Introduction

Apolipoprotein E (APOE) has been established unequivocally as the most important susceptibility gene for late-onset Alzheimer’s disease (LOAD). The ε4 polymorphism of APOE is the most common genetic risk factor for LOAD [1]. Thus far, APOE is the primary target of numerous studies investigating the disease’s underlying molecular neuropathology, pharmacological therapy, clinical progression, diagnosis, prevention, and treatment response.

Though the mechanism underlying APOE ε4 allele linked modulation of Alzheimer’s disease (AD) development is still not completely understood, emerging data suggest that APOE contributes to AD pathogenesis through a wide range of biological functions, including amyloid beta (Aβ)-dependent pathway and Aβ-independent pathway [2–4]. Some hypotheses suggest that APOE genotype is associated with AD biomarkers, with higher levels of Aβ deposition [5–7], higher degree and faster rate of neurodegeneration [8, 9], changes in the brain volume [10], function and glucose metabolism [11, 12], influence of the cerebrospinal fluid (CSF) measures of amyloid and tau [13, 14], and more severe impaired cognition [15–18], However, some other studies have yielded controversial results as to these biomarkers investigations associated with APOE [19, 20]. Furthermore, although there have been previous investigations of cognitive and neuroimaging differences between mild cognitive impairment (MCI) or AD patients who are APOE ε4 carriers vs. noncarriers, no prior work has brought these lines of research together towards normal cognition (NC), early mild cognitive impairment (E-MCI), late mild cognitive impairment (L-MCI), and AD to identify the underlying neuroanatomical basis of this genetically influenced gene. Therefore, the goal of this study is to evaluate the effect of APOE ε4 allele on neurodegeneration, cognition, CSF, and neuroimaging biomarkers across the AD spectrum.

Materials and Methods

Alzheimer’s Disease Neuroimaging Initiative

Data used in the preparation of this paper were obtained from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database (www.loni.ucla.edu/ADNI). ADNI was launched in 2003 by the National Institute on Aging, the National Institute of Biomedical Imaging and Bioengineering, the Food and Drug Administration, private pharmaceutical companies, and non-profit organizations as a $60 million, 5-year public-private partnership [21]. Subjects have been recruited from over 50 sites across the USA and Canada. To date, these three protocols have recruited over 1500 adults, ages 55 to 90, to participate in the research, consisting of cognitively normal older individuals, early or late MCI, and early AD. Further information can be found at www.adni-info.org for up-to-date information and previous reports [22, 23]. Data for this present analysis were downloaded from the ADNI web site in March 2014. This study was approved by institutional review boards of all participating institutions and written informed consent was obtained from all participants or authorized representatives.

Participants

To examine the APOE genetic influence across different clinical and cognitive status further, according to the Clinical Dementia Rating Scale (Sum of Boxes score, CDR-SB), we divided the AD group to the mild AD (M-AD, CDR-SB < 4.5) and severe AD (S-AD, CDR-SB ≥ 4.5) [24]. In this paper, a total of 1718 (S-AD = 180, M-AD = 156, E-MCI = 305, L-MCI = 561, NC = 516) individuals from the ADNI cohort (ADNI1, ADNI2, and ADNI-Go) whose data met all quality control (QC) criteria were included. Detailed quality control steps for CSF [25] and genotype data [26] have been previously reported. Table 1 lists the detailed demographics of all these subjects. Of the 1718 participants, our study included 461 subjects (including 49 AD, 263 MCI, and 149 NC) in MRI analysis, 1288 subjects (including 239 AD, 664 MCI, and 385 NC) in PET analysis, and 1037 subjects (including 343 NC, 301 E-MCI, 257 L-MCI, and 136 AD) in F18-PET-AV45 analysis.

Table 1 Baseline demographic, clinical, and neuroimaging characteristics of study participants
Full size table

Genotyping, Clinical and Neuropsychological Assessments

APOE genotyping was described in http://www.adni-info.org in detail. The APOE gene is polymorphic with the following three major isoforms: APOE ε2, APOE ε3, APOE ε4. In our analyses that controlled for APOE status (carrier of zero, one, or two APOE ε4 alleles), APOE ε4 carriers were coded as 0, 1, and 2, respectively.

All clinical and neuropsychological test performance data for included participants were downloaded from the ADNI clinical data repository on the Laboratory of Neuro-Imaging (LONI) site. Participants underwent a comprehensive battery of neuropsychological tests, but we only evaluated participant performance on the Mini-Mental State Exam (MMSE), CDR-SB, Alzheimer’s disease Assessment Scale (ADAS), Rey Auditory Verbal Learning Test (RAVLT, total), and Functional Activities Questionnaire (FAQ). Three participants were missing ADAS11 and ADAS13 data and additional five participants were missing FAQ data. Thus, the final samples for neuropsychological testing included 1715 in the ADAS11 and ADAS13 score analysis, and 1713 in the FAQ score analysis.

Cerebrospinal Fluid Data

Levels of Aβ1-42, tau, and phosphorylated-tau (p-tau) were measured from all available CSF samples as described previously [25, 27]. CSF data was downloaded from the LONI site and extracted for all included participants. Of the 1718 included participants, there were only 804 participants with detailed CSF tau, p-tau, and Aβ1-42 data. The final samples for CSF analyses included 221 NC, 192 E-MCI, 264 L-MCI, 62 M-AD, and 65 S-AD participants.

Neuroimaging Data

The neuroimaging data, such as regional volume on MRI, cerebral metabolic rate for glucose (CMRgl) on FDG-PET, and florbetapir cortical standardized uptake values ratios (SUVR) via F18-PET-AV45, were all downloaded from the ADNI dataset. The neuroimaging methods utilized by ADNI have been described in detail previously utilizing the calibration techniques to maintain the consistent protocols across scanners and sites [28].

In our study, we used the regions of interest (ROIs) analysis to calculate differences between APOE genotypes across the AD spectrum. Based on the revised guidelines [29] and our previous meta-analysis [18], we adopted the hippocampal volume and CMRgl in the regional volume to analyze the hippocampal neurodegeneration between the APOE genotypes and AD spectrum. Furthermore, we compared the changed values on hippocampal volume (percent atrophy of hippocampus from baseline) in the follow-up study of 2 years.

Five individual important hypometabolic ROI, including bilateral posterior cingular, left angular gyrus, right angular gyrus, left inferior temporal gyrus, and right inferior temporal gyrus, are downloaded from LONI [30]. These five predefined regions of interest (MetaROIs) are selected based on coordinates cited frequently in other FDG studies comparing AD, MCI, and NCs [31]. We extracted data of florbetapir means of F18-PET-AV45 from four regions (frontal, anterior/posterior cingulate, lateral parietal, and lateral temporal) and global florbetapir SUVR to calculate the amyloid burden.

Statistical Analysis

With each biomarker treated as a continuous scale, differences were calculated by one-way analysis of variance (ANOVA); for categorical marker’s data, differences were tested by Spearman’s correlation analysis. In order to examine the correlation between clinical disease severity and clinical biomarkers, we separate the AD group to the mild AD (M-AD, CDR-SB < 4.5) and severe AD (S-AD, CDR-SB ≥ 4.5) subgroups by their CDR-SB scores [32]. We completed statistics across five subgroups (NC, E-MCI, L-MCI, M-AD, and S-AD) to test the association of APOE ε4 genotypes with the biomarkers in subgroups containing no less than 10 individuals. Furthermore, a Multiple Linear Regression model which considered age, gender, and education as covariates in total sample and subgroups was used to estimate coefficients and the 95 % confidence interval (CI) for testing possible correlation between three APOE ε4 genotypes in these five cognitive groups. To explore this association further, we conducted a multiple linear regression analysis which consider age, gender, and education as covariates in total sample, and calculated the 95 % CI. A Bonferroni-corrected P value (P c) of 0.05 was considered significant, after adjusted for age, sex, and education. We also calculated the 24-month percent changes of cognitive scores, and 12-month and 24-month percent volumetric changes of hippocampus for longitudinal analysis. Sample of CSF and other neuroimaging biomarkers in 24 months was not sufficient for analysis. All statistical analyses were performed by SPSS 19.0 statistics for IBM.

Results

Clinical and Neuropsychometric Findings

The 1718 subjects’ baseline demographic characteristics, neuropsychometric scores, and APOE gene ε4 doses are described in Table 1. As expected, significant effects of diagnosis on the neuropsychometric scores CDR-SB, ADAS11, ADAS13, MMSE, RAVLT, and FAQ were observed (P c < 0.001): AD participants showed a greater CDR-SB, ADAS11, ADAS13, FAQ, as well as lower MMSE and RAVLT scores. Also as expected, the AD and MCI groups had significantly higher proportion of subjects with one or two copies of the APOE ε4 allele (P c < 0.01) than the NC group, with more than 60 % participants (223 out of 336) in AD compared to 28.5 % in NC.

The results of six neuropsychometric scores across AD spectrum are displayed in Supplementary Table S1. In E-MCI, heterozygotes and homozygotes showed significant higher ADAS11 scores than noncarriers (corrected P c = 0.015); in L-MCI, the ADAS11, ADAS13, and RAVLT scores were considerably different across the three genotype groups (P c < 0.01) (Supplementary Table S1 and Fig. S1). APOE ε4 carriers had significantly higher scores on ADAS11 and ADAS13 measures, as well as significantly lower scores on RAVLT measures. MMSE scores and FAQ scores did not show significant difference among the three ε4 allele groups.

CSF Biomarkers

APOE ε4 was significantly associated with lower levels of CSF Aβ1-42 throughout the five diagnostic groups, even after adjusting for age, gender, and education (noncarriers > heterozygotes > homozygotes, Bonferroni-corrected P < 0.01, Fig. 2a, Supplementary Table S2). CSF p-tau and tau protein levels showed remarkable difference in E-MCI and L-MCI subgroups (tau: E-MCI P c < 0.01, L-MCI P c < 0.01; P-tau: E-MCI P c < 0.01, L-MCI P c < 0.01), with the CSF p-tau and tau protein level in noncarriers lower than the others.

Hippocampal Neurodegeneration

As expected, hippocampal volume was indicated to be linearly correlated with the cognitive level (P < 0.001; Table 1). APOE ε4 allele was significantly associated with the hippocampal volumes in the L-MCI and M-AD groups (L-MCI, P c < 0.001; M-AD, P c = 0.015), with both heterozygotes and homozygotes showing more hippocampal atrophy than noncarriers. Besides, the subcortical volume of the left hippocampus was marked associated with the APOE ε4 allele in L-MCI participants (L-MCI, P c < 0.001): the more APOE ε4 dose the more hippocampal atrophy in the L-MCI participants (Supplementary Table S3). In L-MCI and AD groups, APOE ε4 carriers showed more right hippocampal atrophy than noncarriers (L-MCI, P c < 0.001; AD, P c = 0.036). However, no significant difference was shown between APOE ε4 genotypes and hippocampal CMRgl in either cognitive groups.

Neuroimaging biomarkers analysis

In our analysis of ROIs via MRI in baseline, APOE ε4 was significantly associated with several regions across the AD spectrum (see Table 2). In L-MCI, when adjusted for age, gender, and education, the difference remained in several regions: subcortical volume of the right thalamus, subcortical volume of the right ventral DC, subcortical volume of the left amygdala, subcortical volume of the left accumbens area, subcortical volume of the left hippocampus, subcortical volume of the right amygdala, and subcortical volume of the right hippocampus (P c < 0.05, Table 2). Patients with L-MCI who are APOE ε4 carriers exhibit greater atrophy in the right thalamus, right ventral DC, amygdala, left accumbens, and hippocampus.

Table 2 Significant results from analysis of MRI regions of interest with APOE genotypes across AD spectrums
Full size table

In the analysis of CMRgl on FDG-PET, only E-MCI group exhibited positive results. In E-MCI, the heterozygotes and homozygotes showed significantly lower CMRgl than the noncarriers on the four regions (right angular gyrus, left angular gyrus, left inferior temporal gyrus, and right inferior temporal gyrus; Supplementary Table S4). Besides, we did not observe significant difference on other groups.

Interestingly, the summary cortical florbetapir SUVRs were significantly associated with the APOE ε4 across the AD spectrum, with the carriers (one or two alleles) showing higher amyloid deposition than noncarriers (P c < 0.05, Fig. 1b). The mean cortical florbetapir SUVR was also associated with the APOE ε4 carriers across the AD spectrum (Supplementary Fig. S2). In the normal and mild cognitive groups (NC, E-MCI, and L-MCI), the highest SUVR (in frontal, cingulate, lateral parietal, and lateral temporal) in subjects with two alleles of APOE ε4 is increasing as the number of alleles increases (P c < 0.01, Fig. 1b).

Fig. 1
figure 1

CSF Aβ1-42 and summary florbetapir cortical SUVR by PET in APOE ε4-negative, homozygous, and heterozygous AD dementia patients. a CSF Aβ1-42 in APOE ε4-negative, homozygous, and heterozygous AD dementia patients. b Summary florbetapir cortical SUVR by PET in APOE ε4-negative, homozygous, and heterozygous AD dementia patients. Multivariate analysis of variance, with APOE dose as independent variable and age, gender, and education as covariates, showed a main effect for APOE dose (all P c < 0.01, except for the Global SUVR in M-AD P c = 0.075). AD = Alzheimer’s disease; M-AD = mild AD patients with high CDR-SB scores; S-AD = AD patients with low CDR-SB scores; SUVR = standardized uptake value ratio

Full size image

Follow-up Researches

Our results revealed that the APOE ε4 was significantly associated with cognitive decline in L-MCI. Remarkably, APOE ε4 carriers showed significant cognitive decline than noncarriers (P < 0.05, Fig. 2a). APOE ε4 was also significantly associated with the hippocampus volume, with carriers showing more atrophy than the noncarriers (P < 0.05, Fig. 2b). In addition, we detected strong associations on the hippocampus volume percent changes over 24 months from the longitudinal analysis (noncarriers < heterozygotes < homozygotes, L, P < 0.001; R, P < 0.001).

Fig. 2
figure 2

Change from baseline in late mild cognitive impairment (L-MCI). a Change from baseline on cognitive assessment measures in late mild cognitive impairment (L-MCI). Adjusted P value: CDR-SB P = 0.005; MMSE P = 0.13; RAVLT P = 0.002; FAQ P = 0.008. b Percent atrophy of the hippocampus from baseline in late mild cognitive impairment (L-MCI). L m12, left hippocampus volume, 12 months; R m12, right hippocampus volume, 12 months; L m24, left hippocampus volume, 24 months; R m24, right hippocampus volume, 24 months. Adjusted P value: L m12, P < 0.001; R m12, P = 0.004; L m24, P < 0.001, R m24, P < 0.001

Full size image

Discussion

This study provides a comprehensive evaluation of the impact of APOE ε4 status on CSF Aβ, tau, and p-tau levels; cognitive performance; cerebral atrophy; and brain metabolism across AD spectrum. In our study, we have three major findings: (1) APOE ε4 allele dosage was significantly associated with decreased CSF Aβ1-42 and increased cerebral amyloid deposition across the AD spectrum; (2) The APOE ε4 was significantly associated with increased CSF tau and p-tau in E-MCI and L-MCI subgroups. The ε4 carriers showed significantly cerebral hypometabolism than ε4 noncarriers only in E-MCI; (3) Hippocampal atrophy was associated with APOE ε4 allele in L-MCI and M-AD subgroups. In L-MCI, the APOE ε4 was significantly associated with atrophy of several cerebral regions, as well as cognitive decline manifested by higher ADAS11, ADAS13 and lower RAVLT scores.

In the present study, we confirm that APOE ε4 has a powerful dose-dependent effect on cerebral Aβ deposition. Previous researches had provided evidence on the important role of APOE genotypes in Aβ metabolism [33, 34]. Our findings were consistent with the previous reports that participants (including NC, MCI, and AD) carrying APOE ε4 alleles had lower CSF Aβ1-42 levels than those without an ε4 allele [35, 27, 6] and in ADNI 1 cohort [25]. Current study suggests that APOE genotype strongly affect deposition of Aβ in the brain. Cross-sectional studies by PiB-PET or Florbetapir-PET have consistently reported that cortical Aβ levels were increased in APOE ε4 carriers of E-MCI, L-MCI, and AD [25, 6, 7, 36]. Interestingly, the associations between APOE and amyloid-PET had mainly been seen in NC subjects, but less observed in AD dementia [37]. Although several studies in AD patients have produced paradoxical results against our positive results [38], it should be noted that these studies have focused on clinically manifest AD or severe AD, which may affect the pathologic status significantly.

Nevertheless, the accordance of the both measures of Aβ, Aβ imaging and direct measuring of CSF Aβ concentrations, indicates that cerebral Aβ deposition is the major pathobiological phenotype of APOE ε4 genotype, which would persist during the dementia development. APOE ε4-associated cerebral Aβ deposition could be interpreted by the hypothesized effect of APOE on Aβ clearance and aggregation. Biochemical evidence had shown that the APOE genotype could affect Aβ clearance rate, with the APOE ε4 isoform showing significantly slower clearance [39, 40]. Neuropathological evidence also suggested that APOE ε4 dosage is associated with increased Aβ, Aβ oligomers, and plaque accumulation in the brain [41].

With the analysis across the spectrum of AD, our data indicate that APOE ε4 affects other biomarkers mainly in prodromal stages (E-MCI and L-MCI) of AD. It seems to affect CSF tau, p-tau and regional FDG metabolism in E-MCI stage, and affect the pattern of regional brain atrophy and hippocampal atrophy in L-MCI. These relationships all direct to the APOE-related neurodegenerative function [42].

Our evidence of significant brain Aβ deposition, either from amyloid imaging or CSF Aβ1-42 concentrations, precedes the cerebral neurodegeneration and clinical cognitive changes. Interestingly, the results presented herein are completely consistent with the hypothetical model of dynamic biomarkers for the progression of neuropathology associated with AD (Fig. 3) [42, 27]. These support the hypothesis that the APOE ε4 has a leftward (during the disease progression) shift function of both the Aβ and neurodegenerative biomarkers cascades.

Fig. 3
figure 3

Modulator of APOE ε4 on hypothetical model of biomarkers change across the cascade of Alzheimer’s disease (AD) pathologic progression. Aβ is identified by CSF Aβ42 or PET amyloid imaging. Tau-mediated neuronal injury and dysfunction is identified by CSF tau or fluorodeoxyglucose-PET. Brain structure is measured by use of structural MRI. Aβ = β-amyloid. MCI = mild cognitive impairment

Full size image

The strength of this study is that it shows relative changes in AD processes across the spectrum of AD. However, interpretations of these results are not certain because the current analyses are based on cross-sectional data, which do not represent individual longitudinal changes. Although relatively large for this study, one limitation of this report is the special ADNI cohort, which is not a population-based cohort. Nevertheless, studies including larger samples are needed to further examine the effect of APOE functions. In addition, not all known biomarkers of AD, including PET-PIB and advanced MRI techniques, were included in this study because of the small sample.

On the basis of our results, we propose that cerebral Aβ deposition is highly associated with APOE genotype, which initiate the pathological cascade of preclinical AD. These were majorly consistent with a recently proposed biomarker model [42], which was hypothesized that accumulation of amyloid-β initiates a cascade of downstream effects, such as neuronal dysfunction and neurodegeneration, which would be amplified by the gene mutation. Therefore, these findings underline the importance of APOE for the disease conversion and progression, which are mediated by its effects on Aβ deposition and a cascade of downstream effects. Elucidating the contribution of APOE ε4 to the neuropathology of AD is a considerable challenge for us, but it provides great support for combating AD.