Ultrastructural differences in pretangles between Alzheimer disease and corticobasal degeneration revealed by comparative light and electron microscopy

Changes that occur in relevant molecules before they become organized into disease-specific inclusions in human brains are attracting increasing attention [1]. The pretangle is an example of such an early change; it was originally defined under light microscopy (LM) as diffuse and granular tau immunoreactivity in the cytoplasm and neurites of otherwise intact neurons in brains from patients with Alzheimer disease (AD) [2]-[4]. Mature neurofibrillary tangles (NFTs), which are hallmarks of AD, are readily identified as bundles of paired helical filaments (PHFs) under electron microscopy (EM) [5]. However, it is difficult to identify pretangles under EM because their faint tau immunoreactivity suggests that their ultrastructure is less distinct. Although putative ultrastructures of pretangles in AD have been reported, it is not yet clarified whether they really represent neurons containing diffuse and granular tau immmunoreactivity as defined under LM [2]. Similar granular tau immunoreactivity has also been observed in corticobasal degeneration (CBD), a four-repeat tauopathy that causes degeneration of the cerebral cortex, basal ganglia, and substantia nigra. Because they appear similar to AD-pretangles under LM, this granular tau immunoreactivity is also known as pretangles. Pretangles are considered one of the most important neuronal cytopathologies in CBD [6],[7] but are also found in argyrophilic grain disease or progressive supranuclear palsy [8].

The aim of this study was to visualize the ultrastructures of LM-defined AD- and CBD-pretangles and thereby to identify possible differences between them. For this purpose, it is necessary to directly compare LM and EM images of the same pretangle, an approach that is now named “correlative light and electron microscopy (CLEM)”. Although CLEM protocols have been developing [9]-[17], they usually allow LM/EM comparisons for only small fields (the size of the EM preparation). This limitation makes it practically impossible to capture pretangles for immunoEM study because pretangles are not sufficiently frequent to be included by chance in such tiny preparations. Therefore, it is necessary to excise tissue containing a pretangle from the LM sample before it can be prepared for EM.

Quantum dots (QDs) are fluorescent, electron-dense semiconductor nanocrystals of uniform size with a core of cadmium selenide [18]. On EM examination, QDs also display a characteristic peripheral halo [19]. These dual optical properties allow QDs to be identified under both LM and EM and therefore permits labeled LM structures to be compared directly with their ultrastructures [16]. Using QDs, we recently established three dimension (3D) - oriented immunoelectron microscopy [19],[20]. In this method, a thick floating section from the formalin-fixed human brain is incubated with the primary and QD-conjugated secondary antibodies. After a target neuron is examined with fluorescent LM (confocal microscopy), landmarks are punched out around the neuron using laser microdissection. Then, this floating section is processed for EM preparation. The advantage of this stepwise LM-EM approach is that the neuron of interest can be observed closely on confocal microscopy prior to the EM examination, and its EM findings can be supplemented with confocal images because the same reporter (i.e., QDs) can been seen under both LM and EM immunostaining.

Although QDs provide a powerful bridge between LM and EM, their electron density is lower and their contour is less distinct than those of gold particles, leading to doubts about QD use as an immunolabeling material for EM. We previously overcame this problem using energy dispersive X-ray (EDX) spectrometry, which demonstrated parallel peaks corresponding to selenium (Se) and cadmium (Cd) on the pixels for definitive confirmation of QDs on EM preparations [19]. Because it is possible to obtain EDX spectrum for each pixel, we extended this pixel-based EDX analysis to plot the entire EM field pixel by pixel in this study. Operational display of pixels containing Cd peak or those containing Se peak highlighted QD particles based on their elemental composition with different colors. When it was overlaid onto the conventional gray-scale EM image, this EDX color mapping clearly distinguished QDs from background structures such as ribosomes.

With these methods, it is possible to examine the ultrastructural details of AD- and CBD-pretangles and to elucidate their similarities and differences at the EM level [2]-[4],[6]. Using this LM/EM correlation with novel mapping method, we obtained an EM image of the early stage of neuronal tau deposition in AD-pretangles and found essential differences between AD- and CBD-pretangles at the EM level. This is the first successful demonstration of their ultrastructural differences.

The name ‘ pretangle’ was originally used to describe the premature stage of NFT formation in AD [2]-[4]. However, in CBD, similar structures (also called ‘pretangles’) are more prevalent than NFTs in the cerebral cortex [6]. It has been unclear whether the pretangles of CBD represent a premature stage before NFT formation and whether they are different from AD-pretangles. Because pretangles are defined only by LM findings [2]-[4], it would be helpful to compare the corresponding pretangle ultrastructures between the two diseases. However, CBD pretangles in the cortex have not previously been described at the EM level [22]-[28]. To address this issue, we used a method of correlative light and electron microscopy with QD immunolabeling [19]. This procedure allowed us to observe not only the features of filamentous structures of inclusions but also their intracellular distribution and relationship with cellular organellae.

Using correlated LM and EM images, we observed a distinctive EM feature of AD-pretangles: specifically, a strong tendency to form bundles as a precursor to NFTs. Even in the earliest stages of tau accumulation, small pieces of NFTs could already been seen on the background of diffuse and granular tau staining on LM (Figure 3A) [3],[29]. Correlation of the LM and EM images revealed that the granular cytoplasmic staining on LM represented straight filaments (sparsely distributed in the neuronal cytoplasm), and the small tangles represented small bundles of parallel filaments (Figure 3D). Very similar EM findings were reported by Bancher [2], although it remains unclear whether their EM findings really represented LM-defined pretangles or not. Because our method of 3D-oriented immunoEM not only identify pretangles on LM, it is quite sure that our immuno EM findings represent LM-identified pretangles. Moreover, it is further possible to correlates the EM plane (Figure 3C) and its exact counterpart of the corresponding LM plane (Figure 3B), even a small aggregates (Figure 3B arrow) not identifiable on 3D stack image (Figure 3A) can be examined with EM (Figure 3C) for comparison with LM at the extreme accuracy. Another feature of AD-pretangles was perinuclear accumulation of tau. Correlation with its EM counterpart showed that a small number of tau-positive 15-nm straight filaments were present around the nuclear membranes of AD-pretangles (Figure 3G-J). It has been reported that PHF or tau-like immunoreactivity may be present in close proximity to the nuclear membrane of mature NFTs in AD [29]-[32], and this report is the first demonstration of tau-immunolabeled filaments around the nuclear membrane of AD-pretangles. Although intranuclear processes such as aberrant cell cycling may be related to the pathogenesis of AD [33],[34], it is unclear how these processes are related to the perinuclear or cytoplasmic deposition of tau.

A similar approach to CBD-pretangles of the cerebral cortex revealed several findings that differed from the features of AD-pretangles and NFTs: (i) random and diffuse distribution of 14–20 nm straight filaments and (ii) paucity of PHFs and fibrillary bundle formation. These ultrastructural architectures may explain the reticular or diffuse tau immmunoreactivity of CBD-pretangles seen on LM (Figures 4 and 5). In dendrites, a small number of straight filaments were observed lying parallel to the dendritic shaft (Figure 4D), similar to the previous reports of dendritic lesions in CBD [8],[22]. Even in CBD-pretangles with abundant tau filaments, this random and diffuse distribution of straight filaments was maintained with little NFT formation (Figure 5D). Indeed, this architecture was maintained even in Pick-like inclusions, where tau filaments were randomly assembled and were composed mainly of straight filaments and, to a lesser extent, PHFs with a periodicity of 130–180 nm (Figure 6D-E). Thus, so-called CBD-pretangles are a random accumulation of tau-positive straight filaments, rarely evolving into so-called NFTs even when the filament density is increased. These findings, especially regarding the filamentous structures themselves, were similar to previous findings in CBD (15–20 nm straight filaments or PHFs with a periodicity of 120–180 nm), which were observed in Pick-like inclusions [22]-[24],[27], ballooned neurons [24],[28], neuronal inclusions in the brainstem [22],[26],[28], or an in vivo study using CBD brains [25]. However, this study is the first to clarify the EM structures of cortical pretangles in CBD by accurately correlating them with LM images. Authentic Pick bodies in Pick body disease were more solid than pretangles on LM, where abundant tau-positive fibrils, 15 nm in diameter, were randomly arranged without forming PHF [19].

In this study, we greatly enhanced the reliability of pre-embedding immunoEM using QDs. Although QDs are considered suitable for CLEM, their reliability as a reporter for ultrastructural immunolabeling has been debated. The penetration of QD labeling is reported to be limited to several micronmeters from the sample surface [15]. However, we were successful in immunolabeling 25-μm-thick free-floating sections by increasing the incubation time with QD-conjugated secondary antibodies to 8 hr at room temperature. This procedure enabled us to label the entire thickness of floating sections with QDs so that each tau filament was sufficiently labeled (Figures 3 and 4). Consequently, confocal images and immunoEM images could be tightly correlated.

Other disadvantages of QDs are that they have a lower electron density and less distinct contours than colloidal gold for immunolabeling. We previously used EDX spot analysis with STEM to demonstrate the presence of Se and Cd on a pixel basis [19]. This EDX spot analysis, now extended to map the entire EM field, resulted in clear visualization of the position and form of each QD particle. When the corresponding EM image was overlaid, QDs could be readily differentiated from the grayscale cellular backgrounds (e.g., ribosomes) (Figure 2). Similar elemental mapping of Cd has been reported using electron energy loss spectrography (EELS) to detect QDs in ultrathin EM samples [35]. However, compared with EELS, EDX is more suitable for the detection of heavy metals, such as Cd or Se [19],[36]. Moreover, because EELS is performed without electron staining, it is difficult to gain sufficient contrast in EM images [35]. Therefore, combined with pre-embedding Q-dot immunoEM and EDX mapping, the use of QDs is one of the most sensitive and distinct ultrastructural immunolabeling techniques available and might be particularly suitable for the correlation of LM/EM images.

Accurate identification of pretangles on LM, followed by EM examination of their exact counterpart was achieved through tau immunolabeling with QD, fluorescent nanocrystals, which are detectable with LM (fluorescence signal) and with EM (electron dense particles with halo). EDX spot analysis to confirm the identity of QD on EM section by showing energy peaks for Cd and Se is now extended to map the entire EM field to highlight QD particles. This improved method with EDX mapping clearly demonstrated for the first time that AD-pretangles showed a strong tendency to form fibrillary tangles even at an early stage, whereas pretangles or Pick-like inclusions in tissue from patients with CBD did not even at an advanced stage. This novel strategy is useful to clarify how molecules other than tau are organized into ultrastructures in the early stages of disease-specific lesions.