Subjects

Autopsies were carried out after informed consent was obtained from family members, and all experiments in this study were approved by the ethical committees of the Okayama University Graduate School of Medicine, Dentistry and Pharmaceutical Sciences, the National Hospital Organization Minami-Okayama Medical Center, Zikei Institute of Psychiatry, the Tokyo Metropolitan Institute of Medical Science, and Niigata University. From 1,125 autopsy cases who had died in psychiatric hospitals or neurological departments of general hospitals and were registered in the database at the Department of Neuropsychiatry, Okayama University Graduate School of Medicine, Dentistry and Pharmaceutical Sciences as of the end of December 2022, we first selected 475 cases for which the data on Saito AG stage were available. In our laboratory, AGs in the central nervous system were routinely evaluated on sections stained with Gallyas silver stain (Gallyas method), and the Saito AG stage was determined in all cases [6, 53]. Further, AGs were confirmed by being labeled with AT8 and an anti-4R tau antibody but not with an anti-3R tau antibody. From the database, other fundamental pathological data in all of these cases, i.e., Braak NFT stage [9], Thal Aβ phase [61], CERAD neuritic plaque score [38], pathological subtypes of Lewy body disease (LBD) [36, 63], Braak Parkinson’s disease (PD) stage [8], the semiquantitative scores of tufted astrocytes, astrocytic plaques, and GFAs [37, 67], the stage of limbic-predominant age-related TDP-43 encephalopathy pathologic change (LATE-NC) [43, 44], histological subtypes of frontotemporal lobar degeneration with TDP-43-positive inclusions [11, 27, 30], the presence or absence of fused in sarcoma (FUS) pathology [29], the presence or absence of C9ORF72 mutation-related p62-positive inclusions in the cerebellar granular cell layer [31, 47], the semiquantitative data regarding the severity of neuronal loss with gliosis in representative anatomical regions in the neocortex, basal ganglia, brain stem nuclei, cerebellum, and spinal cord (defined later), and the data of vascular lesions were extracted. The pathological assessments of all cases were routinely carried out by a senior neuropathological researcher (OY) with or without another neuropathological researcher (TM, HI, or ST).

Among 475 cases, 77 cases had AGs. To minimize the influence of various pathological conditions except for AGs on the analysis of neuronal loss, we excluded 47 AGD cases that had at least one of the following pathologies: NFTs with Braak stages V–VI, Lewy body disease (LBD), tufted astrocytes, astrocytic plaques, frontotemporal lobar degeneration with TDP-43-positive inclusions (FTLD-TDP) [11, 27, 30], amyotrophic lateral sclerosis with TDP-43-positive inclusions (ALS-TDP) [11], FTLD or ALS with FUS-positive pathologies [11], globular glial tauopathy [2], and other various established neurodegenerative diseases (e.g., multiple system atrophy, spinocerebellar degeneration, myotonic dystrophy, Huntington’s disease, dentatorubral–pallidoluysian atrophy, post-encephalitic parkinsonism, Bechet’s disease, leukodystrophies, Alexander disease, neuronal ceroid lipofuscinoses, and Marchiafava–Bignami disease). Cases having findings of global ischemia were also excluded, but cases bearing LATE-NC were not. Finally, 30 cases having AGs were extracted, and we regarded these cases as pAGD (Table 1).

Table 1 Demographic data in pAGD and control cases

Then, we also extracted 34 control cases that had NFTs in Braak stages I to IV with no or mild Aβ deposits with Thal phases 1 or 2 from our autopsy series, corresponding to the definition of primary age-related tauopathy [12]. Cases bearing AGs or other neurological diseases noted above were not included in this group. In these 34 control cases, no case had LATE-NC (Table 1).

The data regarding dementia in all pAGD and control cases were extracted from our data base. The presence of dementia was determined based on clinical diagnosis, cognitive impairment with the necessity of support in instrumental activity of daily living that was noted in the clinical summary, and/or the retrospective interview of physicians who had provided long-term treatment at least in the late stage of the clinical course [3]. These clinical data were known before the pathological examinations of each case. The data regarding dementia were available in all 30 pAGD cases and 33 of 34 control cases.

Neuropathological examination

Brain tissue samples were fixed post mortem with 10% formaldehyde and embedded in paraffin. Ten-μm-thick sections from the frontal, temporal, parietal, occipital, insular, and cingulate cortices, hippocampus, amygdala, basal ganglia, midbrain, pons, medulla oblongata, and cerebellum were prepared. These sections were stained with hematoxylin–eosin (H&E), Klüver–Barrera (KB), the Gallyas method, and modified Bielschowsky silver methods.

Paraffin sections were immunostained by the immunoperoxidase method using 3, 3′-diaminobenzidine tetrahydrochloride. Six-μm-thick paraffin sections were immunostained by the immunoperoxidase method using 3,3′-diaminobenzidine tetrahydrochloride. Deparaffinized sections were incubated with 1% H2O2 in methanol for 20 min to eliminate endogenous peroxidase activity in the tissue. Sections were treated with 0.2% TritonX-100 for 5 min and washed in phosphate-buffered saline (PBS, pH 7.4). After blocking with 10% normal serum, sections were incubated overnight at 4 °C with one of the primary antibodies (Supplementary Table 1) in 0.05 M Tris–HCl buffer, pH 7.2, containing 0.1% Tween and 15 mM NaN3. After three 10-min washes in PBS, sections were incubated in biotinylated anti-rabbit, -mouse, or -pig secondary antibody for 1 h, and then in avidin-biotinylated horseradish peroxidase complex (ABC Elite kit, Vector, Newark, CA, USA) for 1 h. The peroxidase labeling was visualized with diaminobenzidine as the chromogen.

Semiquantitative assessment of the quantity of AGs

AGs were semi-quantitatively assessed in the representative anatomical regions in all pAGD and control cases. Regions examined were the ambient gyrus, amygdala, hippocampal CA1, transentorhinal cortex, subiculum, anterior portion of the superior temporal gyrus, lateral occipitotemporal gyrus, insular cortex, inferior temporal gyrus, middle frontal gyrus, primary motor cortex, inferior parietal lobule, peristriate region, striate region, caudate nucleus, putamen, globus pallidus, periaqueductal gray, substantia nigra, pontine tegmentum, medullary tegmentum, and spinal anterior horns (Table 2). Sections stained with the Gallyas method were employed. The grading system of the quantity of AGs used in this study was a modified version of that used in an original report of Saito AG stage [53] (i.e., one to 19 AGs per × 400 visual field was defined as ± .): −, no grain; ± , more than one to 19 AGs per × 400 visual field; + , 20 to 50 AGs per × 400 visual field; +  + , 51 to 100 AGs per × 400 visual field; +  +  + , more than 100 AGs per × 400 visual field. Control cases had no AG in any region.

Table 2 Distribution of AGs in pAGD casesStaging of the distribution of AGs (modified Saito staging system)

The distribution of AGs was assessed on sections stained with the Gallyas method using the Saito AG stage in all pAGD cases [53]. In the original description of the Saito AG stage, the handling of anatomical regions having a small number of AGs (i.e., one to 19 AGs per × 400 visual field) in the determination of the Saito stage was not described [53]. Therefore, in this study, we operationally classified pAGD cases into each Saito AG stage:

Saito AG stage 0: there was no argyrophilic grain in any region in the cerebrum and brain stem.

Saito AG stage I: one to 50 AGs per × 400 visual field were present in the limbic region (i.e., the ambient gyrus, amygdala, and/or the anterior portion of hippocampal CA1). There were fewer than 20 AGs per × 400 visual field in the superior temporal gyrus at the level of the temporal tip, lateral occipitotemporal gyrus, transentorhinal cortex, and subiculum.

Saito AG stage II: there were AGs that fit stage I, and 20 or more AGs per × 400 visual field are additionally present in at least one of the following four regions: (i) the superior temporal gyrus at the level of the temporal tip, (ii) the lateral occipitotemporal gyrus at the level of the amygdala or hippocampus, (iii) the transentorhinal cortex, and (iv) the subiculum. There were no or fewer than 20 AGs in the insular cortex.

Saito AG stage III: there were AGs that fit stage II, and 20 or more AGs per × 400 visual field were additionally present in the insular cortex.

The diffuse form of pAGD was defined as one or more AGs in the limbic system, temporal cortex, and all of the following regions: (i) all of the middle frontal gyrus, motor cortex, inferior parietal lobule, and occipital cortex, (ii) the striatum (the caudate nucleus and/or putamen), (iii) the midbrain (the periaqueductal gray and/or substantia nigra), (iv) the pontine tegmentum, and (v) the tegmentum in the medulla oblongata. As a result, all of our diffuse form pAGD cases also fit the definition of Saito AG stage III.

Semiquantitative assessment of neuronal loss with glial proliferation

In all pAGD and control cases, the severity of neuronal loss with gliosis in the cerebral cortex was assessed on H&E- and KB-stained sections according to the four-point staging system employed in our previous studies (Supplementary Fig. 1) [64, 65]: stage 0, neither neuronal loss not gliosis was observed; stage 1, slight neuronal loss and gliosis were observed only in the superficial layers; stage 2, obvious neuronal loss and gliosis were found in cortical layers II and III, and status spongiosis and relative preservation of neurons in layers V and VI were often present; and stage 3, pronounced neuronal loss with gliosis was found in all cortical layers, and adjacent subcortical white matter exhibits prominent fibrous gliosis.

In the basal ganglia and brainstem nuclei, the degree of neuronal loss with gliosis was assessed on H&E- and KB-stained sections according to the four-point staging system employed in our previous studies [64, 65] (Supplementary Fig. 2): stage 0, neither neuronal loss nor gliosis was observed; stage 1, mild gliosis and mild neuronal loss were present; stage 2, neuronal loss and gliosis were moderate, but tissue rarefaction was absent; and stage 3, severe neuronal loss, severe fibrous gliosis, and tissue rarefaction were observed.

The severities of the degeneration of the corticospinal tract at the level of the cerebral peduncle and the medulla oblongata and that of the frontopontine tract at the level of the cerebral peduncle were assessed as follows: stage 0, neither loss of myelin nor glial proliferation; stage 1, slight myelin loss and gliosis without atrophy of the tract; stage 2, evident myelin loss and gliosis with slight atrophy of the tract; stage 3, evident myelin loss and gliosis with severe atrophy of the tract.

Semiquantitative assessment of NFTs and GFAs

NFTs in the representative anatomical regions were assessed according to a four-point staging system in cases of diffuse form pAGD: stage 0, no lesion in the anatomical region; stage 1, more than one lesion in the anatomical region but less than one lesion per × 200 visual field; stage 2, one to ten lesions per × 200 visual field; stage 3, 11 to 20 per × 200 visual field; stage 4, 21 to 30 per × 200 visual field; and stage 5, 31 or more per × 200 visual field.

GFAs in the middle frontal cortex, caudate nucleus, putamen, and amygdala were semiquantitatively assessed according to a three-point staging system (GFA stage) in cases of diffuse from of pAGD: stage 0, no lesion in the anatomical region; stage 1, one or more lesions in regions examined but less than one lesion per × 200 visual field; and stage 2, one or more lesions per × 200 visual field.

Tau immunoblotting

Frozen brain tissue from the hippocampus, amygdala, inferior temporal gyrus, middle frontal gyrus, and caudate nucleus of the right hemisphere in two diffuse-form pAGD cases (Cases 1 and 2 in Supplementary Table 2 and Supplementary file 1) was available. These samples were used for Western blotting according to methods described previously [59]. Brain samples (0.5 g) from patients were individually homogenized in 20 ml of homogenization buffer (HB: 10 mM Tris–HCl, pH 7.5, containing 0.8 M NaCl, 1 mM EGTA, and 10% sucrose). Sarkosyl was added to the lysates (final concentration: 2%), which were then incubated for 30 min at 37 °C and centrifuged at 27,000 g for 10 min at 25 °C. The supernatant was divided into tubes (each 1.3 ml) and centrifuged at 166,000 g for 20 min at 25 °C. The pellets were further washed with 0.1% sarkosyl in a homogenization buffer (0.5 ml/tube) and centrifuged at 166,000 g for 20 min. The resulting pellets were used as the sarkosyl-insoluble fraction (ppt). The sarkosyl-ppt was sonicated in 50 μl (/tube) of 30 mM Tris–HCl (pH 7.5) and solubilized in 2 × sample buffer. Samples were run on gradient 4–20% polyacrylamide gels and electrophoretically transferred to PVDF membranes. Residual protein-binding sites were blocked by incubation with 3% gelatin (Wako) for 10 min at 37 °C, followed by overnight incubation at room temperature with primary anti-tau antibodies (AT8, mouse, monoclonal, 1:500; T46, mouse, monoclonal, 1:1,000). The membrane was incubated for 1 h at room temperature with anti-mouse IgG (BA-2000, Vector Lab) or anti-rabbit IgG (BA-2000, Vector Lab), then incubated for 30 min with avidin-horseradish peroxidase (Vector Lab), and the reaction product was visualized by using 0.1% DAB and 0.2 mg/ml NiCl2 as the chromogen.

Genetic analysis of MAPT mutation and ApoE genotype

Genomic DNA extracted from autopsy brains (middle frontal gyrus) was used to determine pathological variants of MAPT and APOE genotypes in two diffuse form pAGD cases (cases 1 and 2 in Supplementary Table 2 and Supplementary file 1). Primer sequences and PCR conditions are available upon request. The concentration of extracted genomic DNA was measured using NanoDrop OneC (Thermo Fisher Scientific, Waltham, MA, USA), and its quality control was also performed by an Agilent 4200 TapeStation (Agilent Technologies, Santa Clara, CA, USA). For Sanger sequencing, the BigDye Terminator v3.1 Cycle Sequencing Kit (Thermo Fisher Scientific) was used, and the series of reactions was conducted according to the instruction manual.

Immunoelectron microscopy

Sarkosyl-insoluble fractions extracted from two brains with diffuse form pAGD (cases 1 and 2 in Supplementary Table 2 and Supplementary file 1) were dropped onto carbon-coated nickel grids (Nissin EM, Tokyo, Japan). The grids were immunostained with an anti-phosphorylated tau monoclonal antibody (AT8, mouse, 1:200) and a secondary antibody conjugated to 5 nm gold particles (BBI Solutions, 1:50) as described [14]. Electron micrograph images were recorded with a JEOL JEM-1400 electron microscope (JEOL).

Statistical analysis

Fisher’s exact test was used to compare the variables between two groups. Correlations between variables in pAGD cases were assessed by Spearman’s rank-order correlation test.

To assess the effects of predictor variables on the severity of neuronal loss in representative regions (amygdala, entorhinal cortex, hippocampal CA1, lateral occipitotemporal gyrus, inferior, middle, and superior temporal gyri, insular cortex, cingulate gyrus, middle frontal gyrus, orbital gyrus, and substantia nigra, respectively) in a combined group of 30 pAGD and 34 control cases, we performed multivariate ordered logistic regression analyses with neuronal loss stage (stages 0–3) as the dependent variable and the age at death, Braak NFT stage, Saito AG stage (stage 0, stage I, stage II, and stage III including diffuse form), and LATE-NC stage as independent variables. In the substantia nigra, the data of neuronal loss stages 1 and 2 were combined (stage 0, stages 1–2, and stage 3). We additionally performed multivariate ordered logistic regression analyses with neuronal loss stages in the amygdala and middle frontal gyrus as dependent variables and the age at death, Braak NFT stage, Saito AG stage (stage 0, stage I, stage II, and stage III including the diffuse form), LATE-NC stage, and GFA stage as independent variables. A Brant test was performed to check the proportional odds assumption, which was satisfied (p value ≥ 0.05).

Then, to assess the effects of predictor variables on the occurrence of neuronal loss in the caudate nucleus, putamen, and globus pallidus in the combined group of 30 pAGD and 34 control cases, we performed binomial logistic regression analyses with the presence of neuronal loss (neuronal loss stage 0/stages 1–3) as the dependent variable and the age at death, moderate Braak NFT stage (stages 0–II/stages III–IV), severe AGD (Saito AG stages 0–II/Saito AG stage III including the diffuse form), and moderate LATE-NC (stages 0–1/stage 2. No pAGD or control case had LATE-NC in stage 3) as independent variables. We additionally performed binomial logistic regression analyses with neuronal loss in the caudate nucleus and putamen as dependent variables and the age at death, moderate Braak NFT stage (stages 0–II/stages III–IV), severe AGD (Saito AG stages 0–II/Saito AG stage III including the diffuse form), moderate LATE-NC (stages 0–1/stages 2. No pAGD or control case had LATE-NC in stage 3), and the presence of GFA (GFA stage 0/GFA stages 1–2) as independent variables.

We examined the impacts of various pathological factors on the development of dementia in a combined group of pAGD and control cases using univariate binomial logistic regression analysis. To minimize the effect of vascular lesions on the development of dementia, cases that had one or more large infarctions and/or two or more lacunae in the neocortex or subcortical nuclei were excluded. As a result, 23 pAGD and 28 control cases were included in this analysis. Independent variables were the age at death, moderate Braak NFT stages (stages 0–II/stages III–IV), Aβ deposits (Thal phase 0/Thal phases 1–5), Saito AG stage (stages I, stage II, and stages II and III, respectively), three density ranges of AGs in the amygdala (one to 49, 50 to 99, and 100 or more per × 400 visual field, respectively), AGs in the amygdala (presence or absence), three density ranges of AGs in the hippocampal CA1 (one to 49, 50 to 99, or 100 or more per × 400 visual field, respectively), AGs in the hippocampal CA1 (presence or absence), AGs in the lateral occipitotemporal gyrus (presence or absence), AGs in the inferior temporal gyrus (presence or absence), AGs in the insular cortex (presence or absence), LATE-NC (presence or absence), moderate LATE-NC stage (stage 2, no pAGD or control case had LATE-NC in stage 3), and moderate to severe neuronal loss stage in the amygdala (stages 0–1/stages 2–3).

Then, to examine the independent effect of AGs on the development of dementia in a combined group of 23 pAGD and 28 control cases, we performed multivariate binomial logistic regression analyses with the age at death, Braak NFT stages III–IV (presence or absence), one of the pathological variables regarding AGs, and LATE-NC (presence or absence) as independent variables. A pathological variable regarding AGs submitted into each model as the independent variable was three density ranges of AGs in the amygdala (one to 49 AGs, 50 to 99 AGs, or 100 or more AGs per × 400 visual field, respectively), three density ranges of AGs in the hippocampal CA1 (one to 49 AGs, 20 to 99 AGs, or 100 or more AGs per × 400 visual field, respectively), AGs in the lateral occipitotemporal gyrus (presence or absence), or AGs in the inferior temporal gyrus (presence or absence), respectively.

Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated. A p value < 0.05 was accepted as significant. Statistical analysis was performed using BellCurve for Excel 2.15 (Social Survey Research Information Co., Ltd., Tokyo, Japan) and EZR (Saitama Medical Center, Jichi Medical University, Saitama, Japan), which is a graphical user interface for R (The R Foundation for Statistical Computing, Vienna, Austria).