Atypical Localization and Dissociation between Glucose Uptake and Amyloid Deposition in Cognitively Normal APOE*E4 Homozygotic Elders Compared with Patients with Late-Onset Alzheimer’s Disease

Visual Abstract


Introduction
The disease process of Alzheimer's disease (AD) begins insidiously decades before enough brain damage occurs to require accessing medical care. Increasingly, there is consensus that prevention is critical and that interventions will be best when used early. The two most significant risk

Significance Statement
APOE‫ء‬E4 has the largest single effect size of any common variant for any human disease. APOE‫ء‬E4 homozygotes increase the risk of late-onset Alzheimer's disease (LOAD) 15-fold. Research indicates that interventions for AD must occur decades before onset of symptoms. However, the phenotypic antecedents and pathophysiology of LOAD remain limited in characterization. APOE‫ء‬E4 homozygotes offer a unique opportunity to characterize preclinical AD. Here, neuroimaging of cognitively normal, elderly APOE‫ء‬E4 homozygotes reveals decreased medial temporal metabolism and increased lenticular amyloid deposition in those at high risk for developing LOAD. In comparison to LOAD, an atypical pattern of change in metabolism and amyloid distribution as well as a dissociation between these two measures arose in homozygotes compared to noncarriers. factors for AD are age and the APOE‫ء‬E4 gene (also referred to as E4).
Having two APOE‫ء‬E4 alleles gives an odds ratio for AD of almost 15 (Farrer et al., 1997). APOE‫ء‬E4 is the common variant with the greatest known effect size for association to any human disease. This level of risk is like that for known Mendelian disease-causing mutations such as the breast cancer gene, BRACA (Genin et al., 2011). E4 homozygotes have earlier onset of cognitive decline by ϳ5-7 years (Blacker et al., 1997;Sando et al., 2008). Therefore, studies of carriers of APOE‫ء‬E4 have the potential to reveal insights into the early pathophysiology of AD. Recent advances in technology offer unique opportunities to study these individuals noninvasively.
Early in the development of AD, amorphous amyloid deposits occur throughout the brain initially in the inferior aspects of the frontal, temporal, and occipital lobes, later spreading diffusely throughout the neocortex (Braak and Braak, 1991). Amyloid deposition in the form of neuritic plaques containing the amyloid-␤ (A␤) protein develop with variable consistency, density, and distribution. Tangles and neuropil threads generally develop before plaques. The presence of APOE‫ء‬E4 is associated with amyloid deposition even in cognitively normal elders (Morris et al., 2010). The amount and distribution of neuritic plaques vary widely between individuals at similar disease stages (Braak and Braak, 1991). As many as one-third of E4 noncarriers diagnosed clinically with AD are characterized as amyloid negative on neuropathologic assessment (Monsell et al., 2015). In that study, approximately onehalf of those with a primary diagnosis of mild to moderate AD and low cerebral A␤ had extensive neurofibrillary degeneration.
The amount of fibrillar amyloid throughout the brain identifies both past and future decline in cognition (Doraiswamy et al., 2014;Donohue et al., 2017). PET studies of AD indicate that fibrillar amyloid and hypometabolism arise early in the posterior cingulate cortex/precuneus (Minoshima et al., 1994(Minoshima et al., , 1997Scheff et al., 2015). Although there is often overlap between cortical thinning, amyloid deposition, and hypometabolism, dissociations are not unusual (Murray et al., 2014).
Abnormal tau begins to deposit in the transentorrhinal cortex even in nondemented individuals (Braak and Braak, 1991;Bouras et al., 1994;Braak and Del Tredici, 2015). Subsequently, tau spreads to limbic allocortices including the hippocampal, cingulate, retrosplenial, and orbitofrontal cortices. With advanced AD, tau spreads into the neocortex. Tau staging correlates best with level of cognitive function and appears consistent across individuals with similar symptoms and degrees of clinical dysfunction (Braak and Braak, 1991;Ossenkoppele et al., 2016). APOE‫ء‬E4 status and tau pathology are not correlated (Morris et al., 2010). Although there is frequent overlap between amyloid, atrophy, and tau, dissociation between markers can occur; tau and atrophy tend to co-occur more frequently (Xia et al., 2017).
Biomarker development has advanced greatly the characterization of preclinical and early AD. As many as 30% of cognitively normal individuals Ͼ65 years of age have significant amyloid deposition (Murray et al., 2014). PET imaging has demonstrated amyloid deposition in both cognitively intact and impaired carriers of mutations in PSEN1, PSEN2, and APP, who typically have early onset of AD (Klunk et al., 2007;Remes et al., 2008;Fleisher et al., 2012;Shi et al., 2015;Rodriguez-Vieitez et al., 2016). Similarly, amyloid deposition has been found in those with and without evidence of cognitive decline who have Down syndrome (trisomy 21) with development of AD during middle age (Handen et al., 2012;Lao et al., 2016;Rafii et al., 2017).
Cognitively normal healthy elders homozygous for APOE‫ء‬E4 and therefore at very high risk of developing AD could provide data relevant to preclinical AD. Such information can guide future research and hint at pathophysiological mechanisms, particularly given recent evidence from transgenic mice that APOE‫ء‬E4 may provide a gainof-toxic-function independently of amyloid (Shi et al., 2017). To this end, PET scans of glucose uptake with [ 18 F]fluorodeoxyglucose (FDG), a proxy for regional cerebral metabolism, and of amyloid deposition from [ 18 F]florbetapir scans were downloaded from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database. Scans from E4 homozygotes were compared to scans from APOE‫ء‬E4 noncarriers.

Participants
Data used in the preparation of this article were obtained from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database (adni.loni.usc.edu; RRID:SCR_003007; accessed May 2017). The ADNI was launched in 2003 as a public-private partnership, led by Principal Investigator Michael W. Weiner, MD. The primary goal of ADNI has been to test whether serial magnetic resonance imaging (MRI), positron emission tomography (PET), other biological markers, and clinical and neuropsychological assessment can be combined to measure the progression of mild cognitive impairment (MCI) and early AD.
All E4/E4 subjects (n ϭ 8) included here had no memory complaints or functional impairments and were identified as normal cognitively based on neurologic and neuropsychological testing; Mini-Mental Status Exam (MMSE) score 24 -30; absence of clinically significant findings on screening MRI; Clinical Dementia Rating (CDR) 0; Geriatric Depression Scale (GDS) Յ6; and Hachinski score Ͻ4. This designation was based on the diagnosis closest to the time of the first imaging scan. Imaging results (e.g., cortical thinning, FDG, amyloid) were not used to define normality, so normal volunteers could have abnormalities in imaging scans subsequently. One homozygote did not get an amyloid scan. The comparison group was likewise characterized as normal but lacked E4 carriers. Demographic characteristics and related data of the homozygotes are shown in Table 1.
Additionally, late-onset AD (LOAD) subjects (n ϭ 8) from ADNI with very mild AD (CDR sum of boxes mean ϭ 3.4; range 2-4.5; SD ϭ 1) and similar ages to the E4/E4 group were selected as a patient group for comparison to the observations on asymptomatic homozygotes. As reviewed in the Discussion, there is already an extensive convergent literature on metabolic changes in AD.

Methods
All methods are described in detail at the ADNI database (adni.loni.usc.edu) following procedures approved by the institutional review boards. All scanning manufacturer's corrections were "On" (decay, randoms, scatter, etc.).
Briefly, subjects were injected intravenously with 185 MBq (5 mCi) [ 18 F]FDG for the glucose-uptake scan. The volunteer rested with eyes and ears open in a quiet, dim room for 20 min. Thirty minutes after injection, emission scans were obtained for 30 min. Scans were corrected for measured attenuation using low-dose CT scans.
Amyloid scans were acquired within 2 wk before or after the FDG scans. The subjects were injected intravenously with 370 MBq (10 mCi) [ 18 F]florbetapir. After a 50-min uptake period, an emission scan was obtained for 20 min. Scans were corrected for measured attenuation using low-dose CT scans.

Image analysis
All scans consisted of a 160 ϫ 160 ϫ 96 image grid with a voxel size of 1.5-mm cubic voxels. Scanner-specific filters were used to obtain an image resolution of ϳ8 mm full width at half-maximum (FWHM) regardless of scanner model. Images were inspected visually for potential artifacts. The FDG PET scans were normalized to a wholebrain uptake of 1000 counts. The amyloid scans were normalized based on the cerebellar cortex. All scans were anatomically coregistered to a template using Neurostat (Stereotactic Image Registration; v.7.1; S. Minoshima, University of Utah). Minima and maxima of the z-images were localized and quantitated with in-house software using an averaged, roving, 3-voxel cube. As routine for FDG PET studies at this resolution, a z-score Ն3.0 was defined as significant. The localization of structures was aided through use of the Talairach daemon (Lancaster et al., 2000).

Results
The demographics and related data of the APOE‫ء‬E4 homozygotes are presented in Table 1. The average age was 76 y, range 66 -85, SD 8. Three volunteers were women. The average of MMSE scores was 29. SUVR ranged from 0.93 to 1.53 (mean 1.02; SD 0.04); four were classified as amyloid positive by ADNI criteria. Hachinski scores ranged from 0 to 1 with average of 0.5. The average GDS rating was 1.5. The reference control group (i.e., non-E4 carriers; including E2 and E3 genotypes) was matched for age (75 y old, range 60-94, SD 6) and educational level (17 y). The reference group for FDG PET included 282 subjects; 144 males, 138 females; education range 8 -20 years, mean ϭ 16.6; 254 European Americans, 17 African Americans; 6 Asian Pacific Americans; and 1 Native American. The reference control group for amyloid imaging included 263 subjects; 133 males, 130 females; 235 European Americans, 17 African Americans, 6 Asian Pacific Americans; and 1 Native American. Fig. 1 (top) shows the peak regions of hypometabolism based on [ 18 F]FDG when comparing subjects who were APOE‫ء‬E4 homozygotes versus APOE‫ء‬E4 noncarriers. Both medial temporal cortices were symmetrically hypometabolic (left greater than right). The peak of hypometabolism mapped to left parahippocampal gyrus, Brodmann 37, at (-39, -37, -11) with z ϭ -3.0. Although below the significance threshold, the second most hypometabolic peak localized to the left hippocampus at (-33, -15, -14) with z ϭ -2.6. No other regions were hypometabolic including the putamen or ACC. No regions showed increased metabolism in the contrast between homozygotes and noncarriers. Fig. 1 (bottom) shows the results for the same metabolism contrast (LOAD vs. noncarrier). There is medial temporal hypometabolism similar in magnitude to that seen in E4/E4. As expected, large foci of hypometabolism localized to the PCC, ACC, and lateral parietal and inferior temporal cortices. Fig. 2 (top) shows the loci of peak deposition of amyloid based on [ 18 F]florbetapir in APOE‫ء‬E4 homozygotes when compared to the E4 noncarrier group. The peaks with highest magnitude localized to the lenticular nuclei, particularly the bilateral putamen. A broad region of amyloid deposition occurred in the ACC as well as medial and middle prefrontal gyri, inferior temporal, and occipital gyri. Other foci are listed in Table 2. No amyloid mapped to the PCC.  (LOAD vs. noncarrier). Most of the cortex shows extensive deposition of amyloid. The bilateral putamen show the greatest magnitude of deposition along with very heavy deposits in PCC/precuneus, ACC, as well as prefrontal, lateral parietal, and lateral temporal cortices. As reviewed below, these results converge with extensive literature but are presented for direct comparison with the homozygotes.

Novel findings in E4 homozygotes
This report provides several new findings. The literature relevant to these results based on metabolic and amyloid imaging as well as limitations is presented in the following sections.
(1) Medial temporal lobe (MTL) metabolism was reduced in E4 homozygotes, like the results found FDG uptake in E4 homozygotes contrasted with E4 noncarrier reference group. Transverse sections (top left inset) are taken from left to right at z ϭ -11, -18, and -25 mm below the intercommissural plane, respectively. The peak (red; z ϭ -3.0) is in the left parahippocampus. Note sequential inferiorly directed medial temporal lobe structures. Larger sections from lower left to upper right: z ϭ -2, y ϭ 12, y ϭ -60, x ϭ -10, respectively. Note minimal hypometabolism throughout including striatum. For illustrative purposes, the threshold was set to z ϭ -3. Note minimal or no hypoactivity in PCC and parietal cortex in E4 homozygotes. Bottom: FDG uptake in very mild AD contrasted with E4 noncarrier reference group. Transverse sections (top left inset) are taken from left to right at z ϭ -11, -18, and -25 mm below the intercommissural plane, respectively. Note sequential inferiorly directed medial temporal lobe structures. Larger sections from lower left to upper right: z ϭ -2, y ϭ 12, y ϭ -60, x ϭ -10, respectively. For illustrative purposes, the threshold was set to z ϭ -3. Note marked hypometabolism in bilateral lateral parietal cortex and PCC/precuneus (dashed circle). Left side of brain is on right side of image (radiologic convention).
here in LOAD, albeit the latter had more extensive lateral temporal changes. Nevertheless, reports of MTL metabolism in relation to E4 status have been mixed. Although cognitively intact E4 homozygotes are widely considered to show patterns of hypometabolism similar to those of patients with early LOAD, there are also numerous regions where they differ; e.g., E4 homozygotes show more extensive hypometabolism in the prefrontal cortex than those with early AD.
(2) E4 homozygotes did not show the anticipated parietal hypometabolism found by others, as reviewed below. Examining all hypometabolic foci, a PCC region occurs at z ϭ -2.0, far below the significance cutoff, which could be a type II error given the small sample size. However, similar sample sizes in MCI or AD show robust PCC hypometabolism as seen in LOAD here. The E4 homozygotes in this study were older than in other studies. The absence of parietal hypometabolism could reflect a degree of resilience For illustrative purposes, the threshold was set to z ϭ -3. Bottom panel: Amyloid deposition in very mild AD contrasted with E4 noncarriers. Transverse sections through MTL (top left inset) are from left to right at z ϭ -11, -18, -25, respectively. Larger sections from lower left to upper right: z ϭ -2, y ϭ 12, y ϭ -60, x ϭ -10, respectively. Note extensive amyloid deposits throughout the cortex, especially PCC/precuneus, and the striatum. Left side of brain is on right side of image (radiologic convention); color bar indicates z-scores.
against AD despite carrying the risk alleles (i.e., sampling bias). (3) E4 homozygotes showed the greatest degree of amyloid deposition in bilateral putamen. Numerous studies as noted below have shown that the PCC/ precuneus and prefrontal cortices have high amyloid loads in cognitively normal adults, early-onset AD, LOAD, and those carrying AD-relevant mutations. However, mutation carriers show highest amyloid deposition in striatum. (4) The E4 homozygotes showed no hypometabolism in the putamen, the principal site of amyloid deposition. This is consistent with studies of preclinical and clinical carriers of AD-relevant mutations. (5) Extensive accumulation of amyloid in E4 homozygotes localized to several additional regions including the ACC, medial frontal, middle frontal, inferior temporal, middle temporal, superior temporal, and occipital regions.

Hypometabolism in AD, E4, and autosomal dominant carriers
Metabolic changes in AD FDG PET remains among the best methods to evaluate functional brain decline in cognitively intact E4 carriers as well as in both preclinical and clinical early-onset AD (EOAD) and sporadic LOAD. EOAD is classified typically as onset before age 65 y. Many studies of EOAD do not systematically screen to exclude carriers of known mutations. Across studies, the greatest consistency and degree of hypometabolism localizes to the PCC/precuneus (see below). It was unanticipated that the MTL was not the principal region of hypometabolism given its role in memory, site of tau deposition, and atrophy in normal aging and AD. However, the large partial volume effects and interslice distances in early PET scanners likely reduced sensitivity to detection. Kim et al. (2005) studied 74 EOAD (Ͻ65 y; CDR 0.5) using FDG PET. They reported more severe hypometabolism in parietal, frontal, and subcortical (basal ganglia and thalamus) areas compared with LOAD, interpreted as a more rapid deteriorating course.  reported EOAD (Ͻ65 y) had more severe deficits in working memory and attention. EOAD had more severe PCC and bilateral temporoparietal hypometabolism than LOAD; LOAD did not show relative decreases in metabolism compared to EOAD. EOAD had lower metabolism after atrophy correction than LOAD in bilateral precuneus and right angular gyrus; no regions in LOAD were less metabolic than in EOAD. There was a positive correlation between age of onset and hypometabolism in the precuneus, lateral parietal, and occipital regions.
The initial region of hypometabolism in early LOAD localized typically to the PCC/precuneus with subsequent extension into biparietal and inferior temporal regions (Minoshima et al., 1994(Minoshima et al., , 1997. Despite the noteworthy absence of MTL hypometabolism in early studies, most subsequent studies using FDG PET on higher resolution scanners or with interleaf acquisition demonstrated hypometabolism in the MTL along the continuum from normal to MCI to AD; this hypometabolism predicted subsequent cognitive decline Lehmann et al., 2014).

Metabolic changes with carriage of E4 allele
Several studies examined alterations in metabolism in presymptomatic subjects and LOAD relative to E4 allele status. Small et al. (1995) compared FDG PET scans from cognitively intact subjects (age ϳ55 y) with mild memory complaints both with (n ϭ 12) and without (n ϭ 19) the E4 allele. They found that E4 carriers had marked hypometabolism in both parietal lobes. Reiman et al. (2001) compared metabolic decline over a 2-y interval in cognitively normal APOE carriers versus noncarriers (50 -63 y). They demonstrated, despite the absence of cognitive change during follow-up, significantly less metabolism in carriers localized to lateral temporal cortex, PCC, lateral prefrontal cortex, basal forebrain, parahippocampal/lingual gyri, and thalamus. Langbaum et al. (2009) compared FDG PET scans from elder normal control subjects and patients with amnestic MCI and AD. Despite higher resolution (8 mm FWHM) scans, medial temporal hypometabolism was detected bilaterally only in the AD patients compared with control subjects. There was hypometabolism in bilateral precuneus and left lateral parietal lobe in normal subjects with E4 allele compared with those without E4 (n ϭ 21 and 61 in each group, respectively).
In a landmark publication, Reiman et al. (1996) contrasted (1) FDG PET of 11 cognitively intact E4 homozygotes with 22 healthy controls without the E4 allele (mean group age: 55 and 56 y, respectively); and (2) FDG PET of 37 probable AD patients with 22 healthy controls (each group of average 64 y old). The probable AD patients relative to matched controls showed three broad regions of hypometabolism: PCC/precuneus as well as bilateral parietal and bilateral inferior temporal lobes. In addition, there were several small foci of hypometabolism in the prefrontal, occipital, and lateral temporal regions. Medial temporal hypometabolism was not reported in the AD patients. In the contrast involving E4 homozygotes versus those without E4 allele, several hypometabolic foci converged with the AD hypometabolic regions, including the PCC/precuneus, parietal, and inferior temporal cortices. Of note, the E4 homozygotes also had broad regions of significant hypometabolism in the prefrontal cortices, as well as smaller foci distributed throughout the cortex and cerebellum, not seen in the AD group. In a subsequent study,  used FDG PET to examine healthy middle-aged adults (20 -39 y). Twelve E4 heterozygotes compared with 15 E4 noncarriers showed some overlap with hypometabolic regions seen in AD, particularly in PCC/precuneus as well as in parietal and inferior temporal regions. Medial temporal changes were not reported. The heterozygotes showed several regions of hypometabolism in prefrontal cortex beyond those seen in AD patients. So, hypometabolism can arise both within and outside AD-affected regions, even in much younger healthy subjects.
De  noted in a longitudinal study that those having cognitive decline from normal status at baseline showed lateral temporal lobe hypometabolism dependent on E4 status; the entorhinal cortex did not show this effect. Mosconi et al. (2004) compared using FDG PET AD patients with and without the E4 allele. They noted an age-by-genotype interaction in the anterior cingulate and medial frontal cortices. The results were interpreted as indicating an age-dependent aggravation in metabolic decline in AD related to the E4 allele status.
To investigate the effects of ethnicity on the relationship of APOE status to glucose metabolism, cognitively intact, middle-aged (mean ϳ55 y) Latino Mexican-Americans were studied with FDG PET . The left hippocampus had decreased metabolism in E4 carriers versus noncarriers (n ϭ 11 and 16 per group, respectively). There was some convergence of medial and lateral parietal hypometabolism in the contrast between E4 carriers versus noncarriers with the hypometabolism seen in LOAD. However, the E4 carriers showed less hypometabolism in the traditional precuneus/PCC regions with more extensive involvement of the ACC.
In contrast, Protas et al. (2013) examined with defined regions of interest a large series of healthy subjects (mean age ϳ56 y) who carried either 0, 1, or 2 E4 alleles (n ϭ 76, 42, and 31, respectively). Those authors noted a highly significant difference across groups in PCC metabolism; no difference in hippocampal metabolism or volume was found. Differences in subject groups, technologies, or analysis methods could account for the divergence in results from those reported here. First, the much smaller sample of homozygotes in the present study compared to that of Protas et al. (2013) could explain the failure to detect PCC hypometabolism; even the changes in the MTL reported here were not large, although there was some subthreshold hypometabolism in the contralateral hippocampus. The left hippocampal hypometabolism in normal E4 carriers reported here does converge with the observation of . Second, the earlier-generation scanner used by Protas et al. (2013) had an interslice distance of 3.375 mm; this slice thickness without 3D acquisition or volume reconstruction would decrease recovery of the thin strip of MTL extending inferiorly from posterior to anterior as seen in Fig. 1 in the present report. Third, the mean age of the homozygotes here was much greater than that in Protas et al. (2013). Fourth, the present analysis used a voxel-wise approach, whereas Protas et al. (2013) used defined regions of interest. Finally, Protas et al. (2013) did not report amyloid deposition or longitudinal outcomes that could impact metabolism across subjects and studies.

Metabolic changes in autosomal dominant mutation carriers
Several studies have examined with FDG PET the metabolic changes in mutation carriers including APP dosage effects (trisomy 21/Down syndrome; APP duplication; exon deletion variant) as well as mutations in PSEN1, PSEN2, and APP. Villemagne et al. (2009) found that the classic PCC and biparietal pattern of hypometabolism seen in sporadic AD was not as evident in mutation carriers (n ϭ 8; CDR 0 -3.0). No distinct pattern of FDG hypometabolism characterized the various PSEN1 versus APP mutation carriers; three individuals showed an almost normal pattern of uptake including the striatum, the principal locus of amyloid deposition. There was no relation between the hypometabolic pattern in the cortex or striatum and disease severity, type of mutation, or cognitive status. Schöll et al. (2012) reported on two APParc mutation carriers showing AD-typical patterns of hypometabolism.  reported in Down syndrome without dementia (mean age ϳ36 y) mostly hypometabolism and decreased gray matter atrophy in the ACC not PCC. Therefore, considerable variability in metabolism is seen in mutation carriers.

Amyloid deposits in EOAD, LOAD, E4 carriers, and autosomal mutation carriers Amyloid deposits in AD
Neuropathologically, nondemented control subjects do not show amyloid deposition or neurofibrillary changes (Braak and Braak, 1991). In contrast, patients with AD show not only extensive fibrillar amyloid, but also an abundance of amyloid deposits that are diffuse without neurofibrillary changes or glial reaction (Braak and Braak, 1991;Brilliant et al., 1997). The striatum develops senile neuritic plaques and neurofibrillary tangles later in the progression of AD (Braak stage V-VI; Braak and Braak, 1991).
Amyloid imaging labels fibrillar amyloid (Ikomonovic et al., 2008;Clark et al., 2011;Curtis et al., 2015). Other forms of amyloid (diffuse, fleecy, deposits in AD cerebellum, amyloid oligomers) are not detected. In AD, the greatest fibrillar amyloid deposition localizes to the parietal cortex (both precuneus/PCC and lateral parietal); parietal fibrillar amyloid deposition often shows a correlation with declining parietal metabolism and synaptic markers Li et al., 2008;Scheff et al., 2015). Subsequently, fibrillar amyloid deposits spread throughout the neocortex including ACC, lateral prefrontal cortex, striatum, and the temporal lobe. Differences in amyloid deposition between EOAD and LOAD have depended on the region of interest.  studied 18 EOAD and compared them to 16 LOAD. They found no effect on amyloid deposition but decreasing metabolism in posterior brain regions depending on age of onset. Youn et al. (2017) studied nine EOAD, 11 LOAD, and 8 normal controls. EOAD patients were screened negative for the common AD-associated mutations. EOAD patients showed greater amyloid deposition only in the thalamus and basal ganglia compared to those with LOAD.

Amyloid deposits and APOE‫ء‬E4
The major locus of amyloid deposition found here in E4 homozygotes localized to the putamen as well as several other regions (ACC, medial frontal, middle frontal, inferior temporal, middle temporal, superior temporal, and occipital). Unlike what is seen in typical LOAD, neither hypometabolism nor amyloid deposition localized to the PCC in these E4 homozygotes. It is noteworthy that the localization of fibrillar amyloid in healthy elders and in AD does not reflect the patterns of cortical thinning, hypometabolism, or clinical phenotypes; these appear to converge with the localization of tau Xia et al., 2017;Lockhart et al., 2017;Pontecorvo et al., 2017).
The pattern of amyloid deposition noted here is not inconsistent with that reported by . They studied a younger group (mean age ϳ63 y) of 8 homozygotes, 12 heterozygotes, and 12 noncarriers of E4. They reported a significant association with "ADaffected mean cortical, frontal, temporal, posterior cingulate-precuneus, parietal, and basal ganglia ROIs." Based on an estimation of their basal ganglia data, reanalysis specifically comparing E4 homozygotes to noncarriers, analogous to that done here, produced p Ͻ 0.01 (twotailed, between-sample t test; df ϭ 18; heteroscedastic correction) indicating convergence in findings. Given the older age of the current homozygotes, amyloid deposition here appeared grossly greater overall than that reported by .

Amyloid deposits in autosomal dominant mutations
Genetic familial effects beyond those associated with the E4 allele have been reported with APP gene dose (trisomy 21/Down syndrome; APP duplication) as well as mutations in PSEN1, PSEN2, and APP. Putaminal amyloid deposits, which can correlate with age early in the course, were seen in Down syndrome without dementia (Handen et al., 2012;Lao et al., 2016). Braak stage (based on tau scans), amyloid deposition, and cognitive decline were correlated with age in amyloid-positive, nondemented subjects with Down syndrome; glucose hypometabolism and tau did not localize to the striatum, although regions with hypometabolism overlapped areas with tau deposition (Rafii et al., 2017). High levels of amyloid in the putamen were reported also in various cases of preclinical and clinical AD related to PSEN1, PSEN2, and APP mutations (Klunk et al., 2007;Remes et al., 2008;Koivunen et al., 2008;Villemagne et al., 2009;Fleisher et al., 2012;Shi et al., 2015;Rodriguez-Vieitez et al., 2016). Like observations here about E4 homozygotes, AD mutation carriers with putaminal amyloid deposits did not show hypometabolism in the putamen (Rodriguez-Vieitez et al., 2016).
The clinical significance of striatal amyloid deposition is unclear. The co-occurrence of Parkinson's in AD is well known. AD variants with mutations such as APP can also have various motor manifestations such as hyperactive reflexes, extremity weakness, and spastic paraparesis. An increased incidence of Parkinson's in subjects with Down syndrome with or without dementia remains controversial (Hestnes et al., 1997).

Limitations
The major limitation of this study is the small number of homozygotes available in the ADNI database, not surprising given their low frequency. The ALZGENE database indicates an E4/E4 prevalence of ϳ2% in controls and 15% in AD cases (Alzforum, 2017). Furthermore, the frequency of E4/E4 decreases with aging, most likely owing to differential survival related to cardiovascular disease. In an Australian community sample of those 70 y or older surviving a follow-up period of 3 y, the E4 allele frequency was 13% (Henderson et al., 1995). Of these 638 subjects, only 10 E4 homozygotes were recruited, and none were older than 90 y. Larger samples will require pooling across multiple databases. The limited sample described here risks type II errors; thus, the present findings must be considered preliminary, particularly regarding the absence of changes in various regions. Nevertheless, the present findings are noteworthy given the differences from LOAD and similarities to AD-related mutations and Down syndrome.
Another issue concerns whether the homozygotes represent typical preclinical AD or whether the sample is biased (e.g., survival bias). Despite normal cognitive status, approximately half were already positive for amyloid by the usual criteria (ADNI's templated SUVR). Their mean age of ϳ76 y is near the peak age (60 -75 y) for the highest risk of conversion from normal cognition to MCI or AD (Bonham et al., 2016). The predicted age for onset of LOAD in E4 homozygotes has been estimated at 5-7 y earlier than for noncarriers (Blacker et al., 1997;Sando et al., 2008).
No atrophy correction was made for the measurements of glucose metabolism in this study. The between-group comparison minimized age-related atrophy across groups by matching on age. The rationale for not performing atrophy correction considered that depending on the algorithm, amplification of noise compounded by the small sample size can introduce additional concerns. The question of PCC and medial temporal atrophy in cognitively normal E4 carriers is controversial given mixed results from different studies. For example, Li et al. (2016) found greater atrophy of the left hippocampus in E4 carrier versus noncarrier groups without dementia (i.e., control and MCI combined) from ADNI (n ϭ 212 and 242 per group, respectively). Haller et al. (2017) studied adults (n ϭ 282) dwelling in the community who were cognitively intact and followed for 18 mo. They found a significant effect of the E4 allele on PCC atrophy but only for those who decline on follow-up; no changes were found in the MTL. The MTL in elderly with intact cognition is known to have tau tangles typically associated with neurodegeneration, whether amyloid positive or negative, and may explain the observed hypometabolism found in this study (Bouras et al., 1994;Braak and Del Tredici, 2015).

Summary
In conclusion, cognitively normal, elderly E4/E4 show an atypical pattern of both hypometabolism and amyloid deposition compared with that known to occur in LOAD. Metabolism is dissociated in localization from amyloid deposition. The region of greatest amyloid deposition localizes to the putamen, as is seen in Down syndrome and early-onset AD arising from mutations. The mechanisms for protein deposition remain unclear. The difference in biomarker phenotypes between E4 homozygotes and those with LOAD suggest divergent pathophysiological processes, unshared environmental effects, or residuals of resilience to AD in a high-risk group.