Studies in cultured cells have demonstrated the existence of higher-order epigenetic mechanisms, determining the relationship between expression of the gene and its position within the cell nucleus. It is unknown, whether such mechanisms operate in postmitotic, highly differentiated cell types, such as neurons in vivo. Accordingly, we examined whether the intranuclear positions of Bdnf and Trkb genes, encoding the major neurotrophin and its receptor respectively, change as a result of neuronal activity, and what functional consequences such movements may have. In a rat model of massive neuronal activation upon kainate-induced seizures we found that elevated neuronal expression of Bdnf is associated with its detachment from the nuclear lamina, and translocation toward the nucleus center. In contrast, the position of stably expressed Trkb remains unchanged after seizures. Our study demonstrates that activation-dependent architectural remodeling of the neuronal cell nucleus in vivo contributes to activity-dependent changes in gene expression in the brain.

The detailed architectural examination of the neuronal nuclei in any brain region, using confocal microscopy, requires quantification of fluorescent signals in three-dimensional stacks of confocal images. An essential prerequisite to any quantification is the segmentation of the nuclei which are typically tightly packed in the tissue, the extreme being the hippocampal dentate gyrus (DG), in which nuclei frequently appear to overlap due to limitations in microscope resolution. Segmentation in DG is a challenging task due to the presence of a significant amount of image artifacts and densely packed nuclei. Accordingly, we established an algorithm based on continuous boundary tracing criterion aiming to reconstruct the nucleus surface and to separate the adjacent nuclei. The presented algorithm neither uses a pre-built nucleus model, nor performs image thresholding, which makes it robust against variations in image intensity and poor contrast. Further, the reconstructed surface is used to study morphology and spatial arrangement of the nuclear interior. The presented method is generally dedicated to segmentation of crowded, overlapping objects in 3D space. In particular, it allows us to study quantitatively the architecture of the neuronal nucleus using confocal-microscopic approach.

No abstract

This paper presents a system used for SemEval-2021 Task 5: Toxic Spans Detection. Our system is an ensemble of BERT-based models for binary word classification, trained on a dataset extended by toxic comments modified and generated by two language models. For the toxic word classification, the prediction threshold value was optimized separately for every comment, in order to maximize the expected F1 value.