Research Article: Automated classification of synaptic vesicles in electron tomograms of C. elegans using machine learning

Date Published: October 8, 2018

Publisher: Public Library of Science

Author(s): Kristin Verena Kaltdorf, Maria Theiss, Sebastian Matthias Markert, Mei Zhen, Thomas Dandekar, Christian Stigloher, Philip Kollmannsberger, Jianhua Xu.

http://doi.org/10.1371/journal.pone.0205348

Abstract

Synaptic vesicles (SVs) are a key component of neuronal signaling and fulfil different roles depending on their composition. In electron micrograms of neurites, two types of vesicles can be distinguished by morphological criteria, the classical “clear core” vesicles (CCV) and the typically larger “dense core” vesicles (DCV), with differences in electron density due to their diverse cargos. Compared to CCVs, the precise function of DCVs is less defined. DCVs are known to store neuropeptides, which function as neuronal messengers and modulators [1]. In C. elegans, they play a role in locomotion, dauer formation, egg-laying, and mechano- and chemosensation [2]. Another type of DCVs, also referred to as granulated vesicles, are known to transport Bassoon, Piccolo and further constituents of the presynaptic density in the center of the active zone (AZ), and therefore are important for synaptogenesis [3].

Partial Text

Caenorhabditis elegans is a well-studied model organism. Its small nervous system, consisting of 302 neurons [4,5] allows for studies of the connectome in its entirety [5–7]. A focus of research is on C. elegans young adult hermaphrodites, which makes them a reference system for comparison with other developmental states in the life cycle. The most remarkable of all C. elegans developmental states is the dauer larva.

Automated methods for image analysis are of great importance to reduce manual labor in the long term and prevent biased analysis of data, hence making double blind analysis unnecessary. This is confirmed by our own experience, where one re-analysis of the same tomogram by the same expert revealed one differential interpretation of CCV and DCV and 6 other non-matching assignments on n = 134 vesicles in total. Such deviations in manual analysis may have a great influence on the outcome. For this reason, we had two experts manually assign all vesicles independently. In case of inconsistency, both persons came to a mutual agreement or consulted a third expert in the decision process. Still unassignable vesicles were excluded from analysis. Since this evaluation process is very elaborate and time-consuming, a long-term automated solution is inevitable. As far as we know there is no other approach for the specific task of automated vesicle classification in electron tomograms. We present here a new methodological approach that is applied subsequent to a modified version of our previously published Fiji macro 3D ART VeSElecT [22]. First, a Fiji macro was written to read out important characteristics. Those were, together with manual labelling, used to train machine learning classifiers in python. Extracted machine learning parameters of an SVM completed the new Fiji macro to classify vesicles and visualize classes through 3D color labelling. Our tool can thus be used for automated vesicle classification. We used our ground truth label to identify differences between two types of vesicles in two different developmental states of C. elegans, young adult hermaphrodites and dauer larvae. We then use these data to train an automated classification workflow to show that resulted algorithm can reproduce these manually obtained biological results.

We developed a new automated approach to classify vesicles and to quantify their properties from electron tomograms. We combined machine learning with an extension of our previously developed vesicle segmentation workflow 3D ART VeSElecT to reliably distinguish clear core from dense core vesicles using image-based features. We apply this method to electron tomograms of C. elegans NMJs in young adult hermaphrodites and dauer larvae. Using our ground truth data, we find an increased fraction of dense core vesicles (~ 16% vs ~ 9%) in tomograms matching certain characteristics (cholinergic synapses with > 100 vesicles of a central slice through the AZ area), as well as significantly reduced vesicle size and a higher distance of dense core vesicles to the active zone in dauer larvae compared to young adult hermaphrodites. Our approach is not limited to C. elegans and can be easily adapted to study differences in synaptic and other vesicles in electron tomograms in various settings and many other model organisms, e.g. in Danio rerio [22].

 

Source:

http://doi.org/10.1371/journal.pone.0205348

 

0 0 vote
Article Rating
Subscribe
Notify of
guest
0 Comments
Inline Feedbacks
View all comments