Pathogenic tau modifications occur in axons before the somatodendritic compartment in mossy fiber and Schaffer collateral pathways

Formalin-fixed temporal lobe free-floating sections (40 μm) from ND (n = 31) and MCI (n = 13) cases were obtained de-identified from the Banner Sun Health Research Institute Brain and Body Donation Program (Sun City, AZ) [7]. Most subjects received annual standardized cognitive assessments prior to death. If death occurred prior to a standard assessment, cognitive status was determined by postmortem informant telephone questionnaire and private medical records review followed by a consensus diagnostic conference of Program neuropsychologists and neurologists. Global neuropathological examination methodology was previously described [7] and included assignment of Braak stage, CERAD neuritic plaque density [55], and National Institute on Aging-Reagan Institute AD probability level (0 = not AD; 1 = low; 2 = intermediate) [57]. Semi-quantitative scores (0–3) for global amyloid plaque and neurofibrillary tangle densities were obtained for frontal, parietal, and temporal neocortex as well as entorhinal and hippocampus regions; the sum of all regional scores are termed “global plaque density” and “global tangle density” scores. Table 1 summarizes the clinical, demographic, and global measures of neuropathology. The donation program also features a standing 24–7 response team that allows exceptionally low post-mortem intervals (PMIs) with a median PMI of 3.2 h for all 1900+ cases collected since 1988 [7].

Temporal lobe sections were immunohistochemically stained as previously described [18, 40, 41] to visualize the pattern of AT8 phosphorylation, PAD exposure, and Aβ pathologies using the monoclonal AT8 (Thermo MN1020), TNT2 (Kanaan lab) [18, 19], and MOAB2 (Kanaan lab, originally created by Dr. Lester Binder at Northwestern) [77] antibodies, respectively. Primary antibodies were diluted in tris-buffered saline (TBS; 150 mM NaCl, 50 mM Tris, pH 7.4) containing 2% goat serum and 0.1% Triton X-100 at 1:16,000 for AT8, 1:400,000 for TNT2, and 1:4000 for MOAB2. Immunoreactivity was detected using biotinylated goat-anti-mouse IgG (H + L) secondary antibody (Jackson ImmunoResearch Laboratories 115–065-166) diluted in TBS + 2% goat serum + 0.1% Triton X-100, VectaStain Elite ABC-HRP Kit (Vector Laboratories PK-6100), and 3,3′-diaminobenzidine supplemented with 0.25% ammonium nickel (II) sulfate hexahydrate (Sigma A1827). All sections were counterstained with cresyl violet before being mounted on microscope slides and coverslipped with Cytoseal 60 (Thermo Scientific, #8310–16). Tissue sections from each case were processed simultaneously for each antibody to eliminate inter-run staining variability. Primary antibody delete controls were run using the same protocol with the exception that the primary antibody was omitted. As expected, the primary deletes produced no staining (Additional file 1: Figure S1).

The unbiased stereological spaceballs probe was used to estimate the total length of neurites in single hippocampal body sections from each case stained with AT8 and TNT2 in the CA3 Str. Luc. layer (i.e. mossy fibers) and the CA1 Str. Rad. layer (i.e. Schaffer collaterals). The CA3 Str. Luc. was defined using fiduciary neuroanatomical landmarks, including the CA3 pyramidal cell layer dorsally, Str. Rad. of CA3 ventrally, the CA2 medially, and hilus laterally. The CA3 pyramidal layer was defined using fiduciary neuroanatomical landmarks, including the CA3 Str. Luc. dorsally, stratum oriens ventrally, CA2 medially, and hilus laterally. The CA1 Str. Rad. was defined using fiduciary neuroanatomical landmarks, including the CA1 pyramidal cell layer dorsally, stratum lacunosum-moleculare ventrally, subiculum medially, and CA2 laterally. The DG granule cell layer was defined using fiduciary neuroanatomical landmarks, including the hilus dorsally and the molecular layer ventrally, and is clearly defined by cresyl violet staining due to cell density and size. If specific subregions were not reliably identifiable within the sections, the case was not used for analyses requiring that region (5, 2, 0, and 5 cases were excluded from the CA3 Str. Luc., DG, CA1 Str. Rad. and CA3 analyses, respectively). A hemisphere probe with a radius of 8 μm was used to sample sites throughout each region. Mounted tissue thicknesses ranged from ~ 11–14 μm (≥70% shrinkage in the z-plane is typical after similar processing of free-floating sections [24, 26, 54]) across all cases and regions analyzed. A 4x objective was used to outline each contour and a 60x oil immersion objective (numerical aperture = 1.35) was used for making the stereological measurements. Local neurite density was calculated by dividing the estimated total axon length by the volume of the region of analysis, and neurite density was used for comparisons. Local somata staining was quantified by total enumeration in the CA3 pyramidal cell layer (i.e. Schaffer collateral pathway) and DG layer (i.e. mossy fiber pathway) of the same sections from above for neurite analyses using 10x magnification and manually counting the cell bodies displaying immunoreactivity. Total cell numbers were used for comparisons. Brightfield images were acquired on a Nikon Eclipse 90i microscope equipped with a Nikon DS-Ri1 camera and processed using Nikon NIS-Elements software.

Hippocampal sections from a subset of the 44 cases (n = 12 cases with high and low tau pathology) were double labeled with TNT2 (mouse IgG1, 1:8000) and biotinylated AT8 (mouse IgG1-biotin, 1:800, Thermo Scientific MN1020B) using methods similar to those previously published [39, 73]. Each primary antibody incubation period was overnight at 4 °C. TNT2 immunoreactivity was detected using AlexaFluor 647-conjugated goat-anti-mouse IgG (H + L) Fab fragments (Jackson ImmunoResearch Laboratories 115–547-003 and 115–607-003) and AT8 immunoreactivity was detected using AlexaFluor 568-conjugated streptavidin (ThermoScientific Pierce S11226). Sections were blocked with unconjugated goat-anti-mouse IgG (H + L) Fab fragments (Jackson ImmunoResearch Laboratories 115–007-003) to prevent cross-labeling of the secondary antibodies between each primary and subsequent secondary antibody incubations. Nuclei were counterstained in the sections by including DAPI (1 μg/ml; ThermoFisher, D1306) in the first rinse after the last detection secondary antibody step. Control sections included omission of each individual primary antibody. As expected, the individual primary delete sections did not produce cross-reaction of signals in the deleted antibody channel (Additional file 2: Figure S2).

In a second and third staining series, hippocampal sections were triple-labeled with either biotinylated AT8 (as above) or TNT2 (as above) and both SMI-312 (mouse IgG1, 1:1000, Biolegend 837,904) and MAP2 (rabbit polyclonal, 1:300, Cell Signaling 8707) to colocalize each of these tau pathologies with an axonal (SMI-312) and dendritic (MAP2) marker. All tissues were stained using a similar protocol to those previously published [39, 73]. Briefly, all sections were incubated overnight at 4 °C in SMI-312 and MAP2 primary antibodies and the following day these primaries were labeled with AlexaFlour 568 goat anti-mouse IgG (H + L) (1:500; ThermoFisher, A11031) and AlexaFlour 647 goat anti-rabbit (1:500; ThermoFisher, A21245) secondary antibodies, respectively. The tissue sections used for the TNT2/SMI-312/MAP2 series were blocked in 2% mouse serum (Invitrogen, 10,410) to saturate open binding sites on the first anti-mouse secondary antibody) for 1 h, followed by an hour incubation in goat anti-mouse whole molecule (1:50, Jackson ImmunoResearch, 115–008-003) to block binding sites on mouse IgGs. After blocking was completed, the tissue sections were incubated in TNT2 primary antibody overnight at 4 °C. The AT8/SMI-312/MAP2 series sections were incubated in biotinylated AT8 antibody (biotinylation precludes the need for the above blocking) overnight at 4 °C. The following day the TNT2 or AT8-biotin primary antibodies were labeled with goat anti-mouse IgG1-specific AlexaFlour 488 (1:500; ThermoFisher, A21121) or streptavidin conjugated to AlexaFluor 488 (1:500; ThermoFisher, S11223). After immunolabeling the tissues were counterstained with DAPI as above before mounting and coverslipping. Control sections included omission of each tau primary antibody. As expected, omission of the primary antibodies did not produce cross-reaction of signals in the deleted antibody channel confirming the tau localization with SMI312 and MAP2 was due to specificity of tau labeling (Additional file 3: Figure S3). Colocalization between tau markers (AT8 or TNT2) and either SMI-312 or MAP2 in neurites within the CA3 Str. Luc. and CA1 Str. Rad. was determined using these sections.

After staining, sections were mounted on microscope slides and autofluoresence of the tissue was blocked by treating with 2% Sudan Black B before coverslipping with VectaShield Hard Set mounting medium (Vector Laboratories H-1000). All IF images were obtained using a Nikon A1+ scanning confocal microscope system. Z-stacks were acquired in 0.5 μm steps at 60x magnification and images for figures were generated with a maximum intensity projection or slices view (for cross-sectional analysis for colocalization) using NIS-Elements software.

All data were analyzed using Prism v7.0 software (GraphPad). Stereological estimate outcomes were analyzed for normality using the D’Agostino and Pearson normality test. All data sets were not normally distributed, and subsequently, non-parametric statistical analyses were performed. Cell number and Aβ plaque data sets were Log(x + 1) transformed for correlations because values of zero were present. All correlation comparisons were performed using the Spearman rank correlation and significant p-values were adjusted to ≤0.005 from ≤0.05 because 10 comparisons were performed within a given pathway (i.e. DG-mossy or CA3-Schaffer) for each immunostain (i.e. AT8 or TNT2) (Tables 4 and 7). Demographic variables were compared between clinical diagnostic groups using Mann-Whitney, Fisher’s exact, or the Chi-squared test. Statistical significance was set at p ≤ 0.05. Adobe Photoshop and Adobe Illustrator programs were used to compile images, graphs and text into final figures.

Demographic, clinical, and global neuropathological results for the 44 cases used in this study are summarized in Table 1. Notably, no significant differences were observed between the ND and MCI groups for age (p = 0.13), sex (p > 0.99), postmortem interval (PMI: mean = 2.7 + 0.6 h, range = 1.5–4.8 h, p = 0.84), Mini-Mental State Examination score (MMSE, p = 0.34), Braak stage (p = 0.85), NIA-Reagan AD probability level (p = 0.86) or CERAD plaque density (p = 0.96).

We measured the amount of AT8+ immunoreactivity (Fig. 1) in the DG-mossy fiber pathway of the hippocampus. In this pathway, the granule cells (where cell body pathology was measured) give rise to axonal projections, known as mossy fibers, that terminate in the CA3 Str. Luc. (where neurites were measured). All cases displayed AT8+ neuropil threads in the CA3 Str. Luc., however, AT8+ staining in the corresponding cell bodies of the DG was less common (Fig. 1a). Specifically, 16.7% of cases (7 of 42) displayed no observable AT8+ staining in the DG cell bodies, but all of these cases showed AT8+ neurite staining in the CA3 Str. Luc. (Table 2). Axonal AT8 neurite density showed a significant positive correlation with the number of AT8+ cell bodies of the mossy fiber pathway (Spearman r = 0.640, p < 0.0001, Fig. 1b). In cases lacking AT8+ cell bodies in the DG, the observable mossy fiber pathology ranged from sparse (Fig. 1c) to moderate (Fig. 1d), and cases with cell body staining were never without neurite staining.

Next, we analyzed the amount of TNT2+ immunoreactivity in the DG-mossy fiber pathway (Fig. 2 and Table 3) as a measure of PAD-exposed tau, an early pathological event in tauopathies [18]. The majority of cases displayed TNT2+ staining in the CA3 Str. Luc. (83%; 34 of 41; Fig. 2a). Additionally, 17 cases (43.6%) contained no observable TNT2+ DG neurons, and among these cases 10 contained TNT2+ neurite pathology in the CA3 Str. Luc. mossy fibers and 7 did not contain TNT2+ neurites. In contrast to AT8 pathology, there were cases that contained no observable TNT2+ pathology in the DG cell bodies or CA3 Str. Luc. layer (17%; 7 of 41). Axonal TNT2+ neurite staining in the mossy fiber pathway displayed a significant positive correlation with the number of TNT2+ DG cell bodies (Spearman r = 0.702, p ≤ 0.0001, Fig. 2b). In cases lacking TNT2+ cell bodies in the DG, the observable mossy fiber pathology ranged from sparse (Fig. 2c) to moderate (Fig. 2d), and cases with cell body staining were never without neurite staining.

The local AT8+ or TNT2+ neurite and cell densities in the DG-mossy fiber pathway were correlated with demographic or global neuropathology scores (Table 4). Local AT8+ and TNT2+ neurite densities showed a significant positive correlation with age (AT8+ and TNT2+, NIA-Reagan (AT8+ only), global tangle density (TNT2+ only) and Braak Stage (TNT2+ only) in the mossy fiber pathway. Local AT8+ and TNT+ neurite densities in the mossy fiber pathway were not correlated with MMSE or PMI (data not shown). In the DG cell layer, AT8+ cell number showed a significant positive correlation with Braak stage and global tangle density, while TNT2+ cell number did not correlate with any measure. Finally, no statistical differences were observed between AT8+ or TNT2+ neurite density in the CA3 Str. Luc. when compared by either diagnosis states (ND or MCI) or sex groups (Additional file 4: Figure S4).

We measured the amount of AT8+ immunoreactivity (Fig. 3) in the CA3-Schaffer collateral pathway of the hippocampus. In this pathway, the CA3 pyramidal neurons (where cell body pathology was measured) give rise to axonal projections, known as Schaffer collaterals, that terminate in the CA1 Str. Rad. (where neurites were measured). All cases displayed AT8+ neuropil threads within the CA1 Str. Rad., however, AT8+ cell bodies within the CA3 pyramidal layer were less common (Fig. 3a). All cases showed AT8+ neurite staining in the CA1 Str. Rad. region, but in 12.8% of cases (5 of 39) there was no observable AT8+ cell bodies in the CA3 pyramidal layer (Table 5). Axonal AT8 neurite staining displayed a significant positive correlation with the number of AT8+ cell bodies of the Schaffer collateral pathway (Spearman r = 0.648, p ≤ 0.0001, Fig. 3b). Importantly, in the absence of observable cell body pathology, the terminal region contained a range of tau pathology from relatively sparse (Fig. 3c) to relatively dense neurite staining (Fig. 3d), and cases with cell body staining were never without neurite staining.

Next, we analyzed the amount of TNT2+ immunoreactivity in the CA3-Schaffer collateral pathway (Fig. 4) as a measure of PAD-exposed tau [18]. All but one case displayed TNT2+ neurite staining in the CA1 Str. Rad. (Fig. 4a). Specifically, 25.6% of cases (10 of 39) displayed TNT2+ neurite staining in the CA1 Str. Rad., but no observable TNT2+ staining in the CA3 pyramidal cell body layer (Table 6). Axonal TNT2+ neurite staining in the Schaffer collaterals showed a significant positive correlation with the number of TNT2+ CA3 cell bodies (Spearman r = 0.719, p ≤ 0.0001, Fig. 4b). In cases displaying no TNT2+ cell bodies, the TNT2+ neurite pathology in the terminal region of this pathway contained pathology ranging from relatively sparse (Fig. 4c) to relatively dense (Fig. 4d). Cases with TNT2+ cell bodies staining were never without neurite staining.

Finally, the local AT8+ or TNT2+ neurite and cell densities in the CA3-Schaffer collateral pathway were correlated with demographic or global neuropathology scores (Table 7). Local AT8+ and TNT2+ neurite densities showed a significant positive correlation with Braak stage, NIA-Reagan, and global tangle density in the Schaffer collateral pathway, while AT8+ neurites were correlated with age but TNT2+ neurites were not. Local AT8+ and TNT+ neurite densities in the Schaffer collateral pathway were not correlated with MMSE or PMI (data not shown). In the CA3 pyramidal cell layer, AT8+ cell number showed a significant positive correlation with Braak stage, global tangle density and NIA-Reagan score, while TNT2+ cell number did not correlate with any measure. No statistical differences were observed between AT8+ or TNT2+ neurite density in the CA1 Str. Rad. when compared to either diagnosis states (ND or MCI) or sex (Additional file 5: Figure S5).

To confirm that the AT8+ and TNT2+ pathologies colocalized in the current tissue cohort we used multi-label immunofluorescence. As previously observed [41], extensive colocalization was observed between AT8 and TNT2 in hippocampal pyramidal neurons (Fig. 5). Here, we demonstrate that extensive colocalization occurs between AT8 and TNT2 in both CA3-mossy fiber and CA1-Schaffer collateral neurites within the hippocampus. Primary delete control sections where one tau antibody was omitted produced the expected pattern of staining confirming the specificity of each label (Additional file 2: Figure S2).

Next, we used multi-label immunofluorescence to determine whether AT8 and TNT2 pathologies in the DG-mossy fiber and CA3-Schaffer collateral pathway were axonal (i.e. SMI-312+ neurites) or dendritic (i.e. MAP2+ neurites). The colocalization between these markers was performed using high magnification z-stacks that were taken in the CA3 Str. Luc (i.e. mossy fibers) and CA1 Str. Rad (i.e. Schaffer collaterals). In the CA3 Str. Luc. region (mossy fibers), AT8+ (Fig. 6) and TNT2+ (Additional file 6: Figure S6a-b) neurites showed moderate colocalization with SMI-312, an axonal marker. Similarly, AT8+ (Additional file 6: Figure S6c-d) and TNT2+ (Fig. 7) neurites showed moderate colocalization with SMI312, in the CA1 Str. Rad. region (Schaffer collaterals). In contrast, little to no colocalization was observed between AT8 or TNT2 and MAP2+ neurites in the images analyzed within the CA3 Str. Luc. and CA1 Str. Rad. regions. It is noteworthy that numerous AT8+ (Figs. 6d) and TNT2+ (Figs. 7d) neurites did not colocalize with either SMI-312 or MAP2.

Finally, we stained for Aβ plaques in the CA3-Schaffer collateral and DG-mossy fiber pathway regions using the MOAB2 antibody [77] and counted local plaques using total enumeration (Fig. 8a). The majority of cases did not contain detectable MOAB-2+ Aβ pathology in the local regions of the hippocampal formation assessed (i.e. DG cell layer, CA3 Str. Luc, CA3 cell layer, and CA1 Str. Rad.). Specifically, 84.2% contained zero plaques in the CA3 Str. Luc. and 59% contained zero plaques in the DG, while 55.8% had zero plaques in the CA1 Str. Rad. and 78.9% contained zero plaques in the CA3 pyramidal cell layer (Table 8). Intracellular Aβ pathologies (e.g. monomeric or soluble oligomeric Aβ species) were not observed in the DG or CA3 neurons of any case used in this study.

Interestingly, neither AT8 phosphorylation nor PAD exposed tau correlated with the presence of local Aβ plaques in both the Schaffer collateral (Spearman r = − 0.182, p = 0.242 for AT8, and r = 0.026, p = 0.867 for TNT2, Fig. 8b, c) and mossy fiber pathways (Spearman r = 0.293, p = 0.149 for AT8, and r = 0.086, p = 0.614 for TNT2, Fig. 8d, e). The lack of correlation between AT8+ or TNT2+ neurite pathology in these regions remained even when cases with no global Aβ pathology (i.e. non-PART) were excluded (data not shown), with exception of a significant negative correlation between AT8+ neurite density and local plaque number in the Schaffer collaterals (Spearman r = − 0.379, p = 0.039). The extent of both AT8+ and TNT2+ tau pathology in some cases was remarkably robust in regions devoid of observable Aβ plaques (Fig. 8f-h). Importantly, the cases used here display the expected ranges of global Aβ pathologies consistent with aged ND and MCI cases, with the vast majority of cases displaying some degree of global Aβ pathology accumulation (31 of 44 cases; see Tables 1, 4 and 7). Taken together, these results indicate that AT8 phosphorylation and PAD exposure occur independently of the presence of Aβ plaque pathology in the Schaffer collateral and mossy fiber pathways of the hippocampus.

Recent work described PART where brains contain AD-like NFT pathology, but no detectable Aβ pathology [21]. The current cohort of cases provides an opportunity to evaluate the earliest deposition of AT8+ and TNT2+ tau pathology deposition in the DG-mossy fiber and CA3-Schaffer collateral pathways of the hippocampus. Cases were separated into PART (global plaque density = 0) or non-PART groups (global plaque density > 0). Both PART and non-PART groups contained 9 Braak I-II stage cases, while the PART group contained only 4 Braak stage III and the non-PART group contained 22 Braak stage III cases. Non-PART cases showed significantly more AT8+ neurites in the CA3 Str. Luc. (mossy fibers; Fig. 9b; p = 0.0005) and CA1 Str. Rad. (Schaffer collaterals; Fig. 9d; p = 0.0095), as well as AT8+ cells in the DG (Fig. 9a; p < 0.0001) and CA3 (Fig. 9c; p = 0.0011) neuron layers (Mann-Whitney U tests). Non-PART cases showed significantly more TNT2+ neurites in the CA3 Str. Luc. (mossy fibers; Fig. 9f; p = 0.0021) and CA1 Str. Rad. (Schaffer collaterals; Fig. 9h; p = 0.0001), but TNT2+ cells in the DG (Fig. 9e; p = 0.0965) and CA3 (Fig. 9g; p = 0.4557) neuron layers were not significantly different between PART and non-PART cases (Mann-Whitney U tests). Interestingly, DG neurons appear relatively spared in PART cases compared to non-PART cases as several lacked detectable somata pathology.