lable at ScienceDirect

Neurobiology of Aging 35 (2014) 1473e1482

Contents lists avai

Neurobiology of Aging

journal homepage: www.elsevier .com/locate/neuaging

Genetic and neuroanatomic associations in sporadic frontotemporallobar degeneration

Corey T. McMillan a,b,*, Jon B. Toledo c, Brian B. Avants d, Philip A. Cook d,Elisabeth M. Wood c, Eunran Suh c, David J. Irwin a,b,c, John Powers a,b,Christopher Olm a,b, Lauren Elman a, Leo McCluskey a, Gerard D. Schellenberg c,Virginia M.-Y. Lee c, John Q. Trojanowski c, Vivianna M. Van Deerlin c,Murray Grossman a,b

aDepartment of Neurology, University of Pennsylvania Perelman School of Medicine, Philadelphia, PA, USAb Penn Frontotemporal Degeneration Center, University of Pennsylvania Perelman School of Medicine, Philadelphia, PA, USAcDepartment of Laboratory and Pathology Medicine, Center for Neurodegenerative Disease Research, University of Pennsylvania Perelman School ofMedicine, Philadelphia, PA, USAdDepartment of Radiology, Penn Image Computing and Science Laboratory, University of Pennsylvania Perelman School of Medicine, Philadelphia, PA,USA

a r t i c l e i n f o

Article history:Received 7 November 2013Received in revised form 18 November 2013Accepted 27 November 2013Available online 2 December 2013

Keywords:Frontotemporal lobar degenerationNeuroimagingGeneticsBiomarkers

* Corresponding author at: Department of Neurologcine, University of Pennsylvania, 3400 Spruce Street, 319104, USA. Tel.: þ1 215 349 5863; fax: þ1 215 349 8

E-mail address: mcmillac@mail.med.upenn.edu (C

0197-4580/$ e see front matter � 2014 Elsevier Inc. Ahttp://dx.doi.org/10.1016/j.neurobiolaging.2013.11.029

a b s t r a c t

Genome-wide association studies have identified single nucleotide polymorphisms (SNPs) that aresensitive for tau or TDP-43 pathology in frontotemporal lobar degeneration (FTLD). Neuroimaginganalyses have revealed distinct distributions of disease in FTLD patients with genetic mutations.However, genetic influences on neuroanatomic structure in sporadic FTLD have not been assessed. In thisreport, we use novel multivariate tools, Eigenanatomy, and sparse canonical correlation analysis toidentify associations between SNPs and neuroanatomic structure in sporadic FTLD. Magnetic resonanceimaging analyses revealed that rs8070723 (MAPT) was associated with gray matter variance in thetemporal cortex. Diffusion tensor imaging analyses revealed that rs1768208 (MOBP), rs646776 (nearSORT1), and rs5848 (PGRN) were associated with white matter variance in the midbrain and superiorlongitudinal fasciculus. In an independent autopsy series, we observed that rs8070723 and rs1768208conferred significant risk of tau pathology relative to TDP-43, and rs646776 conferred increased risk ofTDP-43 pathology relative to tau. Identified brain regions and SNPs may help provide an in vivo screenfor underlying pathology in FTLD and contribute to our understanding of sporadic FTLD.

� 2014 Elsevier Inc. All rights reserved.

1. Introduction

There is increasing neuroimaging evidence that genetic factorsinfluence graymatter (GM) andwhitematter (WM) neuroanatomyin Alzheimer’s disease (AD; Jahanshad et al., 2013; Shen et al.,2010). Genome-wide association (GWA) studies of autopsy-confirmed neurodegenerative disease cases have related severalsingle nucleotide polymorphisms (SNPs) to the risk of accumu-lating specific types of histopathologic abnormality. In this study,

y, Perelman School of Medi-West Gates, Philadelphia, PA464..T. McMillan).

ll rights reserved.

we combine our knowledge of the genetic basis for frontotemporallobar degeneration (FTLD; DeJesus-Hernandez et al., 2011;Höglinger et al., 2011; Renton et al., 2011; Van Deerlin et al.,2010) with neuroimaging in an effort to identify novel geneticand neuroanatomic associations that may be used to improvethe diagnostic accuracy of FTLD. We additionally introduce a data-driven technique for investigating genetic and neuroanatomicassociations that provides a novel approach for biomarkerdiscovery.

FTLD is a common cause of early-onset neurodegenerative de-mentia. Approximately 20% of familial FTLD patients have agenetically identified mutation (Wood et al., 2013). Autopsystudies of FTLD have demonstrated that the vast majority of pa-tients have either tau inclusions (FTLD-tau) or a TDP-43 protein-opathy (FTLD-TDP; Mackenzie et al., 2010). SNPs have been

C.T. McMillan et al. / Neurobiology of Aging 35 (2014) 1473e14821474

identified through case-control GWA or other association studiesof autopsy-confirmed FTLD-tau (Höglinger et al., 2011) and FTLD-TDP (Renton et al., 2011; Van Deerlin et al., 2010) but have notbeen evaluated comparatively in FTLD-tau relative to FTLD-TDP.Neuroimaging studies suggest distinct neuroanatomic distribu-tions of disease in FTLD-tau and FTLD-TDP (Seltman andMatthews,2012; Whitwell et al., 2011b), and small sample studies haveshown distinct anatomic distributions of GM and WM diseaseassociated with genetic mutations (Rohrer et al., 2010; Whitwellet al., 2012). Diffusion tensor imaging (DTI) studies of WM ach-ieve high sensitivity and specificity for predicting FTLD-tau orFTLD-TDP in patients with known pathology or genetic mutations,and these findings were validated in a detailed neuropathologicexamination (McMillan et al., 2013). This observation is alsoconsistent with previous neuropathologic reports suggesting thatFTLD-tau has relatively increased WM disease compared withFTLD-TDP (Forman et al., 2002; Geser et al., 2009).

In this study, we combine SNPswith GMandWMneuroimagingto evaluate the hypothesis that genetic risk factors, or SNPs, may bereflected in differential brain morphology of sporadic FTLD-tau orFTLD-TDP. Although traditional approaches to biomarker discoverytypically involve retrospective studies of gold standard autopsy-proven cases, we use a data-driven prospective approach thattakes advantage of high-powered, multivariate statistics (see Fig. 1for a schematic diagram). Our novel approach for biomarker dis-covery integrates neuroimaging and genetic markers in FTLD using2 approaches. Eigenanatomy uses dimensionality reduction toidentify anatomically constrained correlated voxels that accountfor the greatest variance in the entire data set and thus minimizesmultiple comparison problems that are common in voxelwiseneuroimaging studies (Avants et al., 2012; McMillan et al., 2013).We also use sparse canonical correlation analysis (SCCAN) for themultivariate integration of imaging and genetics by identifyingcorrelations across independent matrices of data (Avants et al.,2010). We first evaluate the hypothesis that GM and WM neuro-anatomic structure are related to genetic variation in sporadicFTLD patients. We then evaluate the hypothesis that SNPs associ-ated with neuroanatomic structure confer risk for a specifichistopathologic subtype of FTLD pathology in a large independent,autopsy-confirmed cohort of sporadic FTLD.

2. Methods

2.1. Neuroimaging participants

Ninety-two patients were recruited from the Penn Fronto-temporal Degeneration Center at the University of Pennsylvaniaand diagnosed with a FTLD-spectrum neurodegenerative diseaseby a board-certified neurologist using published criteria (seeSupplementary Table 1 for clinical phenotypes). The patient cohortcomprised 37 women and 55 menwho had an overall mean age of63.20 years (SD ¼ 8.51), mean disease duration of 4.09 years (SD ¼2.54), and mean education of 15.43 years (SD ¼ 2.95). All patientsand their caregivers participated in an informed consent proce-dure approved by University of Pennsylvania Institutional ReviewBoard.

All patients selected for this study were screened for researchparticipation using an autopsy-validated cerebrospinal fluid ratioof total-tau to beta-amyloid<0.34, which has been cross-validatedacross 2 independent autopsy series and achieves 95.5% accuracyof screening FTLD and AD (Irwin et al., 2012b). To investigatesporadic FTLD, we additionally excluded patients who had aknown genetic mutation that has been associated with FTLD-TDP,including GRN (Baker et al., 2006) and C9orf72 expansions(DeJesus-Hernandez et al., 2011; Renton et al., 2011), or with an

FTLD-tau associated MAPT mutation (Hutton et al., 1998). Wefurther classified our cases using a previously published pedigreeclassification criteria (Wood et al., 2013): cases with a “medium”

(n ¼ 9) or low (n ¼ 7) family history were negative for C9orf72expansions, GRN, and MAPT and have a <12% chance of having amutation detected (Wood et al., 2013); only 3 of our cases had a“high” family history and were negative when screened for 43genetic mutations previously associated with neurodegenerativediseases; the remaining cases were either rated as “apparentsporadic” or had too small of a family to accurately determinefamily history. By omitting cases with genetic mutations, we alsominimized overlap of cases previously reported in DTI and GManalyses of individuals with genetic or autopsy-confirmed FTLD(McMillan et al., 2013; only 3 autopsy cases from the previousreport were included in the current study: 1 corticobasal degen-eration [CBD]; 1 FTLD-TDP, and 1 FTLD-amyotrophic lateral scle-rosis [ALS]).

2.2. Independent autopsy series

We queried the Penn Brain Bank for autopsy samples that hada primary neuropathologic diagnosis of FTLD-tau, including pro-gressive supranuclear palsy (PSP), CBD, Pick disease, and argyr-ophilic grain disease or a diagnosis of FTLD-TDP, including FTLDwith TDP-43 inclusions or ALS. Neuropathologic diagnoses wereestablished according to consensus criteria (Mackenzie et al., 2010)by an expert neuropathologist (JQT) using immunohistochemistrywith established monoclonal antibodies specific for pathogenictau (mAb PHF-1; Otvos et al., 1994) and TDP-43 (mAbs p409/410 or171; Lippa et al., 2009; Neumann et al., 2009). Patients who wereincluded in the neuroimaging analysis were excluded from theindependent autopsy series analysis. We further excluded caseswith a secondary neuropathologic diagnosis (e.g., AD, vasculardisease) or a known FTLD genetic mutation: all FTLD-tau caseswere screened for MAPT mutations; all FTLD-TDP patients werescreened for GRNmutations and a C9orf72 expansion. This resultedin 153 sporadic FTLD-spectrum patients, FTLD-tau (n ¼ 62) andFTLD-TDP (n ¼91; see Supplementary Table 2).

2.3. Genetic analysis

We selected 21 SNPs from a custom-designed Pan-Neurode-generative Disease-oriented Risk Allele panel (PANDoRAVersion 1;Table 1) previously associated with FTLD-TDP or FTLD-tau in case-control GWA studies (Carrasquillo et al., 2010; Höglinger et al.,2011; Van Deerlin et al., 2010) or previously implicated in FTLD(Rademakers et al., 2008, 2005). The panel was designed usingMassARRAYAssay design software in 2multiplex reactions with 27and 24 SNV respectively. See Supplementary Data for detailedgenotypingmethods. Each SNPwas coded using an additivemodel,where 0¼ homozygous for the nonrisk allele, 1¼ heterozygous forthe risk allele, and 2 ¼ homozygous for the risk allele. “Risk allele”refers to the allele previously associated with disease risk in pre-vious case-control studies.

2.4. Neuroimaging analysis

High-resolution volumetric (1 mm3) magnetic resonance im-aging volumes and diffusion-weighted images were acquired andpreprocessed using a previously described pipeline with ANTssoftware (see Supplementary Data for details; Avants et al., 2011;McMillan et al., 2013). To analyze GM density and fractionalanisotropy (FA) of WM, we used Eigenanatomy (available for freedownload in ANTs; https://github.com/stnava/sccan; Avants et al.,2012; McMillan et al., 2013). Eigenanatomy involves identifying

Fig. 1. Schematic illustration of sparse canonical correlation analysis approach. (A) All individual patients’ volumetric magnetic resonance imaging scans are represented in a matrixcontaining all voxels (M) � all patients (N). Colors in the matrix refer to gray matter (GM) density, and it can be observed that there are different bands of density observed incolumns. (B) Each GM voxel is correlated with each other GM voxel to generate a correlation matrix. Scatterplots illustrate correlations between voxels (Voxel1 w Voxel2.). Forillustration purposes, we displayed a 1000 � 1000 voxel matrix (center) and then a 10� zoom of a subcomponent of the matrix (right). (C) Clusters of anatomically constrainedcorrelated voxels are identified using Eigenanatomy, a sparse singular value decomposition method that identifies volumes of interest (VOI) that account for the most variance inthe brain. (D) Represents a matrix of patients (N) � 21 single nucleotide polymorphisms (SNPs) used in this analysis, and colors refer to 0-2 risk allele copies. (E) Eigenanatomy VOIsand the SNP matrix are analyzed using sparse canonical correlation analysis, in which each matrix is correlated with one another, and a sparsity constraint is applied to push lowermodel weights to zero and thus remove from the overall model. Abbreviation: Coef, coeffecient.

C.T. McMillan et al. / Neurobiology of Aging 35 (2014) 1473e1482 1475

volumes of interest (VOIs) composed of correlated voxels thatmaximally account for the greatest variance in the entire data set(see Fig. 1B). By reducing the dimensionality of the data from over1 M voxels to a much smaller number of Eigenanatomy VOIs, wecan perform high-powered statistics. In the current study, weidentified 20 VOIs for each modality (GM and FA) that in totalaccounted for >95% of the variance in the data set of each mo-dality. To identify these VOIs, all normalized images for each

modality were first transformed into a number-of-subjects (N) bynumber-of-voxels matrix in which voxels were selected to liewithin an explicit mask (see Fig. 1A). For GM and FA, we used athreshold of �0.4 to define our mask. Sparse singular valuedecomposition was then used to identify the first sparse eigen-vectors from the data matrix that account for the greatestamount of variance, the second sparse eigenvectors that accountfor the second most amount of variance, and so on. The ANTs

Table 1SNPs previously associated with FTLD-tau or FTLD-TDP included in the neuroimaging analysis and their risk allele frequencies

Disorder Ch SNP Gene Risk allele Risk allele frequency

FTLD-tau 17 rs1052553 MAPT G 18.51% (30)FTLD-tau (PSP) 12 rs11568563 SLCO1A2 G 9.52% (16)FTLD-tau (PSP) 6 rs12203592 IRF4 T 22.58% (35)FTLD-tau (PSP) 1 rs1411478 STX6 A 24.11% (34)FTLD-tau (PSP) 3 rs1768208 MOBP T 45.83% (66)FTLD-tau (PSP) 10 rs2142991 BMS1 C 20.25% (32)FTLD-tau (PSP) 17 rs242557 MAPT A 31.34% (36)FTLD-tau (PSP) 6 rs2493013 EXOC2 A 8.97% (14)FTLD-tau (PSP) 2 rs6547705 CD8B G 23.57% (37)FTLD-tau (PSP) 1 rs6687758 Intergenic G 25.16% (39)FTLD-TAU (PSP) FTLD-tau (PSP) 4 rs6852535 IL2/IL21 A 16.00% (24)FTLD-tau (PSP) 2 rs7571971 CD8B/EIF2AK3 T 26.45% (41)FTLD-tau (PSP) 17 rs8070723 MAPT A 79.75% (130)Tau 2 rs4499362 EPC2 T 22.50% (36)FTLD-TDP 7 rs1020004 TMEM106B C 36.91% (55)FTLD-TDP 7 rs1990622 TMEM106B A 27.40% (40)FTLD-TDP/ALS 9 rs2814707 C9orf72 T 5.16% (8)FTLD-TDP 7 rs3173615 TMEM106B C 72.60% (106)FTLD-TDP/ALS 9 rs3849942 C9orf72 T 5.19% (8)FTLD-TDP 17 rs5848 GRN T 18.57% (26)FTLD-TDP 1 rs646776 SORT1/CELSR2/PSRC1 C 27.89% (41)

Key: ALS, amyotrophic lateral sclerosis; Ch, chromosome; FTLD, frontotemporal lobar degeneration; PSP, progressive supranuclear palsy; SNP, single nucleotidepolymorphisms.

C.T. McMillan et al. / Neurobiology of Aging 35 (2014) 1473e14821476

implementation of Eigenanatomy employs a sparseness penaltyon the eigenvectors such that (1) eigenvectors are both sparse (i.e.,have many zero entries) and nonnegative and (2) the nonzerovoxels are clustered and exceed a cluster extent threshold (>100adjacent voxels). The sparseness and nonnegativity allows theeigenvectors to be interpreted as weighted averages of the originaldata, resembling a distributed version of a traditional region ofinterest (Fig. 1C). We refer to each of these distributed regions asan Eigenanatomy VOI.

To relate the identified Eigenanatomy VOIs to SNPs, we per-formed a SCCAN for each neuroimaging modality (Fig. 1E), aspreviously reported (Avants et al., 2010). SCCAN is based on classiccanonical correlation (CCA) analysis, which can be used to evaluatethe multivariate association between 2 data sets, such as geneticsand Eigenanatomy VOIs. SCCAN methods, like classical CCA,compute eigenvectors that maximize the Pearson correlation be-tween the input modalities. However, unlike classical CCAmethods that can achieve very low weight values for variates thataccount for minimal variance, sparse methods apply a penalty inan effort to push variates that account for minimal variance to zeroand therefore output a subset of significant variates that contributeto a statistical model. In our case, the input modalities comprise 1matrix of Eigenanatomy VOIs that represent a neuroimaging mo-dality data set and 1 matrix of additive SNP values that representthe genetic risk factors. In this study, we did not constrain thepolarity of the SCCAN eigenvectors so that we could determineboth positive and negative correlations between neuroimagingand genetics. We performed independent SCCAN analyses to relateGM to SNPs and then to relate FA to SNPs. The significance ofSCCAN results was tested by permuting the imaging input matrix(FA or GM) over the SNP matrix for 1000 permutations and reportthe results that survive p < 0.001. The correlation value producedby the original ordering of the data is then compared with thecorrelations produced over a set of random permutations and thusis controlled for risk of type I error rates in a manner similar tocorrection for multiple comparisons. For each neuroimagingmodality, we report the canonical weights for sparse correlationsthat, like univariate regression beta weights, provide a measure ofthe relative contribution of each variate to the overall canonicalmodel. To increase interpretability of these canonical weights we

also plot the mean FA or GM for each significant VOI and SNP in ourSCCAN models.

3. Results

3.1. Integration of DTI and SNPs

SCCAN revealed a significant association between 2 VOIs and 3SNPs (p < 0.001; canonical weights are reported following eachfeature). VOIs were located in the midbrain (8.7E-02; Fig. 2B) and inthe right superior longitudinal fasciculus (�4.7E-02; SLF; Fig. 2A).The SNPs included rs1768208 (�8.6E-02), rs646776 (�4.3E-02), andrs5848 (�4.5E-02). The canonical weights suggest that rs1768208 isthe most highly associated SNP with WM that and the midbrain isthemost highly associatedWMVOI with the 3 SNPs. The bar graphsin Fig. 2Ce2H summarize the direction of the associations for eachSNP and WM VOI. In the midbrain, increased copy numbers of riskalleles for each SNP are associated with decreased FA. In the SLF,increased rs5848 risk allele copies are associated with reduced FA,whereas fewer risk allele copies of rs1768208 and rs646776 areassociated with reduced FA. A post hoc multivariate regressionrevealedno significant association betweenSNPs andWMVOIswithdemographic factors (Age þ Education þ Gender; F[3,85] ¼ 1.863;p ¼ 0.142). We also observed no association of clinical phenotypewith SNPs and WM VOIs (F[7,84] ¼ 0.736; p ¼ 0.642).

3.2. Integration of GM density and SNPs

SCCAN revealed an association between a single SNP,rs8070723 (1.1E-01), and 4 GM VOIs (p < 0.001): bilateral poste-rior ventral temporal cortex (5.3E-02; Fig. 3A; green), left tem-poroparietal cortex (3.2E-02; Fig. 3A; yellow), left ventralfrontotemporal (1.5E-02; Fig. 3A; magenta), and right tempor-ooccipital cortex (2.4E-02; Fig. 3A; green). The bar graphs in Fig. 3(panels BeE) illustrate that fewer rs8070723 risk allele copiesare associated with decreased GM volume. A post hoc multivariateregression did not reveal any associations between SNPs andWM with demographic factors (F[3,85] ¼ 1.448; p ¼ 0.235)or any associations with clinical phenotype (F[7,84] ¼ 0.516;p ¼ 0.821).

Fig. 2. White matter regions that are significantly associated with frontotemporal lobar degeneration (FTLD) single nucleotide polymorphisms (SNPs). (A) Superior longitudinalfasisculus (SLF) and (B) the midbrain were related to rs1768208 (MOBP), rs646776 (near SORT1), and rs5848 (GRN). Fig. 2Ce2H summarize the direction of the associations for eachSNP and WM VOI: (C) fewer copies of rs178208 risk alleles are associated with reduced fractional anisotropy (FA) in SLF; (D) more rs178208 risk allele copies are associated withreduced FA in midbrain; (E) more rs5848 risk allele copies are associated with reduced FA in SLF; (F) more rs5848 risk allele copies are associated with reduced FA in midbrain; (G)fewer rs646776 risk allele copies are associated with reduced FA in SLF; and (H) more rs646776 risk allele copies are associated with reduced FA in midbrain. For (G) and (H), weillustrate a dominant model because only a few cases were homozygous for the risk allele.

C.T. McMillan et al. / Neurobiology of Aging 35 (2014) 1473e1482 1477

Fig. 3. Gray matter (GM) regions that are significantly associated with frontotemporal lobar degeneration single nucleotide polymorphisms. (A) Four GM volumes of interest wererelated to rs8070723 (MAPT). Panels B through E illustrate that fewer rs8070723 risk allele copies are associated with reduced GM.

C.T. McMillan et al. / Neurobiology of Aging 35 (2014) 1473e14821478

3.3. Neuropathologic confirmation of risk factors

We evaluated the odds ratio (OR) of each significant SNP fromour neuroimaging analysis in FTLD-TDP (n ¼ 91) relative to FTLD-

tau (n ¼ 62) in an independent autopsy series of sporadic FTLDcases (see Table 2 for risk allele frequencies). Carriers of rs1768208risk alleles have a significantly increased risk of FTLD-tau relative toFTLD-TDP (OR ¼ 1.69; p ¼ 0.030). The SNP rs8070723 also confers

Table 2Risk allele frequencies in the independent autopsy series of SNPs identified in theneuroimaging analysis and odds ratio (95% confidence intervals) of FTLD-tau

SNP FTLD-tau FTLD-TDP tau odds ratio p value

rs1768208 43.55% (54) 31.32% (57) 1.69 (1.02e2.79) 0.030rs646776 14.52% (18) 25.27% (46) 0.50 (0.26e0.95) 0.031rs5848 36.29% (45) 33.52% (61) 1.13 (0.68e1.87) 0.627rs8070723a 91.12% (113) 75.27% (137) 3.36 (1.62e7.56) 0.0005

Key: FTLD, frontotemporal lobar degeneration; SNP, single nucleotidepolymorphisms.

a For all SNPs, we report minor allele frequencies, except rs8070723 for which themajor allele (A/A) has previously been associated with MAPT risk.

C.T. McMillan et al. / Neurobiology of Aging 35 (2014) 1473e1482 1479

an increased risk of FTLD-tau relative to FTLD-TDP (OR ¼ 3.36; p ¼0.005). Conversely, rs646776 risk allele carriers have about half thelikelihood of developing FTLD-tau relative to FTLD-TDP (OR ¼ 0.50;p ¼ 0.030). rs5848 does not confer significant risk of a specific FTLDsubtype (OR ¼ 1.13; p ¼ 0.627).

4. Discussion

SNPs have been associated with GM and WM neuroanatomicstructure in AD (Jahanshad et al., 2013; Shen et al., 2010), butthese associations have rarely undergone independent validationat autopsy and have not been assessed in FTLD. Because FTLD ispredominantly caused by a single misfolded protein, either tauor TDP-43, it is an ideal target for pharmaceutical development(Boxer et al., 2013). This is in contrast to diseases such as ADthat result from both tau tangles and amyloid plaque inclusions.It is therefore critical to improve our ability to diagnose FTLDusing quantitative biomarkers and to better understand theunderlying biological mechanisms that contribute to FTLD neu-rodegeneration. Our novel approach identified several geneticand imaging associations in living FTLD patients, and wedetermined that these associations confer risk of a specific FTLDhistopathologic subtype in an independent autopsy series. Thesefindings validate the importance of specific regional WM andGM abnormalities in screening for FTLD; and these novel geneticand neuroanatomic associations emphasize the influence ofSNPs on neuroanatomic structure and provide a proof of conceptfor using the current approach in ongoing biomarker discoveryresearch.

The SNP rs8070723 associated with GM is located in themicrotubule-associated protein tau (MAPT) gene. It tags the H1haplotype, which is a common variant of MAPT (Höglinger et al.,2011). The H1 haplotype has consistently been associated with4-repeat tauopathies, including PSP (Baker et al., 1999) and CBD(Di Maria et al., 2000), relative to controls, although we are un-aware of a direct comparison between FTLD-tau and FTLD-TDPthat evaluated the specificity of this haplotype. The currentstudy indicates that the major risk-conferring allele (A/A) is nearlyalways present in FTLD-tau (>92%).

The GM associations with rs8070723 suggest that MAPT H1(A/A) is protective of GM atrophy compared with carriers of �1H2 (G) alleles. The direction of this association is consistent withprevious reports that the MAPT H2 haplotype may be a moreaggressive form of neurodegeneration. For example, H2 has beenassociated with reduced brain weight at autopsy (Connelly et al.,2011), lower SNAP-25 synaptic transmission (Connelly et al.,2011), and reduced glucose metabolism (Laws et al., 2007). TheH2 haplotype has also been associated with increased neurofi-brillary tangle burden in Alzheimer’s and Lewy body diseases(Wider et al., 2012), but to our knowledge pathologic burden ofMAPT haplotype has not been investigated in FTLD. The locus of

the rs8070723 GM association included bilateral posteriorventral temporal, left temporoparietal, left ventral fronto-temporal, and right temporooccipital cortices. Our autopsyanalysis demonstrated that H1 confers increased risk of FTLD-tau, and in this comparative study H2 therefore confersincreased FTLD-TDP risk; it is thus possible that this GM distri-bution reflects FTLD-TDP disease. Indeed, FTLD-TDP appears tobe associated with a temporal distribution of disease, asobserved in the semantic variant of primary progressive aphasia,which is statistically biased toward TDP-43 pathology(Grossman, 2010), where FTLD-tau is associated more with afrontal distribution (Whitwell et al., 2009).

We also observed that another SNP, rs1768208, confersincreased risk of developing FTLD-tau relative to FTLD-TDP. In aprevious GWA study of PSP patients, rs1768208 conferredincreased risk of FTLD-tau relative to case controls (Höglingeret al., 2011). This observation suggested that rs1768208 is asso-ciated with the presence of tau pathology, but the specificity ofthe association was not evaluated. The current study performed acomparative evaluation relative to patients with FTLD-TDP todetermine the specificity of rs1768208 as a marker of FTLD-tau. Inanother study, rs1768208 was not associated with late-onset ADrisk relative to case controls (Liu et al., 2013), providing addi-tional evidence that this SNP is not a nonspecific marker of anyneurodegenerative disease.

The presence of more risk allele (T) copies of rs1768208 isassociated with reduced FA in the midbrain. The midbrain iscommonly reported in WM neuroimaging and neuropathologicstudies of PSP (Dickson et al., 2010). Volumetric WM and DTIstudies have revealed reduced FA in the midbrain of PSP patientsrelative to healthy adults (Whitwell et al., 2011a, 2011c). Theobservation that rs1768208 has a WM correlate in the midbrain isconsistent with this SNP’s location in the region on chromosome 3coding for myelin-associated oligodendrocyte basic protein(MOBP). MOBP is involved in protecting the integrity of the myelinsheath (Yamamoto et al., 1999) and is highly expressed in the WMof midbrain regions (Montague et al., 2006). We additionallyobserved that fewer copies of rs1768208 risk alleles is associatedwith reduced FA in the SLF. Whereas SLF has previously beenassociated with FTLD-tau (McMillan et al., 2013), it is possible thatthe 4-repeat subtype of tau found in PSP and related to rs1768208has distinct neural correlates relative to 3-repeat tau subtypes.However, our post hoc analyses demonstrated that clinical syn-drome is not driving any of the observed neuroanatomic and ge-netic associations.

In our comparative autopsy analysis, we observed that the SNPrs646776 conferred a 2-fold increased risk of FTLD-TDP relative toFTLD-tau. rs646776 is located near sortilin (SORT1) as well asCELSR2 and PSRC1 and has been associated with regulation ofplasma progranulin levels: each copy of the risk allele yieldednearly a 15% decrease in progranulin (Carrasquillo et al., 2010).The current study suggests that risk allele copies for rs646776 areassociated with increased FA in SLF, which is consistent withprevious neuroimaging and neuropathologic observations sug-gesting that SLF is relatively preserved in FTLD-TDP (McMillanet al., 2013).

Another SNP, rs5848, is located in the 30 untranslated region ofGRN. A previous study suggested that homozygous risk allelecarriers of rs5848 have a 3.2-fold increased risk of FTLD-TDPcompared with case controls (Rademakers et al., 2008). Riskallele carriers of rs5848, like rs646776, have been reported tohave lower progranulin plasma levels in a clinically definedcohort (Hsiung et al., 2011). We observed that risk allele carriersof rs5848 have reduced FA in the SLF. However, the direction ofthis pattern is contrary to a previous neuroimaging and

C.T. McMillan et al. / Neurobiology of Aging 35 (2014) 1473e14821480

neuropathologic report suggesting that SLF is relatively spared inFTLD-TDP compared with FTLD-tau (McMillan et al., 2013).However, the role of rs5848 in FTLD-TDP has been controversial.Plasma progranulin levels were not affected in a subset of spo-radic patients with presumed FTLD-TDP pathology due to FTLD-ALS or presumed FTLD-tau (e.g., CBS, PSP) (Hsiung et al., 2011),and several studies have suggested that rs5848 may not beassociated with sporadic FTLD-TDP (Rollinson et al., 2011; Simón-Sánchez et al., 2009). Our comparative autopsy study also sug-gests that rs5848 may not specifically increase the risk for spo-radic FTLD-TDP because risk allele carriers in our cohort areequally likely to have FTLD-TDP or FTLD-tau pathology.

The observation that SNPs related to FTLD-TDP have WM as-sociations is consistent with neuropathologic studies reportingmicroglial activation, myelin changes, and gliosis in GRN muta-tion carriers (Kelley et al., 2009). GRN mutations have beenstrongly associated with FTLD-TDP pathology (Gass et al., 2006).Also, progranulin is expressed in microglia (Ahmed et al., 2007)and in the midbrain of ALS patients (Irwin et al., 2009) who haveFTLD-TDP pathology. A DTI study of WM reported that GRNmutation carriers have reduced FA relative to sporadic FTLD pa-tients, but these analyses were restricted to a priori regions ofinterest in the corpus callosum (Bozzali et al., 2013). In anotherDTI study, FA reductions in WM of the extreme capsule, inferiorfrontal-occipital fasciculus and uncinate were reported inasymptomatic GRN mutation carriers relative to control non-carriers. Moreover, volumetric measures of WM and GM did notreveal any differences across carriers and noncarriers (Borroniet al., 2008). This suggests that the DTI modality used in ourstudy may provide a more sensitive preclinical marker of FTLDpathology.

Because WM abnormalities in the midbrain were associatedwith FTLD-tau and FTLD-TDP SNPs, this region may be used toscreen for the presence of any FTLD-related disease on an indi-vidual subject basis, and then SNP analyses can be used to eval-uate whether these WM changes are associated with risk for aspecific FTLD spectrum histopathologic abnormality. Themidbrain may be an optimal region to evaluate presence of dis-ease in WM because this region is hypothesized to be affected inearly stages of TDP-43 (Geser et al., 2011) associated with ALS andmay be an early locus of pathologically modified tau associatedwith PSP or CBD (Irwin et al., 2012a).

The biomarker discovery approach proposed in this reportidentified neuroanatomic and genetic associations in a livingcohort of FTLD patients and then validated the meaningfulness ofthese associations in an independent autopsy series. Rather thanfocusing on the neuroanatomic correlates of a single SNP, ouranalysis procedure additionally benefits by supporting selection ofa subset of meaningful SNPs with neuroimaging correlates from alarge panel of SNPs. Other approaches have included voxel-wisestudies in which a GWA study is performed at each individualvoxel (k ¼ 31,622) of the entire brain (Stein et al., 2010). Althoughquantitative trait analyses such as voxel-wise measures can yieldincreased statistical power relative to categorical case-controlGWA studies (Potkin et al., 2009), these studies require largenumbers of cases which is less feasible in uncommon but infor-mative neurodegenerative disorders like FTLD. Our data-drivenVOI method, Eigenanatomy, also allows us to reduce neuro-imaging comparisons to a small set of statistically determinedregions that yields increased statistical power and minimizesmultiple comparison problems.

In sum, we demonstrate that GM and WM neuroanatomicstructure in sporadic FTLD is related to genetic risk factors and thatthese SNPs confer risk specifically for FTLD-tau or FTLD-TDP his-topathologic abnormalities. These observations also emphasize

that SNPs modify neuroanatomic structure and underline thepractical contribution of SNPs and neuroimaging to biomarkerdiscovery. This report also demonstrates a novel method forevaluating neuroanatomic correlates of genetic risk factors asso-ciated with neurodegenerative disease.

Disclosure statement

The authors have no actual or potential conflicts of interest.

Acknowledgements

Supported in part by the National Institutes of Health (grantsAG043503, T32-AG000255, AG017586, NS044266, AG032953,AG010124) and the Wyncote Foundation. J.B.T. is supported by theAlfonso Martín Escudero Foundation.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.neurobiolaging.2013.11.029.

References

Ahmed, Z., Mackenzie, I.R.A., Hutton, M.L., Dickson, D.W., 2007. Progranulin infrontotemporal lobar degeneration and neuroinflammation. J. Neuroinflamm. 4,7.

Avants, B., Dhillon, P., Kandel, B.M., Cook, P.A., McMillan, C.T., Grossman, M., Gee, J.C.,2012. Eigenanatomy improves detection power for longitudinal cortical change.Med. Image Comput. Comput. Assist. Interv. 15, 206e213.

Avants, B.B., Cook, P.A., Ungar, L., Gee, J.C., Grossman, M., 2010. Dementia inducescorrelated reductions in white matter integrity and cortical thickness: amultivariate neuroimaging study with sparse canonical correlation analysis.NeuroImage 50, 1004e1016.

Avants, B.B., Tustison, N.J., Song, G., Cook, P.A., Klein, A., Gee, J.C., 2011.A reproducible evaluation of ANTs similarity metric performance in brain imageregistration. NeuroImage 54, 2033e2044.

Baker, M., Litvan, I., Houlden, H., Adamson, J., Dickson, D., Perez-Tur, J., Hardy, J.,Lynch, T., Bigio, E., Hutton, M., 1999. Association of an extended haplotype in thetau gene with progressive supranuclear palsy. Hum. Mol. Genet. 8, 711e715.

Baker, M., Mackenzie, I.R., Pickering-Brown, S.M., Gass, J., Rademakers, R.,Lindholm, C., Snowden, J., Adamson, J., Sadovnick, A.D., Rollinson, S., Cannon, A.,Dwosh, E., Neary, D., Melquist, S., Richardson, A., Dickson, D., Berger, Z.,Eriksen, J., Robinson, T., Zehr, C., Dickey, C.A., Crook, R., McGowan, E., Mann, D.,Boeve, B., Feldman, H., Hutton, M., 2006. Mutations in progranulin cause tau-negative frontotemporal dementia linked to chromosome 17. Nature 442,916e919.

Borroni, B., Alberici, A., Premi, E., Archetti, S., Garibotto, V., Agosti, C., Gasparotti, R.,Di Luca, M., Perani, D., Padovani, A., 2008. Brain magnetic resonance imagingstructural changes in a pedigree of asymptomatic progranulin mutation car-riers. Rejuvenation Res. 11, 585e595.

Boxer, A.L., Gold, M., Huey, E., Gao, F.-B., Burton, E.A., Chow, T., Kao, A., Leavitt, B.R.,Lamb, B., Grether, M., Knopman, D., Cairns, N.J., Mackenzie, I.R., Mitic, L.,Roberson, E.D., Van Kammen, D., Cantillon, M., Zahs, K., Salloway, S., Morris, J.,Tong, G., Feldman, H., Fillit, H., Dickinson, S., Khachaturian, Z., Sutherland, M.,Farese, R., Miller, B.L., Cummings, J., 2013. Frontotemporal degeneration, thenext therapeutic frontier: molecules and animal models for frontotemporaldegeneration drug development. Alzheimers Dement. 9, 176e188.

Bozzali, M., Battistoni, V., Premi, E., Alberici, A., Giulietti, G., Archetti, S., Turla, M.,Gasparotti, R., Cercignani, M., Padovani, A., Borroni, B., 2013. Structural brainsignature of FTLD driven by granulin mutation. J. Alzheimers Dis. 33, 483e494.

Carrasquillo, M.M., Nicholson, A.M., Finch, N., Gibbs, J.R., Baker, M., Rutherford, N.J.,Hunter, T.A., DeJesus-Hernandez, M., Bisceglio, G.D., Mackenzie, I.R.,Singleton, A., Cookson, M.R., Crook, J.E., Dillman, A., Hernandez, D.,Petersen, R.C., Graff-Radford, N.R., Younkin, S.G., Rademakers, R., 2010. Genome-wide screen identifies rs646776 near sortilin as a regulator of progranulin levelsin human plasma. Am. J. Hum. Genet. 87, 890e897.

Connelly, S.J., Mukaetova-Ladinska, E.B., Abdul-All, Z., Alves da Silva, J., Brayne, C.,Honer, W.G., Mann, D.M.A., 2011. Synaptic changes in frontotemporal lobardegeneration: correlation with MAPT haplotype and APOE genotype. Neuro-pathol. Appl. Neurobiol. 37, 366e380.

DeJesus-Hernandez, M., Mackenzie, I.R., Boeve, B.F., Boxer, A.L., Baker, M.,Rutherford, N.J., Nicholson, A.M., Finch, N.A., Flynn, H., Adamson, J., Kouri, N.,Wojtas, A., Sengdy, P., Hsiung, G.-Y.R., Karydas, A., Seeley, W.W., Josephs, K.A.,Coppola, G., Geschwind, D.H., Wszolek, Z.K., Feldman, H., Knopman, D.S.,Petersen, R.C., Miller, B.L., Dickson, D.W., Boylan, K.B., Graff-Radford, N.R.,

C.T. McMillan et al. / Neurobiology of Aging 35 (2014) 1473e1482 1481

Rademakers, R., 2011. Expanded GGGGCC hexanucleotide repeat in noncodingregion of C9ORF72 causes chromosome 9p-linked FTD and ALS. Neuron 72,245e256.

Di Maria, E., Tabaton, M., Vigo, T., Abbruzzese, G., Bellone, E., Donati, C., Frasson, E.,Marchese, R., Montagna, P., Munoz, D.G., Pramstaller, P.P., Zanusso, G., Ajmar, F.,Mandich, P., 2000. Corticobasal degeneration shares a common genetic back-ground with progressive supranuclear palsy. Ann. Neurol. 47, 374e377.

Dickson, D.W., Ahmed, Z., Algom, A.A., Tsuboi, Y., Josephs, K.A., 2010. Neuropa-thology of variants of progressive supranuclear palsy. Curr. Opin. Neurol. 23,394e400.

Forman, M.S., Zhukareva, V., Bergeron, C., Chin, S.S.-M., Grossman, M., Clark, C.,Lee, V.M.-Y., Trojanowski, J.Q., 2002. Signature tau neuropathology in gray andwhite matter of corticobasal degeneration. Am. J. Pathol. 160, 2045e2053.

Gass, J., Cannon, A., Mackenzie, I.R., Boeve, B., Baker, M., Adamson, J., Crook, R.,Melquist, S., Kuntz, K., Petersen, R., Josephs, K., Pickering-Brown, S.M., Graff-Radford, N., Uitti, R., Dickson, D., Wszolek, Z., Gonzalez, J., Beach, T.G., Bigio, E.,Johnson, N., Weintraub, S., Mesulam, M., White, C.L., Woodruff, B., Caselli, R.,Hsiung, G.-Y., Feldman, H., Knopman, D., Hutton, M., Rademakers, R., 2006.Mutations in progranulin are a major cause of ubiquitin-positive frontotemporallobar degeneration. Hum. Mol. Genet. 15, 2988e3001.

Geser, F., Martinez-Lage, M., Robinson, J., Uryu, K., Neumann, M., Brandmeir, N.J.,Xie, S.X., Kwong, L.K., Elman, L., McCluskey, L., Clark, C.M., Malunda, J.,Miller, B.L., Zimmerman, E.A., Qian, J., Van Deerlin, V., Grossman, M., Lee, V.M.-Y.,Trojanowski, J.Q., 2009. Clinical and pathological continuum of multisystemTDP-43 proteinopathies. Arch. Neurol. 66, 180e189.

Geser, F., Prvulovic, D., O’Dwyer, L., Hardiman, O., Bede, P., Bokde, A.L.W.,Trojanowski, J.Q., Hampel, H., 2011. On the development of markers for path-ological TDP-43 in amyotrophic lateral sclerosis with and without dementia.Prog. Neurobiol. 95, 649e662.

Grossman, M., 2010. Primary progressive aphasia: clinicopathological correlations.Nat. Rev. Neurol. 6, 88e97.

Höglinger, G.U., Melhem, N.M., Dickson, D.W., Sleiman, P.M.A., Wang, L.-S., Klei, L.,Rademakers, R., de Silva, R., Litvan, I., Riley, D.E., van Swieten, J.C., Heutink, P.,Wszolek, Z.K., Uitti, R.J., Vandrovcova, J., Hurtig, H.I., Gross, R.G., Maetzler, W.,Goldwurm, S., Tolosa, E., Borroni, B., Pastor, P., PSP Genetics Study Group,Cantwell, L.B., Han, M.R., Dillman, A., van der Brug, M.P., Gibbs, J.R.,Cookson, M.R., Hernandez, D.G., Singleton, A.B., Farrer, M.J., Yu, C.-E., Golbe, L.I.,Revesz, T., Hardy, J., Lees, A.J., Devlin, B., Hakonarson, H., Müller, U.,Schellenberg, G.D., 2011. Identification of common variants influencing risk ofthe tauopathy progressive supranuclear palsy. Nat. Genet. 43, 699e705.

Hsiung, G.-Y.R., Fok, A., Feldman, H.H., Rademakers, R., Mackenzie, I.R.A., 2011.rs5848 polymorphism and serum progranulin level. J. Neurol. Sci. 300, 28e32.

Hutton, M., Lendon, C.L., Rizzu, P., Baker, M., Froelich, S., Houlden, H., Pickering-Brown, S., Chakraverty, S., Isaacs, A., Grover, A., Hackett, J., Adamson, J.,Lincoln, S., Dickson, D., Davies, P., Petersen, R.C., Stevens, M., de Graaff, E.,Wauters, E., van Baren, J., Hillebrand, M., Joosse, M., Kwon, J.M., Nowotny, P.,Che, L.K., Norton, J., Morris, J.C., Reed, L.A., Trojanowski, J., Basun, H., Lannfelt, L.,Neystat, M., Fahn, S., Dark, F., Tannenberg, T., Dodd, P.R., Hayward, N., Kwok, J.B.,Schofield, P.R., Andreadis, A., Snowden, J., Craufurd, D., Neary, D., Owen, F.,Oostra, B.A., Hardy, J., Goate, A., van Swieten, J., Mann, D., Lynch, T., Heutink, P.,1998. Association of missense and 5’-splice-site mutations in tau with theinherited dementia FTDP-17. Nature 393, 702e705.

Irwin, D., Lippa, C.F., Rosso, A., 2009. Progranulin (PGRN) expression in ALS: animmunohistochemical study. J. Neurol. Sci. 276, 9e13.

Irwin, D.J., Cohen, T.J., Grossman, M., Arnold, S.E., Xie, S.X., Lee, V.M.-Y.,Trojanowski, J.Q., 2012a. Acetylated tau, a novel pathological signature in Alz-heimer’s disease and other tauopathies. Brain 135, 807e818.

Irwin, D.J., McMillan, C.T., Toledo, J.B., Arnold, S.E., Shaw, L.M., Wang, L.-S., VanDeerlin, V., Lee, V.M.-Y., Trojanowski, J.Q., Grossman, M., 2012b. Comparison ofcerebrospinal fluid levels of tau and Ab 1-42 in Alzheimer disease and fronto-temporal degeneration using 2 analytical platforms. Arch. Neurol. 69,1018e1025.

Jahanshad, N., Rajagopalan, P., Hua, X., Hibar, D.P., Nir, T.M., Toga, A.W.,Jack, C.R., Saykin, A.J., Green, R.C., Weiner, M.W., Medland, S.E.,Montgomery, G.W., Hansell, N.K., McMahon, K.L., de Zubicaray, G.I.,Martin, N.G., Wright, M.J., Thompson, P.M., Alzheimer’s Disease Neuro-imaging Initiative, 2013. Genome-wide scan of healthy human connectomediscovers SPON1 gene variant influencing dementia severity. Proc. Natl.Acad. Sci. U.S.A 110, 4768e4773.

Kelley, B.J., Haidar, W., Boeve, B.F., Baker, M., Graff-Radford, N.R., Krefft, T.,Frank, A.R., Jack, C.R., Shiung, M., Knopman, D.S., Josephs, K.A., Parashos, S.A.,Rademakers, R., Hutton, M., Pickering-Brown, S., Adamson, J., Kuntz, K.M.,Dickson, D.W., Parisi, J.E., Smith, G.E., Ivnik, R.J., Petersen, R.C., 2009. Prominentphenotypic variability associated with mutations in Progranulin. Neurobiol.Aging 30, 739e751.

Laws, S.M., Perneczky, R., Drzezga, A., Diehl-Schmid, J., Ibach, B., Bäuml, J., Eisele, T.,Förstl, H., Kurz, A., Riemenschneider, M., 2007. Association of the tau haplotypeH2 with age at onset and functional alterations of glucose utilization in fron-totemporal dementia. Am. J. Psychiatry 164, 1577e1584.

Lippa, C.F., Rosso, A.L., Stutzbach, L.D., Neumann, M., Lee, V.M.-Y., Trojanowski, J.Q.,2009. Transactive response DNA-binding protein 43 burden in familial Alz-heimer disease and Down syndrome. Arch. Neurol. 66, 1483e1488.

Liu, Q.-Y., Yu, J.-T., Miao, D., Ma, X.-Y., Wang, H.-F., Wang, W., Tan, L., 2013. Anexploratory study on STX6, MOBP, MAPT, and EIF2AK3 and late-onset Alz-heimer’s disease. Neurobiol. Aging 34, 1519.e13e1519.e17.

Mackenzie, I.R.A., Neumann, M., Bigio, E.H., Cairns, N.J., Alafuzoff, I., Kril, J.,Kovacs, G.G., Ghetti, B., Halliday, G., Holm, I.E., Ince, P.G., Kamphorst, W.,Revesz, T., Rozemuller, A.J.M., Kumar-Singh, S., Akiyama, H., Baborie, A., Spina, S.,Dickson, D.W., Trojanowski, J.Q., Mann, D.M.A., 2010. Nomenclature andnosology for neuropathologic subtypes of frontotemporal lobar degeneration:an update. Acta Neuropathol. 119, 1e4.

McMillan, C.T., Irwin, D.J., Avants, B., Powers, J., Cook, P., Toledo, J.B., Wood, E.M., VanDeerlin, V.M., Lee, C.M.-Y., Trojanowski, J.Q., Grossman, M., 2013. White matterimaging helps dissociate tau from TDP-43 in frontotemporal lobar degeneration.J. Neurol. Neurosurg. Psychiatry 84, 949e955.

Montague, P., McCallion, A.S., Davies, R.W., Griffiths, I.R., 2006. Myelin-associatedoligodendrocytic basic protein: a family of abundant CNS myelin proteins insearch of a function. Dev. Neurosci. 28, 479e487.

Neumann, M., Kwong, L.K., Lee, E.B., Kremmer, E., Flatley, A., Xu, Y., Forman, M.S.,Troost, D., Kretzschmar, H.A., Trojanowski, J.Q., Lee, V.M., 2009. Phosphorylationof S409/410 of TDP-43 is a consistent feature in all sporadic and familial formsof TDP-43 proteinopathies. Acta Neuropathol. 117, 137e149.

Otvos, L., Feiner, L., Lang, E., Szendrei, G.I., Goedert, M., Lee, V.M., 1994. Monoclonalantibody PHF-1 recognizes tau protein phosphorylated at serine residues 396and 404. J. Neurosci. Res. 39, 669e673.

Potkin, S.G., Turner, J.A., Guffanti, G., Lakatos, A., Torri, F., Keator, D.B., Macciardi, F.,2009. Genome-wide strategies for discovering genetic influences on cognitionand cognitive disorders: methodological considerations. Cogn. Neuropsychiatry14, 391e418.

Rademakers, R., Eriksen, J.L., Baker, M., Robinson, T., Ahmed, Z., Lincoln, S.J.,Finch, N., Rutherford, N.J., Crook, R.J., Josephs, K.A., Boeve, B.F., Knopman, D.S.,Petersen, R.C., Parisi, J.E., Caselli, R.J., Wszolek, Z.K., Uitti, R.J., Feldman, H.,Hutton, M.L., Mackenzie, I.R., Graff-Radford, N.R., Dickson, D.W., 2008. Commonvariation in the miR-659 binding-site of GRN is a major risk factor for TDP43-positive frontotemporal dementia. Hum. Mol. Genet. 17, 3631e3642.

Rademakers, R., Melquist, S., Cruts, M., Theuns, J., Del-Favero, J., Poorkaj, P.,Baker, M., Sleegers, K., Crook, R., De Pooter, T., Bel Kacem, S., Adamson, J., Vanden Bossche, D., Van den Broeck, M., Gass, J., Corsmit, E., De Rijk, P., Thomas, N.,Engelborghs, S., Heckman, M., Litvan, I., Crook, J., De Deyn, P.P., Dickson, D.,Schellenberg, G.D., Van Broeckhoven, C., Hutton, M.L., 2005. High-density SNPhaplotyping suggests altered regulation of tau gene expression in progressivesupranuclear palsy. Hum. Mol. Genet. 14, 3281e3292.

Renton, A.E., Majounie, E., Waite, A., Simón-Sánchez, J., Rollinson, S., Gibbs, J.R.,Schymick, J.C., Laaksovirta, H., van Swieten, J.C., Myllykangas, L., Kalimo, H.,Paetau, A., Abramzon, Y., Remes, A.M., Kaganovich, A., Scholz, S.W.,Duckworth, J., Ding, J., Harmer, D.W., Hernandez, D.G., Johnson, J.O., Mok, K.,Ryten, M., Trabzuni, D., Guerreiro, R.J., Orrell, R.W., Neal, J., Murray, A.,Pearson, J., Jansen, I.E., Sondervan, D., Seelaar, H., Blake, D., Young, K.,Halliwell, N., Callister, J.B., Toulson, G., Richardson, A., Gerhard, A., Snowden, J.,Mann, D., Neary, D., Nalls, M.A., Peuralinna, T., Jansson, L., Isoviita, V.-M.,Kaivorinne, A.-L., Hölttä-Vuori, M., Ikonen, E., Sulkava, R., Benatar, M., Wuu, J.,Chiò, A., Restagno, G., Borghero, G., Sabatelli, M., ITALSGEN Consortium,Heckerman, D., Rogaeva, E., Zinman, L., Rothstein, J.D., Sendtner, M., Drepper, C.,Eichler, E.E., Alkan, C., Abdullaev, Z., Pack, S.D., Dutra, A., Pak, E., Hardy, J.,Singleton, A., Williams, N.M., Heutink, P., Pickering-Brown, S., Morris, H.R.,Tienari, P.J., Traynor, B.J., 2011. A hexanucleotide repeat expansion in C9ORF72 isthe cause of chromosome 9p21-linked ALS-FTD. Neuron 72, 257e268.

Rohrer, J.D., Ridgway, G.R., Modat, M., Ourselin, S., Mead, S., Fox, N.C., Rossor, M.N.,Warren, J.D., 2010. Distinct profiles of brain atrophy in frontotemporal lobardegeneration caused by progranulin and tau mutations. NeuroImage 53,1070e1076.

Rollinson, S., Rohrer, J.D., van der Zee, J., Sleegers, K., Mead, S., Engelborghs, S.,Collinge, J., De Deyn, P.P., Mann, D.M.A., Van Broeckhoven, C., Pickering-Brown, S.M., 2011. No association of PGRN 3’UTR rs5848 in frontotemporal lobardegeneration. Neurobiol. Aging 32, 754e755.

Seltman, R.E., Matthews, B.R., 2012. Frontotemporal lobar degeneration: epidemi-ology, pathology, diagnosis and management. CNS Drugs 26, 841e870.

Shen, L., Kim, S., Risacher, S.L., Nho, K., Swaminathan, S., West, J.D., Foroud, T.,Pankratz, N., Moore, J.H., Sloan, C.D., Huentelman, M.J., Craig, D.W.,Dechairo, B.M., Potkin, S.G., Jack, C.R., Weiner, M.W., Saykin, A.J., Alzheimer’sDisease Neuroimaging Initiative, 2010. Whole genome association study ofbrain-wide imaging phenotypes for identifying quantitative trait loci in MCI andAD: a study of the ADNI cohort. NeuroImage 53, 1051e1063.

Simón-Sánchez, J., Seelaar, H., Bochdanovits, Z., Deeg, D.J.H., van Swieten, J.C.,Heutink, P., 2009. Variation at GRN 3’-UTR rs5848 is not associated with a risk offrontotemporal lobar degeneration in Dutch population. PLoS One 4, e7494.

Stein, J.L., Hua, X., Lee, S., Ho, A.J., Leow, A.D., Toga, A.W., Saykin, A.J., Shen, L.,Foroud, T., Pankratz, N., Huentelman, M.J., Craig, D.W., Gerber, J.D., Allen, A.N.,Corneveaux, J.J., Dechairo, B.M., Potkin, S.G., Weiner, M.W., Thompson, P.,Alzheimer’s Disease Neuroimaging Initiative, 2010. Voxelwise genome-wideassociation study (vGWAS). NeuroImage 53, 1160e1174.

Van Deerlin, V.M., Sleiman, P.M.A., Martinez-Lage, M., Chen-Plotkin, A., Wang, L.-S.,Graff-Radford, N.R., Dickson, D.W., Rademakers, R., Boeve, B.F., Grossman, M.,Arnold, S.E., Mann, D.M.A., Pickering-Brown, S.M., Seelaar, H., Heutink, P., vanSwieten, J.C., Murrell, J.R., Ghetti, B., Spina, S., Grafman, J., Hodges, J.,Spillantini, M.G., Gilman, S., Lieberman, A.P., Kaye, J.A., Woltjer, R.L., Bigio, E.H.,Mesulam, M., Al-Sarraj, S., Troakes, C., Rosenberg, R.N., White, C.L., Ferrer, I.,Lladó, A., Neumann, M., Kretzschmar, H.A., Hulette, C.M., Welsh-Bohmer, K.A.,Miller, B.L., Alzualde, A., Lopez de Munain, A., McKee, A.C., Gearing, M.,Levey, A.I., Lah, J.J., Hardy, J., Rohrer, J.D., Lashley, T., Mackenzie, I.R.A.,

C.T. McMillan et al. / Neurobiology of Aging 35 (2014) 1473e14821482

Feldman, H.H., Hamilton, R.L., Dekosky, S.T., van der Zee, J., Kumar-Singh, S.,Van Broeckhoven, C., Mayeux, R., Vonsattel, J.P.G., Troncoso, J.C., Kril, J.J.,Kwok, J.B.J., Halliday, G.M., Bird, T.D., Ince, P.G., Shaw, P.J., Cairns, N.J.,Morris, J.C., McLean, C.A., DeCarli, C., Ellis, W.G., Freeman, S.H., Frosch, M.P.,Growdon, J.H., Perl, D.P., Sano, M., Bennett, D.A., Schneider, J.A., Beach, T.G.,Reiman, E.M., Woodruff, B.K., Cummings, J., Vinters, H.V., Miller, C.A., Chui, H.C.,Alafuzoff, I., Hartikainen, P., Seilhean, D., Galasko, D., Masliah, E., Cotman, C.W.,Tuñón, M.T., Martínez, M.C.C., Munoz, D.G., Carroll, S.L., Marson, D.,Riederer, P.F., Bogdanovic, N., Schellenberg, G.D., Hakonarson, H.,Trojanowski, J.Q., Lee, V.M., 2010. Common variants at 7p21 are associated withfrontotemporal lobar degeneration with TDP-43 inclusions. Nat. Genet. 42,234e239.

Whitwell, J.L., Avula, R., Master, A., Vemuri, P., Senjem, M.L., Jones, D.T., Jack, C.R.,Josephs, K.A., 2011a. Disrupted thalamocortical connectivity in PSP: a resting-state fMRI, DTI, and VBM study. Parkinsonism Relat. Disord. 17, 599e605.

Whitwell, J.L., Jack, C.R., Parisi, J.E., Knopman, D.S., Boeve, B.F., Petersen, R.C.,Dickson, D.W., Josephs, K.A., 2011b. Imaging signatures of molecular pathologyin behavioral variant frontotemporal dementia. J. Mol. Neurosci. 45, 372e378.

Whitwell, J.L., Jack, C.R., Senjem, M.L., Parisi, J.E., Boeve, B.F., Knopman, D.S.,Dickson, D.W., Petersen, R.C., Josephs, K.A., 2009. MRI correlates of proteindeposition and disease severity in postmortem frontotemporal lobar degener-ation. Neurodegener. Dis. 6, 106e117.

Whitwell, J.L., Master, A.V., Avula, R., Kantarci, K., Eggers, S.D., Edmonson, H.A.,Jack, C.R., Josephs, K.A., 2011c. Clinical correlates of white matter tract degen-eration in progressive supranuclear palsy. Arch. Neurol. 68, 753e760.

Whitwell, J.L., Weigand, S.D., Boeve, B.F., Senjem, M.L., Gunter, J.L., DeJesus-Hernandez, M., Rutherford, N.J., Baker, M., Knopman, D.S., Wszolek, Z.K.,Parisi, J.E., Dickson, D.W., Petersen, R.C., Rademakers, R., Jack, C.R., Josephs, K.A.,2012. Neuroimaging signatures of frontotemporal dementia genetics: C9ORF72,tau, progranulin and sporadics. Brain 135, 794e806.

Wider, C., Ross, O.A., Nishioka, K., Heckman, M.G., Vilariño-Güell, C., Jasinska-Myga, B., Erketin-Taner, N., Rademakers, R., Graff-Radford, N.R., Mash, D.C.,Papapetropoulos, S., Duara, R., Uchikado, H., Wszolek, Z.K., Farrer, M.J.,Dickson, D.W., 2012. An evaluation of the impact of MAPT, SNCA and APOE onthe burden of Alzheimer’s and Lewy body pathology. J. Neurol. Neurosurg.Psychiatry 83, 424e429.

Wood, E.M., Falcone, D., Suh, E., Irwin, D.J., Chen-Plotkin, A.S., Lee, E.B., Xie, S.X., VanDeerlin, V.M., Grossman, M., 2013. Development and validation of pedigreeclassification criteria for frontotemporal lobar degeneration. JAMA Neurol. 70,1411e1417.

Yamamoto, Y., Yoshikawa, H., Nagano, S., Kondoh, G., Sadahiro, S., Gotow, T.,Yanagihara, T., Sakoda, S., 1999. Myelin-associated oligodendrocytic basic pro-tein is essential for normal arrangement of the radial component in centralnervous system myelin. Eur. J. Neurosci. 11, 847e855.