanography, La Jolla, CA, United States, (3)University of California San Diego, Collaborative Mass Spectrometry Innovation Center, La Jolla, CA, United States, (4)Friedrich Schiller University of Jena, Chair for Bioinformatics, Jena, Germany
Abstract:
The complexity of DOM has challenged comprehensive analysis of its composition, yet, composition must be understood to accurately elucidate the role of DOM in structuring microbial communities and in mediating carbon and nutrient cycling. Therefore, multiple approaches to isolating and characterizing DOM, even selective techniques, can provide valuable insights into DOM biogeochemistry. We have applied tandem mass spectrometry (MS/MS) coupled to liquid chromatography (LC) in a nontargeted approach, to study marine DOM extracted using PPL resin. Our extraction method isolates <40% of the total DOC reservoir and the analysis is limited by our MS ionization method – electrospray. Yet, we still recover 1000s of features per sample, some of which correlate with community composition or physical and chemical features. Therefore, relevant and important information is contained in the compounds detected within this analytical window. As such, we have been refining downstream data analysis techniques to maximize the molecular-level information that can be gained from this dataset. Here, we will use samples collected in marine, coastal waters to demonstrate how comprehensive data processing of multiple samples has enabled us to refine molecular-level identifications. These implementations include feature detection, consensus spectra generation, ion identity and molecular networking with analog search (GNPS), as well as multiple molecular formula assignment (SIRIUS, ZODIAC) and machine learning tools (CSIFinger ID and CANOPUS). When tested with reference spectra, our data analysis pipeline performs significantly better than traditional nontargeted LC MS/MS methods. As expected, our ability to increase the assignment of chemical structures to detected features in marine DOM has also improved. In addition, capitalizing on the Lagrangian approach used to acquire some DOM samples, together with molecular networking enabled clustering of related molecules, we are able to postulate potential chemical transformations within an aging water parcel. As we will show with examples, many of the identified compounds appear to be involved in structuring marine microbial interactions.