Genome-Wide Association Identifies a Common
Variant in the Reelin Gene That Increases
the Risk of Schizophrenia Only in Women
Sagiv Shifman1*¤, Martina Johannesson1, Michal Bronstein2, Sam X. Chen3,4, David A. Collier5, Nicholas J. Craddock6,
Kenneth S. Kendler3,4, Tao Li5,7,8, Michael O’Donovan6, F. Anthony O’Neill9, Michael J. Owen6, Dermot Walsh10,
Daniel R. Weinberger11, Cuie Sun3,4, Jonathan Flint1, Ariel Darvasi2
1 Wellcome Trust Centre for Human Genetics, University of Oxford, Oxford, United Kindgdom, 2 Department of Genetics, The Institute of Life Sciences, The Hebrew
University of Jerusalem, Jerusalem, Israel, 3 Department of Psychiatry, Virginia Commonwealth University, Richmond, Virginia, United States of America, 4 Virginia Institute
for Psychiatric and Behavior Genetics, Virginia Commonwealth University, Richmond, Virginia, United States of America, 5 Social, Genetic and Developmental Psychiatry
Centre, Institute of Psychiatry, King’s College London, London, United Kingdom, 6 Department of Psychological Medicine, School of Medicine, Cardiff University, Cardiff,
United Kingdom, 7 Psychiatric Laboratory and Department of Psychiatry, West China Hospital, Sichuan University, The People’s Republic of China, 8 State Key Laboratory of
Biotherapy, West China Hospital, Sichuan University, The People’s Republic of China, 9 The Department of Psychiatry, The Queens University, Belfast, Northern Ireland, 10 The
Health Research Board, Dublin, Ireland, 11 Clinical Brain Disorders Branch, Genes, Cognition, and Psychosis Program, Intramural Research Program, National Institute of
Mental Health, National Institutes of Health (NIH), United States Department of Health and Human Services, Bethesda, Maryland, United States of America
Sex differences in schizophrenia are well known, but their genetic basis has not been identified. We performed a
genome-wide association scan for schizophrenia in an Ashkenazi Jewish population using DNA pooling. We found a
female-specific association with rs7341475, a SNP in the fourth intron of the reelin (RELN) gene (p ¼ 2.9 3 10?5in
women), with a significant gene-sex effect (p ¼ 1.8 3 10?4). We studied rs7341475 in four additional populations,
totaling 2,274 cases and 4,401 controls. A significant effect was observed only in women, replicating the initial result (p
¼2.1310?3in women; p¼4.2310?3for gene-sex interaction). Based on all populations the estimated relative risk of
women carrying the common genotype is 1.58 (p ¼ 8.8 3 10?7; p ¼ 1.6 3 10?5for gene-sex interaction). The female-
specific association between RELN and schizophrenia is one of the few examples of a replicated sex-specific genetic
association in any disease.
Citation: Shifman S, Johannesson M, Bronstein M, Chen SX, Collier DA, et al. (2008) Genome-wide association identifies a common variant in the reelin gene that increases the
risk of schizophrenia only in women. PLoS Genet 4(2): e28. doi:10.1371/journal.pgen.0040028
Schizophrenia (181500) is a common psychiatric disorder
of unknown aetiology. Individual twin studies and meta-
analyses of twin studies  estimate that the heritability of
schizophrenia is approximately 80%. Analysis of family,
adoption and twin data indicate that inheritance acts in a
complex fashion, in combination with the environment, to
mediate the risk of developing the illness . However,
despite the relatively large heritability of schizophrenia,
efforts to identify the molecular risk factors have so far
yielded equivocal results (reviewed in ).
Sex differences in the risk of a disorder can provide clues
about its pathogenesis. For schizophrenia, the age of onset,
premorbid functioning, symptomatic characteristics, and
course of illness differ significantly between men and women
. Two systematic reviews have demonstrated a sex differ-
ence in the risk of developing schizophrenia [5,6]; both
studies report that the male to female risk ratio is 1.4. Sex-
specific associations with schizophrenia have previously been
reported for a number of loci [7–11], but the robustness of
these claims is open to doubt; results have yet to be
corroborated [8–10] or replication has not been found with
the same single nucleotide polymorphism (SNP) in the same
direction in the same sex (e.g. [7,11]). This difficulty has
afflicted attempts to establish sex-specific association in other
diseases. An empirical assessment of 432 published sex
differences in genetic association studies for different
conditions found a single valid interaction that was consis-
tently replicated in at least two other studies .
In the present study, we carried out a genome-wide
association study using DNA pools of cases and controls
constructed separately for men and women to allow the
identification of sex-specific effects. Several studies have
shown that DNA pooling detects the most promising loci with
considerable savings in time and costs [13–18]. We previously
scanned a three Mb region spanning the 22q11 microdeletion
for association with schizophrenia using DNA pools. Our
previous study [7,19] showed that the pools are representative
of the allele frequencies in the sample and are adequate for
Editor: Greg Gibson, North Carolina State University, United States of America
Received October 5, 2007; Accepted December 17, 2007; Published February 15,
Copyright: ? 2008 Shifman et al. This is an open-access article distributed under
the terms of the Creative Commons Attribution License, which permits unrestricted
use, distribution, and reproduction in any medium, provided the original author
and source are credited.
Abbreviations: CEU, Utah residents with ancestry from northern and western
Europe; LD, linkage disequilibrium; MAF, minor allele frequency; OR, odds ratio;
SNP, single nucleotide polymorphism
* To whom correspondence should be addressed. E-mail: email@example.com
¤ Current address: Department of Genetics, The Institute of Life Sciences, The
Hebrew University of Jerusalem, Jerusalem, Israel
PLoS Genetics | www.plosgenetics.org February 2008 | Volume 4 | Issue 2 | e280001
detection of even modest association signals (odds ratios of
Four separate pools of DNA were constructed as follows,
419 male cases, 241 female cases, 1,807 male controls and 964
female controls. Increasing the number of controls for a
given number of cases significantly increases the power of the
experiment. For example, our experiment with 660 cases and
2,771 controls is equivalent in power to about 1,100 cases and
1,100 controls . Since estimating allele frequencies in
pooled DNA samples introduces measurement errors, each of
the four pools was independently analyzed in ten replicates
with Affymetrix 500K SNP arrays. The results from the ten
replicates were used to rank the SNPs (the 1,000 top rank-
ordered SNPs are listed in Table S1).
Our previously reported findings with several SNPs around
both the COMT (116790) and DGCR2 (600594) genes in
Chromosome 22 acted as a positive control for the perform-
ance of the current study. The association of multiple SNPs in
the vicinity of both genes was clearly identified using our
method (Figure 1). The best SNPs were selected for individual
genotyping using an integrative approach taking into account
their statistical significance, ranking and potential biological
relevance (see Methods). As a result, a total of 194 SNPs were
selected for individual genotyping.
The 194 selected SNPs were individually genotyped. At this
stage, we used a sample of 759 controls and 745 patients from
the Ashkenazi Jewish population. We genotyped a subset of
the control sample used in the pools and enlarged the sample
of cases (compared to the sample used in the pools), in order
to differentiate in a cost-effective way between SNPs with real
differences in allele frequencies between cases and controls
and SNPs showing spurious differences due to technical
anomalies in the DNA pooling procedure. Fifty-two SNPs out
of 167 SNPs that passed our quality control (including a
manual examination of genotype clusters and test of Hardy
Weinberg equilibrium) showed p-values below 0.05 at any of
the tests (male, female and combined) and nine SNPs had p-
values below 0.005 (specific results for all SNPs are presented
in Table S2).
The lowest p-value was found for SNP rs7341475 (G!A),
for women only. SNP rs7341475 is located in Chromosome 7
(bp position ¼ 103192051, NCBI Build 36), in the fourth
intron of the reelin gene (RELN (600514)). This particular SNP
was prioritized for individual genotyping because it resides
within a gene previously studied for association with
schizophrenia, in addition to having a high rank in the pools
result (top 99.98% in the female pools). The common
genotype of this SNP (GG) has a higher frequency in women
with schizophrenia (75.5%) relative to female controls
(59.3%) with an odds ratio (ORGG) of 2.1 (pgenotype¼ 9.8 3
10?5), and a significant gene-sex interaction (pinteraction¼5.33
10?3). In men, however, no effect was observed (ORGG¼ 1.1,
pgenotype ¼ 0.47, GG frequency ¼ 60.6%). Since the AA
genotype is relatively rare, the genotype distribution in cases
and controls was analyzed with the GA and AA genotypes
grouped together. Similarly, the G allele was significantly
overrepresented in women with schizophrenia (frequency ¼
86.6%) relative to female controls (frequency ¼ 76.2%). The
odds ratio for the G allele was ORG¼ 2.0 with a correspond-
ing allelic statistical significance of pallele¼1.9310?5. Again in
men no effect was observed (ORG¼1.1, pallele¼0.38). The most
significant result was obtained when the allele frequencies in
female cases were compared to a combined sample of male
and female control individuals (pallele¼ 4.8 3 10?7). While a p
, 5 3 10?7cannot on its own be considered statistically
significant under a stringent Bonferroni correction, it is
suggestive evidence for true association under some assump-
tions , and therefore was further studied. We expanded
the size of the Ashkenazi Jewish sample by increasing the
number of controls (656 female and 1,988 males). We found
increased evidence for the genotype association (pgenotype¼
2.92310?5; ORGG¼2.0; Table 1) and the gene-sex interaction
(pinteraction¼ 1.8 3 10?4). The association for males remained
non-significant (pgenotype¼ 0.62).
We considered whether the result could be due to
Figure 1. Association Signal on Chromosome 22 in a Sliding-window of
Nine Consecutive SNPs
The ?log10 values are the negative logarithm (base 10) of the sliding
windows ranks divided by the number of windows on Chromosome 22.
The combined rank across men and women is based on a silhouette
statistic . The inset shows a magnification of the region indicated by
a vertical line and the position of DGCR2 and COMT genes within this
PLoS Genetics | www.plosgenetics.orgFebruary 2008 | Volume 4 | Issue 2 | e280002
Reelin and Schizophrenia
Schizophrenia is a complex mental disease, which includes
symptoms of delusions, hallucinations, disorganized speech, aber-
rant behavior, lack of emotional expression, diminished motivation,
and social withdrawal. The cause of schizophrenia is unknown, but
there is extensive evidence that genetics play a significant role in its
aetiology. We studied the genetic basis of schizophrenia by
analyzing around 500,000 genetic variants distributed across the
whole human genome in DNA from schizophrenic patients and
controls. We analyzed separately the DNA from men and women,
and identified a genetic variant that increases the risk of developing
schizophrenia in women only. The genetic variant is estimated to
increase the risk of schizophrenia for women carrying the risk variant
by 1.4-fold. The genetic variant is in a gene called reelin, which is
known to play a part in brain development. However, it is still
unclear how this genetic variant predisposes to schizophrenia nor
why it is specific to women only.
population stratification effect. We categorized parents of the
Ashkenazi Jews individuals by their country of origin and
compared the allele frequencies of rs7341475. We examined
subjects whose parents were from the same country (77% of
all individuals) and focused on those countries with more
than 100 individuals (80% of individuals): Argentina, Ger-
many, Russia, Poland, Ukraine, and USA (Figure S1). There
was no significant difference in allele frequency between
Ashkenazi Jewish individuals originated from different areas
of the world (p ¼ 0.9405).
The association results and the linkage disequilibrium (LD)
structure of RELN are presented in Figure 2. Based on
HapMap LD data (CEU), there are no other SNPs in high LD
(maximum r2¼ 0.4) with rs7341475 on the 500K Affymetrix
array. One other SNP (rs2106173) in the RELN gene was also
tested by individual genotyping but no statistically significant
difference was found. The association signals for other SNPs
in RELN (based on DNA pools) were moderate (maximum
rank: for men ¼ 1,062, for women ¼ 1,969; Figure 2). We
studied the patterns of LD in RELN in the Ashkenazi Jewish
population using the genotypes of 129 SNPs (distributed
between 102783336 and 103533335, NCBI Build 36) in 197
unrelated control individuals (obtained in the course of
another study ). We found that the block structure in our
sample from the Ashkenazi Jewish population is very similar
to that of HapMap CEU. High correlations (r2. 0.8) are
observed between rs7341475 and other SNPs distributed up
to 39.8 kb upstream to rs7341475. The SNPs that are
highly correlated with rs7341475 (rs10435342, rs6951931,
rs17290575, rs6954479 and rs39327) are all located in the
third or fourth intron of the gene. There is a substantial
correlation (r2. 0.5) between rs7341475 and other SNPs in a
77.4 kb interval (103188857–103266305). There are no SNPs
in neighboring genes in high correlation with rs7341475.
To confirm the female-specific association between
rs7341475 and schizophrenia, we tested rs7341475 in four
other sample sets, three of European ancestry (UK, Ireland
and USA) and one Chinese. In this replication study we tested
a total of 2,274 cases (768 women and 1,506 men) and 4,401
controls (2,194 women and 2,207 men). We separated all
samples by sex and tested the association for male and female
subjects. Based on the association in the Ashkenazi Jewish
population, in the replication samples we tested the
prediction that the frequency of the GG genotype is
increased in women with schizophrenia. The female-specific
association of rs7341475 with schizophrenia was replicated in
the UK case-control group with an effect in the same
direction (ORGG ¼ 1.85; pgenotype ¼ 1.8 3 10?3), and a
significant sex effect (pinteraction ¼ 3.2 3 10?3). All other
populations showed an effect in the same direction, although
individually the effects were not significant (Figure 3; Table
1). Combining all replication samples yielded a genotype
ORGG of 1.41 (95% CI ¼ 1.13–1.76) with a corresponding
pgenotypeof 2.1310?3for women and ORGGof 0.97 (95% CI¼
0.83–1.15, pgenotype¼0.76) for men. The odds ratio for women
in the combined replication set was significantly higher than
men (pinteraction¼4.2310?3). The association in the combined
replication samples remained significant (pgenotype¼ 0.045)
even after excluding the UK sample, which shows that the
result of the meta-analysis is not attributable to a single
highly significant sample.
The female specific ORGG, from all samples, including the
initial Ashkenazi Jewish sample is 1.58 (1.31–1.89) with a
corresponding pgenotypeof 8.8 3 10?7, pinteractionof 1.6 3 10?5,
and a female-population attributable risk of 50%. Note that
the estimated risk effect including the Ashkenazi Jewish
sample might be slightly inflated, because rs7341475 was
selected as one of the best performing SNPs from a large
number of SNPs examined. Thus an unbiased estimate for the
risk of this SNP, should be based on the replication samples
(ORGG¼ 1.41, or 40% for the female attributable risk), even
though there was no statistical evidence of heterogeneity of
the odds ratios across the studies (p¼0.20). The frequency of
the GG genotype, however, varies between populations (Table
1) – highest in the Chinese sample (frequency ¼ 82.8%) and
lowest in the Ashkenazi Jewish population (frequency ¼
We have carried out a genome-wide association analysis of
schizophrenia, using pooled DNA. We identified one SNP in
the fourth intron of RELN that confers a sex-specific risk of
schizophrenia. Although the significance of the association
between rs7341475 and schizophrenia in the Ashkenazi
Jewish sample would not withstand a Bonferroni correction
for multiple testing, we were able to replicate the sex-specific
association in other samples from different populations that
Table 1. Frequency of GG Genotype (Sample Size)
MenWomen Men Women
2.9 3 10?5
1.8 3 10?3
AJ, Ashkenazi Jews; PGG, significance level for differences in the frequency of the GG genotype between female cases and controls, calculated with a v2statistics. For men, PGGis not
significant for all samples.
PLoS Genetics | www.plosgenetics.org February 2008 | Volume 4 | Issue 2 | e28 0003
Reelin and Schizophrenia
were tested for this specific SNP. The same allele and
genotype are overrepresented in women, but not men, with
schizophrenia in three different populations, Ashkenazi Jews,
Europeans and Chinese (although the overrepresentation is
not independently statistically significant in all populations
tested). This observation, together with the fact that the
association in the combined replication samples is significant
and robust (even to the removal of the sample showing the
most significant association) increases our confidence that we
have found a genuine association.
The association observed in this study is unlikely to be the
result of population stratification. The samples from the UK
showing robust replication of the initial association were
individually genotyped for SNPs across the whole genome. As
such, this sample was rigorously evaluated for possible
population stratification . We also found no evidence
for stratification in the Ashkenazi sample, indicating that the
increase in allele frequency observed in women with
schizophrenia cannot be caused by population structure in
the Ashkenazi Jewish sample.
Genetic association studies have so far failed to report any
consistent association between Reelin gene polymorphisms
and schizophrenia [23–27]. However the gene has not so far
been systematically screened. According to HapMap data, 183
tag SNPs would be needed to capture common variation at
the gene (with r2. 0.8, MAF . 5%), while the SNPs on the
Affymetrix 500K arrays capture 60% of SNPs. Most studies
report data on a SNP in the promoter region, or on a CGG
repeat polymorphism in the 59-untranslated region; none
have tested rs7341475. The two previously reported poly-
morphisms are located in LD blocks that do not contain
Our finding is important on two counts: first, it supports
the hypothesis of a neurodevelopmental origin for schizo-
phrenia, assuming that the genetic association reflects
variation in the function of the RELN gene. Reelin, the
Figure 2. Association Signal for Schizophrenia in the Reelin Gene
(I) The ?log values are the negative logarithm (base 10) of the rank of each SNP divided by the number of SNPs on the array. The rank is based on a
silhouette statistic , calculated separately for DNA pools of men (red) and women (blue) using the GenePool software. Arrows indicate the most
significant SNPs, genotyped individually. The signal from rs2106173 was found to be an artifact of the DNA pooling experiment. (II) The structure and
location of the reelin gene is from the Genome Browser of the University of California, Santa Cruz. Blue boxes represent the gene exons. (III) Similar
patterns of linkage disequilibrium (LD) in Ashkenazi Jewish population (ASH) and individuals with European ancestry from HapMap (CEU). The blue
arrows indicate the position of SNP rs7341475. Figures were generated using the Haploview software (http://www.broad.mit.edu/mpg/haploview/),
with the standard color scheme. Pairwise LD levels between the SNPs in the region are represented by the color of the squares, which increase from
white to blue to red (white, disequilibrium coefficient (D’) , 1 and LOD score , 2; blue, D’ ¼1 and LOD score , 2; pink or light red, D’ , 1 and LOD
score ? 2; and bright red, D’ ¼ 1 and LOD score ? 2).
PLoS Genetics | www.plosgenetics.orgFebruary 2008 | Volume 4 | Issue 2 | e28 0004
Reelin and Schizophrenia
protein product of RELN, is a serine protease  that acts
via a number of receptor-mediated pathways on neurons .
It plays a key part in corticogenesis, as demonstrated by the
cytoarchitectural abnormalities of the null mutant reeler
(rl?/?) mouse . RELN mutations in humans are associated
with an autosomal recessive form of lissencephaly , a
mental retardation syndrome that does not include psychosis.
Second, while a sex difference in the risk for schizophrenia
has been found, its molecular basis has so far been unclear.
Here we establish a replicated sex-specific association.
Intriguingly, there is prior evidence for sexual dimorphism
at the reelin locus. Sex effect have been noted in one study,
reporting that RELN expression was higher in women
compared to men (in layer I neurons) and a reduction in
RELN expression observed only in men with schizophrenia
(in the superficial interstitial white matter neurons). How-
ever, as the authors acknowledge, these have not been noted
in other brain regions . In mice, loss of Purkinje cell was
observed in male mice with only one functional reelin gene
(rl þ/?) but not in reelin-deficient female mice .
Finally we note that sex hormones are likely to mediate
changes in RELN expression: for example, administration of
testosterone decreases reelin expression in the brains of male
European starlings . Our result of a female-specific
association of RELN with schizophrenia may suggest a
possible pathway where sex hormones modulate gene
expression, which by altering cortical structure, increases
susceptibility to psychosis.
Materials and Methods
Ashkenazi Jewish sample. Samples from individuals diagnosed with
schizophrenia were collected from hospitalized patients from seven
medical centers in Israel (described in ). Schizophrenia was
diagnosed by Diagnostic and Statistical Manual criteria (DSM-IV,
American Psychiatric Association). Ashkenazi controls were collected
from volunteers in blood banks. All four grandparents of each subject
were of Ashkenazi Jewish origin, and all subjects (or the subject’s legal
representative) signed an informed-consent form.
Other studied samples. Chinese Han schizophrenia patients, their
families and control subjects were recruited from Sichuan Province,
SW China and consisted of 415 unrelated patients (222 males and 193
females) and 458 normal Han Chinese controls (229 males and 229
females). All patients were interviewed by an experienced psychiatrist
using the SCID, and a diagnosis of schizophrenia was made according
to DSM-III-R or DSM-IV criteria. Information was also collected from
examination of medical records and all other available sources.
Genomic DNA was extracted from peripheral blood according to
standard phenol-chloroform methods. This study was approved by
the South London and Maudsley Trust ethical committee and
informed consent was obtained from all patients and control
The Irish Case – Control Study of Schizophrenia (ICCSS) samples
were collected in Northern Ireland and the Republic of Ireland.
Individual genotypes were obtained for 980 affected cases (669 males
and 311 females) and 582 controls (337 males, 245 females). The
affected subjects were selected from in-patient and outpatient
psychiatric facilities in the Republic of Ireland and Northern Ireland.
Subjects were eligible for inclusion if they had a diagnosis of
schizophrenia or poor-outcome schizoaffective disorder by DSM-III-
R criteria, which were confirmed by a blind expert diagnostic review.
Excluding the relatively small number of schizoaffective female
patients (n ¼ 36) did not significantly alter the female-specific
association result. Controls, selected from several sources, including
blood donation centers, were included if they denied a lifetime
history of schizophrenia. Both cases and controls were included only
if they reported all four grandparents as being born in Ireland or the
The UK sample consisted of unrelated subjects with schizophrenia
(320 males and 155 females). All were white and born in the British
Isles. All patients had a consensus diagnosis of schizophrenia
according to DSMIV criteria made by two independent raters
following a semi-structured interview by trained psychiatrists or
psychologists using the Present State Examination or the Schedules
for Clinical Assessment in Neuropsychiatry (SCAN) interview and
review of case records. Cases were screened to exclude substance-
induced psychotic disorder or psychosis due to a general medical
condition. The mean age at first psychiatric contact was 23.6 (SD 7.7)
years and the mean at ascertainment was 41.7 (SD 14.6) years.
Multicentre and Local Research Ethics Committee approval were
obtained, and all subjects gave written informed consent to
participate. The control sample used by the Wellcome Trust Case
Control Consortium (WTCCC) is described in detail elsewhere .
Briefly, controls (n ¼ 2938) came from two sources, the 1958 British
Birth Cohort (58C) and UK blood donors. At a genome wide level, the
two groups do not significantly differ with respect to allele
frequencies, justifying their use as a single control group. Individuals
included in the study were living within England, Scotland and Wales.
Individuals (n ¼ 26) with non-Caucasian ancestry as determined by
multidimensional scaling (MDS) were previously removed by the
WTCCC from the sample. Controls were not screened for psychosis,
but given the expected modest frequency in an unscreened sample,
this has little effect on power. For the genome-wide association study
from which these data were taken (manuscript submitted), k was
estimated at 1.06 genome-wide, which is at the lower end observed for
other phenotypes when compared with the same controls. The call
rate at this test locus for subjects in the genome-wide association
study was .99.5% in each of cases and controls.
The USA sample consisted of unrelated subjects with schizophre-
nia (295 males and 105 females) and unrelated controls (202 males
and 232 females) selected from a family-based sample that was
Figure 3. Risk Effects of the GG Genotype of rs7341475 Estimated by Odds Ratio Separately for Men and Women across Samples
The point estimate of the odds ratio for each sample is represented by a square with a size proportional to the weight of study. The 95% confidence
interval (95% CI) for each study is represented by a horizontal line. The confidence interval for the combined data is represented by a diamond.
PLoS Genetics | www.plosgenetics.orgFebruary 2008 | Volume 4 | Issue 2 | e28 0005
Reelin and Schizophrenia
ascertained as part of the Clinical Brain Disorders Branch/National
Institute of Mental Health Sibling Study. All subjects were diagnosed
using the Structured Clinical Interview (SCID). Probands met DSM-IV
criteria for broad schizophrenia diagnosis. Control individuals were
ascertained from the National Institutes of Health Normal Volunteer
Office and were screened by SCID diagnosis for psychiatric disorders
and excluded also if they had a first degree relative with a
schizophrenia spectrum diagnosis. All participants gave informed
consent and self-identified as Caucasian of European ancestry. All
subjects also underwent extensive physical and laboratory screening
to rule out complicating medical conditions and substance abuse.
DNA pooling. The DNA pools were constructed using DNA
samples from 660 schizophrenic patients (419 males and 241 females)
and 2,771 controls (1,807 males and 964 females). Pools were created
by using equal aliquots of each sample, as described in ref. [7,19]. The
pools were constructed using DNA samples from both sexes
separately. Each pool (n ¼ 4) was allelotyped ten times with the
500K Affymetrix arrays sets as previously described .
SNP selection for individual genotyping. We selected 194 unique
SNPs using three different ranking methods (Figure S2), in addition
to prioritizing SNPs in genes previously studied for association with
schizophrenia (schizophrenia candidate genes). We used the Schizo-
phreniaGene database  to identify genes that were previously
tested for association with schizophrenia (515 genes were listed at the
time of access [11/05/07]). SNPs with minor allele frequencies below
5%, according to HapMap (CEU) were excluded from further analysis
due to the limited power to detect a significant association and the
decrease in the accuracy of DNA pooling with rare variants.
Eighty SNPs were selected by the following criteria (some SNPs
were selected by more than one method). We calculated for each SNP
a modified chi-squared statistic, which includes the error variance
introduced by the DNA pooling procedure [18,19] . SNPs were ranked
based on logarithm (base 10) of the p-value (logP). Sixty first-ranked
SNPs (30 from each array type, Nsp & Sty) were selected for
individual genotyping. Additional twenty SNPs with logP . 3 were
selected because they reside in candidate genes.
Fifty-five SNPs were selected by the following criteria. We used the
GenePool software with silhouette statistic and Manhattan distance as
the distance measure . The SNPs were ranked based on the
silhouette statistic separately for males and females patients, ranging
from 1 for the highest silhouette score to 510,552 for the lowest score.
We generated a combined rank for all samples based on the average
rank in males and females weighted for the different size of samples.
The first ten ranked SNPs with a minor allele frequency above 5%
(based on HapMap, CEU) were selected for individual genotyping.
This was done for each of three sets: male, female and combined.
Twenty-five additional SNPs in candidate genes for schizophrenia
were also selected for individual genotyping if they were also ranked
in the first 1,000 SNPs in any of the sets.
Eighty-two SNPs were selected by the following criteria: first, we
calculated the mean rank score in a sliding-window of nine
consecutively neighboring SNPs. For each SNP we assigned a rank
based on the mean score in the window showing the largest score out
of the different possible windows. The sliding window method was
used to identify regions where neighboring SNPs consistently show
differences in allele frequency between cases and controls. This
method minimizes spurious differences arising from technical
anomalies in the analysis of pooled DNA , but could preferentially
select SNPs in large blocks of LD. The analysis was carried out
separately for male, females and the combined sets. SNPs in the first
1,000 ranked windows were sorted based on the SNP’s individual rank
score. The first 60 SNPs for the combined sample and the first ten for
males and females samples were selected for individual genotyping.
Two additional SNPs in candidate genes for schizophrenia that were
also ranked in the first 1,000 SNPs in the combined analysis were also
selected for individual genotyping.
Individual genotyping. DNA samples from the Ashkenazi Jewish
and Chinese populations were genotyped using the Sequenom iPLEX
system. LD in the Ashkenazi Jewish control sample was assessed using
genotypes obtained from an Illumina HumanHap300 BeadChip in
the course of another study . Samples from the Irish Case –
Control Study of Schizophrenia and from the US were genotyped
using a Taqman 59-exonuclease allelic discrimination assay. Quality
checks for the Sequenom system included concordance rate (.98%)
for genotypes of 83 DNA samples from HapMap cell lines, call rate
(.95%), Hardy-Weinberg equilibrium in control (p . 0.001), manual
checks of genotypes clusters and concordance rate (.98%) for re-
genotyping. Genotypes for the UK sample were taken from a genome-
wide association study performed in concert with the WTCCC study
of common diseases  based upon the GeneChip 500K Mapping
Array Set. Samples were genotyped at the Affymetrix service
laboratory in San Francisco (USA) using the same pipeline as the
WTCCC disease and controls samples.
Statistical analysis. All data analysis was performed using the R
language and environment for statistical computing (http://www.
r-project.org/). Single-SNP analysis for the individual genotyping data
was carried out using a v2test on allele and genotype counts. For the
replication samples a one-tailed test was employed, since we tested a
specific hypothesis of an increase in GG frequency in female patients
and not other possible directions of association. We combined the
rare homozygote genotype with the heterozygotes for the genotype
association analysis. Hardy-Weinberg was assessed using the v2
statistic with one degree of freedom. To test for a gene by sex
interaction, a z-score was calculated using a ratio between the
difference in the natural logarithm of the odds ratio between males
and females and the square root of the variance of the difference. We
used the Mantel-Haenszel method to combine the data of different
populations with a fixed effect model and Cochran’s Q statistic to test
for heterogeneity as implemented in the R package ‘meta’ (version
0.5). LD was calculated using the Haploview software package.
Population attributable risk of rs7341475 was calculated for women
as (K ? 1)/K, where K ¼Pfi3 gi; fiis the frequency of the i genotype,
assuming multiplicative model.
and gi is the estimated genotype relative risk of the i genotype
Figure S1. Allele Frequency of rs7341475 in Ashkenazi Jews from
The height of each bar represents the frequency of the G allele of
rs7341475 in Ashkenazi Jews originated from different countries.
Error bars are 95% confidence interval for the frequency. The
vertical solid line is the frequency of the G allele in schizophrenia
female patients and the dashed lines are the 95% confidence interval
for that frequency.
Found at doi:10.1371/journal.pgen.0040028.sg001 (42 KB DOC).
Figure S2. SNP Selection for Individual Genotyping
The diagram shows the method used to select SNPs for individual
genotyping. SNPs were selected based on three ranking methods.
SNPs were also prioritized in each ranking system if they reside
within genes previously studied for association with schizophrenia
(see text for more details).
Found at doi:10.1371/journal.pgen.0040028.sg002 (28 KB DOC).
Table S1. Results of DNA Pooling
The 1,000 top rank-ordered SNPs based on a silhouette statistic
calculated using the GenePool software for men, women, and
Found at doi:10.1371/journal.pgen.0040028.st001 (321 KB XLS).
Table S2. Results of Individual Genotyping
Found at doi:10.1371/journal.pgen.0040028.st002 (52 KB XLS).
The Online Mendelian Inheritance in Man (OMIM) (http://www.ncbi.
nlm.nih.gov/Omim/) accession numbers for disease and genes
mentioned in the paper are Schizophrenia (181500), RELN
(600514), COMT (116790), and DGCR2 (600594).
Author contributions. The study was conceived by AD and
designed by SS. SS and MJ performed the genome-wide association
experiment. SS analyzed the data. SXC, DAC, NJC, KSK, TL, MO,
FAO, MJO, DW, DRW, and CS performed the replication experi-
ments. MB, DAC, TL, JF, and AD contributed reagents/materials/
analysis tools. SS, JF, and AD wrote the paper.
Funding. Funding for this work was provided by the Wellcome
Trust. Sagiv Shifman was supported by the European Molecular
Competing interests. The authors have declared that no competing
PLoS Genetics | www.plosgenetics.orgFebruary 2008 | Volume 4 | Issue 2 | e280006
Reelin and Schizophrenia
1. Cardno AG, Gottesman II (2000) Twin studies of schizophrenia: from bow-
and-arrow concordances to star wars Mx and functional genomics. Am J
Med Genet 97: 12–17.
2.McGue M, Gottesman II, Rao DC (1985) Resolving genetic models for the
transmission of schizophrenia. Genet Epidemiol 2: 99–110.
3. Riley B, Kendler KS (2006) Molecular genetic studies of schizophrenia. Eur J
Hum Genet 14: 669–680.
4.Leung A, Chue P (2000) Sex differences in schizophrenia, a review of the
literature. Acta Psychiatr Scand Suppl 401: 3–38.
5.Aleman A, Kahn RS, Selten JP (2003) Sex differences in the risk of
schizophrenia: evidence from meta-analysis. Arch Gen Psychiatry 60: 565–
6.McGrath J, Saha S, Welham J, El Saadi O, MacCauley C, et al. (2004) A
systematic review of the incidence of schizophrenia: the distribution of
rates and the influence of sex, urbanicity, migrant status and methodology.
BMC Med 2: 13.
7. Shifman S, Bronstein M, Sternfeld M, Pisante-Shalom A, Lev-Lehman E, et
al. (2002) A highly significant association between a COMT haplotype and
schizophrenia. Am J Hum Genet 71: 1296–1302.
8.Tan EC, Chong SA, Wang H, Chew-Ping Lim E, Teo YY (2005) Gender-
specific association of insertion/deletion polymorphisms in the nogo gene
and chronic schizophrenia. Brain Res Mol Brain Res 139: 212–216.
9. Chen X, Wang X, Hossain S, O’Neill FA, Walsh D, et al. (2007) Interleukin 3
and schizophrenia: the impact of sex and family history. Mol Psychiatry 12:
10. Chen Q, Wang X, O’Neill FA,, Walsh D, Fanous A, et al. (2007) Association
study of CSF2RB with schizophrenia in Irish family and case - control
samples. Mol Psychiatry. In press.
11. Chen X, Wang X, Hossain S, O’Neill FA, Walsh D, et al. (2006) Haplotypes
spanning SPEC2, PDZ-GEF2 and ACSL6 genes are associated with
schizophrenia. Hum Mol Genet 15: 3329–3342.
12. Patsopoulos NA, Tatsioni A, Ioannidis JP (2007) Claims of sex differences:
an empirical assessment in genetic associations. JAMA 298: 880–893.
13. Macgregor S, Visscher PM, Montgomery G (2006) Analysis of pooled DNA
samples on high density arrays without prior knowledge of differential
hybridization rates. Nucleic Acids Res 34: e55.
14. Sham P, Bader JS, Craig I, O’Donovan M, Owen M (2002) DNA Pooling: a
tool for large-scale association studies. Nat Rev Genet 3: 862–871.
15. Meaburn E, Butcher LM, Liu L, Fernandes C, Hansen V, et al. (2005)
Genotyping DNA pools on microarrays: tackling the QTL problem of large
samples and large numbers of SNPs. BMC Genomics 6: 52.
16. Butcher LM, Meaburn E, Liu L, Fernandes C, Hill L, et al. (2004)
Genotyping pooled DNA on microarrays: a systematic genome screen of
thousands of SNPs in large samples to detect QTLs for complex traits.
Behav Genet 34: 549–555.
17. Kirov G, Nikolov I, Georgieva L, Moskvina V, Owen MJ, et al. (2006) Pooled
DNA genotyping on Affymetrix SNP genotyping arrays. BMC Genomics 7:
18. Shifman S, Bhomra A, Smiley S, Wray NR, James MR, et al. (2007) A whole
genome association study of neuroticism using DNA pooling. Mol
Psychiatry. In press.
19. Shifman S, Levit A, Chen ML, Chen CH, Bronstein M, et al. (2006) A
complete genetic association scan of the 22q11 deletion region and
functional evidence reveal an association between DGCR2 and schizo-
phrenia. Hum Genet 120: 160–170.
20. McGinnis R, Shifman S, Darvasi A (2002) Power and efficiency of the TDT
and case-control design for association scans. Behav Genet 32: 135–144.
21. WTCCC (2007) Genome-wide association study of 14,000 cases of seven
common diseases and 3,000 shared controls. Nature 447: 661–678.
22. Salonen JT, Uimari P, Aalto JM, Pirskanen M, Kaikkonen J, et al. (2007)
Type 2 diabetes whole-genome association study in four populations: the
DiaGen consortium. Am J Hum Genet 81: 338–345.
23. Wedenoja J, Loukola A, Tuulio-Henriksson A, Paunio T, Ekelund J, et al.
(2007) Replication of linkage on chromosome 7q22 and association of the
regional Reelin gene with working memory in schizophrenia families. Mol
Psychiatry. In press.
24. Huang CH, Chen CH (2006) Absence of association of a polymorphic GGC
repeat at the 59 untranslated region of the reelin gene with schizophrenia.
Psychiatry Res 142: 89–92.
25. Goldberger C, Gourion D, Leroy S, Schurhoff F, Bourdel MC, et al. (2005)
Population-based and family-based association study of 59UTR poly-
morphism of the reelin gene and schizophrenia. Am J Med Genet B
Neuropsychiatr Genet 137: 51–55.
26. Akahane A, Kunugi H, Tanaka H, Nanko S (2002) Association analysis of
polymorphic CGG repeat in 59 UTR of the reelin and VLDLR genes with
schizophrenia. Schizophr Res 58: 37–41.
27. Chen ML, Chen SY, Huang CH, Chen CH (2002) Identification of a single
nucleotide polymorphism at the 59 promoter region of human reelin gene
and association study with schizophrenia. Mol Psychiatry 7: 447–448.
28. Quattrocchi CC, Wannenes F, Persico AM, Ciafre SA, D’Arcangelo G, et al.
(2002) Reelin is a serine protease of the extracellular matrix. J Biol Chem
29. Rice DS, Curran T (2001) Role of the reelin signaling pathway in central
nervous system development. Annu Rev Neurosci 24: 1005–1039.
30. D’Arcangelo G, Miao GG, Chen SC, Soares HD, Morgan JI, et al. (1995) A
protein related to extracellular matrix proteins deleted in the mouse
mutant reeler. Nature 374: 719–723.
31. Hong SE, Shugart YY, Huang DT, Shahwan SA, Grant PE, et al. (2000)
Autosomal recessive lissencephaly with cerebellar hypoplasia is associated
with human RELN mutations. Nat Genet 26: 93–96.
32. Eastwood SL, Harrison PJ (2003) Interstitial white matter neurons express
less reelin and are abnormally distributed in schizophrenia: towards an
integration of molecular and morphologic aspects of the neurodevelop-
mental hypothesis. Mol Psychiatry 8: 769, 821–831.
33. Hadj-Sahraoui N, Frederic F, Delhaye-Bouchaud N, Mariani J (1996)
Gender effect on Purkinje cell loss in the cerebellum of the heterozygous
reeler mouse. J Neurogenet 11: 45–58.
34. Absil P, Pinxten R, Balthazart J, Eens M (2003) Effects of testosterone on
Reelin expression in the brain of male European starlings. Cell Tissue Res
35. Allen NC, Bagade S, Tanzi R, Bertram L. The SchizophreniaGene Database.
Schizophrenia Research Forum. Available at: http://www.schizophreniaforum.
org/res/sczgene/default.asp. Accessed: 5 November 2007.
36. Pearson JV, Huentelman MJ, Halperin RF, Tembe WD, Melquist S, et al.
(2007) Identification of the genetic basis for complex disorders by use of
pooling-based genomewide single-nucleotide-polymorphism association
studies. Am J Hum Genet 80: 126–139.
37. Hanson RL, Craig DW, Millis MP, Yeatts KA, Kobes S, et al. (2007)
Identification of PVT1 as a candidate gene for end-stage renal disease in
type 2 diabetes using a pooling-based genome-wide single nucleotide
polymorphism association study. Diabetes 56: 975–983.
PLoS Genetics | www.plosgenetics.orgFebruary 2008 | Volume 4 | Issue 2 | e28 0007
Reelin and Schizophrenia