Cellular 3D-reconstruction and analysis in the human cerebral cortex using automatic serial sections
Life science aims at a better understanding of multiple biological functions, such as healthy organ development with cellular proliferation, migration and organization, tumor formation, and general pathology. Several techniques have been developed to study biological structure in 3D like serial block-face scanning electron microscopy (EM), focused ion beam scanning EM, serial-section transmission EM, automatic tape-collecting ultramicrotome-scanning EM, and many types of light microscopy with or without tissue clearing1,2,3,4. In life science, EM is widely used to explore the subcellular components, which are several orders of magnitude smaller than the spatial neuron networks. Clearing techniques can be very helpful in attempting to explain neuron networks in greater brain volumes, such as the adult mouse brain, in an entire state without disassembly5,6. However, some of the disadvantages for tissue clearance are that immunostaining of archival tissues is usually complicated as a result of the antigen masking due to formaldehyde protein cross-linking7. Furthermore, practicing immunostaining with tissue clearing remains difficult in human tissues due to factors such as inadequate antibody penetration depth, physicochemical properties, and tissue composition7,8.
A typical human neuron has thousands of complex connections with neighboring neurons, which is essential for normal function, yet the organization of these neurons is still under debate9. The cellular organization in the human neocortex has been described as a local network of vertical columns containing neurons. Neurons with similar functions are grouped together and according to different theories, these cortical columns may contain smaller columns known as minicolumns, which are the smallest unit capable of processing information in the cerebral cortex10,11. Cortical columns are radially oriented cell bodies that span through the laminar pattern perpendicular to the pial surface and can be seen using regular Nissl preparations or another cell body-revealing histological procedures. The introduction of minicolumns was in response to studies of the patterning of apical dendrites of pyramidal cells with somata situated in layers II, III, and V12,13. Studies have attempted to characterize and analyze the morphology of minicolumns with a 2D computerized method designed to detect subtle differences among patient groups such as schizophrenia, autism, and Alzheimer14,15,16,17. As a result, much of our understanding of cellular organization is focused on 2D histological images, which could potentially misrepresent biological structures and malpositioning of cells in 3D-space.
This paper aims to create a method that is accessible to the broader science community and analyze 3D tissue organization through the use of archival tissue to make detailed inferences about pathologies. In the present study, we developed Automatic Collector of Ultrathin Sections for Light microscopy (AutoCUTS-LM) to measure the neuronal cell morphology and their spatial organization in 3D-space of archived tissue. This is accomplished by modifying and adjusting the original AutoCUTS, which was designed for scanning EM array tomography18,19,20, to image archival human brain tissues (~20 years) in layer III of Brodmann Area 46 (BA46). BA46 was chosen since it involves working memory, attention and has been the subject of studies related to mental disorders like schizophrenia and depression21,22,23,24,25,26,27,28. Myelinated axon bundles are potentially cortical efferents that originated in layer II/III pyramidal cells as these bundles descend to the white matter29. Pyramidal neurons in layer III project to other cortical areas and play a key role in cortical and thalamic cortical circuits, and have been found to be the most affected layer by these disorders in BA4630,31,32.
Here, we report the applied technical workflow that is able to uncover morphological properties of pyramidal neurons in human brain autopsy tissue: First, we identified BA46 and applied advanced sampling procedures of biopsies. After embedding biopsies in resin, the AutoCUTS-LM cut them automatically into semi-thin sections (300 and 400 nm), and collected them on tape. Sections were stained and mounted on glass slides, and a library for each sample was organized. The digitalization of sections was stored, aligned, and stacked as a volumetric structure. Then, we detected neurons using the UNetDense architecture. Finally, the 3D spatial arrangement and structural parameters of pyramidal neurons in layer III of BA46 were analyzed in three human brains applying recently developed methods.
Our findings present valuable insight into neuronal morphology and architecture by characterizing pyramidal neurons in 3D from old archived human brain tissue. We discovered that pyramidal cells are not randomly distributed but are clustered in small columnar structures, which may be relevant for understanding the formation and function of the cortical network.