Spatiotemporal characterization of cellular tau pathology in the human locus coeruleus–pericoerulear complex by three-dimensional imaging

Tau pathology of the noradrenergic locus coeruleus (LC) is a hallmark of several age-related neurodegenerative disorders, including Alzheimer’s disease. However, a comprehensive neuropathological examination of the LC is difficult due to its small size and rod-like shape. To investigate the LC cytoarchitecture and tau cytoskeletal pathology in relation to possible propagation patterns of disease-associated tau in an unprecedented large-scale three-dimensional view, we utilized volume immunostaining and optical clearing technology combined with light sheet fluorescence microscopy. We examined AT8+ pathological tau in the LC/pericoerulear region of 20 brains from Braak neurofibrillary tangle (NFT) stage 0–6. We demonstrate an intriguing morphological complexity and heterogeneity of AT8+ cellular structures in the LC, representing various intracellular stages of NFT maturation and their diverse transition forms. We describe novel morphologies of neuronal tau pathology such as AT8+ cells with fine filamentous somatic protrusions or with disintegrating soma. We show that gradual dendritic atrophy is the first morphological sign of the degeneration of tangle-bearing neurons, even preceding axonal lesions. Interestingly, irrespective of the Braak NFT stage, tau pathology is more advanced in the dorsal LC that preferentially projects to vulnerable forebrain regions in Alzheimer’s disease, like the hippocampus or neocortical areas, compared to the ventral LC projecting to the cerebellum and medulla. Moreover, already in the precortical Braak 0 stage, 3D analysis reveals clustering tendency and dendro-dendritic close appositions of AT8+ LC neurons, AT8+ long axons of NFT-bearing cells that join the ascending dorsal noradrenergic bundle after leaving the LC, as well as AT8+ processes of NFT-bearing LC neurons that target the 4th ventricle wall. Our study suggests that the unique cytoarchitecture, comprised of a densely packed and dendritically extensively interconnected neuronal network with long projections, makes the human LC to be an ideal anatomical template for early accumulation and trans-neuronal spreading of hyperphosphorylated tau. Supplementary Information The online version contains supplementary material available at 10.1007/s00401-022-02477-6.


Light sheet fluorescence microscopy (LSFM): detailed scanning parameters
Neuromelanin-content of noradrenergic (NA) neurons diffracts the laser that illuminates the focal plane. This is a major obstacle in light sheet microscopy acquisitions of the human locus coeruleus (LC), which we compensated with bilateral illumination of samples scanned in the coronal plane or by scanning the samples in the horizontal plane. Overall, to obtain the required X/Y/Z resolution, homogenous illumination within the entire focal plane and minimal photo-bleaching, each block was routinely scanned with two approaches. More homogenous illumination and moderate file sizes allowed large-scale quantifications after the first method, while higher resolution provided by the second image acquisition pipeline enabled the detailed examination of cellular structures.
In addition, selected blocks and selected regions were scanned for TH + AT8 volume co-staining in 'multicolour acquisition' mode with 6.3x zoom body (0.479 µm x 0.479 µm x 2.0 µm voxel dimensions).

3D segmentation of LC/pericoerulear (PC) complex subregions
LC core and shell, pars cerebellaris (A4) as well as subcoeruleus were 3D delineated and segmented in Imaris 9.2.1. based on anatomical information in 250-500 µm virtual coronal slices, using the manual surface creation function (Supplementary fig. 1) in both the horizontal and the coronal orientations scans. Principles and further details of segmentation are reported in the Results (3D delineation and segmentation of the human LC/PC complex). A surface (3D volume) with 4.83 µm grain size was created to cover each subregion, and new TH and AT8 channels were created for each subregion in each scan, while keeping the original voxel dimensions. The volume of each subregion in each block (in µm 3 , converted to mm 3 ) was determined.

Dorso-ventral 3D segmentation of the LC core
The segmented tube-like 'LC core' was further segmented manually to equal-size dorsal and ventral halves, using manual segmentation in 2.5 µm virtual coronal slices in each 250 µm of the horizontal orientation scans. Then, new surfaces with 4.83 µm grain size were created covering the dorsal and ventral segments, and new TH and AT8 channels were created for each segment, with keeping the original voxel dimensions (Supplementary fig. 5a-b). These subsurfaces and the respective channels were used for further quantifications.

Quantitative determination of TH + cell number in LC/PC complex subregions and in LC core segments
In order to reliably quantify TH + cell numbers in 3D, a four-step image processing pipeline was created using the ImageJ macro language to subtract local background as well as to balance TH signal intensities in the original scans (raw data). These steps were the following: (i) Adaptive background subtraction and local contrast enhancement. (ii) Images were classified as of low or high intensity, based on their mean pixel intensities. Then, adaptive background subtraction was optimized for the dynamic range of each image. (iii) The TH + immunosignal was further segmented based on a mask defined by the 25% highest intensity pixels, which included cell bodies, processes and the scattered light haloes surrounding them. (iv) Finally, based on the mask and a variance cell edge detection, 'unsharpen' function was used to reduce the haloes within the raw images, appropriately separating individual neurons (including neighbouring 'hugging' neurons), as well as enhancing the soma intensity (full script is reported as Supplementary [TH channel image processing Fiji_ImageJ script]; illustration of the process is documented in Supplementary fig. 4a-d).
The processed image stacks were then 3D cropped following the borders of the manually segmented LC/PC subregions. The ultimate identification of TH + cells was performed by the 'spot detection' function of Imaris 9.2.1., using 30 µm estimated spot diameter without background subtraction, and 'quality' filter . TH + cell density (per mm 3 subregion volume), the mean 3D nearest neighbour distance ('spot-to-spot closest distance' MATLAB XTension) and the proportional distribution of TH + cells among the subregions was determined.
Regarding LC core segments, scenes of 'dorsal' and 'ventral' core subdivisons were imported into the IMS files with processed LC core Z stacks, the TH channel was masked for the imported surfaces one-by-one, and finally spot detection was applied in the TH channel, in each segment, with exactly the same parameters as used for analysis of the entire LC core. TH + cells density (per mm 3 segment volume) and the proportional distribution of TH + cells between 'dorsal' vs. 'ventral' segments were determined.

Measuring the length and width of TH + cell bodies in the various LC/PC complex subregions
These measurements were applied in 100 randomly selected TH + cell bodies in the caudal and rostral ends of the LC core, as well as in 50 randomly selected TH + cells from the 'shell', 'A4' and 'subcoeruleus' subregions, in four randomly selected Braak 0 scans with horizontal orientation. The length and width of cells bodies were determined in Imaris 9.2.1 in 2.5 µm-thick planes in 'slice' mode, in the position of the cells when the nucleus was visible with a maximal diameter. Importantly, in each subregion, the spatial orientation of the elongated somas was considered, namely, the 'length' of cell bodies was measured in the anatomically horizontal (X/Y) plane in case of LC core, shell and A4 (in rostro-caudal orientation of cell bodies), while in the anatomically coronal (X/Z) plane in case of the subcoeruleus (in dorso-ventral orientation of cell bodies).

Quantification of the full AT8 + immunosignal volume in LC/PC complex subregions and in LC core segments
The IMS files containing surfaces and TH + AT8 channels for all subregions were downsampled and a new high-resolution surface (3D volume) was created to fully cover the AT8 immunosignal in each subregion (with 2 µm grain size without background subtraction; 'full tau' surface). The full AT8 immunosignal volume was determined as % of the entire subregion volumes. In addition, the full AT8 immunosignal volume was determined in both LC core segments, and its proportional distribution between 'dorsal' vs. 'ventral' segments was determined.

Segmentation and quantification of AT8 + cell body volumes vs. process volumes
In order to segment AT8 + cell body volumes (vs. the volume of processes), the combination of 'volume' and 'oblate ellipticity' filters were applied on the high-resolution surfaces that covered the total AT8 immunosignal volume. This selection was duplicated as a new surface, which overall covered only the AT8 + cell bodies but not the processes (though short initial dendrite parts were occasionally included in the 'cell body' surface) (Figs. 2a and 3a,h). Subtraction of 'cell body' volume from the 'full AT8 immunosignal volume' resulted in the 'process volume'. The proportion of 'cell body' and 'process' volumes was then determined in each subregion in each scan.
Determination of the number of AT8 + cell bodies in LC/PC complex subregions and in LC core segments First, surfaces covering the AT8 + cell bodies were created as described in the previous paragraph. Then, 'cell body' surfaces were masked both to AT8 and TH immunochannels, and 'spot detection' function of Imaris 9.2.1. was applied in the AT8 channel, using 30 µm estimated spot diameter without background subtraction, and 'quality' filter. Visual comparison of the newly masked AT8 and TH channels showed that virtually all AT8 + cells were also TH + in the examined LC/PC subregions. The number of AT8 + cells (number of spots) was determined and was normalized for segment volume (AT8 + cells / mm 3 volume) and for TH + cell number (AT8 + cell number as percentage of TH + cell number) in each segment. In addition, AT8 + cells were counted in all LC core segments and their proportional distribution between 'dorsal' vs. 'ventral' segments was determined, with normalization both for segment volume and TH + cell number.

Quantitative characterization of AT8 + cellular structures
In order to quantitatively characterize the AT8 + cellular structures, 15 cells were randomly selected in each category and 2 µm grain size 3D-volumes ('surfaces' in Imaris software) were created to cover (i) only cell bodies or (ii) full cell volumes including cell bodies and all traceable processes. Based on these volumes, several three-dimensional quantitative parameters were determined and statistically evaluated (Supplementary fig. 10). The descriptions below were created with the application of the Imaris software reference manual (http://www.bitplane.com/download/manuals/ReferenceManual9_2_0.pdf). Illustrations and equations provided below are all reported in the same manual.
1./ Oblate and prolate ellipticity of cell body. These numerical parameters reflect for the shape of the somas, as follows: Ellipsoid (spheroid) is a type of quadric that is a higher dimensional analogue of an Ellipse.
Illustration and equation of a standard Ellipsoid in an x-y-z Cartesian coordinate system are: Accordingly, formulas to calculate oblate ellipticity (e oblate) and prolate ellipticity (e prolate) are the following: 2./ Sphericity of cell body (Y). This numerical parameter reflects for the shape of the soma.
Sphericity is a measure of how spherical an object is. It is the ratio of the surface area of a sphere (with the same volume as the given particle) to the surface area of the particle: (Vp = volume of the particle; Ap = surface area of the particle.) The sphericity of an ideal sphere is 1.000. The closer the value of Y is to 1, the nearer the shape of the object to an ideal sphere becomes.
3./ Immunostaining intensity of cell body. The sum of the voxel intensities that are enclosed the 3D volume of the segmented cell body.

4./ Cell body volume in µm 3 .
5./ Full cell surface area in µm 2 . The surface area of the entire cell, including all traceable processes (axon and dendrites).
6./ Full cell volume in µm 3 . The volume of the entire cell, including the cell body and all traceable processes (axon and dendrites).
7./ Process volume in µm 3 . The volume of all traceable processes (axon and dendrites).

Spatial distribution of AT8 + processes in proximity of AT8 + cell bodies
Downsampled IMS files with AT8 + cell body detection (spots) and 'full tau' surface (covering the full AT8 + immunosignal volume) were opened in Imaris 9.7.2., and the 'objectobject statistics' function was activated both for the spots (representing the AT8 + cell bodies) and the surfaces (representing the full tau volumes). Then, the filter 'shortest 3D distance to spots' was applied on the 'full tau' surfaces with 50-150 µm, 150-250 µm, 250-350 µm, 350-450 µm, 450-550 µm, 550-650 µm and 650-750 µm ranges. All these selections were then duplicated as new surfaces and were colour-coded in case for each subregion. These new surfaces divided the AT8 + process volume to concentric sphere-like spatial zones around the cell bodies (the 0-50 µm spatial range was excluded, since this contained the actual cell body volumes). Surface volumes (µm 3 ) in all spatial ranges were determined, and the proportion of tau process volumes in these spatial zones was calculated as percentage of the full tau process volume. In case of the shell, A4 and subcoeruleus, Braak 0 blocks without detected AT8 + cell bodies were excluded.

Spatial distribution analysis of AT8 + cells I: calculation of 'nearest neighbour index' (NNI) in
the LC core LC core scans with AT8 + cell body detection were opened in Imaris 9.2.1. X/Y/Z coordinates of spots representing the geometrical centre of AT8 + cell bodies were extracted, and NNI was calculated for each dataset in MATLAB (full script is reported as Supplementary [NNI calculation MATLAB script]. NNI is an estimate of spatial distribution and indicates whether a point pattern is dispersed (NNI > 1), or clustered (NNI < 1) [1]. NNI was calculated as the ratio of the 'actual (real) average nearest-neighbour distance' to a 'simulated average nearest-neighbour' distance. The 'simulated average nearest-neighbour distance' was calculated by randomly distributing the same number of cells as the actual cell population over the same 3D reference frame (Monte Carlo simulation), followed by calculating the average nearest-neighbour distance based on 1000 repeated simulations [2]. . Paired Student's t-test was applied for the evaluation of simulated vs. actual nearest neighbour distances.

Spatial distribution analysis of AT8 + cells II: determination of 'dense cells', and in-depth
cluster analysis in the LC core LC core scans with AT8 and TH channels as well as with AT8 + and TH + cell body spot detections were opened in Imaris 9.7.2. The 3D nearest neighbour distance for both AT8 + and TH + cells (represented by spots) were determined by the 'spot-to-spot shortest distance' MATLAB XTension.
Then, first, those AT8 + cells that were closer to each other than 75% of the average nearest neighbour distance were segmented. These cells were defined as 'dense AT8 + cells'.
Then, 'dense cells' were colour-coded and their spatial distribution was further explored and characterized by the 'split spot' MATLAB XTension. Segmentation and 3D visualization of dense AT8 + cells revealed 'duo cells' and 'minigroups' (i.e., 3-9 AT8 + cells where each cell is closer to its nearest neighbour than the 75% of the average nearest neighbour distance). The number of 'duos' and 'minigroups' was determined in each scan and were statistically evaluated by one-way ANOVA.
Second, we determined the subset of those AT8 + cells that are in equal or shorter distance to the nearest neighbour AT8 + cell than the average nearest neighbour distance of the TH + cells in the actual scan. Since virtually all AT8 + cells were also TH + in the LC core of the examined scans, this subset defines the (statistically) immediate neighbouring AT8 + NA neurons ('neighbouring cells'). The number of 'neighbouring cells' per 100 AT8 + cells, as well as the percentage of 'neighbouring cells' among all 'dense cells' were determined and were statistically evaluated by one-way ANOVA.

Tracing long AT8 + axons in Braak stage 0 cases
Long AT8 + axons in scans from the Braak 0 subjects #3 and #5 were traced and 3D reconstructed by the filament tracer module of Imaris 9.2.1., using the 'autopath' algorithm with 40 µm 'starting point diameter' and 5.66 µm 'seeding point diameter'. Cell bodies were defined as starting points of the created filaments.

Supplementary fig. 1 Three-dimensional delineation and segmentation of the LC/PC complex.
Subregions were manually segmented based on cell densities and soma orientations, using 250-500 µm-thick coronal optical sections with TH immunochannel. (a-b) segmentation of the LC core; blue circles represent the borders of the LC core in every 250 µm. (c) 3D reconstructed subregions in a representative block; segmented subregions were handled as separated volumes (surfaces) indicated here by different colours. Within each surface, new channels for AT8 and TH volume staining were created in Imaris. (d-e) The same block is shown from lateral view before segmentation (d) and after segmentation (e). Arrows in d and e indicate an artery entering the LC core. Scale bars are indicated in each micrograph. A subcoeruleus NA axon that joins the dorsal NA bundle. Red dot denotes a cell body that sends its axon (arrowheads) in the rostral direction. (d). Dorso-ventral soma orientation of subcoeruleus neurons (250 µm-thick coronal optical slice) (e). Subcoeruleus neurons are often organized in dorso-ventral 'columns' (arrowheads in a 100 µm-thick coronal optical slice) (f). (g-j) Demonstration of A4 (pars cerebellaris of the LC). A4 subregion (red surface) is located dorsally from the LC core (blue surface) (g). A4 neuron cell bodies and processes around the mesencephalic trigeminal tract (me5), horizontal view (h). Fusiform A4 neurons (blend 3D rendering mode) (i). 'Hugging' A4 neurons, blend 3D rendering mode (j). Scale bars are indicated in each micrograph.

Supplementary fig. 8 Exploring the potential impact of age on the accumulation of tau cytoskeletal pathology in the LC core.
(a) Age of subjects at death. N = 6 (B0), N = 6 (B1-2), N = 5 (B3-4), N = 3 (B6). Data are expressed as mean ± SEM; *P < 0.05. (b) Summary table of ANCOVA analysis data. Significant results are represented in red. (c-h) Pearson R correlation analyses between the subjects' age at death and (i) the full AT8 + immunostaining volume normalized for LC core volume (c-d); (ii) the number of AT8 + cells normalized for LC core volume (e-f); as well as (iii) the number of AT8 + cells normalized for TH + cell number (g-h). Correlation analyses were performed with the involvement of all subjects from all Braak stages (c, e and g) and with the involvement of Braak 0 subjects only (d, f and h). Green dots: Braak 0 subjects; blue dots: Braak 1-2 subjects; orange dots: Braak 3-4 subjects; red dots: Braak 6 subjects.   fig. 9 Representative overview of AT8 volume immunostaining in the LC core. (a-d) 1 x 1 x 2 mm large-scale 3D crops from a Braak 0 (a), a Braak 1 (b), a Braak 3 (c) and a Braak 6 (d) subject. Note the increased density of AT8 + cells and their processes, associated with the more advanced Braak stages. Also note the clustering tendency of AT8 + cells, which is well recognizable in b and c. Scale bars are indicated in each micrograph.    (a-e) Cells with intact processes and inhomogeneously filled somas: variations for the somatic distribution of AT8 immunoreactivity. Strong, granular AT8 staining along the dendrite/axon initial parts (a), along the somatic cell membrane (b), as granules throughout the somatic cytoplasm (c), as large granules in the somatic cytoplasm (d), and as a large compact granule that almost fills up the entire somatic cytoplasm (e). (f-k) Proposed transition forms and variations of AT8 + cellular structures. Proposed transition between 'cell with low intensity staining' and 'cell with inhomogeneously filled soma'. Arrowheads: axon (f). Cell with partially fragmented, somatic protrusions (in the middle of the micrograph); *: a mature tanglelike cell (g). The same cell as in the middle of g but as 2D projection at higher magnification. Arrows: dendrites; arrowheads: somatic protrusions (h). Cell with inhomogeneously filled soma and fragmenting dendrites. Arrowheads: fragmenting dendrites (i). Cell with heavily atrophic dendrites and fragmenting somatic protrusions (j). Cell with severely atrophic processes close to a matured tangle-like cell in morphology (k). (l) Semiquantitative assessment of AT8 + cellular structures in the pericoerulear area (shell, A4 and subcoeruleus). Scale bars are indicated in each micrograph.  LC shell exhibits low TH + cell density but a dense plexus of TH + processes, mainly dendrites.

Supplementary fig. 14 Visual explanation for segmentation and clustering of 'dense AT8
Ventrally from the sparse shell TH + neurons, long TH + axons of the dorsal noradrenergic bundle are shown.
A4 consisted of sparse TH + cells with distinctly elongated somas. Their processes form a complex plexus around the mesenchephalic trigeminal tracts (me5). These processes are shown between 0:15 -0:30. in details. Me5 is segmented and represented by a white surface between 0:00 -0:07.

Video #4 Tau cytoskeletal pathology in the LC of a Braak NFT stage 0 brain.
Representative 3D crop from the LC core. AT8 channel. 2.42 µm x 2.42 µm x 2.50 µm voxel dimensions.
Sparse AT8 + cells with inhomogeneously or homogeneously filled somas and intact processes are dominating. AT8 + axon of a tau-bearing neuron is followed until the end of the 3D crop in between 0:47 -1:06.

Video #5 Tau cytoskeletal pathology in the LC of a Braak NFT stage 1 brain.
Representative 3D crop from the LC core. AT8 channel. 1.51 µm x 1.51 µm x 2.00 µm voxel dimensions.
Strongly filled cells with partially atrophic dendritic tree are dominating. However, several transition forms are noticed in this sample.

Video #6 Tau cytoskeletal pathology in the LC of a Braak NFT stage 3 brain.
Representative 3D crop from the LC core. AT8 channel. 1.51 µm x 1.51 µm x 2.00 µm voxel dimensions.
Strongly filled cells with severely atrophic processes are dominating. Zoom to examples of strongly filled cells with severely atrophic processes between 0:55 -1:00.

Video #8 Dorso-ventral distribution of AT8 + volume immunostaining in the LC.
Segmented LC core from a Braak NFT stage 3 brain. TH (gray) and AT8 (red)  halves of the LC core, respectively. Blue surface covers the entire LC core volume. Note the more abundant AT8 volume staining in the dorsal half.

Video #9 Demonstration of dense cells.
3D crop from the LC core of a Braak NFT stage 2 brain. TH (gray) and AT8 (red) channels.