Alzheimer’s disease (AD) is a slowly progressive disease that leads to the degeneration of brain cells. It is the major type of dementia, characterized by the decline of thinking ability and independence of daily activities [1]. On the other hand, mild cognitive impairment (MCI) is a disorder in which subjects exhibit objectively cognitive dysfunction and their ability to engage in activities of daily living is minimally affected [2, 3]. The apolipoprotein E (APOE) is a central regulator of cholesterol and is closely related to AD pathology due to the homeostasis of lipid and protein [4, 5]. The APOE gene has three alleles (ε4, ε3, and ε2) responsible for three major APOE subtypes (APOE4, APOE3, and APOE2) [6]. The APOE ε4 allele is the most common genetic risk factor for AD [7], and it is related to increased production of an amyloid-$\beta$ (A$\beta$) [8] other than reduced clearance of cerebral A$\beta$ compared to ε2 and ε3 alleles [9, 10]. Consequently, subjects with APOE ε4 demonstrate increased cerebral A$\beta$ deposition [11], and APOE ε4 carriers have amyloid positive onset earlier than non-carriers [12]. In contrast, other subtypes of APOE are supposed to be protective (APOE2) or neutral (APOE3) for AD risk [13, 14, 15].

Tau pathology is a crucial aspect of AD, and the tau burden can predict cognitive decline in AD [16]. MCI individuals with high tau levels show an increased risk of cognitive decline [17]. However, the relationship between APOE and tau pathology is less clear and controversial. [18] has reported a significant physiological link between cerebrospinal fluid (CSF) levels of APOE and CSF tau in neurologically healthy, cognitively intact individuals. In contrast [19], other studies have reported no effect of APOE ε2 or ε4 on CSF tau in cognitively normal aging. Post-mortem evaluations suggested that APOE ε2 and ε4 alleles were not related to paired helical filament (PHF) tau tangles in the absence of A$\beta$ [20]. However, there was evidence that APOE ε4 significantly influenced tau-mediated neurodegeneration independently of A$\beta$ in a mouse model of tauopathy [21]. Recent studies have shown that the ε4+ group has a higher rate of tau accumulation, and the enhanced effect of APOE ε4 on tau accumulation still exists after adjusting the A$\beta$ load in the cortex [22]. So far, there is no study on the relationship between APOE ε3 and tau pathology. In addition, there were no studies that explored the effect of APOE alleles on tau as measured by CSF dependently or independently of A$\beta$ in a group of individuals that spans the spectrum of cognition.

Similarly, the relationship between cognition and APOE allele status is also controversial. Previous researches reported a positive association [23, 24, 25, 26, 27, 28, 29]. These findings were generally interpreted to suggest that the influences of APOE ε4 on late-life cognitive impairment were mediated by the cascade of APOE that was APOE ε4 led A$\beta$ deposition, then tau tangles, finally cognitive dysfunction [30]. However, other studies showed no relationship between cognition and APOE ε4 [31, 32, 33, 34, 35]. There were few studies on the relationship between cognition and APOE ε2 and ε3 in MCI and AD.

Is there a new pathological cascade that explains the cognitive impairment in the AD continuum? Therefore, the associations of APOE alleles with tau and cognition and whether A$\beta$ mediates these associations need to be further elucidated. In this article, we test hypothesis that the associations of APOE alleles status with CSF tau and cognitive function differ according to the presence and absence of A$\beta$ deposition.

This work evaluated the effects of different APOE allele statuses on T-tau, P-tau, and cognition in relation to A$\beta$ deposition in a large cohort of subjects. We have the following main findings: Firstly, there were significant differences between APOE allele carriers and noncarriers in the measures of T-tau, P-tau, and ADAS-cog scores in MCI, but not in CN and AD. Secondly, there was an interaction between APOE ε4 and ε3 and the presence of A$\beta$. Finally, APOE ε4 and APOE ε3 were associated with CSF tau and cognition in MCI participants with A$\beta$ deposition, but not in AD participants with A$\beta$ deposition.

Compared with noncarriers, previous studies reported APOE ε4 carriers had higher deposition of A$\beta$ in the cerebral cortex in late-onset AD [42, 43]. A low CSF A$\beta$ level is considered a marker of A$\beta$ deposition in AD patients’s brains [44]. Consistent with the report by Vemuri et al. [45], within CN, MCI, and AD group, APOE ε4 carriers had lower CSF A$\beta$42 than noncarriers. In addition, in CN and MCI groups, results demonstrated that APOE ε4 was more common in individuals with significant A$\beta$ deposition than in subjects without significant A$\beta$ deposition. There were no individuals without significant A$\beta$ deposition in the AD group, suggesting that APOE ε4 may relate strongly to CSF A$\beta$ in the different phases of cognitive damage. On the contrary, APOE ε3 carriers had higher CSF A$\beta$42 than noncarriers in any group. However, APOE ε2 carriers had higher CSF A$\beta$42 than noncarriers only in the CN group. APOE ε3 and ε2 were widespread in individuals with significant A$\beta$ deposition in the AD group, and they were prevalent in participants without significant A$\beta$ deposition in the CN group. This phenomenon of APOE ε3 and ε2 in the AD group may be related to A$\beta$ deposition in all AD patients. Relative to APOE ε4, we speculate that APOE ε3 and ε2 may have opposite effects in CN subjects.

There was no significant difference in T-tau and ADAS-cog scores between APOE allele carriers and noncarriers among CN. Among MCI, T-tau, P-tau, and ADAS-cog scores were significantly different between APOE allele carriers and noncarriers. Interestingly, there was not a single difference between APOE allele carriers and noncarriers in the measures of T-tau, P-tau, and ADAS-cog scores in AD subjects. Our data show significant differences in CSF A$\beta$42 levels between APOE allele carriers and noncarriers in all clinical groups. Still, there are no significant differences in T-tau values between APOE allele carriers and noncarriers in CN and AD individuals. In patients with clinically diagnosed cognitive impairment, the effect of APOE genotype on cognitive decline is the most consistent in MCI patients but not in AD patients. This is not to say that APOE genotypes are not associated with neuropathological parameters. When all individuals are combined, APOE ε4 significantly increases the risk of more severe clinical damage and has higher levels of P-tau and T-tau. However, APOE ε3 and ε2 have opposite effects. APOE genotype is not deterministic because of many ε4 carriers without dementia and many ε4 noncarriers with dementia [45]. In contrast, there are many ε3 and ε2 carriers with dementia and many ε3 and ε2 noncarriers without dementia.

In 2012, there was a change in the diagnostic criteria for AD neuropathology [46], requiring the presence of A$\beta$ deposition for the neuropathological diagnosis of AD. However, the previous view shows that even in the absence of A$\beta$, the appearance of neurofibrillary tangles (NFT) is the earliest neuropathological manifestation of AD [47].Therefore, it has been argued that tau tangles are a pathophysiological process different from AD in the absence of A$\beta$ [20, 48]. Several studies revealed a relationship between APOE and A$\beta$ pathology and tau pathology, indicating that the association between APOE and tau pathology may be mediated by A$\beta$ [20, 49]. We found an interaction between APOE ε4 and the presence of A$\beta$ such that the associations of APOE ε4 with T-tau and P-tau were much more robust in persons with A$\beta$. When we considered all subjects as a whole, there was a significant association between APOE ε4 and increased CSF T-tau and P-tau concentrations in individuals with significant A$\beta$ deposition. There is no similar phenomenon in individuals without significant A$\beta$ deposition. In the stratified analyses regressing within CN, MCI, and AD groups, we found that APOE ε4 was significantly related to increased CSF T-tau and P-tau concentrations in MCI but not in AD to A$\beta$ status. Few studies have tested the relationship between APOE ε3 and tau pathology. However, there was also an interaction between APOE ε3 and the presence of A$\beta$ such that the associations of APOE ε3 with T-tau and P-tau were much more robust in persons with A$\beta$, and it revealed that APOE ε3 was associated with decreased concentrations of CSF T-tau and P-tau in individuals with A$\beta$ deposition. In the stratified analyses regression within CN, MCI, and AD groups, the APOE ε3 allele was significantly associated with decreased CSF T-tau and CSF P-tau levels in the MCI with significant A$\beta$ deposition. These results were not observed in individuals without significant A$\beta$ deposition. Some studies reported that APOE ε2 carriers had reduced NFT [50, 51], though inconsistent findings exist [52, 53]. We did not find an interaction between APOE ε2 and the presence of A$\beta$ related to tau. APOE ε2 was only associated with decreased levels of CSF T-tau in MCI individuals with significant A$\beta$ deposition. Our results show that APOE ε4 and ε3 may only affect tau pathology in MCI patients, and A$\beta$ mediates this effect. This work indirectly supports the concept that APOE alleles influence tau pathology dependently on A$\beta$, and tau pathology without A$\beta$ may reflect a different pathological process from MCI.

A longitudinal study has reported that the relationship between APOE and global cognitive decline was mediated by A$\beta$ and tau [54]. It was also found that the effects of APOE on a decline in episodic memory and non-episodic cognition were mediated by A$\beta$ [30]. However, these findings did not divide the subjects according to the severity of cognitive impairment. We found an interaction between APOE ε4 and ε3 and the presence of A$\beta$ such that the associations of APOE ε4 and APOE ε3 with ADAS-cog were much more robust in persons with A$\beta$. When we considered all participants as a whole, there was a significant correlation between APOE ε4 and increased ADAS-cog scores and between APOE ε3 and decreased ADAS-cog scores in persons with significant A$\beta$ deposition but not in persons without significant A$\beta$ deposition. However, APOE ε2 was not associated with ADAS-cog in individuals with and without significant A$\beta$ deposition. In the stratified analyses regressing within CN, MCI, and AD groups, we revealed that APOE ε4 was only significantly associated with increased ADAS-cog scores in the MCI individuals with significant A$\beta$ deposition, and APOE ε3 was only significantly associated with decreased ADAS-cog scores in the MCI individuals with significant A$\beta$ deposition. APOE ε2 was not associated with ADAS-cog in the MCI and AD individuals with or without significant A$\beta$ deposition. Our work suggests that the effect of APOE ε4 and APOE ε3 on cognitive decline is only observed in MCI, and A$\beta$ also mediates this effect. In addition, it demonstrates that APOE ε3 has a protective effect on MCI but not AD, and APOE ε2 has no protective effect on MCI and AD. These seem to differ from previous conclusions that APOE ε3 is considered neutral and APOE ε2 is protective of AD risk. We do not know what the reason is, but we believe it is an interesting question for further research.

Our data suggest that the APOE genotype may only influence CSF tau and cognition in MCI participants. Just as we know, APOE ε4 likely predates the onset of A$\beta$ deposition [45], then A$\beta$ deposition initiates the cascade. Once A$\beta$ triggers the downstream process is, other factors will lead to the AD’s complete pathologic/clinical manifestations [55]. Therefore, we speculate that tau pathology and cognition in AD may be more affected by other factors, such as inflammatory factors, loss of cells, synapses, and dendrites and so on. The other possibility is that the groups are defined by being in a specific cognitive range, and the effect may not be noticed. However, future work is needed to determine why APOE genotype is only related to tau pathology and cognition in MCI patients. In addition, APOE ε3 was associated with lower amyloid (higher CSF A$\beta$42). Thus, it perhaps slows the trajectory of conversion from MCI to AD. However, its downstream signaling mechanism is still unknown, which may be an exciting topic in future research.

There are a few limitations. First of all, it lacks longitudinal data, so it cannot observe the dynamic impact of APOE on CSF tau and cognition. Secondly, it did not contain non-AD neurodegenerative disorders. Finally, the ADNI database consists of self-selected, highly educated volunteers interested in participating in AD research, which may concern their cognition. As such, our findings will benefit from replication in another population-based cohort.

We found that APOE ε4 and ε3 were associated with CSF tau and ADAS-cog. However, APOE ε4 and ε3 only affect tau pathology and cognitive function in MCI patients, and A$\beta$ mediates these effects. Thus, in addition to positron emission tomography (PET) data for A$\beta$ and tau, our findings highlight the need for future longitudinal studies examining the effects of APOE on tau and ADAS-cog.

A$\beta$, amyloid-$\beta$; AD, Alzheimer’s disease; ADAS-cog, Alzheimer’s disease assessment scale-cog; ADNI, Alzheimer’s disease Neuroimaging Initiative; APOE, Apolipoprotein E; CDR, Clinical Dementia Rating scale; CN, cognitively normal; CSF, cerebrospinal fluid; MCI, mild cognitive impairment; MMSE, Mini-mental State Examination; NFT, neurofibrillary tangles; PET, positron emission tomography; PHF, paired helical filament.

FX: manuscript drafting and composition of figures. TM: analysis of data. JT: collection of data. JL: interpretation of data. HZ: concept and supervision of the research.

The Institutional Review Boards approved the ADNI study of all the participating institutions. In addition, informed written consent was obtained from all participants at every center.

Data collection and sharing for this project was funded by the Alzheimer’s Disease Neuroimaging Initiative (ADNI) (National Institutes of Health Grant U01 AG024904) and DOD ADNI (Department of Defense award number W81XWH-12-2-0012). ADNI is funded by the National Institute on Aging, the National Institute of Biomedical Imaging and Bioengineering, and through generous contributions from the following: AbbVie, Alzheimer’s Association; Alzheimer’s Drug Discovery Foundation; Araclon Biotech; BioClinica, Inc.; Biogen; Bristol-Myers Squibb Company; CereSpir, Inc.; Cogstate; Eisai Inc.; Elan Pharmaceuticals, Inc.; Eli Lilly and Company; EuroImmun; F. Hoffmann-La Roche Ltd and its affiliated company Genentech, Inc.; Fujirebio; GE Healthcare; IXICO Ltd.; Janssen Alzheimer Immunotherapy Research & Development, LLC.; Johnson & Johnson Pharmaceutical Research & Development LLC.; Lumosity; Lundbeck; Merck & Co., Inc.; Meso Scale Diagnostics, LLC.; NeuroRx Research; Neurotrack Technologies; Novartis Pharmaceuticals Corporation; Pfizer Inc.; Piramal Imaging; Servier; Takeda Pharmaceutical Company; and Transition Therapeutics. The Canadian Institutes of Health Research is providing funds to support ADNI clinical sites in Canada. Private sector contributions are facilitated by the Foundation for the National Institutes of Health (https://www.fnih.org/). The grantee organization is the Northern California Institute for Research and Education, and the study is coordinated by the Alzheimer’s Therapeutic Research Institute at the University of Southern California. ADNI data are disseminated by the Laboratory for Neuro Imaging at the University of Southern California.

This research was funded by the Medical Research Project of Chongqing Healthy Committee, grant number 2018MSXM058.

The authors declare an interest in the Alzheimer’s Disease Neuroimaging Initiative.

The datasets used and/or analyzed in this study may be obtained from the corresponding author on reasonable request.