Environmental High-content Fluorescence Microscopy (e-HCFM) of Tara Oceans Samples Provides a View of Global Ocean Protist Biodiversity
Environmental High-content Fluorescence Microscopy (e-HCFM) of Tara Oceans Samples Provides a View of Global Ocean Protist Biodiversity
Abstract:
Protists are responsible for much of the diversity in the eukaryotic kingdom
and are crucial to several biogeochemical processes of global importance (e.g.,
the carbon cycle). Recent global investigations of these organisms have relied
on sequence-based approaches. These methods do not, however, capture the
complex functional morphology of these organisms nor can they typically capture
phenomena such as interactions (except indirectly through statistical means).
Direct imaging of these organisms, can therefore provide a valuable complement
to sequencing and, when performed quantitatively, provide measures of
structures and interaction patterns which can then be related back to sequence
based measurements. Towards this end, we developed a framework, environmental
high-content fluorescence microscopy (e-HCFM) which can be applied to
environmental samples composed of mixed communities. This strategy is based on
general purposes dyes that stain major structures in eukaryotes. Samples are
imaged using scanning confocal microscopy, resulting in a three-dimensional
image-stack. High-throughput can be achieved using automated microscopy and
computational analysis. Standard bioimage informatics segmentation methods
combined with feature computation and machine learning results in automatic
taxonomic assignments to the objects that are imaged in addition to several
biochemically relevant measurements (such as biovolumes, fluorescence
estimates) per organism. We provide results on 174 image acquisition from Tara
Ocean samples, which cover organisms from 5 to 180 microns (82 samples in the
5-20 fraction, 96 in the 20-180 fraction). We show a validation of the approach
both on technical grounds (demonstrating the high accuracy of automated
classification) and provide results obtain from image analysis and from
integrating with other data, such as associated environmental parameters
measured in situ as well as perspectives on integration with sequence
information.
and are crucial to several biogeochemical processes of global importance (e.g.,
the carbon cycle). Recent global investigations of these organisms have relied
on sequence-based approaches. These methods do not, however, capture the
complex functional morphology of these organisms nor can they typically capture
phenomena such as interactions (except indirectly through statistical means).
Direct imaging of these organisms, can therefore provide a valuable complement
to sequencing and, when performed quantitatively, provide measures of
structures and interaction patterns which can then be related back to sequence
based measurements. Towards this end, we developed a framework, environmental
high-content fluorescence microscopy (e-HCFM) which can be applied to
environmental samples composed of mixed communities. This strategy is based on
general purposes dyes that stain major structures in eukaryotes. Samples are
imaged using scanning confocal microscopy, resulting in a three-dimensional
image-stack. High-throughput can be achieved using automated microscopy and
computational analysis. Standard bioimage informatics segmentation methods
combined with feature computation and machine learning results in automatic
taxonomic assignments to the objects that are imaged in addition to several
biochemically relevant measurements (such as biovolumes, fluorescence
estimates) per organism. We provide results on 174 image acquisition from Tara
Ocean samples, which cover organisms from 5 to 180 microns (82 samples in the
5-20 fraction, 96 in the 20-180 fraction). We show a validation of the approach
both on technical grounds (demonstrating the high accuracy of automated
classification) and provide results obtain from image analysis and from
integrating with other data, such as associated environmental parameters
measured in situ as well as perspectives on integration with sequence
information.