The PSEN1, p.E318G Variant Increases the Risk of Alzheimer's Disease in APOE-ε4 Carriers
p.E318G Variant Increases the Risk of Alzheimer's Disease in APOE-e4 Carriers. PLoS
Genet 9(8): e1003685. doi:10.1371/journal.pgen.1003685
The PSEN1 , p.E318G Variant Increases the Risk of Alzheimer's Disease in APOE -e4 Carriers
Bruno A. Benitez 0 1 2 3 4 5 6
Celeste M. Karch 0 1 2 3 4 5 6
Yefei Cai 0 1 2 3 4 5 6
Sheng Chih Jin 0 1 2 3 4 5 6
Breanna Cooper 0 1 2 3 4 5 6
David Carrell 0 1 2 3 4 5 6
Sarah Bertelsen 0 1 2 3 4 5 6
Lori Chibnik 0 1 2 3 4 5 6
Julie A. Schneider 0 1 2 3 4 5 6
David A. Bennett 0 1 2 3 4 5 6
Alzheimer's Disease 0 1 2 3 4 5 6
Neuroimaging Initiative (ADNI)" 0 1 2 3 4 5 6
Genetic 0 1 2 3 4 5 6
Environmental Risk for Alzheimer's Disease Consortium 0 1 2 3 4 5 6
(GERAD 0 1 2 3 4 5 6
Anne M. Fagan 0 1 2 3 4 5 6
David Holtzman 0 1 2 3 4 5 6
John C. Morris 0 1 2 3 4 5 6
Alison M. Goate 0 1 2 3 4 5 6
Carlos Cruchaga 0 1 2 3 4 5 6
Amanda J. Myers, University of Miami, Miller School of Medicine, United States of America
0 School , Boston , Massachusetts, United States of America, 4 Program in Medical and Population Genetics, Broad Institute of Harvard University and M.I.T. , Cambridge
1 Genomics, Institute for the Neurosciences Department of Neurology, Brigham and Women's Hospital , Boston , Massachusetts, United States of America , 3 Harvard Medical
2 1 Department of Psychiatry, School of Medicine, Washington University , St. Louis , Missouri, United States of America , 2 Program in Translational NeuroPsychiatric
3 Washington University , St. Louis, Missouri , United States of America
4 Protein Aggregation and Neurodegeneration, Washington University St. Louis, Missouri, United States of America, 8 Department of Genetics, School of Medicine
5 United States of America, 6 Department of Neurology, School of Medicine, Washington University , St. Louis , Missouri, United States of America, 7 Hope Center Program on
6 Massachusetts, United States of America, 5 Rush Alzheimer's Disease Center and Department of Neurological Sciences, Rush University Medical Center , Chicago, Illinois
The primary constituents of plaques (Ab42/Ab40) and neurofibrillary tangles (tau and phosphorylated forms of tau [ptau]) are the current leading diagnostic and prognostic cerebrospinal fluid (CSF) biomarkers for AD. In this study, we performed deep sequencing of APP, PSEN1, PSEN2, GRN, APOE and MAPT genes in individuals with extreme CSF Ab42, tau, or ptau levels. One known pathogenic mutation (PSEN1 p.A426P), four high-risk variants for AD (APOE p.L46P, MAPT p.A152T, PSEN2 p.R62H and p.R71W) and nine novel variants were identified. Surprisingly, a coding variant in PSEN1, p.E318G (rs17125721-G) exhibited a significant association with high CSF tau (p = 9.261024) and ptau (p = 1.861023) levels. The association of the p.E318G variant with Ab deposition was observed in APOE-e4 allele carriers. Furthermore, we found that in a large casecontrol series (n = 5,161) individuals who are APOE-e4 carriers and carry the p.E318G variant are at a risk of developing AD (OR = 10.7, 95% CI = 4.7-24.6) that is similar to APOE-e4 homozygous (OR = 9.9, 95% CI = 7.2.9-13.6), and double the risk for APOE-e4 carriers that do not carry p.E318G (OR = 3.9, 95% CI = 3.4-4.4). The p.E318G variant is present in 5.3% (n = 30) of the families from a large clinical series of LOAD families (n = 565) and exhibited a higher frequency in familial LOAD (MAF = 2.5%) than in sporadic LOAD (MAF = 1.6%) (p = 0.02). Additionally, we found that in the presence of at least one APOE-e4 allele, p.E318G is associated with more Ab plaques and faster cognitive decline. We demonstrate that the effect of PSEN1, p.E318G on AD susceptibility is largely dependent on an interaction with APOE-e4 and mediated by an increased burden of Ab deposition.
Funding: This work was supported by grants from AstraZeneca, NIH (P30 NS069329-01, R01 AG035083) and the Barnes-Jewish Hospital Foundation. The authors
thank the Clinical and Genetics Cores of the Knight ADRC at Washington University for clinical and cognitive assessments of the participants and for APOE
genotypes and the Biomarker Core of the Adult Children Study at Washington University for the CSF collection and assays. Samples from the National Cell
Repository for Alzheimers disease (NCRAD), which receives government support under a cooperative agreement grant (U24 AG21886; U24: 5U24AG026395 and
1R01AG041797) were used in this study. Data collection and sharing for this project was funded by the Alzheimers Disease Neuroimaging Initiative (ADNI)
(National Institutes of Health Grant U01 AG024904). ADNI is funded by the National Institute on Aging, the National Institute of Biomedical Imaging and
Bioengineering, and through generous contributions from the following: Abbott; Alzheimers Association; Alzheimers Drug Discovery Foundation; Amorfix Life
Sciences Ltd.; AstraZeneca; Bayer HealthCare; BioClinica, Inc.; Biogen Idec Inc.; Bristol-Myers Squibb Company; Eisai Inc.; Elan Pharmaceuticals Inc.; Eli Lilly and
Company; F. Hoffmann-La Roche Ltd and its affiliated company Genentech, Inc.; GE Healthcare; Innogenetics, N.V.; IXICO Ltd.; Janssen Alzheimer Immunotherapy
Research & Development, LLC.; Johnson & Johnson Pharmaceutical Research & Development LLC.; Medpace, Inc.; Merck & Co., Inc.; Meso Scale Diagnostics, LLC.;
Novartis Pharmaceuticals Corporation; Pfizer Inc.; Servier; Synarc Inc.; and Takeda Pharmaceutical Company. The Canadian Institutes of Health Research is
providing funds to support ADNI clinical sites in Canada. Private sector contributions are facilitated by the Foundation for the National Institutes of Health (www.
fnih.org). The grantee organization is the Northern California Institute for Research and Education, and the study is Rev March 26, 2012 coordinated by the
Alzheimers disease Cooperative Study at the University of California, San Diego. ADNI data are disseminated by the Laboratory for NeuroImaging at the University
of California, Los Angeles. This research was also supported by NIH grants P30 AG010129 and K01 AG030514. GERAD1 Acknowledgements: Cardiff University was
supported by the Wellcome Trust, Medical Research Council (MRC), Alzheimers Research UK (ARUK) and the Welsh Assembly Government. ARUK supported
sample collections at the Kings College London, the South West Dementia Bank, Universities of Cambridge, Nottingham, Manchester and Belfast. The Belfast
group acknowledges support from the Alzheimers Society, Ulster Garden Villages, N.Ireland R&D Office and the Royal College of Physicians/Dunhill Medical Trust.
The MRC and Mercers Institute for Research on Ageing supported the Trinity College group. The South West Dementia Brain Bank acknowledges support from
Bristol Research into Alzheimers and Care of the Elderly. The Charles Wolfson Charitable Trust supported the OPTIMA group. Washington University was funded
by NIH grants, Barnes Jewish Foundation and the Charles and Joanne Knight Alzheimers Research Initiative. Patient recruitment for the MRC Prion Unit/UCL
Department of Neurodegenerative Disease collection was supported by the UCLH/UCL Biomedical Centre. LASER-AD was funded by Lundbeck SA. The Bonn
group was supported by the German Federal Ministry of Education and Research (BMBF), Competence Network Dementia and Competence Network
Degenerative Dementia, and by the Alfried Krupp von Bohlen und Halbach-Stiftung. The GERAD1 Consortium also used samples ascertained by the NIMH AD
Genetics Initiative. Replication analysis in the Religious Orders Study and Rush Memory and Aging Project cohorts was supported by grants from the National
Institutes of Health [R01 AG30146, P30 AG10161, R01 AG17917, R01 AG15819, K08 AG034290], the Illinois Department of Public Health, and the Burroughs
Wellcome Fund. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing Interests: The authors have declared that no competing interests exist.
" Data used in preparation of this article were obtained from the Alzheimer s Disease Neuroimaging Initiative (ADNI) database (adni.loni.ucla.edu). As such, the
investigators within the ADNI contributed to the design and implementation of ADNI and/or provided data but did not participate in analysis or writing of this
report. A complete listing of ADNI investigators can be found at: http://adni.loni.ucla.edu/wp-content/uploads/how_to_apply/ADNI_Acknowledgement_List.pdf
` Data used in the preparation of this article were obtained from the Genetic and Environmental Risk for Alzheimers disease (GERAD1) Consortium. As such, the
investigators within the GERAD1 consortia contributed to the design and implementation of GERAD1 and/or provided data but did not participate in analysis or
writing of this report. A full list of GERAD investigators can be found in Supporting Information.
Dementias are complex, polygenic and genetically
heterogeneous disorders . The most common form of dementia is
Alzheimers disease (AD), which affects more than 5.3 million
people in the US . Late-onset AD (LOAD) is the most common
form of dementia. However, the current model of AD
pathogenesis is based on the genetic findings in rare and phenotypically
extreme AD cases . LOAD heritability varies from 58% to 79%
 and, despite the tremendous progress in AD genetics in the last
twenty years, the total proportion of phenotypic variance
explained by all the combined variants (including APOE genotype
and genome wide association studies [GWAS] signals) is estimated
to be 23% , which suggests a large proportion of the heritability
of AD still remains unexplained. Three important factors may
account for the missing heritability in AD; first, the clinical
heterogeneity of AD remains a significant confounding variable in
case-control studies , second, much of the unexplained variance
of complex phenotypes may be attributed to low frequency or rare
alleles  and third, gene by gene or gene by environment
interactions . Quantitative intermediate phenotypes have
helped to overcome some of these obstacles in complex diseases
[9,10]. Endophenotype-oriented approaches have greater
statistical power, less clinical heterogeneity and offer important insights
into the mechanisms by which genetic variants modulate the
disease phenotype [6,9,10,11].
The primary constituents of plaques (Ab42/Ab40) and
neurofibrillary tangles (tau and phosphorylated forms of tau [ptau]) are
the current leading diagnostic and prognostic cerebrospinal fluid
(CSF) biomarkers for AD . Recently, it was shown that CSF
biomarker abnormalities typically precede clinical AD symptoms
by decades and reflect the timing and magnitude of
pathophysiological changes . These findings suggest that a better
understanding of the genetic contribution to the variance in these
CSF biomarkers can provide important information about
susceptibility to AD. In fact, the two most important known risk
factors for AD, APOE genotype and age account for 13% and 14%
of the variance in CSF Ab42 and tau levels, respectively .
Likewise, pathogenic mutations in the most important causal genes
for familial AD, amyloid-beta precursor protein (APP), and presenilin 1
and 2 (PSEN1, PSEN2) alter CSF Ab42 levels [13,15,16].
Additionally, some genetic variants initially discovered by their
association with CSF biomarkers have recently been proven to be
modifiers of risk, age at onset (AAO) or rate of AD progression
[17,18,19]. Likewise, it was recently described that carriers of
PSEN1 mutations exhibit very low CSF Ab42, and high tau or
ptau levels [13,20,21,22]. Similar CSF biomarker level profiles
have been described in sporadic AD cases . However, the
genetic variants responsible for CSF changes in sporadic AD have
not been found yet. Together, these results suggest that CSF
biomarker levels as quantitative traits are useful tools in
uncovering genetic variants that are closely related to the
physiopathological mechanisms underlying AD.
Rare or low frequency coding and non-coding variants have
been predicted to be enriched in functional alleles and to exhibit
strong effect size [7,10]. Recently, a rare (minor allele frequency
[MAF] = 0.02) coding variant in TREM2 gene p.P47H was found
to confer a high risk for AD (Odd ratios from 2 to 5) [24,25,26].
Two recent studies analyzed the association of genetic variants of
APP, PSEN1, PSEN2, MAPT, and GRN on risk for AD [27,28]. One
study was focused on common variants in sporadic AD  while
the other focused on the identification of very rare coding variants
in familial LOAD . However, the impact of low-frequency
coding variants of APP, PSEN1, PSEN2, GRN and MAPT on
sporadic LOAD has not been well studied. Identification of low
frequency variants associated with disease remains challenging
because standard case-controls design requires very large sample
sizes. To overcome this problem we have used quantitative
phenotypes. Previously, we identified a pathogenic mutation in a
family with LOAD within the PSEN1 gene by selecting the top and
bottom 5% from the distributions of Ab40, Ab42, and Ab42/40
ratio  In the present study, we sequenced individuals with
extremes levels of CSF-based biomarkers in order to identify
variants in APOE, APP, PSEN1, PSEN2, GRN and MAPT genes
associated with the CSF biomarker levels. This approach allowed
us to identify known pathogenic variants, AD risk factors and
identify a low frequency variant that increases risk for AD in a
gene-gene interaction mode.
Rare variants found by targeted-pooled-DNA and Next
We hypothesized that the coding variants found in individuals at
the extremes of the phenotypic distribution of CSF biomarker
levels are more likely to have a functional impact on CSF
biomarker levels. In order to identify rare or low frequency
variants that affect the CSF levels of Ab42, tau and ptau levels, we
used a two-stage extreme phenotype sequencing design (Figure
S1). A 10-fold difference between the lowest and highest raw
values in Ab42, tau and ptau CSF levels in each series was found
among individuals in these studies. The individuals were selected
regardless of their clinical status (based on the clinical dementia
rating [CDR]) (Table 1). We combined both series (WU-ADRC
[n = 475] and ADNI [n = 259]) by normalizing the CSF Ab42, tau
and ptau levels and adjusting for covariates [17,18]. We selected
212 individuals from the top and bottom 15% for each phenotype
(Table 1). The 212 samples were divided in two pools (Pool 1 and
Alzheimers disease (AD) is the most common
neurodegenerative disease affecting more than 5.3 million people
in the US. AD-causing mutations have been identified in
APP, PSEN1 and PSEN2 genes. Heterozygous carriers of
APOE-e4 allele exhibit a 3-fold increased risk for developing
AD, while homozygous carriers show a 10-fold greater risk
than non-carriers. Here, we sequenced individuals with
extreme levels of well-established AD cerebrospinal fluid
(CSF) biomarkers in order to identify variants in APOE, APP,
PSEN1, PSEN2, GRN and MAPT genes associated with AD
risk. This approach allowed us to identify known
pathogenic variants, additional AD risk genetic factors and
identify a low frequency variant in PSEN1, p.E318G
(rs17125721-G) that increases risk for AD in a gene-gene
interaction with APOE. These findings were replicated in
three large (.4,000 individuals) and independent datasets.
This finding is particularly important because we
demonstrated that a currently considered non-pathogenic variant
is associated with higher levels of neuronal degeneration,
and with Ab deposition, more Ab plaques and faster
cognitive decline in an APOE-e4-dependent fashion.
APOEe4 heterozygous individuals who carry this variant are at
similar AD risk as APOE-e4 homozygous individuals.
2, respectively); targeted and pooled-sample sequencing was
performed. All the validated variants were genotyped in the total
CSF sample and tested for association with each CSF biomarker.
Linear regression (assuming an additive genetic effect) was utilized
for each variant by adjusting for significant covariates (age, gender,
CDR and site [WU-ADRC or ADNI]) (Table S1 in Text S1)
A greater than 30-fold coverage per allele at all positions within
the 62 amplicons designed to cover the protein coding regions of
the APP, APOE, PSEN1, PSEN2, MAPT and GRN were obtained
(Table S2 in Text S1). After adjusting for the sensitivity and
specificity parameters of the base-calling algorithm (SPLINTER)
using negative and positive controls, a total of 396 and 369
variants were called and perfectly annotated in the targeted
genomic regions of Pool 1 and 2, respectively. 73% of these
variants were intronic, 8% were missense, 5% were
codingsynonymous, 1% were at splicing sites, 12% were located at the
untranslated regions (UTR) and 2% were called to be near-gene
(Table S2b in Text S1) We focused on missense and
splicingaffecting variants with a predicted minor allele frequency (MAF)
below 5% (by SPLINTER) in each pool.
A total of 27 rare or low frequency non-synonymous variants
were validated by direct genotyping in the discovery samples (both
pools). 33% of these variants identified (9/27) were novel. Seven of
nine (77%) are located in highly conserved nucleotide (GERP.4)
and 88% (8/9) are predicted to be damaging for the respective
protein (SIFT and polyphen2 algorithms) . As expected, 48%
of these variants are singletons (9/27) or doubletons (4/27)
Among the 18 previously reported variants; we found one
known pathogenic mutation PSEN1 p.A426P. PSEN1 p.A426P
(rs63751223) was reported in a five members of a family with
autosomal dominant AD .We also found four high-risk
variants for LOAD (APOE, p.L46P; MAPT, p.A152T; PSEN2,
p.R62H and p.R71W) [28,32,33], six variants that were previously
reported in families with AD or frontotemporal dementia (FTD),
but classified as non-pathogenic (GRN, p.R433W, p.P458L,
p.R19W; MAPT, p.Q230R; PSEN1, p.R35Q and p.E318G)
, and seven variants that have been recently reported in
public databases with no clear role in human disease to date
(APOE, p.E37K; GRN, p.C231W; MAPT, p.G107S, p.S318L,
p.V224G; PSEN2, p.E317G and p.V300G) (A detailed description
of each variant can be found in the supporting material in Text
These results highlight the relative enrichment of rare and low
frequency variants in six genes involved in AD and FTD among
individuals at the extremes of the CSF biomarker distribution .
Association with CSF biomarker levels
Next, we tested whether any of the variants identified by an
endophenotype-based approach could improve our understanding
of both the genetic architecture and pathophysiology of LOAD
[17,18]. We ran a linear regression analysis for single SNP using
CSF biomarkers as quantitative traits, but we failed to find
significant association with CSF tau, ptau or Ab42 levels for most
of the identified variants, even after we collapsed all of the
potentially damaging variants in each gene and analyzed the dataset
for carriers vs. non-carriers of these variants (Table 3). Surprisingly,
a low frequency coding variant in PSEN1, p.E318G (rs17125721)
(MAF = 0.02 for Europeans Americans, Exome Variant Server
EVS: http://evs.gs.washington.edu/EVS/), whose pathogenic role
is currently debated  exhibited a statistically significant
association (multiple test correction threshold, p = 7.061023) with
CSF tau (p = 9.261024, Beta = 0.14) and ptau levels (P = 1.861023,
Beta = 0.12), but not with Ab42 (p = 0.14, Beta = 20.05).
Interestingly, it has been reported that the combination of Ab42 and tau or
ptau as a ratio provides the best discriminative value to date for AD
cases [35,36] and predict the conversion from non-dementia clinical
status to dementia . p.E318G exhibited a significant association
with the ratio of ptau:Ab42 (p = 9.561025, Beta = 0.08) and
tau:Ab42 (p = 2.061024, Beta = 0.06) (Figure 1AC, 2A) suggesting
that the association of p.E318G with CSF biomarker levels may be
an association with clinical AD.
In order to confirm this association with CSF biomarkers and to
determine whether this or any other SNP in linkage disequilibrium
(LD) was driving the association, we combined genotype and
imputed data from 895 individuals (WU-ADRC, n = 501, and
ADNI, n = 394, this dataset constitute the same CSF series that we
genotyped (Table 1) plus additional 161 individuals) to perform a
dense fine mapping analysis of PSEN1 genomic region. The number
of independent tests (Meff = 317) was calculated based on the
number of SNPs after correcting for LD structure (r2 = 0.8) within
the genomic region (250 Kb in each side) . We performed linear
regression assuming an additive genetic model to test the association
between each SNP and CSF biomarker levels by adjusting for age,
gender and the first three principal components from the population
stratification analysis. We confirmed a significant association
(multiple-testing threshold = 1.661024) between an intronic SNP,
rs76342307 (MAF = 0.016) and CSF ptau (p = 8.061025,
Beta = 0.14), tau (p = 8.461023, Beta = 0.10), and Ab42 levels
(p = 0.02, Beta = 20.06) (Figure 1DF) for the PSEN1 genomic
region. Rs76342307 is located 0.2 Mb 39 upstream from the PSEN1
gene. We used data from the HapMap and the 1000 Genomes
Project to identify all of the SNPs in linkage disequilibrium (LD,
r2.0.8) with rs76342307. Six SNPs (rs76342307, rs17856583,
rs1110058, rs117946815, rs117236337 and rs2091912) were found
to be in strong LD (r2 = 0.95, D9 = 1) with rs76342307 spanning
0.3 Mb (Figure 1G, H). 100% and 97% concordance rates were
observed among the directly typed and imputed results for
rs76342307 and rs117236337, respectively. Interestingly,
rs117236337 is an intronic SNP in PSEN1 gene, which is also
associated with extreme CSF tau (p = 0.02, Beta = 0.08), ptau
Total CSF Samples:
APOE e4+ (%)
Ab42 Low (%)
APOE e4+ (%)
Ab42 Low (%)
APOE e4+ (%)
Ab42 Low (%)
APOE e4+ (%)
Ab42 Low (%)
APOE e4+ (%)
Age at lumbar puncture (LP), percentage of females, percentage of APOE4 allele
carriers, clinical dementia rating (CDR) at LP date for each sample and
percentage of individuals with Low (L) levels of Ab42 normalized for each site
(see methods). For each phenotype the median in pg/ml and the range is
shown. Charles F. and Joanne Knight Alzheimers Disease Research Center at
University of Washington (WU-ADRC) and Alzheimers Disease Neuroimaging
Initiative (ADNI). Cerebrospinal Fluid (CSF).
(p = 5.761024, Beta = 0.09) and Ab42 levels (p = 0.01,
Beta = 20.06). Next, we tested whether PSEN1, p.E318G was in LD
with the SNPs identified by the fine mapping analysis. In fact,
rs17125721 (PSEN1, p.E318G) is in moderate LD with all of them
(R2 = 0.68, D9 = 1) (Figure 1H). To analyze whether the p.E318G
and rs76342307 are two independent signals, we ran a conditional
analysis including both SNPs (rs76342307 and rs17125721) in the
model. When one of the SNPs was included in the model, the
association from the other SNP disappeared, suggesting that the
association in this locus is driven by a single signal (Figure 1I).
Effect of PSEN1, p.E318G on Ab deposition is APOE
We observed that in the subset of individuals with Ab deposition
(CSF Ab42 levels lower than 500 pg/ml in WU-ADRC, and
192 pg/ml in ADNI) [35,39], the frequency of p.E318G carriers
(4.2%, 21/500) was higher than in individuals without Ab
deposition (2.5%, 11/427), although this difference did not
achieve statistical significance (p = 0.18, OR = 1.6,
95%CI = 0.783.4) (Table 3, 4). In addition, we observed that
93% (15/16) of the individuals carrying PSEN1, p.E318G along
with APOE e4 exhibited low CSF Ab42 levels, while only 45% (9/
20) of the individuals carrying PSEN1, p.E318G but do not carry
the APOE e4 allele showed low CSF Ab42 levels, suggesting that
APOE e4 allele is modifying the profile of Ab deposition in PSEN1,
p.E318G carriers (Table 4 and Figure 2A). APOE e4 is strongly
associated with CSF Ab42 levels (Table 4) [14,18], and APOE
genotype has been reported to modify disease expression in
individuals with mutations in PSEN1  and PSEN2  genes.
However, previous reports have not found any significant
interaction between APOE and PSEN1 p.E318G, most likely due
to the low frequency of PSEN1, p.E318G and small sample sizes
[42,43,44]. To analyze whether there was an APOE-dependent
effect on this variant, we tested the association of p.E318G with
CSF Ab42 levels by stratifying it in the presence (+) or absence (2)
of the APOE e4 allele. We found that the risk of having Ab
deposition is greater for carriers of PSEN1, p.E318G and APOE e4
together (OR = 18.3 CI = 2.0166.8, p = 3.561023) than those
carrying APOE e4 allele alone (OR = 4.5, CI = 3.46.0,
p,1.061025) (Table 4). These individuals are more likely to have
a CSF biomarker profile similar consistent with AD (low CSF
Ab42, and high tau or ptau levels) (Figure 2A). p.E318G carriers
who also carry APOE e4+ allele (n = 20) exhibited significantly
higher CSF tau (p = 0.04) and ptau (p = 0.01) levels and
significantly lower CSF levels of Ab42 (p = 0.02) compared to
those that are p.E318G carriers but do not carry the APOE e4
allele (Figure 2 A, B). We also found a significant interaction
AA Substitution dbSNP ID
Gene: official Symbol provide by HGNC; dbSNP: variants with or without rs numbers. AA Substitution: amino acid change resulting from the observed variant; dbSNP ID:
rs# for variants present in dbSNP 135, Novel for variants not present in dbsnp, 1000 genome or Exome Variant Server; GERP score: Genomic Evolutionary Rate Profiling
score; Protein prediction: based on SIFT/Polyphen2 analysis of the predicted effect of the substitution on protein function; MAF in ESV: Minor allele frequency in Exome
Variant Server; Total # Hets: Number of carriers of the variant in the total sample; Total MAF: Minor allele frequency in all sample genotyped. Clinical Interpretation:
Clinical interpretation is based on AD&FTD mutation database and published papers.
1dbSNP 135 appears as validation pending
(p = 0.03) between APOE e4+ and p.E318G in individuals with
increased burden of Ab deposition. Taken together, the results of
the biomarker analyses suggest that PSEN1, p.E318G is associated
with higher levels of neuronal loss (reflected by CSF tau and ptau
levels) and with Ab deposition (low Ab42 CSF levels) in an APOE
Replication of the PSEN1, p.E318G-APOE interaction in
large case-control datasets
Because the purpose of this endophenotype-based approach is
to identify variants implicated in disease, we tested whether the
PSEN1, p.E318G is associated with AD risk, tau/Ab pathology or
rate of cognitive decline in an APOE dependent manner.
Analyses of the association between PSEN1 p.E318G and
clinical AD status in an independent AD case-control series
(n = 1,855, WU series) revealed that the risk of AD is
significantly higher for p.E318G/APOE e4 carriers (OR = 9.9
CI = 2.637.5, p = 1.761024) compared to individuals carrying
APOE e4 alone (OR = 5.1, CI = 4.16.3, p = 3.2610259)
(Table 5). This finding was replicated in an independent
sample from the GERAD consortium (n = 4,058). In this
dataset, the association of p.E318G with AD case-control
status in the presence of at least one APOE e4 allele
(OR = 10.3, 95% CI = 4.125.5, p = 4.161028) was double the
risk for AD in the presence of APOE e4 alone (OR = 4.1, 95%
CI = 3.54.8, p = 1.1610279). In the joint-analysis of these two
independent series (5,161 individuals), the risk of developing
AD in the p.E318G/APOE e4 carriers (OR = 10.1, 95%
CI = 4.820.9, p = 9.0610212) is two-fold the AD risk of those
that carry APOE e4 allele alone (OR = 4.4, 95% CI = 3.95.0,
p = 6.86102139) (Table 5).
In fact, we found that individuals who are APOE e4
heterozygous and also carry the p.E318G variant are at similar
AD risk (OR = 10.7, 95% CI = 4.724.6, p = 2.5610210) as APOE
e4 homozygous (OR = 9.9, 95% CI = 7.2.913.6, p = 5.5610276)
and are at double the AD risk compared to APOE e4 heterozygous
that are not carrying p.E318G (OR = 3.9, 95% CI = 3.44.4,
p = 2.86102106) (Table 6, Figure 2C).
In an independent analysis leveraging two prospective cohorts,
the Religious Orders Study and Rush Memory and Aging Project,
we confirmed a significant interaction between APOE4 and
p.E318G with burden of neuritic plaques at autopsy (n = 748;
P = 0.01) but we failed to detect any significant association with
neurofibrillary tangles (p = 0.47). Interestingly, the effect of APOE
e4 allele alone on neuritic plaques (n = 748, p = 4.5610224,
Beta = 0.39) was increased by two fold the presence of p.E318G
(n = 204, p = 0.08, Beta = 0.74). p.E318G has previously associated
with lower cognitive performance . We tested whether the
interaction between APOE4 and p.E318G affect the episodic
memory. We found that there is trend between interaction
between APOE4 and p.E318G with episodic memory decline
(p = 0.08).Furthermore, the significant effect of APOE e4 allele on
episodic memory decline (p = 1.7610216, Beta = 20.06) was
modified by the presence of p.E318G (p = 0.14, Beta =
20.16).However, these interactions showed the predicted direction of
effects for these phenotypes based on the results of the biomarker
data: In the presence of at least one APOE-e4 allele, p.E318G is
associated with more Ab plaques, faster cognitive decline and
higher risk for AD.
Family based and segregation analysis
The p.E318G variant has been associated with familial AD in
different populations [42,44,46]. However, this association has not
been consistently replicated [43,47,48,49]. Our previous analyses
indicate that in sporadic AD cases the effect of the p.E318G
variant can be detected only in presence of the APOE e4 allele. We
wanted to analyze whether the same effect is found in familial
cases. We genotyped probands from 565 total LOAD families and
found the presence of PSEN1 p.E318G in 30 families
(MAF = 2.5%). PSEN1 p.E318G exhibited a higher frequency in
individuals with familial LOAD than those with sporadic LOAD
(MAF = 1.6%, n = 3,989, p = 0.02) and a group of age matched
control subjects (MAF = 1.5%, n = 830, p = 0.03). Next, we tested
whether the association with familial LOAD was due to the
interaction of p.E318G with APOE-e4 allele. The presence of
APOE-e4 allele in p.E318G carriers in familial AD (70%, 21/30)
was higher than that in sporadic AD (65%, 84/129) but not
statistically significant (p = 0.61). On the other hand, APOE-e4/
p.E318G carriers in familial AD were significantly higher
(p = 4.061024) than those in the control group (15%, 10/69).
Therefore, the risk conferred by APOE-e4 and p.E318G carriers in
familial AD (OR = 16.4, 95% CI = 5.648.2, p = 5.861028)
compared to the control group was higher than the risk associated
with sporadic AD (OR = 10.1, 95% CI = 4.820.9,
p = 9.0610212). These results suggest that higher risk of the
p.E318G variant in familial cases is mostly due to the high
frequency of APOE e4 allele in this population .
Interestingly, the p.E318G variant has been reported in
multigenerational families with AD [42,50]. However, PSEN1
p.E318G is not considered pathogenic in part due to the absence
of conclusive evidence for cosegregation with AD [34,43,47,48].
We observed 8 families (with more than two affected individuals
carrying p.E318G) in which p.E318G segregates with disease
(Figure 2D), even in the absence of APOE-e4 allele (two families)
(Figure S2). These families do not carry any other mutations in
APP, PSEN1, PSEN2, GRN and MAPT genes . In three
additional families the cosegregation p.E318G with AD was
inconclusive because only a few family members had been
sampled and/or because p.E318G carriers were below the mean
age of onset for AD in their respective families. Thus, using the
largest sample of familial LOAD screened to date for the role of
p.E318G in AD, we have demonstrated that minor allele p.E318G
increases the risk of familial LOAD. Furthermore, p.E318G
cosegregates with AD in 26% of all the familial LOAD carriers.
Effect on age at onset of AD
Carriers of PSEN1, p.E318G have been reported across a wide
range of ages (45 to 93 yrs.) [42,44,46,50]. Thus, we tested
whether PSEN1, p.E318G affects AAO regardless of the APOE
genotype; we found that PSEN1, p.E318G carriers have a lower
AAO than non-carriers (73.9 yr. vs. 78.2 yr.; p = 0.01) (Figure S3).
Resequencing genes in individuals from the extremes of the
biomarker distribution constitutes a powerful and efficient strategy
to identify functional sequence variants associated with complex
traits . CSF-based biomarker profiles have proven to be
powerful tools in endophenotype-oriented approaches, by which
we have been able to identify common genetic variants associated
with the rate of progression, AAO or the risk of AD
[11,14,17,18,51]. Previously, we identified a pathogenic mutation
in a family with LOAD within the PSEN1 gene by selecting the top
and bottom 5% from the distributions of CSF levels of Ab40,
Ab42, and Ab42/40 ratio . Here, we have used a novel and
powerful approach by using next-generation sequencing to
sequence individuals with extreme phenotypes: individuals from
the bottom and top 15% of Ab42, tau, or ptau CSF levels.
Pathogenic mutations and high-risk AD variants
Previous data have suggested that mutations in APP, PSEN1,
and PSEN2 genes only cause early-onset familial AD. However,
this study and previous studies from our group [28,52] indicate
that pathogenic mutations in these genes can be also found in
lateonset familial and sporadic AD cases. In this study, we observed a
known and confirmed pathogenic mutation (PSEN1 p.A426P,
rs63751223) in one individual (57 years old) without a clear family
history of dementia, out of 258 individuals (CDR.0), which
constitutes 0.3% of AD cases.
In a previous study, Cruchaga et al, found that 2.3% of families
with multiple members affected by LOAD carried pathogenic
mutations . In this study, we expanded our analyses to
sporadic cases, which constitute 95% of the total number of AD
cases. Although we found only one case with a pathogenic
mutation (0.3%), this could be an underestimate because both of
the novel mutations, PSEN2: p.G270S and MAPT p.T263P were
found in single cases that met biomarker criteria for AD. A novel
variant in GRN, p.C247Y and a known variant in PSEN1, p.R35Q
were found in demented individuals with a non-AD CSF profile
suggesting another type of dementia. However, without
segregation analyses, additional functional studies are required to
determine the potential pathogenicity of these variants.
The classification of mutations as not pathogenic, possibly
pathogenic, probably pathogenic and definitely pathogenic based
on segregation analyses, amino acid conservation, effects on Ab
metabolism in in vitro studies, association studies and presence in
healthy individuals has been useful in prioritizing mutations and
their likelihood of affecting risk for disease . However, this
classification is likely to miss variants with a smaller but real effect
(OR.2.0) on risk for sporadic AD. The variant GRN, p.P458L is
classified as non-pathogenic  due to fact that it was reported in
an ALS/FTD patient and in 25 out of 492 controls (MAF = 2.5%)
. However, this variant is not reported in the EVS server
(6,515 exomes) (EVS-v.0.0.18, (February 8, 2013) or in our control
population of 824 samples (Table 2). Here, this variant was found
Lack of Ab deposition
in an individual with early onset dementia and with typical
biomarker criteria for AD. PSEN2, p.R71W has been classified as
non-pathogenic because it was reported in controls and EOAD
cases . However, in a previous study the frequency of the
p.R71W variant in AD cases was significantly higher than in
controls (n = 3,152, p = 9.061024 OR = 6.45; 95%CI = 1.95
21.39) and carriers have a significantly earlier age at onset than
affected non-carriers (p.R71W: 70.2 vs. 76.7, p = 5.061024),
suggesting that this variant could be a modifier of LOAD risk .
Here, we found the same trend, PSEN2 p.R71W was also found to
be present more frequently in clinical cases than in controls
(p = 0.03, OR = 10.3, 95%CI = 1.196.2). However, it did not
reach statistical significance in individuals with Ab deposition
(p = 0.27, OR = 3.4, 95%CI = 0.3830.7).
PSEN1 p.E318G increases the risk of AD in APOE e4 allele
The PSEN1, p.E318G variant has been considered to be a
nonpathogenic variant, because it has been found in non-demented
individuals [43,48,49] and the absence of conclusive evidence for
cosegregation with AD . However, it has been suggested that
phenocopies, potential presymptomatic individuals, reduced
penetrance and gene by gene interactions complicate the
interpretation of the p.E318G variant in familial and sporadic LOAD
[42,44]. This is the first study to systematically screen the presence
of PSEN1 p.E318G in a large (n = 565) clinical series of
wellcharacterized families densely affected by LOAD with no
mutations in APP, PSEN2, GRN or MAPT genes. PSEN1
p.E318G was found in 5.3% and cosegregated with the disease
in 1.4% of all families. We also found that PSEN1 p.E318G
exhibited a higher frequency in familial LOAD than in sporadic
LOAD (p = 0.025), supporting earlier findings that the p.E318G
variant has higher frequencies among AD cases with a family
history of AD in different populations [42,44,46]. Additionally, our
analyses indicate that PSEN1 p.E318G carriers have an average
age at onset that is 4.3 years earlier than that in non-carriers
(73.9 yr. vs. 78.2 yr). Putative pathogenic variants in genes that
cause late-onset rather than early-onset dementia could have a less
severe effect on protein function due to genetic or environmental
modifiers . Our CSF biomarker analyses suggested that PSEN1
p.E318G was associated with higher levels of neuronal loss
(reflected by high CSF tau and ptau levels) and with Ab deposition
(low Ab42 CSF levels) in an APOE e4-dependent fashion.
Furthermore, in the largest AD case-control series (n = 5,161)
analyzed for the interaction between PSEN1 p.E318G and APOE
e4 allele to date, we found that the presence of p.E318G and
APOE e4 doubles the risk for AD (OR = 10.3, 95% CI = 4.125.5)
compared to the risk with the presence of APOE e4 alone
(OR = 4.1, 95% CI = 3.54.8). There are several reports of
variants that modify the risk of AD in APOE e4 carriers such as
a-1-antichymotrypsin (ACT) gene (APOE e4/ACT, [OR = 6.4,
non 95% CI reported]) , Cholesteryl ester transfer protein
(CETP) gene (APOE e4/CETP  C allele [OR 7.12, non
95% CI reported]) , GRB-associated binding protein 2
(GAB2) gene (APOE e4/rs2373115 genotype GG [OR = 2.36,
95% CI 1.553.58]) , CUG triplet repeat, and RNA binding
protein 2 (CUGBP2) gene (APOE e4/e4/rs62209 [OR = 1.75,
95% CI 1.272.41]) . However, all these variants have a
modest effect increasing the risk due to APOE e4 allele. Here, we
provided evidence of a low frequency variant in PSEN1 gene with
a significant effect on the AD risk in APOE e4 carriers (OR = 10.7,
95% CI = 4.724.6) comparable only to the effect of a second
APOE e4 allele (OR = 9.9, 95% CI = 7.2.913.6). Moreover, we
also found that in the presence of at least one APOE e4 allele,
p.E318G is associated with more Ab plaques and faster cognitive
decline, as recently reported for a low frequency variant in
complement receptor 1 (CR1)  In addition, p.E318G has
previously associated with lower cognitive performance, which
support our findings of cognitive decline . The interaction of
the p.E318G with APOE e4 allele was replicated in four different
datasets: the CSF dataset (discovery set), WU_ADRC case-control
dataset, GERAD1 and the Religious Orders Study and Rush
Memory and Aging Project, indicating that this association and
interaction is not a type I error, but a real association. All these
results together support the role of PSEN1 p.E318G as one of the
most important modifiers of the risk of LOAD reported to date.
Functional studies, especially concerning the effect on Ab
metabolism in vitro, have further questioned the pathogenicity of
the p.E318G variant. One study showed no alteration in the
production of Ab42 induced by p.E318G . However, a recent
study using skin fibroblasts from individuals with the p.E318G
variation showed an increase in the production of Ab40, a
decrease in Ab42 and a subsequent significant reduction in the
Ab42/Ab40 ratio compare to non-carriers , along with a lack
of an inhibitory effect of the exon 9 loop in the presence of the
p.E318G variant reported by an independent study . It has
been proposed that the activation of c-secretase results from a
cleavage-induced conformational change that relieves the
inhibitory effect of the intact exon 9 loop, which is mediated by
occupying the substrate-binding site on the immature enzyme
before it is cleaved . It was reported that p.E318G abolishes
the inhibitory effect of the intact exon 9 loop, which favors the
production of Ab40 . It was also reported that p.E318G affects
the processing of PSEN1 by reducing the amount of N-terminal
fragment that is generated after cleavage , and augments levels
of neuronal cell death after overexpression . We suggest that
another approach to test the impact of pathogenic mutations on
Ab metabolism is to examine the effect on the CSF biomarker
levels. Most of the published data about CSF biomarkers reveal
that PSEN1 gene mutation carriers display a typical AD biomarker
signature with low CSF levels of Ab42 and high CSF tau levels
[13,20]. There is no published data on the levels of CSF
biomarkers for PSEN1, p.E318G carriers. Here, for the first time
we demonstrate that PSEN1, p.E318G/APOE e4 carriers have a
CSF biomarker profile similar to AD cases.
In summary, these results highlight the relative enrichment of
low frequency variants in six genes involved in AD and FTD that
are at the extremes of the distribution of CSF biomarker levels
. We provide evidence that the PSEN1, p.E318G variant
increases the risk for AD in APOE e4 heterozygous, equivalent to
that of APOE e4 homozygous. We also found that p.E318G
increases the risk of familial LOAD and cosegregates with AD in
26% of all the familial LOAD carriers. All these findings have
important implications for genetic counseling since PSEN1,
p.E318G is currently considered a non-pathogenic variant .
By using CSF biomarker levels as a quantitative trait, we were
able to identify a low frequency variant associated with AD risk,
PSEN1, p.E318G. This association is mediated by a SNP-by-SNP
interaction, which has not been found using the standard
casecontrol design [43,48,49]. Together, these results indicate that
there are potentially many more low frequency variants associated
with complex disease, and that the association results from
complex interactions. We were able to identify the association of
PSEN1, p.E318G with risk for AD and its interaction with the
APOE e4 allele because both genes are known to be associated with
AD. However, the identification of such an association and
interactions in a genome-wide approach remains still challenging
and requires novel, powerful approaches.
We believe that this endophenotype-based approach is a good
alternative to case-control studies and can allow us to gain a better
understanding of both the genetic architecture and
pathophysiology of LOAD [17,18]. In terms of genetics and factors that may
explain some of the missing hereditability of complex diseases,
these results are important because they are a clear example of low
frequency variants that are associated with disease and how such
associations are due to epistatic gene by gene interactions.
Materials and Methods
The Institutional Review Board (IRB) at the Washington
University School of Medicine in Saint Louis approved the study.
Prior to their participation, a written informed consent was
reviewed and obtained from family members. The Human
Research Protection Office (HRPO) approval number for our
ADRC Genetics Core family studies is 93-0006.
Two CSF series were used for this study. A total sample of 475
individuals enrolled in longitudinal studies at the Alzheimers
disease Research Center at Washington University School of
Medicine (ADRC) and 259 participants of the Alzheimers disease
Neuroimaging Initiative (ADNI) were used in this study. A subset
of 145 participants from ADRC and 67 from ADNI were included
in the discovery series (two DNA pools). CSF samples were from
individuals of European descent. In the WU-ADRC-CSF series:
60% of sample is female, ranging from 3791 years of age. 73% of
the sample has a clinical dementia rating (CDR) of 0 (cognitively
normal) and 39% of the individuals carry at least one APOE e4
allele. In the ADNI-CSF series: 44% of sample is female, ranging
from 5691 years of age. 60% of the sample has a CDR higher
than 0 (demented) and 47% are APOE e4 allele positive. Table 1
summarizes the demographic data for the CSF series.
Covariateadjusted residuals of CSF Ab42, tau and p-tau were used to define
the pools (see statistical analysis, Table S3 in Text S1). 114
individuals in the bottom 15% of CSF Ab42 levels or individuals
in the top 15% of CSF tau or p-tau levels were included in a pool.
The second pool consisted of 98 individuals in the top 15% of CSF
Ab42 or individuals in the bottom 15% of tau and p-tau181 levels
The Religious Orders Study (ROS) and the Rush Memory and
Aging Project (MAP) recruit participants without known dementia
who agree to annual clinical evaluations and sign an Anatomic
Gift Act donating their brains at death. The full cohort with
genotype data included 1,708 subjects (817 ROS and 891 MAP).
The mean age at enrollment was 78.5 years and 69.1% were
female. At the last evaluation, 24.9% met clinical diagnostic
criteria for AD and 21.8% had mild cognitive impairment. The
summary measure of global cognitive performance was based on
annual assessments of 17 neuropsychiatric tests. A nested autopsy
cohort consisted of 651 deceased subjects (376 ROS and 275
MAP); mean age at death was 81.5 years and 37.6% were male.
Proximate to death, 40.9% of subjects included in the autopsy
cohort met clinical diagnostic criteria for AD. Bielschowsky silver
stain was used to visualize neurofibrillary tangles in tissue sections
from the midfrontal, middle temporal, inferior parietal, and
entorhinal cortices, and the hippocampal CA1 sector. A
quantitative composite score for neurofibrillary tangle pathologic burden
was created by dividing the raw counts in each region by the
standard deviation of the region specific counts, and then
averaging the scaled counts over the 5 brain regions to create a
single standardized summary measure. Additional details of the
ROS and MAP cohorts as well as the cognitive and pathologic
phenotypes are described in prior publications [58,62]. Follow-up
series included 1,031 sporadic AD cases, 824 unrelated elderly
cognitively normal controls and a single case from NIA-LOAD
families (n = 595) . All these samples are independent of the
CSF samples. Cases received a diagnosis of dementia of the
Alzheimers type (DAT), using criteria equivalent to the National
Institute of Neurological and Communication Disorders and
Stroke-Alzheimers Disease and Related Disorders Association for
probable AD [63,64]. Controls received the same assessment as
the cases but were non-demented. All individuals were of
European descent and written consent was obtained from all
DNA from ROS and MAP subjects was extracted from whole
blood, lymphocytes or frozen post-mortem brain tissue and
genotyped on the Affymetrix Genechip 6.0 platform, as previously
described . Following standard QC procedures, imputation
was performed using MACH software (version 1.0.16a) and
HapMap release 22 CEU (build 36) as a reference.
Statistical and association analyses
Association of Ab42, tau and p-tau181 with genetic variants was
analyzed as previously reported [14,17,18]. Briefly, Ab42, tau and
p-tau181 values were log transformed to approximate a normal
distribution. Because the CSF biomarker levels were measured
using different platforms (Innotest plate ELISA vs. AlzBia3
beadbased ELISA, respectively) we were not able to combine the raw
data. For the combined analyses we standardized the mean of the
log transformed values from each dataset to zero. A stepwise
discriminant analysis identified CDR, APOE genotype, gender and
age as significant covariates in both series (Table S1b in Text S1)
[17,18]. No significant differences in the transformed and
standardized CSF values for different series were found (Table
S1b in Text S1).
CSF biomarker levels were used as a quantitative trait for most
analyses. It has been shown that CSF Ab42 is an accurate
predictor of brain amyloid burden regardless of clinical diagnosis
. Therefore, the Ab plaque deposition was assumed using the
biomarker levels as a dichotomous variable (low and high CSF
Ab42). Levels of CSF biomarkers were as follows: for the
ADNICSF series the cut-off was Ab42,192 pg/mL . In the
WUADRC-CSF series, we used CSF Ab42,500 pg/mL as the cut-off
We used Plink (http://pngu.mgh.harvard.edu/,purcell/plink/
) to analyze the association of variants (individually or collapsed by
gene) with CSF biomarker levels. Age, gender and site were
included as covariates. In order to determine whether the
association of variants with CSF biomarker levels was driven by
case-control status we included clinical dementia rating (CDR) or
CSF Ab42 levels as a covariate in the model or stratified the data
by case control status. We also performed analyses including
APOE genotype as a covariate. Association with AAO was carried
out using the Kaplan-Meier method and tested for significant
differences, using a log-rank test .
Fishers exact test was used to compare the frequency of each
variant and collapse by gene in the case control series defined by
CDR or CSF Ab42 levels (Table 3). All variants were included in
the model independent of their pathogenicity.
Analyses of SNP effects on global cognitive decline in ROS and
MAP were performed as in prior publications . Briefly, we first
fit linear mixed effects models using the global cognitive summary
measure in order to characterize individual paths of change,
adjusted for age, sex, years of education, and their interactions
with time. At least two longitudinal measures of cognition were
required for inclusion in these analyses, for which data on 1,593
subjects was available. We then used these person-specific, residual
cognitive decline slopes as the outcome variable in our linear
regression models, with each SNP of interest coded additively
relative to the minor allele, and further adjusted for study
membership (ROS vs. MAP) and the first 3 principal components
from population structure analysis. For analyses of neurofibrillary
tangle burden, linear regression was used to relate SNPs to the
pathologic summary measure, adjusting for age at death, study
membership, and 3 principal components. Because the data were
skewed, square root of the scaled neurofibrillary tangle burden
summary score was used in analyses.
Pooled sequencing analysis
Pooled-DNA sequencing was performed, as previously
described by Druley TE et al. [28,52,65,66]. Briefly, equimolar
amounts of individual DNA samples were pooled together after
being measured using Quant-iT PicoGreen reagent. Two different
pools with 100 ng of DNA from 114 and 98 individuals were
made. The coding exons and flanking regions (a minimum of
50 bp each side) were individually PCR amplified using specific
primers and Pfu Ultra high-fidelity polymerase (Stratagene). An
average of 20 diploid genomes (approximately 0.14 ng DNA) per
individual were used as input into a total of 62 PCR reactions that
covered 46,319 bases from the 6 genes. PCR products were
cleaned using QIAquick PCR purification kits, quantified using
Quant-iT PicoGreen reagent and ligated in equimolar amounts
using T4 Ligase and T4 Polynucleotide Kinase. After ligation,
concatenated PCR products were randomly sheared by sonication
and prepared for sequencing on an Illumina Genome Analyzer IIx
(GAIIx) according to the manufacturers specifications.
pCMV6XL5 amplicon (1908 base pairs) was included in the reaction as a
negative control. As positive controls, ten different constructs (p53
gene) with synthetically engineered mutations at a relative
frequency of one mutated copy per 250 normal copies was
amplified and pooled with the pcr products. Six DNA samples
heterozygous for previously known mutants in GRN, PSEN1,
MAPT genes were also included.
Single reads (36 bp) were aligned to the human genome
reference assembly build 36.1 (hg18) using SPLINTER .
SPLINTER uses the positive control to estimate sensitivity and
specificity for variant calling. The wild type: mutant ratio in the
positive control is similar to the relative frequency expected for a
single mutation in one pool (1 chromosome mutated in 125
samples = 1/250). SPLINTER uses the negative control (first
900 bp) to model the errors across the 36-bp Illumina reads and
to create an error model from each sequencing run of the
machine. Based on the error model SPLINTER calculates a
pvalue for the probability that a predicted variant is a true
positive. A p-value at which all mutants in the positive controls
were identified was defined as the cut-off value for the best
sensitivity and specificity. All mutants included as part of the
amplified positive control vector were found upon achieving
.30-fold coverage at mutated sites (sensitivity = 100%) and only
,80 sites in the 1908 bp negative control vector were predicted
to be polymorphic (specificity = ,95%). The variants with a
pvalue below this cut-off value were considered for follow-up
confirmation. All rare missense or splice site variants (with an
estimated allelic frequency less than 5%) were then validated by
Sequenom and KASPar genotyping in each individual included
in the pools [28,52,66]. The validated SNPs were then
genotyped in all members of the WU-ADRC-CSF and
ADNICSF series. Common variants (.5%) and synonymous variants
were not followed up.
An average coverage of 30X-fold per allele per pool is the
minimum coverage necessary to get an optimal positive predictive
value for the SNP-calling algorithm . The necessary number
of lanes to obtain a minimum of 30-fold coverage per base and
sample were run (Table S2 in Text S1).
The WU-ADRC samples were genotyped with the Illumina 610
or OmniExpress. The ADNI samples were genotyped with the
Illumina 610 chip. Prior to association analysis, all samples and
genotypes underwent stringent quality control (QC). Genotype
data were cleaned applying a minimum call rate for SNPs and
individuals (98%) and minimum minor allele frequencies (0.02).
SNPs not in Hardy-Weinberg equilibrium (P,161026) were
excluded. The QC cleaning steps were applied for each
genotyping array separately. We tested for unanticipated
duplicates and cryptic relatedness using pairwise genome-wide
estimates of proportion identity-by-descent. When a pair of
identical samples or a pair of samples with cryptic relatedness
was identified, the sample from the WU-ADRC or samples with a
higher number of SNPs passing QC were prioritized. Eigenstrat
was used to calculate principal component factors for each sample
and confirm the ethnicity of the samples . The 1000 Genome
Project data (June 2011 release) and Beagle software were used to
impute up to 6 million SNPs. SNPs with a Beagle R2 of 0.3 or
lower, a minor allele frequency (MAF) lower than 0.05, out of
Hardy-Weinberg equilibrium (p,1610-6), a call rate lower than
95% or a Gprobs score lower than 0.90 were removed. A total of
5,815,690 SNPs passed the QC process.
We used PLINK to select the list of SNPs in the gene region
(approximately 250 kb of flanking sequence each side) from the
imputed data. These SNPs were pruned with an r2 cutoff of 0.8..
The simpleM method  was used to calculate the number of
informative SNPs within the genomic region for each gene. This
measure was then used in a Bonferroni adjustment to estimate the
significance threshold. Significant SNPs that were imputed or have
a MAF,10% were directly genotyped in all the samples to
confirm the association.
The AD&FTD mutation database  was used to identify
sequence variants previously found in other studies of early onset
familial dementia and to determine whether or not they are
considered to be disease-causative variants. The sequencing data
from the 1,000 Genome Project and the Exome Variant Server
data base (http://evs.gs.washington.edu/EVS/) were used to
estimate the frequency of novel and rare (minor allele frequency
less than 5%) missense, nonsense and splice site variants in
samples unselected for studies of AD. Conservation was
determined by using the GERP score, which calculates the
conservation of each nucleotide in multi-species alignment. A site
was called conserved when the GERP score was greater than or
equal to 4 [69,70].
ADNI material and methods
Data used in the preparation of this article were obtained from
the ADNI database (www.loni.ucla.edu\ADNI). The ADNI was
launched in 2003 by the National Institute on Aging, the National
Institute of Biomedical Imaging and Bioengineering, the Food and
Drug Administration, private pharmaceutical companies and
nonprofit organizations, as a $60 million, 5-year public-private
partnership. The Principal Investigator of this initiative is Michael
W. Weiner, M.D. ADNI is the result of efforts of many
coinvestigators from a broad range of academic institutions and
private corporations, and subjects have been recruited from over
50 sites across the U.S. and Canada. The initial goal of ADNI was
to recruit 800 adults, ages 55 to 90, to participate in the research
approximately 200 cognitively normal older individuals to be
followed for 3 years, 400 people with MCI to be followed for 3
years, and 200 people with early AD to be followed for 2 years.
For up-to-date information see www.adni-info.org.
GERAD data information
Data used in the preparation of this article were obtained from
the Genetic and Environmental Risk for Alzheimers disease
(GERAD1) Consortium . The GERAD1 sample comprised up
to 3941 AD cases and 7848 controls. A subset of this sample has
been used in this study, comprising 3333 cases and 1225 elderly
screened controls genotyped at the Sanger Institute on the
Illumina 610-quad chip. These samples were recruited by the
Medical Research Council (MRC) Genetic Resource for AD
(Cardiff University; Kings College London; Cambridge
University; Trinity College Dublin), the Alzheimers Research Trust (ART)
Collaboration (University of Nottingham; University of
Manchester; University of Southampton; University of Bristol; Queens
University Belfast; the Oxford Project to Investigate Memory and
Ageing (OPTIMA), Oxford University); Washington University,
St Louis, United States; MRC PRION Unit, University College
London; London and the South East Region AD project
(LASERAD), University College London; Competence Network of
Dementia (CND) and Department of Psychiatry, University of
Bonn, Germany and the National Institute of Mental Health
(NIMH) AD Genetics Initiative. All AD cases met criteria for
either probable (NINCDS-ADRDA, DSM-IV) or definite
(CERAD) AD. All elderly controls were screened for dementia
using the MMSE or ADAS-cog, were determined to be free from
dementia at neuropathological examination or had a Braak score
of 2.5 or lower.
Figure S2 Pedigree a family with p.E318G carriers illustrating
the segregation analysis and the absence of APOE e4. A/G is the
genotype for p.E318G variant and 3/3, is the APOE genotype. *
Symbol means confirmed AD by autopsy.
Figure S3 Survival curves comparing age at onset of LOAD
between the different genotypes of Psen1, p.E318G. Survival
fractions were calculated using the Kaplan-Meier method and
significant differences were calculated by Log-rank test.
Association with age at onset was calculated in 21 families with at least
two AD cases carrier.
Information about the known variants.
The authors thank the Clinical and Genetics Cores of the Knight ADRC at
Washington University for clinical and cognitive assessments of the
participants and for APOE genotypes and the Biomarker Core of the Adult
Children Study at Washington University for the CSF collection and
assays, as well as participants and their families, whose help and
participation made this work possible.
Conceived and designed the experiments: BAB AMG CC. Performed the
experiments: BAB CC CMK SCJ BC DC. Analyzed the data: BAB CC
SCJ YC SB LC. Contributed reagents/materials/analysis tools: JAS DAB
ADNI GERAD1 AMF DH JM. Wrote the paper: BAB CC
1. Holtzman DM , Morris JC , Goate AM ( 2011 ) Alzheimer's disease: the challenge of the second century . Sci Transl Med 3 : 77sr71 .
2 . ( 2009 ) 2009 Alzheimer's disease facts and figures . Alzheimers Dement 5 : 234 - 270 .
3. Goate A , Chartier-Harlin MC , Mullan M , Brown J , Crawford F , et al. ( 1991 ) Segregation of a missense mutation in the amyloid precursor protein gene with familial Alzheimer's disease . Nature 349 : 704 - 706 .
4. Gatz M , Mortimer JA , Fratiglioni L , Johansson B , Berg S , et al. ( 2006 ) Potentially modifiable risk factors for dementia in identical twins . Alzheimers Dement 2 : 110 - 117 .
5. So HC , Gui AH , Cherny SS , Sham PC ( 2011 ) Evaluating the heritability explained by known susceptibility variants: a survey of ten complex diseases . Genet Epidemiol 35 : 310 - 317 .
6. Shulman JM , Chibnik LB , Aubin C , Schneider JA , Bennett DA , et al. ( 2010 ) Intermediate phenotypes identify divergent pathways to Alzheimer's disease . PLoS One 5 : e11244 .
7. Tennessen JA , Bigham AW , O'Connor TD , Fu W , Kenny EE , et al. ( 2012 ) Evolution and functional impact of rare coding variation from deep sequencing of human exomes . Science 337 : 64 - 69 .
8. Liu Y , Maxwell S , Feng T , Zhu X , Elston RC , et al. ( 2012 ) Gene, pathway and network frameworks to identify epistatic interactions of single nucleotide polymorphisms derived from GWAS data . BMC Syst Biol 6 Suppl 3 : S15 .
9. Nolan DK , Sutton B , Haynes C , Johnson J , Sebek J , et al. ( 2012 ) Fine mapping of a linkage peak with integration of lipid traits identifies novel coronary artery disease genes on chromosome 5 . BMC Genet 13 : 12 .
10. Cohen JC , Kiss RS , Pertsemlidis A , Marcel YL , McPherson R , et al. ( 2004 ) Multiple rare alleles contribute to low plasma levels of HDL cholesterol . Science 305 : 869 - 872 .
11. Kauwe JS , Cruchaga C , Bertelsen S , Mayo K , Latu W , et al. ( 2010 ) Validating predicted biological effects of Alzheimer's disease associated SNPs using CSF biomarker levels . J Alzheimers Dis 21 : 833 - 842 .
12. Holtzman DM ( 2011 ) CSF biomarkers for Alzheimer's disease: current utility and potential future use . Neurobiol Aging 32 Suppl 1 : S4 - 9 .
13. Bateman RJ , Xiong C , Benzinger TL , Fagan AM , Goate A , et al. ( 2012 ) Clinical and biomarker changes in dominantly inherited Alzheimer's disease . N Engl J Med 367 : 795 - 804 .
14. Cruchaga C , Kauwe JS , Nowotny P , Bales K , Pickering EH , et al. ( 2012 ) Cerebrospinal fluid APOE levels: an endophenotype for genetic studies for Alzheimer's disease . Hum Mol Genet 21 : 4558 - 4571 .
15. Scheuner D , Eckman C , Jensen M , Song X , Citron M , et al. ( 1996 ) Secreted amyloid beta-protein similar to that in the senile plaques of Alzheimer's disease is increased in vivo by the presenilin 1 and 2 and APP mutations linked to familial Alzheimer's disease . Nat Med 2 : 864 - 870 .
16. Fortea J , Llado A , Bosch B , Antonell A , Oliva R , et al. ( 2011 ) Cerebrospinal fluid biomarkers in Alzheimer's disease families with PSEN1 mutations . Neurodegener Dis 8 : 202 - 207 .
17. Cruchaga C , Kauwe JS , Mayo K , Spiegel N , Bertelsen S , et al. ( 2010 ) SNPs associated with cerebrospinal fluid phospho-tau levels influence rate of decline in Alzheimer's disease . PLoS Genet 6 : e1001101 .
18. Kauwe JS , Cruchaga C , Mayo K , Fenoglio C , Bertelsen S , et al. ( 2008 ) Variation in MAPT is associated with cerebrospinal fluid tau levels in the presence of amyloid-beta deposition . Proc Natl Acad Sci U S A 105 : 8050 - 8054 .
19. Kauwe JS , Wang J , Mayo K , Morris JC , Fagan AM , et al. ( 2009 ) Alzheimer's disease risk variants show association with cerebrospinal fluid amyloid beta . Neurogenetics 10 : 13 - 17 .
20. Wallon D , Rousseau S , Rovelet-Lecrux A , Quillard-Muraine M , GuyantMarechal L , et al. ( 2012 ) The French series of autosomal dominant early onset Alzheimer's disease cases: mutation spectrum and cerebrospinal fluid biomarkers . J Alzheimers Dis 30 : 847 - 856 .
21. Balasa M , Vidal-Pineiro D , Llado A , Antonell A , Bosch B , et al. ( 2012 ) PSEN1 mutation carriers present lower cerebrospinal fluid amyoid-beta42 levels than sporadic early-onset Alzheimer's disease patients but no differences in neuronal injury biomarkers . J Alzheimers Dis 30 : 605 - 616 .
22. Ringman JM , Gylys KH , Medina LD , Fox M , Kepe V , et al. ( 2011 ) Biochemical, neuropathological, and neuroimaging characteristics of early-onset Alzheimer's disease due to a novel PSEN1 mutation . Neurosci Lett 487 : 287 - 292 .
23. Jack CR , Jr., Knopman DS , Jagust WJ , Shaw LM , Aisen PS , et al. ( 2010 ) Hypothetical model of dynamic biomarkers of the Alzheimer's pathological cascade . Lancet Neurol 9 : 119 - 128 .
24. Guerreiro R , Wojtas A , Bras J , Carrasquillo M , Rogaeva E , et al. ( 2013 ) TREM2 variants in Alzheimer's disease . N Engl J Med 368 : 117 - 127 .
25. Jonsson T , Stefansson H , Steinberg S , Jonsdottir I , Jonsson PV , et al. ( 2013 ) Variant of TREM2 associated with the risk of Alzheimer's disease . N Engl J Med 368 : 107 - 116 .
26. Benitez BA , Cooper B , Pastor P , Jin SC , Lorenzo E , et al. ( 2013 ) TREM2 is associated with the risk of Alzheimer's disease in Spanish population . Neurobiol Aging 34 : 1711 e1715- 1717 .
27. Gerrish A , Russo G , Richards A , Moskvina V , Ivanov D , et al. ( 2012 ) The role of variation at AbetaPP, PSEN1, PSEN2, and MAPT in late onset Alzheimer's disease . J Alzheimers Dis 28 : 377 - 387 .
28. Cruchaga C , Haller G , Chakraverty S , Mayo K , Vallania FL , et al. ( 2012 ) Rare variants in APP, PSEN1 and PSEN2 increase risk for AD in late-onset Alzheimer's disease families . PLoS One 7 : e31039 .
29. Kauwe JS , Jacquart S , Chakraverty S , Wang J , Mayo K , et al. ( 2007 ) Extreme cerebrospinal fluid amyloid beta levels identify family with late-onset Alzheimer's disease presenilin 1 mutation . Ann Neurol 61 : 446 - 453 .
30. Adzhubei IA , Schmidt S , Peshkin L , Ramensky VE , Gerasimova A , et al. ( 2010 ) A method and server for predicting damaging missense mutations . Nat Methods 7 : 248 - 249 .
31. Poorkaj P , Sharma V , Anderson L , Nemens E , Alonso ME , et al. ( 1998 ) Missense mutations in the chromosome 14 familial Alzheimer's disease presenilin 1 gene . Hum Mutat 11 : 216 - 221 .
32. Scacchi R , Gambina G , Ferrari G , Corbo RM ( 2003 ) Screening of two mutations at exon 3 of the apolipoprotein E gene (sites 28 and 42) in a sample of patients with sporadic late-onset Alzheimer's disease . Neurobiol Aging 24 : 339 - 343 .
33. Coppola G , Chinnathambi S , Lee JJ , Dombroski BA , Baker MC , et al. ( 2012 ) Evidence for a role of the rare p.A152T variant in MAPT in increasing the risk for FTD-spectrum and Alzheimer's diseases . Hum Mol Genet 21 : 3500 - 3512 .
34. Cruts M , Theuns J , Van Broeckhoven C ( 2012 ) Locus-specific mutation databases for neurodegenerative brain diseases . Hum Mutat 33 : 1340 - 1344 .
35. Shaw LM , Vanderstichele H , Knapik-Czajka M , Clark CM , Aisen PS , et al. ( 2009 ) Cerebrospinal fluid biomarker signature in Alzheimer's disease neuroimaging initiative subjects . Ann Neurol 65 : 403 - 413 .
36. Welge V , Fiege O , Lewczuk P , Mollenhauer B , Esselmann H , et al. ( 2009 ) Combined CSF tau, p-tau181 and amyloid-beta 38/40/42 for diagnosing Alzheimer's disease . J Neural Transm 116 : 203 - 212 .
37. Fagan AM , Roe CM , Xiong C , Mintun MA , Morris JC , et al. ( 2007 ) Cerebrospinal fluid tau/beta-amyloid(42) ratio as a prediction of cognitive decline in nondemented older adults . Arch Neurol 64 : 343 - 349 .
38. Gao X , Becker LC , Becker DM , Starmer JD , Province MA ( 2010 ) Avoiding the high Bonferroni penalty in genome-wide association studies . Genet Epidemiol 34 : 100 - 105 .
39. Fagan AM , Head D , Shah AR , Marcus D , Mintun M , et al. ( 2009 ) Decreased cerebrospinal fluid Abeta(42) correlates with brain atrophy in cognitively normal elderly . Ann Neurol 65 : 176 - 183 .
40. Pastor P , Roe CM , Villegas A , Bedoya G , Chakraverty S , et al. ( 2003 ) Apolipoprotein Eepsilon4 modifies Alzheimer's disease onset in an E280A PS1 kindred . Ann Neurol 54 : 163 - 169 .
41. Wijsman EM , Daw EW , Yu X , Steinbart EJ , Nochlin D , et al. ( 2005 ) APOE and other loci affect age-at-onset in Alzheimer's disease families with PS2 mutation . Am J Med Genet B Neuropsychiatr Genet 132B : 14 - 20 .
42. Albani D , Roiter I , Artuso V , Batelli S , Prato F , et al. ( 2007 ) Presenilin-1 mutation E318G and familial Alzheimer's disease in the Italian population . Neurobiol Aging 28 : 1682 - 1688 .
43. Dermaut B , Cruts M , Slooter AJ , Van Gestel S , De Jonghe C , et al. ( 1999 ) The Glu318Gly substitution in presenilin 1 is not causally related to Alzheimer disease . Am J Hum Genet 64 : 290 - 292 .
44. Taddei K , Fisher C , Laws SM , Martins G , Paton A , et al. ( 2002 ) Association between presenilin-1 Glu318Gly mutation and familial Alzheimer's disease in the Australian population . Mol Psychiatry 7 : 776 - 781 .
45. Laws SM , Clarnette RM , Taddei K , Martins G , Paton A , et al. ( 2002 ) Association between the presenilin-1 mutation Glu318Gly and complaints of memory impairment . Neurobiol Aging 23 : 55 - 58 .
46. Helisalmi S , Hiltunen M , Mannermaa A , Koivisto AM , Lehtovirta M , et al. ( 2000 ) Is the presenilin-1 E318G missense mutation a risk factor for Alzheimer's disease ? Neurosci Lett 278 : 65 - 68 .
47. Guerreiro RJ , Baquero M , Blesa R , Boada M , Bras JM , et al. ( 2010 ) Genetic screening of Alzheimer's disease genes in Iberian and African samples yields novel mutations in presenilins and APP . Neurobiol Aging 31 : 725 - 731 .
48. Mattila KM , Forsell C , Pirttila T , Rinne JO , Lehtimaki T , et al. ( 1998 ) The Glu318Gly mutation of the presenilin-1 gene does not necessarily cause Alzheimer's disease . Ann Neurol 44 : 965 - 967 .
49. Aldudo J , Bullido MJ , Frank A , Valdivieso F ( 1998 ) Missense mutation E318G of the presenilin-1 gene appears to be a nonpathogenic polymorphism . Ann Neurol 44 : 985 - 986 .
50. Goldman JS , Johnson JK , McElligott K , Suchowersky O , Miller BL , et al. ( 2005 ) Presenilin 1 Glu318Gly polymorphism: interpret with caution . Arch Neurol 62 : 1624 - 1627 .
51. Kauwe JS , Cruchaga C , Karch CM , Sadler B , Lee M , et al. ( 2011 ) Fine mapping of genetic variants in BIN1, CLU, CR1 and PICALM for association with cerebrospinal fluid biomarkers for Alzheimer's disease . PLoS One 6 : e15918 .
52. Jin SC , Pastor P , Cooper B , Cervantes S , Benitez BA , et al. ( 2012 ) Pooled-DNA sequencing identifies novel causative variants in PSEN1, GRN and MAPT in a clinical early-onset and familial Alzheimer's disease Ibero-American cohort . Alzheimers Res Ther 4 : 34 .
53. Schymick JC , Yang Y , Andersen PM , Vonsattel JP , Greenway M , et al. ( 2007 ) Progranulin mutations and amyotrophic lateral sclerosis or amyotrophic lateral sclerosis-frontotemporal dementia phenotypes . J Neurol Neurosurg Psychiatry 78 : 754 - 756 .
54. Kamboh MI , Sanghera DK , Ferrell RE , DeKosky ST ( 1995 ) APOE*4- associated Alzheimer's disease risk is modified by alpha 1-antichymotrypsin polymorphism . Nat Genet 10 : 486 - 488 .
55. Rodriguez E , Mateo I , Infante J , Llorca J , Berciano J , et al. ( 2006 ) Cholesteryl ester transfer protein (CETP) polymorphism modifies the Alzheimer's disease risk associated with APOE epsilon4 allele . J Neurol 253 : 181 - 185 .
56. Reiman EM , Webster JA , Myers AJ , Hardy J , Dunckley T , et al. ( 2007 ) GAB2 alleles modify Alzheimer's risk in APOE epsilon4 carriers . Neuron 54 : 713 - 720 .
57. Wijsman EM , Pankratz ND , Choi Y , Rothstein JH , Faber KM , et al. ( 2011 ) Genome-wide association of familial late-onset Alzheimer's disease replicates BIN1 and CLU and nominates CUGBP2 in interaction with APOE . PLoS Genet 7 : e1001308 .
58. Keenan BT , Shulman JM , Chibnik LB , Raj T , Tran D , et al. ( 2012 ) A coding variant in CR1 interacts with APOE-epsilon4 to influence cognitive decline . Hum Mol Genet 21 : 2377 - 2388 .
59. Knappenberger KS , Tian G , Ye X , Sobotka-Briner C , Ghanekar SV , et al. ( 2004 ) Mechanism of gamma-secretase cleavage activation: is gamma-secretase regulated through autoinhibition involving the presenilin-1 exon 9 loop? Biochemistry 43 : 6208 - 6218 .
60. Murayama O , Murayama M , Honda T , Sun X , Nihonmatsu N , et al. ( 1999 ) Twenty-nine missense mutations linked with familial Alzheimer's disease alter the processing of presenilin 1 . Prog Neuropsychopharmacol Biol Psychiatry 23 : 905 - 913 .
61. Hashimoto Y , Tsukamoto E , Niikura T , Yamagishi Y , Ishizaka M , et al. ( 2004 ) Amino- and carboxyl-terminal mutants of presenilin 1 cause neuronal cell death through distinct toxic mechanisms: Study of 27 different presenilin 1 mutants . J Neurosci Res 75 : 417 - 428 .
62. De Jager PL , Shulman JM , Chibnik LB , Keenan BT , Raj T , et al. ( 2012 ) A genome-wide scan for common variants affecting the rate of age-related cognitive decline . Neurobiol Aging 33 : 1017 e1011- 1015 .
63. McKhann G , Drachman D , Folstein M , Katzman R , Price D , et al. ( 1984 ) Clinical diagnosis of Alzheimer's disease: report of the NINCDS-ADRDA Work Group under the auspices of Department of Health and Human Services Task Force on Alzheimer's Disease . Neurology 34 : 939 - 944 .
64. Berg L , McKeel DW , Jr., Miller JP , Storandt M , Rubin EH , et al. ( 1998 ) Clinicopathologic studies in cognitively healthy aging and Alzheimer's disease: relation of histologic markers to dementia severity , age, sex, and apolipoprotein E genotype. Arch Neurol 55 : 326 - 335 .
65. Druley TE , Vallania FL , Wegner DJ , Varley KE , Knowles OL , et al. ( 2009 ) Quantification of rare allelic variants from pooled genomic DNA . Nat Methods 6 : 263 - 265 .
66. Haller G , Druley T , Vallania FL , Mitra RD , Li P , et al. ( 2012 ) Rare missense variants in CHRNB4 are associated with reduced risk of nicotine dependence . Hum Mol Genet 21 : 647 - 655 .
67. Vallania FL , Druley TE , Ramos E , Wang J , Borecki I , et al. ( 2010 ) Highthroughput discovery of rare insertions and deletions in large cohorts . Genome Res 20 : 1711 - 1718 .
68. Price AL , Patterson NJ , Plenge RM , Weinblatt ME , Shadick NA , et al. ( 2006 ) Principal components analysis corrects for stratification in genome-wide association studies . Nat Genet 38 : 904 - 909 .
69. Cooper GM , Goode DL , Ng SB , Sidow A , Bamshad MJ , et al. ( 2010 ) Singlenucleotide evolutionary constraint scores highlight disease-causing mutations . Nat Methods 7 : 250 - 251 .
70. Cooper GM , Stone EA , Asimenos G , Green ED , Batzoglou S , et al. ( 2005 ) Distribution and intensity of constraint in mammalian genomic sequence . Genome Res 15 : 901 - 913 .
71. Harold D , Abraham R , Hollingworth P , Sims R , Gerrish A , et al. ( 2009 ) Genome-wide association study identifies variants at CLU and PICALM associated with Alzheimer's disease . Nat Genet 41 : 1088 - 1093 .