Reduced hippocampal activation during episodic encoding in middle-aged individuals at genetic risk of Alzheimer's Disease: a cross-sectional study

Family history of dementia and the apolipoprotein E (APOE) genotype are significant risk factors for the development of Alzheimer's disease (AD). The APOE gene, found on chromosome 19, has three allele variants (i.e., ε2, ε3 and ε4) and six possible genotypes, with the ε3/3 genotype being most prevalent in the general population (but see [1, 2]). The presence of the ε4 allele significantly increases the risk and reduces the age of onset in people with the late-onset form of AD, the most common form of this disorder [1, 3, 4]. Previous studies have reported estimated cumulative lifetime risk percentages based on APOE genotype in individuals with a family-history of AD to be 30%, 46% and 61% for ε3/3 homozygotes, ε3/4 heterozygotes and ε4/4 homozygotes, respectively [5].

Impairment in the encoding and retrieval of episodic memories, presumably caused by neurodegeneration of the hippocampus and other medial temporal lobe (MTL) structures [6, 7], is one of the earliest symptoms in AD [8]. Recent evidence suggests that healthy ε4 carriers (mean age = 56 years) with a family history of AD show a greater longitudinal, age-related episodic memory decline than non-carriers [9–11]. Furthermore, some structural MRI studies have found reduced hippocampal volume in older cognitively normal ε4 carriers [12–16] (but see [17–19]). Taken together, these results suggest that there may be an age-related phenotype for ε4 carriers involving cognitive decline and brain atrophy prior to the onset of AD. An important question is how early these changes occur and whether they can be detected in vivo using existing functional imaging methods.

The use of neuropsychological measures and brain imaging to examine cognitively normal individuals with AD risk factors such as family history or APOE genotype could potentially yield valuable information about preclinical alterations in neural function that precede the symptomatic stages of AD. Functional MRI (fMRI) is a non-invasive brain imaging technique that has been used successfully to demonstrate hippocampal and MTL activation during several tasks thought to reflect aspects of episodic memory [20–25]. The few studies that have examined brain activation differences between ε4 carriers and non-carriers have reported inconsistent findings on measures of episodic memory.

One episodic encoding paradigm that has been used to demonstrate robust hippocampal activation is novelty detection, in which individuals discriminate between events that were previously learned from events that are novel [26–31] (see [32, 33] for reviews). Tulving and Kroll [34] have suggested that episodic encoding processes are more evident during the processing of novel information versus previously learned (or familiar) information. Therefore, novelty detection appears to be an important cognitive process involved in MTL-dependent episodic memory formation. Novel/familiar discrimination paradigms might therefore yield valuable information about functional changes in the MTL that precede overt changes in cognitive function in individuals at risk of AD.

The purpose of the present study was to use fMRI to test the hypothesis that cognitively normal ε3/4 heterozygotes with a family history of AD would display reduced activation in the hippocampus during a novel picture-encoding paradigm. We also used voxel-based morphometry (VBM) of T1 weighted images to determine whether any observed differences in fMRI activation between the groups were the result of differences in modulated GM volume. In addition, we compared hippocampal activation against a neuropsychological measure of episodic learning ability. Because prior research has shown that brain changes precede cognitive changes in this population, and because carriers of the ε4 allele may have altered hippocampal function, we hypothesized that there would be less systematic shared variance, disrupting the normal relationship between MTL activation and neuropsychological status in the ε3/4 heterozygotes as compared to ε3/3 homozygotes.

The present study examined the effects of APOE genotype on brain activation patterns in the MTL during an episodic encoding task in cognitively normal individuals with a family history of AD who were on average 15–20 years younger than the age at which AD symptoms typically develop. We found that ε3/4 heterozygotes displayed reduced fMRI activation compared to ε3/3 homozygotes in the right hippocampus and entorhinal cortex for the contrast of novel relative to familiar pictures. Importantly, there were no fMRI activation differences between the groups for the reverse contrast (i.e. familiar relative to novel), suggesting that the reduced activation found in the right hippocampus and MTL of ε3/4 heterozygotes was not caused by greater fMRI activation to the previously learned items in the ε3/4 heterozygotes, at least when measured relative to novel items. In subsequent whole-brain analyses, no brain regions were found to display greater activity in the ε3/4 heterozygotes than in the ε3/3 homozygotes for either contrast.

There were also no significant differences between the groups in age, education or memory function, and neuropsychological performance was within the normal range for both groups. This suggests that the reduced MTL activation to novel items (relative to familiar items) in ε3/4 heterozygotes was not caused by impaired cognitive function, and that the observed neurobiological changes in MTL function precede the onset of measurable decline in cognitive function.

We also found no evidence for differences in regional GM volume as measured by VBM, suggesting that the observed activation differences were not caused by reduced MTL GM volume in our cohort of middle-aged subjects. Previous volumetric studies have reported inconsistent findings regarding the effects of the ε4 allele on regional brain volume. Reiman et al. [56] reported nonsignificant trends towards smaller left and right hippocampal volumes in ε4/4 homozygotes (mean age = 58 years), and smaller hippocampal volumes were associated with reduced long-term memory ability. Den Heijer et al. [13] found that elderly ε4 carriers (mean age = 72 years) displayed significantly greater hippocampal and amygdalar atrophy and poorer memory ability relative to ε3/3 homozygotes. Moffat et al. [15] found that older ε4 carriers (mean age = 69 years) displayed a significantly greater rate of hippocampal volume loss over a 3-year follow-up period, although ε4 carriers were also found to display mild decline in memory ability over the same time frame. In a recent large-scale (N = 750) VBM study, Lemaitre et al. [14] found significantly reduced MTL (including hippocampus) volume in elderly (age range 63–75 years) ε4/4 homozygotes compared to both ε3/4 heterozygotes and non-carriers, whereas no significant differences were found between the ε3/4 heterozygotes and the non-carriers. Furthermore, these authors also found that the relative risk of cognitive impairment over a 4-year follow-up period was substantially greater in ε4/4 homozygotes relative to both ε3/4 heterozygotes and non-carriers. Other studies have also failed to demonstrate reduced regional brain volume in ε4 carriers [17–19]. The results of the present study are consistent with these findings.

Since the risk of memory impairment and AD is significantly greater in elderly ε4 carriers relative to non-carriers and younger ε4 carriers [57], cognitive status and age might interact with regional changes in brain volume. Therefore, it is quite possible that the MTL volume reductions found by previous studies in elderly populations were caused by the inclusion of APOE ε4 carriers who were more likely to be in the early stages of AD relative to non-carriers. This interpretation is supported by the findings of previous studies that reduced MTL volume in ε4 carriers was associated with poorer memory performance and increased risk for AD. In any case, the results of the present study failed to demonstrate significant differences in regional GM volume in middle-aged APOE ε3/4 heterozygotes, further supporting the notion that reduced fMRI activation in ε3/4 heterozygotes preceded overt changes in hippocampal volume. More studies are needed to determine the conditions under which the APOE ε4 allele results in reduced MTL volume.

In the present study, we also found that greater encoding ability on a neuropsychological measure of learning (i.e. RAVLT) was positively associated with fMRI activation in the left anterior MTL in the ε3/3 homozygotes but not in the ε3/4 heterozygotes. This signifies that among the noncarriers (in whom the MTL is presumably more intact) the strength of the MTL response was closely matched with better learning ability, whereas among the ε3/4 heterozygotes this relationship was disrupted. The amygdala and anterior hippocampus relationships found in ε3/3 homozygotes are consistent with a facilitative role for the amygdala in processing novel episodic information [58]. There was no significant difference in the magnitude of the correlations in the amygdala and hippocampus of the ε3/3 homozygotes, so conclusions regarding any differential contribution of these regions to encoding ability are limited. Future studies examining amygdala-hippocampal connectivity during encoding, as well as psychological factors (e.g. context, affective valence) that may modulate neural activity in these regions would be required to determine the role of the amygdala in encoding of novel visual information in people at risk of AD. Nevertheless, this finding does suggest that the left anterior MTL was recruited to a greater extent in ε3/3 homozygotes, whereas ε3/4 heterozygotes with better encodingability may have used encoding strategies that did not actively involve the anterior MTL. The fact that the correlation was found in the left hemisphere in ε3/3 homozygotes is consistent with previous studies that have reported primarily left MTL activation for encoding of verbal material (see [59] for review but see also [60]), suggesting that subjects with greater activation may have used a verbal encoding strategy. In contrast, previous studies in cognitively healthy young individuals (mean age = 30) have reported correlations between memory ability and right hippocampal signal change [60]. The exact reason for the divergent findings is not known. More studies examining correlations between memory ability and fMRI signal change in the left and right hippocampus during episodic encoding are needed to clarify these findings.

To date, six studies have examined brain activation differences between ε3/3 homozygotes and ε4 carriers using fMRI, but only two of these used an activation task that had an episodic encoding component. In a series of studies, Smith et al. [61–63] compared fMRI activation between a group of cognitively healthy subjects with a high risk of AD (i.e. ε4 carriers with a family history of AD) versus a low risk group (ε3/3 homozygotes with no family history of AD) during letter fluency and object naming tasks relative to a low-level, resting baseline condition (responding to a grayscale square that randomly changed in intensity). These authors found that ε4 carriers displayed increased activation in the left parietal region during a letter fluency task, and reduced activation in the inferotemporal cortex during the object-naming task. Longitudinal follow-up of a smaller subset (N = 25) of these participants four years later with the same object naming task revealed a greater longitudinal decline in fMRI activation in the inferotemporal cortex of the ε4 carriers compared to ε3/3 homozygotes. Since object naming and letter fluency are generally affected after episodic memory symptoms appear in AD [64], it may be that an episodic memory task would be more helpful in assessing the primary area of early pathology in AD – the MTL [6].

Bookheimer et al. [65] compared brain activation differences between cognitively normal ε4 carriers and ε3/3 homozygotes (age range: 47–82 years) using a paired-associate task as a probe of episodic memory. In their task, subjects were to encode seven unrelated word pairs over six learning trials, in which each learning trial was followed by 30-second periods of rest. The encoding phase was followed by six recall trials in which subjects heard the first word of each pair and were asked to recall the second word silently. The major contrast used to examine group differences was encoding + recall relative to the resting baseline. Whole-brain analysis revealed greater activation in the left prefrontal cortex, bilateral orbitofrontal, superior temporal, and inferior and superior parietal regions in ε4 carriers relative to the ε3/3 homozygotes for the contrast of encoding + recall relative to rest. In follow-up ROI analyses, these authors reported that ε4 carriers displayed greater signal change in the left MTL as well. In a follow-up study, Burrgren et al. [66] reported no differences between ε4 carriers and ε3/3 homozygotes during performance of a modified digit-span (forwards) working memory task relative to a baseline condition that was a single digit. They hypothesized that their results reflected a compensatory response in which ε4 carriers require additional cognitive effort to achieve comparable performance during episodic memory encoding tasks.

More recently, Bondi et al. [67] found that cognitively-healthy, older (mean age = 76) ε4 carriers displayed greater activation in the fusiform gyrus, parietal cortex and frontal gyrus compared to ε3/3 homozygotes using a paradigm in which subjects had to discriminate novel pictures from a single repeating picture. Follow-up ROI analysis revealed that ε3/3 homozygotes displayed greater activation in the left MTL compared to ε4 carriers, consistent with the findings of the present study. In contrast, the opposite pattern of results was found in the right MTL (i.e. ε4 carriers displayed greater activation in the right MTL). These authors also reported correlations between memory ability on a word-list learning task and right and left hippocampal activation during picture encoding (i.e. a positive correlation in ε3/3 homozygotes and a negative or zero correlation in carriers of the ε4 allele) that were similar to the correlations found in the present study. They suggested that their results were consistent with the compensatory response hypothesis described by Bookheimer et al. [65].

The exact reason for these contradictory findings is not knownp; however, there are several demographic and methodological differences between our study and the two previous fMRI studies that employed an episodic memory task. First, our subjects were on average 10 to 23 years younger than the participants in Bookheimer et al. [65] and Bondi et al [67], complicating comparisons between studies. Cabeza et al. [68] found that older individuals (mean age = 70 years) displayed reduced hippocampal activation during episodic retrieval, but greater activity in the prefrontal and parahippocampal cortices compared to young individuals (mean age = 23 years) [see also [69]]. Second, 60% of the subjects in Bookheimer et al. [65] had a family-history of AD, whereas 100% of our subjects had at least one biological parent with AD (Bondi et al. [67] did not report the percentage of subjects in his study with and without a family history of AD). Therefore, differences in these demographic variables might be responsible for the discrepant findings. Longitudinal studies similar to Smith et al. [63] that use episodic memory tasks and employ a 2 × 2 × 2 factorial design with family history of AD and APOE genotype as grouping variables are needed to help clarify these findings.

Third, it is possible that the divergent findings may have resulted from differences in the episodic memory task employed. In the present study, we employed a relatively straightforward novel/familiar discrimination paradigm as the probe of episodic encoding. In contrast, Bookheimer et al. [65] used a paired associate learning/recall task that was presumably more difficult than the paradigm used to define encoding in the present study [see [70]]. The paradigm used in the present study also differed from the encoding task used by Bondi et al. [67]. We presented familiar pictures in two separate trials before the fMRI scan. In contrast, first exposure to the repeating picture was during the fMRI task in the paradigm employed by Bondi et al. [67] and also by others e.g. [30], therefore some reduction in hippocampal signal during the presentation of the repeating novel stimulus may have occurred (see [50]).

Fourth, it is also possible that the compensatory response in ε4 carriers is more evident in elderly ε4 carriers rather than middle-aged ε4 carriers. Presumably, the negative effects of the APOE ε4 allele on hippocampal structure and function accelerate with increasing age. It is therefore possible that reduced hippocampal volume in elderly ε4 carriers [12–16, 19, 56, 71] (but see [18]) might lead to the recruitment of structures to compensate for accumulating neuropathology in the MTL. Smith et al. [62] suggested that increased activation in ε4 carriers might be caused by a disruption in upstream elements of a functional network, resulting in decreased input from regions that are affected early in AD such as the MTL.

Finally, an important methodological difference between our study and previous studies was the baseline condition used. In the present study, we contrasted novel relative to familiar pictures, whereas previous studies reporting increased activation in ε4 carriers used lower level baseline conditions. Since most fMRI designs employ a subtraction method (activation – baseline), the baseline task chosen has a major effect on brain activation patterns observed. Using rest as a baseline condition may augment unintended effects in the experimental task such as language or sensorimotor processes, and increase inter-subject variability in brain activation during the baseline condition since there is less control over the mental state of the subject (i.e. what is the subject thinking about during the resting condition?).

To address this issue, Stark and Squire [72] designed an elegant study to examine the effects of several different baseline conditions including resting fixation on MTL fMRI activation during presentation of novel and familiar pictures. These authors found that MTL activation to novel and familiar pictures was significantly greater when an active task was used as a baseline (in their case, an odd/even number comparison task) than when passive rest was used as the baseline. In fact, several MTL regions were found to display increased activation during rest relative to familiar or novel pictures. In contrast, when the active baseline task was used, both novel and familiar pictures were associated with significant bilateral activity throughout much of the MTL. These authors suggested that using rest as the baseline condition may reduce, eliminate or even reverse the sign of activity during a cognitive task. At the very least, the lack of control over the mental state of subjects during the resting state makes it not ideal to serve as a baseline condition for comparison to cognitive tasks.

More recent evidence suggests that the hippocampus is coactive during the resting state with several cortical structures including precuneus and posterior cingulate cortex [73], and this resting state activation is disrupted in individuals with MCI and AD [74, 75]. These new data make the previous findings of increased task-related activation relative to low-level baseline conditions in ε4 carriers more difficult to interpret, and argue against the use of resting baseline conditions in fMRI studies targeting these brain regions, especially in individuals with AD or at risk for AD.

Several positron emission tomography (PET) imaging studies have also reported reductions in resting state cerebral metabolic rate of glucose (CMRgl) in the MTL and cortical regions of AD patients and cognitively normal ε4 carriers between the ages of 20 and 65 years [76–82], suggesting different baseline neural activity in individuals at risk for AD. Reduced cerebral glucose utilization has also been shown to inhibit the induction of synaptic plasticity in the hippocampus of rats [83] and impair learning and memory on hippocampus-dependent tasks [84]. Prior research has demonstrated that CMRgl is coupled to regional cerebral blood flow (rCBF) [85, 86], and both CMRgl [87] and rCBF [88, 89] are coupled to the neural response. The current finding of reduced hippocampal BOLD signal during episodic encoding in ε3/4 heterozygotes is consistent with these studies, supporting the idea that reduced glucose metabolism would lead to reduced neuronal activation in the MTL during encoding. More research is needed to determine the relationship between reduced resting levels of CMRgl in ε4 carriers and task-related changes in the fMRI BOLD signal.

Reduced hippocampal activation in ε3/4 heterozygotes is consistent with several previous fMRI studies of memory encoding in patients with AD and MCI. For example, AD patients display reduced hippocampal activation compared to elderly controls during memory encoding [90, 91], whereas no differences were found in the motor cortex during a sensorimotor task [90]. Machulda et al. [90] also reported similar findings in MCI patients [see also [92]]. Johnson et al. [50] found that elderly normal subjects displayed a hippocampal adaptation response (i.e. reduction in signal intensity) to repeating unfamiliar faces that was not displayed by age-matched patients with MCI. Taken together, these prior results indicate that MTL activation is reduced in AD and MCI patients with objective memory impairment relative to elderly, cognitively normal controls.

In a recent study, Dickerson et al. [93] examined differences in the extent of brain activation (i.e. number of contiguous as well as non-contiguous voxels), but not the magnitude of activation, during episodic encoding in AD patients and elderly individuals with and without subjective memory complaints who did not meet the clinical criterion for MCI [94]. These authors manually traced bilateral ROIs of the hippocampus and entorhinal cortex from each participant's structural MRI. They found that individuals with memory complaints displayed increased extent of activation in each of the four ROIs relative to age-matched subjects with no memory complaints. AD patients were found to display significantly less extensive activation relative to both groups. Furthermore, APOE ε4 carriers were found to display a greater extent of activation relative to non-carriers when collapsed across the memory complaints factor.

Dickerson et al. [93] suggested that increased extent of MTL activation early in the course of prodromal AD was followed by a subsequent decrease in the extent of activation as the disease progresses. An important caveat to these findings is that the individuals with memory complaints had significantly greater education and total word recall on a word-list learning task relative to individuals without memory complaints. Therefore, increased extent of activation may have been related to the paradoxically better encoding ability in individuals with memory complaints. This interpretation is supported by the finding of the present study that total word recall on the RAVLT was associated with increased magnitude of activation during episodic encoding, at least in the ε3/3 homozygotes. Previous studies have also reported similar findings [60, 95, 96], suggesting that individuals with better episodic memory ability display greater fMRI activation in the MTL during episodic encoding tasks.

Reduced neural function in the MTL of asymptomatic ε4 carriers may also lead to a greater age-related decline in episodic memory over time. Several studies have reported that older ε4 carriers display significantly greater longitudinal age-related decline in episodic memory functions than non-carriers [9–11, 97, 98]. For example, Caselli et al. [97] reported that cognitively normal ε4 carriers (mean age = 60 years) displayed significantly greater longitudinal decline during a three-year follow-up period for total word recall on the RAVLT and delayed recall of complex figures. Our findings that hippocampal activation during a novel-picture encoding task is reduced in ε3/4 heterozygotes, and that memory encoding ability is positively associated with anterior MTL activation in ε3/3 homozygotes but not ε3/4 heterozygotes, are consistent with previous findings that MTL-dependent memory processes decline at a greater rate in individuals who are ε4 carriers and have a family history of AD. These results leave open the possibility that ε4 carriers recruit other brain regions and/or rely on other compensatory psychological processes (e.g. verbal rehearsal of visual information) during encoding to compensate for early pathological changes in MTL structure and function [65, 67].

Studies in transgenic mice expressing a human form of the APOE ε4 allele also support the negative effects of the ε4 allele on hippocampal function. For example, the presence of the APOE ε4 allele increases the production of β-amyloid (the main constituent of the senile plaques in AD) in cultured hippocampal neurons [99], reduces synaptic plasticity in the hippocampus [100], promotes β-amyloid induced blockade of plasticity in the hippocampus [101] and impairs hippocampus-dependent spatial memory [102, 103]. Our results are consistent with these findings as well.

This study has several limitations. We did not include a low-level baseline condition (for the reasons discussed above), limiting our ability to compare and contrast our results directly to previous fMRI studies of APOE genotype. Furthermore, the findings of fMRI activation differences associated with APOE genotype were restricted to ε3/3 homozygotes and ε3/4 heterozygotes with a parental history of AD. Therefore, our results only generalize to individuals with these APOE genotypes and a family history of AD. More studies are needed to determine the effects of other APOE genotypes (e.g. ε4/4, ε2/3) and family history of AD on brain activation during memory encoding. Third, we also did not include a post-scan recognition memory test to determine if the novel pictures were encoded into memory. While conclusions regarding APOE genotype differences in encoding related processes cannot be made directly from this study, they can be inferred on the basis that novel information is more likely to be encoded than familiar information (see also [31]).

Fourth, several of the subjects in this study had elevated total blood cholesterol and/or high blood pressure. Importantly, there were no group differences in the percentage of subjects with these conditions, and these percentages were also no greater than observed prevalence rates in the general US population. Fifth, several subjects in each group were current users of medications that influence brain function and may affect the hemodynamic BOLD response. More studies are needed to determine the effects of these drugs and/or medical conditions on the fMRI signal. Finally, the subjects in this study were predominantly Caucasian, highly educated and female. Additional studies are needed to determine whether our findings can be confirmed in subjects with other demographic characteristics. Despite these limitations, the results of the present study provide converging evidence for the idea that the MTL displays functional decline associated with the APOE ε4 allele in individuals with a family history of AD.

If compromised MTL function continues to be observed in healthy ε4 carriers, this group of subjects may represent a good study population for novel treatments designed to delay the onset of or to prevent AD. More studies are needed to clarify inconsistent findings and to determine the reliability, validity and clinical utility of hippocampal activation paradigms for the early detection of AD. Although the results of the present study indicate cross-sectional differences in MTL activation based on APOE genotype, future studies employing longitudinal designs will be required to determine whether or not differences in MTL activation in individuals at genetic risk for AD can be used to improve the detection of incipient AD.

This study found that APOE ε3/4 heterozygotes with a family-history of AD displayed reduced activation compared to ε3/3 homozygotes in the MTL during a novel picture-encoding task. Importantly, these changes occurred in the absence of cognitive differences between the groups. We also found that greater encoding ability on a neuropsychological measure of learning (i.e. RAVLT) was positively associated with fMRI activation in the left anterior MTL in the ε3/3 homozygotes but not in the ε3/4 heterozygotes. Together with previous studies reporting reduced glucose metabolism and AD-related neuropathology, this study provides convergent validity for the idea that the MTL exhibits functional decline associated with the APOE ε4 allele.