Multiscale three-dimensional imaging of intact human organs down to the cellular scale using hierarchical phase-contrast tomography

Abstract

Human organs are complex, three-dimensional and multiscale systems. Spatially mapping the human body down through its hierarchy, from entire organs to their individual functional units and specialised cells, is a major obstacle to fully understanding health and disease. To meet this challenge, we developed hierarchical phase-contrast tomography (HiP-CT), an X-ray phase propagation technique utilising the European Synchrotron Radiation Facility's Extremely Brilliant Source: the world's first high-energy 4 ^th generation X-ray source. HiP-CT enabled three-dimensional and non-destructive imaging at near-micron resolution in soft tissues at one hundred thousand times the voxel size whilst maintaining the organ's structure. We applied HiP-CT to image five intact human parenchymal organs: brain, lung, heart, kidney and spleen. These were hierarchically assessed with HiP-CT, providing a structural overview of the whole organ alongside detail of the organ's individual functional units and cells. The potential applications of HiP-CT were demonstrated through quantification and morphometry of glomeruli in an intact human kidney, and identification of regional changes to the architecture of the air-tissue interface and alveolar morphology in the lung of a deceased COVID-19 patient. Overall, we show that HiP-CT is a powerful tool which can provide a comprehensive picture of structural information for whole intact human organs, encompassing precise details on functional units and their constituent cells to better understand human health and disease.

Figure 1.. A HiP-CT pipeline for multiscale 3D imaging from whole organ to cellular resolution within large intact soft tissue samples.

A) Flow chart of HiP-CT human organ sample preparation and imaging. B) Experimental configuration, yellow arrows denote possible stage movements, and grey arrow shows propagation distance. The red line indicates the beam. There are two different optics (the two green structures). These provide the range of voxel sizes: 6.5–25 μm is provided by the left hand optic – dzoom optic, and 6–1.3 μm per voxel by the right hand optic- the zoom optic. The sample stage with sample jar can be seen on the right of the schematic and a photograph of the intact human brain mounted in the PET jar with the ethanol agar stabilization is shown in the inset. Ci) Maximum intensity projection of whole human lung with two randomly selected VOIs imaged at 2.45 μm voxel resolution shown in green (VOI1) and blue (VOI2). 3D reconstructions of the two high-resolution VOIs are shown, with 2D slices in the insets. In the 3D high resolution VOIs the fine mesh of pulmonary blood vessels, complex network of pulmonary alveoli and their septations can be seen. Yellow arrows denote occluded capillaries in 2D slices Cii) Image stack histograms for the green (VOI1) and blue (VOI2) high-resolution VOIs respectively (fixed bin width = 0.0001). Intensity distributions are comparable with positive skew (1.82 and 2.68) and kurtosis (6.44 and 11.88) for VOI1 and VOI2 respectively, histogram intersection is 71% ±3% for fixed bin widths range 1×10⁻²-3×10⁻⁴. Ciii) Box whisker plot showing the structural similarity index between 200 pairs of 2D slices taken randomly either from within the same VOI (1–1 and 2–2) or from different VOIs (1–2 and 2–1; one-way ANOVA; p = 0.3217). D) Single representative slices of high-resolution scans from a HiP-CT image of an intact whole human lung lobe and a biopsy taken from the same patient’s contralateral lung. Both VOIs are captured from the upper peripheral region of each upper lung lobe. In HiP-CT images, fine structure of the tissue including the blood capillaries (red arrows) and alveoli (blue arrows) as well as thin alveoli septi (yellow arrows in insets) are depicted.

Figure 2.. HiP-CT enables 3D imaging of organotypic functional units across intact human organs.

HiP-CT of brain (A), lung (B), heart (C) kidney (D) spleen (E), where for each organ (i) shows 3D rendering of the whole organ using the 25 μm per voxel scans. Subsequent 2D slices (ii-iv) show the posisitons of the higher-resolution VOIs relative to the previous scan. (v) Shows a digital zoom of the highest resolution image with annotations depicting characteristic structural features: in the brain (molecular layer, ml; granular layer, gl; Purkinje cell, pc), in the lung (blood capillary, c; epithelial cell/macrophage, ec/m), in the heart (myocardium, mc; coronary artery, ca; adipose tissue, at), in the kidney (efferent/afferent arteriole, e/a; glomerulus, g) and in the spleen (red pulp, rp; white pulp, wp; arteriole, a; splenic sinus, ss).

Figure 3.. HiP-CT analysis of kidney to measure glomeruli morphology and nephron number.

Ai) The three resolution datasets of HiP-CT taken of a human kidney aligned and overlaid. Aii) shows the measurement of the parenchymal volume semi-automatically segmented (green). Aiii) shows the 6μm dataset with the virtual biopsy cylinder in white. 853 glomeruli that were within this cylinder were counted (blue dots) and the parenchymal volume within the cylinder measured. Aiv) the 1.3μm dataset with virtual biopsy (red cylinder) shown. The 13 glomeruli that abutted this cylinder were cropped from the data and segmented (yellow). B) Comparison of HiP-CT with aligned histopathological sections (Haematoxylin and Eosin (H&E) stained) taken after all HiP-CT scanning was finished. The left-hand column shows light micrographs of H&E stained histopathological sections and the right-hand column shows 2D tomograms of HiP-CT, yellow boxes denote images that have been pseudo-coloured.

Figure 4:. HiP-CT with 3D image analysis and morphometry in the lung of a patient with COVID-19

Ai) 3D reconstruction from 25 μm per voxel HiP-CT scanning of the intact upper left lung lobe from the autopsy of a patient decreased from COVID-19-related ARDS. The high-resolution VOIs are shown in red (6 μm per voxel) and blue (2 μm per voxel). Aii) At 25 μm per voxel, high intensity regions are observed in the lung periphery. The yellow dashed line delineates a secondary pulmonary lobule. Aiii) At 6 μm per voxel, heterogeneity in the lung parenchyma including: (1) dilated alveolar ducts and diffuse loss of alveolar structural organisation; and (2) comparatively well-preserved alveolar structure with some oedematous changes. Aiv) At 2 μm per voxel we observed (3) alveolar obstruction, likely representing clotted blood based on its high intensity; and (4) hyaline thickening of alveolar septi. B) 3D reconstruction of COVID-19 lung with segmentation of two adjacent secondary pulmonary lobules with differing degrees of parenchymal deterioration. Ci) 3D reconstruction of segmented acini structure within the SARS-CoV-2-uninfected (control) lung and Cii) the SARS-CoV-2 infected lung. Di)-Diii). 3D reconstruction of representative VOI at the high resolution (2 μm voxels) for the control, COVIDs and COVID_C groups respectively. Duplicated volumes show visual representations of the air-tissue interface in COVID-19 lung, where a smaller distance between a voxel of air and a voxel of tissue is coloured blue whereas larger distances are coded yellow. Div-Dviii) Boxplots showing quantitative comparisons between COVID_S, COVID_C and control, VOIs. Quantification of mean surface area to volume ratio, airspace connectivity, and mean septal thickness are shown respectively (** indicates p<0.001, * indicates p<0.05 and ~ indicates p=0.08) (calculated by anova with Tukey comparison). Dix) Shows the distribution of airspace diameters for all six VOIs in each group modal values for COVID_C, COVID_S and Control = 8.88, 152 and 351 μm respectively.