The dynamic trophic architecture of open-ocean protist communities revealed through machine-guided metatranscriptomics

Proc Natl Acad Sci U S A. 2022 Feb 15;119(7):e2100916119. doi: 10.1073/pnas.2100916119.


Intricate networks of single-celled eukaryotes (protists) dominate carbon flow in the ocean. Their growth, demise, and interactions with other microorganisms drive the fluxes of biogeochemical elements through marine ecosystems. Mixotrophic protists are capable of both photosynthesis and ingestion of prey and are dominant components of open-ocean planktonic communities. Yet the role of mixotrophs in elemental cycling is obscured by their capacity to act as primary producers or heterotrophic consumers depending on factors that remain largely uncharacterized. Here, we develop and apply a machine learning model that predicts the in situ trophic mode of aquatic protists based on their patterns of gene expression. This approach leverages a public collection of protist transcriptomes as a training set to identify a subset of gene families whose transcriptional profiles predict trophic mode. We applied our model to nearly 100 metatranscriptomes obtained during two oceanographic cruises in the North Pacific and found community-level and population-specific evidence that abundant open-ocean mixotrophic populations shift their predominant mode of nutrient and carbon acquisition in response to natural gradients in nutrient supply and sea surface temperature. Metatranscriptomic data from ship-board incubation experiments revealed that abundant mixotrophic prymnesiophytes from the oligotrophic North Pacific subtropical gyre rapidly remodeled their transcriptome to enhance photosynthesis when supplied with limiting nutrients. Coupling this approach with experiments designed to reveal the mechanisms driving mixotroph physiology provides an avenue toward understanding the ecology of mixotrophy in the natural environment.

Keywords: machine learning; metatranscriptomics; microbial ecology; mixotrophy; trophic mode.

MeSH terms

  • Eukaryota / genetics
  • Eukaryota / physiology*
  • Food Chain*
  • Gene Expression Profiling
  • Machine Learning*
  • Models, Biological*
  • Oceans and Seas
  • Plankton / genetics
  • Plankton / physiology*