apoe predicts amyloid-beta but not tau alzheimer pathology in cognitively normal aging

APOE Predicts Amyloid-Beta butNot Tau Alzheimer Pathology in

Cognitively Normal AgingJohn C. Morris, MD,1,2,3,4,5 Catherine M. Roe, PhD,1,2

Chengjie Xiong, PhD,1,6 Anne M. Fagan, PhD,1,2

Alison M. Goate, PhD,1,2,7,8,9 David M. Holtzman, MD,1,2,7

and Mark A. Mintun, MD1,10

Objective: To examine interactions of apolipoprotein E (APOE) genotype with age and with in vivo measures ofpreclinical Alzheimer disease (AD) in cognitively normal aging.Methods: Two hundred forty-one cognitively normal individuals, aged 45–88 years, had cerebral amyloid imagingstudies with Pittsburgh Compound-B (PIB). Of the 241 individuals, 168 (70%) also had cerebrospinal fluid (CSF)assays of amyloid-beta42 (A�42), tau, and phosphorylated tau (ptau181). All individuals were genotyped for APOE.Results: The frequency of individuals with elevated mean cortical binding potential (MCBP) for PIB rose in anage-dependent manner from 0% at ages 45–49 years to 30.3% at 80–88 years. Reduced levels of CSF A�42appeared to begin earlier (18.2% of those aged 45–49 years) and increase with age in higher frequencies (50%at age 80–88 years) than elevations of MCBP. There was a gene dose effect for the APOE4 genotype, withgreater MCBP increases and greater reductions in CSF A�42 with increased numbers of APOE4 alleles. Individualswith an APOE2 allele had no increase in MCBP with age and had higher CSF A�42 levels than individuals withoutan APOE2 allele. There was no APOE4 or APOE2 effect on CSF tau or ptau181.Interpretation: Increasing cerebral A� deposition with age is the pathobiological phenotype of APOE4. Thebiomarker sequence that detects A� deposition may first be lowered CSF A�42, followed by elevated MCBP forPIB. A substantial proportion of cognitively normal individuals have presumptive preclinical AD.

ANN NEUROL 2010;67:122–131

The concept of preclinical Alzheimer disease (AD) pos-tulates that AD lesions accumulate in the brain for

years prior to the appearance of cognitive deficits orsymptoms of dementia.1 This concept developed fromobservations that densities of senile plaques (SPs) andneurofibrillary tangles (NFTs), sufficient to meet his-topathologic criteria for AD, frequently are present inbrains of individuals whose cognition at death was nor-mal2–9 or stable.10,11 Preclinical AD assumes that ADneuropathology in cognitively normal individuals resultsin progressive neuronal deterioration and eventually willculminate in the clinical syndrome of dementia of the

Alzheimer type (DAT), although the time to DAT maybe influenced by variables such as an individual’s degreeof cognitive and brain “reserve.”12–14 To date, however, itis not known whether DAT is inevitable in preclinicalAD. It remains possible that certain individuals with pre-sumptive preclinical AD may never develop DAT, nomatter how long they live.

Imaging and molecular biomarkers for AD now iden-tify in vivo correlates of neuropathological AD15–20 andcan be used as surrogate markers of preclinical AD. In par-ticular, elevations of cerebrospinal fluid (CSF) concentra-tions of tau or phosphorylated tau (ptau) appear to reflect

Published online in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/ana.21843

Received Apr 23, 2009, and in revised form Jul 1. Accepted for publication Jul 1, 2009.

Address correspondence to Dr Morris, Alzheimer Disease Research Center, 4488 Forest Park Avenue, Suite 160, St. Louis, MO 63108.E-mail: [email protected]

From the 1Alzheimer Disease Research Center, 2Department of Neurology, 3Department of Pathology and Immunology, 4Department of PhysicalTherapy, 5Department of Occupational Therapy, 6Division of Biostatistics, 7Department of Developmental Biology, 8Department of Psychiatry,

9Department of Genetics, and 10Department of Radiology, Washington University School of Medicine, St. Louis, MO.

Additional Supporting Information may be found in the online version of this article.

ORIGINAL ARTICLE

122 © 2010 American Neurological Association

NFTs in the brain, and reduced levels of CSF amyloid-beta(A�42) appear to reflect cerebral A� deposition in the formof plaques.21,22 With positron emission tomography (PET),the [11C]benzothiazole radiotracer, Pittsburgh Compound-B(PIB), is retained in greater amounts in individuals withDAT as compared with cognitively normal persons inbrain areas known to have high amounts of fibrillar A�

plaques in AD.23 In postmortem studies, PIB identifiesfibrillar A� typical of AD brain24 and in human frontalcortical biopsy tissue PIB uptake corresponds to paren-chymal A� aggregration.25 The correspondence of CSFA�42 levels with PIB binding values in the same individ-uals suggests that they each are measuring aspects of ce-rebral A� burden.26,27

These biomarkers for AD may identify nondementedindividuals at risk for developing DAT. In persons withmild cognitive impairment (MCI),28 progression to a clin-ical diagnosis of DAT is predicted by CSF A�42, tau levels,and PIB findings.29–31 Limited studies in cognitively nor-mal older adults show that reduced levels of CSF A�42,often combined with elevated levels of tau, predict cogni-tive decline and dementia,27,32,33 but the value of biomar-kers in characterizing preclinical AD antecedent to anymeasurable cognitive deficit (including MCI) is less stud-ied. Moreover, the relationships of these biomarkers inpreclinical AD to established risk factors for AD are un-known.

Increasing age and genetic background are the stron-gest known risk factors for AD.34 The ε4 allele of apoli-poprotein E (APOE), the major genetic susceptibility factorfor late-onset AD, confers dramatically increased risk in agene dose-dependent manner for the development of DATwith an earlier age of onset.35–38 Other isoforms of APOEare considered to be neutral (APOE3) or protective(APOE2) for AD risk.39,40 The increased risk of APOE4 forAD may be mediated by impaired regulation of cerebralA� metabolism.41–43 There is an isoform-dependent pro-pensity (E4� E3�E2) for A� to be deposited as cerebralamyloid plaques in experimental animals,44,45 and in hu-mans APOE4 carriers have increased cerebral amyloid dep-osition in comparison with noncarriers.46–48 The role ofAPOE4 in promoting clinical disease appears to be directlyrelated to its effect on AD pathology, because its associa-tion with clinically diagnosed DAT is nonsignificant aftercontrolling for the densities of SPs and NFTs in autopsiedindividuals.49 Consistent with this premise, an APOE ge-notype effect on A� plaque load recently has been demon-strated in amyloid imaging studies in individuals withmoderate DAT50 and in nondemented individuals.51,52

However, because of small sample sizes, these studies didnot address the influence of age on the APOE4 effect, nor

did they examine the role of APOE2 on A� deposition orutilize CSF biomarkers to determine whether APOE influ-ences tau or A� metabolism.

To more thoroughly examine the interactions ofAPOE with indicators of AD pathology, we examined therelationship of APOE genotype with age in 241 cogni-tively normal individuals, aged 45 to 88 years, with meancortical binding potential for PIB and, in 168 of the 241individuals, with CSF levels of A�42, tau, and ptau. Wealso wished to determine whether the main effect ofAPOE is on A� or tau. We hypothesized that an effect ofage and APOE genotype, the known risk factors for AD,on PIB and CSF biomarkers provides biological evidencefor the relevance of these biomarkers in characterizingpreclinical AD.

Materials and MethodsParticipantsParticipants were community-living volunteers without cognitiveimpairment who enrolled in longitudinal studies of memory andaging at Washington University’s Alzheimer’s Disease ResearchCenter (ADRC). These studies were inaugurated in 1979 in in-dividuals aged �60 years to evaluate the natural history of mildDAT in comparison with nondemented aging; details aboutthe recruitment and assessment methods have been published.53

Individuals with clinically meaningful disorders (eg, disabling ce-rebral infarcts, renal failure requiring dialysis) that could inter-fere with longitudinal follow-up were excluded. A complemen-tary study, the Adult Children Study (ACS), began in 2005 byenrolling community-living individuals without cognitive im-pairment, aged 45–74 years, into 2 groups, 1 in which the bi-ological parent of the participant developed DAT prior to age80 years, the second in which neither biological parent of theparticipant had DAT.54 All participants were assessed in a uni-form manner with identical instruments and procedures with 2exceptions: 1) the ADRC neuropsychological battery was mod-ified for ACS participants to allow for their younger age; and 2)ACS participants were evaluated every 3 years until age 65 years,when they were evaluated annually (identical to ADRC partici-pants).

Inclusion criteria for this study were: 1) completing PETPIB imaging between April 2004 and October 2008; 2) normalcognition at the clinical and cognitive assessment closest in timeto the PET PIB scan; and 3) age 45 years or older at that clin-ical and cognitive assessment. Individuals from families with aknown deterministic mutation for AD were excluded.

EvaluationAll procedures were approved by Washington University’s Hu-man Research Protection Office. Written informed consent wasobtained from all participants and their collateral sources (infor-mants).

At entry and at each follow-up, experienced cliniciansconducted semistructured interviews with the participant and

Morris et al: Biomarkers in Preclinical AD

January, 2010 123

separately with their informant to determine possible decline inthe participant’s cognitive ability to function in everyday activ-ities.53,55 Included in the protocol were demographic informa-tion, a health history, a depression inventory, an aphasia battery,and medication history. The Mini Mental State Examination(MMSE)56 was also obtained. Following a neurological evalua-tion, the clinician determined whether dementia was present orabsent based on the principle of intraindividual decline in theperformance of accustomed activities because of cognitive loss.55

The absence of dementia corresponds to a Clinical DementiaRating (CDR) of 0; very mild dementia is indicated by CDR0.5, and mild, moderate, and severe dementia by CDR 1, 2, and3.57 Diagnosis of dementia etiology (eg, DAT) was made in ac-cordance with standard criteria and methods.58 The CDR de-termination and dementia diagnosis were made without refer-ence to the participant’s performance on the neuropsychologicaltest battery.

Within a few weeks of the clinical assessment, a 1.5-hourneuropsychological test battery55 was administered; psychome-tricians were not informed of the results of the clinical evalua-tion or of any previous psychometric assessments. The indepen-dence of the clinical and the psychometric evaluations alloweddata obtained with each method to be compared without theconfounding that occurs when psychometric performance isused both to classify individuals and to measure outcomes. Theclinical assessment procedures permitted the detection of mini-mal cognitive decline, even when the deficits were too mild tomeet criteria for MCI,55 and resulted in the designation of CDR0.5. The CDR 0 group thus was free of even minimal cognitiveimpairment.59

CSF Collection and AnalysisThe CSF (20–30ml) was collected free from any blood contam-ination in polypropylene tubes at 8:00 AM after overnight fast-ing as previously described.26 Samples were gently inverted toavoid gradient effects, briefly centrifuged at low speed to pelletany cellular elements, and aliquoted (500�l) into polypropylenetubes prior to freezing at �84°C. The samples were analyzed fortotal tau, ptau181, and A�42 by commercial enzyme-linked im-munosorbent assay (ELISA) (Innotest, Innogenetics, Ghent, Bel-gium); CSF A�40 was assayed by ELISA as previously de-scribed.60 For all biomarker measures, samples werecontinuously kept on ice, and assays were performed on samplealiquots after a single thaw following initial freezing.

ImagingDetailed information on the PET PIB imaging and analysis pro-cedures have been reported.61 In brief, brain PET imaging wasconducted using a Siemens 961 HR ECAT PET scanner (CTI,Knoxville, KY) or a Siemens 962 HR� ECAT PET scanner ina darkened, quiet room. A thermoplastic mask was placed tominimize head motion, and participants kept their eyes closedduring the scan. Radiochemical synthesis of [11C]PIB was car-ried out according to published literature.62 After a transmissionscan to measure attenuation, approximately 12mCi of [11C]PIBwas administered intravenously simultaneously with initiation of

a 60-minute dynamic PET scan in 3-dimensional mode (septaretracted; 24 � 5-second frames; 9 � 20-second frames; 10 �

1-minute frame). The measured attenuation factors, scatter cor-rection, and a ramp filter were employed to reconstruct the dy-namic PET images. In addition to PET imaging, all participantsunderwent anatomic magnetic resonance imaging (MRI) usingMPRAGE T1-weighted volume (1mm � 1mm � 1.25mm) ac-quisitions.

Image ProcessingPIB image analysis was performed for specific regions of interest(ROIs) as detailed previously.61 This was achieved by first reg-istering each participant’s structural MRI to a standard atlas tar-get63 that minimizes bias due to atrophy.64 Detailed informa-tion on the boundaries of the ROIs used is available.61 Aftercorrection of head motion during the PET scan and alignmentof the MRI to the PET PIB scan, the ROIs then were applied tounblurred images of the PET dynamic data, yielding high-resolution regional time-activity curves. Time-activity curveswere analyzed for PIB-specific binding by Logan graphical anal-ysis using the cerebellum ROI data as a reference tissue inputfunction with the linear regression step applied to the data from30 to 60 minutes.65 The cerebellum was chosen as the referenceregion because there is little specific binding of PIB in postmor-tem samples of cerebellar cortex even among those with AD atautopsy.62 The Logan analysis yields a tracer distribution vol-ume ratio (DVratio), which was then converted to an estimateof the binding potential (BP) for each ROI: BP � DVratio �

1.61 The BP expresses the regional PIB binding values in a man-ner directly proportional to the number of binding sites. The BPvalues from the prefrontal cortex, gyrus rectus, lateral temporal,and precuneus ROIs were averaged in each subject to calculate amean cortical binding potential (MCBP), as these regions havebeen shown to have high PIB uptake among participants withAD.61

GenotypingDNA was extracted from peripheral blood samples using stan-dard procedures. APOE genotyping was performed as previouslydescribed.40

Statistical AnalysisAge at the time of clinical assessment was used for all analysesinvolving age. Differences in demographic characteristics be-tween participants with and without CSF data were tested usingt tests for independent samples for continuous variables and us-ing chi-square tests for categorical variables. Scatterplots wereused to illustrate associations between each biomarker and age inyears as they relate to APOE4 and APOE2 genotype. TreatingMCBP and CSF A�42, tau, and ptau181 as continuous scales,analysis of covariance was used to model these biomarkers as afunction of age and of number of APOE4 alleles (0, 1, or 2),APOE2 status (positive vs negative), sex, race (white vs non-white), and education in years. Rather than number of APOE2alleles, APOE2 status was used because only 3 individuals had 2APOE2 alleles. This method of analysis included all APOE ge-

ANNALS of Neurology

124 Volume 67, No. 1

notypes and allowed examination of APOE4 dose effect whileadjusting for all other variables in the model, including theAPOE2 effect, allowed examination of the APOE2 effect whileadjusting for APOE4 dose effects and the other variables in themodel, and disentangled the effects of APOE4 and APOE2.

Interactive effects on biomarkers among these factors wereexamined. Different error variances were fitted to the regressionmodel depending on age to test and, if present, account for in-creasing variance of the biomarkers as a function of age. Satter-thwaite’s approximation was used to estimate the denominatordegrees of freedom in F or t tests. These analyses were imple-mented using PROC MIXED/SAS.66

ResultsFrom April 2004 through October 2008, 266 CDR 0participants had at least 1 clinical assessment. Of these,241 were eligible for and completed PET PIB imaging, allwithin 2 years of the assessment confirming their CDR 0status. Of the 241 participants, 168 had CSF assays forA�42, tau, and ptau181, also all within 2 years of confir-mation of CDR 0 status. The mean interval (standard de-viation) between the clinical assessment and PET PIBscan was 0.52 years (0.42 years), and for CSF collection itwas 0.43 years (0.39 years). The mean interval betweenPET PIB scan and CSF collection was 0.56 years (0.52years). Data from many of the 241 participants have ap-peared in previous reports from our ADRC.20,61,67

Demographic information is shown in the Table.The participants ranged in age from 45.2 to 88.6 years atthe time of clinical assessment. Participants with only PIBdata were older than participants who had both PIB andCSF data (mean age 71.0 years vs 64.9 years; p �

0.0001), but there were no significant differences betweenparticipants with and without CSF data with regard to sex(p � 0.2691), minority race (p � 0.3371), presence ofAPOE4 (p � 0.5276) or APOE2 (p � 0.1840), or meanMMSE scores (p � 0.1849).

We previously found that participants with DATtypically show MCBP values for PIB �0.18,61 and haveCSF A�42 levels �500pg/ml, CSF tau levels �500pg/ml,and CSF ptau181 levels �80pg/ml20,27; our threshold ofCSF A�42 �500pg/ml is very similar to the optimumcutoff level for antemortem CSF A�42 of 515pg/ml asindexed to neuritic plaque burden at autopsy.68 (In ourexperience, neither CSF levels of A�40 nor plasma levelsof A�40 or A�42 discriminate DAT from healthy aging,27

and these markers are not further discussed here.) In theentire sample of 241, 18.3% had MCBP values at orabove 0.18. Of the 168 nondemented participants withboth PIB and CSF data, the percentage with biomarkervalues similar to those typically found among participants

with DAT were 15.5% for PIB, 28.0% for A�42, 6.6%for tau, and 4.2% for ptau181.

Unadjusted DataFor the entire sample, the unadjusted relationships be-tween MCBP for PIB with age as a function of the pres-ence or absence of at least 1 APOE4 allele are shown inFigure 1A and as a function of the presence or absence ofat least 1 APOE2 allele in Figure 1B. For the 168 partic-ipants with both PIB and CSF data, the unadjusted asso-ciations between CSF A�42, CSF tau, and CSF ptau181

with age as a function of APOE4 genotype are shown inFigure 2 and as a function of APOE2 genotype in Figure3. MCBP increased with age at a faster rate for partici-pants with at least 1 APOE4 allele compared with thosewithout an APOE4 allele (Fig 1A). Individuals with atleast 1 APOE2 allele had no increase in MCBP with age(Fig 1B). Participants with at least 1 APOE4 allele hadlower CSF A�42 values with age than those withoutAPOE4 (Fig 2A), and those with at least 1 APOE2 allelehad higher CSF A�42 values with age (Fig 3A). There wasno effect of APOE4 or APOE2 on CSF tau or ptau181

(Figs 2B–C and 3B–C). Figure 4 shows the data pre-sented above when dichotomized into PIB-positive(MCBP � 0.18) and PIB-negative individuals (n � 241)and into CSF A�42-positive (�500pg/mL) and CSFA�42-negative individuals (n � 168).

TABLE: Characteristics of 241 NondementedParticipants

Age, mean y (SD) 66.8 (10.7)

Women, No. (%) 164 (68.1)

Race, No. (%)

European American 215 (89.2)

African American 23 (9.5)

Other 3 (1.2)

Education, y (SD) 15.9 (2.7)

MMSE, mean (SD) 29.1 (1.2)

APOE genotype, No. (%)

ε2/ε2 3 (1.2)

ε2/ε3 26 (10.8)

ε2/ε4 5 (2.1)

ε3/ε3 129 (53.5)

ε3/ε4 66 (27.4)

ε4/ε4 12 (5.0)

SD � standard deviation; MMSE � Mini Mental StateExamination, with possible range of scores from 30 (best)to 0 (worst); APOE � gene encoding for apolipoprotein E.


January, 2010 125

For the entire sample, there was an age-related in-crease in the frequency of abnormal values for MCBP(�0.18) and CSF A�42 (�500pg/mL). The frequency ofindividuals with elevated MCBP for PIB was 0% at age45–49 years, 5.7% at 50–59 years, 19.0% at 60–69years, 25.8% at 70–79 years, and 30.3% at 80–89 years.The frequency of individuals with reduced levels of CSFA�42 was 18.2% at age 45–49 years (only 11 individualscontributed data in this age range), 14.0% at 50–59years, 27.1% at 60–69 years, 34.2% at 70–79 years, and50.0% at 80–89 years (only 14 individuals contributeddata in this age range).

FIGURE 1: Address correspondence to Dr Morris, AlzheimerDisease Research Center, 4488 Forest Park Avenue, Suite160, St. Louis, MO 63108. E-mail: [email protected] cortical binding potentials for Pittsburgh Compound-B(PIB) in 241 cognitively normal participants as a function ofage at clinical assessment and of: (A) the presence (redsquares) or absence (black squares) of the �4 allele of apoli-poprotein E (APOE) or (B) the presence (green squares) orabsence (black squares) of the �2 allele of APOE.

FIGURE 2: Unadjusted associations with age at clinical as-sessment and the presence (red squares) or absence (blacksquares) of apolipoprotein E4 (APOE4) in 168 cognitivelynormal participants for cerebrospinal fluid measures of: (A)amyloid-beta42 (A�42); (B) tau; and (C) tau phosphorylatedat the threonine 181 position (ptau181).

ANNALS of Neurology


Adjusted AnalysesAs suggested by the scatterplot of MCBP versus age inFigure 1, the variation of MCBP was higher among indi-viduals �55 years old when compared with those �55years old (2 [1] � 68.35, p � 0.0001). MCBP valuesincreased with greater numbers of APOE4 alleles; adjustedmean (standard error [SE]) MCBP values were 0.033(0.013), 0.198 (0.021), and 0.417 (0.055) for 0, 1, or 2APOE4 alleles (F2,182 � 41.05; p � 0.0001). A mediansplit was used to divide the MCBP values into high versuslow groups. For individuals with an APOE4 allele versusthose without an APOE4 allele, the odds ratio for thelikelihood of a high MCBP group was 5.04 (95% confi-dence interval, 2.53–10.2). There was a differential asso-ciation between MCBP and age as a function of APOE4dosage (F2,171 � 28.07, p � 0.0001). For participantswith 2 APOE4 alleles, the MCBP increased with age at anestimated mean of 0.020/year (SE � 0.003), whereas theestimated increase was slower for those with 1 APOE4allele at 0.013/year (SE � 0.001), and the slowest in-crease in MCBP with age was demonstrated by individu-als lacking an APOE4 allele at 0.003/year (SE � 0.001).None of the other factors, including education, sex, and

FIGURE 3: Unadjusted associations with age at clinical as-sessment and the presence (green squares) or absence(black squares) of apolipoprotein E4 (APOE4) in the cere-brospinal fluid of 168 cognitively normal individuals formeasures of: (A) amyloid-beta42 (A�42); (B) tau; and (C) tauphosphorylated at the threonine 181 position (ptau181).

FIGURE 4: (A) Frequency by age group for individuals withmean cortical binding potential for Pittsburgh Compound-B>0.18 (PIB-positive) in 241 cognitively normal individualsas a function of the presence (red bars) or absence (blackbars) of apolipoprotein E4 (APOE4). (B) Frequency by agegroup for individuals with cerebrospinal fluid amyloid-beta42 <500pg/ml (CSF A�42-positive) in 168 cognitivelynormal individuals as a function of the presence (red bars)or absence (black bars) of APOE4.


January, 2010 127

APOE2 status, were significantly associated with MCBPor affected the detected association between MCBP,APOE4, and age among nondemented individuals.

APOE4 was associated with lower CSF A�42 levels,even after adjusting for age and APOE2 (F2,162 � 11.74;p � 0.0001). There was a differential association betweenCSF A�42 levels and age in individuals as a function ofAPOE2 status (F1,162 � 8.3, p � 0.0045), such that CSFA�42 levels increased nonsignificantly with age in thepresence of an APOE2 allele (4.56pg/ml per year; SE �4.59) and decreased significantly in the absence of anAPOE2 allele (�7.38pg/ml per year, SE � 1.93). For in-dividuals without an APOE2 allele, the adjusted meanCSF A�42 levels were reduced with APOE4 in a dose-dependent manner; the adjusted mean CSF A�42 levelwith 0 copies of APOE4 was 671pg/ml (SE � 24.3), with1 copy of APOE4 was 588pg/ml (SE � 33.1), and with 2copies of APOE4 was 327pg/ml (SE � 73.1). None ofthe other factors, including education and sex, was signif-icantly associated with CSF A�42 or affected the associa-tions between A�42, APOE2, APOE4, and age.

The variation in both CSF tau and ptau181 levels washigher for individuals aged �55 years compared with thoseaged �55 years (for CSF tau, 2[1] � 9.01, p � 0.0027;for CSF ptau181, 2[1] � 8.65, p � 0.0033). Age was as-sociated with increasing levels of CSF tau (F1,130 � 39.62,p � 0.0001) and CSF ptau181 (F1,128 � 21.64, p �0.001). None of the other main effects (APOE4, APOE2,sex) was significant for CSF tau and CSF ptau181.

DiscussionAge and APOE genotype interact to increase the frequencyof cerebral A� deposition in cognitively normal olderadults. As measured by reductions in CSF A�42 levels andby increased MCBP for PIB, we found that cerebral dep-osition of A� begins in middle age and increases in fre-quency such that 34.2% of individuals aged 70–79 yearsand 50.0% of individuals aged 80–89 years in this studyhad lowered CSF A�42 levels; 25.8% of individuals aged70–79 years and 30.3% of individuals aged 80–89 yearshad elevated MCBPs. These percentages are comparable tothe age-related frequency of neuropathological AD on post-mortem examination of cognitively normal older adults9

and correspond to the prevalence of clinically manifestDAT in these age groups,69 as might be expected if cerebralA� deposition in asymptomatic older adults denotes pre-clinical AD that eventually will be expressed as DAT.

We confirm that APOE4 has a powerful dose-dependent effect on cerebral A� deposition with age. (Itis clear that age effects also extend to factors other thanAPOE4, as individuals lacking this allele also had age-

related elevations in MCBP for PIB and lowered CSFA�42 levels, although at much lower frequencies thanAPOE4 carriers.) The APOE4 effect persists for the olderage groups (�80 years), an unexpected finding given thatAPOE4 carriers generally are reported to have an earlieronset of DAT. Possibly APOE4 has previously unrecog-nized effects for the expression of DAT in older agegroups; alternatively, these older individuals may neverbecome symptomatic with DAT despite their APOE4-associated increases in cerebral A� deposition, althoughthis seems at odds with the comparable frequencies of A�

deposition with the prevalence of DAT at these ages.Prior studies of APOE4 in cognitively normal aging

have reported associations with poorer cognitive perfor-mance,70–72 reduced cerebral metabolic rates for glu-cose,73 and smaller regional74,75 and whole brain76 vol-umes. Recent PET imaging studies with molecular tracersfor fibrillar A� also report increased cerebral binding ofthe tracer in normal older adults who carry at least 1APOE4 allele.51,52 Ours is the first study, however, todemonstrate an APOE4 effect as a function of age on ce-rebral A� and to observe the effect with both molecularA� imaging and direct measures of CSF A�42 concentra-tions. The concordance of the findings with both mea-sures of A� but not with measures of tau or ptau181 in-dicates that cerebral A� deposition is the pathobiologicalphenotype of the APOE4 genotype.

To our knowledge, we also provide the first demon-stration in cognitively normal older adults of a protectiveeffect of APOE2 against A� deposition as measured byboth PIB and CSF A�42 levels. Because APOE2 protectsagainst development of DAT, this finding supports apathogenic role for cerebral A� deposition in AD, at leastin its preclinical stages. In contrast, CSF levels of tau andptau181 increase as a function of age in cognitively normalolder adults but are not affected by APOE genotype. Themechanisms by which APOE4 exerts its dramatically in-creased risk for AD thus appear to involve A� but not taumetabolism, consistent with observations that brain atrophyand longitudinal cognitive decline in nondemented agingcorrelate with elevated MCBP for PIB and with abnormal-ities in CSF A�42 but not CSF tau or ptau.20 Hence, A� iscentral to the initial detectable pathological changes in pre-clinical AD, with changes in tau likely occurring later.5

The initial detection of perturbed cerebral A� me-tabolism may be reflected by reduced CSF A�42 levels,followed by elevated MCBP values for PIB. The data inFigure 4 suggest that CSF A�42 reductions are shifted toan earlier age and persist in greater frequencies with agethan elevations in MCBP for PIB, but interpretative cau-tion is indicated by the cross-sectional nature of the data.

ANNALS of Neurology


Lowered concentrations of CSF A�42 perhaps reflect ini-tial A� deposition in the brain in the form of diffuse SPs.These initial A� deposits, which may be downstream ofother A� toxic species, such as dimers and oligomers,77

may be largely nonfibrillar and hence unable to bind PIBin concentrations sufficient for detection by PET. Thus,these data suggest that the biomarker sequence for detec-tion of preclinical AD may be an initial reduction of CSFA�42 levels, followed by elevated MCBP for PIB after theA� deposits become fibrillar, but longitudinal studies areneeded for a definitive determination of the biomarkersequence for AD.

This study has several strengths. To our knowledge,this is the largest sample (n � 241) of older individuals(demented, cognitively normal, or combined) for whichPET PIB findings are reported. The A� imaging datawere combined with CSF levels of A�42, tau, and ptau181

in 70% of the sample. The sample size permitted the firstexamination of APOE2 effects on markers of preclinicalAD. The wide age span (from 45 to 88 years) allowedexamination of the relationships of APOE and the ADbiomarkers with age. Finally, the 241 participants werecarefully characterized to avoid contamination with indi-viduals experiencing even minimal cognitive deficits.

The study also has limitations. We used a conveniencesample of individuals willing to be followed longitudinallyand intensively (eg, PET PIB, CSF collection), possibly re-ducing the generalizability of the findings to the larger pop-ulation. Interpretative caution is recommended for results in-volving some cells with small sample sizes, such as thenumber of individuals (n � 34) carrying an APOE2 allele(the size of our APOE2 subsample alone, however, is com-parable to the size of the total samples used in recent reportsof APOE4 effects on amyloid imaging).50–52 Finally, theconcept of preclinical AD must remain speculative untilthere is enough evidence that cognitively normal older adultswith reduced CSF A�42, elevated CSF tau, elevated MCBPfor PIB, or other indicators for preclinical AD have dispro-portionately greater risk for developing DAT than thosewithout biomarker abnormalities.

AD is a complex disorder, and its pathogenesis al-most certainly cannot be explained simply by abnormalmetabolism of A�. However, we find powerful evidencethat cerebral A� deposition with age is the pathobiologi-cal phenotype for APOE4, the strongest genetic risk factorfor late-onset AD, and that the protective effect ofAPOE2 against developing AD is mediated by its effectsagainst cerebral A� deposition. We also find evidence thatA� abnormalities, but not tau abnormalities, initiate thepathological cascade of preclinical AD, consistent withour previous findings that reduced levels of CSF A�42 are

associated with low whole brain volumes in nondementedindividuals but not those with DAT, whereas elevatedCSF tau levels are not associated with low whole brainvolumes in nondemented persons but are for those withDAT.20 Finally, reduced levels of CSF A�42 and elevatedMCBP for PIB identify presumptive preclinical AD in asubstantial number of cognitively normal older adults andprovide the opportunity to identify these individuals forlongitudinal studies to determine their risk for DAT.

This work was supported by National Institute on Aginggrants: P50 AG05681, P01 AG03991, and P01AG026276 (J.C.M.); the Charles and Joanne Knight Alz-heimer Research Initiative (J.C.M.) of the WashingtonUniversity Alzheimer’s Disease Research Center, St. Louis,MO; and the Postdoctoral Program of 1UL1RR024992-01from the National Center for Research Resources (C.M.R.).

The authors thank the investigators and staff of the Alzhei-mer’s Disease Research Center’s Clinical (participant assess-ments) and Genetics Cores (genotyping) and the investigatorsand staff of the Adult Children Study’s Biomarker Corefor cerebrospinal fluid analytes.

References1. Morris JC, Price JL. Pathologic correlates of nondemented aging,

mild cognitive impairment, and early stage Alzheimer’s disease. JMol Neurosci 2001;17:101–118.

2. Katzman R, Terry R, DeTeresa R, et al. Clinical, pathological, andneurochemical changes in dementia: a subgroup with preservedmental status and numerous neocortical plaques. Ann Neurol1988;23:138–144.

3. Troncoso JC, Cataldo M, Nixon RA, et al. Neuropathology ofpreclinical and clinical late-onset Alzheimer’s disease. Ann Neurol1998;43:673–676.

4. Hulette CM, Welsh-Bohmer KA, Murray MG, et al. Neuropatho-logical and neuropsychological changes in “normal” aging: evi-dence for preclinical Alzheimer disease in cognitively normal in-dividuals. J Neuropathol Exp Neurol 1998;57:1168–1174.

5. Price JL, Morris JC. Tangles and plaques in nondemented agingand “preclinical” Alzheimer’s disease. Ann Neurol 1999;45:358–368.

6. Schmitt FA, Davis DG, Wekstein DR, et al. Preclinical AD revis-ited. Neuropathology of cognitively normal older adults. Neurol-ogy 2000;55:370–376.

7. Knopman DS, Parisi JE, Salviati A, et al. Neuropathology of cog-nitively normal elderly. J Neuropathol Exp Neurol 2003;62:1087–1095.

8. Bennett DA, Schneider JA, Arvanitakis Z, et al. Neuropathologyof older persons without cognitive impairment from twocommunity-based studies. Neurology 2006;66:1837–1844.

9. Price JL, McKeel DW, Buckles VD, et al. Neuropathology of non-demented aging: presumptive evidence for preclinical Alzheimerdisease. Neurobiol Aging 2009;30:1026–1036.

10. Wisniewski KE, Wisniewski HM, Wen GY. Occurrence of neuro-pathological changes and dementia of Alzheimer’s disease inDown’s syndrome. Ann Neurol 1985;17:278–282.

11. Cork LC. Neuropathology of Down syndrome and Alzheimer dis-ease. Am J Med Genet 1990;37:282–286.


January, 2010 129

12. Stern Y. Cognitive reserve and Alzheimer disease. Alz Dis AssocDisord 2006;20:112–117.

13. Mortimer JA, Borenstein AR, Gosche KM, et al. Very early de-tection of Alzheimer neuropathology and the role of brain re-serve in modifying its clinical expression. J Geriatr PsychiatryNeurol 2008;18:218–223.

14. Roe CM, Mintun MA, D’Angelo G, et al. Alzheimer disease andcognitive reserve: variation of education with carbon 11-labeledPittsburgh Compound B uptake. Arch Neurol 2008;65:1467–1471.

15. Rowe CC, Ng S, Ackermann U, et al. Imaging �-amyloid burdenin aging and dementia. Neurology 2007;68:1718–1725.

16. Ewers M, Zhong Z, Burger K, et al. Increased CSF-BACE I activity isassociated with ApoE-E4 genotype in subjects with mild cognitiveimpairment and Alzheimer’s disease. Brain 2008;131:1252–1258.

17. Jack CR, Lowe VJ, Senjem ML, et al. 11C PIB and structural MRIprovide complementary information in imaging of Alzheimer’sdisease and amnestic mild cognitive impairment. Brain 2008;131:665–680.

18. Shaw LM, Vanderstichele H, Knapik-Czajka M, et al. Cerebrospi-nal fluid biomarker signature in Alzheimer’s disease neuroimag-ing initiative subjects. Ann Neurol 2009;65:403–413.

19. Bakkour A, Morris JC, Dickerson BC. The cortical signature ofprodromal AD: regional thinning predicts mild AD dementia.Neurology 2009;72:1048–1055.

20. Fagan AM, Head D, Shah AR, et al. Decreased cerebrospinalfluid A�42 correlates with brain atrophy in cognitively normalelderly. Ann Neurol 2009;65:176–183.

21. Sunderland T, Linker G, Mirza N, et al. Decreased �-amyloid1-42 and increased tau levels in cerebrospinal fluid of patientswith Alzheimer disease. JAMA 2003;289:2094–2103.

22. Clark CM, Xie S, Chittams J, et al. CSF tau and beta-amyloid:how well do these biomarkers reflect autopsy confirmed demen-tia diagnoses? Arch Neurol 2003;60:1696–1702.

23. Klunk WE, Engler H, Nordberg A, et al. Imaging brain amyloid inAlzheimer’s disease with Pittsburgh Compound-B. Ann Neurol2004;55:306–319.

24. Ikonomovic MD, Klunk WE, Abrahamson EE, et al. Post-mortemcorrelates of in vivo PiB-PET amyloid imaging in a typical case ofAlzheimer’s disease. Brain 2008;131:1630–1645.

25. Leinonen V, Alafuzoff I, Aalto S, et al. Assessment of beta-amyloid in a frontal cortical brain biopsy specimen and bypositron emission tomography with carbon 11-labeled PittsburghCompound B. Arch Neurol 2008;65:1304–1309.

26. Fagan AM, Mintun MA, Mach RH, et al. Inverse relation betweenin vivo amyloid imaging load and cerebrospinal fluid Ab42 inhumans. Ann Neurol 2006;59:512–519.

27. Fagan AM, Roe CM, Xiong C, et al. Cerebrospinal fluid tau/�-amyloid42 ratio as a prediction of cognitive decline in nonde-mented older adults. Arch Neurol 2007;64:343–349.

28. Petersen RC, Smith GE, Waring SC, et al. Mild cognitiveimpairment: clinical characterization and outcome. Arch Neurol1999;56:303–308.

29. Hansson O, Zetterberg H, Buchhave P, et al. Association be-tween CSF biomarkers and incipient Alzheimer’s disease in pa-tients with mild cognitive impairment: a follow-up study. LancetNeurol 2006;5:228–234.

30. Simonsen AH, McGuire J, Hansson O, et al. Novel panel ofcerebrospinal fluid biomarkers for the prediction of progressionto Alzheimer dementia in patients with mild cognitive impair-ment. Arch Neurol 2007;64:366–370.

31. Forsberg A, Engler H, Almkvist O, et al. PET imaging of amyloiddeposition in patients with mild cognitive impairment. NeurobiolAging 2008;29:1456–1465.

32. Gustafson DR, Skoog I, Rosengren L, et al. Cerebrospinal fluidbeta-amyloid 1-42 concentration may predict cognitive decline inolder women. J Neurol Neurosurg Psychiatry 2007;78:461–464.

33. Li G, Sokal I, Quinn JF, et al. CSF tau/Abeta42 ratio for in-creased risk of mild cognitive impairment: a follow-up study.Neurology 2007;69:631–639.

34. Blennow K, deLeon MJ, Zetterberg H. Alzheimer’s disease. Lan-cet 2006;368:387–403.

35. Corder E, Saunders A, Strittmatter W, et al. Gene dose of apo-lipoprotein E type 4 allele and the risk of Alzheimer’s disease inlate onset families. Science 1993;261:921–923.

36. Farrer L, Cupples LA, Haines JL, et al. Effects of age, sex, andethnicity on the association between apolipoprotein E genotypeand Alzheimer disease. APOE and Alzheimer Disease Meta Anal-ysis Consortium. JAMA 1997;278:1349–1356.

37. Bertram L, Lange C, Mullin K, et al. Genome-wide associationanalysis reveals putative Alzheimer’s disease susceptibility loci inaddition to APOE. Am J Hum Genet 2008;83:623–632.

38. Strittmatter WJ, Saunders AM, Schmechel D, et al. Apolipopro-tein E: high-affinity binding to beta-amyloid and increased fre-quency of type 4 allele in late-onset familial Alzheimer disease.Proc Natl Acad Sci U S A 1993;90:1977–1981.

39. Corder EH, Saunders AM, Risch NJ, et al. Protective effect ofapolipoprotein E type 2 allele for late onset Alzheimer disease.Nat Genet 1994;7:180–184.

40. Talbot C, Lendon C, Craddock N, et al. Protection against Alzhei-mer’s disease with apoE e2 [letter]. Lancet 1994;343:1432–1433.

41. Jiang Q, Lee DYD, Mandrekar S, et al. ApoE promotes the pro-teolytic degradation of A�. Neuron 2008;58:681–693.

42. Ma J, Yee A, Brewer HB Jr, Das S, Potter H. Amyloid-associatedproteins alpha 1-antichymotrypsin and apolipoprotein E promoteassembly of Alzheimer beta-protein into filaments. Nature 1994;372:92–94.

43. DeMattos RB, Cirrito JR, Parsadanian M, et al. ApoE and clus-terin cooperatively suppress A� levels and deposition: evidencethat apoE regulates extracellular A� metabolism in vivo. Neuron2004;41:193–202.

44. Holtzman D, Bales KR, Tenkova T, et al. Apolipoprotein Eisoform-dependent amyloid deposition and neuritic degenerationin a mouse model of Alzheimer’s disease. Proc Natl Acad Sci US A 2000;97:2892–2897.

45. Fagan AM, Watson M, Parsadanian M, et al. Human and murineapoE markedly influence Abeta metabolism before and afterplaque formation in a mouse model of Alzheimer’s disease. Neu-robiol Dis 2002;9:305–318.

46. Polvikoski T, Sulkava R, Haltia M. Apolipoprotein E, dementia,and cortical deposition of �-amyloid protein. N Engl J Med1995;333:1242–1247.

47. Schmechel DE, Saunders AM, Strittmatter WJ, et al. Increasedamyloid beta-peptide deposition in cerebral cortex as a con-sequence of apolipoprotein E genotype in late-onsetAlzheimer disease. Proc Natl Acad Sci U S A 1993;90:9649 –9653.

48. Rebeck GW, Reiter JS, Strickland DK, et al. Apolipoprotein E insporadic Alzheimer’s disease: allelic variation and receptor inter-actions. Neuron 1993;11:575–580.

49. Bennett DA, Wilson RS, Schneider JA, et al. Apolipoprotein E E4allele, AD pathology, and the clinical expression of Alzheimer’sdisease. Neurology 2003;60:246–252.

50. Drzezga A, Grimmer T, Henriksen G, et al. Effect of APOE ge-notype on amyloid plaque load and gray matter volume in Alz-heimer disease. Neurology 2009;72:1487–1494.

51. Reiman EM, Chen K, Liu X, et al. Fibrillar amyloid-beta burden incognitively normal people at 3 levels of genetic risk for Alzhei-mer’s disease. Proc Natl Acad Sci U S A 2009;106:6820–6825.

ANNALS of Neurology


52. Small GW, Siddarth P, Burggren AC, et al. Influence of cognitivestatus, age, and APOE-4 genetic risk on brain FDDNP positron-emission tomography imaging in persons without dementia. ArchGen Psychiatry 2009;66:81–87.

53. Berg L, McKeel DW Jr, Miller JP, et al. Clinicopathologic stud-ies in cognitively healthy aging and Alzheimer disease: rela-tion of histologic markers to dementia severity, age, sex, andapolipoprotein E genotype. Arch Neurol 1998;55:326 –335.

54. Coats MA, Morris JC. Antecedent biomarkers of Alzheimer’sdisease: the Adult Children Study. J Geriatr Psychiatry Neurol2005;18:242–244.

55. Storandt M, Grant EA, Miller JP, et al. Longitudinal course andneuropathological outcomes in original versus revised MCI andin PreMCI. Neurology 2006;67:467–473.

56. Folstein MF, Folstein SE, McHugh PR. Mini-mental State: a prac-tical method for grading the cognitive state of patients for theclinicians. J Psychiatr Res 1975;12:189–198.

57. Morris JC. The Clinical Dementia Rating (CDR): current versionand scoring rules. Neurology 1993;43:2412–2414.

58. Morris JC, Weintraub S, Chui HC, et al. The Uniform Data Set(UDS): clinical and cognitive variables and descriptive data fromAlzheimer disease centers. Alz Dis Assoc Disord 2006;20:210–216.

59. Storandt M, Grant EA, Miller JP, et al. Rates of progression inmild cognitive impairment and early Alzheimer’s disease. Neurol-ogy 2002;59:1034–1041.

60. Cirrito JR, May PC, O’Dell MA, et al. In vivo assessment of braininterstitial fluid with microdialysis reveals plaque-associatedchanges in amyloid-beta metabolism and half-life. J Neurosci2003;23:8844–8853.

61. Mintun MA, LaRossa GN, Sheline YI, et al. [11C] PIB in a non-demented population: potential antecedent marker of Alzheimerdisease. Neurology 2006;67:446–452.

62. Mathis CA, Wang Y, Holt DP, et al. Synthesis and evaluation of11C-labeled 6-substituted 2-arylbenzothiazoles as amyloid imag-ing agents. J Med Chem 2003;46:2740–2754.

63. Talairach J, Tournoux P. Co-planar Stereotaxic Atlas of the Hu-man Brain: 3-D Proportional System: An Approach to CerebralImaging. Stuttgart, Germany: Thieme Medical Publishers; 1988.

64. Buckner RL, Head D, Parker J, et al. A unified approach for morpho-metric and functional data analysis in young, old, and demented adultsusing automated atlas-based head size normalization: reliability and val-

idation against manual measurement of total intracranial volume. Neu-roimage 2004;23:724–738.

65. Logan J, Fowler JS, Volkow ND, et al. Distribution volume ratioswithout blood sampling from graphical analysis of PET data.J Cereb Blood Flow Metab 1996;16:834–840.

66. Littell RC, Milliken GA, Stroup WW, et al. SAS System for MixedModels. Cary, NC: SAS Institute; 1996.

67. Kauwe JSK, Wang J, Mayo K, et al. Alzheimer’s disease riskvariants show association with cerebrospinal fluid amyloid beta.Neurogenetics 2009;10:13–17.

68. Tapiola T, Alafuzoff I, Herruka SK, et al. Cerebrospinal fluid�-amyloid 42 and tau proteins as biomarkers of Alzheimer-typepathologic changes in the brain. Arch Neurol 2009;66:382–389.

69. Hebert LE, Scherr PA, Bienias JL, et al. Alzheimer disease in theUS population. Arch Neurol 2003;60:1119–1122.

70. Yaffe K, Cauley J, Sands L, et al. Apolipoprotein E phenotypeand cognitive decline in a prospective study of elderly commu-nity women. Arch Neurol 1997;54:1110–1114.

71. Bretsky P, Guralnik JM, Launer L, et al. The role of APOE-e4 inlongitudinal cognitive decline; MacArthur studies of successfulaging. Neurol 2003;60:1077–1081.

72. Haan MN, Shemanki L, Jagust WJ, et al. The role of APOE e4 inmodulating effects of other risk factors for cognitive decline inelderly persons. JAMA 1999;282:40–46.

73. Reiman EM, Caselli RJ, Yun LS, et al. Preclinical evidence ofAlzheimer’s disease in persons homozygous for the epsilon 4allele for apolipoprotein E. N Engl J Med 1996;334:752–758.

74. Wishart HA, Saykin AJ, McAllister TW, et al. Regional brain at-rophy in cognitively intact adults with a single APOEE4 allele.Neurology 2006;67:1221–1224.

75. Schuff N, Woerner N, Boreta L, et al. MRI of hippocampal vol-ume loss in early Alzheimer’s disease in relation to ApoE geno-type and biomarkers. Brain 2009;132:1067–1077.

76. Chen K, Reiman EM, Alexander GE, et al. Correlations betweenapolipoprotein E E4 gene dose and whole brain atrophy rates.Am J Psychiatry 2007;164:916–921.

77. Shankar GM, Li S, Mehta TH, et al. Amyloid-beta protein dimersisolated directly from Alzheimer’s brains impair synaptic plasticityand memory. Nat Med 2008;837–842.


January, 2010 131

apoe predicts amyloid-beta but not tau alzheimer pathology in cognitively normal aging

Documents