Background
Alzheimer’s disease (AD), the leading cause of age-related dementia in humans, occurs due to the convergence of diverse and still ill-understood pathological processes [
1,
2]. Toxic protein deposition, synapse loss, and microglial infiltration precede neuronal death, leading to progressive cognitive decline. As with many neurodegenerative diseases, AD occurs secondary to neuronal protein misfolding. Aggregates of misfolded oligomerized amyloid-β (Aβ) protein accumulate in both dense mesoscale extra-cellular plaques and intracellular fibrils. In addition, hyper-phosphorylated microtubule-associated tau (p-tau) cytoskeletal protein aggregates into dense intra-neuronal neurofibrillary tangles (NFTs). Moreover, cellular systems that regulate protein folding (e.g., chaperone proteins) and degradation (e.g., proteasomal and lysosomal systems) are also known progressively become altered in brain cells during aging [
3] (e.g., in lipofuscin granula, see Suppl.
Introduction).
An in-depth characterization of AD neurodegeneration has been achieved using established immunohistochemistry (IHC) approaches [
4], and extensive PET-tracer development is helping effective differential diagnosis [
5]. Still, modern imaging tools come short of delivering the volumetric cellular-level visualizations, which would be necessary to reliably detect and differentiate individual subtle protein-based cellular lesions or abnormal age-related intracellular accumulations in vivo, especially in the likely crucial pre-symptomatic early phases of AD. Moreover, brain-wide cellular-level monitoring of cellular aging and AD neurodegeneration progression is especially challenging for neuroimaging. An unequivocal AD diagnosis still can only be reached postmortem via histopathologic demonstration of the two hallmark protein lesions (Aβ and p-tau) in nervous system tissue. And even in the context of postmortem small animal imaging, current cutting-edge experimental neuroimaging methods lack an unbiased high-throughput 3D imaging technology sensitive to neuronal proteopathy that allows cellular resolution, full-organ brain coverage, and unbiased detection mechanisms (see Suppl.
Introduction). This limited diagnostic power significantly hinders our monitoring of normal neuronal aging processes, and especially our understanding of early-stage AD etiology and our ability to discover disease-modifying drugs for AD.
Amongst novel neuroimaging techniques, X-ray phase-contrast computed tomography [
6] (X-PCI-CT) represents a set of 3D microscopy techniques, offering enhanced image contrast compared to traditional absorption CT [
7] and thereby enable label-free soft-tissue imaging for diverse biomedical investigations [
8]. Its simplest implementation, propagation-based X-PCI-CT [
9,
10], exploits the physical mechanisms of coherent X-ray refraction, propagation, and Fresnel diffraction to determine local electron density within probed samples. Applied postmortem, different X-PCI-CT methods deliver semi-quantitative to fully quantitative label-free and dissection-free density-based 3D morphological neuroimaging complementary to other brain mapping techniques [
11‐
13]. Modern synchrotron-radiation setups reach the spatial resolution needed to impact mesoscale neuroimaging [
14‐
17], permitting the volumetric exploration of intracellular cyto-architecture within single neurons in deep rodent brain regions [
13,
18,
19]. Furthermore, they can be used to carry out unbiased anatomically dense high-throughput quantifications of cellular and vascular structure within large nervous-tissue samples at histological resolution [
20‐
22]. With regard to the detection of neurodegeneration, X-PCI-CT was shown to be sensitive to extra-cellular Aβ build-up in several AD animal models postmortem [
23‐
29]. Still, previous studies have come short of a multiscale and multimodal characterization of early-stage Aβ and p-tau-driven NFT build-up in intracellular compartments. Furthermore, prior studies have fallen short of applying X-PCI-CT within experimental neuroscience protocols.
Here, we present a postmortem X-PCI-CT-based multiscale organ-level to cellular-level analysis of brain cellular hyperdensity, focusing on intracellular abnormal protein and metal accumulation in aging and AD animal models. We performed micro-X-PCI-CT to nano-X-PCI-CT and X-ray fluorescence microscopy on extracted brain samples from aged wild-type (WT) and triple-transgenic 3xTgAD mice [
30], an experimental AD model that develops both amyloid and tau pathology [
31]. After analyzing multiscale X-PCI-CT results by comparison to more-established neuroimaging modalities, including MRI, TEM, and immunohistochemistry (IHC), we used this methodology to quantify differences in intracellular hyperdensity within key AD-linked hippocampal and cortical brain cell layers.
Neuroprotective effects, which counteract the degeneration in aging and diseased neuro populations, are being intensely investigated. Group II metabotropic glutamate (mGlu2 and mGlu3) receptors, for example, are interesting potential pharmacological targets, as they do not mediate but modulate glutamatergic neurotransmission. These receptors are located in presynaptic terminals where they negatively modulate adenylate cyclase with ensuing inhibition of glutamate release [
32]. mGlu3 receptors are also expressed in astrocytes, and activate mitogen-activated protein kinase and phosphatidylinositol-3-kinase pathways with ensuing increased production of neurotrophic factors such as transforming growth factor-β, glial-derived neurotrophic factor, nerve growth factor, and S-100β protein [
33]. Activation of astrocytic mGlu3 receptors also upregulates the expression of the glutamate transporter 1, which contributes to a reduction in extra-cellular glutamate concentrations, thus limiting excitotoxicity [
34]. Here, we tested the hypothesis that chronic pharmacological activation of mGlu2/3, by using the selective orthosteric agonist LY379268 [
35], could exert a neuroprotective effect in 3xTgAD mice. By collecting X-PCI-CT data from aged WT and 3xTgAD mice, systemically treated with saline or LY379268, we demonstrated for the first time a proof-of-principle application of X-PCI-CT for the morphologic and quantitative evaluation of local and intracellular neurodegeneration/neuroprotection processes.
Discussion
We developed a mesoscale (organ-scale to cellular-scale) probing system for volumetric postmortem morphological neuroimaging based on state-of-the-art, multiscale synchrotron-based X-PCI-CT. X-PCI-CT image contrast is generated via a label-free mechanism and quantitatively describes intra-sample X-ray phase variations proportional to local electron density. Their freedom from labels and stains qualifies these images as anatomically dense and unbiased direct measurements of nervous-tissue structure. This 3D imaging tool was applied to the brain-wide study of cellular hyperdensity in aged WT and aged 3xTgAD mice, an experimental AD mouse model. It was shown that X-PCI-CT provides micro-resolution to nano-resolution 3D neuroanatomical representations of deep neuronal and glial populations within extended rodent brain tissue samples after little manipulation beyond standard sample fixation and paraffin embedding, with invasive sample sectioning necessary only for the 0.13 voxel images. Therefore, high-resolution X-PCI-CT maps provide a means to virtually visualize 3D brain histology with electron density-based coloring.
The collected X-PCI-CT maps, in conjunction with morphological observations, enabled discrimination between different neuron types and different glial populations. Moreover, the X-PCI-CT technique was sensitive to small intra-neuronal density differences, enabling the detection of hyperdense neurons (ICHD) in brain tissues from both aged WT and aged 3xTgAD. The X-PCI-CT-based brain density maps located ICHD in somatic, nuclear, and dendritic/axonal cellular compartments of neurons in key AD-linked cell layers. The observation of ICHD was interpreted as abnormal intra-neuronal protein accumulation or abnormal lysosomal deposits likely arising due to cell aging-related processes (in WT mice) or neurodegeneration-related cellular processes (in 3xTgAD mice). This imaging method could thus be established as a tool for full-organ mapping of aging-associated and AD-associated intracellular lesions.
Since X-PCI-CT signal is biologically nonspecific, ICHD-bearing cells could represent a mixture of hyperdense normally aging or degenerated cell types. A biological characterization of the ICHD particles in 3xTgAD mice was achieved here via extensive multi-technique comparative analysis, involving fluorescence histology, IHC, XFM, MRI, and TEM. All results confirmed that the ICHD signal in 3xTgAD mice is highly likely representative of age-related AD-associated neurodegenerative processes of co-localized Aβ and p-tau intracellular deposition. Notably, both tissue-level and cellular-level patterns of ICHD positivity (Fig.
3) matched cellular markings in collected ThioS-dyed sections and patterns of amyloid and tau immunoreactivity in collected IHC sections (Figs.
1,
4). Somatic compartmentalization of Aβ and dendritic/axonal localization of p-tau fluorescent signal in IHC section suggests a spatially differentiated multi-peptidic agglomeration of AD-linked proteins in ICHD-bearing cells. Aged WT mice also presented ICHD (Fig.
2) and some level of ThioS fluorescence (Fig.
1). The very low levels of immunoreactivity for Aβ and p-tau in the WT mice (Suppl. Figure
9) suggest the ICHD observed in WT animals be likely related to normal neuron aging processes (e.g., lipofuscin granula formation).
Compared to high-field MRI (Suppl. Figures
10–
11), which achieves pre-cellular spatial resolution [
67] after tens of hours of scanning time in the absence of contrast agent, X-PCI-CT with brilliant synchrotron-radiation X-rays presents clear advantages in terms of spatial resolution (sub-cellular) and measurement durations (minutes to a few hours [
37], label-free). The subtlest sub-cellular lesions and protein deposits typical of early AD phases, such as axonal p-tau NFT lesions, are, in fact, especially elusive to high-field MRI [
68]. Compared to TEM, in practice a 2D imaging method for ultrathin sections, X-PCI-CT can capture 3D cellular neuroanatomy within extended un-sliced brain samples.
What stands out in the collected XFM data is the observed perisomatic near-nuclear Fe and Ca increased concentrations in ICHD cells compared to normal cells (Fig.
5). While the nuclear Fe aggregates in normal cells are consistent with anti-fibrillarin IHC nuclear staining of normal neurons [
69], diffuse somatic hyper-accumulation of Ca and Fe metals within ICHD-labeled neurons represents a sign of likely cellular dysfunction, arising either in direct connection to toxic protein clumping (e.g., by protein chelation), or due to cell aging and cell death-related processes. Disturbances in iron metabolism have been coupled to several neurodegenerative diseases [
70] and, indeed, iron, either directly bound to amyloid and tau lesions [
71] or associated to cytoplasmic RNA [
72], is known to play a toxic role in AD pathology [
62], and cause oxidative damage and neurodegeneration. Aβ aggregates, in turn, can induce Ca dyshomeostasis [
73] and lead to cellular synaptic dysfunction, neurodegeneration [
74], and cell death (apoptosis [
75], necrosis, or autophagy [
76]). Overall, these multi-technique comparative analyses established a clear correlation between the ICHD signal, detected via X-PCI-CT, and multiple concurrent and possibly co-localizing intracellular processes related to aging in WT mice, and to amyloid-driven and tau-driven AD-linked cellular proteopathy and neurodegeneration in 3xTgAD mice.
The attained characterization of the ICHD signal as a nonspecific cellular biomarker of neuronal aging and degeneration allowed the application of multiscale X-PCI-CT for the postmortem brain-wide 3D detection and quantification of cellular hyperdensity within small animal brain samples, establishing a novel platform for quantitative cell-by-cell screening and quantification. As a proof-of-concept, we studied the response of aged WT and 3xTgAD mice to a systemic chronic treatment with LY379268, a potential neuroprotective drug. LY379268 activates mGlu2 and mGlu3 metabotropic glutamate receptors and, thus, may reduce excitatory synaptic transmission, and consequently excitotoxicity by reducing glutamate release and increasing glutamate clearance [
34,
77,
78]. Moreover, mGlu3 receptors may induce neuroprotective effects by increasing the production of neurotrophic factors and enhancing glutamate clearance [
33]. Moreover, chronic treatment with LY379268, leading to an increased production of glial-derived neurotrophic factor, does not lead to the development of tolerance, making this treatment optimal [
79]. The analysis of X-PCI-CT data (Fig.
6) pointed out regional differences in ICHD and, most notably, significant lower levels of agglomerates in ventral CTX layers of 3xTgAD mice treated with the drug. Our estimates of lesion tissue load, in the 1–5% range, are in good agreement with a similar X-PCI-based AD-linked protein-deposit quantification performed on the 5xFAD genetic animal model, which reported neocortical amyloid load levels of around 2% in similarly aged animals [
25]. Overall, these results are encouraging and provide a rationale for a future larger-scale study.
In summary, the multiscale characterization of intracellular hyperdensity, obtained by 3D micro-to-nano-imaging of deep neuronal populations vulnerable to proteopathy in key brain layers, cross-validated by other imaging modalities, represents a novel methodological advance of this study. The intracellular hyperdensity in cortical and hippocampal neurons of 3xTgAD mice was characterized as hyper-accumulation of Aβ and p-tau proteins, with the involvement of several key bio-elements, including calcium and iron. Furthermore, the obtained structural-morphological discrimination of glial vs. neuronal cells demonstrated that a precise focusing of the neuropathological evaluation is also possible via this methodology. The viability of this approach for an evaluation of experimental neuroprotective strategies was demonstrated by the efficient unbiased label-free screening of full-organ rodent brains achieved in the “Part
III” section of this study.
The X-PCI-CT virtual-histological approach presents the evident benefit over traditional histology and IHC that full-organ structural analyses without regional bias can be performed in a sample-preserving fashion. State-of-the-art IHC approaches to study brain neuroanatomy, in fact, permit rather cumbersome 3D cellular imaging, but are limited in terms of organ coverage, involve sample sectioning and complicated reconstructions with stitching and aligning issues, and require the use of various intrinsically biased labels (histologic stains, immuno-labels, contrast agents). Recently emerging super-resolution 3D neuroimaging technologies, mainly based on tissue clearing [
80] and expansion [
81] or two-photon microscopy [
82], enable brain-wide cellular resolution structural and functional investigations of entire cell populations [
83], and down to single cells [
84] and single intracellular molecules [
85]. These methods can be used for organ-level transcriptomics and connectomics, and to investigate complex biological processes such as aging and neurodegeneration [
86]. Still, they are based on fluorescence light microscopy, thereby falling short of delivering completely unbiased and anatomically dense visualizations of neural tissue [
87], due to the notorious issue of sparse labeling [
88] by means of antibodies or small molecule tags. Super-resolution electron microscopy (EM) techniques, such as transmission EM (TEM), serial block-face scanning EM (SBEM), or focused ion beam scanning EM (FIB-SEM), are based on label-free cyto-architecture-detection mechanisms and therefore enable anatomically dense characterization of both intra-neuronal amyloid and tau pathology at synaptic resolution by using heavy metal staining. However, these techniques rely on ultrathin sectioning and ablation, generate very large data sets [
89] requiring alignment and stitching, have to overcome the hurdle of long acquisition times [
90], and are limited to very small tissue volumes, far from rodent 3D whole-brain throughput capabilities.
The X-PCI-CT method, instead, affords 3D-morphological measurements comparable to what light microscopy and TEM approaches can provide on 2D thin sections. An evident benefit of a dissection-free multiscale imaging methodology is its ability to collect pre-cellular to cellular-level maps at variable spatial resolutions and from different deep brain regions, simply by running an appropriately aimed sequence of less invasive local CT scans, without having to go through a series of error-prone mechanical sample slicing operations. Furthermore, while one-shot histological workup requires strategic pre-planning before data collection, X-PCI-CT allows for repeated sample interrogations, admits unexpected observations, and can provide a morphological database on which to plan further data collection. Sample-preserving X-PCI-CT imaging in fact is conveniently compatible with most other postmortem neuroimaging analyses. Both the XFM and TEM analyses presented in our work, for example, were carried out on the same brain samples previously imaging via X-PCI-CT. In addition, though the histological and IHC analyses were performed on contralateral hemispheres for practical and time sparing reasons contingent to this multi-institutional study, X-PCI-CT is also compatible with post-imaging same-sample histological and IHC work [
12,
29]. The only postmortem imaging technique that needs to precede X-PCI-CT is MRI, for which a fully hydrated sample is necessary.
The high radiation doses as well as the scan times, used in high-resolution (micro/nano-scale) X-PCI-CT in this work, are not compatible with the imaging of living humans. This imaging approach, though, can be readily applied to human tissues postmortem, e.g. by using autoptic brain samples, e.g. from hippocampus and cerebral cortex tissues. Bioptic material could also be studied if collectable, especially in other organ systems. This methodology could e.g. be used in postmortem human brain samples of people affected by Alzheimer’s disease who had been treated with drugs to reduce the neurodegeneration. This would allow quantifying the potential neuroprotective effect of drugs administered at earlier times in patients showing precocious symptoms of disease or mild cognitive impairment. Moreover, this approach could also be used e.g. in postmortem human brain samples of brain cancer patients to confirm and quantify the antineoplastic effect of anticancer drug treatments. As in the small animal case, human brain tissues should be analyzed after their extraction from the bony skull to avoid the formation of scattering artifacts and disturbances of the X-ray wave front (needed for optimal image quality). Sample size limitations become more important as the spatial resolution increases, just as in the case of the small animal samples used in this work. Tailored multiple local-tomography scan protocols may enable full human brain sample coverage and micrometric (3.03 to 0.73 micron voxel) resolution in the future. Dataset sizes may become the limiting factor in this case. The nano-holotomography setup instead currently limits the sample size to few cubic millimeters.
Summarizing, the multiscale X-PCI-CT approach adds the following information to a cellular-level analysis of an AD mouse brain sample: a volumetric mapping and quantification of sub-cellular-level neurodegenerative lesions in all brain regions (full-organ imaging), an unbiased load calculation, and a quantitative characterization of the sizes and the shapes of these lesions within a single tissue-preserving examination. A more in-depth discussion of the main themes of this study is included as supplementary material (see Suppl.
Discussion).
Publisher's note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
The content of the manuscript is original, has not already been published in a journal, and is not currently under consideration by another journal. All authors agree to the terms of the BioMed Central Copyright and License Agreement.