Comprehensive monitoring of tissue composition using in vivo imaging of cell nuclei and deep learning

Abstract

Comprehensive analysis of tissue composition has so far been limited to ex vivo approaches. Here, we introduce NuClear (Nucleus-instructed tissue composition using deep learning), an approach combining in vivo two-photon imaging of histone 2B-eGFP-labeled cell nuclei with subsequent deep learning-based identification of cell types from structural features of the respective cell nuclei. This allowed us to classify all cells per imaging volume (0.25 mm3 containing ~25000 cells) and identify their position in 3D space in a non-invasive manner using only a single label. NuClear opens a window to study changes in relative abundance and location of all major brain cell types in individual mice over extended time periods, enabling comprehensive studies of changes in cellular composition in physiological and pathophysiological conditions.

Loss of secondary motor cortex neurons in chronic neuropathic pain

“…1B). Tissue composition was determined by imaging Histon2B-GFP-labelled nuclei (Asan et al 2021) and subsequent classification of the cell type derived from specific features of the nucleus (Das-Gupta et al 2022)(Das-Gupta et al 2022). Subcellular changes were investigated by repetitively imaging individually labeled or small groups of labeled neurons in their entirety.…”

“…2B). Using the NuClear approach (Das-Gupta et al 2022), a method allowing for non-biased neuronal network-based identification of all major cell types residing within the imaging volume solely relying on a set of radiomic features of the cell nucleus (Das-Gupta et al 2022)we determined 3D matrices of the localization of all neurons, astroglia, oligodendroglia, microglia and endothelial cells present in the volume (Fig. 2C).…”

“…NuClear analysis for identification of cell types belonging to individual H2B nuclei Details of this method are published elsewhere (Das-Gupta et al 2022). Briefly, automated segmentation of all nuclei was performed with the StarDist neuronal network using a high performance cluster environment (bwForCluster MLS&WISO).…”

“…Hence, to address structural mechanisms underlying chronic neuropathic pain states, monitoring the cellular composition of cortical pain areas after induction of chronic pain is a promising avenue to follow. We recently developed an approach called "NuClear" that allows longitudinal, comprehensive, monitoring of cell type-composition in living animals (Das-Gupta et al 2022).…”

“…Hence, to address structural mechanisms underlying chronic neuropathic pain states, monitoring the cellular composition of cortical pain areas after induction of chronic pain is a promising avenue to follow. We recently developed an approach called “NuClear” that allows longitudinal, comprehensive, monitoring of cell type-composition in living animals (Das-Gupta et al 2022). The resulting 3D matrix of identified cells and their location can be repetitively acquired from the identical cortical area and changes in number and location of any cell type can be identified, for example after establishing chronic neuropathic pain using the spared-nerve injury (SNI) model (Decosterd and Woolf 2000).…”