Abstract
The APOE4 allele is the strongest genetic risk factor for sporadic Alzheimer's disease (AD). Although APOE4 is strongly associated with amyloid-β (Aβ), its relationship with tau accumulation is less understood. Studies evaluating the role of APOE4 on tau accumulation showed conflicting results, particularly regarding the independence of these associations from Aβ load. In this study, we examined three independent longitudinal cohorts (BioFINDER-1, BioFINDER-2 and WRAP), in which participants had cross-sectional and longitudinal measures of tau tangles [tau-PET; temporal meta-region of interest (meta-ROI) and entorhinal] or soluble p-tau (p-tau217), Aβ-PET and APOE genotype. The study included a total of 1370 cognitively unimpaired subjects and 449 with mild cognitive impairment, followed longitudinally with tau-PET and p-tau217. APOE4 carriers accounted for 40.2%-50% of the cohorts. Different linear regressions (cross-sectional) and linear mixed-effect models (longitudinal) with tau measures as outcomes were fitted to test the effect of APOE4 both as an independent predictor and in combination with baseline Aβ load (including interaction). All models included age, sex and cognitive status as covariates. We found no independent effects of APOE4 carriership on insoluble tau in either cohort (BioFINDER-2 or WRAP), on both cross-sectional and longitudinal tau-PET in the temporal meta-ROI, when Aβ was present in the model (P = 0.531-0.949). Aβ alone was the best predictor of insoluble tau accumulation, with no interaction between APOE4 and Aβ on tau-PET. In BioFINDER-2, there was a significant interaction between APOE4 and Aβ (b = 0.166, P < 0.001) in the entorhinal cortex at baseline. However, the interaction was not present in WRAP PET. No independent effects of the APOE4 carriership on baseline (P = 0.683-0.708) and longitudinal (P = 0.188-0.570) soluble p-tau217 were observed when Aβ was included in the model in BioFINDER-1 and WRAP. Likewise, no interaction between APOE4 and Aβ on soluble p-tau217 was observed. Mediation analysis revealed that Aβ load fully mediated most associations between APOE4 and tau (46%-112%, either cross-sectional or longitudinal tau-PET or soluble p-tau217). In the largest cohort (BioFINDER-2), looking at APOE4 groups by the number of ε4 alleles, we found an interaction between APOE4 homozygotes and Aβ on tau-PET levels at baseline and over time in the temporal meta-ROI, whereas in the entorhinal cortex this effect was observed only at baseline. In conclusion, although APOE4 is strongly associated with Aβ aggregation, it seems to be associated minimally with longitudinal changes in soluble or insoluble p-tau levels at a given level of Aβ pathology, confirming the primacy of Aβ in driving tau pathology.