Virtual histology for a multi-modal human brainstem atlas

What is the best imaging approach to build a human brainstem atlas? Dr. Gisela Hagberg, a senior scientist at the University Hospital and Faculty of Medicine, University of Tübingen, affiliated with the Max Planck Institute for Biological Cybernetics in Tübingen, shares insights for a selection of multimodal imaging techniques allowing detailed imaging of ex vivo brain tissue. She highlights propagation based XPCT methods developed by the Trieste-Tübingen Team at the SYRMEP Beamline, Elettra Facility in Trieste, Euro-BioImaging’s Phase Contrast Imaging Flagship Node, which make it possible to go beyond the constraints of conventional microCT techniques for human brain measurements.

MicroCT offers detailed imaging of ex vivo brain tissue. With the use of osmium stains, exquisite contrast between myelin containing tissue and its surrounding can be achieved, revealing ultra fine structure of axonal connections in the macaque brain¹. In preclinical imaging, staining of the vasculature using dedicated perfusion techniques allows high resolution microCT and complete mapping of the vascular tree, from the ascending arteries, arterioles, and capillaries to the venules and veins of the entire mouse brain has been obtained².

Challenges of working with human brain tissue

Such high-quality results are challenging to obtain in human brain tissue, where perfusion is performed post mortem. Due to fine capillaries collapsing during the post mortem interval (PMI), before formalin fixation has begun, vascular penetration of stains is limited. More in general, when larger tissue blocks are used – to increase coverage and obtain a wider depiction of the underlying morphology - tissue penetration of stains is particularly challenging. 

Micro-CT techniques of unstained tissue is therefore an interesting alternative. Owing to the additional contrast available with phase-sensitive techniques, an increasing number of research groups are looking into this possibility. Propagation based XPCT methods enhance the intrinsic contrast between tissue compartments through local differences in refractive index beyond what is possible with conventional microCT techniques. Subtle, additional XPCT contrast variations can be achieved through the specific tissue preparation steps chosen for the study³. For instance, ethanol immersion leads to differences in tissue hydration and can give good white to grey matter contrast for whole brain XPCT measurements⁴.

A team for multimodal imaging of human brainstem samples

We performed multimodal imaging on human brainstem samples obtained from the donor programme of the Institute for Clinical Anatomy and Cell Analysis (led by Bernhard Hirt), Eberhard Karl’s University of Tübingen (Fig 1) described previously⁵. The samples had postmortem intervals PMI<10h and were fixated with 4% formalin in phosphate buffered saline. MRI with quantitative techniques and isotropic voxelsizes between 75-300µm was performed prior to sectioning (Fig 1A-B), dehydration and paraffin embedding. XPCT was performed at SYRMEP, Elettra (Phase Contrast Imaging Flagship Node Trieste) which has a flexible setup to allow coverage of extended samples with multiresolution scanning and long propagation distances (Fig 1C). With cone-beam CT as an intermediate step, co-registration with MRI could be performed (Fig 1D). Without the use of any staining procedure, the vasculature can have a patchy appearance⁶. In our measurements, we found that this patchiness is due to the non-uniform distribution of haemoglobin containing red blood cells, which are trapped only at some locations within the vessels (Fig 1E-F). In XPCT these iron-containing compartments will absorb slightly more of the synchrotron light, leading to hyperintensities. Such vessel segments with trapped iron are interspersed by dark, hypointense segments. These are likely caused by an immobile parenchyma surrounding the vessel walls that can collapse, either during the PMI or during the ethanol dehydration steps prior to paraffin embedding⁷. Despite such highly different XPCT signal variations along the vascular tree, a combination of classical imaging filters could be used for segmentation of the microvasculature (Fig 1G)⁸. A hierarchical approach which included imaging with two different pixel sizes (full coverage with 5µm and zooms with 1µm) allowed detailed characterization of the vascular tree, including vessel diameter mapping (Fig 1H). 

Acknowledgement: The access to the node was kindly provided by Euro-Bioimaging. Besides the anonymous donors we would like to thank the technical staff in Tübingen and Elettra for their professional support of the project, and the Max Planck Institute for Biological Cybernetics for the MRI scanner access.