Impact of APOE on amyloid and tau accumulation in argyrophilic grain disease and Alzheimer’s disease

Abstract

Alzheimer’s disease (AD), characterized by the deposition of amyloid-β (Aβ) in senile plaques and neurofibrillary tangles of phosphorylated tau (pTau), is increasingly recognized as a complex disease with multiple pathologies. AD sometimes pathologically overlaps with age-related tauopathies such as four repeat (4R)-tau predominant argyrophilic grain disease (AGD). While AGD is often detected with AD pathology, the contribution of APOE4 to AGD risk is not clear despite its robust effects on AD pathogenesis. Specifically, how APOE genotype influences Aβ and tau pathology in co-occurring AGD and AD has not been fully understood. Using postmortem brain samples (N = 353) from a neuropathologically defined cohort comprising of cases with AD and/or AGD pathology built to best represent different APOE genotypes, we measured the amounts of major AD-related molecules, including Aβ40, Aβ42, apolipoprotein E (apoE), total tau (tTau), and pTau181, in the temporal cortex. The presence of tau lesions characteristic of AD (AD-tau) was correlated with cognitive decline based on Mini-Mental State Examination (MMSE) scores, while the presence of AGD tau lesions (AGD-tau) was not. Interestingly, while APOE4 increased the risk of AD-tau pathology, it did not increase the risk of AGD-tau pathology. Although APOE4 was significantly associated with higher levels of insoluble Aβ40, Aβ42, apoE, and pTau181, the APOE4 effect was no longer detected in the presence of AGD-tau. We also found that co-occurrence of AGD with AD was associated with lower insoluble Aβ42 and pTau181 levels. Overall, our findings suggest that different patterns of Aβ, tau, and apoE accumulation mediate the development of AD-tau and AGD-tau pathology, which is affected by APOE genotype.

Introduction

Alzheimer’s disease (AD) is pathologically characterized by the extracellular deposition of amyloid-β (Aβ) in senile plaques and the intracellular accumulation of tau neurofibrillary tangles (NFT). However, proteinopathies caused by α-synuclein and by TDP-43 as well as vascular lesions are frequently observed in AD brains [6]. The presence of these additional neuropathological changes is predicted to impact AD phenotypes and progression [3]. Argyrophilic grain disease (AGD) is a common sporadic age-related primary tauopathy, which often coexists with AD (Fig. 1). AGD is defined by the presence of spindle- or comma-shaped argyrophilic grains in the neuropil of several brain regions, including the entorhinal cortex, hippocampus, and amygdala [6, 36]. Argyrophilic grains are neurofibrillary lesions enriched in 4-repeat (4R) tau, in contrast to AD neurofibrillary tangles composed of both 3R and 4R tau aggregates [37]. AGD is detected in approximately 5% of dementia cases [35, 36]. Intriguingly, a neuropathological study has reported that AD patients with AGD have lower scores of amyloid and tau pathologies than those without AGD [35]. While APOE gene coding apolipoprotein E (apoE) is the most significant genetic modifier for AD risk, APOE is also significantly tied with the occurrence of AGD. Among the three major APOE alleles, APOE2 has been demonstrated to increase the risk for AGD onset [8], which is in contrast to its protective effect in AD [17]. Although APOE4 is associated with a dose-dependent risk for AD with a 15-fold increased risk in APOE4 homozygotes [26], a lack of relationship between APOE4 and AGD onset has also been reported [36, 38].

Fig. 1

Representative images of AGD and of co-occurring AD with AGD. A Phosphorylated-tau immunohistochemistry (CP13) of the amygdala of a 94-year-old male patient with AGD. Arrows indicate balloon neurons, arrowheads indicate coiled bodies, and white triangle indicate grains. B, C Phosphorylated-tau immunohistochemistry (CP13) of the amygdala (B) and insula cortex (C) of a 91-year-old male patient with AGD and AD. Arrows indicate balloon neurons, arrowheads indicate coiled bodies, and white triangle indicate grains. Dashed circle shows neuritic plaque. Scale bar: 20 µm

In this study, using a large, neuropathologically defined cohort of postmortem brain samples with different APOE genotypes (N = 353), we biochemically investigated how APOE genotype is associated with the levels of major AD-related molecules, including Aβ40, Aβ42, total tau (tTau), phosphorylated tau 181 (pTau181), and apoE, in the presence of AGD and/or AD pathologies. Our findings revealed that the presence of neuropathologically defined lesions characteristic of AGD-tau pathology is associated with lower levels of Aβ40 and p-tau181 in mixed AD cases, with the association between APOE4 and the AD-related molecules levels being less pronounced in the presence of AGD-tau.

Materials and methods

Human neuropathological assessment

Postmortem brain tissue from non-Hispanic White donors was obtained from a group of 437 autopsied study participants identified from the Alzheimer’s Disease Research Center (P30 AG062677) and Mayo Clinic Study of Aging (U01 AG006786) with inclusion criteria requiring antemortem diagnosis within one year of death of clinical continuum of AD (i.e., normal, mild cognitive impairment, probable/possible AD), frozen tissue availability, tissue blocks availability, and lack of primary tauopathy (e.g. progressive supranuclear palsy, corticobasal degeneration, Pick’s disease, globular glial tauopathy). Of the 437 identified, a total of 353 autopsied non-Hispanic White donors were selected to best span the different APOE genotypes. Standard genotyping methods on blood samples was used to determine APOE allele status (ε2, ε3, ε4) [12]. The APOE2 group includes APOE ε2/ε2 (N = 1) and APOE ε2/ε3 (N = 45). The APOE3 group consists of APOE ε3/ε3 (N = 162) genotype. The APOE4 group includes APOE ε3/ε4 (N = 114) and APOE ε4/ε4 (N = 43). Cases with APOE ε2/ε4 (N = 22) were excluded because their limited representation in our dataset raises concerns about statistical robustness. Additionally, MMSE score was available in 133 subjects. Neuropathological examinations of brain tissue were performed in accordance with standardized protocols approved by Mayo Clinic Institutional Review board, as previously described [23]. These include neuropathologic evaluation using immunohistochemistry for antibodies against Aβ (Clone 6F/3D, DAKO), tau (AT8, ThermoFisher), TDP-43 (p409/410; Cosmo Bio), and α-synuclein (LB 509, Abcam). The diagnosis of AD neuropathologic change (ADNC) was conducted using the 2012 NIA-AA criteria [21], a well-established neuropathologic consensus criteria which include examination of AD-related pathologies such as Thal phase for Aβ plaques [34], and Braak NFT stage [1, 2]. Neuritic plaque semiquantitative scores were employed in our analyses: 0 = None; 1 = Sparse; 2 = Moderate; 3 = Frequent [21].

AGD was first screened using tau (AT8) immunohistochemistry in the amygdala, and later confirmed with 4R tau isoform (RD4, clone 1E1/A6, Millipore) immunohistochemistry and Bielschowsky silver stain, in conjunction with histomorphologic findings of ballooned neurons and other pertinent features on H&E-stained sections. Sections from the hippocampus, amygdala, and anterior cingulate are frequently screened and utilized for the diagnosis of AGD (Fig. 1). For analyses purposes, we have defined the following criteria: “AD-tau negative” (AD-tau = 0) corresponds to Braak stage < 4; “AD-tau positive” (AD-tau = 1) corresponds to Braak stage ≥ 4 [1]; “AGD-tau negative” (AGD-tau = 0) indicates the absence of AGD comorbidity with AD; “AGD-tau positive” (AGD-tau = 1) indicates the presence of AGD comorbidity with AD.

Sample preparation

Dissected tissues from the temporal cortex (100 mg) were pulverized and subjected to three-step extraction to isolate proteins according to their solubility in Tris-buffered saline (TBS), detergent-containing TBS, or formic acid (FA), as detailed previously [18, 33]. Briefly, samples were homogenized in 10 volumes (w/v) of ice-cold TBS supplemented with a protease inhibitor cocktail (Roche Diagnostics) and a phosphatase inhibitor (Roche Diagnostics) by Polytron homogenizer (KINEMATICA). Brain homogenates were centrifuged at 100,000 × g for 60 min at 4 ºC. The supernatant (soluble fraction) was collected, and the residual pellet was resuspended in 10 volumes of TBS containing 1% Triton-X (TBSX), supplemented with protease and phosphatase inhibitors. Following sonication, samples were incubated at 4 °C for 30 min with end-over-end agitation and centrifuged as described above. The resulting supernatant (detergent-soluble fraction) was retrieved, and the resulting pellet was re-solubilized in 70% FA. Samples were sonicated, incubated overnight at 4 °C with end-over-end agitation, and centrifuged as above. The final supernatant (insoluble fraction) was recovered and neutralized 20-fold with 1 M Tris-buffer (pH 11). All collected fractions were aliquoted and stored at − 80 °C until use.

Quantification of AD-related proteins

Amounts of Aβ40, Aβ42, apoE, tTau, and pTau181 in soluble, detergent-soluble, and insoluble fractions were determined by enzyme-linked immunosorbent assay (ELISA). Aβ40 and Aβ42 were measured using sandwich ELISA with antibodies produced in-house at Mayo Clinic, as previously described [5]. Briefly, end-specific monoclonal antibodies (13.1.1 for Aβ40 and 2.1.3 for Aβ42) were used as capture antibodies, and a horseradish peroxidase (HRP)-conjugated monoclonal antibody (Ab5-HRP) was used for detection. ApoE were quantified by sandwich ELISA with a polyclonal antibody directed against apoE (AB947, Millipore) used as capture antibody and a biotin-conjugated polyclonal anti-apoE antibody (K74180B, Meridian Life Sciences) used as detection antibody. An HRP-streptavidin conjugate was used to bind the biotinylated detection antibody [18]. For sandwich ELISA for tTau, monoclonal tau antibody (HT7; ThermoFisher Scientific) and a biotin-conjugated monoclonal anti-tau antibody (BT2; ThermoFisher Scientific) were utilized as capture and detection antibodies, respectively. An HRP-streptavidin conjugate was added to interact with the biotinylated detection antibody [18]. Color development for in-house sandwich ELISAs was initiated by addition of 3,3′,5,5′- tetramethylbenzidine (TMB) substrate, and the reaction was stopped with 1 M sulfuric acid. Absorbance was measured at 450 nm using a Synergy HT microplate reader (BioTek). Target protein levels were calculated using respective standard curves. For pTau181, a commercially available ELISA kit (ThermoFisher Scientific) was used according to the manufacturer’s instructions. All protein levels measured by ELISA were normalized against total protein concentration quantified using Pierce Detergent Compatible Bradford assay kit (ThermoFisher Scientific). Due to limits of detection in our ELISA assays, analytes could not be measured for a small amount (< 15%) of brain lysates samples.

Statistical analysis

Comparisons of characteristics according to APOE genotype group, and also according to combination of AD-tau pathology and AGD-tau pathology, were made using a Kruskal–Wallis rank sum test or a Wilcoxon rank sum test (continuous and ordinal variables) or Fisher’s exact test (categorical variables). Associations of presence of APOE2 or APOE4 with amyloid score, AD-tau, and AGD-tau were evaluated using unadjusted and age/sex-adjusted proportional odds logistic regression models (amyloid score) and binary logistic regression models (AD-tau and AGD-tau); p-values < 0.0167 were considered as statistically significant after applying a Bonferroni correction for multiple testing for the three outcome measures that were assessed.

Associations of demographic and neuropathological characteristics with MMSE score and AD-related molecules were evaluated using linear regression models. First, unadjusted models were assessed. Second, models were adjusted for age and sex only. Finally, a full multivariable analysis was performed adjusting for age, sex, and also any other variable with a P-value < 0.05 in unadjusted analysis for the given outcome (MMSE score or the AD-related molecule). P-values < 0.005 were considered as statistically significant after applying a Bonferroni correction for multiple testing for the 10 characteristics that were assessed for association with each outcome. AD-related molecules were examined on the square root, cube root, or natural logarithm scales in all regression analysis due to the presence of skewed distributions. Interactions with AD-tau and AGD-tau were also assessed in age/sex-adjusted linear regression models, where P-values < 0.0056 were considered as significant after Bonferroni correction.

Comparisons of AD-related molecules between APOE groups (APOE2 vs. APOE3 and APOE4 vs. APOE3) were made using unadjusted and age/sex-adjusted linear regression models. AD-related molecules were examined on the square root, cube root, or natural logarithm scales in all regression analysis due to the presence of skewed distributions; interactions with combination of AD-tau and AGD-tau were also assessed. Associations of MMSE score with AD-related molecules were also examined using unadjusted and age/sex-adjusted linear regression models. P-values < 0.01 were considered significant after applying a Bonferroni correction separately for each fraction. All statistical tests were two-sided. Statistical analysis was performed using R Statistical Software (version 4.1.2; R Foundation for Statistical Computing, Vienna, Austria).

Results

APOE genotype influences neuropathology in the elderly

We investigated postmortem brain samples from our study cohort consisting of 353 subjects (174 males and 179 females) chosen to best represent different APOE genotypes, with a mean age at death of 89 years in the APOE2 group (range: 69–101 years), 89 years in the APOE3 group (range: 59–100 years), and 84 years in the APOE4 group (range: 54–103 years). When comparing demographic and select neuropathological characteristics between the three APOE genotype groups (Table 1), we found that APOE genotype predominantly influenced both amyloid and tau pathology, with more severe scores of Braak stage and Thal phase detected in the APOE4 group. In more detailed analysis of neuritic plaque score, AD tau pathology, and AGD-tau pathology (Table 2), neuritic plaque score was significantly (P < 0.0167 considered significant after multiple testing correction) lower in the presence of APOE2 (OR = 0.49, p = 0.015), but higher in the presence of APOE4 (OR = 4.86, p < 0.001) when adjusting for age and sex. Additionally, the presence of APOE4 was associated with a higher likelihood of AD-tau pathology (OR = 6.34, p < 0.001), while although not quite significant, AGD-tau prevalence was lower in the presence of APOE4 (OR = 0.49, p = 0.041). APOE2 was associated with a significantly lower odds than APOE4 of AD-tau occurrence (OR = 0.40, p = 0.007); however, it was not associated with AGD-tau occurrence (OR = 1.89, p = 0.11).

Table 1 Subject characteristics according to APOE genotype

Table 2 Associations of APOE2 and APOE4 with neuritic plaque score, AD-tau, and AGD-tau

AD-tau pathology is associated with cognitive impairment

Among the cases available for MMSE score in the cohort, we investigated the association between demographic and neuropathological measures and MMSE scores (Additional file 1: Table S1). After correcting for multiple testing (P < 0.005 considered significant), significant negative associations with MMSE scores were observed for both older age (β = − 1.33, p < 0.001) and the presence of AD-tau lesions (β = − 2.11, p < 0.001) in analysis that was adjusted for age and sex. Furthermore, findings remained significant in full multivariable analysis when additionally adjusting for all variables with a p-value < 0.05 in unadjusted analysis (AD-tau) for both age (β = − 1.07, p = 0.004) and the presence of AD-tau (β = − 2.03, p < 0.001). When potential interactive effects of presence of AD-tau and AGD-tau with demographic/neuropathological characteristics were examined regarding associations with MMSE score, with adjustment for age and sex, no significant interactions were identified after correcting for multiple testing (Additional file 1: Table S2). Of note, the presence of AD-tau lesions was significantly associated with lower MMSE scores only in the absence of AGD-tau pathology (β = − 2.56, p < 0.001), with a weaker and non-significant association for subjects with AGD-tau pathology (β = − 0.24, p = 0.82); however, this interaction did not reach statistical significance (p = 0.042).

APOE genotype influences the levels of AD-related molecules in the temporal cortex

We compared the brain levels of AD-related molecules including Aβ40, Aβ42, apoE, tTau, and pTau181 in the soluble (TBS), detergent-soluble (TBSX), and insoluble (FA) fractions of brain lysate between APOE genotype groups (Table 3, Additional file 1: Table S3). Following adjustment for age and sex and after correcting for multiple testing (P < 0.01 considered as significant), we found significantly higher soluble apoE levels in the APOE2 group compared to the APOE3 group (β = 0.61 p < 0.001). We also found numerous differences between APOE4 and APOE3 groups. Aβ40 levels were higher in the APOE4 group than in the APOE3 group in the soluble, detergent-soluble, and insoluble fractions (TBS: β = 0.82, p = 0.005; TBSX: β = 0.99, p < 0.001; FA: β = 14.04, p < 0.001). Aβ42 levels were also higher in all three fractions in the APOE4 group compared to the APOE3 group (TBS: β = 2.29, p < 0.001; TBSX: β = 2.51, p < 0.001; FA: β = 24.96, p < 0.001). Compared to APOE3, APOE4 was associated with increased levels of insoluble apoE (FA: β = 1.24, p < 0.001) and insoluble pTau181 (FA: β = 0.72, p < 0.001), and decreased detergent-soluble tTau levels (TBSX: β = -0.18, p = 0.002). Additionally, among the measured analytes, there was only a positive association between MMSE score and insoluble tTau levels after adjusting for age and sex (FA: β = 0.274, p = 0.007) (Additional file 1: Table S4).

Table 3 Comparisons of AD-related molecules between APOE genotype groups

Neuropathological measures are associated with the levels of AD-related molecules in the temporal cortex

Regression analyses were conducted to examine the independent associations of neuropathological measures for VaD, CAA, amyloid score, TDP-43, synuclein, AD-tau, and AGD-tau with AD-related molecules (Table 4 [FA], and Additional file 1: Tables S5 [TBS] and S6 [TBSX]). In full multivariable analysis adjusting for age, sex, and any other measure that was associated with the given AD-related molecule with P < 0.05 in unadjusted analysis, neuritic plaque score was significantly associated (p < 0.005 considered as significant) with increased levels of soluble, detergent-soluble, and insoluble Aβ42 (TBS: β = 1.23, p < 0.001; TBSX: β = 1.33, p < 0.001; FA: β = 13.03, p < 0.001), insoluble apoE (FA: β = 0.65, p < 0.001) and insoluble pTau-181 (FA: β = 0.31, p < 0.001). The presence of CAA was also associated with increased levels of soluble, detergent-soluble, and insoluble Aβ40 (TBS: β = 2.30, p < 0.001; TBSX: β = 1.41, p = 0.001; FA: β = 20.94, p < 0.001) in full multivariable analysis. AD-tau pathology was positively associated with the levels of insoluble Aβ42 (FA: β = 12.27, p < 0.001) and pTau181 (FA: β = 0.41, p = 0.001), as well as negatively associated with levels of detergent-soluble tTau (TBSX: β = -0.23, p < 0.001), in full multivariable analysis. There were no associations that withstood correction for multiple testing between AD-related molecules and VaD, TDP-43, synuclein, or AGD-tau in full multivariable analysis.

Table 4 Associations of neuropathological measures with AD-related molecules (FA)

When segregating cases based on the presence or absence of AD-tau or AGD-tau pathology (Additional file 1: Table S7), cases with AD-tau pathology had higher levels of insoluble Aβ40, Aβ42, apoE, and pTau181 compared to those without AD-tau, particularly in the absence of AGD-tau (Fig. 2). Though not a significant interaction (P < 0.0056 considered as significant), it is worth noting that significant positive associations of AD-tau with insoluble Aβ40 (FA: β = 12.97, p < 0.001), insoluble apoE (FA: β = 1.05, p < 0.001), and insoluble pTau181 (FA: β = 1.06, p < 0.001) were observed in AGD-tau negative cases, but not in positive cases (Additional file 1: Table S8). AD-tau was also positively associated with insoluble Aβ42 levels, regardless of the AGD stratification (AGD-tau negative: β = 34.95, p < 0.001; AGD-tau positive: β = 28.34, p = 0.001). However, AD-tau association to insoluble Aβ42 levels is slightly weaker in the presence of AGD-tau pathology. Although not quite significant, the presence of AGD-tau was negatively associated with insoluble levels of Aβ40 (FA: β = − 9.96, p = 0.070), Aβ42 (FA: β = − 13.29, p = 0.046) and pTau181 (FA: β = − 0.68, p = 0.026) after adjusting for age and sex in AD-tau positive cases. There were no significant interactions between AD-tau and AGD-tau. Neuritic plaque score significantly interacted with AD-tau for insoluble pTau181 levels (Additional file 1: Table S9).

Fig. 2

Insoluble AD-related molecule levels according to AD-tau and AGD-tau pathology. Dot plots and the median for insoluble Aβ40 (A), Aβ42 (B), apoE (C), tTau (D), and pTau181 (E) levels in FA fraction are shown according to AD-tau and AGD-tau pathology. Measures of AD-related molecules were normalized by corresponding protein concentrations in each sample. P-values result from linear regression models that were adjusted for age and sex

When investigating multivariate correlations among the insoluble AD-related molecules, we found differences in the strength and direction of the associations depending on tau pathology status. Levels of insoluble pTau181 were positively correlated with levels of Aβ40, Aβ42, and apoE in the AD-tau only pathology group. The strength of these associations was weaker in the no tau pathology group and in the AD-tau/AGD-tau group. While insoluble pTau181 and insoluble Aβ40 remained positively corelated in the AGD-tau only group, insoluble pTau181 levels were inversely correlated with levels of insoluble Aβ42 and apoE. Although a positive association was detected between the levels of insoluble tTau and the levels of insoluble Aβ42 in the AD-tau only group, this association was weaker in the no tau pathology group and in the AGD-tau only group, and it was reversed in the AD-tau/AGD-tau group. Overall, the strength of the associations between the insoluble AD-related molecules are modest in the AD-tau/AGD-tau group compared to the AD-tau only group (Fig. 3).

Fig. 3

Correlation matrices of insoluble AD-related molecule levels according to AD-tau and AGD-tau pathology. Heatmap of Spearman correlation among insoluble Aβ40, Aβ42, apoE, tTau, and pTau181 levels in FA fraction are shown by stratifying to groups for (A) no tau pathology, (B) AD-tau pathology only, (C) AGD-tau pathology only, and (D) AD-tau and AGD-tau pathology. Correlation coefficients are visualized with blue-red gradients (− 1.0 to 1.0) and the numbers in the cells represent Spearman’s r

APOE4 is associated with AD-related molecules in the absence of AGD-tau

We then examined the effects of APOE genotype according to AD-tau/AGD-tau pathology on insoluble levels of AD-related molecules through linear regression analyses adjusted for age and sex (Table 5). In the group without AD-tau and AGD-tau pathology, APOE4 was significantly associated with higher levels of soluble, detergent-soluble, and insoluble Aβ42 compared to APOE3 (TBS: β = 2.68, p < 0.001; TBSX: β = 2.18, p = 0.002; FA: β = 21.79, p = 0.006). In the AD-tau positive group without AGD-tau pathology, higher levels of insoluble Aβ40 (FA: β = 14.15, p < 0.001), apoE (FA: β = 1.41, p < 0.001), and pTau181 (FA: β = 0.60, p = 0.001) as well as detergent-soluble Aβ40 (TBSX: β = 1.31, p = 0.002), Aβ42 (TBSX: β = 1.07, p = 0.004) and apoE (TBSX: β = 0.22, p = 0.002) were observed in the APOE4 group compared to the APOE3 group. Weaker associations between APOE4 and the levels of insoluble Aβ40 and of pTau 181 were in the AGD-tau pathology positive group.

Table 5 Interactions of APOE with tau pathology regarding associations with AD-related molecules

APOE4 remained significantly associated with detergent-soluble and insoluble Aβ42 levels compared to APOE3 in the mixed tau pathology group (TBSX: β = 2.98, p = 0.004; FA: β = 26.52, p = 0.008), which may be driven by AD-tau positivity as, out of 13 cases, 12 are AD-tau positive (Table 5).

It is however important to note that no significant interactive effects were reached between APOE4 and tau pathology, aside from a significant interaction between levels of detergent-soluble pTau-181 and APOE4. Further, there were no significant differences in levels of AD-related molecules between APOE2 and APOE3 groups irrespective of tau pathology stratification (Table 5).

Discussion

Carrying APOE4 significantly increases the risk of AD and age-related cognitive decline [30, 40]. While APOE genotype appears to influence AD pathogenesis through multiple pathways, the predominant effect in modulating amyloid pathology has been implicated as a major mechanism impacting AD risk [19]. A meta-analysis in non-dementia cohorts has shown that amyloid positivity, determined through amyloid PET imaging and CSF biomarkers, is exacerbated during aging in an APOE genotype-dependent manner (ε4/ε4 > ε3/ε4 = ε2/ε4 > ε3/ε3 > ε2/ε3 > ε2/ε2) [16]. In addition to Aβ [31], APOE4 has been implicated to influence proteinopathies involving tau, α-synuclein, and TDP-43 [7, 9]. Indeed, in this study we also found that APOE4 is associated with prevalence of AD-tau pathology as well as worsen amyloid score in our cohort composed of cognitively unimpaired individuals, individuals with mild cognitive impairment, and AD cases. Moreover, major AD-related molecules including insoluble Aβ40, Aβ42, apoE, and pTau181 were significantly increased in the presence of APOE4. However, associations of Aβ40 and pTau181 with APOE4 were no longer evident in the presence of AGD-tau. Consistent with previous studies [36, 38], our cohort had a lower percentage of APOE4 carriers in cases with AGD-tau pathology. Although AGD is a common tauopathy frequently detected in AD [37, 41], there is likely a distinct role of APOE4 in tau pathogenesis between AD-tau and AGD-tau. On the one hand, APOE4 may facilitate the shift from AGD-tau to AD-tau, while it is also possible that AGD-tau somehow mitigates the deleterious APOE4 effects exacerbating AD-related pathology. On the other hand, there was a trend increase of AGD-tau pathology in the presence of APOE2 which is consistent with a previous report [8]. Interestingly, polymorphisms in α2-macroglobulin (A2M) and low-density lipoprotein receptor-related protein 1 (LRP1) genes are also associated with AGD risk [10]. While LRP1 functions as a receptor for apoE and α2M, it has also been shown to mediate the cellular uptake and propagation of tau [25]. Thus, the apoE-LRP1 axis may be involved in the molecular mechanism mediating the development of AD-tau or AGD-tau pathology.

Synergistic effects between Aβ and tau in AD pathogenesis have been compellingly recognized [4]. We also found positive associations among insoluble Aβ40, Aβ42, apoE, and pTau181 levels in AD-tau positive cases without AGD-tau. However, these associations were either weakened in co-occurring AGD-tau and AD-tau cases, or even reversed to negative associations in the presence of only AGD-tau. Since the tauopathy negative group (without both AD-tau or AGD-tau) also showed associations among Aβ, apoE, and pTau181, their interactions are likely diminished through unknown mechanisms in AGD-tau positive cases. Weaker associations between Aβ40, Aβ42, and apoE are observed in the presence of AGD-tau, possibly indicating that AGD might cause tauopathy independently of Aβ. The balance of Aβ-apoE-tau interaction may be a key factor influencing the development of either AD-tau or AGD-tau pathology, or their co-occurrence. The conflicting APOE4 effects on AD-tau and AGD-tau pathologies may be due to its proneness facilitating the proteinopathy in the brain. The structure of 4R-tau filaments in AGD differs from those from AD [32]. The tau properties of AGD-tau may induce the suppressive effects on Aβ and apoE aggregation although further studies are needed. In addition, Aβ has been shown to accelerate tau propagation from the entorhinal cortex and medial temporal lobe into limbic system and neocortex through the hippocampal cingulum bundle [4, 14, 15]. In most AGD cases, tauopathy is detected in ambient gyrus, hippocampus, anterior entorhinal area and amygdala (Stage I), but spreads into medial temporal lobe and subiculum (Stage II), and to anterior temporal, cingulate gyrus, rectus gyrus, septum, accumbens nucleus, insular and orbitofrontal cortices, and hypothalamus (Stage III) [29]. Since the AGD stages are not associated with Braak stages and Thal phase [29], AGD tauopathy is predicted to propagate through an Aβ-independent manner. Co-occurrence of AGD and AD may affect the nature of tau properties and consequently its spread. Of note, a recent study has identified APOE as one of the top-ranked genes whose expression is associated with the spatial spreading of tau [20]. Thus, apoE amounts as well as APOE genotype may also differently influence the development of AD-tau and AGD-tau pathologies. In addition, co-occurring limbic predominant age-related TDP-43 encephalopathy neuropathological changes (LATE-NC) in AD has been suggested to associate with elevated tau levels [39]. However, our biochemical analyses in the medial temporal cortex did not reveal significant correlations between tau levels (tTau or pTau181) and neuropathologically defined TDP-43 pathology. This discrepancy may be due to our measurements differing both in brain region (medial temporal cortex as opposed to entorhinal and frontal cortex) and phosphorylated tau isoform (p-Tau 181 as opposed to p-Tau 199). Moreover, the lack of TDP-43 biochemical measures in our study emphasizes the need for even more comprehensive investigations across varied brain regions and tau isoforms to further explore the relationship between LATE-NC and AD.

Our study showed that the presence of AD-tau or AGD-tau pathologies differentially influences the cognitive functions assessed by MMSE. The occurrence of AD-tau pathology was negatively associated with MMSE scores. However, the significant association between AD-tau and MMSE score was weakened in the presence of AGD-tau. This result is in line with another study reporting that cognitive status is not affected by the presence of AGD [13]. Since Aβ and tau synergically cause synaptic damage and neurodegeneration [4], lower Aβ accumulation and lack of Aβ-tau interaction in AGD-tauopathy may be involved in the benign effects on cognitive function even in the presence of AD-tau. In addition, tau acetylation at K274 residue was not detected in AGD-tau, while this specific posttranslational modification was generally identified in other tauopathies [11]. Since tau K274 acetylation exacerbates tau aggregation and cytotoxicity [24], the unique nature of AGD-tau may mitigate AD-tau toxicity. However, cognitive function is likely impaired in severe AGD cases. At AGD stage III, 71.2% of cases have been reported to have dementia with the Clinical Dementia Rating (CDR) ≥ 1 [29]. Since Braak stages and Thal phase are milder in dementia cases with AGD compared to AD [35], the mechanisms of neuronal damages caused by AGD-tau should differ from those of AD-tau. While AGD-tau may be preventive against AD-related phenotypes by lowering tau aggregation and propagation at Stage I, the wide-spread AGD-tau at Stage III may cause cognitive decline independently of the more common amyloid and tau pathology detected in AD. It is worth noting that although the lack of a relationship found between the presence of AGD and cognitive impairment agrees with past literature, we did not specifically apply Saito staging to evaluate regional involvement of argyrophilic grains [13, 22, 27, 28].

In summary, we demonstrated that APOE4 increases the risk of AD-tau pathology, but not AGD-tau pathology, accompanied with exacerbated accumulation of insoluble Aβ40, Aβ42, apoE and pTau181. In the presence of AGD-tau, the effect of AD-tau on cognitive impairment became modest with lower insoluble AD-related molecule levels and a lack of association amongst those molecules. Our study provides a comprehensive analysis into how APOE genotype influences the trajectory of AD-tau and AGD-tau pathologies by incorporating biochemical measures, thus supplementing, and enriching our understanding of the neuropathological studies previously published. However, with our study predominantly presenting association data, experimental validation in future work will strengthen the robustness of our findings. One limitation of our study is that we subjectively built the cohort based on APOE genotype. Since APOE2 and APOE4 carriers in our cohort are more prominent than in the general population, their effects may be over- or under-estimated in our study. There is also a possibility of a false-negative error due to the relatively small sample numbers. Further studies should define interactions among APOE, AGD-tau and AD-tau by including various brain regions and assessing other phosphorylated tau species, ideally in a larger cohort with different stages of AGD, spanning different ages and APOE genotypes.