letters

Identification of common variants influencing risk of the

tauopathy progressive supranuclear palsy
Günter U Höglinger1,30, Nadine M Melhem2,30, Dennis W Dickson3,30, Patrick M A Sleiman4,30, Li-San Wang5,
Lambertus Klei2, Rosa Rademakers3, Rohan de Silva6, Irene Litvan7, David E Riley8, John C van Swieten9,
Peter Heutink10, Zbigniew K Wszolek11, Ryan J Uitti11, Jana Vandrovcova6, Howard I Hurtig12, Rachel G Gross12,
Walter Maetzler13,14, Stefano Goldwurm15, Eduardo Tolosa16, Barbara Borroni17, Pau Pastor18–20,
© 2011 Nature America, Inc. All rights reserved.

PSP Genetics Study Group29, Laura B Cantwell5, Mi Ryung Han5, Allissa Dillman21, Marcel P van der Brug22,
J Raphael Gibbs6,21, Mark R Cookson21, Dena G Hernandez6,21, Andrew B Singleton21, Matthew J Farrer23,
Chang-En Yu24,25, Lawrence I Golbe26, Tamas Revesz27, John Hardy6, Andrew J Lees6,27, Bernie Devlin2,
Hakon Hakonarson4, Ulrich Müller28,30, Gerard D Schellenberg5,30

Progressive supranuclear palsy (PSP) is a movement disorder with PSP is a rare neurodegenerative movement disorder clinically char­
prominent tau neuropathology. Brain diseases with abnormal tau acterized by falls, axial rigidity, vertical supranuclear gaze palsy,
deposits are called tauopathies, the most common of which is bradykinesia and cognitive decline. Though PSP is rare (with
Alzheimer’s disease. Environmental causes of tauopathies include a prevalence of 3.1–6.5 per 100,000 people1), after Parkinson’s
repetitive head trauma associated with some sports. To identify ­disease, PSP is the second most common cause of degenerative
common genetic variation contributing to risk for tauopathies, parkinsonism 2. PSP is a tauopathy with abnormal accumulation
we carried out a genome-wide association study of 1,114 of tau protein within neurons as neurofibrillary tangles, primarily
individuals with PSP (cases) and 3,247 controls (stage 1) in the basal ganglia, diencephalon and brainstem, with neuronal
followed by a second stage in which we genotyped 1,051 cases loss in the globus pallidus, subthalamic nucleus and substantia
and 3,560 controls for the stage 1 SNPs that yielded P ≤ 10−3. nigra. Abnormal tau also accumulates within oligodendroglia and
We found significant previously unidentified signals (P < 5 × astrocytes3. In Alzheimer’s disease, even though all affected indivi­
10−8) associated with PSP risk at STX6, EIF2AK3 and MOBP. duals have neurofibrillary ­tangles, Aβ plaques are closely tied to
We confirmed two independent variants in MAPT affecting risk the primary disease process, and, thus, Alzheimer’s ­disease is a sec­
for PSP, one of which influences MAPT brain expression. The ondary tauopathy. PSP is a primary tauopathy because tau is the
genes implicated encode proteins for vesicle-membrane fusion major abnormal protein observed. Both environmental insults and
at the Golgi-endosomal interface, for the endoplasmic reticulum inherited factors contribute to the risk of developing ­tauopathies4.
unfolded protein response and for a myelin structural component. Repetitive brain trauma, associated with certain sports, can cause

1Department of Neurology, Philipps-Universität, Marburg, Germany. 2Department of Psychiatry, University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania,
USA. 3Department of Neuroscience, Mayo Clinic, Jacksonville, Florida, USA. 4Center for Applied Genomics, Children’s Hospital of Philadelphia, Philadelphia,
Pennsylvania, USA. 5Department of Pathology and Laboratory Medicine, University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania, USA. 6Reta
Lila Weston Institute, University College London (UCL) Institute of Neurology, London, UK. 7Department of Neurology, Division of Movement Disorders, University
of Louisville, Louisville, Kentucky, USA. 8Department of Neurology, University Hospitals, Case Western Reserve University, Cleveland, Ohio, USA. 9Department of
Neurology, Erasmus University Medical Center, Rotterdam, The Netherlands. 10Department of Clinical Genetics, Vrije Universiteit (VU) Medical Center, Section
Medical Genomics, Amsterdam, The Netherlands. 11Department of Neurology, Mayo Clinic, Jacksonville, Florida, USA. 12Department of Neurology, University
of Pennsylvania Health System, Philadelphia, Pennsylvania, USA. 13Center of Neurology, Department of Neurodegeneration, Hertie Institute for Clinical Brain
Research, University of Tübingen, Tübingen, Germany. 14German Center for Neurodegenerative Diseases, University of Tübingen, Tübingen, Germany. 15Parkinson
Institute, Istituti Clinici di Perfezionamento, Milano, Italy. 16Neurology Service, Centro de Investigación Biomédica en Red sobre Enfermedades Neurodegenerativas
(CIBERNED), Hospital Clínic, Institut d’Investigacions Biomèdiques August Pi i Sunyer (IDIBAPS), Universitat de Barcelona, Barcelona, Spain. 17Department of
Medical and Surgical Sciences, Institute of Neurology, University of Brescia, Brescia, Italy. 18CIBERNED, Instituto de Salud Carlos III, Madrid, Spain. 19Neurogenetics
laboratory, Division of Neurosciences, University of Navarra Center for Applied Medical Research, Pamplona, Spain. 20Department of Neurology, University of
Navarra, Clínica Universidad de Navarra, Pamplona, Spain. 21Laboratory of Neurogenetics, National Institute on Aging, National Institutes of Health, Bethesda,
Maryland, USA. 22Department of Neuroscience, The Scripps Research Institute, Jupiter, Florida, USA. 23Department of Medical Genetics, University of British
Columbia, Vancouver, British Columbia, Canada. 24Department of Medicine, University of Washington School of Medicine, Seattle, Washington, USA. 25Geriatric
Research, Education, and Clinical Center (GRECC), Veterans Affairs Puget Sound Health Care System, Seattle, Washington, USA. 26Department of Neurology,
University of Medicine and Dentistry of New Jersey, Robert Wood Johnson Medical School, New Brunswick, New Jersey, USA. 27Department of Molecular
Neuroscience, Queen Square Brain Bank for Neurological Disorders, UCL Institute of Neurology, University College London, London, UK. 28Institut for Humangenetik,
Justus-Liebig-Universität, Giessen, Germany. 29A list of members appears at the end of the paper. 30These authors contributed equally to this work. Correspondence
should be addressed to G.D.S. (gerardsc@mail.med.upenn.edu) or U.M. (ulrich.mueller@humangenetik.med.uni-giessen.de).
Received 29 November 2010; accepted 16 May 2011; published online 19 June 2011; doi:10.1038/ng.859

Nature Genetics VOLUME 43 | NUMBER 7 | JULY 2011 699

 letters

Table 1 Characteristics of the samplesa

Male Onset age Age at death Disease duration
Total sample Mean Mean Mean
Cohort analyzed Percent nb age Range s.d. n age Range s.d. n durationd Range s.d. n
PSP stage 1c 1,114 55 599 68 41–93 8.2 827 75 45–99 8.0 1,070 7.4 1–21 3.2 827
PSP stage 1 1,069 55 570 68 41–93 8.3 794 75 45–99 8.0 1,025 7.4 1–21 3.1 794
European ancestry
PSP stage 2e 1,051 53 530 65 40–91 7.3 913 75 57–94 7.4 118 8.0 <1–18 3.3 42
aControls were young healthy subjects recruited from the Children’s Hospital of Philadelphia Health Care Network (see Online Methods for details). There were 3,287 controls in stage 1 and 3,560
controls in stage 2. bn is the number of samples with available data. Values of n for each type of analysis do not add up to the total samples used because of missing values. cStage 1 consisted
of autopsy-confirmed cases. dDuration in years. eThe stage 2 dataset included 130 cases with autopsy data. All stage 2 samples (cases and controls) were independent of stage 1 samples.

chronic traumatic encephalopathy associated with tau deposits5. compared the control allele frequencies at significant and strongly
Viral encephalitis, associated with subsequent parkinsonism, is suggestive SNPs to those of older controls (N = 3,816) from three
also associated with tau neuropathology. In PSP, neurotoxins4 and datasets from the US National Institutes of Health (NIH) repository
low education levels6 may also contribute to risk. Genetic risk for Database for Genotypes and Phenotypes (Supplementary Table 3).
PSP is in part determined by variants at a 1-Mb inversion poly­ Only SNPs with no significant differences in allele frequencies
morphism that contains a number of genes, including MAPT, the between old and young controls are presented in Table 2.
gene that encodes tau7. The inversion variants are called H1 and H2 Stage 1 P values (P1) for SNPs in three regions crossed the signi­
‘haplotypes’, with H1 conferring risk for PSP8. H1 also contributes ficance threshold of P < 5 × 10−8 (Table 2 and Fig. 1). At 1q25.3,
to risk for corticobasal degeneration 9,10 and Guam amyotrophic a SNP in STX6 crossed this threshold (P1 = 1.8 × 10−9). Another
© 2011 Nature America, Inc. All rights reserved.

lateral sclerosis/parkinsonism dementia complex 11, both of which SNP at 3p22.1 in MOBP crossed this threshold (P1 = 1.0 × 10−9).
are rare tauopathies. H1 does not contribute to risk for Alzheimer’s The third significant region was 17q21.31, in which 58 SNPs had
disease. Notably, H1 is also a risk factor for Parkinson’s disease12, P1 < 5 × 10−8 (Table 2 and Fig. 2a). This focus of association is the
a movement disorder with clinical features that overlap those of approximately 1-Mb H1/H2 inversion polymorphism containing
PSP, yet in Parkinson’s disease there are no neuropathologically MAPT15.
recognizable tau-containing lesions. We selected SNPs for stage 2 from the original set if they yielded
We performed a genome wide association study (GWAS) of PSP P1 < 10−3. We assessed 4,099 SNPs for association in 1,051 cases,
to identify genes that modify risk for this primary tauopathy. We most of which were living subjects clinically diagnosed with PSP
­performed a two-stage analysis to maximize efficiency while maintain­ (Supplementary Table 4), and 3,560 control subjects, all of European
ing power13,14. For stage 1, we used only autopsied cases (n = 1,114), ancestry. We also included 197 ancestry informative markers16 to
thereby essentially eliminating incorrect diagnoses. These cases were evaluate population substructure. Clinically diagnosed PSP17 is rea­
contrasted with 3,287 controls; 96% of cases and 90% of controls sonably concordant with autopsy results18. We estimated the diag­
were of European ancestry (Table 1 and Supplementary Table 1). nostic misclassification rate as 12%, which has only a small impact
We assessed association between genotypes at 531,451 SNPs and on power (Online Methods).
PSP status among subjects of all ancestries (Supplementary Table 2) We replicated all three loci associated in stage 1 by joint analysis
and those of only European ancestry (Table 2 and Supplementary (Table 2 and Figs. 1 and 2). A joint analysis revealed two new loci
Fig. 1) using an additive model. Results from both ancestry groups with joint P values (PJ) below the genome-wide significance threshold.
were similar. Because our controls were younger than our cases, we One of these loci was at 2p11.2 within EIF2AK3 (PJ = 3.2 × 10−13).

Table 2 Results from stage 1, stage 2 and the joint analysis among subjects of European ancestry
Stage 1 Stage 2 Joint P
Gene or
SNP nearby MAFa MAF ORb MAF MAF
Chr. band SNP location gene case control (95% CI) P1 case control OR (95% CI) P2 OR (95% CI) PJ
1q25.3 rs1411478 179,229,155 STX6 0.50 0.42 0.73 1.8 × 10−9 0.46 0.43 0.85 1.5 × 10−3 0.79 2.3 × 10−10
(0.65–0.81) (0.77–0.94) (0.74–0.85)
2p11.2 rs7571971 88,676,716 EIF2AK3 0.31 0.26 0.75 7.4 × 10−7 0.31 0.25 0.75 8.7 × 10−8 0.75 3.2 × 10−13
(0.66–0.84) (0.67–0.83) (0.69–0.81)
3p22.1 rs1768208 39,498,257 MOBP 0.36 0.29 0.70 1.0 × 10−9 0.35 0.29 0.74 1.3 × 10−8 0.72 1.0 × 10−16
(0.63–0.79) (0.67–0.82) (0.67–0.78)
17q21.31 rs8070723 41,436,651 MAPT 0.05 0.23 5.50 2.1 × 10−51 0.06 0.23 4.74 4.8 × 10−67 5.46 1.5 × 10−116
(4.40–6.86) (3.92–5.74) (4.72–6.31)
rs242557 41,375,823 MAPT 0.53 0.35 0.48 2.2 × 10−37 0.50 0.36 0.54 5.0 × 10−35 0.51 4.2 × 10−70
(0.43–0.53) (0.48–0.59) (0.47–0.55)
rs242557/ – MAPT – – 0.66 1.3 × 10−11 – – 0.74 6.3 × 10−8 0.70 9.5 × 10−18
rs8070723c (0.58–0.74) (0.67–0.83) (0.65–0.76)
Shown are SNPs significant at P < 5 × 10−8 in the joint analysis.
aMAF, minor allele frequency. bThe OR is based on major allele. crs242557 controlling for rs8070723. P1, stage 1 P; P2, stage 2 P; PJ, joint P; STX6 encodes syntaxin 6; EIF2AK3 encodes
eukaryotic translation initiation factor 2-α kinase 3; MOBP encodes myelin-associated oligodendrocyte basic protein; MAPT encodes microtubule associated protein tau. A summary of the func-
tion of each gene listed is in Supplementary Table 9. We determined associations using an additive genetic model. Exploratory analyses (results not shown) of PSP using dominant and recessive
models did not produce new loci, although some of the associations in 17q21.31 were also consistent with these non-additive models. These less parsimonious models did not fit the data
significantly better than the additive model. By evaluating 5,000 SNPs with the smallest P values in more complicated models involving main effects and interactions38 we uncovered no
noteworthy gene-gene interactions. There were additional SNPs in the regions for the above loci that were significant or strongly suggestive for association; however, these SNPs were no longer
significant after controlling for the most significant SNP in the region (Supplementary Table 5). All loci significant in the joint analyses remained so after controlling for the MAPT inversion
(Supplementary Table 10).

700 VOLUME 43 | NUMBER 7 | JULY 2011 Nature Genetics

 letters

a 12 b 14 c

Recombination rate (cM/Mb)

Recombination rate (cM/Mb)

Recombination rate (cM/Mb)

32 rs7571971 50 18 rs1768208 55
rs1411478 Stage I 12 16
10 Stage II 26 40 44
10 14
8 Joint results
19 30 12
8 33
–log10 P

–log10 P

–log10 P
6 10
13 6 20 8
4 22
4 6
2 6 10 4 11
2
2
0 0 0 0 0 0
KIAA1614 STX6 MR1 IER5 FOXI3 EIF2AK3 RPIA SLC25A38 RPSA MOBP

179,150 179,200 179,250 179,300 179,350 88,400 88,500 88,600 88,700 88,800 88,900 39,400 39,450 39,500 39,550 39,600
1q25.3 (kb) 2p11.2 (kb) 3p22.1 (kb)

Figure 1 Regional association plots. (a) Association results for 1q25.3 STX6. (b) Association results for 2p11.2 EIF2AK3. (c) Association results for
3p22.1 MOBP regions. −log10 P values are shown for stages 1 and 2 and for the joint analyses. The recombination rate, calculated from the linkage
disequilibrium (LD) structure of the region, was derived from HapMap3 data. LD, encoded by the intensity of the colors, is the pairwise LD of the most
highly associated SNP in stage 1 with each of the SNPs in the region. Transcript positions are shown below each graph.

Another, rs12203592 (PJ = 6.2 × 10−15), at 6p25.3, highlighted IRF4, samples was comparable to the stage 1 OR, which is evidence that the
with a neighboring SNP in EXOC2, rs2493013 (PJ = 6.0 × 10−7); clinically and autopsy diagnosed cohorts are similar in composition.
rs2493013 was significant after controlling for rs12203592 at P < 1 × If all of the risk from 17q21.31 were associated with H1/H2, con­
10−3 (Supplementary Table 5). However, the allele frequencies for trolling for H1/H2 (using rs8070723 as a proxy) should be sufficient
rs12203592 and rs2493013 in older controls were significantly differ­ to make association at all other loci in this region non-significant.
ent from those of our controls (Supplementary Table 3). Curiously, That was not the case; instead, certain SNPs remained associated, with
© 2011 Nature America, Inc. All rights reserved.

the older control datasets were all significantly different from each the maximally associated SNP falling in MAPT (rs242557) (Table 2,
other. Whereas rs12203592 alleles frequencies vary widely across Fig. 2 and Supplementary Table 5). No other 17q21.31 SNPs showed
Europe19, we could not ascribe these fluctuations amongst controls association after controlling for rs8070723 and rs242557 genotypes.
to either ancestry or genotyping artifacts. In the joint analysis, three rs242557 was previously identified as a key regulatory polymorphism
other loci reached suggestive association (an intergenic region at influencing MAPT expression21. Note that rs242557 accounts for only
1q41, PJ = 2.8 × 10−7; BMS1, PJ = 4.9 × 10−7; SLCO1A2, PJ = 1.9 × part of the total risk associated with H1/H2 (Table 2).
10−7; Supplementary Table 6 and Supplementary Fig. 2). The SNPs used to detect a genome-wide association signal are not
In the MAPT region, most of the PSP-associated SNPs mapped necessarily the risk-causing variants. For STX6 and EIF2AK3, there
directly or closely to H1/H2, producing very small P values (for are non-synonymous SNPs in close proximity to and highly corre­
­example, P1 = 2.1 × 10−51 and PJ = 1.5 × 10−116 for rs8070723). H1 lated with the top genome-wide associated SNPs (Supplementary
confers risk, and 95% of PSP subject chromosomes have H1 as com­ Table 7), making these coding changes candidates for the pathogenic
pared to 77.5% of control chromosomes. In the stage 1 autopsy cases, change. To evaluate the possibility that some risk variants regulate
the odds ratio (OR) is 5.5 (95% confidence interval (CI) 4.4–6.86, gene expression, we analyzed the correlations between gene expres­
Table 2), which is stronger than the OR for the APOE ε3/ε4 genotype sion levels from two brain regions of 387 normal subjects and SNP
as a risk locus for Alzheimer’s disease20. The OR for the stage 2 PSP genotypes for the regions listed in Table 2. Two regions showed strong
genotype-expression associations (Fig. 3). SNPs falling in or near
MOBP have some effect on MOBP expression but are more strongly
a 130 50 correlated with SLC25A38 expression, which is 70 kb from MOBP
Stage I
120 Stage II
rs8070723
(Fig. 3a). This effect on SLC25A38 is seen in the cerebellum but is
Recombination rate (cM/Mb)

110 Joint results 40

100 weaker in the frontal cortex.
90
80 30
The second region showing a strong genotype-expression corre­
lation is the MAPT inversion region. SNP alleles across the entire
–log10 P

70
60
50 20 H1/H2 inversion and flanking regions show strong correlation with
40
30
not only MAPT expression (P = 8.71 × 10−28 for multiple SNPs) but
10
20 also with ARL17A (P = 9.2 × 10−22), PLEKHM1 (P = 1.0 × 10−9) and
10
0 0 LRRC37A4 (P = 2.2 × 10−35)12. Note that although MAPT expression
MAP3K14 PLEKHM1
is correlated with SNPs across the entire inversion region, the SNPs
NSF
FMNL1 ARHGAP27 MAPT WNT3

40,437 40,637 40,837 41,037 41,237 41,437 41,637 41,837 42,037 42,237 influencing ARL17A are associated with a subset of regional SNPs
17q21.31 (kb) and these are not identical to the SNPs affecting MAPT expression.
Expression of CRHR1 and KIAA1267, genes that are in the inversion
b 20 rs242557 Stage I
50
region and that flank MAPT, is not correlated with H1/H2 SNPs.
Stage II
Recombination rate (cM/Mb)

Joint results
15
40
To distinguish between the effects on gene expression of the inver­
30
sion versus other independent effects, we controlled for H1/H2, as was
–log10 P

10
20 Figure 2 Regional association results for the MAPT region of chromosome 17.
(a) Association results for the 17q21.31 H1/H2 inversion polymorphism
5
10 (40,974,015–41,926,692 kb) and flanking segments. (b) Association results
for 17q21.31 controlling for H1/H2. Results are shown for stages 1 and 2
0 0 and the joint analyses. The recombination rate, calculated from the linkage
MAP3K14 PLEKHM1 NSF
FMNL1 ARHGAP27 MAPT WNT3 disequilibrium (LD) structure of the region, was derived from HapMap3 data.
40,437 40,637 40,837 41,037 41,237 41,437 41,637 41,837 42,037 42,237 LD, encoded by intensity of the colors, is the pairwise LD of the most highly
17q21.31 (kb) associated SNP in stage 1 with each of the SNPs in the region.

Nature Genetics VOLUME 43 | NUMBER 7 | JULY 2011 701

 letters

Figure 3 Effects of genotypes on gene expression

for the MOBP region of chromosome 3 and for the
a 6
Cerebellum
6
Frontal cortex

MAPT region of chromosome 17. (a) Association 5 5

results for the relationship between SNP genotypes 4 4

–log10 P

–log10 P
and mRNA transcripts from the cerebellum and 3 3
frontal cortex for the SLC25A38-MOBP region. 2 2
(b) Association results for the relationship between 1 1
SNP genotypes and mRNA transcripts from the 0 0
cerebellum and frontal cortex for the H1/H2 –1
RPSA
SLC25A38
MOBP
–1
RPSA
SLC25A38
MOBP

inversion polymorphism region. (c) Association 39,300 39,400 39,500 39,600 39,700 39,300 39,400 39,500 39,600 39,700
results for the relationship between SNP genotypes Chr. 3 physical position (kb) Chr. 3 physical position (kb)
and mRNA transcripts from the cerebellum and
frontal cortex for the H1/H2 inversion polymorphism b 40 MapT 1710903
MapT 2310814
Inversion region
40
MapT 1710903
MapT 2310814
Inversion region
ARL 17A ARL 17A
region controlling for H1/H2. The color of the circle NSF NSF
corresponds to the color assigned to each gene, and 30 PLEKHM1 30 PLEKHM1

–log10 P

–log10 P
LRRC37A4
we tested each SNP against multiple cis transcripts.
20 20
The data presented here are independent samples
from those used previously12. 10 10

0 0
done for association with PSP (Table 2). After 40,500 41,000 41,500 42,000 42,500 40,500 41,000 41,500 42,000 42,500
controlling for H1/H2, all significant geno­ Chr. 17 physical position (kb) Chr. 17 physical position (kb)
type-expression correlation for MAPT and
© 2011 Nature America, Inc. All rights reserved.

LRRC37A4 disappeared (Fig. 3c), showing that

c 40 Inversion region
40
Inversion region

either the orientation of this region or a poly­ 30 30

–log10 P
–log10 P

morphism that maps to H1/H2 determines

MAPT expression. In contrast, controlling for 20 20

H1/H2 had no effect on ­genotype-expression 10 10

correlations for ARL17A. Potential SNPs that
0 0
affect expression (eSNPs) for ARL17A include
40,500 41,000 41,500 42,000 42,500 40,500 41,000 41,500 42,000 42,500
rs242557 (Table 2), which is highly associated Chr. 17 physical position (kb) Chr. 17 physical position (kb)
with PSP but which is more modestly corre­
lated with ARL17A expression, and rs8079215, which is highly cor­ the endoplasmic reticulum (ER) unfolded protein response (UPR).
related with ARL17A expression but not as strongly correlated with When excess unfolded proteins accumulate in the ER, PERK is acti­
risk for PSP. Statistical modeling of these data produced the following vated and protein synthesis is inhibited, allowing the ER to clear
conclusions: haplotypes involving H1 and rs242557 alleles predict a misfolded proteins and return to homeostasis. The UPR is active
highly significant portion of the variability of ARL17A expression, in PSP30, Alzheimer’s disease31 and Parkinson’s disease32. In PSP,
however, essentially all of that variance can be explained by alleles activated PERK is in neurons, oligodendrocytes and astrocytes30. In
at rs8079215, which are correlated with H1/H2 and rs242557 alleles; Alzheimer’s disease, activated UPR components are found in pre-
and that alleles at rs8079215 cannot predict risk for PSP independent tangle neurons in a number of brain regions31. In Parkinson’s disease,
of H1/H2 status even though they are excellent predictors of ARL17A UPR activation occurs in neuromelanin-containing dopaminergic
expression. In sum, risk for PSP does not rise and fall with ARL17A neurons in the substantia nigra32. How the UPR contributes to PSP
expression. The global MAPT brain region expression analyzed here pathogenesis is unclear because the primary misfolded protein in
does not explain how rs242557 alleles confer risk to PSP. Yet this SNP PSP, tau, is not a secreted protein and thus is not expected to traffic
or a correlated polymorphism is assumed to have a regulatory effect through the ER.
because there are no coding variants in MAPT brain isoforms that The PSP susceptibility gene STX6 encodes syntaxin 6 (STX6), a
are candidate pathogenic variants. One possible explanation is that SNARE-class protein. SNARE proteins are part of the cellular machin­
rs242557 alleles could affect alternative splicing without altering total ery that catalyzes the fusion of vesicles with membranes 33. STX6 is
MAPT expression levels22,23. localized to the trans-Golgi network and endosomal structures34.
Because Alzheimer’s disease and PSP are tauopathies, and because Because our work implicates ER stress in PSP pathogenesis, genetic
H1 is a shared risk factor for PSP and Parkinson’s disease, we deter­ variation at STX6 could influence movement of misfolded proteins
mined whether any confirmed Alzheimer’s disease24–28 or Parkinson’s from the ER to lysosomes through the endosomal system.
disease29 loci also produced suggestive evidence for PSP association MOBP (PJ = 1 × 10−16), like MBP, the myelin basic protein gene,
(Supplementary Table 8). Besides the overlap between Parkinson’s dis­ encodes a protein (MOBP) that is produced by oligodendrocytes and
ease and PSP at MAPT, the single noteworthy result was from rs2075650 is present in the major dense line of central nervous system myelin.
in TOMM40, which yielded PJ = 1.28 × 10−5 for association with PSP. MOBP is highly expressed in the white matter of the medulla, pons,
TOMM40 is adjacent to APOE, and rs2075650 tags the Alzheimer’s dis­ cerebellum and midbrain35, regions affected in PSP. Our findings sug­
ease risk allele, ε4, in APOE. The effect in PSP is opposite of that seen gest that myelin dysfunction or oligodendrocyte misfunction contrib­
in Alzheimer’s disease: ε4 frequency is elevated in Alzheimer’s disease utes to PSP pathogenesis.
and is diminished in PSP (for rs2075650, the estimated minor allele Our work generates a testable translational hypothesis based on the
frequency (MAF) in cases was 0.11 versus a MAF of 0.15 in both our results for EIK2AK3. Our work suggests that perturbation of the UPR
young and older controls; r2 between rs2075650 and ε4 was 0.33). can influence PSP risk, and that the UPR is not just a downstream
Our work suggests a number of intriguing insights into PSP. One consequence of neurodegeneration. Thus pharmacologic modulation
comes from EIF2AK3, a gene that encodes PERK, a component of of the UPR is a potential therapeutic strategy for PSP36,37.

702 VOLUME 43 | NUMBER 7 | JULY 2011 Nature Genetics

 letters

Methods (NIGMS). The samples used for PMRP analyses were obtained with funding from
Methods and any associated references are available in the online Marshfield Clinic, Health Resources Service Administration Office of Rural Health
Policy grant number D1A RH00025 and Wisconsin Department of Commerce
­version of the paper at http://www.nature.com/naturegenetics/. Technology Development Fund contract number TDF FYO10718. Funding
Note: Supplementary information is available on the Nature Genetics website. support for genotyping, which was performed at Johns Hopkins University, was
provided by the NIH (U01HG004438). Assistance with phenotype harmonization
Acknowledgments and genotype cleaning was provided by the eMERGE Administrative Coordinating
We thank the subjects and their families that participated in this study. This work Center (U01HG004603) and the National Center for Biotechnology Information
was funded by grants from the CurePSP Foundation, the Peebler PSP Research (NCBI). The datasets used for the analyses described in this manuscript were
Foundation and US National Institutes of Health (NIH) grants R37 AG 11762, obtained from dbGaP at http://www.ncbi.nlm.nih.gov/gap through dbGaP
R01 PAS-03-092, P50 NS72187, P01 AG17216 (National Institute on Aging accession number phs000170.v1.p1.
(NIA)/NIH), MH057881 and MH077930 (National Institute of Mental Health
(NIMH)). Work was also supported in part by the NIA Intramural Research AUTHOR CONTRIBUTIONS
Program, the German National Genome Research Network (01GS08136-4) and the Co-first authors G.U.H., N.M.M., D.W.D. and P.M.A.S. and senior authors U.M.
Deutsche Forschungsgemeinschaft (HO 2402/6-1), Prinses Beatrix Fonds (JCvS, and G.D.S. contributed equally to this project. G.U.H. and U.M. initiated this
01-0128), the Reta Lila Weston Trust and the UK Medical Research Council (RdS: study and consortium, drafted the first grant and protocol, coordinated the
G0501560). The Newcastle Brain Tissue Resource provided tissue and is funded European sample acquisition and preparation, contributed to data interpretation
in part by a grant from the UK Medical Research Council (G0400074), by the and contributed to the preparation of the manuscript. N.M.M. conducted the
Newcastle National Institute for Health Research (NIHR) Biomedical Research analyses and contributed to the preparation of the manuscript. D.W.D. contributed
Centre in Ageing and Age Related Diseases to the Newcastle upon Tyne Hospitals to study design, data interpretation and preparation of the manuscript. P.M.A.S.
National Health Service Foundation Trust and by a grant from the Alzheimer’s contributed in the selection of controls for both phases of the experiment, data
Society and Alzheimer’s Research Trust as part of the Brains for Dementia Research quality control, data analysis and content curation for the replication phase custom
Project. We acknowledge the contribution of many tissue samples from the array. L.-S.W. participated in the initial association analysis, eSNP and pathway
Harvard Brain Tissue Resource Center. We also acknowledge the ‘Human Genetic analysis and functional annotation of SNPs in the top genes. L.K. participated
© 2011 Nature America, Inc. All rights reserved.

Bank of Patients affected by Parkinson Disease and Parkinsonism’ (http://www. in genotype quality control and analysis. R.R. and R.d.S. participated in study
parkinson.it/dnabank.html) of the Telethon Genetic Biobank Network, supported design, sample preparation and revising the manuscript for content. I. Litvan,
by TELETHON Italy (project no. GTB07001) and by Fondazione Grigioni per il D.E.R., J.C.V.S., P.H., Z.K.W., R.J.U., J.V., H.I.H., R.G.G., W.M., S.G., E.T., B.B.,
Morbo di Parkinson. The University of Toronto sample collection was supported P.P. and the PSP Genetics Study Group (R.L.A., E.A., A.A., M.A., S.E.A., J.A., T.B.,
by grants from Wellcome Trust, Howard Hughes Medical Institute and the S.B., D.B., T.D.B., N.B., A.J.W.B., Y.B., A.B., H.B., M.C., W.Z.C., R.C., C.C., P.P.D.,
Canadian Institute of Health Research. Brain-Net-Germany is supported by the J.G.D., L.D.K., R.D., A. Durr, S.E., G.F., N.A.F., R.F., M.P.F., C.G., D.R.G., T.G.,
Federal Ministry of Education and Research (BMBF) (01GI0505). R.d.S., A.J.L. and M. Gearing, E.T.G., B.G., N.R.G.R., M. Grossman, D.A.H., L.H., M.H., J.J., J.L.J.,
J.A.H. are funded by the Reta Lila Weston Trust and the PSP (Europe) Association. A.K., H.A.K., I. Leber, V.M.L., A.P.L., K.L., C. Mariani, E.M., L.A.M., C.A.M.,
R.d.S. is funded by the UK Medical Research Council (Grant G0501560) and N.M., B.L.M., B.M., J.C.M., H.R.M., C. Morris, S.S.O., W.H.O., D.O., A.P., R.P.,
Cure PSP+. Z.K.W. is partially supported by the NIH/NINDS 1RC2NS070276, G.P., S.P.B., W.P., A. Rabano, A. Rajput, S.G.R., G.R., S.R., J.D.R., O.A.R., M.N.R.,
NS057567, P50NS072187, Mayo Clinic Florida (MCF) Research Committee CR G.S., W.W.S., K. Seppi, L.S.M., S.S., K. Srulijes, P.S.G., M.S., D.G.S., S.T., W.W.T.,
programs (MCF #90052030 and MCF #90052030) and the gift from C.E. Bolch C. Trenkwalder, C. Troakes, J.Q.T., J.C.T., V.M.V., J.P.G.V., G.K.W., C.L.W., P.W.,
Jr. and S.B. Bolch (MCF #90052031/PAU #90052). The Mayo Clinic College of C.Z. and A.L.Z.) participated in characterization, preparation and contribution
Medicine would like to acknowledge M. Baker, R. Crook, M. DeJesus-Hernandez of samples from individuals with PSP. L.B.C. coordinated the project, sample
and N. Rutherford for their preparation of samples. P.P. was supported by a grant acquisition and selection and managed phenotypes. M.R.H. conducted eSNP
from the Government of Navarra (‘Ayudas para la Realización de Proyectos de and pathway analysis. A. Dillman performed mRNA expression experiments in
Investigación’ 2006–2007) and acknowledges the ‘Iberian Atypical Parkinsonism human brain. M.P.v.d.B. and D.G.H. performed mRNA expression experiments
Study Group Researchers’: M.A. Pastor, M.R. Luquin, M. Riverol, J.A. Obeso in human brain and contributed to the design of eQTL experiments. J.R.G.
and M.C. Rodriguez-Oroz (Department of Neurology, Clínica Universitaria performed computational and statistical analysis of the eQTL data and contributed
de Navarra, University of Navarra, Pamplona, Spain), M. Blazquez (Neurology to the design of eQTL experiments. M.R.C. and A.B.S. were responsible for overall
Department, Hospital Universitario Central de Asturias, Oviedo, Spain), A. Lopez supervision, design and analysis of eQTL experiments. J.C.V.S., M.J.F., L.I.G., J.H.
de Munain, B. Indakoetxea, J. Olaskoaga, J. Ruiz, J. Félix Martí Massó (Servicio and A.J.L. participated in study design and data analysis discussions. C.-E.Y. and
de Neurología, Hospital Donostia, San Sebastián, Spain), V. Alvarez (Genetics T.R. participated in the initial design of experiments. B.D. supervised analyses
Department, Hospital Universitario Central de Asturias, Oviedo, Spain), T. Tuñon and contributed to the writing of the manuscript. H.H. supervised genotyping
(Banco de Tejidos Neurologicos, CIBERNED, Hospital de Navarra, Navarra, and platform and sample selection, participated in analyses and reviewed the
Spain), F. Moreno (Servicio de Neurología, Hospital Ntra. Sra. de la Antigua, manuscript. G.D.S. led the consortium, supervised study design, coordinated the
Zumarraga, Gipuzkoa, Spain), A. Alzualde (Neurogenétics Department, Hospital US sample acquisition and preparation, contributed to data interpretation and
Donostia, San Sebastián, Spain). E.T. wishes to acknowledge the Banco de Tejidos wrote and coordinated assembly of the manuscript.
Neurológicos de la Universidad de Barcelona-Hospital Clinic, which provided
many tissue samples for the project. We also acknowledge E. Loomis for providing COMPETING FINANCIAL INTERESTS
technical support. The authors declare competing financial interests: details accompany the full-text
The datasets used for older controls were obtained from Database for Genotypes HTML version of the paper at http://www.nature.com/naturegenetics/.
and Phenotypes (dbGap) at http://www.ncbi.nlm.nih.gov/gap/. Funding support
for the ‘Genetic Consortium for Late Onset Alzheimer’s Disease’ was provided Published online at http://www.nature.com/naturegenetics/.
through the Division of Neuroscience, NIA. The Genetic Consortium for Late Reprints and permissions information is available online at http://www.nature.com/
Onset Alzheimer’s Disease (study accession number: phs000168.v1.p1) includes reprints/index.html.
a genome-wide association study funded as part of the Division of Neuroscience,
NIA. Assistance with phenotype harmonization and genotype cleaning, as well as
1. Hoppitt, T. et al. A systematic review of the incidence and prevalence of long-term
with general study coordination, was provided by Genetic Consortium for Late neurological conditions in the UK. Neuroepidemiology 36, 19–28 (2011).
Onset Alzheimer’s Disease. Funding support for the ‘CIDR Visceral Adiposity 2. Litvan, I. Update on progressive supranuclear palsy. Curr. Neurol. Neurosci. Rep.
Study’ (study accession number: phs000169.v1.p1) was provided through the 4, 296–302 (2004).
Division of Aging Biology and the Division of Geriatrics and Clinical Gerontology, 3. Dickson, D.W., Rademakers, R. & Hutton, M.L. Progressive supranuclear palsy:
NIA. The CIDR Visceral Adiposity Study includes a genome-wide association pathology and genetics. Brain Pathol. 17, 74–82 (2007).
study funded as part of the Division of Aging Biology and the Division of Geriatrics 4. Stamelou, M. et al. Rational therapeutic approaches to progressive supranuclear
and Clinical Gerontology, NIA. Assistance with phenotype harmonization and palsy. Brain 133, 1578–1590 (2010).
5. McKee, A.C. et al. TDP-43 proteinopathy and motor neuron disease in chronic
genotype cleaning, as well as with general study coordination, was provided by
traumatic encephalopathy. J. Neuropathol. Exp. Neurol. 69, 918–929 (2010).
Heath ABC Study Investigators. Funding support for the Personalized Medicine 6. Golbe, L.I. et al. Follow-up study of risk factors in progressive supranuclear palsy.
Research Project (PMRP) was provided through a cooperative agreement Neurology 47, 148–154 (1996).
(U01HG004608) with the National Human Genome Research Institute (NHGRI), 7. Stefansson, H. et al. A common inversion under selection in Europeans. Nat. Genet.
with additional funding from the National Institute for General Medical Sciences 37, 129–137 (2005).

Nature Genetics VOLUME 43 | NUMBER 7 | JULY 2011 703

 letters

8. Baker, M. et al. Association of an extended haplotype in the tau gene with progressive 23. Myers, A.J. et al. The MAPT H1c risk haplotype is associated with increased
supranuclear palsy. Hum. Mol. Genet. 8, 711–715 (1999). expression of tau and especially of 4 repeat containing transcripts. Neurobiol. Dis.
9. Cruchaga, C. et al. 5′-upstream variants of CRHR1 and MAPT genes associated 25, 561–570 (2007).
with age at onset in progressive supranuclear palsy and cortical basal degeneration. 24. Harold, D. et al. Genome-wide association study identifies variants at CLU and
Neurobiol. Dis. 33, 164–170 (2009). PICALM associated with Alzheimer’s disease. Nat. Genet. 41, 1088–1093 (2009).
10. Houlden, H. et al. Corticobasal degeneration and progressive supranuclear palsy 25. Lambert, J.C. et al. Genome-wide association study identifies variants at CLU and
share a common tau haplotype. Neurology 56, 1702–1706 (2001). CR1 associated with Alzheimer’s disease. Nat. Genet. 41, 1094–1099 (2009).
11. Sundar, P.D. et al. Two sites in the MAPT region confer genetic risk for Guam 26. Seshadri, S. et al. Genome-wide analysis of genetic loci associated with Alzheimer
ALS/PDC and dementia. Hum. Mol. Genet. 16, 295–306 (2007). disease. J. Am. Med. Assoc. 303, 1832–1840 (2010).
12. Simon-Sanchez, J. et al. Genome-wide association study reveals genetic risk 27. Naj, A.C. et al. Common variants at MS4A4/MS4A6E, CD2AP, CD33 and EPHA1
underlying Parkinson’s disease. Nat. Genet. 41, 1308–1312 (2009). are associated with late-onset Alzheimer’s disease. Nat. Genet. 43, 436–441
13. Chanock, S.J. et al. Replicating genotype-phenotype associations. Nature 447, (2011).
655–660 (2007). 28. Hollingworth, P. et al. Common variants at ABCA7, MS4A6A/MS4A4E, EPHA1,
14. Skol, A.D., Scott, L.J., Abecasis, G.R. & Boehnke, M. Joint analysis is more efficient CD33 and CD2AP are associated with Alzheimer’s disease. Nat. Genet. 43,
than replication-based analysis for two-stage genome-wide association studies. Nat. 429–435 (2011).
Genet. 38, 209–213 (2006). 29. Nalls, M.A. et al. Imputation of sequence variants for identification of genetic risks
15. Zody, M.C. et al. Evolutionary toggling of the MAPT 17q21.31 inversion region. for Parkinson’s disease: a meta-analysis of genome-wide association studies. Lancet
Nat. Genet. 40, 1076–1083 (2008). 377, 641–649 (2011).
16. Tian, C. et al. Analysis and application of European genetic substructure using 300 30. Unterberger, U. et al. Endoplasmic reticulum stress features are prominent in
K SNP information. PLoS Genet. 4, e4 (2008). Alzheimer disease but not in prion diseases in vivo. J. Neuropathol. Exp. Neurol.
17. Litvan, I. et al. Clinical research criteria for the diagnosis of progressive supranuclear 65, 348–357 (2006).
palsy (Steele-Richardson-Olszewski syndrome): report of the NINDS-SPSP 31. Hoozemans, J.J.M. et al. The unfolded protein response is activated in pretangle neurons
international workshop. Neurology 47, 1–9 (1996). in Alzheimer’s disease hippocampus. Am. J. Pathol. 174, 1241–1251 (2009).
18. Osaki, Y. et al. Accuracy of clinical diagnosis of progressive supranuclear palsy. 32. Hoozemans, J.J. et al. Activation of the unfolded protein response in Parkinson’s
Mov. Disord. 19, 181–189 (2004). disease. Biochem. Biophys. Res. Commun. 354, 707–711 (2007).
19. Duffy, D.L. et al. Multiple pigmentation gene polymorphisms account for a 33. Jahn, R. & Scheller, R.H. SNAREs–engines for membrane fusion. Nat. Rev. Mol.
substantial proportion of risk of cutaneous malignant melanoma. J. Invest. Dermatol. Cell Biol. 7, 631–643 (2006).
© 2011 Nature America, Inc. All rights reserved.

130, 520–528 (2010). 34. Wendler, F. & Tooze, S. Syntaxin 6: the promiscuous behaviour of a SNARE protein.
20. Farrer, L.A. et al. Effects of age, sex, and ethnicity on the association between Traffic 2, 606–611 (2001).
apolipoprotein E genotype and Alzheimer disease. A meta-analysis. APOE and 35. Montague, P., McCallion, A.S., Davies, R.W. & Griffiths, I.R. Myelin-associated
Alzheimer Disease Meta Analysis Consortium. J. Am. Med. Assoc. 278, 1349–1356 oligodendrocytic basic protein: a family of abundant CNS myelin proteins in search
(1997). of a function. Dev. Neurosci. 28, 479–487 (2006).
21. Rademakers, R. et al. High-density SNP haplotyping suggests altered regulation of 36. Scheper, W. & Hoozemans, J.J.M. Endoplasmic reticulum protein quality control in
tau gene expression in progressive supranuclear palsy. Hum. Mol. Genet. 14, neurodegenerative disease: the good, the bad and the therapy. Curr. Med. Chem.
3281–3292 (2005). 16, 615–626 (2009).
22. Caffrey, T.M., Joachim, C., Paracchini, S., Esiri, M.M. & Wade-Martins, R. Haplotype- 37. Paschen, W. & Mengesdorf, T. Cellular abnormalities linked to endoplasmic reticulum
specific expression of exon 10 at the human MAPT locus. Hum. Mol. Genet. 15, dysfunction in cerebrovascular disease—therapeutic potential. Pharmacol. Ther.
3529–3537 (2006). 108, 362–375 (2005).

Roger L Albin31,32, Elena Alonso33,34, Angelo Antonini35,36, Manuela Apfelbacher37, Steven E Arnold38,
Jesus Avila39, Thomas G Beach40, Sherry Beecher41, Daniela Berg42, Thomas D Bird43, Nenad Bogdanovic44,
Agnita J W Boon45, Yvette Bordelon46, Alexis Brice47–49, Herbert Budka50, Margherita Canesi35,
Wang Zheng Chiu45, Roberto Cilia35, Carlo Colosimo51, Peter P De Deyn52, Justo García de Yebenes53,
Laura Donker Kaat45, Ranjan Duara54, Alexandra Durr47–49, Sebastiaan Engelborghs52, Giovanni Fabbrini51,
NiCole A Finch55, Robyn Flook56, Matthew P Frosch57, Carles Gaig58, Douglas R Galasko59, Thomas Gasser42,
Marla Gearing60, Evan T Geller41, Bernardino Ghetti61, Neill R Graff-Radford62, Murray Grossman63,
Deborah A Hall64, Lili-Naz Hazrati65, Matthias Höllerhage66, Joseph Jankovic67, Jorge L Juncos68,
Anna Karydas69, Hans A Kretzschmar70, Isabelle Leber47–49, Virginia M Lee41, Andrew P Lieberman71,
Kelly E Lyons72, Claudio Mariani35, Eliezer Masliah59,73, Luke A Massey74, Catriona A McLean75,
Nicoletta Meucci35, Bruce L Miller69, Brit Mollenhauer76,77, Jens C Möller66, Huw R Morris78, Chris Morris79,
Sean S O’Sullivan74, Wolfgang H Oertel66, Donatella Ottaviani51, Alessandro Padovani80, Rajesh Pahwa72,
Gianni Pezzoli35, Stuart Pickering-Brown81, Werner Poewe82, Alberto Rabano83, Alex Rajput84, Stephen G Reich85,
Gesine Respondek66, Sigrun Roeber70, Jonathan D Rohrer86, Owen A Ross55, Martin N Rossor86,
Giorgio Sacilotto35, William W Seeley69, Klaus Seppi82, Laura Silveira-Moriyama74, Salvatore Spina61,
Karin Srulijes42, Peter St. George-Hyslop65,87, Maria Stamelou66, David G Standaert88, Silvana Tesei35,
Wallace W Tourtellotte89, Claudia Trenkwalder77, Claire Troakes90, John Q Trojanowski41, Juan C Troncoso91,
Vivianna M Van Deerlin41, Jean Paul G Vonsattel92, Gregor K Wenning82, Charles L White93, Pia Winter94,
Chris Zarow95 & Anna L Zecchinelli35

704 VOLUME 43 | NUMBER 7 | JULY 2011 Nature Genetics

 letters

31Department of Neurology, University of Michigan, Ann Arbor, Michigan, USA. 32Geriatrics Research, Education, and Clinical Center, Veterans Affairs (VA) Ann Arbor
Health System, Ann Arbor, Michigan, USA. 33CIBERNED, Instituto de Salud Carlos III, Madrid, Spain. 34Neurogenetics Laboratory, Division of Neurosciences,
University of Navarra Center for Applied Medical Research, Pamplona, Spain. 35Parkinson Institute, Istituti Clinici di Perfezionamento, Milan, Italy. 36Department for
Parkinson’s Disease, Istituto Di Ricovero e Cura a Carattere Scientifico (IRCCS) San Camillo, Venice, Italy. 37Institute of Legal Medicine, University of Würzburg,
Würzburg, Germany. 38Department of Psychiatry, Center for Neurobiology and Behavior, University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania,
USA. 39Centro de Biologia Molecular Severo Ochoa (CSIC-UAM), Campus Cantoblanco, Universidad Autonoma de Madrid, Madrid, Spain. 40Civin Laboratory for
Neuropathology, Banner Sun Health Research Institute, Sun City, Arizona, USA. 41Department of Pathology and Laboratory Medicine, University of Pennsylvania
School of Medicine, Philadelphia, Pennsylvania, USA. 42Center of Neurology, Department of Neurodegeneration, Hertie Institute for Clinical Brain Research,
University of Tübingen and German Center for Neurodegenerative diseases (DZNE), Tübingen, Germany. 43Geriatrics Research Education and Clinical Center, Veterans
Affairs Puget Sound Health Care System, Seattle, WA, USA. 44Department of Neurobiology, Care Sciences and Society, Karolinska Institutet, Hudding University
Hospital, Stockholm, Sweden. 45Department of Neurology, Erasmus University Medical Center, Rotterdam, The Netherlands. 46Department of Neurology, University of
California Los Angeles, Los Angeles, California, USA. 47Centre de Recherche de l’Institut du Cerveau et de la Moelle épinière, Université Pierre et Marie Curie, Paris,
France. 48Institut National de la Santé et de la Recherche Médicale, Paris, France. 49Centre National de la Recherche Scientifique, Paris, France. 50Institute of
Neurology, Medical University Vienna, Vienna, Austria. 51Dipartimento di Scienze Neurologiche e Psichiatriche, Sapienza Università di Roma, Rome, Italy.
52Department of Biomedical Sciences, University of Antwerp, Antwerp, Belgium. 53Department of Neurology, Hospital Ramón y Cajal, Madrid, Spain. 54Wien Center

for Alzheimer’s Disease and Memory Disorders, Mt. Sinai Medical Center, Miami Beach, Florida, USA. 55Department of Neuroscience, Mayo Clinic, Jacksonville,
Florida, USA. 56Centre for Neuroscience, Flinders University and Australian Brain Bank Network, Victoria, Australia. 57C.S. Kubik Laboratory for Neuropathology,
Massachusetts General Hospital and Harvard Medical School, Boston, Massachusetts, USA. 58Neurology Service, Centro de Investigación Biomédica en Red sobre
Enfermedades Neurodegenerativas (CIBERNED), Hospital Clínic, IDIBAPS, Universitat de Barcelona, Barcelona, Spain. 59Department of Neurosciences, University of
California San Diego, La Jolla, California, USA. 60Department of Pathology and Laboratory Medicine, Emory University School of Medicine, Atlanta, Georgia, USA.
61Department of Pathology and Laboratory Medicine, Indiana University School of Medicine, Indianapolis, Indiana, USA. 62Department of Neurology, Mayo Clinic,

Jacksonville, Florida, USA. 63Department of Neurology, University of Pennsylvania Health System, Philadelphia, Pennsylvania, USA. 64Department of Neurological
Sciences, Rush University, Chicago, Illinois, USA. 65Tanz Centre for Research in Neurodegenerative Disease, University of Toronto, Toronto, Ontario, Canada.
66Department of Neurology, Philipps University, Marburg, Germany. 67Department of Neurology, Baylor College of Medicine, Houston, Texas, USA. 68Department of

Neurology, Emory University, Atlanta, Georgia, USA. 69Department of Neurology, Memory and Aging Center, University of California, San Francisco, San Francisco,
© 2011 Nature America, Inc. All rights reserved.

California, USA. 70Institut für Neuropathologie, Ludwig-Maximilians-Universität and Brain Net Germany, Munich, Germany. 71Department of Pathology, University of
Michigan Medical School, Ann Arbor, Michigan, USA. 72Department of Neurology, University of Kansas Medical Center, Kansas City, Kansas, USA. 73Department of
Pathology, University of California San Diego, La Jolla, California, USA. 74Reta Lila Weston Institute, UCL Institute of Neurology, University College London, London,
UK. 75Victorian Brain Bank Network, Mental Health Research Institute, Victoria, Australia. 76Department of Neurology, Georg-August University, Goettingen, Germany.
77Paracelsus-Elena-Klinik, University of Goettingen, Kassel, Germany. 78Medical Research Council (MRC) Centre for Neuropsychiatric Genetics and Department of

Neurology, School of Medicine, Cardiff University, Cardiff, UK. 79Newcastle Brain Tissue Resource, Newcastle University, Institute for Ageing and Health, Newcastle
upon Tyne, UK. 80Department of Medical and Surgical Sciences, Institute of Neurology, University of Brescia, Brescia, Italy. 81Neurodegeneration and Mental Health
Research Group, Faculty of Human and Medical Sciences, University of Manchester, Manchester, UK. 82Department of Neurology, Innsbruck Medical University,
Innsbruck, Austria. 83Department of Neuropathology and Tissue Bank, Fundación Centro Investigación Enfermedades Neurológicas (CIEN), Instituto de Salud Carlos
III, Madrid, Spain. 84Division of Neurology, Royal University Hospital, University of Saskatchewan, Saskatchewan, Canada. 85Department of Neurology, University of
Maryland School of Medicine, Baltimore, Maryland, USA. 86Department of Neurodegenerative Disease, Dementia Research Centre, UCL Institute of Neurology, UCL,
London, UK. 87Cambridge Institute for Medical Research and Department of Clinical Neurosciences, University of Cambridge, Cambridge, UK. 88Department of
Neurology, Center for Neurodegeneration and Experimental Therapeutics, University of Alabama at Birmingham, Birmingham, Alabama, USA. 89Human Brain and
Spinal Fluid Resource Center, Veterans Affairs West Los Angeles Healthcare Center, Los Angeles, California, USA. 90Department of Clinical Neuroscience, MRC Centre
for Neurodegeneration Research, King’s College London, London, UK. 91Department of Neurology, Johns Hopkins University School of Medicine, Baltimore, Maryland,
USA. 92Department of Pathology and the Taub Institute for Research on Alzheimer’s Disease and the Aging Brain, Columbia University, New York, New York, USA.
93Department of Pathology, University of Texas Southwestern Medical Center, Dallas, Texas, USA. 94Institute of Human Genetics, Justus-Liebig University, Giessen,

Germany. 95Rancho Los Amigos National Rehabilitation Center, University of Southern California, Downey, California, USA.

Nature Genetics VOLUME 43 | NUMBER 7 | JULY 2011 705

 ONLINE METHODS rate >99.9% and with an inter-SNP distance more than 0.4 Mb. SpectralGEM40
Sample. The stage 1 sample consisted of 1,150 autopsy-confirmed cases and estimated eight significant dimensions of ancestry (Supplementary Fig. 3).
3,658 population-based controls. Individuals with clinical parkinsonism in Subsequent clustering on these dimensions resulted in 11 clusters: 4 clusters of
life and neuropathological confirmation of PSP were identified from brain European ancestry with N = 691, N = 682, N = 939 and N = 1,116 ­individuals
banks, research hospitals and neuropathologists (Supplementary Table 1). (Supplementary Fig. 4). The other clusters reflected other major ancestral
The neuropathological diagnosis was made according to the NINDS diag­ groups, largely of African and African-American ancestry. Using the samples
nostic criteria38,39. DNA was extracted from brain tissue from cases who had of European ancestry to determine MAF and evaluate HWE, 24,132 SNPs were
consented for brain donation. removed because MAF < 0.01 and 1,919 failed HWE.
The stage 2 sample consisted of 1,244 individuals diagnosed with PSP and As determined by genotypes, all cases and controls for stage 2 were inde­
3,654 control subjects, none of whom were included in stage 1. The majority pendent of those used in stage 1. Of the 5,283 genotyped SNPs, 4,617 could be
of cases were clinically diagnosed (1,113 cases), although some were auto­ aligned with control genotypes: most (3,203 SNPs) were identified by CHOP’s
psied (131 cases). Clinical cases were recruited through specialized movement or the University of Pittsburgh’s stage 1 analyses as SNPs showing log-additive
disorder hospitals and clinics and referrals from neurologists (Supplementary association at P1 ≤ 0.001, as well as a subset of SNPs correlated with the original
Table 4). All cases fulfilled the clinical diagnosis of possible or probable PSP17. associated SNPs; AIMs (197)42; or SNPs identified by exploratory modeling
The diagnostic assessments involved a review of the individual’s full medical to identify recessive and dominant or interactive risk loci (896). Sixty control
history, neurological examinations and mental status evaluations. The demo­ samples were genotyped on both BeadChip and iSelect platforms to assess
graphic data included gender, age at onset of motor symptoms, disease dura­ concordance, which was 100%.
tion and final clinical diagnosis. Informed consent was obtained for blood Controls were selected from a larger CHOP control dataset to match cases
collection and genetic analysis. The University of Pennsylvania assembled all in terms of genetic ancestry (by P.M.S.). To determine ancestry, multilocus
DNA samples identified for the study. genotype data were decomposed by eigenvector analysis as implemented in
Controls for stage 1 and 2 were young subjects recruited from the Children’s Eigensoft43 into a small set of vectors representing ancestry using smartpca, a
Hospital of Philadelphia (CHOP) Health Care Network by CHOP clinicians part of the EIGENSTRAT package. To do so, a trimmed set of 3,579 SNPs was
© 2011 Nature America, Inc. All rights reserved.

and nursing staff. Written informed consent was obtained from all subjects. selected from those on the iSelect panel, excluding all SNPs in LD (pairwise
The stage 1 cohort was largely of European ancestry (89.5%), whereas all of r2 < 0.5) and all SNPs in the 17q12 inversion. After eigenvector analysis, cases
the stage 2 controls were of European ancestry. Females comprised 47.7% and were 1:3 matched to controls using three principal components and a distance
47.0% of the stage 1 and 2 samples, respectively; the mean ages for the stage 1 threshold of 0.071 (ref. 44).
and 2 cohorts were 7.8 years and 8.8 years, respectively. The advantage of
using these controls is that all were genotyped at the same center using the Association analysis. In stage 1, we contrasted the genotypes at 531,451 SNPs
same protocols as the cases. Although the controls were not selected for an of PSP subjects and genetically matched controls using conditional logistic
absence of neurodegenerative disease, the low population frequency of PSP regression and a log-additive model. Full matching44 of cases and controls
ensures a negligible number of controls will get PSP later in life. The disadvan­ resulted in 1,114 strata, each of which contained one case and one or more
tage of young controls is that some loci could also have a positive or negative controls. To analyze chromosome X, the data were matched conditional on
impact on survival. To diminish this disadvantage, for SNPs significantly or gender. After removing individuals not in close proximity to a discordant (in
suggestively associated with PSP, we compared the control allele frequencies diagnosis) individual of the same gender, there were 628 PSP cases and 1,686
of our younger samples with those from a set of older controls (N = 3,816) controls in the male-only set and 485 PSP cases and 1,441 controls in the
obtained from the NIH repository Database for Genotypes and Phenotypes. female-only set for a total of 1,113 cases and 3,127 controls. Cases were full
Only SNPs for which there was no difference between young and old frequen­ matched in 1,113 distinct strata, each of which contained one case and one
cies were reported in Table 2. or more multiple controls of the same gender. For the analysis of subjects of
European ancestry, data consisted of 1,069 PSP cases and 2,964 controls.
Genotyping. Stage 1 cases were genotyped by the Center for Applied Genomics Primary analyses were conducted among subjects of European ancestry;
at CHOP using Human 660W-Quad Infinium BeadChips. Control samples the resulting quantile-quantile plot is shown in Supplementary Figure 5.
were genotyped using the Illumina Human HapMap550 Infinium BeadChip. Genotypes for any SNPs showing genome-wide significant association, or
Stage 2 cases were genotyped for 5,283 SNPs using Infinium HD iSelect Custom nearly so, were manually inspected for valid genotype clustering. SNPs show­
BeadChip, most of which were identified by CHOP or at the University of ing poor clustering were excluded. Exploratory analyses included domi­
Pittsburgh at stage 1 as SNPs showing association by the log-additive model at nant and recessive models, as well as evaluation of gene-gene interaction
P ≤ 0.001. Ancestry informative markers, or AIMs, were also included. Stage 2 by using a model selection procedure called ‘screen and clean’45. Analyses
controls were selected from a larger CHOP control dataset to match cases in using subjects of any ancestry were also conducted and are reported in
terms of genetic ancestry. Supplementary Table 2.
At stage 2, association of SNP genotype and diagnostic status was assessed
Quality control. Quality control procedures were performed at the individual for 4,099 SNPs remaining after quality control, again by using conditional
and then at the SNP level. At the individual level, gender miscalls based on logistic regression. After the quality control analysis described above, 1,051
chromosome X and Y genotypes, duplicate samples and highly related sam­ PSP cases and 3,560 controls remained for the association analysis, all of
ples were excluded. At the SNP level, a genotype completion rate of ≥98% was European ancestry.
required. Hardy-Weinberg equilibrium (HWE) was evaluated in samples of
European ancestry, and SNPs failing HWE were excluded (P < 1 × 10−4). To Study design and statistical significance. Following the definition previously
control potential confounding caused by variation in genetic ancestry, cases used14, markers considered genome-wide significant were those that met the
and controls were matched for ancestry based on genetic data using methods threshold for genome-wide significance from the joint analysis of stages 1 and 2.
previously described40,41. All quality control and subsequent association analy­ We took P ≤ 5 × 10−8 (joint z ≥ 5.44) as genome-wide significant and 5.7 × 10−7
ses were performed independently at CHOP by P.M.S. and at the University ≥ P > 5 × 10−8 as strongly suggestive (joint z = 5.0 for the larger bound).
of Pittsburgh by N.M.M and L.K. Results were then compared for agreement.
The analytic methods and results described are those used at the University Effect of diagnostic misclassification on power. Samples from stage 1 con­
of Pittsburgh unless otherwise noted. sisted of autopsy-confirmed subjects, whereas a large fraction of the stage 2
For stage 1, there were three gender inconsistencies (one case and two con­ samples were diagnosed clinically. For stage 2, then, a fraction 1 – π of subjects
trols), 32 samples were duplicates (all cases), and 12 samples had genotyping could be misdiagnosed as having PSP when they have some other diseases,
completion rate <98%. These samples were eliminated. For SNPs, 4,222 were often Parkinson disease. Here we assume all misdiagnoses are Parkinson’s
monomorphic, had uncalled genotypes or had genotyping completion rate <98%. disease, whereas in truth it is most likely a mixture of diseases; most of these
Ancestry for cases and controls was determined using 6,490 SNPs with a call diseases, just like Parkinson’s disease, are unlikely to share risk loci with PSP.

Nature Genetics doi:10.1038/ng.859

 For this reason, for most SNPs, denoted by j, the expected odds ratio OR = 1 RNA from the cerebellum (CRBLM) and frontal cortex (FCTX) was
and θ = log(ORPDj), which is assumed to be normally distributed with mean 0 obtained, and expression was measured using Illumina HumanHT-12 v3
and variance σPD2, the usual variance of the log odds ratio. Expression BeadChips. DNA was genotyped using the Illumina genotyping
The distribution of the data in stage 2 is a normal mixture. As noted above, arrays (HumanHap550 v3, Human610-Quad v1 or Human660W-Quad v1).
typically θPDj is 0; however, the inversion polymorphism of the MAPT locus Processing and analysis of these data were performed in a manner similar
is known to be a risk factor for PSP and Parkinson’s disease, with an estimated to methods previously described for a study including 150 of these 387 sub­
OR of 5.6 for PSP, stage 1 (θP = 1.72), 4.7 for PSP, stage 2 (θP = 1.56), and 1.32 ject46. In brief, expression profiles were adjusted for the following covariates:
(θP = 0.28) for Parkinson’s disease12. Of note, the expected value of θ at stage 2 subject age, gender, post-mortem interval, originating tissue bank, principal
for any locus j is πθPj + (1 – π)θPDj. This observation gives us a way to estimate components 1 and 2 based on identity-by-state pairwise genetically estimated
the fraction of misdiagnosis 1 – π by using the estimated parameters for the distances between subjects47, and the mRNA sample preparation and hybridi­
MAPT locus and the method of moments. Plugging in estimates of log(OR) zation batch. eQTL analysis was then performed using linear regression of the
and solving for 1 – π, our estimate is 0.12, meaning that 12% of the stage 2 residuals for every trait in each brain tissue with the allele dosage. For each trait
sample is likely to have Parkinson’s disease. Note that if the misclassification analyzed, only SNPs that were cis to the trait were considered in the analysis,
were not all because of Parkinson’s disease, it would likely diminish the esti­ where cis is the region near the gene ± 1 Mb from the mRNA transcript start
mated fraction of subjects misclassified. or end site. SNPs associated with PSP by joint analysis were then evaluated for
We can also use this result to ask what impact 12% misclassification would their impact on expression of genes proximate to the SNPs.
have on the OR in stage 2. For a true OR = 1.4 for a pure PSP sample, the
38. Hauw, J.J. et al. Preliminary NINDS neuropathologic criteria for Steele-Richardson-
adjusted odds ratio, assuming the locus has no impact on risk for Parkinson’s Olszewski syndrome (progressive supranuclear palsy). Neurology 44, 2015–2019
disease or other misclassified diagnoses, would be 1.34; for an OR = 1.3, it (1994).
would be 1.26. Because we are more concerned about the OR from the joint 39. Dickson, D.W. et al. Office of rare diseases neuropathologic criteria for corticobasal
analysis of stages 1 and 2, an OR = 1.4 is expected to be reduced to 1.37 in the degeneration. J. Neuropathol. Exp. Neurol. 61, 935–946 (2002).
40. Lee, A.B., Luca, D., Klei, L., Devlin, B. & Roeder, K. Discovering genetic ancestry
joint analysis and an OR = 1.3 would be reduced to 1.28 by 12% misclassifica­
© 2011 Nature America, Inc. All rights reserved.

using spectral graph theory. Genet. Epidemiol. 34, 51–59 (2010).

tion. Applying the CaTS power calculator14 for an ∝ = 5 × 10−8 and an OR 41. Crossett, A. et al. Using ancestry matching to combine family-based and unrelated
reduction from 1.4 to 1.37, the power for the joint analysis is reduced from samples for genome-wide association studies. Stat. Med. 29, 2932–2945 (2010).
97% to 92%; when reducing the OR from 1.3 to 1.28, the power is reduced 42. Tian, C., Plenge, R.M., Ransom, M., Lee, A. & Villoslada, P. Analysis and application
of European genetic substructure using 300K SNP information. PLoS Genet. 4, e4
from 61% to 48%. Thus misclassification does not have a large impact on (2008).
power in this setting. 43. Price, A.L. et al. Principal components analysis corrects for stratification in genome-
wide association studies. Nat. Genet. 38, 904–909 (2006).
Expression quantitative trait loci (eQTL) analysis. The sample for eQTL 44. Luca, D. et al. On the use of general control samples for genome-wide association
studies: genetic matching highlights causal variants. Am. J. Hum. Genet. 82,
analysis was composed of 387 neurologically normal subjects of European 453–463 (2008).
ancestry with no clinical history of neurological or cerebrovascular disease 45. Wu, J., Devlin, B., Ringquist, S., Trucco, M. & Roeder, K. Screen and clean: a tool
or a diagnosis of cognitive impairment (249 brains from the University of for identifying interactions in genome-wide association studies. Genet Epidemiol 34,
Maryland Brain Bank; 36 from Johns Hopkins University Brain Bank; 25 from 275–285 (2010).
46. Gibbs, J.R. et al. Abundant quantitative trait loci exist for DNA methylation and
the Baltimore Longitudinal Study on Aging; 24 from the University of Miami gene expression in human brain. PLoS Genet. 6, e1000952 (2010).
Brain Bank; and 78 from the Sun Health Brain Bank). All these individuals 47. Purcell, S. et al. PLINK: a tool set for whole-genome association and population-
were of non-Hispanic European ancestry. based linkage analysis. Am. J. Hum. Genet. 81, 559–575 (2007).

doi:10.1038/ng.859 Nature Genetics