Abstract
The apolipoprotein E (ApoE) ɛ4 allele is a well-established genetic risk factor for sporadic Alzheimer’s disease. Some evidence suggests a negative role of the ApoE ɛ4 allele for cognitive performance in late life, while beneficial effects on cognition have been shown in young age. We investigated age-related effects of the ApoE gene on brain function by assessing cognitive performance, as well as functional activation patterns during retrieval of Face-Name pairs in a group of young (n = 50; age 26.4±4.6 years, 25 ɛ4 carriers) and old (n = 40; age 66.1±7.0 years, 20 ɛ4 carriers) participants. A cross-sectional factorial design was used to examine the effects of age, ApoE genotype, and their interaction on both cognitive performance and the blood oxygenation level dependent (BOLD) brain response during retrieval of Face-Name pairs. While there were no genotype-related differences in cognitive performance, we found a significant interaction of age and ApoE genotype on task-related activation bilaterally in anterior cingulate gyrus and superior frontal gyrus, as well as left and right insula. Old age was associated with increased activity in ɛ4 carriers. The increased BOLD response in old ɛ4 carriers during retrieval could indicate a neurocognitive disadvantage associated with the ɛ4 allele with increasing age. Furthermore, recruitment of neuronal resources resulted in enhanced memory performance in young ɛ4 carriers, pointing to a better neurofunctional capacity associated with the ApoE4 genotype in young age.
INTRODUCTION
Presence of the apolipoprotein E (ApoE) ɛ4 allele on chromosome 19 and a higher age significantly increase the risk of developing Alzheimer’s disease (AD) [1].
Converging evidence suggests that ApoE ɛ4 is involved in amyloid-β (Aβ) aggregation, Aβ clearance, Aβ fibrillization, and tangle formation thereby contributing to AD pathogenesis [2, 3].
Neuroimaging studies of nondemented older adults utilizing event-related functional magnetic resonance imaging (fMRI) paradigms have shown that ɛ4 carriers more strongly recruit task-related brain regions than non-carriers. This has been interpreted as an effective compensatory mechanism that counters the cognitive deficits associated with neurobiological decline in older persons genetically predisposed to dementia [4, 5]. While compensatory recruitment of task-related brain regions in ɛ4 carriers may initially be effective for maintaining cognitive performance, it might no longer be sufficient once AD pathology increases and cognitive decline ensues. The impact of the ɛ4 allele on cognitive functioning becomes present in the sixth decade and is reflected in poorer performance of ɛ4 carriers especially in memory tests [6–10]. The impact of ApoE ɛ4 on cognitive function in young adults is equivocal. While some studies indicate better cognitive performance in young ɛ4 carriers [11–15], other studies have shown no ApoE genotype dependent differences in cognitive performance [16–19].
The findings of poorer performance in ɛ4 carriers after the sixth decade of life together with the findings of a beneficial effect of ApoE ɛ4 in younger age have led to the suggestion that expression of ApoE ɛ4 could be an example of antagonistic pleiotropy [13, 21]. Antagonistic pleiotropy refers to a gene that expresses a beneficial effect in young age, but has a more deleterious effect later in life [22]. Up to now it is unclear if the hypothesis of antagonistic pleiotropy in regard to ApoE ɛ4 holds true. There is some evidence for (e.g., [10, 23–25]) and against (e.g., [26–28]) this hypothesis. One problem in determining the validity of the hypothesis is a difference in methodological approaches across studies that have either assessed young or elderly subjects. Thus, although some studies have shown beneficial effects of ɛ4 in young subjects and other studies have shown deleterious effects of ɛ4 in elderly subjects, this could be a consequence of different cognitive tests and fMRI protocols rather than an expression of antagonistic pleiotropy.
To our knowledge, there are only two studies that have used the same task-associated imaging protocol (fMRI and MRI) in young and old ɛ4 carriers and matched controls [15, 29]. Using an fMRI paradigm for episodic encoding, Filippini et al. [29] found that overactivity of brain function in young ɛ4-carriers was disproportionately reduced with advancing age. More specifically, increasing age was associated with decreases in blood oxygenation level dependent (BOLD) signal during encoding in ɛ4-carriers, whereas non-carriers showed a reverse pattern of decreased BOLD signal during encoding with increasing age. The age-associated change in neuronal activity did not have an influence on cognitive performance. Young and old ɛ4-carriers performed equally well in a follow up task outside the scanner when they were asked to recognize the images they had just seen during the fMRI task. Also, there was no effect of genotype on recognition performance: ɛ4 carriers and non-carriers performed equally well, irrespective of age. Thus, although a significant genotype×age interaction for the BOLD response was found in several brain areas, the nature of the interaction (decreased BOLD signal with increasing age), as well as the lack of performance differences provide no evidence for antagonistic pleiotropy. Nichols et al. [15] also used an fMRI paradigm for episodic memory (encoding and retrieval of neutral scenes) to investigate genotype x age interactions in 133 participants aged between 19–77 years. They found that the ApoE ɛ4 genotype was associated with decreased hippocampal activations during encoding and retrieval in young age, whereas hippocampal activations were increased in ɛ4 carriers compared to non-carriers in old age. Interestingly, they also found a cognitive advantage of young ɛ4 carriers in a delayed recall task outside the scanner.
The goal of the present study was to test age related effects of the ApoE ɛ4 genotype on both brain activation and cognitive performance. We used the same fMRI paradigm in young and elderly ɛ4 carriers consisting of a Face-Name association task and analyzed the BOLD signal associated with successful retrieval of Face-Name pairs. Additionally, we used an extensive neuropsychological assessment battery that captured performance of young and old participants in several cognitive domains such as verbal and visual memory, working memory and attention. By applying the same methods and assessment instruments to all participants, we aimed to shed light on the impact of ApoE ɛ4 on brain function across the lifespan.
To our knowledge, this is the first study to use an fMRI paradigm across different age groups that is especially sensitive for AD pathology in order to investigate age-related effects of the ApoE genotype. The Face-Name paradigm used in this study has been shown to reliably activate the hippocampus and related structures in the medial temporal lobe that are among the first structures to be affected by AD pathology [30]. Furthermore, by combining the fMRI results with the outcomes of a thorough neuropsychological assessment in both age groups, we aimed to shed light on the impact of ApoE ɛ4 on brain function across the lifespan. In accordance with the hypothesis of antagonistic pleiotropy, we expected to find a neurocognitive advantage of ɛ4 carriers in young age and a reversal of the ɛ4 effect in old age, with ɛ4 being associated with poorer neurofunctional capacity and poorer cognitive performance.
MATERIALS AND METHODS
Participants
We examined a subset of 90 cognitively intact participants (young group: mean age 26.4±4.6 years; old group: mean age 66.91±7.0 years) drawn from a larger cohort of 250 subjects. The larger cohort included 117 young (age 20 to 39 years) and 133 old participants (age 55 to 80 years) without any history of neurological or psychiatric disease. Participants were recruited through various advertisements in local and national newspapers. The majority of old participants were recruited via a cooperation with the Memory Clinic at Goethe-University Frankfurt am Main. The selection of the 90 participants analyzed here was based on the genotype. 24 subjects who were heterozygote for ApoE ɛ4 (ɛ3/ɛ4) and one subject who was homozygote for ApoE ɛ4 were included into the young ɛ4 positive (ɛ4+) group. 25 ɛ4 negative subjects (ɛ3/ɛ3), matched for age, sex, and education were included into the young ɛ4 negative (ɛ4–) group. 18 subjects who were heterozygote for ApoE ɛ4 (ɛ3/ɛ4) and two subjects who were homozygote for ApoE ɛ4 were included into the old ɛ4+ group. 20 ɛ4 negative subjects (ɛ3/ɛ3), matched for age, sex, and education were included into the old ɛ4- group. The local ethics committee of the Goethe University Frankfurt approved the study. All subjects declared that they understood the experimental procedure and signed a written informed consent. The study was undertaken in accordance with the Code of Ethics of the World Medical Association (Declaration of Helsinki) [31]. All subjects were right-handed (Edinburgh Handedness Inventory) [32].
All participants underwent ApoE genotyping. DNA was extracted from whole-blood samples. The DNA extraction and genotyping process was conducted at bio.logis laboratories (Frankfurt a.M., Germany). ApoE genotyping of the two determinating variants rs7412 and rs429358 was amplified using PCR and analyzed with pyrosequencing. The resulting sequences were compared to established sequence variants of the ApoE allele.
Verbal learning and memory were assessed using the German Version of the California Verbal Learning Test (CVLT [33]; German version by [34]). Additionally, measures of working memory and attention were obtained using the Letter Number Span (LNS) [35], the Spatial Span of the Wechsler Memory Scale (WMS-SS [36]; Germany version by [37]) and Trail Making Test A (TMT-A) [38]. The verbal IQ was tested with a German verbal intelligence test: Mehrfachwahl Wortschatz Test-B (MWT-B) [39]. Additionally, the CERAD–NP (Consortium to Establish a Registry for Alzheimer’s Disease) [40] was done with the old group to screen for incipient dementia. Participants were excluded from the current study if they scored at least 1.5 standard deviations below the norm (adjusted for sex, age, and education) in any of the CERAD’s subtests to rule out mild cognitive impairment or dementia. Depressive symptoms were assessed using the German Version of the Beck Depression Inventory (BDI II [41]; German adaptation by [42]). Group characteristics are summarized in Table 1.
Experimental task
We used event-related fMRI to identify brain activation associated with encoding and recognition of Face-Name pairs. The design of our paradigm was based on a Face-Name association task originally developed by Sperling et al. [43]. The stimuli consisted of 30 grey-scale photographs of forward-facing, neutral faces taken from the AR face database [44]. Additionally, 120 popular first names were selected from a German online name database (http://www.beliebtevornamen.de). Thirty names were randomly paired with study faces (during encoding and retrieval) and 90 were used as distractors (during retrieval only). Encoding and retrieval tasks were alternated across six runs, with each retrieval run testing memory for Face-Name pairs encoded in the previous run. The interstimulus interval (ITI) consisted of a checkerboard pattern. The duration of the ITI was kept variable (from 8 s to 12 s) in order to “jitter” the onset times of trials and thus minimize multicolinearity of event-related fMRI analyses. During encoding, 30 face–name pairs were presented for 8 s each. During retrieval, the previously shown Face-Name pairs were presented together with three distractor names in a pseudo-randomized order. Subjects were instructed to indicate the correct name by a button press. Each face was presented together with four names (one target name, three distractors) for 8 s. Participants performed three encoding runs and three retrieval runs, each lasting approximately 4 min. All responses were logged using Presentation Software (Version 10.3 Neurobehavioral Systems Inc.).
Twenty min after MRI-scanning participants received a sheet with the same Face-Name pairs (together with three distractor names) as in the scanner and were asked to mark the correct names. This additional task was introduced in order to assess the differential effects of ApoE genotype on short-range episodic memory as well as on delayed retrieval performance.
Behavioral data analysis
Since the neuropsychological test variables and the behavioral data acquired in the scanner were not normally distributed, a non parametric test (Mann-Whitney U-test) was applied for group comparisons. For categorical variables a chi-square test was applied.
For statistical analyses that comprised more than one between-subject factor (age and ApoE genotype), two methods were applied, depending on whether normal distribution or homoscedasticity were present. If the assumptions for parametric tests were met, a mixed Analysis of Variance (ANOVA) was applied. If the assumptions for parametric tests were not met, the van-Elteren Test, a stratified Wilcoxon-Mann-Whitney-U-Test was applied.
Performance (reaction times and their influence on accuracy) for the retrieval of Face-Name pairs was analyzed using a mixed ANOVA with ApoE and age as between-subject factors and reaction time and accuracy (repeated measures) as within-subjects factors.
MRI hardware and procedure
All MR images were acquired using a Trio 3-T scanner (Siemens, Erlangen, Germany) with a standard head coil for radiofrequency transmission and signal reception. Participants were outfitted with protective earplugs to reduce scanner noise and a hand-held response device. Stimuli were presented and button press logged using Presentation Software (Version 10.3 Neurobehavioral Systems Inc.). For T1 weighted structural brain imaging, an optimized 3D modified driven equilibrium Fourier transform (3D MDEFT) sequence [45] with the following parameters was conducted: acquisition matrix = 256×256, repetition time (TR) = 7.92 ms, echo time (TE) = 2.48 ms, field of view (FOV) = 256 mm, 176 slices, 1.0 mm slice thickness. Functional images were subsequently acquired using a T2* weighted Echo-Planar-Imaging (EPI) sequence with the following parameters: acquisition matrix = 64×64, repetition time (TR) = 2000 ms, echo time (TE) = 30 ms, field of view = 192 mm, flip angle = 90°. Thirty 3-mm thick axial slices with a 0.6 mm gap were acquired within each TR. The fMRI time series consisted of 3 encoding and 3 retrieval runs varying in length between 116 and 121 images.
Image processing
Brain Voyager QX 2.3 (Brain Innovation Maastricht, the Netherlands) was used to analyze the fMRI data [46]. Anatomical data was preprocessed with intensity inhomogeneity correction and transformed into Talairach space. Preprocessing of functional data included three-dimensional motion correction to overcome minor head movement during the scan, slice scan time correction, spatial smoothing with a 4 mm Gaussian kernel (full width at half-maximum) to accommodate inter-subject anatomical variability, temporal high-pass filtering to remove low-frequency non-linear drifts of three or fewer cycles per time course and linear trend removal. The complete set of functional data of each subject was co-registered with the anatomical scans, transformed into Talairach space and resampled to a voxel size of 3×3×3 mm3. The first two scans per run were discarded to permit T1 equilibration effects.
General linear models (GLM) were calculated separately for the older group, the younger group and the whole sample, yielding three different GLMs. Each GLM was modeled as a z-transformed separate subject predictors GLM with otherwise default settings. A brain mask (grey matter mask) was used to restrict the analysis to grey matter only.
Successful and unsuccessful retrieval activity for Face-Name associations was identified in the test-phase. Each GLM for retrieval included three conditions as separate predictors (successful retrieval, unsuccessful retrieval, ITI). Trial-related fMRI activity was modeled by convolving the predictors with a canonical hemodynamic response function. In the first level of a random effects analysis, condition effects for each subject (beta-values) were estimated. The obtained beta values served as input for the calculations of statistical comparisons between experimental conditions. Activations associated with successful retrieval (successful retrieval > ITI) were calculated for all subjects using linear contrasts (t-statistics). False discovery rate (FDR) was used for the correction of multiple comparisons. Between groups comparisons (ApoE ɛ4+ versus ApoE ɛ4– and Young versus Old) were performed for the same contrasts. For the between groups comparison of the retrieval contrasts (successful retrieval > ITI), an initial voxel level threshold was set to p < 0.05 uncorrected, resulting in a cluster level of 918 mm3 (34 contiguous voxels) for genotype and 1080 mm3 (40 contiguous voxels) for age after 1000 iterations at a corresponding false positive probability of Alpha 5% or less. The resulting statistical maps were visualized on surface reconstructions of the MNI template brain (courtesy of the Montreal Neurological Institute).
In order to test for potential interaction effects of age and ApoE genotype on the BOLD signal for the retrieval contrast (successful retrieval > ITI), a two factorial ANOVA with age and genotype as between subject factors was computed. For the analysis, an initial voxel level threshold was set to p < 0.05 uncorrected, resulting in a cluster level of 2241 mm3 (83 contiguous voxels). The areas showing a significant (p < 0.05, cluster threshold corrected) interaction effect of age and ApoE genotype were visualized on surface reconstructions of the MNI template brain (courtesy of the Montreal Neurological Institute).
Statistical analysis: correlation
In order to investigate the association between brain activity and memory performance, patterns of correlation between individual differences in retrieval-related brain response and psychometric measures were examined for the older ɛ4 carriers, older non-carriers, young ɛ4 carriers and young non-carriers separately. For each brain region that showed significant age x genotype interactions (p < 0.05, cluster threshold corrected) in the BOLD signal for the “successful retrieval versus ITI” contrast, measures of brain activity (β-values) were extracted and correlated with Face-Name recognition accuracy (during scanning and 20 min post-scanning) as well as immediate and delayed recall scores of the CVLT using Kendall’s tau correlation.
Anatomic imaging and voxel-based morphometry analysis
The T1-weighted volume was segmented using SPM8 (Wellcome Department of Cognitive Neurology, London, UK) running under Matlab 8 (Mathworks, Sherborn, MA, USA) and used for a voxel-based morphometry (VBM) analysis of grey matter volume (GMV) to determine whether there were any regional GMV differences that could confound the interpretation of the BOLD response differences with regard to ApoE ɛ4 status. Data preprocessing included spatial smoothing with a Gaussian Kernel of 10 mm full width at half maximum (FWHM) and spatial normalization using the DARTEL toolbox (SPM8). Voxel-wise GLM was applied to test for effects of genotype, age and age by genotype interaction using familywise error (FWE) correction for multiple comparisons.
RESULTS
Neuropsychological test results
The Mann-Whitney U-test was used to analyze differences across the ApoE groups (ApoE ɛ4+ versus ApoE ɛ4–) and age groups (young versus old) in neuropsychological test variables (CVLT, TMT-A, LNS, WMS-SS, MWT-B, CERAD). The van-Elteren test did not reveal any significant interaction effects for age and ApoE genotype across all neuropsychological test variables.
Also, there were no significant differences for the ApoE genotype in any of the test variables. With respect to age, significant differences were found across all tests for working memory and attention (TMT-A, WMS-SS, LNS), verbal memory (CVLT), and intelligence (MWT-B) where the younger participants scored significantly higher than the older participants. See Table 1 for details. CERAD test variables were not examined for age difference since only the old group performed the CERAD. Based on the CERAD results none of the older participants showed a cognitive performance indicative of dementia or mild cognitive impairment. The van-Elteren test did not reveal any significant interaction effects for age and ApoE genotype across all neuropsychological test variables.
Behavioral results
Associative retrieval of Face-Name pairs was tested both inside the scanner (immediate retrieval) and outside the scanner (delayed retrieval). The van-Elteren test did not reveal any significant interaction effects for age and ApoE genotype across these test variables.
Both ɛ4 carriers and non-carriers performed equally well when retrieving Face-Name pairs inside the scanner (71.7% accuracy in ɛ4+ subjects and 73.9% in ɛ4– subjects) and also subsequently after the scanning session (64.7% accuracy in ɛ4+ subjects and 66.1% in ɛ4– subjects). Within the young and the old groups, the ApoE genotype had also no effect on retrieval performance (Table 2). Age had a strong effect on immediate and delayed retrieval performance: young participants showed a significantly higher accuracy for immediate retrieval (82.1% accuracy in young participants versus 61.2% accuracy in elderly participants, p < 0.001) and delayed retrieval (78% accuracy in young participants versus 49.7% accuracy in elderly participants).
Reaction times were analyzed with a linear mixed model. Age and ApoE genotype groups as well as the correctness of the answer were used as fixed effects (main effects and interaction) and the intercept as a random effect. Post-hoc comparison of groups was performed with applied Bonferroni correction. All other options in SPSS were left at default, including scaled identities as covariance matrix and Restricted Maximum Likelihood as estimation approach.
Both age and correctness showed a significant interaction [F(1) = 29.84, p < 0.001]. While young participants showed a shorter reaction time for correct answers (M = 3530.34 ms) than for incorrect answers (M = 4428.96 ms), the older participants showed slightly longer reaction times for the correct (M = 5025.61 ms) than for the wrong answers (M = 4992.22 ms). No effect for an interaction of the ApoE genotype was observed.
The model showed a significant influence of age [F(1) = 49.61, p < 0.001] and correctness [F(1) =25.71, p < 0.001] on reaction time. Post-hoc testing revealed that young participants (M =3668.97 ms) had a shorter reaction time than the older participants (M = 5016.25 ms). Equally, participants had shorter reaction times when responding correctly (M = 4277.98 ms) compared to when responding incorrectly (M = 4710.59 ms).
fMRI results
Functional activations during successful retrieval
For all subjects, successful retrieval of Face-Name pairs (successful retrieval > ITI) was associated with increases in BOLD signal in a number of brain regions, including: bilateral dorsolateral prefrontal cortex, ventrolateral prefrontal cortex, middle frontal gyrus, anterior cingulate gyrus, fusiform gyrus, regions of the medial temporal lobe (parahippocampal gyrus and hippocampus), thalamus, right angular gyrus, and left amygdala. Decreases in BOLD signal during successful retrieval were located bilaterally in the posterior cingulate cortex medial prefrontal cortex and bilateral precuneus (p < 0.0001, FDR corrected). There was a main effect of age and genotype as well as an age x genotype interaction on brain activation.
Age x genotype interaction effects on brain activation
The 2 factorial ANOVA with Age and ApoE genotype as between-group factors revealed significant age x genotype interactions (p < 0.05, cluster threshold corrected) in the BOLD signal for the “successful retrieval versus ITI” contrast (Fig. 1, Table 3) in four brain regions: right and left insula, bilateral cingulate gyrus, and the bilateral superior frontal gyrus. Plotting the β-values (measures for brain activity) for each region revealed that aging was associated with increased brain activity in ɛ4 carriers (but not in non-carriers). Regarding the effect of age within genotype groups, all brain regions showed an increase of the BOLD signal with age in ɛ4 carriers. In non-carriers, only the superior frontal gyrus showed an increase in the BOLD signal with increasing age. For right and left insula and the cingulate gyrus, old ɛ4 carriers showed stronger activations than non-carriers, while this pattern was reversed in young participants. Taken altogether, there was a pattern of increased activation with age in the ɛ4 carriers, while this was not the case in non-carriers. Moreover, old ɛ4 carriers showed stronger activations during Face-Name recognition than all other groups.
Main effect of age on brain activation
Compared to elderly participants, young participants showed stronger activations in anterior cingulate cortex bilaterally, right middle frontal gyrus, and left caudate nucleus. In contrast, stronger activations in elderly participants compared to young participants were found in right parahippocampal gyrus, bilateral anterior cingulate cortex, bilateral inferior frontal gyrus, right angular gyrus, left fusiform gyrus, left thalamus, and right cerebellum. Stronger deactivations in young participants compared to elderly participants were found in posterior cingulate cortex bilaterally and in right fusiform gyrus. Stronger deactivations in elderly participants compared to young participants were found in left medial frontal gyrus (p < 0.05, corrected for cluster size) (Fig. 2A, Table 3).
Main effect of genotype on brain activation
Between-group statistical comparisons revealed a significant effect of ApoE genotype in several brain regions. Relative to non-carriers, ɛ4 carriers showed increased activity in left hippocampus, left parahippocampal gyrus, bilateral middle and superior temporal gyrus, bilateral anterior cingulate cortex, bilateral superior frontal gyurs, right dorsolateral prefrontal cortex, and left ventrolateral prefrontal cortex during successful retrieval (p < 0.05, corrected for cluster size). Also, ɛ4 carriers showed weaker deactivations in left medial prefrontal cortex during successful retrieval (p < 0.05, corrected for cluster size). Conversely, there was no increased activation in non-carriers compared to ɛ4 carriers during successful retrieval of Face-Name pairs (Fig. 2B, Table 3).
Correlation analysis
There were significant correlations between measures of memory performance and measures of brain activity in all brain regions that showed significant age x genotype interactions. In young ɛ4 carriers, brain activity in all brain regions was significantly positively correlated with memory performance. Activity in the right insula was significantly correlated with CVLT immediate free recall (τ= 0.35, p = 0.016). Brain activity in superior frontal gyrus was significantly correlated with Face-Name recognition performance during scanning (τ= 0.42, p = 0.006), with delayed Face-Name recognition performance 20 min after scanning (τ= 0.29, p = 0.043) and with CVLT delayed free recall (τ= 0.43, p = 0.005). Brain activity in anterior cingulate gyrus was significantly correlated with CVLT immediate free recall (τ= 0.32, p = 0.026). Brain activity in left insula was correlated with CVLT immediate free recall (τ= 0.32, p = 0.026). In old ɛ4 carriers, brain activity in anterior cingulate gyrus was significantly inversely correlated with Face-Name recognition performance 20 min after scanning (τ= –0.38, p = 0.024), with CVLT immediate free recall (τ= –0.41, p = 0.016) and with CVLT delayed free recall (τ= –0.40, p = 0.019). In old non-carriers, brain activity in right insula was significantly inversely correlated with Face-Name recognition performance 20 min after scanning (τ= –0.32, p = 0.027). Brain activity in young non-carriers was not significantly correlated with memory performance. Taken together, increasing brain activity was related to improved memory performance in young ɛ4 carriers, while this pattern was reversed in old ɛ4 carriers.
Voxel based morphometry
There were no significant grey matter volume differences between ɛ4 carriers and non-carriers (FWE correction, p > 0.05). Also, there were no significant interactions between age and genotype (FWE correction, p > 0.05). Significant age-related differences in grey matter volume were detected in brain regions which were previously associated with decreases in brain volume related to increasing age [47, 48] (Fig. 3).
DISCUSSION
We have investigated age related effects of the ApoE ɛ4 genotype on memory performance and the retrieval associated BOLD response. We applied the same neuropsychological assessment tools and fMRI paradigm to groups of young (26.75±5.31 years) and old (65.4±8.19 years) ɛ4 carriers, as well as matched control groups of non-carriers. There was no clear evidence that ɛ4 was associated with cognitive performance in the neuropsychological tests that were used in this study. However, with respect to our fMRI findings, there was some evidence for age-related differences in brain function associated with the ɛ4 allele. Higher age was associated with increased brain activation (BOLD signal) during successful retrieval of Face-Name pairs in ɛ4 carriers compared to non-carriers. This might reflect higher effort during associative retrieval. Furthermore, increased brain activation in older ɛ4 carriers was related to diminished memory performance. Principally this finding could indicate aberrant brain function associated with the ApoE genotype with increasing age.
Effect of age and APOE genotype on cognitive function
Consistent with well established theories of cognitive aging (e.g., [49]), we found a strong effect of age on all cognitive domains (attention, working memory, verbal learning, and memory), with younger participants scoring significantly higher than elderly participants. There was also an effect of age on intelligence as assessed with the MWT-B. The finding of lower intelligence scores in younger participants compared to elderly participants is probably due to the nature of the intelligence test, since the MWT-B captures crystallized intelligence which has been shown to increase with age [50]. ApoE genotype had no effect on cognitive performance. Also, there was no significant age x genotype interaction on cognitive performance. The finding of absent genotype related differences in cognitive performance in the younger group is consistent with a large meta-analysis by Ihle et al. [27]. This meta-analysis including more than 11,000 children, adolescents, and young adults showed no effect of ApoE ɛ4 on cognitive performance. Along these lines, a large study by Jorm et al. [19] investigating over 2,000 young adults with a standard test battery also did not find ApoE ɛ4 dependent group differences in cognitive performance. Contradictory to our findings, there are some studies that point to better cognitive performance in young ɛ4 carriers [11–13, 15].
Similar to young ɛ4 carriers, elderly ɛ4 carriers also showed no genotype related differences in any of the cognitive domains that were assessed in our study. Although some studies have found an effect of ApoE ɛ4 on cognition in elderly participants [6–8, 10], there are also contradictory research findings that did not show an effect of genotype on cognition [17, 51].
Inconsistencies in research findings on the effects of ApoE ɛ4 on cognitive performance in young and elderly cohorts could be due to differences in sample sizes as well as differences in the neuropsychological tests employed [27]. In general, studies with large sample sizes did not detect ApoE ɛ4 related benefits in cognitive performance ([17, 19]; see also Ihle et al. [27] for a review). Studies with small sample sizes tend to either over- or underestimate effect sizes, whereas investigations with large sample sizes estimate effect sizes more accurately [27].
We not only found no effect of ApoE ɛ4 on cognitive function, but there was also no significant age-related change in cognitive function associated with the ɛ4 allele (age x genotype interaction p > 0.05). This finding is in line with the largest study to date on the effects of age and genotype on cognitive function by Bunce et al. [17]. Using a battery of cognitive tests that covered a range of cognitive domains including processing speed, working memory, lexical decision making, as well as immediate and delayed recall, they investigated the effect of ApoE genotype and age on cognitive function in three different age groups (20–24, 40–44, and 60–64 years) each consisting of approximately 1,800 subjects. Unsurprisingly, they found a strong effect of age on cognitive function indicating that overall cognitive performance declined with increasing age. Similar to our results, there was no effect of ApoE genotype on cognitive function, nor importantly, was there a significant age x ApoE interaction. The cognitive tests employed by our study are quite similar to those used by Bunce et al. [17]. Moreover, like Bunce and colleagues, we have investigated cognitive performance with the same cognitive measures in young and old participants, which has not been done by previous studies. However, although we carefully designed our study in choosing appropriate assessment instruments that were applied to both young and old participants, a major shortcoming of our study is the sample size for detecting genotype related effects on cognitive performance. A sample size of 40 participants in the old group and 50 participants in the young group might just not be large enough to detect subtle differences in cognitive performance. However, the fact that Bunce et al. [17] did not find an effect of ApoE genotype on cognitive function, nor an age x genotype interaction in a very large study population could indicate that the small sample of our study was not the only explanation for the absence of ApoE genotype effects.
Effect of age and APOE genotype on brain activation
We found a main effect of both age and ApoE genotype on brain activation. Significant interactions between age and ApoE genotype were found in right and left insula, bilateral anterior cingulate gyrus, and bilateral superior frontal gyrus. We found that among older ɛ4 carriers, both the magnitude and the extent of brain activation during successful retrieval of Face-Name pairs were greater than in all other participants. Our finding of increased brain activation in older ɛ4 carriers during a memory task is in line with numerous other studies investigating the effect of old age and ApoE genotype on memory related brain activation.
Studies that have investigated task specific age effects on brain activation during memory tasks have found an increase in brain activity in parahippocampal gyrus with increasing age (e.g., [52]). This finding is in line with our results of increased activity during Face-Name recognition in right parahippocampal gyrus in old compared to young participants. Along with increased activity in parahippocampal gyrus, reduced down regulation of areas belonging to the default mode network during memory tasks is also a common finding in older subjects [53, 54]. Thus, our finding of reduced deactivation in a key region of the default mode network (posterior cingulate cortex) in old compared to young participants is in good agreement with previous studies on age related changes in brain activation.
Numerous studies with older subjects that have investigated the effect of ɛ4 on brain activation during memory tasks have found a genotype related increase in brain activation [4, 55]. More specifically, Bookheimer et al. [4] investigated genotype related alterations in the BOLD response during associative recall of word pairs. Similar to our findings, they found an increased BOLD response in ɛ4 carriers in left hippocampus, left prefrontal cortex, bilateral anterior cingulate, and bilateral superior temporal gyrus during associative retrieval. Since the type of task that was employed in the Bookheimer et al. study is quite similar to the task used in our study (associative retrieval), the great overlap in the results is especially compelling. An increase of brain activation during cognitively challenging tasks in old ɛ4 carriers has been interpreted as a possible compensatory mechanism counteracting incipient neuropathological processes associated with the ApoE 4 genotype [4, 5]. However, the results of our study do not support the hypothesis of increased brain activation as a compensatory mechanism.
The analysis of age x genotype interaction effects showed a decreased BOLD signal in young ɛ4 carriers compared to non-carriers during associative retrieval of Face-Name pairs in left and right insula, as well as in anterior cingulate cortex, while this pattern was reversed in elderly participants. All of the regions that showed significant age x genotype interaction effects are associated with cognitive processes. The insula is commonly activated during face recognition [56, 57] whereas the anterior cingulate cortex has been associated with cognitive effort, as activations in this brain area increases when the task becomes more challenging [58]. Finally, lesion studies have shown a role of the superior frontal gyrus for higher cognitive functions such as working memory [59].
A possible explanation why aging was associated with increased activation in ɛ4 carriers could be a compensatory mechanism that has been commonly suggested in old ɛ4 carriers [4, 55]. However, the nature of the correlations we have found does not support this assumption. Increased activations in old ɛ4 carriers were associated with diminished task performance, thus not pointing to a compensatory mechanism. This inverse correlation could be interpreted as a sign of underlying pathology in old ɛ4 carriers leading to aberrant neuronal activation and contributing to diminished task performance. There is some evidence that increasing levels of neuropathology could be reflected in higher levels of neuronal activation. A study by Mormino et al. [54] investigating the effect of Aβ deposition on brain activation supports this assumption. Using an episodic encoding paradigm, they found that increased activation among task-positive regions was related to Aβ burden in cognitively normal elderly controls. Further support for the assumption that increased activation might reflect underlying neuropathology is given by the results of Bookheimer et al. [4]. They studied a subgroup of subjects at baseline and after two years and found that the level of brain activation at baseline correlated with longitudinal memory decline. Taken together, the increases in brain activation that we have found in old ɛ4 carriers could be reflective of neuropathological processes such as Aβ deposition in brain areas that showed the most pronounced increases in activity.
A neuroimaging study by Filippini et al. [29] investigating age and ApoE genotype effects on memory found that aging was associated with decreased activity in ɛ4 carriers and increased activity in non-carriers. On the contrary, Nichols et al. [15] showed similar genotype-related effects on the BOLD response as our study using a combined encoding/recognition paradigm. In young age, the ApoE ɛ4 genotype was associated with decreased BOLD response during encoding and recognition of neutral scenes, whereas the pattern was reversed in old age. Taken together, there are contradictory findings regarding age-related genotype effects on brain activation during memory processes. These might be due to methodological differences (encoding versus recognition and type of material that had to be encoded). However, the majority of studies investigating the effect of APOE genotype on brain function in old participants with a similar age range (60–80 years) as in our study showed an increased BOLD response in ɛ4 carriers with respect to non-carriers (see [60] for a review). Thus, our findings of an age-related genotype effect are at least partly corroborated by other studies.
With respect to the hypothesis of antagonistic pleiotropy, our imaging results partly support the assumption that the ApoE ɛ4 allele mediates early lifespan benefits and later lifespan detriments. More specifically, brain regions that have been associated with retrieval effort and recognition memory performance showed increased activation with increasing age in ɛ4 carriers. Thus, it seems like older carriers of the ApoE4 genotype had to make more effort to retrieve the correct name for each face than any other group. On the other hand, the comparatively weak activations in young ɛ4 carriers point to less retrieval effort compared to old ɛ4 carriers. Importantly, brain activity was positively correlated with memory performance in young ɛ4 carriers which indicates that young ɛ4 carriers effectively recruited neuronal resources to enhance memory performance.
A major shortcoming of our study is the small sample size to detect age-related genotype effects on cognitive performance. The finding of no genotype-related differences in cognitive performance, nor age x genotype interactions in cognitive performance might have been due to the relatively small sample sizes and thus might not reflect reality. Furthermore, the young ɛ4 carriers in our study were relatively old (mean age 26.4 years) compared to young ɛ4 carriers investigated by other studies. Studies that have shown a cognitive advantage associated with the ApoE4 genotype included subjects that were aged below 23 years on average [11, 14]. Possibly, a cognitive advantage associated with the ApoE4 genotype is present already in childhood and adolescence [61] and the effect gets less pronounced with advancing age. Thus, including younger subjects (<20 years) into our study could have been favorable for investigating the hypothesis of antagonistic pleiotropy.
Moreover, a cross-over design as used in our study is not the most appropriate design to investigate the hypothesis of antagonistic pleiotropy. Longitudinal measures of cognitive performance and brain function across the lifespan could be more informative to elucidate the hypothesis of antagonistic pleiotropy. Future research on the interaction between age and ApoE genotype on cognition and brain function should include bigger sample sizes and employ longitudinal designs.
CONCLUSION
Our study used the same psychometric assessment instruments and fMRI paradigm in young and old ɛ4 carriers to investigate age-related effects of the ApoE4 genotype on brain function. With regard to cognitive function, there were no genotype-dependent differences in cognitive performance, thereby contradicting the assumption of age-dependent differential effects of the ApoE ɛ4 allele on cognitive performance. During the retrieval of Face-Name pairs, old ɛ4 carriers more strongly engaged brain areas that are associated with retrieval effort compared to any other group, pointing to increased effort to master the task. In contrast, young ɛ4 carriers showed the weakest activations in most of these areas. Unlike in all other groups, neuronal recruitment in young ɛ4 carriers was related to enhanced task performance which could point to a better neurofunctional capacity. Taken together, our results partly support the hypothesis of antagonistic pleiotropy, although this support remains ambiguous, since no genotype dependent differences in cognitive performance were found. More studies combining thorough neuropsychological assessment with neuroimaging that ideally include large sample sizes and follow a longitudinal design are needed to further our understanding of the effects of ApoE ɛ4 across the lifespan.
