3D imaging of SARS-CoV-2 infected hamster lungs by X-ray phase contrast tomography enables drug testing

Abstract

X-ray Phase Contrast Tomography (XPCT) based on wavefield propagation has been established as a high resolution three-dimensional (3D) imaging modality, suitable to reconstruct the intricate structure of soft tissues, and the corresponding pathological alterations. However, for biomedical research, more is needed than 3D visualisation and rendering of the cytoarchitecture in a few selected cases. First, the throughput needs to be increased to cover a statistically relevant number of samples. Second, the cytoarchitecture has to be quantified in terms of morphometric parameters, independent of visual impression. Third, dimensionality reduction and classification are required for identification of effects and interpretation of results. To address these challenges, we here design and implement a novel integrated and high throughput XPCT imaging and analysis workflow for 3D histology, pathohistology and drug testing. Our approach uses semi-automated data acquisition, reconstruction and statistical quantification. We demonstrate its capability for the example of lung pathohistology in Covid-19. Using a small animal model, different Covid-19 drug candidates are administered after infection and tested in view of restoration of the physiological cytoarchitecture, specifically the alveolar morphology. To this end, we then use morphometric parameter determination followed by a dimensionality reduction and classification based on optimal transport. This approach allows efficient discrimination between physiological and pathological lung structure, thereby providing quantitative insights into the pathological progression and partial recovery due to drug treatment. Finally, we stress that the XPCT image chain implemented here only used synchrotron radiation for validation, while the data used for analysis was recorded with laboratory $\mu$ CT radiation, more easily accessible for pre-clinical research.

Figure 1

(a) Seven days after infection of Golden Syrian hamsters by SARS-CoV-2, the animals were sacrificed, lung autopsies were extracted and paraffin embedding was performed. In total, 45 hamsters, divided in nine groups, were examined in this study, including uninfected controls ($\times$3), positive controls ($\times$8), PEG ($\times$3) and PBS controls ($\times$3) as well as five different drug candidates ($\times$8, each). In further analysis, uninfected, PEG and PBS controls are referred to as NEG-CTRL, while positive controls are referred to as POS-CTRL. (b) Schematic depiction of the in-house experimental setup. The cone beam geometry allows for an adjustable magnification and thus effective voxel size depending on the distances x₀₁ and x₁₂. A scintillator-based CCD-detector is recording the incoming photons, followed by a phase- and tomographic reconstruction from the acquired projections. (c) Photo of the experimental setup with the source (left), the sample stage (center, green) and the detector (right, blue). X-rays are created at the anode (orange spot) and propagate through the object, reaching the scintillator where X-rays are converted to visible light and subsequently to an electrical signal. (d,e,f) Schematic illustration of classification by OT. From top to bottom, in color: three synthetic sample distributions $\mu$, $\nu$ and $\rho$; the corresponding inverse cumulative distribution functions used in (2); the corresponding PCA embedding, obtained by projecting onto the two first principal components. From bottom to top, in gray: Some selected points around $\rho$ in the 2d-embedding, along the first principal component; approximately corresponding to vertical translation in the inverse CDF space; approximately corresponding to a translation on $\rho$ (with a slight change in standard deviation), thus making this direction in the embedding interpretable.

Figure 2

Comparison of imaging setups for uninfected control sample (left) and positive control (right): (a) Selected slices from in-house recorded tomograms (entire 3 mm punch). (b) Comparison of two slices recorded at the coherent imaging beamline at DESY (P10, PETRA III) with an effective pixel size of 650 nm and a FOV of $\sim$1.3 mm. (c) Slices from an uninfected (left) and positive control (right) hamster, recorded at the ID16A beamline (ESRF, Grenoble) with an effective pixel size of 90 nm and a FOV of 0.29 mm. All recording parameters are listed in Table 2.

Figure 3

Demonstration of workflow for XPCT. (a) Illustration of data acquisition by using in-house XPCT. Paraffin-embedded sample biopsies are stacked onto each other in a Kapton tube and inserted into a Huber pin. By vertical translation all samples are scanned and the electron densities ${\rho }_{x,y,z}$ are automatically reconstructed (1). Thereafter, a region of interest in the volume is selected (2). (b) (i) shows the selected region of interest, which is binned (ii) after finding a thresholding value (Otsu’s method) and further processed by applying morphological operations (iii), namely a combination of erosion and dilation (opening/closing). Subsequently, chord lengths can be extracted by introducing a few thousand of randomly oriented lines to every slice of a binned volume. A line can pass through alveolar septae (phase 0, black) or alveolar lumen (phase 1, white) and is accordingly divided (iv). The length is determined by the distance between the intersection points (blue dots).

Figure 4

(a) Chord length distribution (CLD) for alveolar septae (left) and lumen (right), shown for all hamsters, including control groups and drug treated animals. The normalized CLD corresponds to the (unweighted) probability distribution function (PDFs) of finding a cord length $L_c$,in the alveolar septae or lumen, respectively. The horizontal black lines divide the different hamster groups in (from top to bottom): UNI-CTRL, POS-CTRL, PEG-CTRL, drug group 1, drug group 2, PBS-CTRL, drug group 3, drug group 4, drug group 5. (b) The PDFs for the septae data of the control groups (negative in blue, positive in red). Outliers are shown in green (H52, negative), orange (H56, positive), and black (H59). For the case of H59, inspection of the sample showed pronounced cracks in the paraffin, i.e. an artifact of sample preparation.

Figure 5

(a) Embedding of the control group into the plane spanned by the first two PCA components. Pathological (positive controls) hamsters are drawn in red while negative controls are in blue. Axes are scaled by the standard deviations ${\sigma }_1$ and ${\sigma }_2$ of the samples along each PCA component. The two subgroups are almost perfectly separated by their coordinate along the first PCA component (PCA1), as emphasized by the grey dashed line. (b) Evolution of the chord length distribution when moving from the origin along PCA1 by one standard deviation in both directions (represented by horizontal solid grey line in a).

Figure 6

PCA1 values (in units of the standard deviation ${\sigma }_1$ along the first principal component) of all five drug trials groups with mean and intervals of one standard deviation of each group indicated by lines. The dashed line indicates $\text{PCA1}=0$ for emphasis.