Preview

PDF, English - main document Download (49MB) | Terms of use

Abstract

In Correlative Light and Electron Microscopy (CLEM), two imaging modalities are combined to take advantage of the localization capabilities of light microscopy (LM) to guide the capture of high-resolution details in the electron microscope (EM). However, traditional approaches have proven to be very laborious, thus yielding a too low throughput for quantitative or exploratory studies of populations. Recently, in the electron microscopy field, FIB-SEM (Focused Ion Beam -Scanning Electron Microscope) tomography has emerged as a flexible method that enables semi-automated 3D volume acquisitions. During my thesis, I developed CLEMSite, a tool that takes advantage of the semi-automation and scanning capabilities of the FIB-SEM to automatically acquire volumes of adherent cultured cells. CLEMSite is a combination of computer vision and machine learning applications with a library for controlling the microscope ( product from a collaboration with Carl Zeiss GmbH and Fibics Inc.). Thanks to this, the microscope was able to automatically track, find and acquire cell regions previously identified in the light microscope. More specifically, two main modules were implemented. First, a correlation module was designed to detect and record reference points from a grid pattern present on the culture substrate in both modalities (LM and EM). Second, I designed a module that retrieves the regions of interest in the FIB-SEM and that drives the acquisition of image stacks between different targets in an unattended fashion. The automated CLEM approach is demonstrated on a project where 3D EM volumes are examined upon multiple siRNA treatments for knocking down genes involved in the morphogenesis of the Golgi apparatus. Additionally, the power of CLEM approaches using FIB-SEM is demonstrated with the detailed structural analysis of two events: the breakage of the nuclear envelope within constricted cells and an intriguing catastrophic DNA Damage Response in binucleated cells. Our results demonstrate that executing high throughput volume acquisition in electron microscopy is possible and that EM can provide incredible insights to guide new biological discoveries.