Neuroinflammation is increased in the parietal cortex of atypical Alzheimer’s disease

Post-mortem brain tissue was obtained from the Netherlands Brain Bank (NBB; Amsterdam, The Netherlands). Donors signed informed consent for brain autopsy, and the use of brain tissue and medical records for research purposes. Neuropathological diagnosis was based on histochemical stainings including hematoxylin and eosin, congo red staining, Bodian or Gallyas and methenamine silver stainings, and immunohistochemical stainings for amyloid-beta, pTau, alpha-synuclein, and p62. These stainings were performed on formalin-fixed paraffin-embedded (FFPE) brain tissue of multiple brain regions including the frontal cortex, temporal pole, superior parietal lobe, occipital pole, amygdala, and the hippocampus. Neuropathological diagnosis of AD was based on Braak stages for NFT and amyloid [7], Thal phases for amyloid-beta [8], and CERAD criteria for neuritic plaques [26].

Between 1996 and 2014, 352 AD cases came to autopsy and were semi-quantitatively scored by two neuropathologists (WK, AR) for the NFT load using Bodian or Gallyas staining in the temporal pole, the frontal, superior parietal, and occipital cortex as previously described by Hoogendijk et al. [27]. The NFT load was scored in a 0.4-mm² area as being absent (0), sparse (1), mild (2; 2 to 3 NFTs), or severe (3; > 3 NFTs) for each brain region separately. From this cohort, we selected cases with an NFT score of 3 in either the temporal or parietal section, or in both sections, resulting in 296 cases (for flowchart, see Fig. 1). For 142 cases, the NFT score was higher in the temporal section than the parietal section. These cases were referred to as having a typical NFT distribution [7]. In 126 of 296 cases, an NFT score of 3 was found in the temporal as well as the parietal section. A higher NFT score in the parietal compared to the temporal section was observed in 28 cases and was defined as a parietal dominant and thus atypical NFT distribution.

To study the distribution of neuroinflammation in typical and atypical AD, we further refined our cohort to include only cases with a concordance between clinical presentation and NFT distribution. The clinical phenotype of 36 cases with atypical and typical NFT pathology was retrospectively assessed. In 18 out of 28 cases with an atypical NFT distribution, the available clinical information was sufficient to come to a retrospective clinical diagnosis (see Table 1 for demographics). To have an equal group for comparison, we randomly chose 18 cases with both typical NFT distribution and sufficient clinical information for clinical phenotyping. Clinical assessment was performed retrospectively and independently by 2 cognitive neurologists (YP and FB) using the NIA-AA criteria [2]. Both clinicians were blinded to the pathological stratification when assessing the clinical phenotype. The clinical stratification was based on (collateral) history and cognitive examination documented by the clinical neurologist. Cases with first-degree relatives affected by EOAD were excluded to minimize the risk of genetic AD. Other exclusion criteria were sepsis, other neurodegenerative or psychiatric diseases, significant cerebrovascular disease, post-mortem interval > 12 h, and prior known genetic mutations. Due to our exclusion criteria and availability of archived brain tissue samples, our inclusion was limited to 9 atypical AD cases and 10 typical AD cases of which the temporal pole and superior parietal lobe were assessed by immunohistochemistry (see Fig. 1 for inclusion flowchart, Table 2 for patient details, and Table 4 for demographics in the “Results” section). Cases were not intentionally matched for disease duration, brain weight, ApoE status, or post-mortem interval.

IHC was performed to detect pTau (AT8); amyloid-beta (N-terminal; IC16); reactive astrocytes (GFAP); microglia (Iba1); activated microglia (CD68 and HLA-DP/DQ/DR); and complement proteins C1q, C3d, C4b, and C5b-9 (Table 3). FFPE sections (5-μm thick) from the temporal pole and superior parietal lobe of the right hemisphere were used.

IHC for pTau, amyloid-beta, GFAP, and Iba1 was performed using the Ventana BenchMark ULTRA staining system (Roche, Basel, Switzerland). Tissue sections were mounted on TOMO adhesive glass slides (Matsunami, Osaka, Japan) and deparaffinized. After blocking for endogenous peroxidase, antigen retrieval was performed by heating sections at 100 °C in Cell Conditioning 1 solution (pH 8.5) (Roche) for different durations per antibody (see Table 3). For detection of primary antibodies with 3,3′-diaminobenzidine tetrahydrochloride (DAB), Optiview DAB IHC detection kit (Roche) was used. Finally, the sections were mounted with Coverslipping film (Sakura Tissue-Tek, Leiden, The Netherlands).

IHC for CD68, HLA-DP/DQ/DR, C1q, C3d, C4b, and C5b-9 was performed manually. The sections were mounted on SuperFrost Plus glass slides (Menzel-Gläser, Braunschweig, Germany) and deparaffinized. Subsequently, the sections were blocked for endogenous peroxidase using 0.3% hydrogen peroxide in phosphate buffer saline (PBS; pH 7.4). The sections were immersed in sodium citrate buffer (10 mM sodium citrate, 5 M NaOH, dH₂O, pH 6.0) and heated to 120 °C in an autoclave for antigen retrieval. Primary antibodies were diluted in normal antibody diluent (ImmunoLogic, Duiven, The Netherlands) and incubated overnight at 4 °C. Primary antibodies were detected using EnVision (Dako, Glostrup, Denmark). Between steps, the sections were washed in PBS. Subsequently, antibodies were visualized with DAB (Dako). After counterstaining with hematoxylin, the sections were mounted with Entellan (Merck, Darmstadt, Germany).

For quantitative assessment, 2 regions of interest (ROI) were randomly selected within non-curved areas of each section containing all 6 cortical layers [28]. Within each ROI, contiguous microscopic fields arranged in columns perpendicular to the cortical surface of the cortex were photographed. Total surface, depending on the width of the cortex, could vary for each ROI and contained at least 2 columns. Images were taken using a × 10 objective on an Olympus BX 41 photomicroscope with a Leica MC 170 HD digital camera. The presence of DAB staining was quantified with ImageJ (NIH) using the color threshold plugin. Our outcome measurement was the percentage of DAB-stained area per marker, also referred to as immunoreactivity. In addition to immunoreactivity, we quantified the number of amyloid-beta and C4b plaques. For the amyloid-beta plaques, diffuse deposits were not taken into account and only dense plaques were quantified, defined as particles with an immunoreactive surface area of 100 μm² or more [29, 30]. C4b-positive deposits of the same surface area were quantified as a measurement of the atypical appearing plaques as described in the “Results” section.

Co-localization of C4b and CD68 or HLA-DP/DQ/DR with amyloid-beta and thioflavine S was visualized in the parietal section of 4 atypical AD cases and the temporal section of 2 typical AD cases. The typical AD cases served as positive controls and reference since localization of complement and microglia in classical-cored plaques is widely described in literature (for CD68 [20]/for complement [16, 19]).

After deparaffinization, the sections were submerged in sodium citrate buffer and heated to 120 °C in an autoclave. Subsequently, the sections were incubated using different combinations of primary antibodies: mouse IgG2a-anti-amyloid-beta (1:200), rabbit-anti-C4b (1:200), and mouse IgG1-anti-HLA-DP/DQ/DR (1:25) or mouse IgG1-anti-CD68 (1:300). Antibodies were diluted in normal antibody diluent (ImmunoLogic) and incubated overnight at 4 °C. Subsequently, the sections were incubated with the following secondary antibodies: goat-anti-mouse IgG2a Alexa Fluor dye 594, goat-anti-mouse IgG1 Alexa Fluor dye 647, and donkey-anti-rabbit Alexa Fluor dye 647 (1:250 dilution, ThermoFisher, Waltham, USA). For visualization of amyloid structures, the sections were counterstained with thioflavine S (1% in dH₂O) and subsequently rinsed in 70% ethanol. Autofluorescence was blocked with 0.1% Sudan black in 70% ethanol for 5 min. Between steps, the sections were rinsed with PBS. Finally, the sections were enclosed with 80% glycerol/20% tris buffered saline. Representative pictures were taken with a Leica DMi8 inverted fluorescent microscope equipped with a Leica DFC300 G camera.

Demographics of the typical and atypical AD groups were compared using Fisher’s exact test for categorical and Mann-Whitney U test for numerical and not normally distributed data. Outcome measures were compared between the 2 AD groups by using linear mixed model analysis. Linear mixed model analysis was used to adjust for the nested observations within cases. In the linear mixed model analyses, the group variable (typical versus atypical AD), the region (temporal versus parietal), and the interaction between group and region were added. Correcting for age and sex made the model less stable and was therefore not performed. An assumption to apply a linear mixed model is that residuals of outcome measurements are normally distributed. To meet this assumption, all outcome variables (pTau, amyloid-beta, GFAP, Iba1, HLA-DP/DQ/DR, CD68, C1q, C3d, C4b, number of amyloid-beta plaques, and number of C4b plaques) were transformed by taking the natural log of the (variable + 1). The covariance structure was set to unstructured. Using the linear mixed model, we answered if the difference in outcome measurement over the 2 regions was different between the 2 AD phenotypes (region × phenotype), also referred to as interaction effect. Both phenotypes showed a similar distribution over the 2 regions if no interaction effect was found. Statistical analysis was performed in IBM SPSS statistics version 22.0 (IBM SPSS Statistics, Armonk, NY, USA). Bonferroni correction was used to correct for multiple testing. Since we tested 11 outcome measurements (amyloid-beta, pTau, GFAP, Iba1, CD68, HLA-DP/DQ/DR, C1q, C3d, C4b, # amyloid-beta plaques, and # C4b plaques), statistical significance was set at p < .0045 (p < .05/11 outcome measurements) for each effect (region × phenotype and region) of the linear mixed model. Statistical significance was set at p < .05 for comparison of baseline characteristics.

The post-mortem cohort used for IHC consisted of 9 atypical AD cases and 10 typical AD cases. For a summary of initial clinical symptoms at presentation for each case, see Table 2. Most atypical AD cases presented with symptoms of aphasia, consisting of word-finding difficulties and spelling mistakes, combined with apraxia. None of our atypical cases retrospectively met the criteria for an isolated primary progressive aphasia [5]. One case presented with Parkinsonism and an alien hand syndrome, fitting a corticobasal syndrome during life. All 10 typical AD cases presented with memory complaints as most prominent initial symptom. Demographic characteristics of both AD phenotypes used for immunohistochemical analysis are shown in Table 4. Similar to the large cohort of 296 AD subjects (see Fig. 1), atypical AD patients selected for IHC analysis were younger at age of death and more often male. The disease duration, brain weight, post-mortem interval, disease severity, and ApoE genotype did not differ between groups.

Immunohistochemistry for pTau showed neuronal inclusions as well as neuritic threads (Fig. 2a). Typical AD cases showed more pTau immunoreactivity in the temporal compared to the parietal section (Fig. 2c). This was contrary to the pTau distribution in atypical AD cases, in which the parietal section showed more immunoreactivity compared to the temporal section. In addition, the distribution of pTau over the 2 regions differed significantly between the 2 phenotypes (Table 5).

Amyloid-beta immunoreactivity was present in the form of diffuse deposits, dense plaques, and classical cored plaques (Fig. 2b). Both the typical and atypical AD group showed more immunoreactivity for amyloid-beta in the temporal than parietal section (Fig. 2d), and no difference in distribution was observed (Table 5). While total immunoreactivity levels did not show differences between the 2 phenotypes, the number of dense plaques with an immunoreactive surface area of 100 μm² or more did show a significant difference in distribution over the 2 regions between the 2 phenotypes (Fig. 2e) (Table 5). In contrast to the typical AD group, more dense plaques were observed in the parietal section of the atypical AD group. Besides a difference in plaque number, we observed a contrast in plaque morphology between the two groups, which will be addressed below.

Staining for GFAP-positive astrocytes showed variably sized star-like GFAP-positive structures in all AD cases (Fig. 3a). Both phenotypes showed higher levels of immunoreactivity for GFAP in the temporal section compared to the parietal section. Atypical AD cases showed relatively more GFAP immunoreactivity in the parietal cortex compared to typical AD cases (Fig. 3c; Table 5).

Iba1 immunostaining showed positivity in both the cell soma as well as the processes of the microglia, mostly in the form of ramified microglia (Fig. 3b). Both AD groups showed a similar temporal dominancy for Iba1.

Activated microglia were stained using CD68 and HLA-DP/DQ/DR. Whereas CD68 positivity was mostly found in the soma of microglia, HLA-DP/DQ/DR showed a prominent staining in the processes of microglia (Fig. 4a, b).

Both markers showed a different distribution over the 2 regions between the 2 AD phenotypes (Fig. 4c, d; Table 5). Atypical AD cases showed more immunoreactivity for CD68 and HLA-DP/DQ/DR in the parietal compared to the temporal section, which was in contrast to the typical AD cases. In addition, the levels of CD68 and HLA-DP/DQ/DR immunoreactivity were relatively high in the parietal section of atypical AD compared to the temporal section of typical AD.

To visualize different parts of the complement cascade, we stained for C1q, C3d, and C4b, representing the start of the cascade, as well as for C5b-9, defining the end-stage of the cascade and forming the membrane attack complex.

C1q immunoreactivity was observed as a weak diffuse staining in the form of plaque-like structures, as punctuate staining of the neuropil, and sometimes in neurons (Fig. 5a). While the plaque-like structures were more often observed in the parietal section, the punctate staining of the neuropil was more prominent in the temporal section of both AD groups. No difference in C1q immunoreactivity was observed between regions or AD phenotypes (Fig. 5d; Table 5).

Compared to C1q, staining for C3d and C4b showed an intense plaque-like staining, which had a morphology resembling that of compact and classical cored plaques (Fig. 5b, c). Levels of immunoreactivity for C3d and C4b in typical AD were higher in the temporal compared to the parietal cortex (Fig. 5e, f; Table 5). In contrast, atypical AD showed higher immunoreactivity levels for both markers in the parietal section compared to the temporal section.

Analysis of C5b-9 exposed very low to no immunoreactivity in both regions of both phenotypes (data not shown). For this reason, no quantification or statistical analysis was performed for C5b-9. The few structures that were C5b-9 positive included parenchymal and meningeal vessels. The serum within these vessels also stained positive for C5b-9.

Our data show a different distribution in temporal and parietal regions between both AD phenotypes for complement factors C3d and C4b. Both complement factors were mostly abundant in the parietal section of atypical AD.

Looking at the morphology of plaques stained by amyloid-beta and C4b, we observed a difference in appearance between the 2 AD phenotypes. Amyloid-beta and C4b immunoreactive plaques in the parietal section of atypical AD cases showed a more granular composition compared to deposits in the temporal section of this phenotype or compared to deposits in both regions of the typical AD cases (Fig. 6). The surface area positive for C4b of these coarse-grained plaques was larger (> 100 μm²) than that of typical plaques. Atypical AD cases had more of these C4b plaques in the parietal compared to the temporal cortex, which was contrary to typical AD cases (Fig. 6i; Table 5). The coarse-grained plaques observed in the parietal cortex of atypical AD triple-stained for C4b, amyloid-beta, and thioflavine S (Fig. 7). This staining pattern was compared with that of classical-cored plaques observed in typical AD. In classical-cored plaques, C4b, amyloid-beta, and thioflavine S co-localized in the core of the plaque, while the corona only stained for amyloid-beta. Compared to cored plaques, coarse-grained plaques showed a fibrillar, less organized morphology with co-localization of C4b, amyloid-beta, and thioflavine S all over the plaque surface. Since an increased presence of activated microglia in atypical AD was observed, the localization of CD68 and HLA-DP/DQ/DR with cored and coarse-grained plaques was compared (Fig. 7). Like cored plaques, coarse-grained plaques were associated with clusters of CD68 and HLA-DP/DQ/DR-positive microglia. In cored plaques, activated microglia were located between the core and corona of the plaque. In coarse-grained plaques, the localization of CD68 and HLA-DP/DQ/DR-positive microglia was less structured and positive microglia appeared throughout the plaque. This data supports a morphological difference between cored and coarse-grained plaques, of which the latter occurs prominently in atypical AD (Fig. 6i).