A

cnhrtfio©

K

1

ncauoMae(

tRT

0d

Neurobiology of Aging 31 (2010) 1122–1131

White matter diffusion alterations in normal women atrisk of Alzheimer’s disease

Charles D. Smith a,b,c,∗, Himachandra Chebrolu b, Anders H. Andersen b,c,David A. Powell b, Mark A. Lovell d, Shuling Xiong d, Brian T. Gold c

a Department of Neurology, University of Kentucky, Lexington, KY 40536, United Statesb Magnetic Resonance Imaging and Spectroscopy Center, University of Kentucky, Lexington, KY 40536, United States

c Department of Anatomy and Neurobiology, University of Kentucky, Lexington, KY 40536, United Statesd Department of Chemistry, University of Kentucky, Lexington, KY 40536, United States

Received 18 May 2008; received in revised form 15 July 2008; accepted 4 August 2008Available online 17 September 2008

bstract

Increased white matter mean diffusivity and decreased fractional anisotropy (FA) has been observed in subjects diagnosed with mildognitive impairment (MCI) and Alzheimer’s disease (AD). We sought to determine whether similar alterations of white matter occur inormal individuals at risk of AD. Diffusion tensor images were acquired in 42 cognitively normal right-handed women with both a familyistory of dementia and at least one apolipoprotein E4 allele. These were compared with images from 23 normal women without either ADisk factor. Group analyses were performed using tract-based spatial statistics. Reduced FA was observed in the fronto-occipital and inferior

emporal fasciculi (particularly posteriorly), the splenium of the corpus callosum, subcallosal white matter and the cingulum bundle. Thesendings demonstrate that specific white matter pathways are altered in normal women at increased risk of AD years before the expected onsetf cognitive symptoms.

2008 Elsevier Inc. All rights reserved.

aging; A

gsirTtbca

eywords: Diffusion tensor; Fractional anisotropy; Magnetic resonance im

. Introduction

There is now strong evidence that significant Alzheimer’seuropathology is present years before any symptoms of mildognitive impairment (MCI) or Alzheimer’s disease (AD)ppear (Braak and Braak, 1997; Bennett et al., 2006). Thenderlying theory of this study is that the first appearancef cognitive symptoms due to AD pathology, e.g., amnesticCI, represents early failure of brain compensatory mech-

nisms and cerebral reserve, a critical point in the furthervolution of neuronal injury produced by that pathologySmith, 2007). We assume that AD neuropathology pro-

∗ Corresponding author at: Magnetic Resonance Imaging and Spec-roscopy Center, Room 62, University of Kentucky Medical Center, 800ose Street, Lexington, KY 40536-0098, United States.el.: +1 859 323 1113; fax: +1 859 323 1068.

E-mail address: csmith@mri.uky.edu (C.D. Smith).

tagriraw

197-4580/$ – see front matter © 2008 Elsevier Inc. All rights reserved.oi:10.1016/j.neurobiolaging.2008.08.006

lzheimer’s disease; Apolipoprotein E

resses with time. But even if a treatment could theoreticallylow or arrest MCI/AD, other processes of aging and vascularnjury could operate on this substrate of critical reserve andesult in further cognitive decline (Schneider et al., 2007).herefore, identifying the brain at risk before MCI is impor-

ant to the goal of preventing the critical point from evereing reached, by using AD disease-modifying treatments,ombined with treatments aimed at the comorbid pathologiesnd enhancement of brain reserve, before symptoms appear.

Several methodologies have been applied, or have poten-ial to be applied, to asymptomatic persons to assess the braint AD risk, including cerebrospinal fluid protein patterning,enetic analysis, resting glucose positron emission tomog-aphy, and functional and structural magnetic resonance

maging. Studies of white matter integrity using magneticesonance diffusion tensor imaging have demonstrated alter-tions in the frontal, temporal and parietal lobes in subjectsith MCI and AD (Medina et al., 2006; Xie et al., 2006;

ogy of A

F2a2tbdmott2fkispwntmktd

2

2

liascblhrhlmLa

matatphd

u

taIVWIatD(TitamffsUp

Apioomhbsqgasat

brrephstandard least-squares model was then developed for the psy-chometric test scores, with risk group, age, education and riskby age interaction as the independent variables versus thefollowing independent variables: MMSE, Logical Memory

Table 1AAge and education of study participants

Group N Age, years Education, years

C.D. Smith et al. / Neurobiol

ellgiebel et al., 2007; Firbank et al., 2007; Huang et al.,007; Stahl et al., 2007; Sydykova et al., 2007; Teipel etl., 2007; Zhang et al., 2007; Ding et al., 2008; Rose et al.,008; Yasmin et al., 2008). There is pathological evidencehat disruption of white matter in Alzheimer’s disease maye an early event (Englund et al., 1988). In line with thisata, there is important recent evidence suggesting that nor-al persons at increased risk for late-onset AD by virtue

f an apolipoprotein-E4 (APOE4) allele show reduced frac-ional anisotropy (FA) in certain brain regions comparedo low-risk groups (Nierenberg et al., 2005; Persson et al.,006a,b). However, these studies have either explored only aew regions with ROI analyses, or used automated techniquesnown to have limitations related to registration and smooth-ng (Bookstein, 2001; Smith et al., 2006). In the presenttudy we applied rigorously validated methods for com-aring group fractional anisotropy maps across the brain’shite matter (Smith et al., 2006) between well-characterizedormal subjects who vary in their risk of AD. The main advan-ages of these methods are improved registration of white

atter, the lack of a need to apply an arbitrary smoothingernel, and a common reference framework (mean FA skele-on). In addition, because the method is voxel-based, arbitraryefinitions of region of interest are not required.

. Methods

.1. Subjects

All subjects are participants in a University of Kentuckyongitudinal imaging study of normal persons who varyn their risk of AD based on family history of dementiand APOE allele status (Smith et al., 1999, 2002). Inclu-ion criteria include generally good health without memoryomplaints, stable medication regimen, right handed femaleetween the age of 40 and 90 years, vision correctable to ateast 20/50, and knowledge of family history (FH). Familyistory was considered positive if one or more first-degreeelatives had late-onset progressive dementia. Only right-anded females were recruited because they are part of aarger multimodal imaging study that includes functional

agnetic resonance imaging during language-based tasks.anguage task-based fMRI is known to demonstrate gendernd handedness activation differences.

Exclusions include current medical evaluation or treat-ent for acute or worsening complaints (e.g., unstable

sthma), stroke or other cerebral injury, psychoactive medica-ions except stable doses of serotonin reuptake inhibitors andbsence of active depression, significant head injury (opera-ionally loss of consciousness greater than 5 min), significantsychiatric history, claustrophobia, metallic implants in the

ead or neck, pacemakers, major surgery within 3 months, oreath of one or both parents before age 65.

Participants who are selected for scanning based on riskndergo standardized psychometric testing which includes

HLp

Sa

ging 31 (2010) 1122–1131 1123

he Mini Mental State Examination (MMSE), Trails A (TA)nd B (TB), FAS Letter Fluency (LF), Wechsler Adultntelligence Scale (WAIS) Block Design (Block Des) andocabulary (Vocab), 15-item Boston Naming Test (BNT),echsler Memory Scale (WMS) Logical Memory I (LM

mm) and II (LM Del), Visual Reproduction immediatend delayed (VR and VR Del), Digit Span (DS), Spa-ial Span (SS), Selective Reminding Test Immediate andelayed (SRT & SRT Del), and Beck depression inventory

Beck). Abbreviations are given for reference to Table 1B.hese tests are given to exclude persons with cognitive

mpairment and to document equivalent normal cognitiveest performance between risk groups. All subjects hadn APOE allele determination. APOE alleles were deter-ined using polymerase chain reaction (PCR) amplification

ollowed by enzymatic digestion and separation of DNAragments by agarose gel electrophoresis. Informed con-ent was obtained from each enrollee under an approvedniversity of Kentucky medical institutional review boardrotocol.

Forty-four high risk (both FH dementia and one or morePOE4 alleles) and 27 low risk (neither FH nor APOE4)articipants underwent diffusion tensor magnetic resonancemaging (DT-MRI) between the years 2006 and 2007. Onef each risk group were excluded because of poor mem-ry performance on cognitive testing (1.5 S.D. below theean on Logical Memory II), and an additional one of the

igh risk and three of the low-risk group were not analyzedecause of poor quality MRI scans. None of the includedcans demonstrated small or large-vessel infarcts. Conse-uently there were 42 in the high-risk and 23 in the low-riskroup available for DT analysis. Demographic informationnd testing results are summarized in Table 1. Psychometriccores that sample general mental ability, processing speed,nd delayed verbal and nonverbal memory are provided inhe table.

Demographic and psychometric scores were comparedetween risk groups. First age and education were entered asesponse variables in a standard least-squares analysis withisk group as the independent variable (Table 1A). Age andducation were chosen as the most likely to variables to affectsychometric test scores. Because all subjects were right-anded females, gender and handedness were not factors. A

igh risk 42 57.7 ± 1.1 15.7 ± 0.4ow risk 23 68.1** ± 1.5 15.9 ± 0.6-Value – <0.0001 0.73

tandard Least Squares analysis using age and education as response vari-bles and risk as regressor (Least square means ± S.E.M.).

1124 C.D. Smith et al. / Neurobiology of Aging 31 (2010) 1122–1131

Table 1BComparison of psychometric test scores between risk groups

MMSE LM Imm LM Del Block Des VR VR Del BNT Vocab

Risk high 29.3 ± 0.3 49.8 ± 1.6 32.6 ± 1.3 36.3 ± 1.7 80.1 ± 1.8 62.8 ± 3.6 14.4 ± 0.2 53.6 ± 1.2Risk low 29.1 ± 0.2 51.4 ± 2.2 32.7 ± 1.8 43.3 ± 2.4 83.1 ± 2.5 65.9 ± 5.1 14.6 ± 0.3 56.8 ± 1.8

p-ValuesRisk 0.49 0.56 0.96 0.02* 0.32 0.63 0.44 0.14Risk × age 0.61 0.04* 0.09 0.77 0.004** 0.19 0.77 0.49Education 0.28 0.34 0.30 0.12 0.10 0.80 0.91 0.08Age 0.58 0.44 0.62 0.0001** (−) 0.0004** (−) 0.009** (−) 0.58 0.13

SRT SRT 30 TA TB LF SS DS Beck

Risk high 103.1 ± 3.4 8.5 ± 0.5 30.8 ± 1.2 70.7 ± 4.1 37.4 ± 2.5 15.5 ± 0.5 17.3 ± 0.7 7.3 ± 1.2Risk low 106.9 ± 4.8 9.2 ± 0.7 25.0 ± 1.8 60.7 ± 5.8 45.6 ± 3.5 17.5 ± 0.7 17.3 ± 1.0 3.2 ± 1.6

p-ValuesRisk 0.52 0.41 0.01** 0.16 0.06 0.02* 0.99 0.05*Risk × age 0.80 0.75 0.68 0.99 0.33 0.77 0.79 0.60Education 0.29 0.99 0.90 0.78 0.07 0.70 0.34 0.05*Age 0.09 (−) 0.05* (−) 0.002** (+) 0.002** (+) 0.06 0.03* (−) 0.61 0.66

High risk: presence of both family history of late-onset progressive dementia in a first-degree relative and possession of at least one APOE4 allele. Low risk:absence of both factors. Standard Least Squares analysis using psychometric raw scores as response variables and risk, education, age, and risk × age interactionas regressors. Least square means ± S.E.M. are entered in the first two rows. p-Values are uncorrected; values between 0.05 down to 0.01 are denoted by asingle asterisk (*), values 0.01 or less by double asterisks (**). Because no correction was made for the 16 multiple comparisons, we chose to conservativelyinterpret * as indicating marginal significance, and ** as significant. Abbreviations: MMSE, Mini Mental State Examination; LM Imm, Logical Memory I;LM Del, Logical memory II; Block Des, Block Design; VR, Visual Reproduction Immediate; VR Del, Visual Reproduction Delayed; BNT, 15-item BostonN te; SRTS orrelati

iitBi

rmsw(chtattgt

2

Wwwlis

2

c(pffmetbpawutp

rwtwsw

aming Test; Vocab, Vocabulary; SRT, Selective Reminding Test Immediapatial Span; DS, Digit Symbol; LF, FAS Letter Fluency. The direction of c

mmediate and delayed, Block Design, Visual Reproductionmmediate and delayed, Boston naming, Vocabulary, Selec-ive Reminding Test immediate and delayed, Trails A and, Letter Fluency, Spatial Span, Digit Symbol, and Beck

nventory score (Table 1B).To test for potential differences between risk groups

elated to vascular factors or estrogen replacement thatight affect white matter, Chi-Square analysis using Pear-

on’s criterion without correction for multiple comparisonsas performed on the following categorical variables

Present/Not Present): current smoking, past smoking,urrent estrogen replacement, past estrogen replacement,ypercholesterolemia, hypertension and diabetes. There werehree subjects with stable diabetes in the study, one in the low-nd two in the high-AD risk group. None of the p-values inhese comparisons was less than p = 0.16, demonstrating thathe frequencies of these factors were the same between riskroups. More detailed data is presented in a supplementaryable on the journal web site (http://www.elsevier.com).

.2. MRI protocol

Scanning was performed on a Siemens Trio 3T instrument.hole-brain (40-slice) diffusion tensor images were acquiredith 12 encoding directions with 3 averages each (B0 images

ere averaged 6 times; TR = 14600 ms, TE = 96 ms; reso-

ution 1.8 mm × 1.8 mm × 3.0 mm) using a fluid attenuatednversion recovery (FLAIR) double-refocused spin echoequence.

wosf

Del, Selective Reminding Test Delayed; TA, Trails A; TB, Trails B; SS,on is indicated by a plus (+) or minus (−) sign when p < 0.05.

.3. Image processing and statistics

Voxelwise statistical analysis of the FA data wasarried out using TBSS (Tract-Based Spatial StatisticsSmith et al., 2006)), part of the publicly available imagerocessing software FSL 4.0 (http://www.fmrib.ox.ac.uk/sl/tbss/index.html). TBSS was used because its algorithmsor alignment of FA images from multiple subjects into a com-on space have been extensively tested and validated (Smith

t al., 2006). In addition, the technique does not require spa-ial smoothing of images, which can produce different resultsased upon the kernel size (Jones et al., 2005). All of therocessing modules referred to below by their acronym arevailable for download at the FSL web site. First, FA imagesere created by fitting a tensor model to the raw diffusion datasing FSL’s Diffusion Toolbox, which consists of a series ofools for low-level diffusion parameter reconstruction androbabilistic tractography.

In order to minimize the extent of warping required foregistration to a common target, a subset of 25 subjectsas used to select a target image for nonlinear registra-

ion. Subjects forming both the whole set and the subsetere scanned over the same time interval. The subset was

elected in two steps. First, the brain volumes of all subjectsere calculated and rank ordered. Second, age-volume means

ithin the interquartile range were arranged in chronologicalrder, and the 25 subjects closest in age to the median wereelected. From the 25 subjects, the optimum target image wasound.

ogy of A

el(Fiasrmrsjitvi

aFmpurdtiFib(

3

gfciLdlbteptemTflcsg

tBScitsmVsbdf

fdtsabwAnhn3pwgeVi

sslmoarw1c

mstswll

C.D. Smith et al. / Neurobiol

The subset subjects’ FA images were aligned withvery other subset subjects’ FA images using the non-inear registration toolkit IRTK (Rueckert et al., 1999)http://www.doc.ic.ac.uk/∼dr/software). The common targetA image for the nonlinear registration was chosen as the FAmage for which the least amount of total nonlinear warpingcross all subset subjects was necessary for registration to theame space. The least amount of warping was determined byegistration of each subject to every other subject, and sum-arization of each warp field by its mean displacement. The

esulting target FA image was then affine registered to MNIpace. A single transformation matrix for each of the 65 sub-ects, combining the nonlinear registration of the subject’s FAmage to the target FA image and the affine registration of thearget FA image to MNI space, was then applied. The indi-idual FA images were resampled at 1 mm × 1 mm × 1 mmsotropic voxels using sinc interpolation.

Next, a mean FA image from all of the subjects was cre-ted and thinned with a FA threshold of 0.2 to create a meanA skeleton which represents the centers of all tracts com-on to the group. Each subject’s aligned FA data was then

rojected onto this skeleton and the resulting data analyzedsing voxelwise cross-subject statistics. A false discoveryate (FDR) threshold scaled to an alpha of p = 0.05 wasetermined via random simulations (using 5000 permuta-ions) to correct for multiple comparisons (implemented asn: http://www.fmrib.ox.ac.uk/fsl/randomise/fdr.html). TheDR was then applied to all voxel-based group compar-

sons. Centered age was used as a covariate in the analysis,ecause the risk groups showed an overall difference in agesee Section 3).

. Results

Demographic and testing characteristics of the studyroups are given in Table 1. Because no correction was madeor the 16 multiple comparisons in Table 1B, we chose toonservatively interpret p between 0.01 and 0.05 as indicat-ng marginal significance, and p less than 0.01 as significant.ow-AD risk subjects were older than high-risk subjectsue to the convenience nature of the sample. We specu-ate that normal working-age high-risk younger subjects maye more motivated to volunteer their time for this studyhan similarly aged low-risk volunteers because of personalxperience with Alzheimer’s disease in family members. Inrevious studies using APOE allele status for risk stratifica-ion, the average age was approximately 67 years (Nierenbergt al., 2005; Persson et al., 2006a,b), compared to approxi-ately 58 years for high-risk subjects in the present study.he time to expected development of MCI or AD is there-

ore about 9 years greater in the present study. Education

evel was not different between risk groups. The frequen-ies of past and present hypertension, hypercholesterolemia,moking and estrogen replacement were similar between riskroups (Supplemental Table).

tcog

ging 31 (2010) 1122–1131 1125

Correlation with age was in the expected positive direc-ion for Trails A and B, and in the negative direction forlock Design, Visual Reproduction Immediate and Delayed,elective Reminding Test Delayed, and Spatial Span. Edu-ation effects were observed only marginally on the Becknventory; the absence of these effects overall is attributedo the similar mean and narrow range of education in thetudy subjects (Table 1B). The age × risk interaction wasarginally significant for LM immediate and significant forR immediate (Table 1B). In the LM immediate interaction,

cores decreased with increasing age in the low-risk grouput not for the high-risk group, whereas VR immediate scoresecreased with increasing age in the high-risk group but notor the low-risk group.

Trails A was the only psychometric test of the 16 per-ormed meeting our threshold p-value of <0.01 for risk groupifference. Trails A is a visuomotor task combining visuospa-ial and motor speed components. High-risk subjects werelower on this test than low risk. High-risk subjects werelso slower on Trails B, but not significantly. The differenceetween Trails B and A is thought to produce a variableith reduced dependence on motor speed. This B minusdifference when compared between groups was not sig-

ificant, suggesting that lower performance on Trails A inigh-AD risk subjects was due to decreased motor speed andot visuospatial factors (high risk 39.9 ± 3.7 S.E.M., low risk5.7 ± 5.2 S.E.M. seconds; p = 0.5). As expected, age wasositively correlated with B minus A (p = 0.02). Spatial spanas lower in high-risk subjects but not within our cutoff forroup difference (Table 1B). In contrast, there were no differ-nces between risk groups in general mental ability (MMSE,ocabulary, BNT) or in memory measures (LM delayed, or

n the more sensitive Selective Reminding Test delayed).The Beck Inventory score was slightly higher in high-risk

ubjects but well below the cutoff of 13 for mild depres-ion (mean 7.3 in high risk and 3.2 in low risk; p = 0.05). Theower Block Design score in high-risk subjects, although only

arginally significant, appeared interesting at first becausef the association with visuospatial intelligence. However, indirect comparison of Block Design scaled score between

isk groups adjusted for age and education, the differenceas not significant (high risk 11.4 ± 3.70.4 S.E.M., low risk2.5 ± 0.5 S.E.M.; p = 0.08), tempering the potential signifi-ance of this finding.

Figs. 1–3 illustrate statistical maps rendered on the com-on FA target image in Montreal Neurologic Institute (MNI)

patial coordinates, using the contrast: FA high-AD risk lesshan FA low risk. The reverse contrast yielded only a fewcattered voxels. Maps are referenced to a standard humanhite matter atlas (Mori et al., 2005). The maps demonstrate

ower FA in the high-AD risk subjects in: (a) the inferiorongitudinal fasciculus of the temporal lobe posteriorly (par-

icularly inferior fronto-occipital fasciculus bilaterally) andingulum bundle posteriorly on the left (Fig. 1), (b) spleniumf the corpus callosum, white matter of the parahippocampalyrus (left), and right mid cingulum (Fig. 2), and (c) infe-

1126 C.D. Smith et al. / Neurobiology of Aging 31 (2010) 1122–1131

Fig. 1. Map showing inferotemporal regions of decreased FA in normal, high-AD risk subjects. The anatomic underlay used for illustration is the MNI-spaceregistered target FA image. The registered average FA skeleton is represented in green. Warmer colors on red-orange scale indicate higher t-values for FAd sagittali indicatf wheadsF

r(ipi

c

TL

L

−2−3−3−7−−−20−4

Ll

ecrease at each voxel (FAhigh risk < FAlow risk). Arrowheads in (A) mark thes indicated by single arrowheads in (C). The lineal patterns of FA decreaseasciculus/inferior fronto-occipital fasciculus (Mori et al., 2005). Double arroA adheres only to white matter regions.

ior fronto-occipital fasciculus/uncinate fasciculus anteriorly

Fig. 3). Decreased white matter FA in these specific regionss seen in the younger high-AD risk group of subjects com-ared to the older low-AD risk subjects, the reverse of thencreased FA expected with aging; therefore the difference

ousr

able 2ocations of voxels with decreased FA in high-AD risk participants

ocation Tract

27, −77, 2 Inferior fronto-occipital fa5, −78, 2 Inferior fronto-occipital fa38, −63, −8 Inferior fronto-occipital fa

9, −64, −8 Inferior fronto-occipital fa37, −55, 0 Inferior longitudinal fascic

6, −59, 5 Inferior longitudinal fascic19, −36, −7 Cingulum bundle L (poster

, −8, 32 Cingulum bundle R (mid)42, −25, −6 Inferior fronto-occipital fa15, 2, 56 Cortico-pontine tract24, 23, 8 Inferior fronto-occipital fa

6, 28, 2 Inferior fronto-occipital fa, −36, 21 Splenium, corpus callosum4, 3, −10 Subcallosal white matter

, 3, −10 Subcallosal white matter

ocations are given in Talairach coordinates in the first Column (Talairach and Touocation in the second Column (Mori et al., 2005 #67), and the nearest neighboring

slice locations on the right (B) and on the left (D). Coronal slice level in Aed by the arrowheads in (B) and (D) correspond to the inferior longitudinalin (C) point to the posterior cingulum bundle on the left. Note that decreased

annot be attributed to aging. A listing of the center locations

f clusters with ≥50 contiguous voxels (at least 0.05 cm3 vol-me) is shown in Table 2. These listings provide a conciseummary of relatively spatially coherent FA decreases fromesults in the figures, which were corrected for multiple com-

Nearest BA

sciculus L (posterior) 19sciculus R (posterior) 19sciculus L (posterior) 19, 37sciculus R (posterior) 19, 37ulus L (posterior) 37ulus R (posterior) 37ior) 27, 30

23, 24sciculus L (posterior) 21

6sciculus L (anterior) –sciculus L (anterior) –

–2525

rnoux, 1988, 1027), the corresponding tract associated with this anatomicBrodmann cortical area (BA) to the location in the third Column.

C.D. Smith et al. / Neurobiology of Aging 31 (2010) 1122–1131 1127

Fig. 2. Decreased FA in the right mid-cingulum bundle and splenium in normal, high-AD risk subjects. Display conventions are given in Fig. 1 legend. Toprow: Arrowheads in (A) and (B) indicate the mid-portion of the cingulum on the right demonstrating decreased FA. A magnified view of the region indicatedb symmeh Arrowhc or cingu

pto

4

hllaadv

ofoppi

tsmp

uSFweF2eeIa

i

y a box in the mid-sagittal slice is illustrated in B. Arrows in (A) denotesave the lineal spatial coherence of the right-sided counterpart. Bottom row:orpus callosum. Double arrows in (D) indicate decreased FA in the posteri

arisons at the significance level but did not impose a clusterhreshold. Cluster thresholding refers to comparisons basedn spatial extent of FA differences.

. Discussion

The most important finding was a clearly delineated andighly selective decrease in FA in the inferior temporalobe white matter bilaterally (inferior longitudinal fascicu-us/inferior fronto-occipital fasciculus). This decrease has

strong linear spatial character suggesting involvement ofspecific group of white matter fibers bridging the amyg-

ala/hippocampal head region anteriorly and the ventralisual association areas posteriorly (Fig. 1).

We also found significant FA decreases in the spleniumf the corpus callosum, white matter of the cingulum, andronto-occipital fasciculus anteriorly. Downward curvature

f the involved fronto-occipital fasciculus shown in Fig. 3,articularly well seen on the left, is characteristic of thisathway, which parallels the uncinate fasciculus in part ofts course. In general our findings were more robust in the

SmcZ

tric cingulate areas with decreased FA on the left, but this region does noteads in (C) and (D) mark the region of decreased FA in the splenium of thelate bundle on the left.

emporal and frontal regions than in parietal lobes, and morepatially coherent in the ventral visual association regions,edial temporal lobe, cingulum and frontal lobe than in the

arietal lobe or occipital lobe.Previous studies using DTI in aging and MCI/AD have

sed several different acquisition and analysis strategies.tudies of normal aging have shown consistently decreasedA and related measures with age, particularly in the frontalhite matter with relative sparing of the temporal, pari-

tal and occipital regions (Pfefferbaum and Sullivan, 2003;urutani et al., 2005; Charlton et al., 2006; Persson et al.,006a,b; Sullivan et al., 2006; Grieve et al., 2007; Kochunovt al., 2007; Madden et al., 2007; Taylor et al., 2007; Abet al., 2008; Hsu et al., 2008; Hugenschmidt et al., 2008).ncreases of FA with age have not been reliably observed inny region.

In contrast, only a few studies have reported selective find-ngs in the frontal lobe in MCI or AD (Choi et al., 2005;

ydykova et al., 2007). The remainder have demonstratedore consistent overlapping findings of decreased FA in the

ingulum bundle (Fellgiebel et al., 2007; Teipel et al., 2007;hang et al., 2007), temporal lobe generally (Xie et al., 2006;

1128 C.D. Smith et al. / Neurobiology of Aging 31 (2010) 1122–1131

Fig. 3. Map showing anterior regions of decreased FA in normal, high-AD risk subjects. Display conventions are given in Fig. 1 legend. Arrowheads in Am al sliceo erior frod

HtHmerF(pte(owSteetfw

dc

aotoaobproacsFo

iasp

ark the sagittal slice locations on the right (B) and on the left (D). Coronf FA decrease indicated by the arrows in (B) and (D) correspond to the infecrease region, best seen on the left in (B), is typical of this tract.

uang et al., 2007; Stahl et al., 2007), parietal lobe and pos-erior cingulate (Medina et al., 2006; Firbank et al., 2007;uang et al., 2007; Ding et al., 2008; Rose et al., 2008),edial temporal white matter (Fellgiebel et al., 2007; Huang

t al., 2007; Yasmin et al., 2008), inferior temporal lobe, supe-ior and inferior fronto-occipital fasciculi (Xie et al., 2006;ellgiebel et al., 2007), and splenium of the corpus callosumStahl et al., 2007). Findings in MCI tend to overlap with AD,articularly in posterior areas. These studies have differed inhat some selected a restricted target region of a priori interest,.g., the medial temporal white matter or cingulate bundlesZhang et al., 2007; Ding et al., 2008; Yasmin et al., 2008),thers picked several general regions of interest within lobarhite matter (Fellgiebel et al., 2004, 2007; Choi et al., 2005;tahl et al., 2007), whereas still others used a voxel-based

echnique (Medina et al., 2006; Rose et al., 2006, 2008; Xiet al., 2006; Firbank et al., 2007; Huang et al., 2007; Sydykovat al., 2007; Teipel et al., 2007). For comparison the analysisechnique we used in this study is voxel-based, and there-ore unbiased with respect to a specific hypothesis regarding

here regional FA differences might be found.It is important to make the distinction between studies

escribed above that examine subjects with symptoms of alinically diagnosed cognitive disorder, e.g., MCI and AD,

td2a

level in A is indicated by single arrowheads in (C). The coherent patternsnto-occipital fasciculus(Mori et al., 2005). Downward curvature of the FA

nd studies of normal persons without symptoms or deficitsn psychometric testing who may later develop these condi-ions. There are many fewer investigations like the presentne that examine the latter issue. Brain FA alterations thatre associated with the presence of a diagnosed clinical dis-rder due to an underlying pathology may be assumed toe different in degree, or perhaps kind, from alterations thatredict later development of those symptoms. But it is cur-ently unknown whether such findings would predict MCIr AD. This is implied but not demonstrated in studies oft-risk normal persons. The first step in the validation of theurrent findings is to follow these currently (c. 2008) normaltudy subjects longitudinally to show whether the observedA alterations are predictive of later clinically diagnosed MCIr AD.

The voxel-based studies have not been entirely consistentn localization of FA decreases in AD, e.g., some find frontalnd internal capsule alterations, but the overall pattern acrosstudies suggests alterations in the white matter of the parahip-ocampal gyrus, inferior longitudinal fasciculus, splenium of

he corpus callosum, parietal white matter, and cingulum bun-le (Medina et al., 2006; Rose et al., 2006, 2008; Xie et al.,006; Firbank et al., 2007; Huang et al., 2007; Sydykova etl., 2007; Teipel et al., 2007).

ogy of A

FmoginfRrapscctoiFapssfia

nrpdbhfaaao

rrrtdiari

ropbbFm

hib

iTscai2cefih1oFss

wbttit

thpdfArstMif

C

A

36660. We thank Agnes Bognar, R.T., Kim Wilson, B.A.,

C.D. Smith et al. / Neurobiol

Of particular relevance, a recent study reported reducedA in the fornix and left frontal WM in a young group ofildly impaired (CDR = 0.5) subjects (mean age = 35 years

ld) at risk for early onset-AD by virtue of fully penetrantenetic alterations (Ringman et al., 2007). Only two stud-es have explored WM integrity differences in cognitivelyormal subjects at risk for the more common late-onsetorm of AD due to possession of an APOE4 allele. UsingOI or SPM techniques, respectively, these studies observed

educed FA in the left parahippocampal WM (Nierenberg etl., 2005), or the splenium of the corpus callosum, and leftosterior hippocampus (Persson et al., 2006a,b). Our resultshow some overlap with findings in these studies, specifi-ally, the findings of decreased FA in the splenium of theorpus callosum and inferomedial temporal lobes. The loca-ion of the present FA decreases in presymptomatic carriersf AD overlap with previous FA reductions observed in ADn the inferior fronto-occipital fasciculi (Xie et al., 2006;ellgiebel et al., 2007), and cingulum bundle (Fellgiebel etl., 2007; Teipel et al., 2007; Zhang et al., 2007). The termresymptomatic in this context is a group characterization andhould not be interpreted to mean that each individual subjecthould be considered presymtpomatic. These findings there-ore suggest that AD-related FA declines are already presentn groups of individuals at high risk for AD, even when theyre thinking, behaving and functioning normally.

The strengths of our study are the confirmation of cog-itive normality and test performance equivalence betweenisk groups using comprehensive psychometric testing of allarticipants, rigorous exclusion of confounding neurologicisease, reduction of FA variance due to gender and cere-ral language representation by exclusive recruitment of rightanded women (Hsu et al., 2008), exclusion of potential con-ounds due to smoking, hypertension, hypercholesterolemiand estrogen replacement, a consistent imaging protocolcross all subjects using a 3T MRI instrument, and use ofn advanced, validated and regionally unbiased DTI method-logy.

Several cautions should be kept in mind in interpreting ouresults. Despite equivalent cognitive performance betweenisk groups, the low-risk subjects were older than the high-ADisk. We used age as a regressor in all our analyses, but sta-istical adjustments are not perfect. Because FA is known toecrease – not increase – with age, a finding of an FA decreasen the younger high-risk group is the reverse of expectation onn age basis, so we believe the difference in age strengthensather than weakens our findings. Nonetheless, the differencen age between groups should be noted.

Another more important caveat is in interpretation ofesults. We can only infer that the alterations we havebserved are related to the risk of AD. The study partici-ants were selected to differ in no respect other than AD risk,

ut it is possible the groups differed in some other unknownut important way for which we could not adjust or control.urthermore, we do not know that presence of these whiteatter alterations is due to underlying AD pathology, a basic

Ltaa

ging 31 (2010) 1122–1131 1129

ypothesis of the study, because we have no independent ver-fication of its degree or presence, e.g., by autopsy or brainiopsy.

The issue of the functional significance of the FA declinesn presymptomatic groups also needs further investigation.his is a tricky issue because presymptomatic high-riskubjects display cognitive function that is, by definition,omparable to that of low-risk subjects. However, there isccumulating evidence that FA is of functional significancen young healthy subjects (Madden et al., 2004; Tuch et al.,005; Gold et al., 2007), and that age-related FA declinesontribute to some cognitive performance declines (Maddent al., 2004; Gold et al., 2008). The view that FA declines areunctionally relevant is also in-line with our previous find-ng of decreased ventral temporal activation in asymptomaticigh-AD risk women in an fMRI naming task (Smith et al.,999). The fMRI under-recruitments in this previous studyf subjects were in regions adjacent to the postero-inferiorA alterations observed in the present study. Although bothtudies were performed using similar subjects selected by theame criteria, the specific individuals studied were different.

Finally, like other studies of presymptomatic participants,e cannot yet claim a role for altered FA in predicting ADecause the participants have not been followed long enougho observe even the earliest symptoms for validation. Longi-udinal follow up of the participants described in this studys in place at our center and should yield relevant results inhe future.

In summary this report represents further progress towardhe goal of identifying the brain at risk of AD using DTI. Weave identified restricted regions of the inferomedial tem-oral lobes bilaterally, and other regions, that demonstrateecreased FA in normal high-AD risk subjects, that may inuture serve as a biomarker for the presence of asymptomaticD pathology. What we have observed here are specific

egions of decreased FA in high-AD risk subjects. The nexttep in validating this work is to follow subjects longitudinallyo show whether the observed changes are truly predictive of

CI or AD, and with what accuracy. Therefore, this methods promising but needs further study before it could be usedor that purpose in trials of preventive agents in AD.

onflict of interest

None.

cknowledgements

This study was supported by NINDS Grant R01 NS-

uAnn Hamon, B.A., Dorothy Ross, B.A., and Barbara Mar-in, B.A. for their invaluable assistance in recruiting, scanningnd testing the participants, and for treating them profession-lly and so well.

1 ogy of A

A

bn

R

A

B

B

B

C

C

D

E

F

F

F

F

G

G

G

H

H

H

J

K

M

M

M

M

N

P

P

P

R

R

R

R

S

S

130 C.D. Smith et al. / Neurobiol

ppendix A. Supplementary data

Supplementary data associated with this article cane found, in the online version, at doi:10.1016/j.eurobiolaging.2008.08.006.

eferences

be, O., Yamasue, H., Aoki, S., Suga, M., Yamada, H., Kasai, K., Masutani,Y., Kato, N., Kato, N., Ohtomo, K., 2008. Aging in the CNS: comparisonof gray/white matter volume and diffusion tensor data. Neurobiol. Aging29, 102–116.

ennett, D.A., Schneider, J.A., Arvanitakis, Z., Kelly, J.F., Aggarwal, N.T.,Shah, R.C., Wilson, R.S., 2006. Neuropathology of older persons withoutcognitive impairment from two community-based studies. Neurology 66,1837–1844.

ookstein, F.L., 2001. “Voxel-based morphometry” should not be used withimperfectly registered images. Neuroimage 14, 1454–1462.

raak, H., Braak, E., 1997. Frequency of stages of Alzheimer-related lesionsof different age categories. Neurobiol. Aging 18, 351–357.

harlton, R.A., Barrick, T.R., McIntyre, D.J., Shen, Y., O’Sullivan, M.,Howe, F.A., Clark, C.A., Morris, R.G., Markus, H.S., 2006. White matterdamage on diffusion tensor imaging correlates with age-related cognitivedecline. Neurology 66, 217–222.

hoi, S.J., Lim, K.O., Monteiro, I., Reisberg, B., 2005. Diffusion tensorimaging of frontal white matter microstructure in early Alzheimer’sdisease: a preliminary study. J. Geriatr. Psychiatry Neurol. 18, 12–19.

ing, B., Chen, K.M., Ling, H.W., Zhang, H., Chai, W.M., Li, X.,Wang, T., 2008. Diffusion tensor imaging correlates with protonmagnetic resonance spectroscopy in posterior cingulate region ofpatients with Alzheimer’s disease. Dement. Geriatr. Cogn. Disord. 25,218–225.

nglund, E., Brun, A., Alling, C., 1988. White matter changes in dementia ofAlzheimer’s type. Biochemical and neuropathological correlates. Brain111 (Part 6), 1425–1439.

ellgiebel, A., Schermuly, I., Gerhard, A., Keller, I., Albrecht, J., Weib-rich, C., Muller, M.J., Stoeter, P., 2007. Functional relevant loss of longassociation fibre tracts integrity in early Alzheimer’s disease. Neuropsy-chologia.

ellgiebel, A., Wille, P., Muller, M.J., Winterer, G., Scheurich, A., Vucurevic,G., Schmidt, L.G., Stoeter, P., 2004. Ultrastructural hippocampal andwhite matter alterations in mild cognitive impairment: a diffusion tensorimaging study. Dement. Geriatr. Cogn. Disord. 18, 101–108.

irbank, M.J., Blamire, A.M., Krishnan, M.S., Teodorczuk, A., English, P.,Gholkar, A., Harrison, R., O’Brien, J.T., 2007. Atrophy is associatedwith posterior cingulate white matter disruption in dementia with Lewybodies and Alzheimer’s disease. Neuroimage 36, 1–7.

urutani, K., Harada, M., Minato, M., Morita, N., Nishitani, H., 2005.Regional changes of fractional anisotropy with normal aging using sta-tistical parametric mapping (SPM). J. Med. Invest. 52, 186–190.

old, B.T., Powell, D.K., Xuan, L., Jiang, Y., Hardy, P.A., 2007. Speed oflexical decision correlates with diffusion anisotropy in left parietal andfrontal white matter: evidence from diffusion tensor imaging. Neuropsy-chologia 45, 2439–2446.

old, B.T., Powell, D.K., Xuan, L., Jicha, G.A., Smith, C.D., 2008. Age-related slowing of task switching is associated with decreased integrityof frontoparietal white matter. Neurobiol. Aging. [Epub ahead of print].

rieve, S.M., Williams, L.M., Paul, R.H., Clark, C.R., Gordon, E., 2007.Cognitive aging, executive function, and fractional anisotropy: a diffu-

sion tensor MR imaging study. Am. J. Neuroradiol. 28, 226–235.

su, J.L., Leemans, A., Bai, C.H., Lee, C.H., Tsai, Y.F., Chiu, H.C., Chen,W.H., 2008. Gender differences and age-related white matter changesof the human brain: a diffusion tensor imaging study. Neuroimage 39,566–577.

S

ging 31 (2010) 1122–1131

uang, J., Friedland, R.P., Auchus, A.P., 2007. Diffusion tensor imaging ofnormal-appearing white matter in mild cognitive impairment and earlyAlzheimer disease: preliminary evidence of axonal degeneration in thetemporal lobe. Am. J. Neuroradiol. 28, 1943–1948.

ugenschmidt, C.E., Peiffer, A.M., Kraft, R.A., Casanova, R., Deibler, A.R.,Burdette, J.H., Maldjian, J.A., Laurienti, P.J., 2008. Relating imagingindices of white matter integrity and volume in healthy older adults.Cereb Cortex. 18, 433–442.

ones, D.K., Symms, M.R., Cercignani, M., Howard, R.J., 2005. The effect offilter size on VBM analyses of DT-MRI data. Neuroimage 26, 546–554.

ochunov, P., Thompson, P.M., Lancaster, J.L., Bartzokis, G., Smith, S.,Coyle, T., Royall, D.R., Laird, A., Fox, P.T., 2007. Relationship betweenwhite matter fractional anisotropy and other indices of cerebral health innormal aging: tract-based spatial statistics study of aging. Neuroimage35, 478–487.

adden, D.J., Spaniol, J., Whiting, W.L., Bucur, B., Provenzale, J.M.,Cabeza, R., White, L.E., Huettel, S.A., 2007. Adult age differences inthe functional neuroanatomy of visual attention: a combined fMRI andDTI study. Neurobiol. Aging 28, 459–476.

adden, D.J., Whiting, W.L., Huettel, S.A., White, L.E., MacFall, J.R.,Provenzale, J.M., 2004. Diffusion tensor imaging of adult age differ-ences in cerebral white matter: relation to response time. Neuroimage21, 1174–1181.

edina, D., DeToledo-Morrell, L., Urresta, F., Gabrieli, J.D., Moseley, M.,Fleischman, D., Bennett, D.A., Leurgans, S., Turner, D.A., Stebbins,G.T., 2006. White matter changes in mild cognitive impairment and AD:a diffusion tensor imaging study. Neurobiol. Aging 27, 663–672.

ori, S., Wakana, S., Nagae-Poetscher, L.M., van Zijl, P.C.M., 2005. MRIAtlas of Human White Matter. Elsevier, Amsterdam.

ierenberg, J., Pomara, N., Hoptman, M.J., Sidtis, J.J., Ardekani, B.A., Lim,K.O., 2005. Abnormal white matter integrity in healthy apolipoproteinE epsilon4 carriers. Neuroreport 16, 1369–1372.

ersson, J., Lind, J., Larsson, A., Ingvar, M., Cruts, M., Van Broeckhoven,C., Adolfsson, R., Nilsson, L.G., Nyberg, L., 2006a. Altered brain whitematter integrity in healthy carriers of the APOE epsilon4 allele: a riskfor AD? Neurology 66, 1029–1033.

ersson, J., Nyberg, L., Lind, J., Larsson, A., Nilsson, L.G., Ingvar, M.,Buckner, R.L., 2006b. Structure–function correlates of cognitive declinein aging. Cereb Cortex. 16, 907–915.

fefferbaum, A., Sullivan, E.V., 2003. Increased brain white matter diffusiv-ity in normal adult aging: relationship to anisotropy and partial voluming.Magn. Reson. Med. 49, 953–961.

ingman, J.M., O’Neill, J., Geschwind, D., Medina, L., Apostolova, L.G.,Rodriguez, Y., Schaffer, B., Varpetian, A., Tseng, B., Ortiz, F., Fitten,J., Cummings, J.L., Bartzokis, G., 2007. Diffusion tensor imaging inpreclinical and presymptomatic carriers of familial Alzheimer’s diseasemutations. Brain 130, 1767–1776.

ose, S.E., Janke, A.L., Chalk, J.B., 2008. Gray and white matter changes inAlzheimer’s disease: a diffusion tensor imaging study. J. Magn. Reson.Imaging 27, 20–26.

ose, S.E., McMahon, K.L., Janke, A.L., O’Dowd, B., de Zubicaray, G.,Strudwick, M.W., Chalk, J.B., 2006. Diffusion indices on magnetic res-onance imaging and neuropsychological performance in amnestic mildcognitive impairment. J. Neurol. Neurosurg. Psychiatry 77, 1122–1128.

ueckert, D., Sonoda, L.I., Hayes, C., Hill, D.L., Leach, M.O., Hawkes, D.J.,1999. Nonrigid registration using free-form deformations: application tobreast MR images. IEEE Trans. Med. Imaging 18, 712–721.

chneider, J.A., Arvanitakis, Z., Bang, W., Bennett, D.A., 2007. Mixed brainpathologies account for most dementia cases in community-dwellingolder persons. Neurology 69, 2197–2204.

mith, C.D., 2007. Mild cognitive impairment is too late: the case forpresymtomatic detection and treatment of Alzheimer’s disease. Cogn.

Sci. 3, 127–177.

mith, C.D., Andersen, A.H., Kryscio, R.J., Schmitt, F.A., Kindy, M.S.,Blonder, L.X., Avison, M.J., 1999. Altered brain activation in cognitivelyintact individuals at high risk for Alzheimer’s disease. Neurology 53,1391–1396.

ogy of A

S

S

S

S

S

T

T

T

T

X

Y

C.D. Smith et al. / Neurobiol

mith, C.D., Andersen, A.H., Kryscio, R.J., Schmitt, F.A., Kindy, M.S.,Blonder, L.X., Avison, M.J., 2002. Women at risk for AD show increasedparietal activation during a fluency task. Neurology 58, 1197–1202.

mith, S.M., Jenkinson, M., Johansen-Berg, H., Rueckert, D., Nichols, T.E.,Mackay, C.E., Watkins, K.E., Ciccarelli, O., Cader, M.Z., Matthews,P.M., Behrens, T.E., 2006. Tract-based spatial statistics: voxelwise anal-ysis of multi-subject diffusion data. Neuroimage 31, 1487–1505.

tahl, R., Dietrich, O., Teipel, S.J., Hampel, H., Reiser, M.F., Schoenberg,S.O., 2007. White matter damage in Alzheimer disease and mild cog-nitive impairment: assessment with diffusion-tensor MR imaging andparallel imaging techniques. Radiology 243, 483–492.

ullivan, E.V., Adalsteinsson, E., Pfefferbaum, A., 2006. Selective age-related degradation of anterior callosal fiber bundles quantified in vivowith fiber tracking. Cereb Cortex. 16, 1030–1039.

ydykova, D., Stahl, R., Dietrich, O., Ewers, M., Reiser, M.F., Schoenberg,S.O., Moller, H.J., Hampel, H., Teipel, S.J., 2007. Fiber connectionsbetween the cerebral cortex and the corpus callosum in Alzheimer’sdisease: a diffusion tensor imaging and voxel-based morphometry study.

Cereb Cortex. 17, 2276–2282.

alairach, J., Tournoux, P., 1988. Co-Planar Stereotaxic Atlas of the HumanBrain. Thieme Medical Publishers, New York.

aylor, W.D., Bae, J.N., MacFall, J.R., Payne, M.E., Provenzale, J.M., Stef-fens, D.C., Krishnan, K.R., 2007. Widespread effects of hyperintense

Z

ging 31 (2010) 1122–1131 1131

lesions on cerebral white matter structure. Am. J. Roentgenol. 188,1695–1704.

eipel, S.J., Stahl, R., Dietrich, O., Schoenberg, S.O., Perneczky, R., Bokde,A.L., Reiser, M.F., Moller, H.J., Hampel, H., 2007. Multivariate networkanalysis of fiber tract integrity in Alzheimer’s disease. Neuroimage 34,985–995.

uch, D.S., Salat, D.H., Wisco, J.J., Zaleta, A.K., Hevelone, N.D., Rosas,H.D., 2005. Choice reaction time performance correlates with diffusionanisotropy in white matter pathways supporting visuospatial attention.Proc. Natl. Acad. Sci. U.S.A. 102, 12212–12217.

ie, S., Xiao, J.X., Gong, G.L., Zang, Y.F., Wang, Y.H., Wu, H.K., Jiang,X.X., 2006. Voxel-based detection of white matter abnormalities in mildAlzheimer disease. Neurology 66, 1845–1849.

asmin, H., Nakata, Y., Aoki, S., Abe, O., Sato, N., Nemoto, K., Arima, K.,Furuta, N., Uno, M., Hirai, S., Masutani, Y., Ohtomo, K., 2008. Diffusionabnormalities of the uncinate fasciculus in Alzheimer’s disease: diffusiontensor tract-specific analysis using a new method to measure the core ofthe tract. Neuroradiology.

hang, Y., Schuff, N., Jahng, G.H., Bayne, W., Mori, S., Schad, L., Mueller,S., Du, A.T., Kramer, J.H., Yaffe, K., Chui, H., Jagust, W.J., Miller,B.L., Weiner, M.W., 2007. Diffusion tensor imaging of cingulum fibersin mild cognitive impairment and Alzheimer disease. Neurology 68,13–19.