To monitor disease development and uncover possible pre-symptomatic disease-related structural and cellular alterations, we used XPCT to obtain 3D images of CNS, gut, and eye in EAE-induced mice at different time points following immunization with MOG35-55, spanning very early asymptomatic stage, i.e. 3 days post induction (dpi), to asymptomatic inflammatory phase (7 dpi) through to disease onset (11-13 dpi).
XPCT allows direct detection of blood-CNS barrier disruption in EAE
We first investigated the disruption of the blood-brain and blood-spinal cord barriers (BBB/BSCB), which is a crucial hallmark in the pathogenesis of MS and EAE. The impairment of the barrier manifests as an alteration in its permeability which favors subsequent infiltration of immune cells and immune mediators into the CNS. We analyzed the brain and spinal cord (SC) in naïve mice as control and in EAE-affected mice, focusing on the lumbar SC region, which is first affected since the temporal pattern of the EAE lesions exhibits a caudo-rostral progression20. Ex-vivo XPCT images were obtained of the organs isolated from mice sacrificed at the selected time points. Fig. 1 demonstrates the progressive degeneration of the vascular system through the appearance of “clouds”, visible as halos around the vessels. These clouds are compatible with extravasated material, which includes inflammatory cells and blood components. The presence and the extension of the clouds can be assumed as a criterion to assess the increasing BBB/BSCB permeability and, consequently, the ongoing inflammatory process. Fig. 1 (a-d) shows representative tomographic images of a sample of lumbar SC in sagittal view for each time point. The tomographic images have a voxel size of 3.05x3.05x3.05 𝜇𝑚3. We applied image segmentation through intensity maximum projections in order to highlight the vasculature. The different grey-levels are proportional to different electron densities inside the sample: structures with higher electron density appear brighter than the surrounding tissue. We have chosen to use non-perfused samples to preserve the blood components inside them. Indeed, because of the presence of proteins which bind and carry metals, such as hemoglobin, transferrin and albumin, blood acts as an endogenous contrast agent, highlighting the blood vessels and the extravasated material21. The micrometric spatial resolution allows the visualization of the distribution of vessels with diameters ranging from 10 to 35 microns.
In all the samples, we can observe vessels arising from the longitudinal anterior spinal artery, visible as a white stripe running along the right vertical edge of the SC. In the lumbar SC of EAE-induced mice at 7 dpi and onset (Fig. 1c, d), these vessels appear surrounded by numerous clouds of an intermediate level of grey, which are not observed in the SC of naive mice or of EAE-induced mice at 3 dpi (Figs. 1a, b).
Quantification of vascular damage at different time points is shown in the bar chart reported in Fig. 1e. We analyzed n=3 samples for each group (naïve, 3 dpi, 7 dpi, onset), defining as damaged vessels those that present with clouds in their surroundings, such as those highlighted in the insets of Fig. 1f, g and indicated by the arrows. The vascular damage is thus expressed as the number of the detected vessels surrounded by clouds over the total number of visible vessels in the whole lumbar SC region. At 3dpi, the first time point tested, damaged vessels are not visible. At 7 dpi, corresponding to an early inflammatory phase of the disease, about 10 % of the vessels were damaged; this value increased to 55% at disease onset.
By measuring vascular damage separately for vessels originating from the central artery and vessels originating from secondary arteries and venules, over time, we aimed at defining how vascular alteration spreads in lumbar SC. As mentioned above, at 7 dpi, vascular degeneration largely involves vessels arising from the anterior spinal artery, as seen in Fig. 1c and f, which present the sagittal and axial views of a representative 7 dpi lumbar SC, respectively. The zoom in Fig. 1f highlights the presence of a small cloud (indicated by the arrow) close to a vessel originating from the central artery. However, at that time, lesions are not visible around vessels arising from other arteries or venules. In contrast, at the onset of the disease, both vessels from the central artery and vessels from the minor arteries appear significantly damaged, as visible in Fig. 1d and g, showing the sagittal and axial views of a representative lumbar SC at onset, respectively. The zooms in Fig. 1g highlight the presence of clouds around numerous vessels (arrows). Note that the clouds are mainly located at the base of the vessels.
Quantification of the number of damaged vessels measured separately for vessels originating from the central artery and for vessels originating from secondary arteries and venules, is reported in Fig. 1h. The vessels from the central artery appear to be almost all involved in the inflammatory process from 7 dpi. On the other hand, for vessels originating from secondary arteries, we observe a sharp increase in damaged vessels from 15% at 7 dpi to 50% at onset, reflecting how the lesion becomes more extensive at this stage.
High resolution tomographic images (voxel size of 0.65×0.65×0.65 𝜇𝑚3) allowed the visualization of the clouds as including a large accumulation of cells localized around the vessels (Fig. 2a), which would be commensurate with infiltrating inflammatory T cells and macrophages typical of an EAE lesion. The rendering reported in Fig. 2b provides a 3D morphological description of the surroundings of this lesion, where the structures of interest are segmented in different colors. We can observe how the clouds of extravasated material (gray), which include the infiltrating cells, completely surround the base of the vessels (yellow). In the neural network, we could distinguish cells with morphology commensurate with nonspecific neuron-like cells and multipolar neuron-like cells, rendered in green and blue respectively.
The tomographic images of the SC displayed in Fig. 2c, e, and f demonstrate the level of detailing achievable with micro-XPCT, which allows a thorough imaging of the micro-environment at neuronal and vascular levels. Fig. 2d shows cells located in the anterior horn of the SC compatible with neurons provided with processes, dendrites, or axons. A 3D-rendering of the neuro-vascular micro-environment is shown in Fig. 2d, where cells compatible with neurons (blue) are surrounded by capillaries (yellow). The 3D representation emphasizes the distribution in space of the neural and vascular components. As observed in Fig. 2e and f (white arrows), dashed structures were detected, which could be compatible with neural axons wrapped by sheaths of myelin, an extension of oligodendrocyte plasma membrane, periodically interrupted by the so-called nodes of Ranvier. Of note, in Fig. 2e, the myelinated axon (white arrow) seems to become less bright and to lose its periodicity as it approaches the EAE vascular lesion (yellow arrow).
Along with the SCs, we measured brains dissected from the same mice, to assess the anatomical progression of the vascular damage throughout the CNS of EAE-induced mice. Fig. 3a shows the posterior brain region of a mouse at onset, where cerebellum, brainstem, and cervical SC regions are visible. Tomographic images in Fig. 3b and c, which represent a portion of brainstem and cervical SC in coronal view, respectively, demonstrate the presence of clouds and cell accumulation (highlighted by arrows) in these sites at the clinical onset. Although present to a lesser extent with respect to the lumbar SC region, BBB/BSCB damage appears to have reached upper SC and brainstem at clinical onset, whereas vascular alterations in these regions are not observed at 7 dpi. These results corroborate the temporal course of BBB dysfunction in EAE, with increased permeability proceeding along from the lumbar SC to the brain20,22, and confirm XPCT as a high-resolution technique able to detect even subtle alterations of the blood barriers.
XPCT allows the detection of variations in cell density in ileal villi in EAE
To investigate whether there is a temporal correlation between CNS and gut alterations, we sought to identify imaging markers for the inflammation processes that involve the intestine of EAE-induced mice. To reveal changes between normal and pathological conditions, the gut morphology needs to be examined, but the complex geometry and structural convolutions of the gut prevent the identification of sample features. Anatomically, the gut looks like a long tube of varying diameter with flexible walls, folded several times on itself, as illustrated by 3D-rendering in Fig. 4a. To extract morphological details, we reduced the geometrical complexity by virtual flattening. The gut cylindrical surface was mapped into a plane by means of flattening algorithms described in Methods. These procedures allow the visualization of structures otherwise very hard to recognize in the 3D volume. Figs. 4b, c, and d illustrate some morphological features of the small intestine wall, displayed on 2D plane upon virtual flattening. We can distinguish the thin longitudinal layer of smooth muscle fibers belonging to the tunica muscularis (Fig. 4b), cells compatible in shape, dimension, and location with neurons of the myenteric plexus23,24 (Fig. 4c), and blood vessels (Fig. 4d) running along the tela submucosa.
XPCT does make possible the multiscale 3D imaging of the tissues, ranging from the intestine as a whole (Fig. 4a) down to the single cell (Fig. 4b), thereby overcoming the limitations of histology/immuno-histochemistry. Indeed, although histology/immuno-histochemistry provides very informative and resolved 2D images of biological tissues, it requires destructive sample preparation to thin the sample down to hundreds of microns and subsequently restricts spatial coverage within a finite depth. However, in contrast to immuno-histochemistry which identifies a cell through its staining of markers recognized by specific antibodies, XPCT cannot identify a cell other than through its morphological aspect, together with its location. Nevertheless, XPCT can reproduce in 3D the morphological features visible in the histological sections with a similar spatial resolution. The comparison between a hematoxylin-eosin stained histological section and an XPCT image of villi from a naïve mouse ileum (Figs. 4e and f, respectively) demonstrates that XPCT can indeed achieve histology-like resolution. A 3D rendering from XPCT data of an ileal villus is shown in Fig. 4g, where blood vessels are segmented in red, lymphatic canals in blue and cell nuclei in yellow.
In some models of EAE3, intestinal barrier permeability, together with morphological alterations, was observed already at 7 dpi and was associated with an increase in potentially pathogenic T cells infiltrating the gut lamina propria3,25. In order to identify an imaging marker of EAE detectable with XPCT, we measured the density of cells located in this area, limited to the region of the villi, at different time points of the disease, i.e. 3 dpi, 7 dpi and EAE onset. We define cell density as the number of cells detected per unit volume.
Because XPCT does not allow the identification of specific cells, if not for morphological or location considerations, to exclude epithelial cells and select only the cells located in the lamina propria we quantified the variation over time in cell density in the area delineated by the red line in Fig. 4h, which shows a 3D-rendering of a villus segmented to highlight the spatial distribution of cell nuclei (in yellow) inside the volume.
We analyzed n=2 ilea for each time point in EAE-induced mice (3 dpi, 7 dpi, EAE onset) or for naïve mice. Measurements of cell density were performed on approximately 20 villi per ileum. Quantification of cell density is shown in Fig. 4i, where we observe an increment at clinical onset.
However, because of the large variability we have detected in the villi population of a single mouse and among different mice, we do not pretend to have obtained determinative or conclusive results. To achieve statistical significance, it will be necessary to measure many more samples so as to limit the variability effect. Nevertheless, our results indicate that XPCT and the described procedures are fully suitable to perform both qualitative and quantitative investigations at high resolution.
XPCT reveals inflammatory infiltrates and atrophy of the optic nerve at EAE onset
Visual deficits are relevant symptoms in MS and EAE. Neuropathological alterations in the retina and the optic nerve occur, including thinning of retinal fiber layers, losses in retinal ganglion cells, demyelination and activation of microglia and astroglia, and inflammatory cell infiltration.
The accessibility of the retina through advanced non-invasive ocular imaging techniques makes it a potential convenient site for the research and diagnosis of diseases affecting the CNS. For this reason, it is essential to identify imaging markers in the eye and to temporally correlate them to the progression of the disease in the other relevant anatomical sites.
Fig. 5a and b shows XPCT images of the optic nerve in longitudinal view, near the point where it enters the retina, partially visible at the top of the images, in a naïve mouse (Fig. 5a) and an EAE-induced mouse (Fig. 5b). The dashed lines define the location of the axial sections reported in Fig. 5c and d. In both longitudinal and axial view of the EAE optic nerve (Fig. 5b and 5d), XPCT reveals clear alterations represented by large accumulations of inflammatory cell infiltrates (highlighted by the arrows) compatible with the occurrence of optic neuritis. Moreover, the longitudinal view in Fig. 5b clearly shows atrophy of the optic nerve.
The XPCT technique provides a feasible means to detect the structure of blood vessels, and we are interested in studying the vasculature of the eye over an area as large as possible. Blood vessels in the eye are confined mainly within very thin spherical layers and optimal visualization of this vascular network is possible only upon virtual flattening. Moreover, deformations can occur upon embedding the samples in the media, as seen in Fig. 5e, which shows a section of naïve mouse eye imaged by XPCT. The outer layers of the eye collapsed onto the lens (the circular shape structure in the center) and curled up since the vitreous contained within the eye dried out. Therefore, to investigate the ocular vascularization, virtual flattening procedures are mandatory. Once the spherical surface of the eye has been virtually flattened, the different layers can be studied by simply performing virtual slicing parallel to the plane of the layers. Representative results of virtual flattening, performed by means of algorithms described in Methods, are shown in Fig. 5f and g, which show a lateral section of the retina and the en face surface of the choroidal layer, respectively. The choroid (Fig. 5g) is a highly vascularized structure that separates the sclera from the retina. XPCT images were quantified and we measured the diameter of the visible vessels, which ranged from 5 to 20 microns.
Sequential evolution of pathological alterations of EAE over time in the investigated anatomical sites
We have monitored pathological features of EAE at pre-symptomatic and early stages of the disease in each of the considered anatomical regions, i.e. brain, spinal cord, gut, and optic nerve. The observed alterations are of different nature and appear at different time points, depending on the location. In the CNS, we observed the impairment of the vascular system through the appearance of clouds revealing the presence of extravasated material due to the pathological increase in BBB permeability; in the gut, XPCT permitted the monitoring of cellular infiltrates increasing over time from disease induction to onset; at the level of the optic nerve, a large accumulation of infiltrating cells and atrophy were detected. Fig. 6a summarizes with a visual schematization the results obtained through XPCT. In the lumbar spinal cord and gut, pathological alterations (PA) were identified as early as the pre-symptomatic stage, whereas changes appeared at a later time in the brain, as expected.
Since we handle heterogeneous data that describe the PA through the quantification of the presence of distinctive markers as described above, to compare the extent of PA involving the different organs we introduced the parameter PAr as the ratio of PA at a certain time point over PA at the onset (see Methods). The PA has been calculated as the percentage of damaged vessels in the CNS, and in the gut as the variation in cell density with respect to the control.
In Fig. 6b, we show the spatial distribution of PAr revealed by XPCT at 7 dpi in the CNS and gut, to emphasize the appearance of imaging markers of EAE as early as the pre-symptomatic stage. At this time point, the PA reflected by the imaging markers of BBB permeability is not yet visible in the brain. At the same stage, however, we observe that both the SC and the gut exhibit some grade of PA expressed through PAr. While it is not surprising to observe changes in the SC already from a pre-symptomatic stage, it might be significant to observe a non-negligible level of alteration at that time point in the gut compared with that measured at onset. In this framework, Fig. 6c depicts the temporal distribution of PA, expressed as PAr, appearing in the gut; thus, PA, already appreciable in the pre-symptomatic stages as early as 5 dpi, increase further at clinical onset.