Virtual correlative light microscopy-EM Time Series of C. elegans: Automated Single Cell Identification in embryos during development

Abstract number
European Microscopy Congress 2020
Corresponding Email
[email protected]
LSA.6 - Applications of correlative microscopy of biological systems
Dr Irina Kolotuev (2), Dr Anthony Santella (1, 3), Dr Zhirong Bao (1, 3)
1. Developmental Biology Program, Sloan Kettering Cancer Center
2. Electron Microscopy Facility Université de Lausanne
3. Molecular Cytology Core, Sloan Kettering Cancer Center

C. elegans, Correlative microscopy, 3D imaging, development

Abstract text

The nematode C. elegans was the first metazoan whose three dimensional structure was completely deciphered in 1979 using serial section EM analysis that revealed that the adult worm has precisely 959 somatic cells. Ever since, the EM analysis of C. elegans has continued, providing an impressive body of biological information as well as facilitating the development of 3D-EM in general via reconstructions from tomography and serial sections. The bottleneck at the moment in 3D- EM analysis is not data acquisition but its rendering and interpretation, although public domain websites such as ‘Worm Atlas’ and ‘Worm Image’ provide an excellent asset for the research community.

Though C. elegans anatomy is relatively simple, many of its organ systems are represented by dozens of cells, which can be complicated to localize at the EM level. The situation is even more challenging when the multiple embryonic stages are considered: The development of the embryo is relatively fast; extensive morphological changes occur between 300-500 min after the first cleavage of the fertilized egg. Remarkably, C. elegans exhibits a programmed cell division pattern, and cells in different animals at the same time point in development have near identical morphology and position.  This lineage has been tracked precisely in previous studies with cells identified based on the invariance of division events. Significant light microscopy (LM) data for wild type and mutant embryos exist. Yet, many questions remain unresolved because of the complicated ultrastructure of these stages. Moreover, the dimensions of the embryos are tiny and each sectioning orientation gives a different image, making it difficult to interpret sections when not linked to the context of the entire embryo.  


Here, we present a robust, indirect CLEM method for classifying individual cells of C.elegans embryos irrespective of their sectioning orientation and of embryonic stage. We start with time-lapse fluorescence microscopy to provide an atlas or a reference map of all cells in all developmental stages. This LM level information is then correlated to ultrastructural level information in serial sections obtained with Focused Ion Beam SEM or using the serial array tomography method.  The precise coordinates of each nucleus in any developmental time point is mapped from LM data to EM using an explicit model of expected cell-cell contacts and cell division timing. This yields a predicted identity for each cell in the EM data set. We validated the accuracy of this method by comparison to a manual ground truth for a subset of cells and found that the predicted identity of the cells were around 90 percent accurate irrespective of their developmental stages. For validation, ground truth identities were independently and manually assigned, based on position and cell morphology, to a subset of cells in our EM data sets. 


Based on this method we have begun the construction of a developmental time series spanning six different embryonic stages. Automated cell identification enables the targeted reconstruction of individual cells and organs to map the sequence of developmental changes, such as cell shape, migration, junctions in systems like the pharynx, muscles, neurons, etc. For the datasets obtained by the Array tomography method, we have applied a two-step analysis: first, the overall embryo stacks were generated with lower resolution acquisition parameters, and single-cell identities were assigned. Subsequently, based of identities micrographs of a precise region of interest were generated to monitor the desired cellular events, such as synaptic connections or cell junctions remodeling in a high-resolution mode.

In addition to providing an extensive database of C. elegans developmental information at the ultrastructural level, this approach has the potential to be applied to other organisms where establishing single-cell alignment between individuals is critical for understanding the extent of single-cell level identity.