APOE genotype and early β-amyloid accumulation in older adults without dementia

Abstract

Objective:

To clarify associations between APOE ε4 allele and age on longitudinal rates of β-amyloid (Aβ) accumulation within Aβ+ and Aβ− older individuals without dementia.

Methods:

We analyzed 595 older adults without dementia classified cross-sectionally as Aβ− (n = 325) and Aβ+ (n = 270) using longitudinal florbetapir PET. The influence of age and APOE genotype on longitudinal accumulation of Aβ was examined with linear mixed models.

Results:

APOE ε4 and older age were associated with higher risk of being classified as Aβ+ at baseline. The annual rate of Aβ accumulation was significantly greater than zero for Aβ− ε3 (0.0021 ± 0.0007 standardized uptake value ratio [SUVR] units) and Aβ− ε4 (0.0044 ± 0.0010 SUVR units), as well as Aβ+ ε3 (0.0141 ± 0.0019 SUVR units) and Aβ+ ε4 (0.0126 ± 0.0018 SUVR units). Aβ accumulation was significantly faster in Aβ− ε4 compared to Aβ− ε3 and Aβ− ε2. Rates of Aβ accumulation did not differ significantly between Aβ+ APOE groups. Older age was associated with higher rates of Aβ accumulation in the Aβ− group.

Conclusions:

APOE ε4 carriage and older age were predictors of longitudinal Aβ accumulation within the Aβ− group but not the Aβ+ group. APOE ε2 carriage was protective against longitudinal Aβ accumulation within the Aβ− group. APOE genotype in conjunction with chronologic age may aid in participant selection for primary prevention trials aimed at halting Aβ accumulation before abnormal levels are reached.

Abnormal levels of β-amyloid (Aβ+) emerge many years before the clinical manifestation of Alzheimer disease (AD) dementia.¹ Characterizing the effect of early Aβ in the stages preceding dementia has been critical for understanding disease development as well as for the design of prevention trials targeting clinically asymptomatic at-risk adults.^2–4

The APOE ε4 allele is the strongest genetic risk factor for AD dementia and is consistently linked to abnormal Aβ aggregation.^5–7 While ε4 has been consistently associated with Aβ+ when measured cross-sectionally, the association between ε4 and longitudinal rates of Aβ accumulation remain unclear, especially after accounting for the strong association between APOE ε4 and baseline Aβ.⁸ However, as rates of Aβ accumulation reach a plateau at higher levels,⁹ it is possible that the effects of ε4 on increasing Aβ levels occur earlier, for example, in older adults whose absolute Aβ levels have yet reached any threshold for abnormality (Aβ−). Thus, it is critical to examine factors that influence longitudinal Aβ accumulation separately within Aβ− and Aβ+ groups without dementia to establish the utility of these factors for identifying at-risk individuals as early as possible. In addition, the effect of the APOE ε2 allele on longitudinal rates of Aβ accumulation is also unclear. Although ε2 carriers are less likely to develop AD dementia and more likely to be classified as Aβ− at baseline,^1,10–13 the relative low frequency of the ε2 allele has prevented previous studies from examining the effect of this protective factor on rates of longitudinal Aβ accumulation among participants without dementia.¹⁴

We therefore examined the effect of APOE genotype on rates of Aβ accumulation assessed using florbetapir PET imaging in 595 older adults who were cognitively normal or diagnosed with mild cognitive impairment (MCI). Importantly, the large sample size enabled exploration across multiple APOE groups (ε2, ε3, ε4), rather than just between ε4 carriers and noncarriers. We hypothesized that rates of Aβ accumulation would be fastest in ε4 carriers and slowest in ε2 carriers, and that an effect of APOE ε4 on longitudinal rates of Aβ accumulation would only be present within the Aβ− group.

METHODS

Participants.

The process of recruitment and inclusion/exclusion criteria for the Alzheimer's Disease Neuroimaging Initiative (ADNI) has been described previously.¹⁵ ADNI participants were included in this study if they were clinically normal or had MCI during the visit closest to their first florbetapir PET, and had completed at least 2 florbetapir PET scans (n = 595; figure e-1 at Neurology.org). Given that florbetapir data was collected during the ADNI2 phase, approximately 15% of participants (n = 89) who had enrolled during the ADNI1 phase received their first PET scan many years following baseline enrollment into the study. We therefore used diagnosis closest to each participant's first PET scan rather than baseline diagnosis.

Standard protocol approvals, registrations, and patient consent.

Institutional review boards of all participating institutions of ADNI provided approval for the study. All participants at each site provided written informed consent.

Outcome measures.

Global summary measures extracted from florbetapir scans were downloaded directly from the LONI website (UCBERKELEYAV45_06_15_16.csv).¹⁶ Summed images 50–70 minutes postinjection were coregistered to each participant's MRI to enable alignment of FreeSurfer regions of interest. PET values were averaged across 4 large regions (frontal, cingulate, parietal, and lateral temporal). Standardized uptake value ratios (SUVR) were derived by dividing the cortical summary measure by the mean uptake from the reference region. While the whole cerebellum reference region is commonly used to assess Aβ cross-sectionally,^16,17 recent data suggest that a larger composite region that incorporates cortical white matter may be superior for assessing longitudinal change in Aβ when measured with florbetapir PET.^18,19 As such, we incorporated both reference regions for classification of individuals using previously published cutoffs for the composite (0.79) and the whole cerebellum reference regions (1.11) (figure e-2).^18,20 Thus, participants were considered Aβ− only if they fell below both reference region cutoffs and Aβ+ if they surpassed both cutoffs. Of the 656 participants without dementia with longitudinal florbetapir data available, 325 were classified as Aβ− across both cutoffs, 270 were classified as Aβ+ across both cutoffs, and 61 were incongruently classified (figure e-2). As our aim was to understand how APOE genotype and age influence Aβ− and Aβ+ groups, we excluded the 61 participants who were incongruently classified to reduce confounds related to misclassification. Of the 325 Aβ− participants, 178 completed 2 florbetapir scans and 147 completed 3 florbetapir PET scans. Of the 270 Aβ+ participants, 175 completed 2 florbetapir scans and 95 completed 3 florbetapir PET scans. Each follow-up PET scan was separated by a median of 2 years. All longitudinal analyses predicting Aβ over time used the summary measure normalized by the composite reference region as the dependent variable.¹⁸

Statistical models.

To examine the effect of APOE group on change in florbetapir SUVR over time, we conducted linear mixed effects models in the statistical program R v3.3.1 (lme4 package), with interactions between baseline age (centered) × time, diagnosis (cognitively normal vs MCI) × time, and APOE group × time specified as fixed factors. Time was coded as time relative to each participant's first florbetapir scan, and was expressed as years. Random intercepts and slopes were included. APOE group was divided into 3 levels: APOE ε2 (comprising ε2/ε2 and ε2/ε3), APOE ε3 (ε3/ε3), and APOE ε4 (comprising ε4/ε4, ε3/ε4, and ε2/ε4). Contrasts were formed to determine (1) whether accumulation over time within each APOE group was significantly different from zero and (2) whether rates of accumulation were significantly different between APOE groups. Analyses were conducted separately in Aβ− and Aβ+ groups.

RESULTS

APOE ε4 carriers were more likely to be classified as Aβ+ at baseline whereas APOE ε2 carriers were less likely to be classified as Aβ+ at baseline (72.9% of ε4 and 15.3% of ε2 were Aβ+, compared to 30.6% for ε3, p < 0.001). Older age was also associated with a greater likelihood of Aβ+ classification at baseline (p = 0.003).

In general, APOE ε4 carriers were significantly younger compared to ε2 and ε3 groups (table 1). Specifically, the Aβ− ε4 group was younger than the Aβ− ε2 (p = 0.005) and Aβ− ε3 groups (p = 0.001). Similarly, the Aβ+ ε4 group was significantly younger than the Aβ+ ε3 group (p < 0.001), but not the Aβ+ ε2 group (p = 0.161). Groups based on joint Aβ and APOE status did not differ significantly on any other demographic or clinical characteristic (table 1).

Table 1.

Demographic and clinical characteristics

In the Aβ− group, ε4 carriers and ε3 homozygotes showed rates of Aβ accumulation significantly greater than zero (table 2). There was no significant increase in Aβ over time in Aβ− ε2 carriers. The rate of Aβ accumulation in Aβ− ε4 carriers was significantly faster than in Aβ− ε3 homozygotes and Aβ− ε2 carriers (table 2, figure 1). Among Aβ− adults, the interaction between diagnosis and time was insignificant, suggesting no difference in the rate of Aβ accumulation between cognitively normal and MCI Aβ− groups (table 3). The difference between the Aβ− ε4 and ε3 groups remained significant after exclusion of the 5 ε4 homozygotes (β = 0.002, p = 0.041). Baseline age was significantly associated with Aβ accumulation, such that older Aβ− participants showed faster rates of accumulation independent of APOE genotype (table 3). Examination of a 3-way interaction among APOE ε4, age, and time was not significant (p = 0.926), suggesting that the effects of APOE genotype and age were additive rather than synergistic.

Table 2.

Rates of β-amyloid (Aβ) accumulation for each Aβ/APOE group, and contrast matrix comparing rates of Aβ accumulation between groups (p values are listed)

Figure 1. Effect of APOE genotype on β-amyloid (Aβ) accumulation over time.

Overall, rates of Aβ accumulation are greater in the Aβ+ (B) group compared to the Aβ− (A) group, regardless of APOE group. Within the Aβ− participants, the APOE ε4 group shows the greatest rates of Aβ accumulation. SUVR = standardized uptake value ratio.

Table 3.

Results from linear mixed model examining longitudinal accumulation of β-amyloid (Aβ) among Aβ− and Aβ+ older adults without dementia in relation to APOE genotype (APOE ε3 as the reference group)

We additionally examined the proportion of Aβ− participants who progressed to an Aβ+ classification at follow-up. At the first follow-up scan (i.e., ∼2 years postbaseline), 7/325 (2.2%) Aβ− progressed to Aβ+ and there were no differences between APOE groups in progression rates (ε2 = 1/50; ε3 = 5/213; ε4 = 1/62), χ² = 0.21, p = 0.902. At the second follow-up scan (i.e., ∼4 years postbaseline), 7/147 (4.8%) Aβ− were classified as Aβ+, with a higher proportion of progressors in the APOE ε4 group (14%; 4/29) compared to the ε2 (4%; 1/25) and ε3 groups (2%, 2/93), χ² = 6.21, p = 0.045.

In the Aβ+ group, ε4 carriers and ε3 homozygotes showed rates of Aβ accumulation that were significantly greater than zero (table 2; figure 1). Aβ accumulation in Aβ+ ε2 carriers did not reach statistical significance (table 2). However, there were no statistically significant group differences in Aβ accumulation between Aβ+ APOE groups (table 2; figure 1). Within the Aβ+ group, the cognitively normal group showed significantly faster rates of Aβ accumulation than Aβ+ MCIs (figure 2, table 3). However, inclusion of a 3-way interaction among diagnosis, APOE genotype, and time was not significant (p = 0.210), suggesting that the pattern of Aβ accumulation across APOE groups did not differ by diagnosis (figure 2).

Figure 2. Annual rate of change in β-amyloid (Aβ), by APOE and diagnostic group.

Error bars represent standard error. A similar pattern related to APOE group is present between normal controls and mild cognitive impairment (MCI) groups, such that there is an effect of APOE ε4 only among participants who are Aβ− at baseline.

Given that Aβ− ε4 shows greater continuous levels of baseline Aβ compared to other Aβ− groups (table 1), it is possible that slightly elevated values within the Aβ− range account for the effect of ε4 within the Aβ− group. We therefore examined the simultaneous effect of APOE group and continuous levels of Aβ at baseline on Aβ accumulation over time within the Aβ− group. These analyses revealed that both ε4 and baseline Aβ were independently associated with Aβ accumulation within the Aβ− group (e-Methods, figure e-3, table e-1).

DISCUSSION

We aimed to determine whether APOE genotype and age were associated with rates of Aβ accumulation in individuals without a clinical diagnosis of AD dementia. Consistent with previous reports, we found that the APOE ε4 allele and older age were associated with a greater likelihood of being classified as Aβ+ at baseline.^1,10,21 Importantly though, we observed that APOE ε4 carriers showed faster rates of Aβ accumulation when compared to ε3 and ε2 groups, but only in Aβ− adults. Furthermore, APOE ε2 carriers were more likely to be Aβ− and showed the slowest rates of Aβ accumulation in the Aβ− group. Older age was associated with increased rates of Aβ accumulation, and this effect was independent of the effect of APOE. When taken together, these results highlight that through serial assessment of Aβ using florbetapir PET imaging, subtle but significant increases in Aβ over time are present among Aβ− older adults with established AD risk factors (i.e., ε4 and older age), and that the protective effect of ε2 against AD dementia may be mediated by slower rates of Aβ accumulation.

An increasing number of studies now show significantly increased rates of Aβ accumulation in Aβ+ older adults without dementia; however, the rate at which Aβ accumulates over time in Aβ− older adults is less clear, with multiple studies reporting no significant increases at the group level.^8,9,22 The lack of significant group level increases among Aβ− older adults may, in part, be due to insufficient follow-up, as well as heterogeneity across Aβ− participants, such that some participants show accumulation whereas others remain stable. Identification of factors that explain variable rates of Aβ accumulation among Aβ− older participants is increasingly relevant as the field moves towards interventions that attempt to delay clinical impairment by targeting Aβ as early as possible,^23,24 especially by targeting the Aβ protein before deposition into fibrillar Aβ plaques.²⁵ Our results suggest that this variability among Aβ− adults can be explained in part by APOE genotype and older age. Thus, consideration of both APOE genotype and age could aid in the selection of Aβ− adults at greatest risk of future Aβ accumulation.

Despite the strong mechanistic link between APOE ε4 and Aβ, the effect of ε4 on longitudinal rates of Aβ accumulation using in vivo biomarkers has remained unclear in previous studies. Recent work examining 2 large cohorts without dementia revealed that while ε4 is strongly associated with Aβ+ at baseline, there was no additional effect of ε4 on longitudinal rates of Aβ accumulation when considering the combined group of Aβ− and Aβ+ participants.⁸ However, it is noteworthy that there is a nonlinear association between baseline Aβ and longitudinal change in Aβ, with a positive relationship in Aβ− adults (e.g., within the Aβ− range, higher continuous levels of Aβ predict greater rates Aβ accumulation) whereas a plateau or even a negative association is observed in Aβ+ adults (e.g., within the Aβ+ range, higher continuous levels of Aβ become uncoupled from rates of Aβ accumulation).⁹ This decoupling between baseline and longitudinal change in Aβ as a function of overall Aβ magnitude implies that rates of Aβ accumulation can be differentially relevant depending on an individual's disease stage. Specifically, a faster rate of Aβ accumulation in Aβ− adults may indicate the very early stages of AD pathophysiology. Conversely, a steady rate of Aβ accumulation in Aβ+ adults may suggest a later disease stage as these participants reach an Aβ plateau. Given this dissociation, it is critical to examine risk factors that predict Aβ accumulation separately within Aβ− and Aβ+ groups. In doing so, we identified an association between ε4 and Aβ accumulation, such that among Aβ− adults, APOE ε4 was associated with increased rates of Aβ accumulation, whereas in Aβ+ adults, there was no difference across APOE genotypes in rates of Aβ accumulation. The differential relationship between ε4 and future Aβ accumulation highlights that the APOE ε4 allele provides additional information about risk of further accumulation specifically among Aβ− participants.

The APOE ε2 allele is a strong protective genetic factor against AD dementia.^11,26 Our results suggest that this protection is at least in part mediated by slower rates of Aβ accumulation. Specifically, we found that ε2 carriers were the only group to not show significantly greater rates of accumulation than zero in both Aβ− and Aβ+ groups. This pattern is consistent with previous reports that have shown that when compared to the ε3 allele, the ε2 allele is associated with increased neuroprotection against AD, through an increased rate of Aβ clearance.⁵ It will be important to establish whether Aβ+ ε2 carriers without dementia are also at lower risk of progressing to AD dementia, which may be due to the slower rates of accumulation in this group even though they have surpassed the Aβ+ threshold. However, our analyses investigating the effects of ε2 within the Aβ+ group were greatly limited by sample size (n = 9) and warrant further research. Nevertheless, a lack of significant Aβ accumulation was also present in the 50 Aβ− ε2 carriers, which lends support to the hypothesis that slower rates of Aβ accumulation may explain the effects of ε2 on lowering the risk of AD dementia.

We found that among older adults without dementia who underwent PET imaging at 3 timepoints (mean follow-up of 4 years), 4.7% of Aβ− adults progressed to Aβ+, and 57% of Aβ− adults who progressed to Aβ+ were ε4 carriers. This is consistent with a previous study that reported that 3.1% of Aβ− adults progressed to Aβ+ after 2.5 years, and 70% of those who progressed were ε4 carriers.²² However, our results are at odds with another study that reported a much higher rate of progression from Aβ− to Aβ+ (13% per year), and did not find any association with APOE genotype.²⁷ Given the small numbers that result when examining the incidence of surpassing the Aβ+ cutoff, and inconsistent results across studies, additional research is needed to investigate incidence of progressing to Aβ+ among Aβ− adults without dementia. It is likely that these estimates will be influenced by differences across cohorts or methodologic factors (e.g., amyloid PET tracer, choice of cutoffs used to define Aβ−). For instance, given our exclusion of ambiguous cases (i.e., participants with inconsistent Aβ classification across reference regions), we are likely underestimating the prevalence of progression from Aβ− to Aβ+ in our study. Given these considerations, and that surpassing a study-defined Aβ cutoff is somewhat arbitrary, it may be more insightful to examine Aβ accumulation over time as a continuous measure rather than as an event that occurs when the Aβ+ threshold is crossed.

Several caveats need to be considered when interpreting the results of our study. First, we focused on a single global summary measure of Aβ while regional measures may be more sensitive during the early stages of accumulation.²⁸ Second, we combined cognitively normal and MCI groups to enable examination across a large sample of Aβ− and Aβ+ older adults divided by APOE genotype. Although the associations we identified between APOE and Aβ accumulation appear consistent across diagnosis, it will be important for future studies with larger sample sizes to determine whether the APOE effect on Aβ accumulation is affected by clinical diagnosis. We did not examine the role of vascular risk factors in influencing Aβ accumulation, which may account for the effect of APOE ε4 or add independent information regarding risk of future Aβ positivity, as suggested by previous cross-sectional data.²⁹ Finally, while we found significant associations between Aβ accumulation and APOE genotype, these annual rates of increase are small, making it unlikely that rates of Aβ accumulation will be used as an endpoint in clinical trials that utilize short follow-ups. Despite this, the association between APOE ε4 and Aβ accumulation among Aβ− adults suggest that considering ε4 status for enrichment in clinical trials will improve the likelihood of including Aβ− adults at greater risk for AD pathophysiologic processes.

These caveats notwithstanding, the ability to measure the effect of ε4 on increasing rates of Aβ accumulation in Aβ− older adults suggests that very early disease processes can be detected with serial assessment of PET imaging. As such, APOE ε4 carriers may be an ideal population for primary prevention trials that aim to halt Aβ accumulation before abnormal levels are reached.

GLOSSARY

Aβ

β-amyloid

AD

Alzheimer disease

ADNI

Alzheimer's Disease Neuroimaging Initiative

MCI

mild cognitive impairment

SUVR

standardized uptake value ratio

AUTHOR CONTRIBUTIONS

Study concept and design: Y.Y.L., E.C.M. Data analysis and interpretation: Y.Y.L., E.C.M. Writing of manuscript: Y.Y.L., E.C.M. Critical revision of the manuscript for important intellectual content: Y.Y.L., E.C.M.

STUDY FUNDING

Data collection and sharing for this project was funded by the Alzheimer's Disease Neuroimaging Initiative (ADNI) (NIH 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 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 NIH (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. Y.Y.L. is supported by the National Health & Medical Research Council–Australian Research Council (NHMRC-ARC) Dementia Research Development Fellowship (APP1111603). E.C.M. has received funding from NIH grant K01 AG051718.

DISCLOSURE

Y. Lim currently serves as a consultant for Biogen and Cogstate Ltd. E. Mormino has also served as a consultant for Biogen and Eli Lilly. Go to Neurology.org for full disclosures.