Oral Abstracts - WCPG 2014 - ISPG

ND World Congress of
Psychiatric Genetics
October 12-16, 2014
Copenhagen, Denmark
Pathways to Therapy and Prevention
Monday, October 13, 2014
11:00 AM - 12:00 PM
Concurrent Oral Sessions
Joanna Martin1, Michael C. O'Donovan1, Anita Thapar1, Kate Langley1, Nigel Williams1
MRC Centre for Neuropsychiatric Genetics and Genomics, Cardiff University
Background Attention deficit hyperactivity disorder (ADHD) is highly heritable. Genome-wide
molecular studies show an increased burden of large, rare copy number variants (CNVs) in children with
ADHD compared with controls. Recent polygenic score analyses have also shown that common variants
can be used in mass to differentiate ADHD cases from controls. The relationship between these common
and rare variants has yet to be explored.
Methods In this study, we tested whether children with ADHD with a large (>500kb), rare (<1%
frequency) CNV (N=60) differ from children with ADHD without such CNVs (N=421) by polygenic
risk scores for ADHD. We also compared ADHD polygenic scores in ADHD children with and without
CNVs to a group of population controls (N=4,670; of whom N=397 with CNVs).
Results The results show that children with ADHD with large, rare CNVs have lower polygenic scores
than children without such CNVs (OR=0.72, p=0.020). Although ADHD children without CNVs had
higher scores than controls (OR=1.19, p=0.0013), this difference was not observed for ADHD children
with CNVs (OR=0.86, p=0.26).
Discussion These results are consistent with a polygenic liability threshold model of ADHD with both
common and rare variants involved.
Christel M Middeldorp1, Maria M. Groen-Blokhuis1, Kees-Jan Kan1, Eveline L. de Zeeuw1, Abdel
Abdellaoui1, Catharina E.M. van Beijsterveldt1, Meike Bartels1, Eric A. Ehli2, Gareth E. Davies2, Paul A.
Scheet3, James J. Hudziak4, Jouke-Jan Hottenga1, Psychiatric Genomics Consortium Subgroup ADHD,
Benjamin M. Neale5, Dorret I. Boomsma1
VU University Amsterdam, Biological Psychology, 2Avera Institute for Human Genetics, 3University of
Texas, MD Anderson Cancer Center, 4University of Vermont, department of Psychiatry, 5Broad Institute
Background With polygenic score analyses we recently demonstrated that genetic risk factors
associated with an ADHD diagnosis predict continuous ADHD scores in the general population in
preschool and school-aged children. ADHD symptoms can persist in adolescence and adulthood and are
frequently co- morbid to aggression, anxiety and depression in childhood and adolescence. We present
the results of polygenic risk scores analyses for these phenotypes.
Methods In participants from the Netherlands Twin Register, polygenic risk scores were calculated
based on the latest results from the childhood ADHD mega-analysis performed in PGC. In a linear
mixed model, taking into account the relatedness between twins, the prediction of the polygenic risk
scores was tested for maternal ratings at age 3, 7, 10 and 12 for the Child Behavior Checklist Scales
(www.aseba.org) of aggression, anxious depression and withdrawn depressed. In adolescence, the
prediction was tested for the self-reports for these scales and attention problems. For each age, data were
available for between 1,000 and 2,000 individuals. We also analyzed the largest childhood and largest
adolescent dataset including 2263 and 3424 individuals respectively.
Results Significant prediction was found for the aggression and for attention problem scores throughout
childhood and adolescence, for maternal ratings as well as self-reports and for each age. There were no
significant predictions from ADHD polygenic scores towards anxious depression and withdrawn
depressed behavior, not even in the largest samples.
Discussion There is consistent overlap in genetic risk factors for a diagnosis of ADHD and continuous
ADHD and aggression scores in childhood. Moreover, this prediction extends into adolescence.
Although the phenotypic correlation between ADHD scores and anxious depression is as high as for
ADHD scores and aggression, no significant predictions were detected from genetic risk scores to a
diagnosis to anxious depression. Modeling of twin data and bivariate genome-wide association or GCTA
analyses need to shed light on the different mechanisms that underlie the phenotypic associations of
childhood ADHD and other behavioral and emotional problems.
Evie Stergiakouli1, Joanna Martin2, Marian Hamshere2, Anita Thapar2, David Evans3, Beate St.
Pourcain4, Nicholas Timpson3, George Davey Smith3
MRC Integrative Epidemiology Unit at the University of Bristol, 2Institute of Psychological Medicine
and Clinical Neurosciences, MRC Centre for Neuropsychiatric Genetics and Genomics, Cardiff
University School of Medicine, 3MRC Integrative Epidemiology Unit at the University of Bristol, 4MRC
Integrative Epidemiology Unit at the University of Bristol, School of Oral and Dental Sciences,
University of Bristol
Background Many psychiatric disorders can be viewed as extremes of dimensional attributes present in
the general population. Polygenic risk scores based on additive effects of common gene variants have
been shown to contribute to psychiatric disorders considered either categorically or as a continuum (Lee
et al. 2012). However, for behavioral traits in the general population genome-wide complex trait analysis
(GCTA) did not show a significant genetic influence despite twin heritability being substantial in the
same sample, suggesting that perhaps the genetic architecture of behavioral traits is different to disorder
(Trzaskowski et al. 2013). For ADHD at least, common genetic risk scores associated with
categorical diagnosis contribute to population trait variation (Martin et al. 2014). We performed a genomewide association study of ADHD symptoms in a general population sample and tested whether polygenic
risk scores for ADHD traits predict diagnostic status and the severity of disorder.
Methods Polygenic risk scores were calculated according to the method described by the International
Schizophrenia Consortium (ISC) (Purcell et al. 2009) for 508 children aged 4-18 years with a confirmed
research diagnosis of ADHD (Cardiff University) and 5,081 comparison subjects from the Wellcome
Trust Case Control Consortium (target sample). The QC procedures, ascertainment of the target sample
and GWAS results have been described in detail previously (Stergiakouli et al. 2012). The discovery
sample consisted of a genome-wide study of ADHD symptoms measured by the Development and WellBeing Assessment (DAWBA) completed by a parent in 5,690 ALSPAC children (mean age 7.7 years
(SD 0.14) (Boyd et al. 2013). SNPs with p<0.5 (after LD pruning) were selected from the discovery
sample to calculate polygenic scores weighted for beta coefficient in the target sample.
Results Polygenic risk scores calculated on 508 children with an ADHD diagnosis and 5,081 controls
were associated with the number of total ADHD symptoms (beta coefficient=0.29 (0.04-0.54), p=0.024)
within cases. ADHD polygenic scores could also distinguish cases from controls (OR=1.15 (1.05-1.26),
Discussion Our results suggest that the same genetic variants that are relevant for the number of ADHD
symptoms in a general population sample without clinical ADHD are also implicated in clinical ADHD
predicting both severity and ADHD diagnostic status. This provides evidence that common genetic
factors contribute to both behavioral traits in the general population and psychiatric disorder at least in
the case of ADHD. Future studies of other behavioral traits will show if this is the case for other
behavioral traits or it is limited only on ADHD symptoms.
Najaf Amin1, Cornelia van Duijn2
Erasmus University Medical Center, 2Department of Epidemiology, Erasmus MC, Rotterdam, the
Background Depression is a common psychiatric disorder with a lifetime prevalence estimated to be
14.6%. Its obscure etiology hinders effective treatment, which at present is based on trial and error and
hampered by the lack of biomarkers. Despite a strong genetic component (40-50%), there is no single
unequivocally identified common variant for major depressive disorder or related phenotypes. This has
raised the question whether relatively rare variants that segregate in families determine the disease in part.
These variants may be identified using exome-sequencing in families.
Methods We sequenced exomes of 1,300 individuals at an average depth of 74x and exome-array
genotyped another 1,500 individuals from a uniquely large Dutch family (Erasmus Rucphen Family;
ERF) that spans 23 generations. All individuals are assessed for depressive symptoms (Hospital anxiety
and depression scale (HADS) and Center for epidemiologic studies depression scale (CESD)) and 369
patients diagnosed with depression. We used several approaches including linkage, haplotyping, filtering,
genome-wide single-variant and gene-based association analyses to identify rare genetic variation that
confer large effects on depression/depressive symptoms in this family.
Results Genome-wide single-variant association analysis identified a significant, although intronic, rare
variant G>A (p-value=9.2*10-08; MAF=1.7%, effect=3.36) in the gene TMEM151A on chromosome
11q13 associated with the HADS scale. The variant also appeared to segregate in the family, connecting
all carriers (N=43) in six generations. Using linkage, haplotyping and filtering approach we identified a
missense variant T>C on chromosome 9p24 (p-value=9*10-04, MAF=1%, effect=2.47) in the gene RCL1
associated with the HADS scale. All carriers (N=34) connected within 6 generations. This variant is
highly conserved (phastcons = 1) and predicted to be damaging (polyphen=0.68). A rare T>G damaging
(polyPhen=1) missense variant in the gene BTNL9 on chromosome 5q35 was suggestively associated (pvalue=1.5*10-05, MAF=1%, effect=3.44) with the HADS depression scale. All carriers (N=35) were
connected to each other in four generations of which, 15 were treated for either major or mild depression.
Discussion Using exome-sequencing and various gene-mapping techniques in a large family from a
genetically isolated population, we have identified several rare genetic variants that segregate and confer
large effects on depression/depressive symptoms. While RCL1 is a novel candidate, TMEM151A lies in
the candidate region discovered in the major depression genome-wide association study by the psychiatric
genetics consortium. Further, TMEM151A is predominantly expressed in brain (primarily subthalamicnucleus: AUC = 1.00, p-value=3*10-09) and predicted to be involved in dopamine and serotonin release
cycle (p-value=2.7*10-11). BTNL9 is expressed in brain and fat tissues and known to be involved in
triglyceride homeostasis. All discovered variants are relatively rare in 1000 genomes and other
populations and usually not well-imputed thus limiting the scope for replication. Most variants are also
not present on the Illumina exome array. We, therefore, plan to perform functional analyses.
Sven Cichon1, Thomas W. MГјhleisen2, Andreas Forstner3, Markus Leber4, Thomas G. Schulze5, Jana
Strohmaier6, Franziska Degenhardt3, Stefan Herms1, Manuel Mattheisen7, Per Hoffmann1, Additional
Members of the MooDS Bipolar Disorder Working Group MooDS BD, Peter Propping8, Tim Becker4,
Marcella Rietschel6, Markus M. Nöthen3
University of Basel, 2Institute of Human Genetics, Life & Brain Center, University of Bonn, Germany;
Institute of Neuroscience and Medicine (INM-1), Research Center Juelich, 3Institute of Human Genetics,
Life & Brain Center, University of Bonn, Germany, 4DZNE Bonn, Germany; Institute of Medical
Biometry, Informatics and Epidemiology (IMBIE), University of Bonn, 5University of Göttingen,
Central Institute of Mental Health, Medical Faculty Mannheim, University of Heidelberg, 7Aarhus
University, 8Institute of Human Genetics, University of Bonn
Background Genome-wide association studies (GWAS) have identified the first common risk variants
for bipolar disorder (BD), in particular ANK3, CACNA1C, NCAN, ODZ4, ADCY2, MIR2113-POU3F2.
The majority of genetic variants influencing BD, however, remains unknown. These variants are expected
to have small effect-sizes and are difficult to detect individually at high statistical stringency by GWAS
using the currently available sample sizes. Pathway-based approaches have been developed, which use
prior biological knowledge on gene function to facilitate a more powerful analysis of GWAS data sets
and get more comprehensive insights into the biology of complex diseases. We employed this strategy in
a large GWAS data set of BD.
Methods For the pathway-based analysis, we used a sample of 9,747 patients with BD and 14,278
controls, comprised of a large cohort of European/Australian descent and the samples of the published
BD-GWAS by the Psychiatric Genomics Consortium. Analysis was performed with INRICH, a software
that tests if association signals in predefined gene sets (pathways) are enriched across independent gene
loci (non-overlapping intervals).
Results Test intervals were constructed in two steps. First, GWAS results on 2.3 million autosomal SNPs
were filtered for strong to moderate signals (P<5E-4), resulting in 5,312 SNPs. Mapping of these SNPs to
the largest gene isoform yielded a basic set of 386 genomic intervals. Secondly, overlapping intervals
were merged to avoid multi-counting of clustered genes. Finally, 359 intervals covering 496 genes were
tested for enrichment in 430 sets with 3,881 genes (Reactome pathway map). We found that a subset of
10 intervals, each covering a single gene, was significantly enriched in a set of 67 genes that form a
pathway for NCAM signaling (P=3.4E-5). The result withstood correction for the total number of sets
tested. Of note, among the 10 interval genes were the voltage-dependent calcium channel gene
CACNA1C, the sulfate proteoglycan gene NCAN, and the transcription factor gene CREB1.
Discussion The present study is an example for the increased power to detect potentially disease-relevant
biological processes by applying pathway-based approaches. The results were generated in the largest
GWAS data set of BD published to date. Our INRICH analysis provides correction-stable evidence that
genetic variation in the NCAM pathway is of likely relevance for the development of BD. Some of the
genes involved in this pathway (CACNA1C, NCAN) had previously been identified at genome-wide
significance in single-marker level GWAS. The INRICH analysis puts these two genes in a broader
biological context now by highlighting an enrichment of association signals in additional genes of that
particular pathway. The NCAM pathway plays an important role in defined cellular processes in the brain,
such as axonal growth and synaptic plasticity.
Wouter J. Peyrot1, Sang H. Lee2, Yuri Milaneschi1, Tonu Esko3, Douglas F. Levinson4, Nicholas G.
Martin5, Dorret I. Boomsma6, Naomi R. Wray2, Brenda WJH Penninx1, PGC-MDD Consortium
VU University Medical Center & GGZ inGeest, Amsterdam, 2The University of Queensland,
Queensland Brain Institute, 3Estonian Genome Center, University of Tartu, Estonia, 4Department of
Psychiatry and Behavioral Sciences, Stanford University, 5QIMR Berghofer Medical Research Institute,
Brisbane, QLD 6Department of Biological Psychology, VU University Amsterdam, Amsterdam, The
Background An association of lower educational attainment (EA) and increased risk for depression
(MDD) has been confirmed in various western countries. This negative phenotypic correlation can result
from multiple, not necessarily independent, effects; including causal, environmental or pleiotropic genetic
effects. This study aims to estimate the contribution of pleotropic genetic effects on the phenotypic
correlation between EA and MDD (genetic correlation).
Methods Data were analyzed from a total of 9,662 MDD cases (with a DSM-IV based diagnoses) and
14,949 controls (with no diagnosis of MDD in lifetime) from the Psychiatric Genomics Consortium with
additional Dutch and Estonian data. Information on EA (years of education) was available for 15,138 of
these individuals. The association of MDD and EA was assessed with logistic regression. With genomewide data on 884,105 common autosomal SNPs, four methods were applied to test for genetic pleiotropy:
(i) bivariate Genomic-Relationship-Matrix Restricted Maximum Likelihood (GREML), polygenic risk
score analyses with (ii) EA as discovery and MDD as target and with (iii) MDD as discovery and EA as
target, and (iv) SNP effect concordance analysis (SECA; Nyholt Bioinformatics 2014). The discovery
sample for (ii) consisted of independent meta-analyses results of Rietveld et al (Science 2013), and the
discovery sample for (iii) was constructed in the current sample with a ten-fold leave-one-out procedure.
Results On the phenotypic level, EA was associated to MDD as expected with an odds ratio of 0.85 per
standard deviation increase in years of education (95%CI 0.82-0.88). A similar association was found
when only cases with an age at onset older than 30 were taken into account. A weak negative genetic
correlation between MDD and EA was suggested by bivariate GREML analyses (i), but this correlation
was not significant. The polygenic risk score analyses showed no evidence for genetic correlation,
because the risk scores of discovery EA did not predict MDD (ii), and the risk scores of discovery MDD
did not predict EA (iii). The SECA analyses (iv) yielded no evidence for negative genetic correlation.
Discussion An association between low EA and MDD was confirmed within this sample from western
countries comprising approximately 25,000 individuals. No consistent evidence was found for pleotropic
genetic effects of genome-wide common SNPs. Hypothetically, pleiotropic genetic effects could exist
amongst genetic variation not taken into account in this study, but more plausible is that environmental
effects are causing the phenotypic correlation between EA and MDD.
Laramie Duncan1, Caroline Nievergelt2, Stephan Ripke1, Jackie Goldstein1, Lynn Almli3, Laura
Bierut4, Louis Fox4, Joel Gerlernter5, Guia Guffanti6, Israel Liberzon7, Mark Logue8, Adam Maihoffer2,
Monica Uddin9, Mark Daly1, Kerry Ressler3, Karestan Koenen6
Broad Institute of MIT and Harvard, 2University of California San Diego, 3Emory University, 4
Washington University, 5Yale University, 6Columbia University, 7University of Michigan, 8Boston
University, 9Wayne State University School of Medicine
Background Post-traumatic stress disorder (PTSD) is a common psychiatric disorder with substantial
unmet treatment need. Though neural circuitry is thought to be well understood in PTSD, the specific
genetic variants contributing to the moderate heritability of PTSD (estimates 30%-72%) are largely or
completely unknown, with few and somewhat inconsistent loci reported to date. To address the
challenge of underpowered individual studies, the PGC-PTSD group was formed, with over 20 groups
contributing data and analysis complete for 19,029 samples.
Methods Standard GWAS quality control procedures were performed on each dataset individually,
accounting for substantial diversity of genetic ancestry by identifying African American, European
American, and other mixed ancestry subsets of each dataset. GWAS was first conducted within each
ancestral group, and then fixed effect, inverse-variance weighted meta-analysis was conducted across
Results One novel locus reached genome-wide significance in the overall meta-analysis (3.76e-8). This
SNP is genic and has the same direction of effect in the current PGC schizophrenia mega-analysis
(p=0.01). The possibility of shared risk loci with schizophrenia is also indicated by an excess of samedirection, nominally associated (p<.05) loci observed when examining 111 SNPs robustly associated with
schizophrenia in these PTSD meta-analytic results (15 same-direction, nominally associated loci observed
and only 6 expected).
Discussion In this collaborative study with sample size larger than any PTSD GWAS published to date,
we identified a novel genic locus associated with PTSD across African American and European
American samples. Preliminary evidence also suggests shared risk loci with schizophrenia. Notably, the
diversity of ancestry in these PTSD samples is substantially greater than all other PGC datasets, with 52%
African Americans and only 37% European Americans. This represents an important resource for the
PGC community as we seek to ensure that results from large-scale genomic studies are equally applicable
across diverse populations.
Michael Benros1, Betina Trabjerg2, Sandra Meier2, Preben Mortensen2, Merete Nordentoft3, Esben
Mental Health Centre Copenhagen, Copenhagen University, 2National Centre for Register-based
Research, Aarhus University, 3Mental Health Centre Copenhagen, University of Copenhagen
Background Several studies have suggested an important role of infections and immune responses in the
etiology of schizophrenia. However, the causal pathway underlying the enhanced risk for schizophrenia in
individuals with infections is still unknown. Genetic studies of individuals with schizophrenia have shown
associations with genes involved in immune processes, suggesting a shared genetic liability towards
infections and schizophrenia. We therefore investigated the effect of the polygenic risk scores for
schizophrenia on the association between infections and the risk of schizophrenia.
Methods We made use of a nested case-control design and analysed a Danish population-based sample
comprising of 823 cases with schizophrenia and 832 controls matched on sex, age and birth year. The
(post-imputed) genomic data was obtained from the Psychiatric Genomics Consortium (PGC) after
samples have been processed from the Danish Neonatal Screening Biobank. Polygenic risk scores based
on the local cases-control sample were calculated using discovery effect size estimates weights from the
latest PGC-GWAS mega-analysis for schizophrenia (excluding the Danish replication sample). All
individuals were linked with nationwide population-based registers with virtually complete registration of
all hospital contacts for infections. Out of the 823 individuals diagnosed with schizophrenia, a total of
332 individuals had a hospital contact with infection prior to the schizophrenia diagnosis (40%).
Results After mutual adjustments for family history of infections and psychiatric disorders, a prior
hospital contact with infection was associated with an increased relative risk of schizophrenia by 1.47
(95%CI=1.16-1.84). Adding the polygenic risk score, which was robustly associated with schizophrenia
in this sample (RR=1.18; 95%CI=1.13-1.23), did not significantly alter the observed association of
hospital contacts with infections and increased risk of schizophrenia (RR=1.53; 95%CI=1.20-1.94). No
significant interaction between the polygenic risk score and infections were observed on the risk of
developing schizophrenia (p=0.938). Neither did we observe any significant effect of the polygenic risk
score on the risk of acquiring infections prior to being diagnosed with schizophrenia in analysis of cases
only (RR=1.01; 95%CI=0.96-1.07). After mutual adjustments for the above variables, a maternal
history of hospitalization for infection increased the risk of schizophrenia (RR=1.34; 95%CI=1.0
Discussion The polygenic risk score and a history of infections have strong independent effects on the
schizophrenia risk. Although we adjusted for an important source of common genetic risk using the
polygenic score, the effect of infections on the risk of schizophrenia remained. The common genetic risk
measured by the polygenic risk score seems not to account for the association with infection. However
the polygenic risk is lacking information on variation in the MHC region and rare variants, which might
have affected the results. Results will additionally be presented from updated analysis on a larger dataset,
including the associations with immune related genes, dose-response associations between the number of
infections, time since the last infection, the severity of the infection, type and localization of the infection
and associations with the polygenic risk score.
H. Simon Xi2, Eric B. Fauman2, Shaoxian Sun3, Sara A. Paciga3, Schizophrenia Working Group, Patricio
O'Donnell4, Jens R. Wendland3
Pfizer, 2Computational Sciences CoE, Pfizer Worldwide Research and Development,
PharmaTherapeutics Clinical Research, Pfizer Worldwide Research and Development, 4Neuroscience
Research Unit, Pfizer Worldwide Research and Development
Background Human genetics is a rational starting point for evidence-based drug discovery. Due to a
paucity of robust genetic findings for brain disorders such as schizophrenia, this approach has found
little application in CNS drug discovery to date. However, recent large-scale analyses have begun to
identify a number of robust genetic loci for schizophrenia and now pose a fundamentally new
opportunity and challenge to derive truly novel drug targets. In this work, we summarize current
strategies for applying human genetics and related �omics data to drug discovery and outline the path
from genetic locus to testable therapeutic mechanism for schizophrenia.
Methods Starting from a list of genome-wide significant GWAS loci identified by the schizophrenia
working group of the Psychiatric Genomics Consortium (PGC), we identified putatively causal genes in
LD-independent loci using linkage disequilibrium and/or distance mapping. Once identified, we
contextualize each causal gene with additional annotation on function and pathway, tissue expression,
pharmacological, and literature knowledge to a) assess the potential causal relationship to schizophrenia,
directionality of effect, cellular context, and potential safety liability, and b) identify a putative
hypothesis for therapeutic intervention. These putative hypotheses were then further prioritized based on
confidence in disease mechanisms, chemical doability, and availability of reagents and tools.
Results Starting from a list of 125 LD-independent loci, we were able to map 107 putatively causal
genes (85.6%). After gene triage and prioritization as outlined above, we were able to select 10-12 high
priority genes within well-defined biological pathways relevant to schizophrenia, including calcium
signaling and homeostasis, synaptic transmission, solute carrier transporters and inflammatory processes.
Several of the identified high priority targets are currently being followed up in exploratory studies to
address key gaps in validating them individually as a truly novel target for schizophrenia.
Discussion Neuropsychiatric disorders, such as schizophrenia, remain one of the defining biomedical
challenges in the 21st century, but are poorly served with new therapeutics. While recent advances in
human genetics, from GWAS studies to rare variants, hold great promise for defining the pathogenesis of
these brain disorders, the path from genetics to new medicines is far from clear. In this work, we have
identified an efficient strategy to generate testable hypotheses to facilitate the translation from human
genetics into new therapeutics for schizophrenia. This strategy is generalizable and applicable to not only
other brain disorders, but furthermore other therapeutic areas where rich genetic substrate is or will be
Younes Mokrab1, James Scherschel1, Lewis Vidler1, Cara Ruble1, Brian Eastwood1, Suzanne
Brewerton1, David Collier1, Schizophrenia Working Group Psychiatric Genomics Consortium
Eli Lilly
Background Genome-wide association studies (GWAS) have been successful in identifying common
variants associated with complex disorders such as schizophrenia, type 2 diabetes, and Alzheimer’s
disease. However, most of the associated variants are non-coding (intra- or intergenic), and each index
variant (i.e. showing the strongest statistical significance) is in linkage disequilibrium (LD) with other
variants in the same locus, often spanning multiple genes. This makes it difficult to identify the variants
that are likely to have a causal biological effect on disease, thus hampering the identification of the
causative gene(s) within most loci. In the present study, we used summary data from the Schizophrenia
Working Group of the Psychiatric Genomics Consortium (PGC2-SCZ) study which recently identified
108 genome-wide significant loci associated with schizophrenia (Ripke et al. 2014, submitted), to
develop a pipeline for functional annotation of associated variants for further biological analysis.
Methods GWAS meta-analysis was performed as described in the PGC2-SCZ study. Briefly, the dataset
contained 51 ancestry matched non-overlapping case-control samples (48 European, 3 East Asian,
36,989 cases and 113,075 controls). In each sample, association was tested using imputed marker
dosages and principal components to control for population stratification. Results were combined using
an inverse- weighted fixed effects model. Next, the tested genetic variants were processed in two stages:
First, QC was performed involving filtering by imputation INFO score ≥ 0.6, MAF ≥ 0.01, and
successfully imputation in ≥ 20 samples producing 9.5 million variants. Second, variants were processed
using an internal pipeline based on Variant Effect Predictor (VEP) (ensembl.org) to annotate the variants
and their associated genes together with publically available data on gene function, regulation,
conservation, expression, disease associations, pathways, known ligands, drug ability and protein
Results LD-independent index variants were defined as those with low LD (r2 < 0.1) to a more
significantly associated variant within a 500 kb window. 128 index variants were found to surpass
genome-wide significance (P ≤ 5x10-8) and are in turn correlated with 3,801 significant variants at r2
<0.6. Chromosomal loci were defined as the physical region containing each index variant and the
associated variants at r2 > 0.6. Associated loci within 250 kb of each other were merged. Thus 108
distinct loci were found, 84 of which have not been previously implicated in schizophrenia. Together
these loci contain 588 genes. 75% of the loci include protein-coding genes (40% a single gene) and a
further 8% are within 20 kb of a gene. In principle, any of the index or close LD friends can constitute
the causal variants affecting specific gene function. Therefore, systematic analysis for all the
significant 3,929 variants was performed using VEP in order to assess the potential impact on gene
Discussion Analysis of VEP results showed that 304 genes are found within 5kb distance from any given
significant variant. These comprise 268 genes in which variants fall outside gene boundaries and 36 genes
for which variants fall inside a gene (exons, introns, 5’-UTR or 3’-UTR). Of the latter a number of genes
were affected by missense variants including SLC39A8, APOPT1, ITIH3 and FES. Currently, further
analysis is being performed to categorize the variants into functional classes and add further regulatory
data to help build specific hypotheses about gene function alterations as disease aetiology in
Xiangning Chen1, Jingchun Chen1, FTND Meta-analysis Consortium
Virginia Commonwealth University
Background The Fagerstrom test for nicotine dependence (FTND) is commonly used in the study of
smoking addiction and nicotine dependence. One of the items, the time to smoke first cigarette in the
morning, or TFC, could be considered a nicotine withdrawal measure. We used these phenotypes to
identify risk genes for nicotine dependence. Those genes identified by TFC analyses may be used for
translational studies using animal models where withdrawal measures are robust.
Methods We organized a consortium and conducted genome-wide association analyses of FTND sum
scores (0-10) and TFC (0-3) phenotypes. The consortium yielded 15 independent datasets with more than
19,000 subjects of European ancestry. Each site conducted genotype imputation and association analyses
separately, and meta-analyses were used to combine data from individual datasets.
Results We found that the CHRNA5-A3-B4 gene cluster on 15q25 was associated with FTND scores
(minimal p at rs14714468, 6.9x10-18). 4 other loci (rs76000782 located between TRIM42 and
SCL25A36, p 1.7x10-8; rs148155309 in PIK3AP1, p 3.0x10-8; rs117029742, p 4.7x10-9; and
rs78824641 in HS3ST4, p 1.1x10-8) reached genome-wide significance. The analyses of the TFC
phenotype identified the CHRNA5-A3-B4 locus (minimal p at rs17487223, 1.1x10-9) and 3 other loci
(rs184042824, p 4.4x10-8, rs117029742, p 1.0x10-8 and rs10133756, p 3.9x10-9). Other candidates
identified by the analyses of both FTND and TFC included CHRNB3 (rs57308096, FTND p 4.8x10-5
and TFC p 4.1x10-7), KIF2B (rs2877510, FTND p 2.5x10-5, TFC p 1.67x10-5), CDH12 (rs4266369,
FTND p 9.6x10-6, TFC p 6.6x10-6), ZNF804A (rs80078811, FTND p 2.2x10-5, TFC p 3.1x10-5), and
MFSD2A (rs34022242, FTND p 2.2x10-5, TFC p 1.3x10-6).
Discussion In addition to the CHRNA5-A3-B4 locus, other significant loci discovered by FTND and
TFC are mutually supportive. The same marker identified in PI3KAP1 by FTND analyses,
rs148155309, is close to genome-wide significance (p = 6.0x10-8) in TFC analyses. rs117029742 is
significant in both FTND and TFC analyses. rs10133756, identified by TFC analyses, has a p value of
2.6x10-6 in FTND analyses. And rs78824641 identified by FTND has a p value of 3.6x10-5 in TFC
analyses. PIK3AP1 encodes a protein involved in the Toll like receptor signaling pathway that plays an
important function in inflammatory responses and other immune functions. There is no known gene near
rs117029742, but an EST, BG182718, with no known function, is nearby. Other top candidate genes
supported by both FTND and TFC analyses have been reported to be associated with smoking behaviors
(CHRNB3), obesity (KIF2B and CDH12), psychiatric disorders (CDH12 and ZNF804A) and cancers
(CDH12, HS3ST4, and MFSD2A).
Sven Stringer1, Camelia Minica2, Karin J.H. Verweij3, Hamdi Mbarek2, International Cannabis
Consortium International Cannabis Consortium, Eske M. Derks1, Nathan A. Gillespie4,
Jacqueline M. Vink2
Academic Medical Center Amsterdam, 2Department of Biological Psychology / Netherlands Twin
Register, VU University Amsterdam, The Netherlands, 3Department of Developmental Psychology and
EMGO Institute for Health and Care Research, VU University, Amsterdam, the Netherlands, 4Virginia
Commonwealth University
Background Cannabis is the most widely produced and consumed illicit drug worldwide. Previous
research has demonstrated the adverse effects of cannabis use. Cannabis use may lead to abuse or
dependence; subsequently causing physical, psychological and social problems. The International
Cannabis Consortium (ICC) was created to combine results of multiple GWA studies in order to identify
genetic variants underlying individual differences in cannabis use phenotypes.
Methods We performed a meta-analysis of 27,788 GWA samples from 12 samples collected in Europe,
the US and Australia. Lifetime cannabis use (i.e., never/ever used cannabis) ranged from 1.26% to 91.6%
with a median of 46.0%. Participating groups performed their own quality control. Imputation and GWA
analysis were performed according to a standardized protocol. All imputations were based on the same
reference panel (1000 genomes phase 1 European (EUR)). All GWA analyses were based on dosage data
and corrected for age, sex, and birth cohort effects and population stratification by controlling for ancestry
PCs. Additionally, we performed gene-based tests of association.
Results Although the QQ-plot clearly indicated enrichment of nominally significant findings, no
genome- wide significant hits were identified. The statistically most significant marker was located on
chromosome 12 (12:30479358) with a p-value=8.6*10-8. This polymorphism is located in an intergenic
region about 30kb from transmembrane and tetratricopeptide repeat containing 1 (TMTC1) and 30kb
from Importin 8 (IPO8). Among the 23,523 genes tested, none reached genome-wide significance
following FDR correction. The lowest p-value from the gene-based test was found for GammaAminobutyric Acid (GABA) A Receptor, Rho 3 (GABRR3) (p=9.46*10-5). We have access to two
independent replication samples, including another ~3500 subjects in which we aim to test for replication
of the top 10 SNPs.
Discussion We present preliminary results of the world’s largest meta-analysis of cannabis use to date.
The QQ-plot of this meta-analysis indicated significant enrichment of nominally significant findings. One
polymorphism nearly reached genome-wide significance and this may be an interesting candidate for
future replication studies. The SNP of interest is located near the gene TMTC1, which has previously
been found to be associated with weight-related phenotypes. The reward system in the brain plays a role
both in eating behaviors and substance use and this gene is therefore an interesting candidate for future
studies. In this regard, it is interesting that the top-result in the gene-based tests was found for GABBR3.
Previous research has suggested that GABA plays a role in addictive behaviors through its involvement
in the reward pathway and even though no significant association was detected, the role of this gene in
cannabis use should be further investigated.
Pamela Sklar1, CommonMind Consortium
Icahn School of Medicine at Mount Sinai
Background Advances in human genetics are reshaping the way we understand schizophrenia (SCZ).
We know infinitely more about disorder-associated DNA changes, specifically, that there are many
variants, rare and common, that are contributors. Using information from gene expression microarrays
and protein interactions databases we have basic outlines of genesets enriched for importance, but none of
this information has led to the identification of specific targets for drug development. The CommonMind
Consortium (CMC, http://commonmind.org) is a public-private pre-competitive consortium that brings
together disease area expertise, large and well-curated brain sample collections, and data management
and analysis expertise. CMC is generating and analyzing large scale data from human subjects with
neuropsychiatric and neurodevelopmental disorders. The consortium consists of five academic groups,
two pharmaceutical companies, and one non-profit group.
Methods RNA sequencing was performed in 554 human post-mortem samples (265 schizophrenia
samples and 289 controls) from the DLPFC (BA9, BA46) as part of the CommonMind Consortium
efforts. Ribozero libraries were constructed to enable detection of non-coding RNAs. Genotype data
from Illumina human core and exome were available on all samples. Covariates were controlled using
surrogate variable analysis. Differential expression analysis was performed using linear models
implemented in a voom/limma analysis pipeline. Gene coexpression networks were constructed using
WGCNA and high-density eQTL analyses were conducted. A variety of publicly available CRE
annotations for promoters, enhancers or open chromatin (DNase hypersensitivity regions) were used.
Furthermore, we used in-house generated CRE (promoter) annotations for neuronal cells sorted from the
DLPFC of controls and cases with schizophrenia. Common and rare variant data from multiple GWAS
and exome sequencing were also used.
Results Differential expression was detected for 15.6% of transcripts in the DLPFC at an FDR of 5%.
Differentially expressed genes were enriched for several categories of DNA variants implicated in risk
for SCZ including rare nonsynonymous DNA mutations previously reported in a Swedish case-control
exome sequencing study and common variants associated with SCZ. WGCNA gene coexpression
analysis identified 37 modules of which 11 are dysregulated in SCZ at FDR 5%. Among those, 3
modules are upregulated (primarily related to neuronal function) and 8 are down regulated (primarily
related to neuronal function, synaptic function, glutamate transmission, PSD and mitochondria/energy
Discussion In this study, we applied a stepwise approach to identify a subset of putative causal SNPs and
genes and then examined their distribution in gene coexpression networks. Overall, the results support the
existence of convergent genetic abnormalities in schizophrenia that could potentially drive the disease
leading to molecular and cellular alterations.
Douglas Ruderfer1, Menachem Fromer1, Giulio Genevese2, Peter Holmans3, Patrick Sullivan4, Steven
McCarroll2, Christina Hultman5, Pamela Sklar1, Shaun Purcell1
MSSM, 2Broad, 3Cardiff, 4UNC, 5Karolinska
Background Large, rare copy number variants (CNV) are associated to schizophrenia (SCZ) risk, in
terms of genome-wide burden as well as at specific loci (e.g. 22q11.2). We have previously documented
the role of rare CNVs in a large Swedish sample, based on single nucleotide polymorphism microarray
data (Szatkiewicz et al., 2014). Here we use next generation exome-sequencing on over 10,000
individuals to study rare CNVs in the same sample, by analysis of sequence read-depth. Compared to
microarrays, CNV detection through sequencing likely has complementary properties, including, in some
cases, greater sensitivity for smaller deletions and duplications that impact single genes.
Methods The Swedish sample compromised 4,978 individuals with SCZ and 6,256 controls; deepcoverage exome-sequence and SNP microarray data were available on all individuals. We used a method
we previously developed (XHMM, Fromer et al., 2012) to detect and genotype CNVs from exomesequencing read-depth; PLINK was used to perform the primary QC, burden and pathway analyses.
Results Using XHMM we detected 6,426 high-confidence rare (<1%) deletions and 9,270 duplications
(~1.3 per person). Of prior microarray-based CNVs that spanned genes, ~80% (4,649/5,929) were also
detected by XHMM, consistent with previous reports of sensitivity. We were able to demonstrate a
significant enrichment of genic deletions in cases compared to controls, especially pronounced for large
(>500kb) deletions (p=1.4x10-5). Of particular note here, however, is the large number of smaller, genic
CNVs (N=3,379 CNV spanning no more than 3 targets, mean size 11kb), most of which were not in the
prior microarray dataset. In the set of CNVs new to this analysis we see significant enrichment in both
deletions (p=0.008) and duplications (p=0.001). We are currently in the process of further geneset and
network analyses of these novel sequence-based CNVs.
Discussion Our results demonstrate that CNVs can be robustly detected by exome-sequencing and we
replicate the previously reported burden of CNVs in schizophrenia cases. XHMM also detects a large
number of novel CNVs that typically impact single genes: although a proportion of these calls are likely
to be false positives, they could also provide a basis to implicate individual risk genes. The next step,
which we will also discuss here, is to test single genes and pathways for association combining all classes
of rare variation obtained from exome-sequencing, including point mutations, short insertion/deletions
and de novo mutations from family-based studies.
Giulio Genovese1, Shaun Purcell2, Jennifer Moran3, Menachem Fromer2, Kimberly Chambert3, Patrick
Sullivan4, Pamela Sklar2, Christina Hultman5, Steven McCarroll3
Broad Institute, 2Mount Sinai, 3Stanley Center, 4University of North Carolina at Chapel Hill, 5Karolinska
Background Because individuals affected with schizophrenia have substantially fewer offspring than
unaffected individuals do, variants of large effect may be rare in populations. While common variant
association studies based on tens of thousands of individuals have implicated many individual genes, rare
variant association studies are at an earlier stage.
Methods We sequenced the exomes of more than 11,000 individuals – 4,954 schizophrenia cases and
6,239 controls – doubling the size of our recent collaborative study (Purcell et al., Nature 506, 185–190
(2014)). We developed a framework for effectively correcting for ancestry effects. We applied burden
tests to find enrichments of mutations in cases. We also describe methods for ascertaining somatic
mutations in exome sequence data.
Results We first show that singleton (observed exactly once in the cohort of 11,200) loss of function
mutations in any gene are enriched in schizophrenia cases relative to controls (p=0.0009, OR=1.055). We
show a clear polygenic signal which is a) concentrated in brain-expressed genes, b) concentrated in ultrarare alleles, and c) more pronounced in disruptive loss of function and damaging missense variants. We
also show how results from population-based studies are converging with trio studies of de novo
mutation. Our results greatly expand the implication of genes whose mRNA transcripts are bound by
FMRP, which in our data show definitive enrichment for loss of function ultra-rare alleles in
schizophrenia cases (p=3.1e-7, OR=1.39).
Discussion Rare variation points to a highly polygenic architecture for schizophrenia. Though this is the
largest exome sequencing study in schizophrenia to date, more than doubling the size of our recent
collaborative study, definitive implication of specific genes will require still-larger sample sizes.
Robert Power1, Stacy Steinberg2, Gyda Bjornsdottir2, G. Bragi Walters2, Engilbert Sigurdsson3,
Augustine Kong2, Daniel Gudbjartsson2, Hreinn Stefansson2, Kari Stefansson2
Institute of Psychiatry, London, 2deCODE genetics, 3University of Iceland
Background Great thinkers of the past, from Aristotle to Shakespeare, have remarked that the same
unleashing of thoughts and imagination characterize the creative genius and insanity. Such views run the
risk of romanticizing psychiatric disorders that present misery to individuals and huge healthcare costs to
society. However, evidence in support of the notion that creativity and psychotic disorders truly overlap
could shed light on the nature of creativity and on psychiatric pathogenic processes that hitherto have
remained elusive. Several attempts have been made at quantifying the overlap between psychiatric
disorders and creativity by studying both patients and their close relatives, for the most part these studies
have documented the existence of the overlap though concerns exist over reporting and ascertainment of
cases within families defined as �creative’. Hence we sought to answer this question using genetic
markers to establish a more objective measure of overlapping heritability.
Methods To do so we made use of the Psychiatric Genomics Consortium’s available data on common
sequence variants that predispose to psychosis, and tested whether they are more commonly found in
creative Icelanders than in population controls and in several carefully selected and matched control
groups (N=92,647). Creative individuals were defined from member lists of organizations and unions in
various fields of arts, including writers, musicians, painters, dancers, and accomplished chess players.
Results We showed that polygenic risk scores for psychotic disorders are significantly greater in
unaffected creative individuals than in population controls (P = 2.6 x 10-9 and 2.7 x 10-7 for
schizophrenia and bipolar risk scores, respectively), but not for 20 diseases chosen as negative controls or
in random individuals matched to creative individuals on age, sex and ancestry. Further it appeared that
the polygenic risk score for bipolar disorder was associated with educational attainment while the
schizophrenia score was not, potentially suggesting different mechanisms by which they affect cognitive
traits such as creativity.
Discussion Our results suggest that creativity that is conferred on individuals by common variants in
their genomes that are also known to increase risk of psychiatric disorders. Hence, the ability to think
and feel differently, a prerequisite for being creative, may come with a compulsion to think and feel
differently, a prerequisite for psychiatric diagnoses. Determining what features set an individual down
one of these paths and not, be it environment or deleterious rare variants, is a crucial next step in
understanding the causes of psychiatric disorders.
Esben Agerbo1, Carsten Pedersen2, Manuel Mattheisen3, Ole Mors4, Sandra Meier5, Anna Kähler6,
Preben Bo Mortensen7
CIRRAU - Centre for Integrated Register-based Research, National Centre for Register-based Research,
Aarhus University, 2CIRRAU - Centre for Integrated Register-based Research, National Centre for
Register-based Research, Aarhus University, 3Department of Biomedicine and Centre for Integrative
Sequencing (iSEQ), Aarhus University, 8000 Aarhus C, Denmark. The Lundbeck Foundation Initiative
for Integrative Psychiatric Research, iPSYCH, Aarhus and Copenhagen, Denmark, 4Centre for
Psychiatric Research, Aarhus University Hospital, Psychiatric Hospital, Denmark, 5National Centre for
Register- Based Research, Aarhus University, Aarhus, Denmark, 6Department of Medical Epidemiology
and Biostatistics, Karolinska Institutet, Stockholm, Sweden, 7Lundbeck Foundation Initiative for
Integrative Psychiatric Research, iPSYCH; Aarhus University, Denmark
Background Genome-based profile risk scores and family history of severe psychiatric illnesses are
strongly associated with the risk of schizophrenia. Few studies, however, have had the opportunity to
evaluate the combined impact of a polygenic risk score and a family history of severe psychiatric illness
simultaneously on the risk of schizophrenia in a nationwide and population-based sample. The aim of this
study is 1) to assess the marginal impacts of polygenic risk scores and family history of severe
psychiatric illness on the risk of schizophrenia in a population-based sample, 2) to quantify the fraction of
subjects with schizophrenia that would not have occurred if the effects associated with the polygenic risk
score and the family history of severe psychiatric illness were absent, 3) to determine the proportion of
the risk associated with family history that is mediated through the polygenic risk score.
Methods We made use of a nested case-control design and analysed a Danish population-based sample
comprising of 866 cases with schizophrenia, 871 controls and their first-degree relatives. Information on
family history was extracted from the National Health Registers and the (post-imputed) genomic data
was obtained from the Psychiatric Genomics Consortium (PGC) after samples have been processed from
the Danish Neonatal Screening Biobank. Risk scores based on the local cases-control sample were
calculated using discovery effect size estimates weights from the latest PGC-GWAS mega-analysis for
schizophrenia (excluding the Danish replication sample). Family history was categorized to indicate
whether the subject's parents or siblings previously had been diagnosed with: schizophrenia or related
psychosis, bipolar affective disorder or any other psychiatric disorder.
Results The risk of schizophrenia was elevated in those with relatives with a schizophrenia-like
psychoses (OR:4.2 [2.6-6.8]), bipolar affective disorder(2.8 [1.9-4.3]) or 3) other psychiatric disorder(2.6
[2.0-3.4]). Based on 24755 markers (p-value threshold of 0.05) in the PGC discovery sample, there was a
dose-response relationship with the risk score and the risk of schizophrenia with an OR of 8.0 (4.5-14.1)
in the upper decile vs the lowest decile. The attributable risks associated with family history and the risk
score were 26% (23%-28%) and 52% (50%-53%). The interaction p-value was 0.03. To assess the part of
family history that was mediated through the polygenic risk score, we calculated the mediating proportion
under interaction, which suggested that 64% (26%-103%) of the effect of a family history with psychosis
among subjects with a family history of psychosis was mediated through the risk score while 24%
(14%-34%) was mediated in subjects without a family history of psychosis.
Discussion The polygenic risk score as well as a family history of severe psychiatric illness represent
strong valid measures to which a sizeable proportion of the excess schizophrenia risk can be attributed.
The effects of the polygenic risk score and family history are dependent. Furthermore, a particular large
proportion of the effect associated with a family history of psychosis is mediated through the polygenic
risk score for subjects with a family history of psychosis leaving some room open for the influence of
environmental factors.
Stephan Ripke1, PGC Schizophrenia Group
Massachusetts General Hospital
Background The PGC (Psychiatric Genomics Consortium) is an international group of researchers
whose major aim is to maximize the utility of extant psychiatric GWAS through mega-analysis. In a
previous study, our first wave of genome-wide schizophrenia association analysis identified multiple loci
involved in this genetically complex and clinically heterogeneous disorder (Nature Genetics, 2011). The
first results from our most recent endeavor were presented at World Congress of Psychiatric Genetics
Methods Here we present an update of the biological insights gained analyzing the results. This
international endeavor, which now comprises 35,476 schizophrenia cases and 46,839 controls coming
from 52 substudies. The presented data is imputed into 1000 Genomes (Aug, 2012) and analyzed using
standard logistic regression with ancestry components as covariates. All index SNPs with a p-value
smaller than 1x10-6 were used for replication lookup in an independent GWAS analysis with 1,500 cases
and 66,000 controls. There will be an estimated additional lookup of approximately 4,000 cases and
10,000 controls genotyped on PsychChip.
Results There are numerous follow up analysis being performed with the more than 100 reliably
associated regions from this newest round of meta-analysis. The loci implicated include prior findings
(MIR137, CACNA1C, ZNF804A) along with a host of new targets. Associations at DRD2 and multiple
genes involved in glutamatergic neurotransmission highlight molecules of known and potential
therapeutic relevance to schizophrenia. Additionally the hypothesized link between the immune system
and schizophrenia is supported by these associations.
Discussion These results are in line with prior predictions and developments in other complex disease
GWAS with sufficiently large samples like Crohn's disease. They continue to provide new insights into
the biology of schizophrenia.
Sarah Marzi1, Emma Meaburn2, Manuela Volta3, Matthew Davies4, Claire Troakes3, Simon Lovestone3,
Leonard Schalkwyk3, Jonathan Mill5
MRC SGDP Research Centre, Institute of Psychiatry, King’s College London, 2Department of
Psychological Sciences, Birkbeck, University of London, 3MRC SGDP Research Centre, Institute of
Psychiatry, King’s College London, 4Department of Twin Research and Genetic Epidemiology, St
Thomas’ Hospital, King’s College London, 5Exeter University Medical School
Background While most DNA methylation is thought to be symmetrical across both alleles throughout
the genome, there are some notable exceptions. Genomic imprinting and X chromosome inactivation are
two well-studied sources of allele-specific (or allelically-skewed) methylation (ASM), but recent research
has indicated a more complex pattern in which genotypic variation can be associated with DNA
methylation in cis.
Methods Given the known heterogeneity of methylation across tissues and cell types we explored inter
and intra-individual variation in ASM across multiple human brain tissues and whole blood from multiple
individuals. We used SNP microarrays to quantitatively assess ASM in amplicons covering ~8% of the
human genome following cleavage with a cocktail of methylation-sensitive restriction enzymes (MSREs).
Results Consistent with previous studies, we find widespread ASM with >4% of the ~220,000 loci
interrogated showing evidence of allelic skewing. A large proportion of ASM appears to be tissuespecific, with ~50% of ASM loci identified within an individual being specific to one tissue, with higher
levels observed in blood compared to brain. Interestingly, cross-tissue ASM is enriched in regions of the
genome associated with lincRNAs.
Discussion These findings contribute to our understanding about the nature of differential DNA
methylation across tissues and have important implications for genetic studies of psychiatric disease.
Chloe Chung Yi Wong1, Neelroop Parikshak2, Laura Lysenko3, Elham Assary3, Claire Troakes3, Joana
Viana4, Daniel Condliffe3, T. Grant Belgard2, Vivek Swarup2, Eilis Hannon4, Leonard Schalkwyk3,
Daniel Geschwind2, Jonathan Mill5
Institute of Psychiatry, King's College London, 2University of California, Los Angeles, 3King's College
London, 4Exeter University, 5King's College London, Exeter University
Background Autism Spectrum disorders (ASD) are a range of complex neurodevelopmental disorders
with heterogeneous aetiological origins. There is now emerging evidence to suggest that in addition to
genetic factors, environmental and epigenetic factors also play a significant role in the aetiology of ASD.
We had previously defined shared RNA co-expression changes in ASD post mortem brain, but aetiology
of many of these changes was undefined.
Methods With the aim to explore the contribution of epigenetic variation to ASD, we have profiled DNA
methylation in a unique and sizeable collection of post-mortem brain tissues (n=202) among three brain
regions including dorsolateral prefrontal cortex, primary auditory cortex and cerebellum. DNA
methylation at over 485,000 CpG sites was quantified using the Illumina Infinium HumanMethylation450
array. Quality control and data pre-processing was undertaken using the WateRmelon R package in
conjunction with an analysis pipeline developed by our group.
Results Our analyses reveal a number of significant ASD-assoicated DNA methylation differences CpG
sites, located in both novel genomic region as well as in the vicinity of several known ASD genes. In
addition to identifying region- specific differentially methylated sites, we have identified regions that are
consistently altered across the two cortical regions using a novel meta- analysis method.
Discussion This study, to our knowledge, represents the most comprehensive epigenomic analysis of
ASD using post-mortem tissues to date. Our epigenome-wide scan identifies several new candidate
genes for ASD and provide further evidence for a role of altered DNA methylation in ASD.
Nadine Provençal1, Tania Carrillo-Roa2, Torsten Klengel3, Christoph Anacker4, Carmine M Pariante5,
Kerry J. Ressler6, Elisabeth Binder7
Max Planck Institute for Psychiatry, 2Dept. Translational Research in Psychiatry Max Planck Institute
for Psychiatry, 3YRK: Behav Neuro & Psych Dis, Emory University, 4Dept. Neurosciences, Columbia
University, 5Psychological Medicine, King's College London, 6Dept. Psychiatry and Behavioral
Sciences, Emory University, 7Dept. Translational Research in Psychiatry, Max Planck Institute for
Psychiatry and Emory University
Background Excessive glucocorticoids (GC) release after early life stress exposure is thought to result
in a long-lasting disruption of the stress hormone system and ultimately to an increase risk for psychiatric
disorders later in life. Stress and GCs are known to regulate hippocampal neurogenesis and to induce
long- lasting changes in DNA methylation in specific loci such as the glucocorticoid receptor (NR3C1)
and FK506 binding protein 5 (FKBP5) in hippocampal and in peripheral blood cells DNAs. Here we aim
to expend these results to multiple loci using whole genomic comprehensive analysis of the epigenetic
effects of GC activation during hippocampal neurogenesis. Specifically, we aim to identify stable
epigenetic modifications induced by GC activation during early neurogenesis stages that are maintained
through cell maturation. Moreover, we hypothesised that part of these epigenetic alterations will be seen
in peripheral blood cells DNA of adults expose to severe child abuse.
Methods We used Illumina arrays to analyse gene expression, CpG methylation and
hydroxymethylation levels of immortalised human hippocampal progenitor cells (HPC) treated with
dexamethasone (Dex) or vehicle at different stages during neurogenesis. Cells were either treated in the
proliferation phase, after the proliferation and differentiation phases (acute), followed by 20 days of
washout after the acute treatment (sustained) or followed by 20 days of washout after a treatment in postdifferentiation (post- diff). We also quantify global changes in DNA methylation and
hydroxymethylation levels using ELISA. CpG methylation profiles from adults exposed or not to severe
child abuse (n=496) were analysed by Illumina 450K arrays. All the analysis where done correcting for
estimated cells counts (neuron/glia or PBMC ratios).
Results Global methylation and hydroxymethylation analysis revealed a significant effect of Dex
treatment on hippocampal neurogenesis that could possibly be explained by differential expression of
DNMT3b and TET1. Overall, the methylation analysis revealed a large number of Dex-induced
differentially methylated CpG sites (Pvalue < 0.05 and FDR < 0.05) following the acute (7222 CpGs),
the sustained (6207 CpGs) and the post-differentiation (14383 CpGs) treatments but not after
Interestingly, 136 CpG sites showed differential methylation followed Dex treatment in the acute phase
and maintained this difference after 20 days of washout, identifying sustained effect of Dex in various
loci including FKBP5. Most of these CpGs sites were not affected by Dex treatment in postdifferentiation. In addition, over 11% of the differentially methylated CpGs in the HPCs following Dex
treatment were also differentially methylated in blood cells of adult exposed to child abuse.
Discussion Preliminary analysis provides evidence of clustered and genome-wide epigenetic effects of
GC activation during hippocampal neurogenesis where the timing of the exposure seems to be critical to
induce long-lasting changes.
Jonathan Mill1, Katie Lunnon1, Rebecca Smith2, Eilis Hannon1, Claire Troakes2, Joe Burrage1, Ruby
Macdonald1, Pavel Katsel3, Vahram Haroutunian3, Zachary Kaminsky4, Catharine Joachim5, John
Powell2, Simon Lovestone5, Leonard Schalkwyk2
University of Exeter, 2King's College London, 3The Icahn School of Medicine at Mount Sinai, 4Johns
Hopkins University School of Medicine, 5University of Oxford
Background Alzheimer’s disease (AD) is a chronic neurodegenerative disorder characterized by
progressive neuropathology and cognitive decline. Although the neuropathological manifestation of
AD is well characterized in post-mortem brain, little is known about the underlying risk factors or
mechanism(s) involved in disease progression. Increasing knowledge about the biology of the genome
implicates an important role for epigenetic variation in human health and disease, and recent
methodological advances mean that epigenome-wide association studies (EWAS) are now feasible for
complex disease phenotypes. We have undertaken the first systematic cross-tissue EWAS analysis of
DNA methylation in AD using a powerful sequential replication design, with the goal of identifying
disease-associated methylomic variation across pathologically-relevant regions of the brain.
Methods The first (�discovery’) stage of our analysis utilized multiple tissues from donors (N = 117)
archived in the MRC London Brainbank for Neurodegenerative Disease. From each donor, genomic
DNA was isolated from four brain regions (entorhinal cortex, superior temporal gyrus, frontal cortex and
cerebellum) and, where available, whole blood obtained pre-mortem. A cortical 'replication' dataset was
generated using DNA isolated from two regions (STG and PFC) obtained from a cohort of brains
archived in the Mount Sinai Alzheimer's Disease and Schizophrenia Brain Bank (N = 144). A third
replication sample was obtained from the Thomas Willis Oxford Brain Collection. DNA methylation was
quantified using the Illumina 450K HumanMethylation array with differentially methylated positions
(DMPs) confirmed using bisulfite-pyrosequencing.
Results We identify a highly significant AD-associated differentially methylated region (DMR) in the
ankrin 1 (ANK1) gene that is strongly associated with neuropathology in the entorhinal cortex, a
primary site of AD manifestation in the brain. This region was confirmed as significantly
hypermethylated in two other cortical regions (superior temporal gyrus and prefrontal cortex) but not in
the cerebellum, a region largely protected from neurodegeneration in AD, or whole blood obtained premortem, from the same individuals. These CpG sites were subsequently found to be significantly
hypermethylated in cortical samples from two independent brain cohorts, providing compelling evidence
for an association between cortex-specific ANK1 hypermethylation and AD-related neuropathology.
Discussion Our study represents the first EWAS of AD employing a sequential replication design
across multiple tissues, and highlights the power of this approach more broadly for the identification of
disease- associated DMRs.
Helen Spiers1, Eilis Hannon2, Leonard Schalkwyk1, Chloe Wong1, Rebecca Smith1, Nick Bray1, Jonathan
King's College London, 2University of Exeter, 3University of Exeter, King's College London
Background Brain development involves the alteration of cellular phenotype in response to
genetically pre-programmed, environmental and stochastic events. The emergence of epigenome-wide
profiling technologies has facilitated the study of epigenetic processes, such as DNA methylation at
CpG dinucleotides, in neurodevelopment. Here we report data from a genome-wide interrogation of
methylomic trajectories across human fetal brain development.
Methods Genome-wide DNA methylation was quantified in a unique cohort of human fetal brain
samples (n=100 male, n=79 female, age range = 23 to 184 days post-conception) using the Illumina
HumanMethylation450 BeadChip. This technology provides a quantitative measure of DNA methylation
level at 485,577 CpG sites, covering 99% of RefSeq genes with an average of 17 CpG sites per gene
Results After adjusting for sex, linear regression identified 28,718 autosomal age-associated
differentially methylated probes (aDMPs) passing genome-wide significance testing (P<1.25E-7),
16,190 of these showed age-associated hypomethylation, whereas 12,528 became hypermethylated.
Weighted gene co-methylation network analysis identified several modules of co-methylated loci
enriched for functional pathways associated with brain development. We also identified considerable sex
differences in DNA methylation during brain development; of note, 521 autosomal probes displayed
significant (P<1.22E-7) sex-specific levels of DNA methylation across brain development and sex-age
interactions were observed at 59 autosomal probes.
Discussion This study provides increased knowledge about the molecular mechanisms that regulate
dynamic gene expression across human brain development, and identifies pathways contributing to sexual
differentiation of the brain. This work has the potential to enhance understanding of the pathogenesis of
neurodevelopmental brain disorders, such as autism and schizophrenia, and may provide a route to
understanding the sexual dimorphism observed in such conditions.
Tempei Ikegame1, Miki Bundo1, Yui Murata1, Hiroko Sugawara1, Harumi Saida1, Fumiko Sunaga1, Jun
Ishigooka1, Tsukasa Sasaki1, Kenji Kondo1, Masashi Ikeda1, Nakao Iwata1, Tadafumi Kato2, Kiyoto
Kasai1, Kazuya Iwamoto1
The University of Tokyo, 2Laboratory for Molecular Dynamics of Mental Disorders, RIKEN Brain
Science Institute
Background To date, a large number of studies have focused on serotonin transporter (5-HTT) as a key
molecule to elucidate the mechanism of mental disorders because aberrant release and reuptake of
serotonin in the brain of patients with mood and anxiety disorders were reported. In a previous study, we
have shown the promoter hypermethylation of SLC6A4 gene, which encodes 5-HTT in monozygotic
twins discordant for bipolar disorder (BD). Furthermore, we have confirmed SLC6A4 promoter
hypermethylation in lymphoblastoid cell lines and postmortem brain tissues of patients with BD. Here we
examined DNA methylation level of SLC6A4 promoter using genomic DNA derived from a relatively
large-scale peripheral blood cell (PBC) samples of healthy controls and patients with BD. In addition, we
analyzed the relationship between DNA methylation level and the short (S) or long (L) allele of serotonin
transporter-linked polymorphic region (5-HTTLPR) of SLC6A4 promoter.
Methods Two micrograms of genomic DNA extracted from PBC samples of patients with BD (n = 449)
and age- and sex-matched healthy controls (n = 456) were treated with sodium bisulfite modification
using an EpiTect 96 Bisulfite Kit (QIAGEN). All subjects were from the Japanese population. A regionspecific PCR with a biotinylated primer was performed for SLC6A4 promoter. The examined CpG sites
were chosen based on the our previous epigenetic studies that reported DNA methylation alterations in
BD. DNA methylation level was measured with the PSQ 96MA instrument (QIAGEN) according to the
manufacturer’s protocol. The 5-HTTLPR was genotyped by using standard PCR and direct amplicon
Results We found a significantly higher methylation level at one of the two validated CpG sites in
patients with BD compared to controls (p < .001). Subsequent analysis revealed significant effect of sex
on DNA methylation level in controls. Data analysis considering sex difference revealed that
hypermethylation was prominent in male patients with BD compared to male controls (p < .01).
Significant hypermethylation in male patients was also observed even when they were classified into
bipolar I and II disorders (p < .05). Finally, subgroup analysis considering 5-HTTLPR revealed a
significant hypermethylation in male patients with BD harboring a particular L allele.
Discussion Hypermethylation of SLC6A4 promoter in patients with BD is consistent with most of the
previous studies on mood disorders and other psychiatric disorders. Hypermethylation in BD was
associated with sex and 5-HTTLPR, suggesting the complex interactions between genetic and
environmental factors contribute to the epigenetic change in patients with BD.
2:30 PM - 4:30 PM
Symposia Sessions
Chair: Stephen Faraone, SUNY Upstate Medical University
Overall Abstract Details This symposium presents new data about the genetics of ADHD. The first two
talks focus on variant discovery; the second two address the biological significance of implicated
variants. Anders Borglum will present a genome-wide association study of a Danish birth-cohort of
ADHD cases and controls based on the Danish Newborn Screening Biobank (DNSB). The Danish
Psychiatric Central Research Register has identified all children born since 1981 diagnosed with ADHD
(~18,000) and selected a large random population based control group (~28,000) that can be individually
matched to the cases. The samples of the identified individuals have been retrieved from the DNSB for
DNA extraction and genotyping on the PsychChip. The register provides access to a broad range of
phenotypic information (e.g. on sub-phenotypes and comorbidities) and environmental risk
factors/exposures, allowing for comprehensive genetic, environmental and GxE analyses. Dr. Borglum
will present the results from the first data freeze consisting of a large subset of the Danish iPSYCH
sample of ADHD cases and controls (in total >45,000 individuals). Ben Neale will present result from the
PsychChip analyses from the ADHD subgroup of the Psychiatric Genomics Consortium (PGC). The
PsychChip incorporates exome chip content, a common variant GWAS backbone for imputation, and a
set of ~50,000 markers selected based on previous evidence in psychiatry. Using the PsychChip, the
current number of ADHD cases with genome-wide association data increases from the approximately
5,000 samples currently to ~15,000. Dr. Neale will also integrate analysis of 23&Me self-report of
ADHD into the meta-analysis to evaluate the effectiveness of their self-report phenotyping. Beth Wilmot
will present the results of a pathway analysis of ADHD that uses individual genotype data from multiple
variants to model a pathway-level effect, thus allowing different underlying models of disease risk to be
tested. This approach is in contrast to those that test for enrichment of individual SNP-level effects. She
will discuss the differences in performance across these methods and their impact on the resulting
determination of significant pathways in relation to the ADHD cohorts within the PGC. Jan Haavik will
review some approaches to investigating the biological functions of ADHD variants, particularly
missense variants in genes found by GWAS and sequencing. This work focusses on regulatory protein
complexes involved in serotonin and catecholamine signaling, including the family of 14-3-3 proteins
that are highly expressed in the nervous system and interact with hundreds of target proteins, implicating
them in a range of different cellular pathways. The functional analysis pipeline includes biochemical
studies, in silico modeling and molecular systems biology approaches. Many computational tools have
been proposed to predict the effects of SNPs and prioritize variants for further studies. While homology
Anders BГёrglum1
Aarhus University
Individual Abstract Attention deficit hyperactivity disorder (ADHD) is a common childhood behavioral
disorder affecting 3-6% of school-age children around the world. The disorder is highly heritable and
several moderately sized genome-wide association studies have been performed, the largest by the
Psychiatrics Genomics Consortium including 2,064 trios, 896 cases, and 2,455 controls. None of these
studies have identified single SNPs reaching genome-wide significance. However, the SNP heritability of
ADHD has been estimated to 0.28, which is similar to what is reported for schizophrenia (0.23),
indicating that common SNPs contribute substantially to ADHD susceptibility and that increasing GWAS
sample sizes is likely to produce significant results. In Denmark, nationwide screening of new-borns for
phenylketonuria (PKU) and other metabolic diseases has been carried out since 1975 and since 1981
surplus of the samples have been stored in the Danish Newborn Screening Biobank (DNSB). DNSB
presently contains dry blood spot samples from more than 2 million individuals, and it continuously
increases with annual screening of around 65,000 new-borns. Thus, the Biobank constitutes a unique
resource of biological material from nationwide birth-cohorts. Moreover, using the unique personal
identification number assigned to all live-born children it is possible to crosslink the DNSB samples to
the comprehensive Danish register system containing detailed longitudinal information on multiple health
and social outcomes. As part of the Danish iPSYCH (Lundbeck Foundation Initiative for Integrative
Psychiatric Research) program and in collaboration with the Broad Institute and the Psychiatric Genomics
Consortium, we have initiated a large-scale study of ADHD based on the samples available in the DNSB.
Coupling the samples with information from the Danish Psychiatric Central Research Register we have
identified all children born in Denmark since 1981 that have been diagnosed with ADHD (?18,000) and
selected a large random population based control group (?28,000) that can be individually matched to the
cases. The samples of the identified individuals have been retrieved from the DNSB for DNA extraction
and genotyping on the PsychChip. Through the register system we have access to a broad range of
phenotypic information (e.g. on sub-phenotypes and comorbidities) and environmental risk
factors/exposures, allowing for comprehensive genetic, environmental and GxE analyses. Here we will
present the results from the first data freeze consisting of a large subset of the Danish iPSYCH sample of
ADHD cases and controls (in total >45,000 individuals).
Benjamin Neale1, PGC-ADHD Group
Massachusetts General Hospital
Individual Abstract Genome-wide association studies for psychiatric illnesses are beginning to yield
robust replicable results, as led by the current efforts in schizophrenia. For Attention
Deficit/Hyperactivity Disorder (ADHD), such associations have yet to be found. However, GCTA
heritability analysis and polygenic scoring results suggest that the search for common variants may not
be in vain. As part of the current Psychiatric Genomics Consortium efforts, we have designed the
PsychChip, a custom genotyping array that incorporates the exome chip content, a common variant
GWAS backbone for imputation, and a set of ~50,000 markers selected based on previous genome-wide
association evidence in psychiatry or on emerging rare variant analysis. Using the PsychChip, we are
expanding the current number of cases with genome-wide association data from the approximately 5,000
samples currently to ~15,000. We will also integrate analysis of 23&Me self-report of ADHD into the
meta-analysis to evaluate the effectiveness of such self-report phenotyping. Primary association results
from the meta-analysis of ADHD will be presented, as will polygenic prediction from the self-report
phenotypes into the clinically ascertained samples. We will also evaluate the extent to which rare
standing coding variation, as assayed by the exome chip content, will inform on risk to ADHD in the
population. Genome-wide association analysis of ADHD holds the potential to reveal robust replicable
results that can inform the biological basis of disease as well as provide insight into the nature of overlap
between ADHD and other psychiatric and developmental phenotypes. Large sample sizes are an
essential component of this landscape and these results will shed further light on how many cases for
ADHD will be necessary to gain a more comprehensive view on the genetic basis of this disease.
Beth Wilmot1, Michael Mooney1, Shannon McWeeney1, Joel Nigg1
Oregon Health & Science University
Individual Abstract Pathway representation can provide a biological context for the interpretation of
genomic variants associated with complex disease. Within Pathway analyses, multiple approaches for
calculating pathway-level association measures exist. A critical aspect of some newer methods is to
utilize individual genotype data from multiple variants to model a pathway-level effect, thus allowing
different underlying models of disease risk to be tested. These methods are in contrast to those that test
for enrichment of individual SNP-level effects. The interpretation of results from pathway analyses is
dependent on the underlying assumptions of the pathway methods and the parameters used for each
algorithm. Factors influencing the analysis include choices for SNP to gene assignment, pathway
definition, summarization across genes and pathways and the model assumptions of the statistical
methods. The challenge in finding the most informative model is how to adequately capture the
complexity of the underlying genomic architecture when calculating pathway-level effects. The widely
used enrichment-type methods are dependent on individual SNPs having independent main effects,
because they first calculate individual SNP association measures and then combine these individual
effects to calculate a pathway-level association measure. Our hypothesis is that methods utilizing multi-
variant models to calculate gene- or pathway-level effects better represent the underlying genetic
complexity. Such methods are underutilized in psychiatric genetics to date. Therefore, we investigated
approaches that utilize the original genotype data as input, rather than individual SNP p-values, to test for
association between pathways and a trait of interest. We will discuss the differences in performance
across these methods and their impact on the resulting determination of significant pathways in relation to
the ADHD cohorts within the Psychiatric Genetics Consortium.
Jan Haavik1
University of Bergen
Individual Abstract Genome wide association studies and DNA sequencing are gradually revealing
genetic markers that are associated with psychiatric disorders. Reported relative risks have been small
and the strongest associations have been reported for large copy number variants spanning many genes
and intergenic or intronic SNPs variants. As each variant contributes small effects across many different
phenotypes, it is unclear how they relate a particular symptom or diagnosis, e.g. if there are any “true”
schizophrenia or ADHD genes. It is often unclear which genes or biological systems that are affected, if
the variants represent gain or loss of function, altered temporal or spatial expression patterns and how
they interact with other genetic or environmental factors. Moreover, given the large number of variants in
the human genome, it is difficult to distinguish relatively “silent” versus pathogenic variants. All these
questions need to be addressed to understand how particular DNA variants trigger events at the
molecular, cellular and organism level that ultimately lead to observable phenotypes and clinical
syndromes. Together, this indicates that elucidation of disease mechanisms constitutes an even larger
challenge than finding the genetic markers. It also implies that a variety of different experimental and
statistical tools are needed, depending on the nature of the variations and which aspects of function that
are addressed. In my talk I will briefly review some of these approaches and show recent examples from
our group where we have investigated the effects of missense variants in ADHD candidate genes found
by GWAS and sequencing. At the molecular level, communication in the nervous system is mainly
performed by proteins and through complex regulation of protein-protein interactions. We have been
studying regulatory protein complexes involved in serotonin and catecholamine signaling, including the
family of 14-3-3 proteins that are highly expressed in the nervous system and interact with hundreds of
target proteins, implicating them in many different cellular pathways. We have established a pipeline of
methodology for analyzing the effects of multiple nonsynonymous variants in these proteins. This
includes biochemical studies, in silico modeling and molecular systems biology approaches. Many
computational tools have been proposed to predict the effects of SNPs and prioritize variants for further
studies. While homology modeling using high resolution NMR- or X-ray structures sometimes can be
used to predict functional outcomes, we have shown that commonly used “mutation prediction” softwares
such as PolyPhen or SIFT etc. are of limited value and may produce misleading results. This is an area
where improved methods are needed to assess the role of genetic variants. As many disease associated
variants act in concert with other macromolecules, particular attention should be on developing tools to
predict effects on molecular interactions.
Chair: Catharina Lavebratt, Karolinska Institutet
Overall Abstract Details Recent findings propose that telomeres are altered in psychiatric disorders.
Telomeres are protective DNA-protein complexes that form the chromosome ends, which shorten
progressively during each cell division. Telomere shortening is a hallmark of aging and has been
associated with oxidative stress, inflammation, and recently, psychological stress and psychiatric
disorders. In psychiatry, telomeres have so far primarily been studied in blood leukocytes and with regard
to length of the telomeres. Shorter blood leukocyte telomere length (LTL) has been reported in
internalizing disorders including depression, anxiety and PTSD. Altered LTL has been reported also in
schizophrenia. This symposium will focus on data from leukocytes and from the hippocampus, a region
with active telomerase which counteracts the telomere shortening by adding oligonucleotides to the
chromosome ends. A recent study revealed that disruption of hippocampal telomerase activity led to
depressive behavior in mice, while the antidepressant fluoxetine unregulated telomerase activity. Also,
elevation of telomerase activity has been reported to correlate with treatment response to SSRI in
depressed patients. Further, a recent study implicated long-term lithium treatment to protect against
telomere shortening in blood leukocytes. Mammalian telomeres are coated by capping proteins, known as
shelterin. Shelterin is a six-protein complex with compelling evidence for involvement in telomere
protection and regulation of telomerase activity. Dysfunctional telomeres may result from either removal
of shelterin components or excessive attrition due to telomerase deficiency. We will present telomere
length studies in large cohorts, but also findings regarding mechanisms regulating telomerase.
Catharina Lavebratt1
Karolinska Institutet
Individual Abstract This talk will review telomere biology in relation to already published findings in
psychiatry as well as to somatic disorders. Studies have during the last 7 years generated compelling
findings proposing telomere dysfunction in mood disorders and schizophrenia. Telomeres are DNA
nucleoprotein complexes capping the ends of linear eukaryotic chromosomes, which protect the latter
from cellular erosion and fusion with each other. However, telomeres erode progressively with each cell
division, partly because of the end replication problem and also because of oxidative stress, finally
signaling cellular senescence and apoptosis. Key factors in the protection against telomere shortening
include the enzyme telomerase which elongates the telomere, and the shelterin protein complex which
'caps' and stabilizes the telomere ends, and regulates telomerase activity. Telomerase activity has been
positively linked to antidepressant effects and hippocampal neurogenesis. Critically short telomeres
cannot recruit their associated proteins, leaving the chromosome ends �uncapped’. Short telomere length
has been associated with mood disorders, increased all-cause mortality and somatic disorders including
diabetes mellitus, cardiovascular diseases, dementia and osteoporosis, whereas lithium treatment of
bipolar patients was proposed to associate with longer leukocyte telomeres. The pathophysiological role,
and mechanisms, of telomere dysregulation in mood disorders and their treatment are yet to be explored.
Brenda Penninx1, Josine Verhoeven1
VU University Medical Center
Individual Abstract Introduction: Patients with Major Depressive and Anxiety disorders have an
increased onset risk of aging-related somatic diseases such as heart disease, diabetes, obesity and
dementia. This might be the consequence of accelerated cellular aging, which can be indexed by a shorter
length of leukocyte telomeres. We determined whether these psychiatric disorders are associated with
shorter telomeres and whether specific disease characteristics influence this association. Methods: Data
are from a total of 2981 subjects (mean age 41.6 years, 66.8% female) from the Netherlands Study of
Depression and Anxiety. The sample consisted of 1881 current MDD and/or anxiety patients, 518
remitted MDD and/or anxiety patients and 582 control subjects without any lifetime psychiatric disorder
based on the Composite International Diagnostic Interview. Telomere length (TL) was assessed as the
telomere sequence copy number (T) compared to a single-copy gene copy number (S) using quantitative
polymerase chain reaction (qPCR). This resulted in a T/S ratio and was converted to base pairs (bp).
Results: Compared to control subjects (mean bp=5506), adjusted TL was shorter among current MDD
patients (p=.03), remitted MDD patients (p=.04) and current anxiety patients (p=.003). Adjustment for
sociodemographics, health and lifestyle variables did not reduce associations. Higher depression severity
(p<.01), longer symptom duration in the past 4 years (p=.01) and higher anxiety severity (p<.01) were
associated with shorter TL. Conclusions: Our results demonstrate that depressed and anxious patients
show accelerated cellular aging according to a “dose-response” gradient: those with the most severe and
chronic symptoms of depression or anxiety showed the shortest telomere length, representing 7-10 years
of advanced cellular aging compared to controls. We also confirmed the imprint of past exposure to
depression as those with remitted MDD had shorter telomere length than controls.
Yabin Wei1, Lina Martinsson1, Yvonne Forsell2, Martin Schalling2, Lena Backlund2, Catharina Lavebratt2
Karolinska University Hospital, 2Karolinska Institutet
Individual Abstract Mammalian telomeres are protective DNA-protein complexes that form the
chromosome ends, which shorten progressively during each cell division. A number of studies reported
shorter blood leukocyte telomere length (LTL) to be associated with major depression, but whether
telomere length (TL) is shorter in brain and whether levels of telomere protective proteins, known as
shelterin, are associated with depression have never been investigated. Further, long-term lithium
treatment was previously reported to protect against LTL shortening, but how remains elusive. In the
hippocampus region, TL, shelterin expression, and telomerase expression and activity were compared
between: 1) a genetic rat model of depression-like behavior (the Flinders Sensitive Line; FSL) and its
controls (the Flinders Resistant Line; FRL), and between 2) naГЇve FSL and lithium-treated FSL. In
addition, we assessed if rs2736100 in hTERT associated with depression and number of depressive
episodes in cohorts of unipolar depression and bipolar disorder (BP), respectively. Finally, by using
human whole saliva DNA we compared the TL between unipolar depression patients and healthy
controls, and tested the TL correlation between whole saliva DNA and blood leukocyte DNA. The naГЇve
FSL, compared to FRL, exhibited shorter hippocampal TL, which associated with down regulation of
Terf2, Rap1 and Tert expression and reduced telomerase activity. Lithium treatment rescued the Tert
expression and telomerase activity in the FSL. Rs2736100 associated with a unipolar depression
diagnosis and with number of depressive episodes in BP. Saliva TL was decreased in depression patients
compared to the healthy controls, and it correlated positively with TL in blood leukocytes. This is the
first report on shelterin in psychiatric disorder and on lithium’s mechanism in protection against telomere
shortening. We also provide the first finding of hTERT variation associated with depression.
Adolfo Sequeira1, Firoza Mamdani1, Marquis Vawter1, William Bunney1
University of California Irvine
Individual Abstract Stress and depression have been associated reduced neurogenesis and hippocampal
volume in animal and in human studies. Telomere shortening has been observed in blood lymphocytes in
depressed patients in some but not all studies while a post-mortem brain study using occipital cortex
tissue did not observe any reduction of telomere length in depression. We hypothesized that because
telomere length is the result of the balance between dividing and non-dividing cell telomere degradation,
variable telomere length might be observed across brain regions and that telomere length might be
reduced in psychiatric disorders as a consequence of stress mediated accelerated cellular aging. Postmortem human brains (N=40; 10 per diagnosis) obtained through the UCI brain bank were dissected and
used to extract DNA. Telomere length was quantified using Q-PCR and compared to a single copy gene
(t/s) in several brain regions (dorsolateral prefrontal cortex (DLPFC), hippocampus, amygdala, nucleus
accumbens and substantia nigra (SN)) in major depressive disorder (MDD), bipolar disorder (BP),
schizophrenia (SZ) and control subjects. We observed significant differences in telomere length across
brains regions, suggesting variable levels of cell aging, with SN and hippocampus having the longest
telomeres and the DLPFC the shortest. Also, a significant decrease in telomere length was observed
specifically in the hippocampus of MDD subjects even after controlling for age. Our results suggest
accelerated cellular aging in depression specifically in the hippocampus.
Chair: Peter Visscher, University Of Queensland
Overall Abstract Details Complex traits and disorders such as schizophrenia are associated with the
effects of multiple genes. These disorders often cluster in families, have no clear-cut pattern of
inheritance, and have a high fraction of phenotypic variance attributable to genetic variance (high
heritability). It is becoming clear that many genes influence most complex traits and disorders. In such a
scenario with a very high number of risk genes, each gene has a tiny effect. This makes it difficult to
determine an individual’s risk, and to identify disease mechanisms that can be used for development of
new effective treatments. Genome-wide association studies (GWAS) have identified many traitassociated single nucleotide polymorphisms (SNPs), but so far these explain only small portions of the
heritability of complex disorders. This “missing heritability” has been attributed to a number of potential
causes. However, it has been shown that a large proportion of the missing heritability exists GWAS data
when associations of SNPs are examined in aggregate. This implies that there are very many common
variants each with small genetic effects. These effects cannot be reliably detected with traditional GWAS
statistical methods given realistic sample sizes. Thus, there is a need for innovative statistical approaches
to identify polygenetic effects and to reduce the proportion of �missing heritability’. We describe recent
GWAS results and novel statistical tools that are designed for polygenic traits. These methods enhance
gene discovery, improve replication rates of discovered risk variants, and improve estimates of polygenic
risk. The basic framework relies on a simple model that assumes a large proportion of loci are either null
(unassociated with the phenotype of interest) or have negligible effects, but that a small proportion have
larger (though still small) effect sizes. The first talk (A. Schork) outlines relevant statistical aspects of this
model as applied to PGC schizophrenia (SCZ) GWAS data. The second talk (O. Andreassen) presents a
genetic pleiotropy-informed method to improve power to identify new loci associated with SCZ. The
third talk (S.H. Lee) presents an extension of a recent two-trait approach to a multiple trait model and
applies multi-trait genomic best linear unbiased prediction (MTGBLUP) for individual risk prediction.
The fourth talk (M. Reimers) describes an empirical Bayes algorithm to integrate various kinds of
genomic data, and selects one or a few specific SNPs within wide loci implicated by GWAS, and further
identifies many more loci than GWAS alone. The Discussant (N. Wray) will provide an overview of how
these statistical methods and results reflect current understanding of complex diseases and potential future
directions of statistical methodological research.
Andrew Schork1, Wesley Thompson1, Yunpeng Wang1, Anders Dale1
University of California, San Diego
Individual Abstract This talk outlines the relevant statistical aspects of our mixed model approach to
statistical inference as applied to GWAS data. This talk also describes a novel resampling algorithm for
estimating posterior effect sizes directly. Model parameters estimated from this resampling algorithm can
also be used to estimate the posterior probability that a given locus is null given its observed test statistic,
termed the local false discovery rate (fdr). Applying this methodology to the Psychiatric Genetics
Consortium SCZ GWAS data demonstrates that a simple scale mixture of normal model fits replication
effect sizes very closely, with strong implications for tagged heritability, power for gene discovery, and
power for estimation of polygenic risk.
Ole Andreassen1
University of Oslo
Individual Abstract We have developed a new statistical framework leveraging genetic pleiotropy to
improve discovery and effect size estimation. This is based on a genetic pleiotropy-informed method to
improve gene discovery using genome-wise association study (GWAS) summary statistics data
(Andreassen et al, 2014). This methodology was used to identify new loci associated with psychiatric
disorders, which are highly heritable disorders with significant missing heritability. Bipolar disorder and
schizophrenia have overlapping clinical characteristics, and are both regarded as polygenic complex
disorders. We applied the new statistical framework to boost the discovery of new genes, using nonoverlapping summary stats results. The new tools provided 3-4 times increased discovery rate of common
gene variants in schizophrenia and bipolar disorders. These discoveries also have a high replication rate.
The new statistical tools can also be used to investigate polygenic overlap between neurological disorders
and psychiatric phenotypes. We leveraged the pleiotropy between multiple sclerosis and schizophrenia
and bipolar disorders, and discovered a strong genetic overlap between multiple sclerosis and
schizophrenia but not bipolar disorders. Follow up analyses revealed that most of the overlap was located
in HLA alleles, possibly distinguishing between bipolar disorder and schizophrenia. Further, the effect of
five overlapping HLA variants were opposite in multiple sclerosis and schizophrenia. Epidemiological
and clinical studies suggest co-morbidity between schizophrenia and cardiovascular disease (CVD) risk
factors, including systolic blood pressure, triglycerides, low and high-density lipoprotein cholesterol,
body mass index, waist-hip-ratio, and type 2 diabetes. Applying a novel conditional false discovery rate
method, we identified more than 25 loci associated with schizophrenia at a conditional fdr level of 0.01.
Of these, 10 loci were associated with both schizophrenia and CVD risk factors, mainly triglycerides, low
and high-density lipoproteins cholesterol, but also waist hip ratio, systolic blood pressure, and body mass
index. Recently, we have applied the new tools to investigate the pleiotropy between immune-mediated
diseases and schizophrenia, providing strong evidence for overlapping genes and thus strengthening the
immune component of schizophrenia pathophysiology. We have also recent results providing strong
evidence for overlapping common variants in schizophrenia and prefrontal cortex area obtained from
MRI GWAS. Together these findings demonstrate an important role of pleiotropy in psychiatric
disorders, and show how we can leverage polygenic pleiiotropy to provide better understanding of the
polygenic architecture of psychiatric disorders. Andreassen OA, Thompson WK, Dale AM. Boosting the
Power of Schizophrenia Genetics by Leveraging New Statistical Tools. Schizophr Bull. 2014
Sang Hong Lee1, Gerhard Moser1, Guo-Bo Chen1, PGC-CDG NA, Naomi Wray1
The University of QueenslandIndividual Abstract Most common diseases are highly polygenic and each
variant explains only a small proportion of the genetic variation. Therefore, the accuracy of risk prediction
is low even when using a polygenic approach. A major factor determining how well a polygenic model can
predict a trait value in an independent sample is the sample size of the training data. Using more
individuals will provide more information about the effect of a specific SNP. Another way to increase
information about SNP effects is to incorporate information from correlated traits. Using a bivariate linear
mixed model, we recently demonstrated significant shared genetic factors across five psychiatric disorders
(schizophrenia, bipolar disorder, major depression, autism and ADHD) from the Psychiatric Genomics
Consortium (PGC). Here we extend our two-trait approach to a multiple trait model and apply multi-trait
genomic best linear unbiased prediction (MTGBLUP) for individual risk prediction. MTGBLUP is
expected to be more powerful as it uses pleiotropy between disorders and simultaneously evaluates
individual risk across disorders. We apply our model to the cross-disorders?) PGC GWAS data and show a
significant increase of prediction accuracy of schizophrenia risk using MTGBLUP. We further
demonstrate a relationship between functional annotated SNPs and increased prediction accuracy of SCZ.
Mark Reimers1, Kenneth Kendler2, Aaron Wolen2
Virginia Institute for Psychiatric & Behavioral Genetics, 2Virginia Commonwealth University
Individual Abstract Risk SNPs for psychiatric disorders are likely to lie in DNA regions with large
effects on gene regulation in the brain, which are a relatively small fraction of the genome. Several
studies have shown that the majority of genetic variants currently implicated by GWAS for many diseases
lie in open chromatin regions in the specific tissue relevant to the disease, or have other epigenetic marks
indicative of function. We describe an empirical Bayes algorithm to integrate various kinds of genomic
data with genetic data that selects one or a few specific SNPs within wide loci implicated by GWAS, and
further identifies many more loci than GWAS alone. This approach makes use of information about
regulatory regions of the genome obtained from the recently released ENCODE epigenetic data as well as
using evolutionary conservation. We adopt an empirical Bayes formalism to accomplish the integration;
the talk will discuss how to estimate the prior and conditional probabilities essential to make the method
work. We illustrate the method on two notoriously difficult psychiatric disorders: schizophrenia and
bipolar disorder, and identify hundreds of specific variants with high probability of association with each.
We validate the predictions of the method on data from independent genetic studies. Follow-up studies
confirm that the Bayesian posterior probabilities of risk for SNPs are indeed accurate; and confidence
intervals are available. We show how to combine GWAS results with several different types of genomic
information in an extensible and flexible manner to obtain much greater power in psychiatric genetic
Chair: Melvin McInnis, University of Michigan
Overall Abstract Details Neurons and glial cells derived from induced pluripotent stem cells (iPSC)
provide an opportunity to study functional cellular models from individuals affected with
neuropsychiatric disorders. One of the major limitations in the study of these conditions is limited access
to the primary organ (brain) tissue, which may be at least partially overcome by iPSC technology. These
models will provide new information on gene expression patterns at sequential stages of differentiation;
developmental patterns and predictors of functional capacity may be identified by studying expression
profiles at specific culture times. Electrophysiological analyses of the cell and cell networks can be
studied and compared; the cellular microenvironment perturbed and analyzed, and the metabolite profile
of the cell culture supernatants examined. The coordinated study of the genetics, biochemistry, and
physiology of iPSC derived neural cells thus provides a window on brain organization. This symposium
will present and discuss the current status of induced pluripotent stem cell modeling in the affective
disorders and psychosis. The genetic basis of such disorders is clearly established; however there is
considerable heterogeneity in the genetic findings. There are no established pathogenesis, etiology, or
anatomical substrates, and there is a critical unmet need for informative models that will lead to more
efficacious treatments. Data and analyses will include expression patterns from iPSC cells and derived
neurons at sequential stages of development, the undifferentiated iPSC and neurons and glial cells
derived from the iPSC. Gene expression between control and affected neurons are remarkably similar,
but neurons differentiated from them are different in their transcription factor expression pattern, in their
expression of membrane receptors and ion channels. Interestingly, TF consistent with ventral neuronal
cell fate (MGE) are increased in BP, and TF that control or maintain dorsal fate increased in neurons
differentiated from control iPSC. Independent electrophysiological studies are emerging that suggest
differences in action potentials in bipolar derived neurons vs controls. Methods and results to be
presented include voltage clamp and calcium transient studies that are suggest altered excitability of the
bipolar derived neuron, and co-culture with lithium affects BP and C neurons differently. Analysis of the
supernatants from BP and C neuronal cultures also consistent identify fundamental biological differences
in the cells. The plasticity of bipolar derived neurons appears to be less compared to control neurons. We
will present data from large scale screening of small molecules in iPSC cells to demonstrate the
feasibility of a large scaled approach to identify patterns of responses that may suggest novel treatment
modalities.This approach will lead to prioritization of small molecules for testing on specific signaling
Roy Perlis1, Jennifer Wang1, Steven Shamah2, Alfred Sun3, Stephen Haggarty1
Massachusetts General Hospital, 2X-Body, Inc., 3Stanford University
Individual Abstract Cellular reprogramming may allow the disease biology of psychiatric disorders to
be investigated using patient skin cells transdifferentiated to neurons. A major challenge in such efforts is
the efficient identification and characterization of relevant cellular or molecular phenotypes. This
presentation will describe the generation and investigation of induced neuron models of lithium response
in bipolar disorder, based on fibroblasts from individuals with bipolar disorder as well as healthy control.
We utilized a high-throughput, label-free imaging assay based on a nanostructured photonic crystal
biosensor to characterize induced neurons from these three groups. In this assay, the signal generated is a
measure of adhesion of the cell to the underlying surface, allowing quantification of changes in cell
count, morphology, and adhesion over time. Cells drawn from lithium-responsive individuals differed
significantly from those drawn from lithium-nonresponsive individuals. In addition to discussing these
results, more general methodologic challenges - including substantial effects of age and sex, as well as
inter individual variability - will be discussed, and their implication for future cellular modeling efforts.
Our results suggest the possibility that this platform may be useful in development of high-throughput
approaches to drug discovery, but also highlight the major methodologic challenges in applying cellular
reprogramming to understand psychiatric illness.
Guo-Li Ming1
Johns Hopkins University
Individual Abstract Schizophrenia and affective disorders are chronic and generally disabling brain
disorders with a prominent genetic basis and with neurodevelopmental origin. A number of susceptibility
genes have been identified, including DISC1, neuregulin, COMT, FEZ1. How dysfunction of these genes
leads to aberrant neural development and contribute to the pathology of the disorder is largely unknown.
DISC1 is by far the best-characterized risk genes for schizophrenia and other major mental disorders, and
almost nothing is known about its function in human neural development. To understand how mutation of
DISC1 gene in patients impacts the development of human neurons, we generated iPSC lines from
multiple patients from one family with a DISC1 mutation and we are able to differentiate these iPSCs into
forebrain neurons in high efficiency. I will discuss our recent findings on the roles of DISC1 in
morphological developmental and synaptic development of human neurons derived from patient specific
Sevilla Detera-Wadleigh1, Liping Hou2, Xueying Jiang2, Nirmala Akula2, David Chen2, Barbara Mallon3,
Nahid Tayebi4, Winston Corona2, Layla Kassem2, Ellen Sidransky5, Marc Ferrer6, Francis McMahon2
National Institute of Mental Health, NIH, 2Human Genetics Branch, NIMH/NIH, 3NIH Stem Cell Unit,
NINDS/NIH, 4Neurogenetics Section, NHGRI/NIH, 5Molecular Neurogenetics Section, NHGRI/NIH, 6
Division of Preclinical Innovation, NCATS/NIH
Individual Abstract Bipolar disorder (BD) is a complex neuropsychiatric disease marked by debilitating
episodes of mania and depression afflicting approximately 1% of studied populations. Recent genomewide association studies on large samples have identified BD-associated genetic variants, potentially
representing risk loci. Two major challenges in understanding the biological underpinnings of BD
include: a) determining how risk loci function to confer aberrations in mood, and b) developing novel,
effective, fast-acting, and better-tolerated therapeutic agents. Recent advances in cell biology offer a
unique and unprecedented opportunity to conduct functional studies on a patient’s own neural cells
generated through induced pluripotent stem cell (iPSC) technology. These new templates permit
discovery of distinct cellular phenotypes that would facilitate the search for novel, fast-acting and bettertolerated therapeutic agents. Novel agents could benefit patients, particularly those unresponsive to the
existing mood stabilizers, such as lithium and valproate (VPA). We have developed an assay system
based on viability of human iPSC-derived neural stem cells after challenge with a drug known to induce
neuropsychiatric disturbances in humans. Unlike terminally-differentiated neurons, neural stem cells are
renewable and expandable. Drug challenge is performed using mefloquine, an anti-malarial drug widely
reported to cause psychiatric disturbances, including mania and psychosis, particularly in individuals with
a preexisting psychiatric condition. In preliminary studies, we found that mefloquine destroys the
viability of iPSC-derived neural stem cells, and that lithium and VPA, at therapeutic levels, exert a
neuroprotective effect on cell viability. To our knowledge, this is the first study of neuroprotection by
mood stabilizers in human iPSC-derived neural stem cells after challenge by a psychiatric symptominducing drug. The assay is amenable to a high throughput format and in collaboration with the National
Center for Advancing Translational Sciences (NCATS) at NIH, we are poised to screen three large
collections of small molecules to identify compounds that mimic the neuroprotective effect of lithium and
VPA in our model system. Gene expression profiles reveal the molecular repertoire related to various
drug treatments in neural stem cells. Compounds that mimic the neuroprotective effects of lithium and/or
VPA will be strong candidates for further evaluation as novel mood stabilizing treatments for patients
with bipolar disorder.
Melvin McInnis1, Sue Oshea1, Monica Bame1, Cindy Delong1, Aislin Williams1
University of Michigan
Individual Abstract Bipolar disorder (BP) affects millions of individuals worldwide, yet progress in
understanding its genesis and improving treatments has been hampered by the lack of viable cell models.
Patient-derived induced pluripotent stem cells (iPSC) now offer the opportunity to study the development
of neural tissues and the prospect of identifying novel disease mechanisms in BP. We have derived iPSC
from three individuals with BP and three healthy controls and differentiated them into telencephalic
neurons. RNAs were extracted from undifferentiated iPSC and neurons derived from them and
microarray analysis carried out. Expression of transcription factors that convey regional neuronal cell fate
was significantly different between the two groups. Neurons derived from control iPSC expressed
transcripts that confer dorsal telencephalic fate, while neurons derived from BP iPSC expressed genes
involved in the differentiation of ventral (MGE) brain regions. During development, the majority of
cortical interneurons originate in the MGE, undergo tangential migrations to their adult cortical locations
where they form a variety of inhibitory GABAergic interneurons. Although only 20% of the total number
of cortical neurons, interneurons play a critical role in maintaining the normal balance in cortical activity
by making local synapses on long-projecting excitatory (glutamatergic) pyramidal neurons. Consistent
with the increase in transcripts involved in interneuron cell fate, GABA expression by BP vs C neurons
was elevated throughout the culture period. To determine if the phenotype could be altered, cells have
been exposed to ventralizing agents (purmorphamine) or dorsalizing agents (lithium) and are being
evaluated using PCR, western blot and immunohistochemistry. We have found that iPSC from both BP
and controls are responsive to patterning cues, increasing expression of NKX2-1 (ventral identity) or
EMX2 (dorsal). Since interneuron dysfunction has been suggested to underlie a number of
neurodevelopmental and neuropsychiatric conditions, alterations in the specification or function of
GABAergic interneurons would be expected to permanently disrupt the excitatory-inhibitory balance in
the cortex, contributing to BP.
6:00 PM - 8:00 PM
Symposia Sessions
Chair: Margarita Rivera, Institute of Psychiatry, King's College London
Overall Abstract Details Major depressive disorder (MDD) or depression is a highly prevalent and
heterogeneous mental disorder. It is a major public health problem and one of the leading causes of
disease burden and disability in adults worldwide. There is growing evidence suggesting that rates of
physical conditions, such as obesity, diabetes, migraine and cardiovascular disorder are increased among
people with depression. Mortality rates are elevated in people with depression, mainly due to comorbid
physical conditions. These people are also less likely to receive standard levels of health care for most of
the comorbid physical conditions. Besides, both depression and physical conditions are associated with
substantial individual and societal economic costs. Shared genetic effects (genetic pleiotropy) may
contribute to the link between depression and comorbid physical conditions. Phenotypic heterogeneity of
MDD may play a role in this relationship, as different clinical subtypes may be characterized by a
partially distinct genetic liability. The relationship between depression and comorbid physical conditions
may therefore differ when considering the specific clinical subtypes as compared to the overall MDD
diagnosis. To further explore the contribution of shared genetic factors to the association between
depression and certain physical conditions (BMI, migraine, inflammation and autoimmune disorders), we
performed several strategies based on genome-wide data including Genome Polygenic Risk Scoring
(GPRS), Genomic-Relatedness-Matrix Restricted Maximum Likelihood (GREML) and SNP Effect
Concordance Analysis (SECA). Our findings have important implications for studies investigating shared
genetic effects of depression and physical conditions, and highlight the importance of using novel
statistical genetic strategies in order to disentangle the observed comorbidity.
Margarita Rivera1, Chi-Fa Hung2, Nick Craddock 3, Michael J. Owen3, Martin Preisig4, Stefan Kloiber5,
Bertram MГјller-Myhsok 5, Susanne Lucae5, Florian Holsboer5, Oliver S.P. Davis 6, Gerome Breen7, Ian
W. Craig7, Cathryn M. Lewis 7, Anne E. Farmer 7, Peter McGuffin7
King's College London, 2Kaohsiung Chang Gung Memorial Hospital and Chang Gung University
College of Medicine, 3Cardiff University, 4Lausanne University Hospital, 5Max-Planck-Institute of
Psychiatry, 6UCL Genetics Institute, University College London, 7MRC Social Genetic and
Developmental Psychiatry Centre, Institute of Psychiatry, King's College London
Individual Abstract Background: Both obesity and major depressive disorder (MDD) are prevalent in
developed countries and cause huge disease burden. Previous studies have shown strong association
between obesity and MDD but why they cluster together remains unclear. Given the high heritability of
both disorders it is worth considering that the clustering of MDD and obesity might be partly mediated by
common genetic factors. We aimed to investigate the phenotypic variance of body mass index (BMI) and
MDD explained by genetic variance captured by genome-wide association studies (GWAS) data and
genetic correlation between BMI and MDD using GREML (Genomic-Relatedness-Matrix Restricted
Maximum Likelihood) analysis. Methods: The sample consists of 3,872 unrelated individuals from the
RADIANT study, and another 1,645 individuals from the Munich depression study. DSM-IV or ICD-10
diagnosis of MDD was ascertained using the Schedules for Clinical Assessment in Neuropsychiatry
(SCAN) interview. The controls were screened for lifetime absence of any psychiatric disorder using a
modified version of the Past History Schedule and the Composite International Diagnostic Screener,
respectively. All the individuals were genotyped using the Illumina HumanHap610-Quad BeadChip
(Illumina, Inc., San Diego, CA, USA). BMI was defined as weight in kilograms divided by height in
meters squared (kg/m2). We performed univariate and bivariate GREML analysis using the GCTA
(Genome-wide Complex Trait Analysis) software package. Results & Conclusions: We found in the
combined Radiant and Munich sample that the phenotypic variance accounted for by common SNPs was
15% for BMI (s.e=0.09, p=0.04) and 33% for MDD (s.e=0.08, p=5x10-6). There was also evidence of
genetic correlation between BMI and MDD (rG=0.40, s.e=0.23, p=0.08) suggesting that a genetic overlap
may contribute to the association between MDD and high BMI. The results confirm that both BMI and
MDD are heritable with a significant proportion of phenotypic variance explained by the additive genetic
effects of common SNPs.
Yuri Milaneschi1, Femke Lamers1, Dorret Boomsma2, Brenda Penninx1
VU University Medical Center/GGZ inGeest, 2VU University
Individual Abstract Introduction Phenotypic heterogeneity of Major Depressive Disorder (MDD) may
contribute to discrepant or blurred effect sizes in large collaborative genetic studies. Studies based on
data- driven techniques have confirmed that depressed populations can generally be divided into a
�typical’ (a.k.a. �melancholic’) and an �atypical’ subtype, differentiated mainly by the direction of change
in vegetative symptoms and associated with distinct biological correlates and specific genetic variants.
We hypothesized that MDD subtypes may be characterized by a partially distinct genetic liability. We
preliminary tested this hypothesis by comparing genomic profile risk scores (GPRS), derived from large
discovery cohorts, for psychiatric (MDD, schizophrenia, bipolar) and somatic (CRP-inflammation, BMI)
disorders across MDD subtypes in a Dutch target sample. Methods: The target sample is represented by
1649 MDD patients from the Netherlands Study of Depression and Anxiety (NESDA) and 1810 screened
controls mainly from the Netherlands Twin Registry (NTR). Autosomal SNPs were genotyped on the
Affymetrix 6.0 Human Genome-Wide SNP Array. Depression diagnoses are based on DSM-IV criteria
and MDD subtypes (severe melancholic and severe atypical) are derived from latent class analyses
applied to MDD endorsed symptoms. GPRS were generated based on meta-analyses results from the
Psychiatric Genomics Consortium (PGC) for MDD, schizophrenia (SCZ) and bipolar disorder (BIP),
from the Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE) Consortium for
CRP, and from the Genetic Investigation of Anthropometric Traits (GIANT) consortium for BMI.
Results: Overall MDD diagnostic status was significantly predicted by the GPRS for psychiatric
disorders but not for CRP and BMI. Results suggested differential patterns of association between
specific GPRS and MDD sub-phenotypes, with SCZ being most strongly associated with the
melancholic subtype and BMI with the atypical subtype. Conclusions Preliminary evidence suggests
differential polygenic signature across MDD clinical sub-phenotypes. We planned to confirm these
findings by using complementary techniques (SNP Effect Concordance Analysis, SECA) and by
including additional cohorts in order to achieve the adequate statistical power to compare single genetic
variants across MDD subtypes and to estimate their (co)heritability via genomic-relationship-matrix
restricted maximum likelihood (GREML) methods. The possibility to identify a more homogenous
MDD phenotype may help future research on the genetic determinants of depression.
Lannie Ligthart1, Dale Nyholt2, Brenda Penninx3, Cathryn Lewis4, Dorret Boomsma1
VU University, 2QIMR Berghofer Medical Research Institute, Brisbane, 3VU Medical Center,
Amsterdam, 4King's College London
Individual Abstract Patients with major depressive disorder (MDD) are known to have high rates of
migraine and other pain symptoms. The mechanisms underlying this comorbidity have been debated for
decades. Twin and family studies have shown there is a genetic correlation between migraine and
depression, which may reflect a variety of underlying mechanisms, including pleiotropy, uni- or
bidirectional causation, or a syndromic relationship. To further investigate these mechanisms, we applied
several strategies based on genetically informative data, including a co-twin control design and crossdisorder prediction using polygenic scores and SNP effect concordance analysis (SECA). Questionnaire
data on migraine, pain symptoms, anxiety and depression, and genotype data were collected in
participants of the Netherlands Twin Registry and the Netherlands Study of Depression and Anxiety
(NESDA). Cross-disorder genetic risk prediction was performed using GWA summary statistics from the
UK RADIANT study on major depressive disorder (MDD) and the Australian Twin Migraine study
(migraine). Our research showed that the presence of pleiotropic genetic effects alone is unlikely to
explain the comorbidity of depression and migraine. The results suggest that the observed comorbidity is
explained by a subset of individuals with both MDD and migraine, in whom the migraine may be a
symptom or consequence of MDD. We speculate that in these patients, migraine might be viewed as a
symptom of MDD, rather than a separate comorbid condition. Furthermore, our observation of a highly
consistent pattern of comorbidity between depression and pain symptoms, regardless of anatomical site, is
consistent with the hypothesis that not only migraine, but pain in general can arise as a symptom of
MDD. Further analysis should elucidate whether our findings with respect to the genetic overlap of MDD
and migraine can be generalized to MDD and pain in general. These findings have important implications
for genetic studies of depression and comorbid pain conditions, and are particularly relevant to studies
investigating cross-disorder genetic effects.
Jack Euesden1, Andrea Danese1, Ian Scott1, Cathryn Lewis1
King's College London
Individual Abstract BACKGROUND Comorbidity, the co-existence of two or more diseases in an
individual, can provide an insight into the aetiology of many common disorders. Major Depressive
Disorder (MDD) shows an unusual pattern of comorbidities; although no risk alleles have been identified
for MDD via genome-wide association studies (GWAS), comorbidity can be a sign of genetic overlap –
pleiotropy – between disorders. MDD is more common in autoimmune disorders such as Rheumatoid
Arthritis (RA), Crohn’s Disease (CD) and Ulcerative Colitis (UC), than would be predicted by chance.
Many theories for the aetiology of MDD have focused on the ability of abnormal inflammatory protein
levels in the blood to influence mood – the �Cytokine Hypothesis’ - and so investigating the genetic
overlap between MDD and autoimmune disorders may illuminate biological pathways underlying MDD.
Furthermore, depression in RA patients is associated with worse pain, more functional disability and
higher rates of healthcare utilization; establishing its cause in RA is therefore an important research goal.
METHODS In order to investigate the genetic overlap between MDD and autoimmune disorders, we
used a number of statistical genetics techniques, including Genome Profile Risk Scoring (GPRS) and
bivariate Genome-Relatedness-Matrix Restricted Maximum Likelihood (GREML). We focused on MDD
and RA, and expanded this to UC and CD, both of which include arthritis in their extra-intestinal
symptoms. We performed GPRS, using GWAS results to construct polygenic risk scores in an
independent test dataset. The RADIANT dataset contains GWAS data on 1,624 MDD cases and 1,588
screened controls. We can therefore test the predictive value of a genetic risk profile for an autoimmune
disorder on MDD status. We calculated risk profiles for RA using GWAS results from the BIRAC and
YEAR consortia, testing their prediction of MDD status using Nagelkerke’s Pseudo R2. We repeated this
calculating risk profiles for Crohn’s Disease and Ulcerative Colitis using results from the Inflammatory
Bowel Disorder Consortium. Finally, we used bivariate GREML to investigate the genetic correlation
between MDD and RA. We used the WTCCC RA dataset in addition to the RADIANT MDD dataset,
and implemented bivariate GREML in the GCTA software package. RESULTS & CONCLUSIONS In
GPRS, there was no evidence for prediction of MDD status from RA polygenic risk scores (minimum pvalue 0.42). The GREML estimate for the genetic correlation between MDD and RA was non-significant
(rG = 0.29, p = 0.094). We find no evidence for shared genetic risk between RA and MDD. This is
supported by both GREML and GPRS, methods, whose assumptions we review. Our results have
implications for the Inflammatory Cytokine hypothesis of MDD and indicate the extent to which GWAS
datasets can be leveraged using newer statistical genetics techniques in order to dissect observed
phenotypic correlations.
Chair: Chunyu Liu, University of Illinois at Chicago
Overall Abstract Details This symposium selected four studies, including imaging genetics of
schizophrenia, candidate gene study of schizophrenia, brain regulatory network study and gene- and
pathway-based secondary analysis of large genome-wide association study (GWAS) data. Dr. Xue has
worked on study of GABRB2 in Chinese schizophrenia for a decade, representing a focused research of
candidate genes. Results of extensive molecular genetics, population genetics, evolutionary genetics, and
epigenetics will be presented about GABRB2's role in schizophrenia and relevant intermediate
phenotypes. Dr. Yao and her collaborators studied the global brain connectivity variations in healthy
siblings, comparing with that of schizophrenia patients and healthy controls, looking into the plasticity
mechanism of those healthy siblings although they may share common genetic risks and environmental
factors with patients. Connectivity variations and alternations in patients and their siblings will be
presented. Dr. Chen and his collaborators used postmortem brain samples to study miRNA-mRNA
regulatory network, and further identified novel regulatory pathway related to psychiatric diseases. Dr. Li
and his collaborators developed gene and pathway-based methods to re-analyze GWAS data of
psychiatric diseases. This study demonstrates that integrative analysis is important to characterizing
genetic risk genes of complex diseases. Besides these studies presented in the Symposium, several posters
about Autism and Schizophrenia studies in China will be presented in the conference. Altogether, they
showcase the current psychiatric genetics researches in China. Chinese investigators welcome more
collaborative researches, contributing not only samples and data but also original, innovative ideas on
studies of psychiatric disorders in Chinese.
Hannah Hong Xue1
Hong Kong University of Science and Technology
Individual Abstract Deciphering the molecular basis of schizophrenia is essential to effective
management of this devastating mental disorder. Over the past decade, my research group has focused on
the basic research on schizophrenia etiology through the discovery and characterization of a
schizophrenia-associated gene – GABRB2, coding for GABAA receptor? 2 subunit. The association
between schizophrenia and single nucleotide polymorphisms (SNPs) in introns 9 and 10 of GABRB2,
first reported by my group (1), has been cross-validated for multiple ethnic groups (2, 3). Functional
impacts of the schizophrenia associated non-coding SNPs in GABRB2 have been demonstrated at both
mRNA and protein levels, viz. genotype-dependent alterations in mRNA expression and splicing, and
effects of genotypes on isoform ratios and electrophysiological attenuation of GABAA receptors (4, 5).
Through extensive molecular genetics, population genetics and evolutionary genetics characterizations,
GABRB2 has been shown by us to be under strong positive selection (6), active recombination (7) as
well as genomic imprinting (8), likely contributed to by a human lineage-specific insertion of an AluY
transposable element. Our work on epigenetic regulation of GABRB2 revealed its developmental control
and disruption in schizophrenia (9). Most recently, we have also extended GABRB2 association from
psychotic disorders to social cognitions (10). Our work has thus improved current understanding of
schizophrenia at molecular level centered at GABRB2, which represents at present one of the best
characterized schizophrenia candidate genes. References: 1. Wing-Sze Lo, Ching-Fun Lau, Zhenyu
Xuan, Anthony Chun-Fung Chan, Guo-Yin Feng, Lin He, Zhong-Chang Cao, Hua Liu, Qing-Ming
Luan, and Hong Xue (2004) Association of SNPs and haplotypes in GABAA receptor ?2 gene with
schizophrenia. Molecular Psychiatry 9(6): 603-608 2. TL Petryshen, et al. (2005) Genetic investigation
of chromosome 5q GABAA receptor subunit genes in schizophrenia. Molecular Psychiatry 10:10741088 3. Wing-Sze Lo, Mutsuo Harano, Micha Gawlik, Zhiliang Yu, Jianhuan Chen, Frank Wing Pun,
Ka-Lok Tong, Cunyou Zhao, Siu-Kin Ng, Shui-Ying Tsang, Naohisa Uchimura, Gerald Stoeber and
Hong Xue (2007) GABRB2 association with schizophrenia: commonalities and differences between
ethnic groups and clinical subtypes. Biological Psychiatry 61:653-660 4. Cunyou Zhao, Zhiwen Xu,
Jianhuan Chen, Zhiliang Yu, Ka-Lok Tong, Wing Sze, Cario Lo, Frank Wing Pun, Siu-Kin Ng, ShuiYing Tsang, and Hong Xue (2006) Two isoforms of GABAA receptor ОІ2 subunit with different
electrophysiological properties: differential expression and genotypical correlations in schizophrenia.
Molecular Psychiatry 11:1092-1105 5. Zhao, CY, Zhiwen Xu, Feng Wang, Jianhuan Chen, Siu-Kin Ng,
Pak-Wing Wong, Zhiliang Yu, Frank W. Pun, Lihuan Ren, Wing-Sze Lo, Shui-Ying Tsang and Hong
Xue (2009) Alternative-splicing in the Exon-10 region of GABAA receptor ОІ2 subunit gene:
Yin Yao1, HongBao Cao2, Jinsong Tang3, Xiaogang Chen3
National Institutes of Health, 2NIMH, 3Xiangya Second Hospital
Individual Abstract Background: Schizophrenia (SCZ) is a complex disease that has been hypothesized
to arise from functional dysconnectivity of the brain. However, study results discovered various
inconsistent abnormal connectivity alterations in SCZ patients. Recent work used both SCZ patients and
their nonpsychotic siblings in SCZ studies to seek common abnormalities as biomarkers of this disease.
To date, no researchers have compared the global alterations amongst SCZ patients, their healthy siblings
and healthy controls. In this study, we investigate the global brain connectivity variations in healthy
siblings, compare with that of SCZ patients and healthy controls, re-examining the compensative
plasticity mechanism in healthy siblings Methods: Resting state functional magnetic resonance imaging
(fMRI) data were collected from 107 study subjects, including 44 healthy controls, and 32 schizophrenia
patients that are treatment-resistant, and 31 of their healthy siblings with no SCZ history. The whole brain
was parcellated into 1000 brain networks using a fix point clustering (FPC) algorithm proposed by us.
Connectivity features in the number of 500500 were analyzed, including both intra- and internetwork/region connectivity. Then the ANOVA analysis was independently conducted for each
connectivity feature with 4 contrasts: C1, Cases, siblings and healthy controls; C2, Case vs. sibling; C3,
Case vs. control; and C4, Sibling vs. control. Results: This work indicates that healthy siblings, while
compared with the SCZ patients, presented connectivity variations involving greater than 60% of the
whole brain regions. At the same level of significance, the healthy siblings exhibited little connectivity
alteration in small brain regions in contrast with healthy controls. On the other hand, SCZ patients have
great than 40% comparable brain regions that demonstrate connectivity variations in contrast with healthy
controls; And those brain regions are partially overlapped with that of siblings/cases study. In addition,
the changes observed between cases and controls in altered connectivity feature numbers are smaller than
that of cases and siblings (less than a half). When evaluated using the multivariate classification
approach, selected connectivity features provided the highest identification accuracies of 84.1%, 92.1%
and 81.3% for C2, C3 and C4 respectively. Conclusion: Healthy siblings of individuals with SCZ
demonstrate functional regulations within large area of brain regions. Although most of those changes are
moderate in comparison with healthy controls, the majority of them show significance in sibling/case
study, indicating that connectivity variations in SCZ patients and their healthy siblings are in different
scales or even opposite directions. We thus postulate that those alterations in healthy siblings may be a
form of virtuous compensative regulations preventing them from becoming ill.
Chao Chen1, Lijun Cheng2, Chunling Zhang3, Judith Badner3, Elliot Gershon3, Chunyu Liu3
Central South University, 2Northwestern University, 3The University of Chicago
Individual Abstract MicroRNAs (miRNAs) are small, non-coding, endogenous RNAs involved in
regulating gene expression and protein translation. One single miRNA can target multiple mRNAs and
a single mRNA can be targeted by multiple miRNAs. We consider that miRNA-mRNA clusters with
statistically significant associations can explore potentially regulatory mechanism and, therefore, of
biological interest. In this study, we collected 89 parietal cortex samples from Stanley Medical Research
Institute (SMRI). After quality control, each sample has 420 miRNA, 19,984 mRNA and more than
1,000,000 SNPs screened. We first constructed scale-free networks including both miRNA and mRNA,
and found one module exhibited differential expression between controls and psychotic patients. In this
module, mir-320e acted as one of the hub nodes. Quantitative Trait Locus (QTL) result also indicated
mir-320e was regulated by genetic variants. Another hub gene, PDLIM5, was validated by five miRNA
binding prediction software. To further investigate the causal relationship between PDLIM5 and
mir-320e, we applied Network Edge Orienting (NEO) and found mir-320e regulates PDLIM5. In
summary, we detected one classic regulation pathway: Genotype ->mir.320e -> PDLIM5 -> gene
-> psychosis, which can be partially explain the etiology of psychiatric disease.
Miaoxin Li1, Shu-Jui Hsu1, Pak Chung Sham1
The University of Hong Kong
Individual Abstract Characterizing genetic risk factors of common psychiatric diseases is far from
complete. Using gene and pathway as analysis units, i.e., the gene- and pathway-based association
approaches, is potentially more powerful than the use of individual Single-nucleotide polymorphisms
(SNPs) to identify weak genetic factors of complex diseases. In the present study, we applied the gene
and pathway-based methods we developed (http://statgenpro.psychiatry.hku.hk/limx/kgg/) to re-analyze
association p-values of SNPs from meta-analysis studies on multiple psychiatric diseases (including
autism spectrum disorder, attention deficit-hyperactivity disorder, bipolar disorder, major depressive
disorder, and schizophrenia) released by Psychiatric Genomics Consortium (PGC). This knowledgebased secondary analysis revealed a number of additional interesting genes and pathways significantly
associated with psychiatric disorders. We also found that these genes show interesting co-expression
patterns in brain-tissues and protein-protein interaction networks. While supporting the polygenic model
of common psychiatric diseases, this study demonstrates that introducing genomic knowledge into
conventional statistical genetic analysis is a powerful strategy to characterizing genetic risk genes of
complex diseases.
Chair: Gary Donohoe, NUI Galway
Overall Abstract Details Large-scale genome-wide association studies (GWAS) have been successful
in identifying high confidence genetic susceptibility loci for schizophrenia, with more than 100 genomewide significant signals yielded to date. GWAS have additionally provided evidence for numerous other
schizophrenia susceptibility loci that fall short of this significance threshold, which can be captured
collectively though methods such as polygenic risk scoring. This symposium will feature four talks that
span the gap between risk genotype and the clinical phenotype of schizophrenia. These will cover
molecular, cellular, neuroimaging and neuropsychological investigations of GWAS signals for
schizophrenia at both the individual variant and polygenic level. The first talk, given by Dr Nick Bray
(Institute of Psychiatry, King’s College London), is entitled �Proximal genetic risk mechanisms for
schizophrenia: effects of genome-wide significant risk variants on gene expression in the human brain.’
This talk will introduce fundamental ways in which genetic risk variants can impact on gene function,
before describing largely unpublished work investigating the effects of several GWS schizophrenia risk
variants (e.g. at AS3MT-CNNM2-NT5C2, TSNARE1, VRK2, CACNA1C and CACNB2) on the
expression of these candidate genes in human brain. The second talk, given by Dr Colm O’Dushlaine
(Broad Institute of Harvard and MIT), is entitled �Functional annotation and cellular phenotyping of
disease-implicated variants’. This presentation will describe computational approaches (e.g. functional
annotation and pathway analyses) and cellular approaches (e.g. assay of isogenic cell lines by multielectrode arrays) to elucidate molecular and cellular mechanisms underlying GWS associations with
schizophrenia. The third talk will be given by Prof. Andreas Meyer-Lindenberg (Heidelberg University)
and is entitled 'Neural mechanisms underlying genome wide significant associations with schizophrenia'.
This will describe the application of imaging genetics to delineate neural correlates of both genome-wide
significant common (ZNF804A, CACNA1C, MHC) and rare variants (with an emphasis on CNV). This
work begins to define convergent mechanisms of risk for psychotic disorders, with an emphasis on
medial prefrontal-limbic interactions. The final talk of the symposium will be given by Prof. Andrew
McIntosh (Edinburgh University) and is titled �Cross-disorder and condition–specific cognitive effects of
polygenic risk for psychosis’. This presentation will describe the effects of GWAS identified
schizophrenia risk variants on cognition and cognitive ageing, and how these compare with the cognitive
effects observed in other psychoses, major depression, ADHD and autism.
Nick Bray1
Institute of Psychiatry, King's College London
Individual Abstract Large-scale genome-wide association studies (GWAS) have identified more than
100 high confidence risk loci for schizophrenia, but the molecular mechanisms that mediate these
associations are largely unknown. In the absence of obvious effects on protein structure, many of these
loci are likely to impact on schizophrenia risk by altering the expression of nearby genes. However,
extensive linkage disequilibrium at many of the risk loci and long-range effects of regulatory elements
make it difficult to confidently determine the genes that are affected. This talk will introduce fundamental
ways in which genetic risk variants can impact on gene function and methods for testing association
between risk genotype and gene expression. It will explain how the effects of genetic risk variants on
gene expression can be specific to brain region and developmental stage, before describing largely
unpublished work investigating the effects of several genome-wide significant schizophrenia risk variants
(e.g. at ZNF804A, AS3MT-CNNM2-NT5C2, TSNARE1, VRK2, CACNA1C and CACNB2) on the
expression of these candidate genes in the human brain.
Colm O'Dushlaine1, Jen Pan1, Ralda Nehme1, Sulagna Ghosh1
Broad Institute
Individual Abstract Technology advances and sample size increases over recent years have given rise to
a large number of genome-wide significant associations to psychiatric disorders. These range from
common (SNP) associations, to rare structural or rare single nucleotide variants. We describe
computational and physical experiments that strive to elucidate causative functional mechanisms
underlying regional associations to schizophrenia. Computational experiments extend from detailed
functional annotation and constraint scoring (for example using scoring metrics such as Polyphen and
CADD) to testing for enrichment of association by gene-based or pathway analyses. We summarize
preliminary experiments extending from this, from applying genome editing technologies such CRISPR to
establish isogenic stem cell lines carrying the identified variants, to developing cellular assays for
phenotypic analysis of the derived neurons, such as using multi-electrode arrays for electrophysiological
Andreas Meyer-Lindenberg1
Central Institute of Mental Health
Individual Abstract Recent advances in psychiatric genetics have provided an unprecedented amount of
Information on Genome-wide significant both common and rare variants associated with schizophrenia,
but the understanding of the neural mechanisms through which these genetic risk factors act is
incompletely understood. In this contribution, we review recent work from our laboratory and
collaborators to delineate neural correlates of both genome-wide significant common (ZNF804A,
CACNA1C, MHC) and rare variants (with an emphasis on CNV). This work begins to define convergent
mechanisms of risk for psychotic disorders, with an emphasis on medial prefrontal-limbic interactions.
Interestingly, this circuit is also impacted by environmental risk factors associated with schizophrenia,
such as urban upbringing and migration, suggesting that this circuit may also support, besides
convergence of risk, gene-environment interactions. In initial support of this hypothesis, we report
unpublished work that indicates that methylation in candidate genes associated with early social adversity
further modulate both activation and connectivity in this circuit.
Andrew McIntosh1, Toni Clarke1, Michelle Lupton2, Ian Deary1, David Porteous1, Lynsey Hall1, Caroline
University of Edinburgh, 2 QIMR Berghofer Medical Research Institute
Individual Abstract Introduction: Polygenic risk of schizophrenia is associated with reduced general
cognitive ability and with a greater relative decline between age 11 and 73. The specificity of these
findings to risk variants affecting schizophrenia is now known however. We sought to test whether risk
variants influencing vulnerability to other forms of major psychiatric disorders (depression, bipolar
disorder, major depression, ADHD and Autism) had convergent effects on cognition and cognitive
ageing. Methods: We used data from a newly-available cohort, Generation Scotland (N=21516), the
Lothian Birth Cohorts of 1921 and 1936 (N=1522) and the Brisbane Adolescent Twins Sample (N=902).
Individuals were profiled using the latest public release from the Psychiatric Genomics Consortium
Cross- Disorder GWAS. Polygenic risk profile scores were then tested for their association with cognition
and cognitive ageing. Results: We found specific effects of polygenic risk for schizophrenia on cognition
and cognitive ageing. In contrast, however, we found either no association for other mental disorders – or
effects in the opposing direction (evidence of a positive association between polygenic risk an cognitive
ability). Conclusions: Polygenic risk for major mental disorder influences cognition, general cognitive
ability (IQ) and cognitive ageing. These effects are not uniform across all mental disorders. This suggests
important differences between factors influencing vulnerability for major mental disorder and differing
effects on brain function.
Chair: Sarah Medland, Queensland Institute of Medical Research
Overall Abstract Details In this session, we will provide an overview of recent findings from the
ENIGMA consortium (http://enigma.ini.usc.edu) examining the genetic determinants of subcortical brain
structure (MRI) and white matter (DTI). The ENIGMA Consortium first presented results to the WCPG
last year in Boston and this year we will present the significant progress we have made since. Sarah
Medland will chair the session and will give an overview of the underlying philosophy and long-term
goals of the consortium. The second major phase of the ENIGMA Consortium meta-analysed GWAS of
subcortical brain structures (nucleus accumbens, amygdala, caudate, hippocampus, pallidum, putamen,
thalamus) and intracranial volume. This study of 29,037 participants, in partnership with the CHARGE
consortium, is now complete, yielding some very exciting results. Derrek Hibar will present these
findings including linking significant genetic variants to functional and behavioural changes. Recently,
there has been intense interest in the ENIGMA Consortium specifically with the intent of expanding the
focus of the consortium to include specific psychiatric and neurological diseases as well as working
groups on imaging modalities. Working groups on schizophrenia, bipolar disorder, major depressive
disorder, and ADHD were formed in order to identify the most robust brain-derived endophenotypes with
the largest possible power. Ole Andreassen will present findings from two of the working groups each of
which is the largest ever study of schizophrenia and bipolar disorder. In addition, will provide updates on
new projects examining disease specific effects in the human cortex. Some major projects in the
consortium have focused on imaging modalities. One such effort is the DTI Working Group, which was
formed with the goal of identifying the most heritable and reliably segmented regions of white matter in
the brain. They have produced a standardized protocol for harmonizing the analysis of white matter (DTI)
across sites around the world. The results of this work and the first GWAS findings will be presented by
Dr. Neda Jahanshad. In order to advance one of the long-term goals of the ENIGMA Consortium and of
particular interest to the WCPG audience, we have established a collaboration with the Psychiatric
Genomics Consortium with the goal of identifying common genetic overlap between the genetic
determinants of brain-derived endophenotypes and risk factors for psychiatric disorders. Our efforts and
initial results examining the genetic overlap between the largest-ever GWAS of subcortical structures
from the ENIGMA Consortium and the 2nd round of results from the PGC Schizophrenia group will be
presented by Barbara Franke (Moderator). The implications of these findings for psychiatric research
using imaging endophenotypes will be discussed by Nick Martin (Discussant).
Derrek Hibar1, Sarah Medland2, Paul Thompson1, ENIGMA Consortium
University of Southern California, 2 QIMR Berghofer Medical Research Institute
Individual Abstract Subcortical brain structures are considered the gateway to the human cortex,
linking neuronal interactions to form complex human behaviors. The human subcortex has diverse
functions, but can largely be split into two major systems: the basal ganglia and limbic system. The basal
ganglia is comprised of the putamen, pallidum, accumbens, and caudate nuclei which play a role in
voluntary movement and procedural memory. The limbic system is comprised of the amygdala,
hippocampus, and thalamus which play a role in memory formation and emotional response. A number
of highly heritable psychiatric and neurological disorders are characterized by disrupted connections in
subcortical regions of the brain. In order to identify genetic variants related to structural changes in seven
subcortical brain regions we examined genome-wide genotyping data and related it to structural MRI
brain changes in 29,037 subjects, the largest-ever study of neuroimaging genetics. Here we report six
novel genetic variants, influencing the volumes of the putamen (14q22.3, P = 1.35 x10-32; 8q21.2, P =
5.61x10-14; 20q11.21, P = 1.62x10-12), caudate nucleus (11q14.1, P = 5.15x10-9), and global head size
(7p11.2, P = 4.41x10-10), and replicated two associations previously found to influence hippocampal
volume (12q24.22, P = 5.51 x10-16 and 12q14.3, P = 3.70 x10-10). One of the novel intergenic loci with
highly replicable influence on putamen volume (14q22.3, an eQTL for KTN1) showed evidence of
altering functional activation of the brain’s reward circuitry, neuronal cell shape and dendritic
complexity. Variants influencing the putamen across cohorts clustered near developmental genes known
to regulate apoptosis, axon guidance, and vesicle transport. Identification of these genetic variants
enables us to begin mapping the genetic architecture of brain development and function, a process that
will help elucidate the dysfunctions that lie at the core of neuropsychiatric and neurological disorders.
Ole Andreassen1
University of Oslo
Individual Abstract As part of the disease specific working groups of the ENIGMA Consortium we
have joined forces with researchers around the world in order to search for the best brain-derived
endophenotypes for the study of bipolar disorder and schizophrenia. Our worldwide effort analyzed brain
MRI scans from 1,745 bipolar patients and 2,613 healthy controls (the largest neuroimaging study of
bipolar disorder to date). We found consistent disease effects on subcortical brain volumes (nucleus
accumbens, amygdala, caudate, hippocampus, pallidum, putamen, thalamus, lateral ventricles), and
intracranial volumes (ICV). Many limbic structures were smaller in bipolar patients compared to controls:
mean (of left + right) hippocampus (Cohen’s d = -0.221 ± 0.049; P = 6.62x10-6), thalamus (d = -0.150 ±
0.051; P = 3.21x10-3), and amygdala (d = -0.143 В± 0.043; P = 9.44x10-4). Bipolar patients had larger
lateral ventricles (d = 0.289 В± 0.066; P = 1.29x10-5) than healthy controls. This profile of limbic deficits
suggests that key subcortical structures are subtly but consistently altered in bipolar cohorts worldwide.
Similarly, we analyzed brain MRI scans from 2,028 schizophrenia patients and 2,540 healthy controls,
assessed with standardized methods at 15 centers worldwide. The profile of brain structural abnormalities
in schizophrenia is still not fully understood, despite decades of research using brain scans. We identified
subcortical brain volumes that differentiated patients from controls, and ranked them according to their
effect sizes. Compared to healthy controls, patients with schizophrenia had smaller hippocampus
(Cohen’s d=-0.46), amygdala (Cohen’s d=-0.31), thalamus (Cohen’s d=-0.31), accumbens (Cohen’s
d=-0.25), and intracranial volumes (Cohen’s d=-0.12) and larger pallidum (Cohen’s d=0.21) and lateral
ventricle volumes (Cohen’s d=0.37). Putamen and pallidum volume exacerbations were positively
associated with duration of illness and hippocampal deficits scaled with the proportion of not-medicated
patients. Cooperative analyses of brain imaging data support a consistent profile of subcortical
abnormalities in schizophrenia. These findings provide the foundation for future work using brain-derived
measures as endophenotypes for genetics analysis as well as understanding the differences and common
factors underlying psychiatric disorders.
Neda Jahanshad1, Peter Kochunov2, Paul Thompson3, David Glahn4, DTI Working Group ENIGMA
Imaging Genetics Center, Institute of Neuroimaging and Informatics, Keck School of Medicine of USC,
University of Maryland, 3University of Southern California, 4Yale University
Individual Abstract Introduction: White matter pathways in the human neural network relay
information to and from the functioning gray matter cortex and the subcortical control centers of the
brain. Diffusion tensor imaging (DTI) allows for insight into the microstructural organization and makeup
of these white matter pathways. The DTI working group within the Enhancing Neuro Imaging Genetics
through Meta-Analysis (ENIGMA) Consortium has actively worked on establishing efficient methods for
delineating reliable and heritable phenotypes of interest from fractional anisotropy (FA) images; these
methods have been shown to be efficient and reliable in over 10 cohorts of different ethnicities, ages, and
familial relatedness. Now that heritable phenotypes have been established and prioritized, the working
group is initializing a worldwide, genome-wide association meta-analysis. An initial mega- and metaanalytical investigation into candidate white matter pathway genes reveals the urgent need for such
systematic discovery procedures. Methods: Over 2,200 healthy subjects from five different family-based
cohorts were used to calculate estimate the heritability of phenotypes extracted using the ENIGMA-DTI
protocols. Heritability estimates were jointly analyzed using two different meta-analytical approaches as
well as combining all data into one large family and performing a mega-analysis of heritability. Specific
genetic variants were then tested for reproducibility and replication across populations. Nearly 2900
subjects from 3 overlapping and two additional cohorts were then combined in the largest DTI-imaging
genetics study to date. 16 candidate SNPs previously associated with FA in a single cohort were
examined across all groups in all 15 prioritized regions (phenotypes). Candidates evaluated include
psychosis variants in DISC1, BDNF, COMT, NRG1, NTKR3, ErbB4, and dementia-risk variants in
APOE, HFE, CLU, as well as other variants found to be “top hits” associated with FA through previously
published single-site genome-wide analyses. Genome-wide association results are Results – Children and
adults of different ethnicities show similar patterns of FA heritability in regions extracted using
standardized protocols. Meta-analysis of candidate genes shows inconsistencies in SNP-FA associations
across cohorts. Meta-analyzed GWAS results show SNPs that associate to brain microstructure.
Conclusion: By harmonizing protocols and initially estimating heritability of the FA across the entire
scan, we are able to prioritize the regions on the FA map that show consistent heritability across multiple
cohorts of children and adults from multiple ethnicities. Many SNPs previously found to associate with
FA may not be generalizable across populations using meta-analysis and genome-wide discovery
methods are essential to find pathway associated genetic markers. A rolling GWAS-meta analysis is now
underway with over 10,000 scans.
Tuesday, October 14, 2014
2:15 PM - 4:15 PM
Concurrent Symposia Sessions
Chair: Anders BГёrglum, Aarhus University
Overall Abstract Details The allelic architecture of complex traits is likely to be underlined by a
combination of multiple common and rare variants. Genome-wide association studies (GWAS) and largescale consortia meta-analysis of GWAS have successfully been applied in the search for common variants
affecting the risk of developing psychiatric disorders. However, these studies are designed to examining
only “the common variant” proportion of the genomic landscape of psychiatric disorders. Due to
increased genetic drift during founding and potential bottlenecks, followed by population expansion,
isolated populations may be particularly useful in identifying rare disease variants, that may appear at
higher frequencies and/or within a more clearly distinct haplotype structure compared to outbred
populations. Small isolated populations also typically show reduced phenotypic, genetic and
environmental heterogeneity, thus making them advantageous in studies aiming to map risk variants
involved in complex traits. These characteristics are complemented by elevated levels of linkage
disequilibrium, which facilitates long-range haplotype-phasing and accurate imputation using population
specific reference data. The extended levels of LD may, however, hamper the ability to discover the
specific variants involved. A unique pattern of LD and haplotype frequencies may also obstruct the use of
imputation algorithms based on HapMap CEU samples, and thus require the development of a population
specific reference sample for imputation. Other potential concerns involve whether findings using
isolated populations will generalize to other populations, when the identified risk alleles are private to the
isolated population. Isolated populations have been applied in psychiatric genetics for decades, but they
are, however, often small and the number of available cases are therefore often substantially lower than
what can be obtained in outbred populations. One unresolved question is thus whether the increase in
power due to genetic and environmental homogeneity is large enough to fully compensate for the limited
number of available cases? The recent developments in the application of �next-generation sequencing’
(NGS) are particularly useful for identifying rare variants, especially when applied in samples from
genetically isolated populations. NGS allows the direct examination of both common and rare alleles and
the characteristics of isolated populations may facilitate the identification of rare variants. The speakers
for this session will discuss approaches to using NGS and GWAS data for identification of risk variants
for psychiatric disorders in isolated populations, presenting specific methodological approaches and
illustrating their use in samples from such populations. They will present their latest research on
psychiatric disorders using isolated populations of varying age, ranging from the relatively old isolated
populations of the Ashkenazi Jews and Finland to the more recently
Hreinn Stefansson1
deCODE genetics
Individual Abstract In a small fraction of patients with schizophrenia or autism, alleles of recurrent
copy-number variants (CNVs) in their genomes are probably the strongest factors contributing to the
pathogenesis of the disease. Some of the CNVs clearly alter fecundity and also cognitive function in
control carriers. The high mutation rate of these CNVs compensates for the reduced fecundity and the
CNV alleles are found in comparable frequencies worldwide. Single nucleotide polymorphisms (SNPs),
conferring high-risk of severe psychiatric disorders, are also likely to be under negative selection pressure
explaining why few founder mutations have been uncovered for psychiatric disorders. Genotyped
samples from isolated populations can be long-range-phased which allows for imputations of low
frequency sequence variants. Using this approach in Iceland variants conferring high-risk and protective
for Alzheimer’s disease have been uncovered.
Aarno Palotie1
SISu Project
Individual Abstract Population isolates provide potential short cuts to identify disease associated low
frequency and rare variants. Finland is the largest population isolate in Europe with its current 5.4
million inhabitants. It was founded thousands of years ago by a small number of settlers and has until
recently had very little immigration. In the 17th century the Swedish King demanded a major internal
migration to populate the Eastern and Northern parts of the country. This resulted in a second bottle neck
effect and a large number of small communities that remained isolated for centuries. For unknown
reasons the prevalence of schizophrenia and cognitive impairment follows the internal migration pattern,
both traits being more frequent in the rural, late settlement regions in the North East of the Country. In a
North Eastern Finnish schizophrenia high risk isolate we recently identified a rare 250kb deletion that
deletes the TOP3B gene and found the deletion to be associated with an increased risk to schizophrenia
and cognitive impairment. Characteristic for a large isolate, we also could identify four cases who were
homozygotes for the TOP3B deletion; all of them had schizophrenia and/or cognitive impairment. The
topoisomerase TOP3B forms a complex with the FMRP protein and thus connects the finding to the
FragileX pathway (Stoll et al 2013). This work demonstrated the potential to identify disease associated
low frequency variants enriched in an isolate. To use the potential of the Finnish isolate to identify low
frequency coding SNPs, relevant for disease traits, the collaborative SISu Project (Sequencing Initiative
Suomi (Suomi is Finland in Finnish) integrates all large scale sequence data produces from Finnish
samples. To understand the overall landscape of coding variants we compared 3000 Finnish exomes from
the SISu project with variants from 3000 non-Finnish European exomes. We could demonstrate that the
genetic bottleneck has resulted depletion of private variants and in a shift of an increased proportion of
population specific variants with a frequency of 0.5-5%, specifically with an increased proportion of loss
of function variants (Lim et al submitted). Large, existing epidemiological cohorts that can be linked to
nationwide health registers enable phenome mining strategies to study potential gene variant associations
to a large number of phenotypes, including longitudinal health data. To test such strategies we analyze 80
loss of function (LoF) variants enriched in Finland in 35 000 Finns and linked these variants to a number
of biochemical parameters and disease endpoints using the National Health Registers. Among these LoF
variants we identified several associations to medically relevant, mostly cardiometabolic, traits. These
findings suggest that similar strategies could be used to study neuropsychatric traits.
Todd Lencz1, Semanti Mukherjee1, Shai Carmi2, Anil Malhotra1, Itsik Pe'er2, Ariel Darvasi3
The Zucker Hillside Hospital, 2Columbia University, 3Hebrew University
Individual Abstract Increasing attention in psychiatric genetics has been paid to the identification of
rare (<1%) variants with relatively high penetrance. Microarrays have provided support for the rare
variant hypothesis, with the identification and replication of several high penetrance copy number
variants associated with schizophrenia, autism, and other neuropsychiatric disorders. The recent advent of
affordable next-generation sequencing provides the opportunity to identify a broader range of rare
variants, but interpretation is hindered by locus and allelic heterogeneity in disease susceptibility, as well
as the large number of mutations observed in healthy genomes. To reduce the “needle-in-the-haystack”
problem, we examined cases and controls in a homogeneous founder cohort: the Ashkenazi Jewish (AJ)
population. Using genomewide SNP data, we applied an identity-by-descent (IBD) approach for disease
mapping in a genetically homogenous cohort of 904 cases and 1640 controls from the Ashkenazi Jewish
population. Shared chromosomal segments (IBD segments) with length greater than 10KB were
identified using the GERMLINE algorithm and clustered using the DASH algorithm. Consistent with our
hypothesis, we observed a greater degree of IBD sharing of rare haplotypes in cases compared to
controls; under a “clan genomics” model, these rare haplotypes are likely to harbor disease susceptibility
variant(s). Whole genome sequencing is currently being performed in n=240 cases of this cohort,
compared with n=300 sequenced AJ controls. Candidate mutations emerging from the sequencing
analysis will then genotyped in additional subjects from the full cohort. Analyses are ongoing, and
detailed results will be presented at the meeting. Because the range of rare of variation in the human
genome is immense, strategies for increasing signal-to-noise are required to more rapidly identify
functional variants associated with psychiatric disease.
Francesco Lescai1, Thomas D. Als1, Mette Nyegaard1, Andrew McQuillin2, Ditte Demontis1, Alessia
Fiorentino2, Niamh O’Brien2, Alexandra Jarra2, Jakob Grove1, Manuel Mattheisen1, Gudrid
Andorsdottir3, Marjun BiskopstГё3, August G. Wang4, Ole Mors5, Jun Wang6, Anders BГёrglum1
Aarhus University, 2University College London, 3Genetic Biobank of the Faroes, Faroe Islands, 4Mental
Health Centre Amager, Denmark, 5Aarhus University Hospital, 6BGI - Beijing Genomics Institute, China
Individual Abstract Isolated populations represent an advantage to identify rare disease variants that
may appear at higher frequencies compared to outbred populations. In this work, we use the Faroese
population to test this hypothesis, and combine low-depth (6X) whole-genome (WGS) and high-depth
(35X) whole exome (WES) sequencing approaches to describe genomic variation, de-novo mutations and
perform association mapping in patients with schizophrenia (SZ) and bipolar disorder (BP). Our sample
consists of 106 SZ cases, 28 BP and 214 controls (344 total): together with unrelated individuals, it
includes 54 complete trios. A total of 17,345,307 and 259,904 variants have been called in WGS and
WES respectively. We discovered 9,130 de-novo mutations in WGS and 417 in WES. Clear differences
emerge between WGS and WES in identifying different types of variants. A specificity of this work is the
combination of WGS and WES in the discovery of de-novo mutations: both approaches discover the
same number of de-novo loss-of-function mutations but high-depth WES is clearly more powerful in
calling coding variants, mostly because of depth filtering criteria. In the association mapping we analysed
variants with increased frequency in the Faroese population and identified a number of significant loci,
which are currently being replicated in 3,300 BP and controls from UCL. The results of this analysis, the
concordance between WES and WGS, and their perspectives will be presented and discussed. Whole
Exome (35X) Whole Genome (6X) count percent count percent All Variants Loss of Function 3,140
1.21% 5,753 0.03% Missense 65,785 25.31% 80,087 0.46% Synonymous 41,904 16.12% 42,989 0.25%
Non coding 149,075 57.36% 17,216,478 99.26% Total 259,904 17,345,307 De Novo Variants Loss of
Function 6 1.44% 6 0.07% Missense 89 21.34% 12 0.13% Synonymous 32 7.67% 6 0.07% Non-coding
290 69.54% 9,106 99.74% Total 417 9,130.
Chair: Cynthia Bulik, University of California at Chapel Hill
Overall Abstract Details Eating disorders (ED), including anorexia nervosa, bulimia nervosa, and binge
eating disorder are associated with high morbidity and mortality. Anorexia nervosa is an enigmatic illness
marked by extreme negative energy balance and maintenance of biologically implausible low body
weights. Only 50% of patients with ED improve with treatment, highlighting the need for new discoveries
and treatment targets. Genetic research into all eating disorders is progressing, with new consortia and
novel large studies yielding insights into the pathogenesis of ED. This symposium will focus on new
findings from varied methodological approaches aimed at identifying genetic factors in ED. We will
present and discuss emerging evidence from across Europe and the US on how the genome, the
epigenome, and the co-action of genes and environment can elucidate risk ED, course, and outcome. The
first presentation (Prof C. Bulik) will focus on a mega-analysis performed as part of the Psychiatric
Genomics Consortium, combining samples from: the Genetic Consortium for Anorexia Nervosa and the
Wellcome Trust Case Control Consortium 3, and the Children’s Hospital of Philadelphia and Price
Foundation samples. The second presentation (Dr Nadia Micali) will focus on the role of geneenvironment interactions (focusing on serotonin, dopamine genes and obesity-related genes and their
interaction with life events and other relevant environmental factors) in increasing the risk for bulimic
behaviours (binge eating and purging) and anorexic behaviours (food restriction and excessive exercise
in the context of low weight) in a large UK population-based cohort of 8,000 adolescents and young
adults. The third presentation (Dr Y Guo) will focus on applying machine learning risk prediction to
anorexia nervosa using genome-wide genotyping data from 3940 cases and 4179 controls of European
The fourth presentation (Prof H. Frieling) will highlight findings from a large genome-wide
methylation study and a candidate gene methylation study in German participants with eating
disorders. Dr G Breen will moderate the symposium, discuss the findings presented, and contextualize
them relative to other psychiatric disorders. He will also discuss translational implications of the
findings highlighting next steps in animal models and how and whether the findings are currently
relevant to clinicians, families, and patients in understanding these pernicious illnesses.
Cynthia Bulik1, Genetic Consortium for Anorexia Nervosa, Wellcome Trust Case Control Consortium,
Anorexia Nervosa Working Group of the Psychiatric Genomics Consortium
University of California at Chapel Hill
Individual Abstract Anorexia nervosa (AN) is a complex and heritable eating disorder characterized by
the maintenance of dangerously low body weight. Two subtypes, restricting—marked by decreased
energy consumption and increased energy expenditure—and binge-eating/purging—marked by the
presence of both low weight and binge eating or purging exist. Both represent extremes of dysregulated
appetite and
weight. Two small genome-wide association studies (GWAS) of AN have been conducted, neither of
which has yielded genome-wide significant findings, as would be expected by sample size. A mega
analysis is underway to combine these two studies to be followed by cross-disorder analyses with
other phenotypes present in the Psychiatric Genomics Consortium (i.e., obsessive-compulsive
disorder, major depression, alcohol and drug dependence, major depression, autism, attention-deficit
hyperactivity disorder, schizophrenia, bipolar disorder, and post-traumatic stress disorder). With the
increasing availability of large GWAS genotyped samples of many psychiatric disorders, we are well
positioned to apply molecular genetic approaches to explain clinical comorbidity. Such analyses,
including genome- wide complex trait analysis (GCTA) enable the calculation of bivariate SNP
heritabilities and determination of genetic correlations between disorders. This talk will set the stage
for the symposium and present up-to-date findings and analyses relevant to the genetics of eating
Nadia Micali1, Marta Cros Bou2, Janet Treasure3, Emily Simonoff3
UCL Institute of Child Health, 2Harvard School of Public Health, 3Institute of Psychiatry
Individual Abstract Eating disorders have a peak of onset in adolescence. Evidence suggests that
eating disorders result from an interplay between genes and environment. The majority of studies so
far, however, have focused on anorexia nervosa and bulimia nervosa. Moreover, few studies have
investigated Gene x environment (GxE) interactions in population-based studies, and fewer have
focused on specific eating disorder behaviours (possible better phenotypes than syndromes). Case
control studies have highlighted an association between genes in the serotonin and dopamine system,
as well as other genes in the appetitive and weight system and eating disorders. G x E interactions
have mainly been studied in relation to bulimia nervosa. The aim of this study was to investigate
previously identified GxE interactions for anorexia nervosa and bulimia nervosa such as: childhood
abuse and life events in interaction with a polymorphism of the serotonin transporter (5-HTTLPR) for
bulimia nervosa behaviours (bingeing and purging); parenting style in interaction with 5-HTTLPR
polymorphisms for anorexia nervosa behaviours (drive for thinness, food restriction in the presence of
low weight); life events and stress in interaction with a SNP in Brain-derived neurotrophic factor
(BDNF) for anorexia nervosa behaviours. We prospectively collected data at three timepoints in
adolescence (14, 16, 18 years) on 7,000 adolescents from the Avon Longitudinal Study of parents and
Children (ALSPAC), a population-based study based in the UK to derive eating disorder behaviours.
Genotype data was obtained using the Illumina Hapmap 550 quad chip on 6,500 adolescents.
Population stratification will be assessed and hidden population stratification will be controlled for.
Data will be analysed using logistic and linear regression models. We will present data on the full
sample and stratified analyses by gender. All analyses will be adjusted for age at assessment and Body
Mass index (for analyses relating to bulimia nervosa behaviours).
Yiran Guo1, Zhi Wei2, Brendan Keating1, GCAN, WTCCC, Hakon Hakonarson1
Children's Hospital of Philadelphia, 2New Jersey Institute of Technology
Individual Abstract Machine learning disease risk prediction is measured by area under ROC (receiver
operating characteristic) curve (AUC), a value between 0.5 and 1 for assessing how well the model can
distinguish cases versus controls, with the higher number indicating better discriminative power. This
method has been used to predict risk for complex human diseases in which genetics plays a part of the
etiology like type 1 diabetes (T1D) and inflammatory bowel disease (IBD), with AUC of around 0.85.
We applied the method to evaluate the risk for Anorexia nervosa, a psychiatric and eating disorder to
which genetics also contribute susceptibility, using genome-wide genotyping data from 3940 cases and
4179 controls of European ancestry. The resulting AUC is 84.7% which indicates comparable ability to
predict disease risk using the genetics data and machine learning method.
Helge Frieling1, Vanessa Buchholz2, Martina de Zwaan2, Stefan Ehrlich3
University of Erlangen-Nurenberg, 2Hannover Medical School, 3Gustav Carus University Dresden
Individual Abstract Eating disorders and especially anorexia nervosa are believed to be highly heritable
with heritability estimates ranging up to 70%. Molecular genetic studies so far have failed to find
convincing risk genes. Twin studies have revealed high concordance rates between persons sharing the
same genotype, still many monozygotic twin pairs discordant for anorexia nervosa exist. This
discordance together with the high heritability estimates poses the question, why some persons with a
genetic risk profile are able to stay healthy. The study of epigenetics, investigating those mechanisms that
control the activity of certain genes by turning them on or off, may help to understand this paradox.
Environmental influences change the epigenetic code of gene regulation throughout the life-span of the
organism, starting even before the implantation of the fertilized egg. Pre-, peri and postnatal adversities
have been shown to program certain vulnerabilites towards an eating disorder and life-style factors later
in life can function as initiating and maintaining factors of the actual disease. Epigenetic mechanism like
DNA methylation and histone modifications can be regarded as the molecular biological foundation of
these processes. In contrast to genetics risk factors, epigenetics are dynamic and prone to modification by
all kind of interventions including changes in life-style and psychotherapy. The talk will focus on how
epigenetic mechanisms are involved in the etiology of eating disorders and show, how these mechanisms
contribute to successful remission of the disorder, thereby showing that genes are not fate.
Chair: Francis McMahon, NIH/NIMH
Overall Abstract Details Genetic testing was once a distant prospect in clinical psychiatry, but is now
increasingly regarded by clinicians and the public as a potential source of information that could help
guide diagnosis, treatment, and family counselling. However, research on the clinical use of genetic
testing in psychiatry is still quite limited, and the issue remains clouded in misinformation, regulatory
flux, and ethical concerns. Recently, the International Society of Psychiatric Genetics (ISPG) released a
consensus statement on the clinical use of genetic testing in psychiatry aimed at providing guidance for
clinicians and the general public. To accomplish this, the ISPG charged a taskforce with the goal of
updating and expanding the Society's previous statement. The taskforce considered a broad range of
scientific, clinical, and ethical issues. In this workshop, members of the taskforce will review and discuss
the evidence and considerations that were used in formulating the latest consensus statement. Several
controversial topics will be highlighted and discussed. Audience participation is highly encouraged in
what promises to be a lively and informative discussion of one of the big issues facing our field today.
Elliot Gershon1, Ney Alliey-Rodriguez1
University of Chicago
Individual Abstract Copy Number Variants (CNVs; chromosomal microdeletions and
microduplications) are the most potent known individual risk factors for Schizophrenia, Autism Spectrum
Disorder, Bipolar disorder, and Intellectual Disability. Rare CNVs at specific locations greatly increase
the likelihood that an individual will have one of these disorders. Another risk is presented by de novo
CNVs, which are mutations found in an individual but not the parents. In the aggregate, these are
common events, occurring in 1% of normals and in 4% to 7% of patients with BD, SZ, or ASD. De novo
CNVs are independent of family history. These findings have profound implications for genetic
counseling, although for some of these findings further replication will be needed before introduction into
clinical counseling practice. The broad spectrum of phenotypes of CNV associations leads to relatively
large cumulative risks of any illness. The risk of BD, SZ, or ASD is 14% in the case of a de novo CNV.
The risk of any of these disorders for a carrier of the rare 22q11 microdeletion CNV is 82%.
Traditionally, people who ask for risk counseling are concerned about a particular illness because of a
family history. As genome-wide scans become more widely used, people will ask for counseling on any
possible illness based on unanticipated genetic test results. With such potent and pluripotent risk factors,
counseling issues can arise: a) abortion or non-implantation of embryos, b) stigmatization of carriers, of
their families, and of their communities, c) family members’ rights to genetic information, d) population
screening, and e) psychological and interpersonal problems arising from results of counseling, which
include an individual’s coming to terms with undesired test results, and conflict within families over
communication of results and over blame for genetic risk.
Daniel Mueller1
University of Toronto
Individual Abstract While few studies have investigated the association between non-pharmacological
treatment forms in psychiatry (e.g., ECT, CBT), there is a growing list of genetic markers associated with
effectiveness and adverse events of various drugs. In some situations, pharmacogenetic markers can
supplement clinical information to help guide treatment decisions for psychiatric disorders, reducing the
risk of treatment failure and serious adverse events. Notably, some pharmacogenetic markers were shown
to have substantial effects sizes and clinically relevant odd ratios, and in particular for drug induced side
effects. For example, in patients of Asian ancestry who receive carbamazepine, the HLA-B*1502 marker
substantially increases risk of serious skin disorders (Stevens Johnson Syndrome and toxic epidermal
necrolysis). Another example would be antipsychotic-induced weight gain, where one marker near the
MC4R gene has shown to account for approximately half of the variance of weight gain. As for serum
drug levels, a similar relationship can be seen for some CYP450 enzymes (e.g., CYP2D6, CYP2C19).
These enzymes are highly involved in metabolism of drugs, including antidepressants and antipsychotics.
Variation in the genes that encode these enzymes can lead to differences in drug metabolism that can be
predicted by genetic markers. Individuals with genetic markers of poor or rapid metabolism may be at
higher risk for non-response, adverse events, or drug-drug interactions. A third pharmacogenetic
phenotype could be delineated which would include gene products targeted specifically for symptoms
reduction, such as the serotonin transporter through antidepressants. Although most studies have
consistently shown an association between the short allele of the serotonin transporter and antidepressant
response, the relationship is less clear and effects are typically smaller in this category. Other gene-drug
pairings are under active investigation. In view of these findings, expert panels have started to publish
guidelines such as for use of CYP450 testing in psychiatry. In a recent statement, the ISPG board
declared to generally concur with some of these guidelines, which do not recommend genetic testing on a
global level, but provide guidance if genotype data are already available. As with most complex
phenotypes, other factors also influence drug outcome (such as diet, use of other medications, or
treatment resistance) which need to be taken into account and studied further. Randomized, double-blind
clinical trials are needed to establish the clinical utility of genetic testing in psychiatric drug treatment.
ISPG recommends clinicians follow good medical practice and stay current on changes to drug labeling
and adverse event reports. One useful (but not necessarily exhaustive) list of pharmacogenetic tests is
maintained by the US Food and Drug Administration.
Marcella Rietschel1
Central Institute of Mental Health
Individual Abstract Incidental findings have been defined as unexpected observations of potential
clinical significance. Genetic technologies permitting genome-wide screens may generate incidental
findings of potential importance for medical conditions unrelated to the clinical complaint for which these
tests were originally performed. Given that some of these events occur with a substantial frequency in the
general population, it may be more appropriate to describe them as secondary rather than incidental.
Irrespective of the terminology applied, such secondary findings may highlight a preventable illness or
one that could benefit from early intervention. Some authorities, such as the American College of
Medical Genetics (ACMG) recommend that clinicians report some secondary findings back to the
individual patients. This recommendation is not generally accepted and while the ISPG in its statement
“Genetic Testing and Psychiatric Disorders” concurs with the ACMG recommendation regarding
reporting of actionable secondary findings to the referring clinician, it also states that a decision to inform
a patient about such finding(s) must weigh the seriousness of the implicated disease, the potential
medical consequences of nondisclosure, the patient’s stated wish to be informed about
secondary/incidental findings (ideally established during pre-test counselling), the patient’s ability to
rationally appreciate the prognostic implications of such finding(s) and participate in any preventive or
therapeutic interventions that might be recommended, and the potential negative impact of disclosure on
the patient’s psychological condition and quality of life. To fulfill these recommendations adequate
expertise in counseling and time resources are needed. This text is largely based on the wording of the
ISPG Statement “Genetic Testing and Psychiatric Disorders”
Jehannine Austin1
University of British Columbia
Individual Abstract We now have a substantial and growing list of genetic variations - including both
single nucleotide polymorphisms and copy number variations - that we can confidently identify as
contributing to the etiology of conditions like schizophrenia, bipolar disorder and depression. This
progress is accompanied by a need for urgent attention to the question of how to apply this knowledge
clinically in such a way as to promote the best possible outcomes for those who live with psychiatric
disorders and their families. In this presentation, the ethical and psychological implications of applying
our developing knowledge of the etiology of psychiatric disorders in the clinical setting will be discussed.
For example, we will explore the psychological importance for people with psychiatric illness and their
families of understanding cause of illness, and in particular, the psychological ramifications of
understanding that there is a genetic contribution to these conditions. The process of risk communication
in relation to psychiatric disorders and its attendant ethical issues will be discussed, and the influence that
applying new genetic knowledge clinically may have on various facets of psychiatric illness-related
stigma. The aim is to open discussion of questions of exactly *how* to apply our developing knowledge
clinically, in order to promote the best possible outcomes for patients and families.
Chair: John Kelsoe, University of California San Diego
Overall Abstract Details Genome sequencing in families is a powerful strategy for gene identification.
Genome sequencing generates a large number of variants that must be filtered on a variety of criteria in
order to identify those most likely associated with disease. Family segregation is a powerful filter in
identifying variants most related to disease. Early work in this area has revealed some intriguing results.
Most families segregate for multiple likely functional susceptibility variants, and many of these variants
are regulatory in nature. Seth Ament will present results from the sequencing of 36 families that supports
these two ideas. Maja Bucan will discuss their efforts in genome sequencing portions of very large Amish
pedigrees. These data suggest that even such large pedigrees from population isolates may still harbor
many susceptibility variants. Bill Byerley will present mutations found in genome sequencing of
extended pedigrees from isolates in Palau and Costa Rica. Lastly, Tadafumi Kato will discuss their work
in exome sequencing of Japanese trios with bipolar disorder.
Seth Ament1, Gustavo Glusman1, Szabolcs Szelinger2, Katherine Rouleau1, Denise Mauldin1, Tatyana
Shekhtman3, Richard Gelinas1, Nathan Price1, Howard Edenberg4, Francis McMahon5, David Craig3,
Leroy Hood1, John Kelsoe4, Jared Roach1
Institute for Systems Biology, 2Translational Genomics Institute, 3University of California at San Diego,
Indiana University School of Medicine, 5National Institute of Mental Health
Individual Abstract We sequenced the genomes of 200 individuals from 40 multiply-affected pedigrees
with bipolar disorder to identify its genetic causes. We show that a minority of these pedigrees can be
explained by kilobase-scale deletions that segregate perfectly with disease and which are predicted to
have large effects on disease risk. Bipolar disorder in the remaining pedigrees is better explained by the
combined effects of multiple small- to moderate-effect risk variants. Intriguingly, both the large-effect
and small-effect risk variants were enriched in the regulatory (non-coding) regions around genes with
neuronal functions. In contrast to recent findings in schizophrenia and autism, we did not find an
enrichment of exonic (coding) variants in these neuronal genes. Based on these results, we selected 30
genes for targeted sequencing in an additional 4000 bipolar disorder cases and 2000 controls. We
confirmed an association in both European Americans and African Americans between bipolar disorder
and rare, non-coding variants near several voltage-gated calcium channels. In addition, we confirmed an
association in European Americans between bipolar disorder and rare variants in MTRNR2L2, which
encodes the neuropeptide humanin. We demonstrate a novel effect of humanin on calcium flux in
neuronal cell culture. Our results support the idea that risk variants for bipolar disorder perturb the
regulatory sequences of genes involved in neuronal excitability.
Maja Bucan1, Benjamin Georgi2, Rachel Kember2, David Craig3, Christopher Brown2, Janice A.
Egeland4, Steven M. Paul5, Maja Bucan2
University of Pennsylvania , 2Department of Genetics, Perelman School of Medicine, University of
Pennsylvania, 3The Translational Genomics Research Institute, 4Department of Psychiatry and Behavioral
Sciences, University of Miami Miller School of Medicine, 5Departments of Neuroscience, Pharmacology
and Psychiatry, Weill Cornell Medical College
Individual Abstract We conducted a comprehensive genomic analysis of bipolar disorder in a large Old
Order Amish pedigree. High-density SNP-array genotypes of 388 subjects were combined with whole
genome sequence data for 80 family members, comprising 30 parent-child trios. This study design
permitted evaluation of candidate variants within the context of haplotype structure by a) resolving the
phase in sequenced parent-child trios and b) by imputation of variants into multiple unsequenced siblings.
Non-parametric and parametric linkage analysis of the entire pedigree as well as on smaller clusters of
families identified nominally significant linkage peaks. We report dozens of predicted deleterious genetic
variants under each linkage peak, in addition to moderately frequent (in the Amish, but rare in 1000
Genomes) variants at the published bipolar and schizophrenia GWAS loci. In addition, we used high
density SNP-array data to address the role of copy-number variation (CNV). Although we find no
evidence for an increased burden of CNVs in BP individuals, we report a trend towards a higher burden
of CNVs in known Mendelian disease loci in bipolar individuals (p=0.06). Dissection of exonic and
regulatory variants in genes identified additional credible candidate genes for functional studies and
replication in population-based cohorts. The striking haplotype and locus heterogeneity we observed
suggest that mechanistic studies on a large number of genes will be necessary to increase our knowledge
about the etiology of bipolar illness and its relationship to other disorders.
Tadafumi Kato1, Nana Matoba2, Muneko Kataoka3, Kumiko Fujii4, Tadafumi Kato2
RIKEN Brain Science Institute , 2Lab for Molecular Dynamics of Mental Disorders, RIKEN Brain
Science Institute, 3Lab for Molecular Dynamics of Mental Disorders, RIKEN Brain Science Institute;
Department of Child Neuropsychiatry, Graduate School of Medicine, The University of Tokyo, 4
Department of Psychiatry, Dokkyo University School of Medicine
Individual Abstract Bipolar disorder is one of major mental disorders characterized by recurrent manic
and depressive episodes. Twin studies showed that heritability of bipolar disorder is around 85%. GWAS
identified a number of common SNPs associated with bipolar disorder, but the effect of each SNP is
weak. We hypothesize that combination of multiple rare damaging variants contribute to bipolar disorder
and the variants detected in patients are enriched in genes within specific functional categories. To test
this hypothesis, we performed exome sequencing in 36 parents-proband trio families of bipolar disorder.
No significant difference in the number of damaging mutations was found between the variants
transmitted from the parents to the proband and those not transmitted to the proband. When gene
ontology enrichment analysis was applied, several functional categories were found to be enriched in
transmitted rare damaging variants than in the un-transmitted variants. We are analyzing independent
bipolar trios as well as control trios to test the reproducibility and specificity of this finding.
Chair: Andrew McIntosh, University of Edinburgh
Overall Abstract Details Endophenotypes are quantitative traits associated with mental disorders that
show co-segregation with clinical disorder within multiply affected families and demonstrate a high
genetic correlation with the target condition irrespective of clinical state. To date, many candidate
endophenotypes have been proposed and in many cases these traits have a polygenic architecture similar
to the clinical disorders with which they are associated. In this symposium we will review the value of
endophenotypes for psychosis, bipolar disorder and depression and present methods for adjudicating their
value. We will also present data that allows endophenotypes to be assessed on the basis of epistasis as
well as an additive genetic component. We will also present evidence demonstrating their value for gene
finding studies using linkage and family-based association studies.
Andrew McIntosh1, David Porteous1, Ian Deary1, Toni Clarke1, Lynsey Hall1, Pippa Thomson1, Caroline
Hayward1, Maria Fernandez1, Chris Haley1, Donald MacIntyre2
University of Edinburgh, 2NHS Lothian
Individual Abstract Introduction Major Depression (MDD) is clinically and causally heterogeneous and
has so far resisted attempts to reveal its underling genetic architecture through genome wide association
studies (GWAS). The use of genetically-correlated quantitative traits represents an alternative means of
identifying risk variants for the condition. Methods We genotyped a large, recently available population
based family cohort 'Generation Scotland' (N=21516). Fourteen thousand individuals were genotypes
using the Human OmniExpress SNP Array with exome variants. Individuals completed a cognitive test
battery, the General Health Questionnaire (GHQ), the Eysenck Personality Questionnaire (EPQ) and
measures of schizotypy and bipolar spectrum disorders. Ranking on the basis of genetic correlation with
MDD, we prioritized variables and used them for GWAS. Results Three of our quantitative traits had
genetic correlations with MDD of >0.3. We performed multivariate GWAS (mvGWAS) using the vector
of these traits. In addition, we performed a principal components analysis of the genetic and phenotypic
correlations between these traits and used the first unrotated principal component for univariate GWAS.
Whereas we found little evidence for significant genome-wide association with the binary clinical trait,
we found a novel association with our derived quantitative traits. A common locus was revealed by both
methodological approaches (mvGWAS and 1st-PCA analysis). Conclusions Our findings demonstrate
that quantification of MDD coupled to studies which prioritize candidate endophenotypes may hold
promise as a means of identifying risk variants.
David Glahn1
Yale & Institute of Living
Individual Abstract The genetic architecture of schizophrenia is complex, involving multiple common
and rare mutations within specific gene pathways. To make progress, it is necessary to determine how
risk variants impact the multifaceted behavioral symptoms that define the illness. Yet, traversing between
genotype and phenotype is difficult, even for simple Mendelian disorders. Endophenotypes can help to
characterize disruptions in gene networks on quantitative traits closely aligned to schizophrenia.
Unfortunately, relatively few schizophrenia endophenotypes are genetically correlated with disease
liability. We present a novel method for discovering endophenotypes in unselected extended pedigrees.
Specifically, we search for neurocognitive and neuroanatomic endophenotypes for schizophrenia in large
unselected multigenerational pedigrees using a novel approach to the estimation of the endophenotypic
ranking value that is closely related to the genetic correlation between endophenotype and disease. Using
a coefficient of relationship approach, a fixed effect test within a variance component analysis was
performed on neurocognitive and cortical surface area traits in 1,606 Mexican-American individuals from
large, randomly ascertained extended pedigrees who participate in the “Genetics of Brain Structure and
Function” study. Despite having sampled just 6 individuals with schizophrenia, our sample provided 233
individuals at various levels of genetic risk for the disorder. We identified three neurocognitive measures
(digit-symbol substitution, facial memory, and emotion recognition) and six medial temporal and
prefrontal cortical surfaces associated with liability for schizophrenia. With our novel analytic approach
one can discover and rank endophenotypes for schizophrenia, or any heritable disease, in randomly
ascertained pedigrees.
Kristin Nicodemus1
The University of Edinburgh
Individual Abstract Although several studies have successfully shown a polygenic component explains
a small but significant amount of variation in endophenotypes for psychiatric disorders such as cognition
and clinical symptomology, epistasis may also play an important role in the underlying genomic
architecture of these complex traits. Recent work1 that has examined variation in cognitive
endophenotypes in psychosis cases explained by the schizophrenia candidate gene ZNF804A and its
putative pathway (defined as differentially expressed genes after ZNF804A knockdown). Of particular
importance was assessing the relative contribution of the polygenic score versus epistasis in explaining
variation in these cognitve endophenotypes. Psychosis patients (N = 424) were assessed in cognitive
function impaired in schizophrenia (e.g., IQ, memory, attention). The polygenic score was created Using
the Psychiatric GWAS Consortium schizophrenia case-control study results within the genes in the
ZNF804A pathway. Increased polygenic scores were associated with poorer performance in psychosis
patients on IQ, memory and social cognition, and the amount of variation explained (R2) by the
polygenic score on these endophenotypes ranged between 1-3%, which is similar to that observed in
other studies.
Using a newly-developed statistical model that simultaneously models both polygenic and epistatic
components, epistasis in the ZNF804A pathway was found to explain 2-3 times more variability in
working memory in psychosis cases than the polygenic score, even after controlling for the contribution
of the polygenic score in the model. This increase was able to be replicated in two independent samples,
including a “narrow psychosis” (p = 0.016) and “broad psychosis” set (p = 0.036) as well as combined
psychosis (p = 0.0012). This method is currently being applied to variation in cognitive endophenotypes
explained by the Fragile X Mental Retardation Protein (FMRP) pathway, which has been associated with
several psychiatric disorders including schizophrenia, major depressive disorder, bipolar disorder and
autism. 1. Nicodemus KK, et al. Epistasis increases the amount of variability in working memory
performance explained by polygenic scores in the ZNF804A pathway. JAMA Psychiatry [in press]
Carrie Bearden1, Nelson Freimer 1, Scott Fears 1, Susan Service 1, Chiara Sabatti 2, Rita Cantor1, Carlos
Lopez-Jaramillo3, Gabriel Macaya3, Victor Reus3, David Glahn4, Julio Molina , Javier Escobar, Juan
David Palacio
University of California, Los Angeles, 2Standford University, 3Universidad de Antioquia, 4Yale &
Institute of Living
Individual Abstract Although genome-wide association studies have now identified the first replicated
loci contributing to risk for bipolar disorder (BP), the small relative risk attributed to these loci may
reflect the significant heterogeneity of the disorder, at both the genetic and phenotypic level. We have
now collected the most extensive set of putative BP component phenotypes yet assessed within any study
sample, in multi-generational pedigrees enriched for severe BP-I disorder. Based on strong heritability
and association with disease in adult members of these pedigrees, we have been able to prioritize a set of
candidate quantitative traits for genetic mapping and further investigation. In particular, neuroimaging
phenotypes [i.e., prefrontal and temporal cortical thickness, volume of medial temporal structures, and
microstructural integrity of the corpus callosum, as measured with diffusion tensor imaging (DTI)], as
well as circadian and sleep phenotypes, were prioritized for further analysis based on these criteria. We
have now extended our investigation to adolescent offspring of the adult members of the pedigrees, in
order to examine developmental expression of these phenotypes. Preliminary findings reveal high rates of
anxiety and inattentive disorders, and impairments in recognition of facial emotional expression and in
frontally- mediated cognitive processes (i.e. inhibitory control) in youth at genetic risk for BP. Greater
stability of daily rhythms, as measured with actigraphy, was associated with lower self-reported daily
stress. Additionally, given significant linkage findings (LOD =5.1) for amygdala volume in adult
pedigree members, we are now investigating structural integrity and functional connectivity of the
amygdala as a candidate biomarker for bipolar risk in adolescents. Longitudinal investigation of a
genetically informative cohort, in which verified risk loci for both the clinical diagnosis of BP and BPassociated endophenotypes have been identified, offers an unprecedented opportunity for connecting risk
genes to brain development and emergent psychopathology in adolescence.
4:30 PM - 6:00 PM
Concurrent Oral Sessions
Steven McCarroll1, Evan Macosko2, Anindita Basu3, James Nemesh4, Melissa Goldman5, Alex Shalek6,
David Weitz7, Aviv Regev6
Stanley Center for Psychiatric Research, Broad Institute of MIT and Harvard, 2Harvard Medical School;
Stanley Center for Psychiatric Research, Broad Institute, 3Harvard University; Broad Institute, 4Stanley
Center for Psychiatric Research, Broad Institute, 5Harvard Medical School, 6Broad Institute, 7Harvard
Background An increasingly critical research direction is to leverage genetic leads to reach insights
about the pathophysiology involved in brain disorders. To do this, we need methods to systematically
relate genes to the specific cell populations in which they are expressed, and to identify altered cellular
states in those cells. The brain’s enormous cellular complexity, which has yet even to be satisfactorily
catalogued, poses immense challenges to this goal.
Methods To address this need, we have been developing a technology called DropSeq, enabling
simultaneous analysis of thousands of single-cell transcriptomes. DropSeq starts with suspended cells,
isolates individual cells in nanoliter-sized aqueous compartments within oil-aqueous reverse emulsions,
massively barcodes these tiny compartments, and generates high-quality 3'-end single-cell cDNA libraries
from thousands of cells, in a process that takes about 12 hours. Library preparation and sequencing occur
in a single bulk reaction. We use “cellular barcodes” to track the cell-of-origin of each transcript;
“molecular barcodes” to distinguish the distinct mRNA molecules from the same cell; and the rest of a
sequence read to identify the gene from which each mRNA transcript arose. This allows expression
profiling of each individual cell from a mixture of thousands of cells.
Results In validation experiments, we have found that DropSeq can detect tens of thousands of unique
mRNA molecules per cell, while accurately tracking the cell-of-origin of each transcript. Because of the
tiny reaction volumes used, we estimate DropSeq library preparation costs to be 3 cents per cell, and
throughput to be 10,000 single-cell libraries per day. We will present single-cell mRNA-seq data from
tens of thousands of cells, including human primary neurons and glia, demonstrating DropSeq's ability to
classify cellular types and subtypes.
Discussion We believe DropSeq has the potential to accelerate progress from psychiatric genetics to
biological insights by enabling the comprehensive cellular-level characterization of complex tissues
throughout the brain. We will describe research strategies for systematically relating findings from
psychiatric genetics to specific neuronal and glial populations.
Christina Halgren1, Niels Tommerup2, Peter Jacky3, Iben Bache4, Allan Lind-Thomsen2, Malene
Boegehus Rasmussen2, Mana Mehrjouy2, Claus Hansen2, Ana Carolina dos Santos Fonseca5, Angela
Vianna Morgante6, Kikue Terada Abe7, Mads Bak2, The International Breakpoint Mapping Consortium
Wilhelm Johannsen Center for Functional Genome Research, Department of Cellular and Molecular
Medicine, University of Copenhagen, 2Wilhelm Johannsen Center for Functional Genome Research,
Department of Cellular and Molecular Medicine, University of Copenhagen, Denmark, 3Kaiser
Permanente Emeritus, 4Wilhelm Johannsen Center for Functional Genome Research, Department of
Cellular and Molecular Medicine, University of Copenhagen, Denmark and Department of Clinical
Genetics, Rigshospitalet, Copenhagen, Denmark, 5Wilhelm Johannsen Center for Functional Genome
Research, Department of Cellular and Molecular Medicine, University of Copenhagen, Denmark and
Departamento de Genetica e Biologia Evolutiva, Universidade de Sao Paulo, Brazil, 6Departamento de
Genetica e Biologia Evolutiva, Universidade de Sao Paulo, Brazil, 7Sarah Network of Rehabilitation
Background Despite the availability of the draft human genome for a decade, we still lack genotypephenotype-information for ~80-90% of our ~22,000 protein-coding genes, and for almost all of the
rapidly growing number of non-coding RNA (ncRNA) genes and regulatory elements. Even with the
prospects of exome and full genome sequencing, it will take decades and tremendous resources to
saturate the exome, transcriptome and regulome with mutations that can be linked to normal and
abnormal phenotypes including neurodevelopmental and neuropsychiatric disorders.
Methods As a supplement to exome and full genome sequencing strategies, we will use already
identified balanced chromosomal rearrangements (BCR) to establish a first, detailed map of mutations
covering a significant fraction of the human genome. In the first clinical and molecular re-examination of
unselected carriers of de novo balanced chromosomal rearrangements (BCRdn) detected by 40 years of
prenatal diagnosis in Denmark, we used high throughput mate-pair sequencing to rapidly map the
chromosomal breakpoints to sequence level.
Results We showed that BCRdn truncate protein coding genes, ncRNA genes, unannotated transcripts
detected by deep sequencing, as well as developmental regulatory genomic landscapes, mimicking
random mutagenesis. Phenotype-genotype associations were guided by information obtained in nationwide Danish medical registries including the Danish Psychiatric Central Research Register. With a mean
follow-up period of 17 years (range 4-34 years), we identified a ~20% morbidity-risk exclusively
involving neurodevelopmental and neuropsychiatric disorders, e.g. intellectual disability, behavioural
disorders, autism spectrum disorders, depression and anxiety. This is likely to represent a conservative
morbidity-risk since many neuropsychiatric disorders manifest in adolescence and adulthood.
Discussion We have initiated clinical re-examination and mapping of all known BCRs in Denmark.
Based on a population of just 5.5 million, this will provide data on >1,200 breakpoints. By international
expansion we will extend this at least 10-fold to reach a proposed first goal of ~10,000 breakpoints.
Unlike other large scale genomic efforts, all countries including undeveloped and developing countries
can participate. We expect that the breakpoint-map will identify and confirm numerous genotypephenotype associations, a majority of which will involve disorders of the brain.
Stephan Ripke1, PGC PsychChip Group
Massachusetts General Hospital
Background The Psychiatric Genomics Consortium (PGC) is an international group of researchers
whose major aim is to maximize the utility of psychiatric genome-wide association studies (GWAS)
through mega-analysis. In recent years, these studies have successfully identified many novel genetic
associations for psychiatric disorders by integrating data from >170,000 subjects. To continue these
efforts, the PGC has developed a custom genotyping array, the PsychChip, and is coordinating
genotyping of over 100,000 samples at the Stanley Center of the Broad Institute and the Mount Sinai
School of Medicine.
Methods The PsychChip consists of three components: a GWAS backbone of ~256k SNPs, ~236k rare
and low-frequency exome variants, and ~50k custom markers tailored to psychiatric disorders. We used
previous psychiatric genetic studies to select markers with the following goals in mind. First, we ensured
any variants showing modest association (P<0.01) were represented on the chip either directly or
indirectly. Second, for all highly associated loci (P<0.00001), we selected a dense set of markers in the
region that can be used for fine-mapping studies. Third, we added extra markers to regions with copy
number variants (CNV) associated with psychiatric disorders. Fourth, we added rare variants discovered
from whole-exome sequencing studies. Finally, we predicted where functional variants were likely to
occur for genes that have been shown to be strongly associated with psychiatric disorders (e.g., CHD8
and autism).
Results As of May 2014, we have genotyped 11,811 samples on the PsychChip array at the Broad
Institute and 2,666 samples at Mount Sinai. For genotype calling, we tested GenCall and Birdseed for
common variants to assess which algorithm generates the most robust data. We found that a consensus
calling approach maximizes the number of SNPs that pass QC and minimizes the number of Mendel
errors and violations of Hardy-Weinberg Equilibrium. We used zCall to recover rare genotypes missed by
GenCall and Birdseed. Preliminary analysis of ~2,000 schizophrenia cases, ~2,000 controls and ~1,100
trios with schizophrenic probands showed that 85% of the previously reported schizophrenia associations
are in a consistent direction (P=2.1x10-14).
Discussion A major limitation in psychiatric genomics has been inadequate sample size. We believe that
the PsychChip will be an important tool due to its low cost and targeted content for psychiatric disorders.
The PsychChip can be purchased by anyone directly from Illumina. By the end of 2014, we project to have
genotyped >100,000 samples, including substantial numbers of new cases for schizophrenia, bipolar
disorder, ADHD, autism, PTSD, OCD, and anorexia nervosa. The PsychChip pilot data demonstrate that
this genotyping platform will be a useful for interrogating the role of common variation in psychiatric
illnesses while also enabling the assessment of rare coding and copy number variation. Substantial data
generation is in progress, and we will present an update at WCPG in 10/2014.
Rebecca Knickmeyer1, Kai Xia1, Shaili Jha1, Fei Zou1, Hongtu Zhu1, Martin Styner1, Pat Sullivan1, John
University of North Carolina Chapel Hill
Background Brain development in the prenatal and perinatal periods is extremely dynamic and may be
critical in the etiology of psychiatric illness. Previous research revealed high heritability of gray and white
matter volumes in neonates, but the specific genetic variants which contribute to this variation remain
unknown. The primary aim of this study was to identify single nucleotide polymorphisms (SNPs)
associated with intracranial and global tissue volumes in neonates.
Methods Buccal cells from a large and well-characterized population sample of infants assessed with
high-resolution MRI of the brain at 2 weeks of age were genotyped with Affymetrix Axiom GenomeWide LAT and Exome arrays. Following rigorous quality control, SNP imputation was performed using
data from the 1000 Genomes project. An automatic, atlas-moderated expectation maximization
segmentation tool was used to classify brain tissue as gray matter (GM), white matter (WM), or cerebral
spinal fluid (CSF). In addition to total tissue volumes, intracranial volume (ICV) and cortical GM and
WM were also calculated. 594 subjects (278 singletons and 316 twins/siblings) with high quality genetic
and neuroimaging data are included in this analysis. To account for the correlation structure between
twins/siblings, linear mixed effect models were used to test a total of 9.5 million SNPs against each
MRI variable.
Results An intergenic SNP in 15q13.3 between KLF13 and OTUD7A was significantly associated with
ICV (rs8030297; p=2.98 x 10-8) and total WM (rs6493639, p=4.24 x 10-8), and marginally associated
with total GM (rs8030297; p=5.25 x 10-7) and cortical WM (rs6493639, p=1.17 x 10-7). Additional
marginally significant SNPs in/near biologically plausible genes were also identified.
Discussion 15q13.3 microdeletion increases the risk of intellectual disability, seizures, behavioral
problems, and psychiatric disorders including schizophrenia. The current results suggest that common
genetic variants in this region are associated with brain volumes in neonates and thus may play a role in
cognitive development and psychiatric risk. We are also actively testing whether the combined effects of
many common variants each with a small effect size predict variation in neonatal brain structure using
pathway analysis and exploring the impact of copy number variants (CNVs) on neonatal brain structure.
Ultimately, identifying genetic variations impacting brain development will significantly improve
diagnosis, guide research efforts into environmental risk factors, and generate new therapeutic
possibilities for individuals with psychiatric conditions.
Emma Sprooten1, Navin Cota2, Emma Knowles1, D. Reese McKay1, Joanne E Curran3, Jack W. Kent 3,
Melanie A. Carless3, Marcio Almeida3, Thomas Dyer3, Rene L. Olvera4, Peter Kochunov5, Laura
Almasy3, Vince D Calhoun6, John Blangero3, Jessica A Turner7, David C Glahn1
Yale University, 2The Mind Research Network, 3Texas Biomedical Institute, 4University of Texas Health
Science Center, 5University of Maryland, 6University of New Mexico, 7Georgia State University
Background Meta-analysis (Glahn et al. 2009; Bora et al., 2011) and multivariate mega-analysis (Turner
et al., 2012) indicate that grey matter density in the insula and the medial prefrontal cortex (mPFC) are
the most consistent and pronounced imaging-based grey matter abnormalities in association with
schizophrenia. Moreover, this grey matter component is also altered in unaffected siblings of patients
(Turner et al., 2012), indicating that it may be a mediating some of the effects of genetic risk for the
disorder. We applied source-based morphometry (SBM) to a large sample of randomly ascertained
extended pedigrees to extract the same insula-mPFC grey matter component, estimate its heritability, and
identify quantitative trait loci (QTL) that influence it in the general population.
Methods As part of the GOBS study, T1-weighted MR images were acquired for 887 individuals from
extended pedigrees of Mexican-American ancestry (532 female; 18-85 years; pedigree size: 1-258
individuals). After normalisation and grey mater segmentation, SBM was applied to extract 21 spatially
independent components. The insula-mPFC component was identified by visual inspection and its
overlap with the schizophrenia-associated component (Cota et al., In Review) was verified using the Dice
coefficient and cross-voxel correlations. To estimate the heritability, the weights on the component,
reflecting the overall grey matter density in the contributing voxels for each individual, were entered into
polygenic analyses in SOLAR (Almasy & Blangero, 1998). Linkage analysis was performed by
extending the polygenic model with location-specific identiy-by-descent information for ~15,000 points
across the genome. Linkage peaks (LOD>2.9) were further examined by performing associations of SNPs
Results The insula-mPFC grey matter SBM component derived from GOBS was very similar to the one
derived from the case-control sample (Cota et al., In Review): Dice coefficient: 0.42, Pearson r = 0.58.
The overall grey matter density across this component was highly heritable, as indicated by the polygenic
model of the weights (h2 = 0.59; p = 1.78*10-15). A QTL was identified on chromosome 12 at
12q24.22-12q24.23, with a highly significant LOD = 3.76. There were 397 common SNPs under the
linkage peak, the strongest association of which was for rs7133582 (p = 7.71*10-4) in a transcription
factor binding site of KSR2, at the 12q24.23 end of the peak, which is in agreement with the maximum
LOD score in this locus.
Discussion There is compelling evidence that gray matter density in the insula and mPFC is reduced in
patients with schizophrenia, and their unaffected relatives. Our findings indicate that genetic variation in
12q24 influences grey matter density in these brain regions. Our QTL has previously been linked to
schizophrenia (Bulayeva at al., 2007) and bipolar disorder (Berettini et al., 2001). Our top SNP under the
peak is located <200kb away from the schizophrenia candidate gene NOS1, and near the single top SNP
in the ENIGMA genome-wide association analysis with hippocampal volume (Stein et al., 2012). In
conclusion, mPFC and insula morphology are likely brain morphological endophenotypes that are coinherited with schizophrenia susceptibility variants at 12q24, and are thus vulnerability markers that can
give further insights into the bioligical mechanisms of the development of schizophrenia and related
Bryan Dechairo1, Josiah Allen1, Joseph Carhart1, Andrew Marshak1, Joel Winner1, Tony Altar1
AssureRx Health
Background Less than half of patients experience complete remission when initially treated with
antidepressants. These medications show comparable efficacy across drug classes, with limited
improvement relative to placebo. By optimizing therapeutic selection for patients, pharmacogenomics can
increase treatment response and decrease healthcare costs. GeneSight Psychotropic is a combinatorial
pharmacogenomic test designed to bridge the translational gap from the bench to the point of care.
GeneSight integrates variations in 8 genes (CYP2D6, CYP2C19, CYP2C9, CYP2B6, CYP1A2,
CYP3A4, SLC6A4, HTR2A) to stratify 38 psychotropic medications into one of three cautionary
categories, based on each medication’s metabolic pathways and mechanisms of action.
Methods In three clinical outcome studies, 258 depressed subjects who had failed at least one
antidepressant medication were enrolled into one of two treatment arms: GeneSight-guided treatment
(results were available to clinicians at the beginning of the trial), or treatment as usual (TAU; results were
withheld until the end of the trial). Study visits occurred at baseline and weeks 2, 4, and 8. In a
retrospective healthcare utilization study, the medical charts of 96 psychiatric patients were reviewed for
healthcare utilization (e.g., outpatient visits, medical absence days, disability claims) and analyzed
according to the GeneSight cautionary category of their medications. In a prospective pharmacy claims
pilot, 2,176 patients who received GeneSight testing were propensity matched to 10,880 non-tested, TAU
patients. Pharmacy data were tracked for 180 days prior to and 365 days following project entry.
Medication costs were compared between the two groups and between cautionary categories.
Results Each prospective study showed improved clinical outcomes for subjects in the GeneSight arm
relative to the TAU arm. In a meta-analysis, a 3.9 additional HAM-D17 point reduction was obtained
from baseline to week 8 for subjects in the GeneSight versus TAU arm, representing a 71% greater
treatment response. Stratification by GeneSight cautionary category within the blinded, TAU arm showed
almost no improvement for subjects on genetically discordant medications, while subjects on genetically
concordant medications showed the most improvement (p = 0.003). In the GeneSight arm, >90% of
subjects were switched from genetically discordant to genetically concordant medications and showed
improved clinical outcomes relative to their blinded, TAU counterparts (p = 0.005).
Discussion Continued from results: In the retrospective healthcare utilization study, subjects on
discordant medications had 69% more total healthcare visits (p = 0.01), 3-fold more medical absence days
(p = 0.04), 4-fold more disability claims (p = 0.003), resulting in an estimated $5,188 higher medical
costs relative to subjects on genetically concordant medications. In the prospective pharmacy claims pilot,
GeneSight-guided patients saved a mean $1,035.60 in annual medication costs compared to unguided
TAU patients (p < 0.0001). Within the GeneSight arm, patients who remained on genetically concordant
medications saved $587.77 more annually relative to patients who remained on genetically discordant
medications (p = 0.007). Conclusions - GeneSight has shown clinical validity by predicting patient
treatment responses. - GeneSight clinical utility is evidenced by a 2.3-fold greater odds of response. GeneSight is estimated to reduce healthcare costs by over $3,000 annually.
James Crowley1, Andrew Morgan1, Randal Nonneman1, Corey Quackenbush1, Cheryl Miller1, Allison
Ryan1, Molly Bogue2, Sur Paredes1, Scott Yourstone1, Ian Carroll1, Thomas Kawula1, Maureen Bower1,
Balfour Sartor1, Patrick Sullivan1
University of North Carolina at Chapel Hill, 2Jackson Laboratory
Background The second-generation antipsychotic olanzapine is effective in reducing psychotic
symptoms but is associated with considerable weight gain. Given the known involvement of the gut
microbiome in obesity, we used a mouse model to evaluate the role of the gut microbiome in olanzapineinduced weight gain.
Methods C57BL/6J mice were randomized to receive either olanzapine (50 mg/kg diet) or placebo
while consuming a high-fat diet ad libitum beginning at 8 weeks of age and body weight was measured
weekly. Results First, we established that oral delivery of olanzapine to C57BL6/J mice on a high fat
resulted in considerable weight gain compared to placebo (p = 1.1 Г— 10-5). Second, we found that mice
raised in germ-free conditions had no significant weight gain while consuming olanzapine (p = 0.48) but
that the same mice had significant weight gain following introduction of gut flora (p = 4.9 Г— 10-3). Third,
we used a randomized controlled crossover design to survey the fecal microbiome before, during, and
after olanzapine treatment by sequencing bacterial 16S ribosomal DNA. Olanzapine potentiated a shift
towards an “obesogenic” microbiota and this shift was correlated with weight gain. Finally, we
demonstrated that olanzapine has antimicrobial activity in vitro against two commensal enteric bacterial
Discussion Taken together, these results provide strong evidence for a mechanism underlying
olanzapine- induced weight gain in mouse. Olanzapine is a subtle antimicrobial, and shifts the gut
microbiome to an obesogenic pattern. This work suggests a hypothesis for clinical translation in human
patients. We note that the effects of olanzapine are analogous to low-dose antibiotic regimens used to
promote growth in livestock.
Xueying Jiang1, Sevilla Detera-Wadleigh 1, Nirmala Akula1, Francis McMahon1
National Institute of Mental Health
Background Genome-wide association studies (GWAS) have identified several risk variants for bipolar
disorder (BD), but the functional consequences of most variants remain undefined. A common variant
(rs9834970) located ~15 kb 3’ of the gene TRANK1 on chromosome 3p22 has shown genome-wide
significant association with BD in several studies [1-4] and nearby markers have been associated with
schizophrenia [5]. Previously, we showed that valproic acid (VPA), an effective treatment for BD,
increased TRANK1 expression in commercial cell lines [1]. In this study, we aimed to confirm the effect
of VPA treatment on TRANK1 expression in induced pluripotent stem cells (iPSc) and in iPSC-derived
neural progenitor cells (NPCs), and to test the effect of the rs9834970 risk allele (G) on TRANK1
expression in both iPSc and NPC cultures.
Methods iPSC lines were generated by lentiviral reprogramming of adult human fibroblast cells from 7
individuals with known genotypes at rs9834970. All 7 iPSC lines were further differentiated into NPCs
with Gibco PSC neural induction medium (Life technology, CA). iPScs and NPCs were validated by
standard immunochemical analysis. RNA was extracted at baseline and after 72h of treatment with VPA
(0.5mM or 1mM) from 4 iPSC and 7 NPC lines. TRANK1 gene expression levels were measured by
quantitative real-time polymerase chain reaction (qRT-PCR), with 3 technical replicates for each
treatment condition. All samples were genotyped on the Illumina Infinium Human OmniExpress Exome
bead array. Statistical significance of gene expression differences was determined by two-way ANOVA.
Results VPA treatment substantially increased TRANK1 expression in both iPSc and NPC lines. Foldchange vs. baseline ranged from 2.76 (0.5mM VPA) to 6.18 (1mM VPA) in iPSc (P<0.01), and from 2.76
(0.5 mM VPA) to 4.06 (1mM VPA) in NPC (p<0.01). Carriers of the risk allele of rs9834970 had lower
baseline TRANK1 expression in NPC lines (fold-change vs non-carriers, 4.98, p<0.05). The decreased
TRANK1 expression in risk allele carriers was normalized by VPA (Number of risk alleles x VPA,
F(2,17)=4.48, p<0.03).
Discussion These results confirm and extend our previous findings, demonstrate that VPA increases
TRANK1 expression in both iPSc and NPC lines, and reveal a previously unknown cis-effect of
rs9834970 on TRANK1 expression that is antagonized by VPA. These findings suggest that VPA
normalizes reduced TRANK1 expression in carriers of the BP risk allele at rs9834970, implying a novel
therapeutic mechanism for VPA in BD. Gene expression studies in iPSc-derived NPCs of risk allele
carriers may prove to be a useful strategy to characterize the tissue-specific functional impact of risk
alleles implicated by GWAS, ultimately enhancing our understanding of etiological mechanisms and
pointing the way toward improved pharmacologic therapies.
Stefanie Malan-MГјller1, Lorren Fairbairn 2, Mahjoubeh Jalali 3, Edward Oakeley 4, Junaid Gamieldien 3,
Martin Kidd 5, Soraya Seedat 2, Sian Hemmings 2
Stellenbosch University, 2Stellenbosch University, Department of Psychiatry, 3University of the Western
Cape, South African National Bioinformatics Institute, 4Novartis Institutes for BioMedical Research,
Biomarker Development - Human Genetics and Genomics, Genome Technologies, 5Stellenbosch
University, Centre for Statistical Consultation
Background Posttraumatic stress disorder (PTSD) is a severe, chronic and debilitating psychiatric
disorder that can occur after a traumatic event. D-cycloserine (DCS), a partial N-methyl-D-aspartate
(NMDA) receptor agonist, has been found to be effective in facilitating fear extinction in both animal and
human studies of anxiety. However, the precise mechanism whereby DCS facilitates fear extinction is
unknown. The aim of this study was to elucidate the molecular mechanism of action of DCS in
facilitating fear extinction in a rat model of PTSD.
Methods The PTSD animal model described by Siegmund and Wotjak (2007) was followed. Rats were
grouped into four groups, Fear + saline (FS), Fear + DCS (FD), Control + Saline (CS) and Control + DCS
(CD). Animal behavioural tests were conducted to determine which rats displayed anxiety-like behaviour.
Next-generation RNA-seq and microRNA (miRNA)-seq and subsequent bioinformatics analyses were
performed on RNA extracted from left dorsal hippocampi (LDH) to identify differentially expressed
genes and miRNAs between the groups which might provide information on how DCS facilitates fear
extinction. Target enrichment analysis was performed to determine whether the differentially expressed
miRNAs targeted any of the differentially expressed genes identified in the RNAseq analysis. A
luciferase assay was performed to functionally verify if the upregulation of rno-miR-31a-5p may have
facilitated the downregulation of its predicted target gene, interleukin-1 receptor antagonist (IL1RN).
Results A total of 424 genes were significantly down-regulated in the FD Well-adapted (FDW) group
compared to the FS maladapted (FSM) group, of which 121 genes were predicted to be biologically
significant to PTSD. Twenty seven genes were significantly upregulated in the FDW group compared to
the FSM group, of which nine genes were predicted to be biologically relevant to PTSD. Genes
transcribing components within the immune, proinflammatory and oxidative stress systems were
downregulated in fear conditioned rats that received DCS. These factors mediate neuroinflammation and
cause neuronal damage. DCS also regulated genes involved in learning and memory processes, genes that
were previously associated with PTSD and disorders that commonly co-occur with PTSD. In addition, 32
miRNAs were differentially expressed between FDW and FSM groups. Functional luciferase analysis
indicated that the upregulation of rno-mi31a-5p could have facilitated the downregulation of IL1RN as
detected in RNAseq.
Discussion Differential gene and miRNA expression analyses in this PTSD animal model enabled us to
identify genes, miRNAs, and networks that might explain how DCS facilitates fear extinction. It is
hypothesised that DCS attenuates neuroinflammation and subsequent neuronal damage, and also
regulates genes involved in learning and memory processes. Gene and miRNA expression alterations may
have mediated optimal neuronal functioning, plasticity, learning and memory which contributed to the
fear extinction process. Furthermore, differentially expressed genes that were associated with other
chronic medical conditions, such as cardiovascular disease and metabolic diseases, might help to explain
the co- occurrence of these disorders with PTSD.Identifying the molecular underpinnings of fear
extinction might bring us closer to understanding and effectively treating PTSD.
Annika Forsingdal1, Jacob Nielsen1, Marcelo Bertalan2, Thomas Werge2
H. Lundbeck A/S, 2Mental Health Center Sct Hans
Background Genome wide association studies have revealed that certain copy number variants (CNVs)
strongly increase the risk of schizophrenia and other psychiatric diseases. One such CNV is a 1.5 MB
long hemizygous deletion located in the 15q13.3 region that covers 6 genes (FAN1, MTMR10, TRPM1,
KLF13, OTUD7A, and CHRNA7). The 15q13.3 microdeletion increases the risk of schizophrenia,
epilepsy and autism (Malhotra and Sebat, 2012). Human cases of homozygous microdeletion carriers
have also been reported, all with severe impairments (Hoppman-Chaney et al., 2013). However, the
mechanisms underlying increased disease risk in the 15q13.3 microdeletion syndromes are unknown. A
mouse model of the human 15q13.3 hemizygous microdeletion syndrome, Df(h15q13)+/-, has been
generated, and characterization of the model identified disease-related phenotypes (Fejgin et al., 2013).
However, those phenotypes were relatively subtle, which complicates mechanistic exploration.
Methods Homozygous 15q13 knockout mice, Df(h15q13)-/-, were bred from Df(h15q13)+/-.
Df(h15q13)-/- mice were characterized by basic physiological and behavioral tests as well as disease
related behavioral tests. Transcriptional changes were assessed by whole transcriptome RNAsequencing of brain and body samples from Df(h15q13)-/- and wildtype mice.
Results Df(h15q13)-/- display physiological and behavioral impairments compared to wildtype mice. A
number of genes are differentially expressed in Df(h15q13)-/- compared to wildtype. Differential
expression is seen both in the central nervous system and in the periphery. Bioinformatic analysis of
RNA sequencing data is ongoing.
Discussion As expected, a number of genes are differentially expressed in the Df(h15q13)-/- mice. The
expression profile of these mice will provide cues to which biological mechanisms that predispose 15q13
deletion carriers to psychiatric diseases and guide future mechanistic exploration.
Abraham Palmer1
University of Chicago
Background The subjective response to d-amphetamine is heritable and may serve as an endophenotype
for a variety of psychiatric disorders, especially those related to dopaminergic signaling.
Methods We performed a Genome Wide Association Study (GWAS) for the subjective responses to
amphetamine using data from 398 non-drug abusing healthy volunteers. Response to amphetamine were
measured using a double-blind, placebo-controlled, within-subjects design. We used sparse factor
analysis to reduce the dimensionality of the data and then performed GWAS using genotypes from Affy
6.0 imputed to 1000 Genomes.
Results We identified several putative associations; the strongest was between a factor reflecting the
positive subjective effects of amphetamine and a SNP (rs3784943) in the 8th intron of cadherin 13
(CDH13; P?=?4.58Г—10-8), a gene previously associated with a number of psychiatric traits, including
methamphetamine dependence. We have examined both CDH13 knockout rats and adiponectin knock out
mice and observed differences in conditioned place preference, which offers additional support for the
role of CDH13 in modulating the sensitivity to the subjectively positive effects of amphetamine.
Additionally, we observed a putative association between a factor representing the degree of positive
affect at baseline and a SNP (rs472402) in the 1st intron of steroid-5-alpha-reductase-?-polypeptide-1
(SRD5A1; P?=?2.53Г—10-7), a gene whose protein product catalyzes the rate-limiting step in synthesis of
the neurosteroid allopregnanolone. This SNP belongs to an LD-block that has been previously associate
Discussion None of the data from CDH13 KO rats or adiponectin knock out mice have been published.
Some ongoing analyses of the human GWAS data will also be discussed.
Lynsey Hall1, Toni-Kim Clarke1, Ana-Maria Fernandez-Pujals1, Pippa Thomson1, Caroline Hayward1,
Donald MacIntyre1, Chris Haley1, David Porteous1, Ian Deary2, Andrew McIntosh1
University of Edinburgh
Background Despite heritability estimates of 31-42% genetic studies of major depressive disorder
(MDD) have failed to identify any robust, replicable genetic risk loci. One method which may aid genetic
discovery is to identify quantitative endophenotypes for depression. The aim of this study was to assess
whether cognitive, mood and personality traits genetically correlated with depression could be used to
generate a quantitative endophenotype for MDD where genetic correlation and heritability were jointly
maximized. We further hypothesized this composite trait would have greater power to identify genetic
risk variants for depression than the binary classification of case/control.
Methods Generation Scotland: A Scottish Family Health Study (GS:SFHS) is a large family and
population-based study with genotypic and extensive phenotypic information, including detailed data on
mental health and ten measures of cognitive function, mood and personality. Bivariate heritability
analysis of 21,340 individuals from GS:SFHS, conducted using social pedigree data, revealed five traits
with a genetic correlation (rg) with depression of > 0.2. The three most highly correlated traits (General
Health Questionnaire (rg=0.69), Mood Disorder Questionnaire (rg=0.62) and Eysenck Personality
Questionnaire for Neuroticism (rg=0.58)) were used to create a composite trait. These were subjected to
principal components analysis and the first unrotated principal component used as a quantitative measure
of depression (h2=0.38, rg=0.79).
Results Genome-wide association analyses of the composite trait (n=9,863; nSNPs=609,002), depression
as a binary variable (n=1,700 cases, 7,634 controls) and recurrent depression (n= 1,147 cases, 7,634
controls) identified no genome-wide significant SNPs (p < 5x10-8). The top GWAS hit for the composite
trait was rs4661250 (p = 1.1x10-7), located in the 3' intron of sushi domain-containing protein 4 (SUSD4)
gene, which is highly expressed in the white matter of oligodendrocytes. White matter integrity is reduced
in MDD suggesting a plausible role for SUSD4 in the aetiology of depression. Using summary data from
the PGC GWAS of MDD, polygenic risk score profiles for MDD were generated in GS:SFHS. These
scores were tested for their association with MDD and the newly derived composite trait. Polygenic score
had a stronger association with and explained more of the variance of the composite trait (p=8.4x10-7,
0.4%) than depression (p=0.001, 0.1%) or recurrent depression (p=7.2x10-4, 0.1%).
Discussion Given that the composite trait is more strongly associated with polygenic risk for MDD,
using summary data from a completely independent sample, it suggests that the trait is successfully
capturing a greater proportion of the heritable component of depression. However, this study is still
underpowered to detect causal variants for depression whether analyzed as a binary or quantitative trait.
The remainder of the GS:SFHS cohort (currently being genotyped), will afford a much larger GWAS
sample (n= ~21,000; ~2750 cases, ~17,400 controls). Replicating these analyses in the larger cohort, in
conjunction with larger collaborative efforts, may yield robust results which will aid the discovery of
genetic variants associated with depression.
Jie Song1, Sarah E. Bergen 2, Ralf Kuja-Halkola 2, Henrik Larsson 2, Mikael LandГ©n 3, Paul Lichtenstein 2
Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, 2Department of Medical
Epidemiology and Biostatistics, Karolinska Institutet, 3Institute of Neuroscience and Physiology, The
Sahlgrenska Academy at Gothenburg University; Department of Medical Epidemiology and Biostatistics,
Karolinska Institutet
Background Bipolar disorder (BPD) shares genetic components with other psychiatric disorders;
however, uncertainty remains about where in the psychiatric spectra BPD falls. To understand the
etiology of BPD, we studied the familial aggregation of BPD and co-aggregation between BPD and
schizophrenia, depression, anxiety disorders, attention deficit/hyperactivity disorder (ADHD), drug
abuse, personality disorders and autism spectrum disorders (ASD).
Methods A population-based cohort was created by linking several Swedish national registers. 54,723
BPD individuals were identified among 8,141,033 offspring from 4,149,748 nuclear families. The relative
risk of BPD in relatives and co-occurrence of other psychiatric disorders in BPD patients and their
relatives were compared to those of matched population controls. Structural equation modeling was used
to estimate the heritability and tetrachoric correlation.
Results The familial risks for relatives of BPD probands were 5.8-7.9 in first degree relatives, and
decreased with genetic distance. Co-occurrence risks for other psychiatric disorders were 9.7-22.9 in
BPD individuals and 1.7-2.8 in full siblings of BPD probands. Heritability for BPD was estimated at
58%. The correlations between BPD and other psychiatric disorders were considerable (0.37-0.62) and
primarily due to genetic effects. The correlation with depression was the highest (0.62), and was 0.44 for
Discussion The high familial risks provides evidence that genetic factors play an important role in the
etiology of BPD, and the shared genetic determinants suggest pleiotropic effects across different
psychiatric disorders. Results also indicate BPD is in both the mood and psychotic spectra, but possibly
more closely related to mood disorders.
Eli Stahl1, PGC Bipolar Working Group
Icahn School of Medicine at Mount Sinai
Background The purpose of the Psychiatric Genomics Consortium (PGC) is to conduct meta-analyses of
genome-wide genetic data for psychiatric disease. Recognizing that individual GWAS studies are too
small to have adequate power for gene discovery, an international PGC Working Group has focused on
extending their meta-analysis of bipolar disorder (BD). Recently, we reported a combined GWAS of
bipolar disorder in a sample of 16,731 individuals that identified two genome wide-significant loci
(Nature Genetics, 2011). Here we present the results of PGC2 Bipolar Disorder GWAS, including data
from 21,035 cases and 28,758 controls.
Methods Data for 13,200 new case samples and 19,508 new controls of European decent were received
from Germany, Bulgaria, Romania, Sweden, Norway, the UK, the US and Mexico. Subphenotype data
acquisition is ongoing, with at least 11,888 cases of bipolar disorder 1 (BD1), 2,359 bipolar disorder 2
(BD2) and 938 schizoaffective bipolar subtype (SAB). Data were prepared by the PGC central analytic
pipeline as described previously. The data were imputed with 1000 Genomes Project data and analyzed
using standard logistic regression with MDS components as covariates. SNP-heritability analyses were
conducted using GCTA, and polygenic scoring analyses using all SNPs were conducted as previously
Results Initial analyses of the entire dataset yield at least eleven genome-wide significant, and at least
three new, bipolar risk loci. We continue to investigate substantial heterogeneity among the sample
cohorts, revealed by leave-on-out BD polygenic risk score profiling. Bivariate analysis of a subset of the
new data reveals that SNP-heritability of BD1 (0.35) is greater than that of BD2 (0.23, P=4x10-3 for
difference from BD1), and greater than BD1-BD2 coheritability (0.23, P=4x10-4 for difference from
BD1), consistent with an incomplete genetic correlation (rG=0.84) between BD1 and BD2 diagnoses.
Polygenic risk scores based on other psychiatric disease GWAS differentiate between BD subphenotypes;
for example, schizophrenia polygenic scores are higher in BD1 than BD2 (P=0.002). We also report
initial lookups of suggestively associated SNPs in preliminary psychChip data, and pathway analyses of
the primary GWAS results.
Discussion In conclusion, we provide support for the importance and utility of continued GWAS
exploration in bipolar disorder in efforts to increase the number of genetic loci with compelling
association to bipolar disorder.
Niamh Mullins1, Robert Power1, Ken Hanscombe1, Helen Fisher1, RADIANT, BACCs and GENDEP
Investigators1, Rudolf Uher2, Anne Farmer1, Peter McGuffin1, Gerome Breen1, Cathryn Lewis1
King's College London, 2King's College London, Dalhousie University
Background Depression is a common and disabling condition which results from a complex interaction
between genetics and environmental factors. The genetic diathesis for major depressive disorder (MDD)
is highly polygenic, resulting from the additive and multiplicative interaction of many genetic variants
with small effect sizes. Adverse experiences such as childhood trauma and stressful life events (SLEs) are
also risk factors. Gene-by-environment interaction studies in depression have typically investigated
candidate genes, but polygenic scores that incorporate thousands of genetic variants simultaneously,
better capture the genetic liability to a complex trait. Here, for the first time we test for an interaction
between polygenic scores and environmental adversity in the etiology of major depressive disorder.
Methods The RADIANT UK sample consists of patients with recurrent MDD and controls screened for
the absence of psychiatric illness. Blood samples were genotyped genome-wide. The List of Threatening
Experiences Questionnaire was used to assess the numbers of SLEs in the 6 months prior to worst episode
of depression (cases) or interview (controls). The number of SLEs was adjusted for age and sex. The
Childhood Trauma Questionnaire was also used to assess exposure to sexual, physical and emotional
abuse, physical and emotional neglect in childhood. Discovery results from a mega-analysis on MDD by
the Psychiatric Genomics Consortium (with the RADIANT UK sample removed) were used to construct
MDD polygenic scores for each individual in the RADIANT UK validation dataset. Ability to predict
case/control status was tested using logistic regression with an interaction between polygenic score and
environmental adversity and principal components as covariates to adjust for population stratification.
Results In 1605 MDD cases and 1064 controls, cases reported a significantly greater number of SLEs
than controls (mean cases= 1.57, mean controls= 0.68, P < 0.001). Polygenic scores for depression
showed significant predictive ability for depression in the RADIANT UK sample (P = 1.8x10-6,
Nagelkerke’s pseudo-R2 = 0.011). Polygenic score-by-SLE interaction showed no predictive ability for
case/control status. Genetic liability to bipolar disorder and schizophrenia will also be examined using
results from the Psychiatric Genomics Consortium as well as testing for interactions between polygenic
scores and childhood trauma.
Discussion Polygenic scoring is an appropriate approach to investigating the genetics of a complex trait.
Including environmental risk along with genetics is important in studying the etiology of major
depressive disorder and using polygenic scores rather than candidate genes may increase statistical
power to detect gene-by-environment interactions.
T. Bernard Bigdeli1, Stephan Ripke2, Silviu-Alin Bacanu3, Roseann E. Peterson3, Ayman H. Fanous4, Li
Qibin5, Yu Xin6, Jonathan Flint7, Kenneth S. Kendler3, Patrick F. Sullivan8, PGC MDD Workgroup9,
VIPBG, 2Analytic and Translational Genetics Unit, Massachusetts General Hospital,
Virginia Institute for Psychiatric and Behavioral Genetics, Department of Psychiatry, Virginia
Commonwealth University School of Medicine, 4Mental Health Service Line, Washington VA Medical
Center, 5Beijing Genomic Institute, 6 Institute of Mental Health, Peking University, Beijing, CN,
Wellcome Trust Centre for Human Genetics, 8Departments of Genetics and Psychiatry, Center for
Psychiatric Genomics, University of North Carolina at Chapel Hill
Background Major depressive disorder (MDD) is a common, complex psychiatric disorder and a
leading cause of disability worldwide. Although modestly heritable (~30-40%), a complex genetic
architecture has hindered efforts to detect robustly associated genetic risk variants. Furthermore, the
extent to which liability to MDD is shared across ancestrally divergent populations is unknown.
Methods We combined single nucleotide polymorphism (SNP) summary statistics from CONVERGE
(China, Oxford and VCU Experimental Research on Genetic Epidemiology) and the Psychiatric
Genomics Consortium (PGC) studies of MDD, representing 11139 Han Chinese (5647 cases, 5492
controls) and 18663 European (9447 cases, 9215 controls) subjects, respectively. We performed genomewide association study (GWAS) meta-analysis, consisting of 15094 cases and 14707 controls. Secondary
GWAS used phenotypes consisting of recurrent MDD (>2 episodes), only females, and only females
with recurrent MDD. For varying P-value thresholds (PT), we determined the fraction of CONVERGE
SNPs that had the same direction of effect as those in the PGC study and assessed the predictive
accuracy of polygene scores constructed from each study’s results.
Results Of the observed associations not previously reported in GWAS of either dataset individually, the
strongest evidence was for females-only recurrent MDD at SNPs immediately upstream of SLC29A4
(7p22.1; P=6.2x10-8). This locus encodes a transmembrane protein known to facilitate reuptake of
serotonin and dopamine into presynaptic neurons. A binomial sign test of the fraction of CONVERGE
SNPs demonstrating a direction of effect consistent with the PGC results was most significant at a PT [for
consistency with above] threshold of .05 (P=2.4x10-5). A PGC-trained polygene score explained less than
0.3% of the variance in MDD risk in CONVERGE (P=1.65x10-5; recurrent MDD; PT>0.4); a
CONVERGE-trained score explained slightly more than a tenth of a percent of the variance in the PGC
(P=8.82x10-3; females-only MDD; PT>0.5).
Discussion We have conducted a large, cross-population meta-analysis of MDD, and the first such
study to combine European and Han Chinese samples. After multiple-testing correction for the number
of GWAS performed, no single variant remained significant at genome-wide levels, though replication
efforts are ongoing. However, the observation that a significant fraction of SNPs exhibit a consistent
direction of effect across European and Chinese studies, taken together with the significant, bidirectional
predictive values of polygene scores, suggests a shared polygenic risk of MDD across these populations.
These findings support a complex etiology for MDD and possible population differences in predisposing
genetic factors, with important implications for future genetic studies.
Margarita Rivera 1, Adam E Locke2, Tanguy Corre3, Darina Czamara4, Christiane Wolf4, Ana ChingLopez5, Yuri Milaneschi6, Dorret I Boomsma7, Stefan Kloiber4, Bertram MГјller-Myhsok4, Brenda WJH
Penninx6, Martin Preisig8, Anne E Farmer9, Cathryn M Lewis9, Gerome Breen9, Peter McGuffin9
Institute of Psychiatry, King's College London, 2Department of Biostatistics and Center for Statistical
Genetics, University of Michigan, 3Department of Medical Genetics, University of Lausanne and Swiss
Institute of Bioinformatics, 4Max-Planck-Institute of Psychiatry, 5Department of Psychiatry, School of
Medicine, University of Granada, 6Department of Psychiatry and EMGO Institute for Health and Care
Research, VU University Medical Center/GGZ inGeest, 7Department of Biological Psychology, VU
University Amsterdam, 8Department of Psychiatry, Lausanne University Hospital, 9MRC SGDP Centre,
Institute of Psychiatry, King’s College London
Background Depression and obesity are highly prevalent major public health problems that
frequently co-occur. Both conditions are major risk factors for chronic (physical) diseases such as type
2 diabetes, cardiovascular disease and hypertension among others. Shared aetiological factors have
been found between depressive disorder and obesity, although the nature of this association remains
Recently, we reported the first study implicating FTO in the association between depression and obesity.
We aimed to confirm these findings by investigating the FTO rs9939609 polymorphism in a metaanalysis of 13,701 individuals.
Methods The sample consists of 6,902 depressed patients and 6,799 controls from five independent
studies (Radiant, PsyCoLaus, GSK, MARS and NESDA/NTR). As common inclusion criteria we looked
for the studies with information available on a lifetime DSM-IV diagnosis of depressive disorder, body
mass index (BMI) and genotype data for the rs9939609 FTO polymorphism. Homogeneous ethnicity
(Caucasian) was also required for each study to reduce the risk of population stratification. In each
individual study, linear regression models for quantitative traits assuming an additive genetic model
were performed to test for association between the rs9939609 and BMI. We also tested for the
interaction between rs9939609 variant and depression status for an effect on BMI. Age, sex and principal
components were controlled for including them as covariates in the models. Fixed-effects and randomeffects meta-analyses based on inverse-variance-weighted effect size were performed using
Results Fixed-effects meta-analyses support a significant association between rs9939609 polymorphism
and BMI in the whole sample (Гџ=0.07, p=1.29x10-12) and in depressive cases (Гџ=0.12, p=6.92x10-12).
No association was found in the control group (Гџ=0.02, p=0.15). Fixed and random-effects metaanalyses further support a significant interaction between FTO, BMI and depressive disorder (fixedeffects: Гџ=0.13, p=3.087x10-7; random-effects: Гџ=0.12, p=0.027), wherein depressed carriers of the risk
allele have increased BMI risk. Subjects with depression have an additional increase of 2.2% in BMI for
each risk allele, over and above the main effect of FTO and disease status.
Discussion This meta-analysis demonstrates a significant interaction between FTO, depression and BMI,
indicating that depression increases the effect of FTO on BMI. Depression-related alterations in key
biological processes may interact with the rs9939609 FTO risk allele to increase obesity. These findings
will have potential implications for predicting which patients with depression are at risk of high-BMI
related disorders, but could also have more general relevance to the population as a whole. The
identification of common causes for its comorbidity can help better clinical awareness and ascertainment
of such comorbid states.
Panagiotis Ferentinos1, Artemis Koukounari 2, Robert Power 3, Margarita Rivera3, RADIANT
Study Investigators, Rudolf Uher 4, Gerome Breen 3, Ian W. Craig3, Anne E. Farmer 3, Peter
McGuffin 3, Cathryn M. Lewis 3
University of Athens, 2nd Department of Psychiatry, 2Department of Biostatistics, Institute of
Psychiatry, King’s College London, London, United Kingdom, 3MRC Social Genetic and
Developmental Psychiatry Centre, Institute of Psychiatry, King’s College London, London, United
Kingdom, 4Dalhousie University
Background Despite extensive research in the field, the genetic architecture of major depressive
disorder (MDD) remains highly elusive. Phenotypic and genetic heterogeneity have been pinpointed as
mainly responsible for as yet unfruitful investigations. Promising strategies to dissect MDD heterogeneity
have mainly relied on subphenotypes, such as age at onset (AAO) and recurrence/ episodicity. Recurrent
and early-onset forms are most consistently associated with higher familiality and heritability of MDD.
Yet, to date evidence on whether these subphenotypes are per se familial or heritable is scarce in MDD.
The aims of this study are, therefore: first, to investigate the familiality of AAO and episode frequency of
MDD; second, to assess the SNP heritability of AAO and episode frequency in unrelated subjects with
Methods For investigating familiality, we used 1498 subjects with recurrent depression from the DeNt
(Depression Network) affected siblings study (691 families with 2-5 affected full siblings). Square root
AAO (sqrtAAO) was fitted into a linear mixed model (LMM) with center and family as nested random
effects. SqrtAAO familiality was measured by the family-level intraclass correlation coefficient (ICC).
Depressive episode count was fitted into a negative binomial generalized linear mixed model with center
as fixed and family as random effects. An ICC was calculated with a recently described formula. For
estimating SNP heritabilities, we used 2695 unrelated MDD cases from the RADIANT studies.
SqrtAAO was adjusted for gender, study and center in a LMM and standardized residuals were saved.
Episode count was similarly adjusted for gender, age, study and center in a negative binomial model and
deviance residuals were saved. Derived residuals were then used with the GREML method in GCTA
software. Results In the DeNt dataset, the ICC for sqrtAAO was estimated at 0.28 (SE 0.054, p<0.001)
and ICC for episode frequency was 0.071 (SE 0.011, p<0.001). Analyses were underpowered for
estimating SNP
heritability of AAO and episodicity in the RADIANT dataset (power 0.66 and 0.07, respectively;
Visscher et al 2014 PLOS Genetics). The SNP-heritability estimates obtained were 0.16 (SE 0.13, p=0.1)
for AAO and 0.095 (SE 0.18, p=0.29) for episodicity.
Discussion Significant familiality of both AAO and episodicity were obtained in the DeNt sibling
sample; a moderate and small proportion, respectively, of their variance could be attributed to family
membership. However, GREML analyses of SNP heritability were underpowered and we were unable to
confirm that common SNPs capture this information; larger samples are required to estimate the SNP
heritability of AAO and episodicity in MDD.
Andrew Pocklington1, Elliott Rees2, David H. Kavanagh2, Jun Han2, Jennifer L. Moran3, George Kirov2,
Steven A. McCarroll3, James T.R. Walters2, Michael J. Owen2, Michael C. O'Donovan2
Cardiff University, 2MRC Centre for Neuropsychiatric Genetics & Genomics, Cardiff University, UK, 3
Stanley Centre for Psychiatric Research, Broad Institute of MIT and Harvard
Background Individuals with schizophrenia have been found to possess an increased burden of large,
rare CNVs compared to matched controls in most studies. Set based analyses applied to case-control data
indicate enrichment of schizophrenia CNVs for synaptic and developmental gene sets, while de novo
case CNVs have been found to be enriched for members of postsynaptic N-methyl-D-aspartate receptor
(NMDAR) and neuronal activity-regulated cytoskeleton-associated (ARC) protein complexes (Kirov et
al., 2012). Further supporting a role for these complexes in schizophrenia, exome sequencing has
uncovered enrichment of both ARC and NMDAR gene sets for rare de novo point mutations (Fromer et
al, 2014). Here we present a detailed functional analysis of CNVs called in 12,492 schizophrenia cases
and 16,996 controls, drawn from three separate studies (ISC, 2008; Levinson et al., 2011; Hamshere et
al., 2013).
Methods Starting from the primary hypothesis that schizophrenia reflects perturbation of brain function
and development, we analyzed the enrichment of a circumscribed set of annotations based on proteomic,
RNA sequencing and functional genetic data. As a secondary, hypothesis-generating analysis, we then
searched for any additional gene-set enrichment in the more comprehensive range of annotations
available from large, freely accessible databases. Gene set enrichment analysis of large, rare CNVs
(>100kb, frequency < 1%) was performed using a logistic regression model with covariates to correct for
CNV size, number of genes hit and the inclusion of multiple studies and genotyping chips.
Results When compared to permuted data a consistent excess of associated CNS gene sets was observed
across multiple significance thresholds. Gene sets linked to learning and memory, synaptic physiology
and postsynaptic protein complexes were all highly associated. In particular, there was strong,
independent evidence of enrichment in glutamatergic PSD-95 and NMDAR complexes. There was no
evidence of any additional enrichment in non-CNS gene sets.
Discussion There is good evidence that the disruption of specific elements of nervous system function is
associated with schizophrenia. Evidence is strongest for behavioral and physiological correlates of
learning and closely related postsynaptic complexes. These associations support the disease relevance of
functional processes previously implicated through the study of de novo CNVs and rare variants.
Brady Maher1, Matthew Rannals1, Stephanie Cerceo-Page1, Morganne Campbell1, Aaron Briley1,
Andrew Jaffe1, Ran Tao1, Thomas Hyde1, Joel Kleinman1, Daniel Weinberger1
Lieber Institute for Brain Development
Background Schizophrenia is a neurodevelopmental disorder with unknown pathophysiology.
Genome- wide association studies (GWAS) have identified a number of loci associated with increased
risk for SZ and several of these risk variants are located within introns of Transcription Factor 4 (TCF4;
E2-2, ITF2). In addition, autosomal dominant mutations in TCF4 result in Pitt Hopkins Syndrome
(PTHS), a rare neurodevelopmental disorder characterized by a spectrum of symptoms including
hyperventilation, seizures, autistic behaviors, intellectual disability, and brain malformations. Currently,
the molecular mechanisms and underlying pathophysiology responsible for these two disorders are not
understood. Our goal is to determine the function of TCF4 during cortical development and to
understand the molecular mechanism of risk that is associated with genetic variants of TCF4.
Methods To test the function of TCF4 in the developing neocortex we altered its expression by
transfecting layer 2/3 pyramidal cells in the rat medial prefrontal cortex by in utero electroporation. We
knockdown TCF4 expression using two shRNA constructs that target independent sequences within the
TCF4 transcript and over-expressed human TCF4 with recombinant TCF4 constructs. Functional
analysis was performed using whole-cell electrophysiology and confocal imaging in acute brain slices.
To gain insights into the molecular mechanisms responsible for increased risk associated with genetic
variants of TCF4, we analyzed RNA sequencing data obtained from the dorsal lateral prefrontal cortex of
postmortem brains from schizophrenia patients (n=107) and controls (n=107).
Results Embryonic knockdown of TCF4 in layer 2/3 pyramidal cells resulted in decreased intrinsic
excitability and the ectopic appearance of spike-frequency adaptation. These phenotypes were associated
with an increase in the afterhyperpolarization (AHP) amplitude and were rescued by manipulating intraand extracellular Ca2+ levels. Embyronic over-expression of the full-length human isoform TCF4B
significantly accelerated neuronal differentiation and migration. In addition, expression of either TCF4B
or a shorter isoform TCF4A, which lacks a nuclear localization sequence, resulted in the abnormal
distribution of cortical columns. Lastly, analysis of human RNA sequencing data from postmortem dorsal
lateral prefrontal cortex (DLPFC) identified a single TCF4 5’ exon that showed significantly decreased
expression in schizophrenia patients compared to controls. This differentially expressed exon is unique to
TCF4H and in utero transfection of this isoform did not produce abnormal cortical columns.
Discussion Our results suggest the dosage of TCF4 is critical to cortical development and neuronal
physiology. Knockdown of TCF4 produces defects in neuronal excitability and over-expression results in
abnormal cortical columns, the presumed microprocessing unit of the cortex. By analyzing RNA
sequencing data in human brain, we have now identified a specific isoform of TCF4 that infers
schizophrenia risk, suggesting that this may be the molecular mechanism of the clinical association with
schizophrenia. We believe these novel data will allow us to model schizophrenia in our rat model with
high fidelity. Future experiments will be designed to specifically knockdown this risk associated TCF4
isoform and determine its effects on cortical development and neuronal physiology.
Peter Visscher1, Jacob Gratten1, Naomi Wray1, Wouter Peyrot2, John McGrath1, Michael Goddard3
The University of Queensland, 2VU University Medical Center, 3University
of Melbourne
Background There is evidence that the offspring of older fathers are at increased risk of psychiatric
disorders such as schizophrenia and autism, and it is well established that the de novo point mutation rate
increases with paternal age, and that gene-disrupting de novo mutations confer risk for psychiatric illness.
A widespread assumption in the field is that a causal relationship exists between these observations - i.e.
that paternal age-related mutations are sufficient to explain the epidemiological findings. However, not
all the evidence is consistent with this conclusion and an alternative explanation is that elevated genetic
liability to psychiatric illness may lead to delayed fatherhood.
Methods We used population genetic models to explore whether recent empirical estimates of the de
novo mutation rate can explain increased rates of psychiatric disorders in those with older fathers
(defined as fathers 10 years older than average), or if other mechanisms (e.g. delayed fatherhood in men
with a high liability to psychiatric disease) are more plausible. We considered four models: (1) a model in
which new mutations are assumed to act independently to cause psychiatric illness, (2) an equivalent
model in which the rate of causal mutations is also influenced by age-related selfish spermatogonial
selection in the testis ("selfish selection"), (3) a model in which new and existing mutations combine
additively in their effect on liability to psychiatric illness, (4) a model in which paternal age-at-first-child
is correlated with liability for psychiatric illness, and/or in which mating is assortative with respect to
disease liability.
Results In models considering age-related de novo mutations (1,3), the relative risk (RR) to children of
older fathers was low (e.g. <1.05) for all parameter combinations consistent with empirical observations
on the RR to siblings. Thus if a disease is heritable, in the sense of a high RR to siblings, few of the
cases are due to de novo mutations and only some of these are due to mutations in older fathers.
Similarly the RR to offspring due to age-related selfish selection was trivial (e.g. ~1.01) for most
plausible combinations of model parameters. Conversely, a modest correlation (e.g. <0.2) between
paternal age-at- first-child and liability to psychiatric illness recapitulated published estimates (i.e. ~1.5)
of the RR to children of older fathers.
Discussion Our models suggest that shared genetic factors, rather than age-related de novo mutations,
are the primary genetic mechanism underlying the relationship between advanced paternal age and risk of
psychiatric disorders in offspring. However, our simple models do not represent the true complexity of
the aetiology of common psychiatric disorders and it is possible that a more complex and nuanced
combination of factors underlie the epidemiological observations. There is a compelling need for more
primary data to investigate this question and to enable well-powered assessments of model predictions.
Joseph McClay1, Hanzhang Xia1, Lin Ying Xie1, Gaurav Kumar1, Douglas Sweet1, Robin Chan1,
Srilaxmi Nerella1, Karolina Aberg1, Patrick Sullivan2, Edwin van den Oord1
Virginia Commonwealth University, 2University of North Carolina at Chapel Hill
Background The TCF4 locus on chromosome 18 has been consistently associated with schizophrenia
(SZ) in meta-analyses of genome-wide studies. The strength of current evidence suggests that functional
characterization of this gene would be timely. TCF4 encodes a basic helix-loop-helix transcription factor
that recognizes an Ebox motif ('CANNTG'). However, this motif is too small and non-specific to predict
TCF4 binding computationally and no study has yet mapped genome-wide TCF4 binding empirically.
Knowledge of TCF4 binding sites would enable us to build TCF4 regulatory networks and could provide
clues to SZ pathogenic processes. Here, we use chromatin-immunoprecipitation coupled with nextgeneration sequencing (ChIP-seq) to systematically map genome-wide binding of TCF4 in CNS cell
Methods Antibody screening followed ENCODE guidelines, with three antibodies exhibiting clean
Western blots, immunoprecipitate (IP) cross-reactivity and presence of TCF4 in IP as confirmed by mass
spectrometry. All three antibodies were used to perform ChIP in SH-SY5Y cells, a commonly used CNS
model. Each ChIP-seq experiment used 12 million cells, with two duplicate assays per antibody. Mock IP
(IgG) and input DNA controls were used. ChIP DNA was sequenced on a SOLiD 5500xl Wildfire (Life
Technologies) using single-end 50 bp reads. Alignment to hg19, followed by stringent quality control,
yielded on average 36 million uniquely aligned reads per duplicate, or 73 million per antibody. This
greatly exceeded the ENCODE standard 20 million minimum. ChIP-seq analysis was performed using
SPP, while pathway analysis used ConsensusPathDB. Genome-wide gene expression array data (n=72
subjects) was from the Stanley Medical Research Institute post-mortem brain collection.
Results Best Minimal Saturated Enrichment Ratios (MSERs) were with antibody ITF-2 (N-16) (Santa
Cruz Biotech) using mock IP control. MSERs for each duplicate were 2.1 and 3.2, indicating good
saturation, with 781 sites implicated at FDR<0.05. DNA input control comparisons also worked well,
with 80-85% of the top 100 sites typically overlapping between duplicates. The top site detected by all
antibodies was chr15: 51922178-89, located < 10 bp from a known Ebox ('CATGTG') and 7.5 kb
upstream from DMXL2. Other robust findings included binding sites at ZEB2 and NRCAM. Pathway
analysis of genes implicated (В±10 kb) by the top 250 binding sites yielded significant results for
monoamine neurotransmitter degradation (p=4.85x10-5, q=0.013), focal adhesion (p=0.002, q=0.08) and
axon guidance (p=0.003, q=0.08). Finally, genes harboring TCF4 binding sites showed significant
expression differences in brain tissue of subjects with psychosis versus controls (p < 0.01, 10,000
Discussion DMXL2 is involved in regulating Notch signaling, ZEB2 directs migration of cortical
neurons and NRCAM encodes the neuronal cell adhesion molecule. These top genes, plus our pathway
analysis, indicate that TCF4 regulates several genes involved in neuronal development and cell adhesion.
Biologically, this suggests a plausible role for TCF4 in SZ etiology. This role is further supported by our
observation that TCF4-regulated genes show altered expression in patients with psychosis. We plan to
extend our ChIP-seq methods to study TCF4 in other CNS cell types.
Simone Berkel1, Ana deSena2, Franziska Degenhardt3, Birgit Weiss2, Ralph Roeth2, Marcella Rietschel4,
Markus Noethen3, Gudrun Rappold2
Institute of Human Genetics Heidelberg, 2Heidelberg, 3Bonn, 4Mannheim
Background The SHANKs are postsynaptic scaffolding proteins at glutamatergic synapses in the brain
that are essential for proper synapse formation and maintenance. The SHANK gene family (comprising
SHANK1, SHANK2 and SHANK3) is linked to a spectrum of neurodevelopmental disorders, including
intellectual disability and autism spectrum disorders (ASD). Schizophrenia (SCZ) is a neuropsychiatric
disease with high variability in the clinical phenotype, characterized by major impairments in perception
of reality and disorganized thought or behavior. Different studies have already pointed to an impairment
of glutamatergic synaptic plasticity as an underlying cause of SCZ pathology.
Methods To elucidate a putative contribution of genetic SHANK3 variants to the etiology of SCZ, we
sequenced the gene in 500 affected individuals and compared the sequencing results to ancestrally
matched controls.
Results Novel SHANK3 missense variants were identified in 1.6 % of the screened individuals, three of
which were predicted as deleterious by different algorithms. We identified disease risk alleles of 3
uncommon variants, with study-wide (c.4947C>T, P=7xE-06) and genome-wide significance (c.2997C>T,
P=1.4xE-08). Combined with previous studies a rare amino acid exchange G>V was found in 2 out of 685
SCZ patients and in 4 out of 1972 individuals with autism spectrum disorders (ASD), but not in 9082
Discussion We conclude that the SHANK3 gene harbors different genetic variations predisposing to
SCZ, ranging from common and uncommon variants to rare deleterious missense mutations. The
SHANK3-G>V variant found in both ASD and SCZ patients, points to an overlapping genetic
contribution of SHANK3 to both neuropsychiatric disorders.
Jonathan Mill1, Ruth Pidsley2, Joana Viana1, Eilis Hannon1, Helen Spiers2, Claire Troakes2, Safa
Al-Saraj2, Naguib Mechawar3, Gustavo Turecki3, Leonard Schalkwyk2, Nicholas Bray2
University of Exeter, 2King's College London, 3McGill University
Background Schizophrenia is a severe neuropsychiatric disorder that is hypothesized to involve
disturbances in early brain development. The neurobiological mechanisms underlying the disorder remain
largely undefined, and molecular evidence for in utero disturbances in schizophrenia is currently lacking.
Here, we describe a systematic study of schizophrenia-associated methylomic variation in the adult brain
and its relationship to changes in DNA methylation during human fetal brain development.
Methods Our �discovery’ cohort comprised prefrontal cortex (PFC) and cerebellum samples from
schizophrenia patients and matched (for sex, age and sample quality markers (e.g. pH)) control donors
archived in the MRC London Brain Bank for Neurodegenerative Disease. We quantified genome-wide
patterns of DNA methylation using Illumina Infinium HumanMethylation450 BeadChip (450K array).
Bisulfite-pyrosequencing was used to validate the 450K array data for three schizophrenia-associated
differentially methylated positions (DMPs). We subsequently generated a 'replication' 450K array PFC
dataset using schizophrenia and control brains archived at the Douglas Bell-Canada Brain Bank,
Montreal, Canada. Finally, schizophrenia-associated DMPs were tested for an association with brain
development using a unique 450K DNA methylation dataset generated by our lab using human fetal brain
tissue (n=179, range 23-184 days post-conception.
Results We identify significant disease-associated differential DNA methylation at multiple loci,
particularly in the prefrontal cortex (PFC), and confirm these differences in an independent set of adult
brain samples. Our data reveal discrete modules of co-methylated loci associated with schizophrenia that
are highly significantly enriched for genes involved in neurodevelopmental processes. Methylomic
profiling in human fetal cortex samples (spanning 23 to 184 days post-conception) showed that
schizophrenia-associated differentially methylated positions (DMPs) are significantly enriched for loci at
which DNA methylation is dynamically altered during human fetal brain development.
Discussion Our data strongly support the hypothesis that the etiology of schizophrenia has an important
early neurodevelopmental component, with epigenetic mechanisms likely contributing to these
Wednesday, October 15, 2014
1:00 PM - 2:30 PM
Concurrent Oral Sessions
Hyun Ji Noh1, Hyun Ji Noh1, Ruqi Tang2, Jason Flannick1, Colm O'Dushlaine1, Ingegerd Elvers1, Ross
Swofford1, Michele Perloski1, Jeremy Johnson1, S. Evelyn Stewart3, James K. Knowles4, Carol
Mathews5, Guoping Feng2, Jeremiah M. Scharf6, Elinor K. Karlsson7, Kerstin Lindblad-Toh8
Broad Institute of MIT and Harvard, 2McGovern Institute , 3British Columbia Mental Health and
Addictions Research Institute, University of British Columbia, 4Department of Psychiatry and The
Behavioral Sciences, Zilkha Neurogenetic Institute, Keck School of Medicine, University of Southern
California, 5University of California San Francisco, 6Massachusetts General Hospital, 7FAS Center for
Systems Biology, Harvard University, 8Science for Life Laboratory, Uppsala University
Background Obsessive-compulsive disorder (OCD) is a common, complex genetic disorder. A prior
genome-wide association study (GWAS) on 1865 cases and 6357 controls implicated glutamate
signaling, but no SNPs were associated at genome-wide significance. In a natural model (dog), we
found four genes (CDH2, CTNNA2, PGCP, ATXN1), all involved in synapse function and
maintenance, through GWAS and targeted re-sequencing of predisposed dog breeds. Notably, many of
the top human and canine SNPs fall in non-coding regions, and in dogs, we have shown that two such
SNPs have regulatory function in a human neural cell line. In an artificial model, mouse, optogenetic
control of the orbitofronto-striatal pathway governs compulsive behavior, implicating this specific
neural circuit in OCD.
Methods We aimed to identify genes and networks that harbor risk alleles by re-sequencing 592 DSMIV OCD cases and 560 matched controls. We targeted the coding and non-coding, evolutionarily
constrained sequence for 608 genes, selected from genetic studies of OCD and related disorders,
including genes found in the canine OCD GWAS and the orbitofronto-striatal pathway implicated in
mouse. In order to increase power and to account for potential different architecture between coding and
regulatory variants, we evaluated the genic burden of all, coding and regulatory variants separately. We
also evaluated a polygenic burden of variants using 989 gene ontology (GO) sets representing our target
set, covering diverse brain-related functions from synaptic transmission and cytoskeleton to receptors and
ion channels.
Results Five genes had excess variant burdens significantly associated with OCD after multiple testing
procedures (MTP). Two had excess coding variants in cases, two had excess regulatory variants, and one
had excess both variant types. Literature suggests that genes with coding variants involve synaptic
composition while genes with regulatory variants involve synaptic maintenance. The gene with both
variant types involves signaling molecules for cell migration and maturation. Considering rare variants
only (allele frequency<0.01), no significant burden was found, suggesting a major role for common risk
alleles. Among the 989 GO sets analyzed for polygenic burden, neuronal migration and developmental
signaling were nominally associated with OCD (MTP in progress). We also identified a number of
candidate rare variants from the top genes that we are currently validating by genotyping.
Discussion This is the first study to report robust genetic associations to OCD (strict MTP). We were
able to achieve this through small-scale sequencing due to several factors: 1) Our target space included
regulatory regions where many risk alleles reside, capturing many causal variants; 2) Association signals
mainly emerged from common alleles, making the burden test sensitive. Our results are consistent with
recent genome complex trait analysis showing OCD heritability largely arises from common alleles
(although for rare variants our sample size may be underpowered); 3) Separate analyses of coding and
regulatory variants revealed 80% of our genes, implying stratification improved power. In short, our
study shows that regulatory and coding variants are both critical in OCD but may affect different types of
genes. Our approach, which combined insights offered by natural and artificial models of OCD with
targeted human genome sequencing, is yielding new insights into the genetic etiology of OCD.
Laura Huckins1, Konstantinos Hatzikotoulas1, Laura Thornton2, Lorraine Southam1, GCAN GCAN,
David Collier3, Patrick Sullivan2, Cynthia Bulik2, Eleftheria Zeggini1
Wellcome Trust Sanger Institute, 2University of North Carolina at Chapel Hill, 3King's
College London
Background Anorexia nervosa (AN) is a neuropsychiatric disorder presenting with extremely low body
weight, and a marked fear of gaining weight. AN has the highest mortality of any psychiatric disorder,
and affects roughly 0.9% of women. Very little is known about the biological mechanisms which
underlie AN; two small GWAS have been completed and have yet to identify genome-wide significant
hits. No effective medications are available, and treatment outcome for AN remains unacceptably poor.
Methods Our study comprises 2,376 female AN cases and >22,000 controls, genotyped on the CoreExome Chip. Samples derive from eight different populations; care has been taken to ensure that cases
and controls are ancestrally matched. The CoreExome Chip enables us to study both common and lowfrequency variants simultaneously; our study is the first to examine the role played by low-frequency and
rare variants in AN.
Results Analysis is currently complete across three of the eight contributing populations: Norway, with
87 cases and 100 controls; Finland, 163 cases, 5,300 controls; and the UK, 181 cases and 10,034
controls. We have performed a meta-analysis across these three populations and thus far have identified
four genome-wide significant signals: exm370124, exm462797, exm464785, exm2116552. These four
variants are all low frequency, mis-sense variants. We looked at the frequency of these SNPs in both
cases and controls. All SNPs were extremely low frequency in the control populations, with highest MAF
between 0.005 and 0.01. SNPs were also low frequency in the cases, with highest MAF between 0.01 and
0.10. Effect sizes for each SNP were high, and the same direction of effect was noted for every SNP in at
least 2/3 populations. Maximum effect sizes for each SNP were between 6.6 and 74.5.
Discussion One of these associated variants (exm464785) lies in RASGRF2, a gene that has previously
been associated with eating disorders (Wade et al 2013), albeit not at a genome-wide significant level.
This is the first genome-wide significant variant that has been associated with AN. We hope that this will
enable further studies into the functional mechanisms underlying AN, and perhaps be a first step towards
establishing effective medications and treatment. Further, all four hits that have been identified are very
low frequency and could not possibly have been identified in previous GWAS studies. This may be a
good indication that low-frequency, Core-Exome chip type studies have potential to reveal new
associated variants across a range of psychiatric disorders.
T. Bernard Bigdeli1, Stephan Ripke2, Silviu-Alin Bacanu3, Richard L. Amdur4, Aiden Corvin5, Cathryn
M. Lewis6, Robert A. Power7, Andrew McQuillin8, S. Hong Lee9, Naomi R. Wray9, Kenneth S. Kendler3,
PGC Cross-disorder Group, PGC MDD Workgroup, PGC Bipolar Disorder Workgroup, PGC
Schizophrenia Workgroup, Ayman H. Fanous4
VIPBG, 2Analytic and Translational Genetics Unit, Massachusetts General Hospital, 3Virginia Institute
for Psychiatric and Behavioral Genetics, Department of Psychiatry, Virginia Commonwealth University
School of Medicine, 4Mental Health Service Line, Washington VA Medical Center, 5Neuropsychiatric
Genetics Research Group, Department of Psychiatry, Trinity College Dublin, 6Medical Research Council's
Social, Genetic, and Developmental Psychiatry Centre, Institute of Psychiatry, King's College London, 7
MRC Social Genetic & Developmental Psychiatry Centre, Institute of Psychiatry, King's College London,
Molecular Psychiatry Laboratory, Windeyer Institute of Medical Sciences, Research Department of
Mental Health Sciences, University College London, 9The Queensland Brain Institute, The University of
Background Age at onset (AAO) of adult psychiatric disorders is an important clinical indicator of
illness course and outcome, with earlier AAO often associated with disease severity and response to
treatment. Whether onset at very early or relatively advanced ages suggests distinct disease entities
remains unclear, as does the relative importance of familial and environmental factors.
Methods Single-nucleotide polymorphism (SNP) and AAO data were available for 8803 bipolar
disorder (BIP), 9380 major depressive disorder (MDD), and 9354 schizophrenia (SCZ) cases from
respective workgroups of the Psychiatric Genomics Consortium. We performed genome-wide association
studies (GWAS) of AAO within each disorder (case-only), combining these results in cross-disorder
meta- analyses of BIP and MDD (N=18183) and BIP and SCZ (N=18157), and culminating in a metaanalysis of all three disorders (N=27537). We obtained estimates of the proportion of variability in AAO
explained by genome-wide SNPs using the Genome-wide Complex Traits Analysis (GCTA) tool.
Results While no single SNP attained genome-wide significance (P<5Г—10-8) in the all-disorder or
BIP/MDD meta-analyses, combining results for SCZ and BIP yielded several SNPs in vicinity of TGFB1
(19q13.2) that exceeded this criterion (P=1.45Г—10-9). That TGFB1 failed to demonstrate meaningful
association with disease risk in either primary analysis is an important observation, as this suggests a
“modifier” effect. Polygenic scores based on AAO in any one disorder did not significantly predict AAO
in either of the other disorders. We estimated for each disorder the variability in AAO attributable to
common SNPs, revealing that genetic factors explain a significant fraction of the heritability of AAO
among BIP and SCZ (16.17% and 9.39%, respectively) but not MDD cases.
Discussion Our mega-analysis of AAO in BIP, MDD, and SCZ represents the largest case-only GWAS
to date to our knowledge. Our efforts have thus far yielded a single robustly associated candidate locus—
albeit one of particular biological relevance, given an established role for TGFB1 in neurodevelopment.
Using molecular methods, we provide additional support for a genetic basis of AAO in BIP and SCZ but,
notably, not MDD. Taken together, these findings have important implications for the study of
psychiatric genetics, suggesting that case only designs based on careful phenotyping can provide insight
into important genetic mechanisms of disease not assessed by standard case-control studies.
Thomas Muehleisen1, Silke Lux2, Stefan Lenzen2, Tatsiana Vaitsiakhovich3, Svenja Caspers2, Per
Hoffmann4, Stefan Herms4, 1000BRAINS Study Group, Tim Becker5, Katrin Amunts2, Sven Cichon2
Genomic Imaging Group, 2Institute of Neuroscience and Medicine (INM-1), 3Institute for Medical
Biometry, Informatics, and Epidemiology, University of Bonn, 4Division of Medical Genetics,
Department of Biomedicine, University of Basel, 5German Center for Neurodegenerative Diseases
Background Normal aging of the brain is characterized by changes of its structure and function
resulting in decreasing cognitive abilities. However, the age-dependent decline of the cognitive
performance (CP) can vary greatly in elderly subjects of the same age range and in both sexes. Formal
genetic data from family and twin studies provide accumulating evidence that the variability of cognitive
endophenotypes is also influenced by a strong genetic component. So far, there are only a few systematic
genetic studies of cognitive aging and therefore little is known about the underlying genetic factors in the
general population. Here we performed a genome-wide association study (GWAS), in which elderly
subjects showing 'good' CP-profiles were compared with elderly subjects showing 'poor' CP-profiles to
detect variants that contribute to these differences. The profiles comprehensively covered four major
cognitive domains: attention, executive functions, language, and memory.
Methods Subjects represent the first release of a new population-based cohort from Germany with a
research focus on the aging brain, the 1000BRAINS-Study. CP was assessed using 13
neuropsychological tests. Test profiles were analyzed by a cluster analysis and a subsequent discrimant
analysis. DNA was genotyped for single-nucleotide polymorphisms (SNPs) using Illumina microarrays.
SNPs were tested for association using a logistic regression model. SNPs were annotated for regulatory
features of the noncoding genome using HaploReg. Expressed quantiative trait (eQTL) analysis in blood
and 12 brain regions was performed using a linear regression model and RNA sequencing data from the
GTEx Consortium.
Results CP-profile analysis of the total sample revealed a group of 323 'high-performers' (HPs) that
cognitively performed better than a group of 159 'low-performers' (LPs). Of 574K SNPs, none reached
genome-wide significance (P<5E-8). However, 28 SNPs showed strong to moderate evidence for
association with CP-differences (P<5E-5). Overall, the most significant association was observed for a
SNP (P=1.94E-6) that is located between the genes SATB1 and KCNH8. The minor allele (MA) of this
SNP is significantly over-represented in the LPs compared to the HPs (frequency: 38% vs. 26%, odds
ratio: 1.77). In a first follow-up investigation, we found evidence that the SNP overlaps with a
regulatory motiv for a transcription factor (TF). In addition, we found that the MA showed a nominally
significant association with lower expression levels of KCNH8 in the amygdala. On the
neuropsychological level, the MA was associated with a reduction of CP in all four domains with a
minimum in the memory domain.
Discussion The main finding of our GWAS suggests that a common variant on chromosome 3p24.3
contributes to CP in the elderly. The protein-encoding gene, which is located closest to the SNP, is
KCNH8. It is primarily expressed in the nervous system and belongs to a family of voltage-gated
potassium channels that likely are involved in modulating the overall excitability of neurons. Evidences
for a TF binding site and an eQTL support a potentially functional role of the SNP. Our study proposes a
promising new candidate gene for CP. Replication in independent samples is necessary to confirm our
GWAS finding.
Silvia Paracchini1, William Brandler2, Susan Ring3, John Stein2, Joel Talcott4, Simon Fisher5, Caleb
University of St Andrews, 2University of Oxford, 3University of Bristol, 4University of Aston, 5Max
Planck Institute, Nijmegen
Background Humans display structural and functional asymmetries in brain organization, strikingly
manifested through language and handedness. Atypical laterality patterns have been associated with
neurodevelopmental disorders, such as dyslexia and schizophrenia. While we understand the biology of
body asymmetries, the molecular basis of behavioural laterality and brain asymmetries remains mainly
unknown. Understanding the biology of laterality and the link between asymmetries and psychiatric
disorders have been two important research questions investigated for over a century.
Methods We conducted a genome-wide association study (GWAS) for a quantitative measure of
handedness and dexterity (pegboard) in individuals with dyslexia (n = 728) and the general population (N
=2666). The pegboard task involves one individual moving pegs from one row of holes to another. The
time difference (PegQ) between completing the task with the right versus the left hand provides a
measure of relative hand skills. PegQ is normally distributed with a positive mean, indicating most
people are more skilled with their right hand. PegQ strongly correlates with hand preference.
Results The most strongly associated variant, rs7182874 (P = 8.68Г—10-9), is located in PCSK6, a gene
known to activate NODAL, which is required to regulate left/right body axis determination. The
association is specific in the dyslexia cohort. A novel approach for GWAS pathway analysis, based on
gene-set enrichment strategies, showed that left/right asymmetry pathways are associated with
handedness in both the dyslexia and a general population cohorts. In particular, genes involved in
corpus callosum development were enriched among the GWAS top hits. Furthermore, different markers
at the PCSK6 locus were found to be associated with a measure of handiness in a completely
independent study. The dyslexia-specific marker-trait association could be the result of an epistatic
Discussion We report the first gene to be associated with handedness at genome-wide significant level.
PCSk6 has an established role in controlling left/right structural asymmetries suggesting that share
biological pathways are implicated in different aspects of laterality. Furthermore, the dyslexia-specific
association has opened the way to novel hypotheses in studying the link between laterality and
neurodevelopmental disorders at molecular level. Recent findings show that dyslexia candidate genes
play a role in ciliogenesis, an important developmental process at the basis of left/right structural
asymmetries determination. We propose cilia function might have an important role in contributing to
disorders by controlling laterality processes important for brain development. We are now investigating
the molecular mechanisms underlying these observations using neuronal stem cell and zebrafish models.
Christie Burton1, Jennifer Crosbie1, Lauren Erdman1, Annie Dupuis1, Laura Park1, Vanessa Sinopoli1,
Andrew Paterson1, Russell Schachar1, Paul Arnold1
Hospital for Sick Children
Background Obsessive-Compulsive disorder (OCD) is a common (1-2% lifetime prevalence),
debilitating and phenotypically heterogeneous disorder which is highly heritable, particularly when
symptoms begin in childhood or adolescence. Obsessive-compulsive (OC) behaviors are quantitative
traits, continuously distributed in the general population and thus ignoring the quantitative nature of OCD
may reduce power of genome wide association studies (GWAS) comparing OCD patients to controls.
The goal of our study was to conduct a GWAS on a quantitative distribution of OC behaviors measured
in children and adolescents in a community-based sample.
Methods The sample consisted of 17,263 children and adolescents recruited from the community. Selfand/or parent-report data on obsessive-compulsive (OC) behaviors using the Toronto OC scale (TOCS)
was collected from all participants. From 7662 unrelated individuals of Caucasian descent, the
individuals with TOCS total scores in the bottom 10% (N=766) and top 10% (N=766) of the distribution
were genotyped first for the GWAS. The remaining samples are currently being genotyped. Genotyping
was conducted using Illumina HumanCoreExome beadchips. Standard quality control analyses were
conducted using PLINK. Logistic regression analyses using principal components to control for
population structure were conducted. The current results are based a subset of the unimputed data (High
TOCS N= 718 and low TOCS N = 720).
Results Ninety-seven percent of the sample passed QC. Although no genome-wide significant hits were
identified in this preliminary analysis, we identified SNPs with P values as low as 5.0 x 10-7.
Discussion This research shows the feasibility and potential power of using quantitative OC traits rather
than a case control strategy for gene discovery. This will be the first report of a genome-wide study of
quantitative OC traits. On-going analyses include genotyping the entire Caucasian sample and analysis of
the imputed data.
Robert Power1, Karin J.H. Verweij2, Mohamed Zuhair3, Grant W. Montgomery4, Anjali K. Henders4,
Andrew C. Heath5, Pamela A.F. Madden5, Sarah E. Medland4, Naomi R. Wray6, Nicholas G. Martin4
Institute of Psychiatry, London, 2Department of Developmental Psychology and EMGO Institute for
Health and Care Research, VU University, 3MRC Social, Genetic & Developmental Psychiatry Centre,
Institute of Psychiatry, King’s College London, 4Queensland Institute of Medical Research, ,
Department of Psychiatry, Washington University School of Medicine, 6Queensland Brain Institute,
The University of Queensland
Background Cannabis is the most commonly used illicit drug worldwide. With debate surrounding the
legalization and control of use, investigating its health risks has become a pressing area of research. One
established association is that between cannabis use and schizophrenia, a key consideration in the debate
about legislating its use. Although considerable evidence implicates cannabis use as a component cause
of schizophrenia, it remains unclear whether this is entirely due to cannabis directly raising risk of
psychosis, or whether the same genes that increase psychosis risk may also increase risk of cannabis use.
Methods In 2,082 healthy individuals from the Queensland Institute of Medical Research sample of
twins, we show an association between an individual’s burden of schizophrenia risk alleles and use of
cannabis. Polygenic risk scores were derived using the Psychiatric Genomics Consortium’s analysis of
schizophrenia in 9,394 cases and 12,462 controls.
Results A significant association was found with schizophrenia polygenic risk scores when
comparing those who have ever vs. never used cannabis (p=2.6x10-4) and for quantity of use within
users (p=3.0x10-3).
Discussion While directly predicting only a small amount of the variance in cannabis use, these findings
suggest that part of the association between schizophrenia and cannabis is due to a shared genetic
aetiology. This is an important consideration when calculating the impact of cannabis use and its health
Bjarni Vilhjalmsson1, Jian Yang2, Hilary Finucane3, Alexander Gusev4, Sara Lindström4, Stephan Ripke5,
Giulio Genovese6, Nikolaos Patsopoulos7, Po-Ru Loh8, Schizophrenia Working Group of the Psychiatric
Genomics Consortium, Shaun Purcell9, Michael Goddard10, Peter Visscher2, Peter Kraft4, Nicholas
Patterson6, Alkes Price4
Harvard School of Public Health, 2The University of Queensland, Brisbane, Queensland, Australia, 3
Massachusetts Institute of Technology, 4Harvard School of Public Health, 5Massachusetts General
Hospital, 6Broad Institute, Cambridge, 7Brigham & Women's Hospital, 8Harvard School of Public
Health, 9Mount Sinai School of Medicine, 10The University of Melbourne
Background In recent years, polygenic risk scores have become an important technique for detecting a
genetic signal, understanding the genetic architecture of common diseases, and predicting disease risk. As
training sample sizes increase their accuracy is expected to approach the limit set by the heritability
(Chatterjee et al., Nat Genet 2013; Dudbridge et al., PLoS Genet 2013). However, linkage disequilibrium
(LD) between causal markers biases the marginally estimated effect estimates from genome-wide
association studies (GWAS. The standard approach of applying a P-value threshold to association
statistics followed by LD-pruning (LD-clumping) does account for this bias, and yields suboptimal
predictions. This negative impact of LD on the prediction accuracy is expected to increase as GWAS
sample sizes continue to grow.
Methods To address this problem, we propose a Bayesian polygenic risk score that estimates LD from a
reference panel and re-weights the effect estimates obtained from GWAS summary statistics. The new
effect estimates are the posterior mean effects sizes, which give optimal (best linear unbiased prediction)
polygenic risk scores when model assumptions hold. These estimates have a closed-form solution under
the prior where every marker is causal and effect sizes are drawn from a mean-zero normal distribution.
Under a prior where a fraction of the markers are causal, the posterior mean effect sizes do not have a
closed form, and we approximate them with a Markov chain Monte Carlo (MCMC) approach, which we
call LDpred. LDpred requires a small independent validation sample to optimize the fraction of causal
markers. Importantly, LDpred is computationally efficient and yields well-calibrated predictions when
model assumptions hold.
Results We compared LDpred to previously proposed approaches that represent the state of the art in
polygenic risk prediction with GWAS summary statistics as training data. We found LDpred to
outperform other methods in simulations with real and simulated genotypes. The relative improvement
over the commonly used LD-pruning/thresholding approach increased with larger GWAS sample sizes.
We applied LDpred to WTCCC diseases and observed improved prediction R2 for all but one disease,
and for autoimmune diseases that improvement was substantial (e.g. the R2 on the observed scale
improved from 28.3% to 35.5% for T1D). Consistent with our simulations, when applied to
Schizophrenia PGC2 GWAS summary statistics, we observe a significant improvement in prediction R2
(Nagelkerke) from 15.7% to 18.3% using ISC as an independent validation sample. The relative
improvement was even larger when predicting in samples of non-European descent: the R2 (Nagelkerke)
went from 2.0% to 2.6%.
Discussion As sample sizes continue grow, the prediction accuracy of polygenic risk scores are expected
to improve. However, unless the LD between causal markers is appropriately accounted for, the prediction
accuracy will fall short of the upper limit determined by the heritability. LDpred addresses this problem,
resulting in improved prediction accuracies for most diseases that we applied it to. Intuitively, since nonEuropean samples have a different LD pattern than European samples, we expect greater relative
improvement in prediction accuracies for non-European validation samples. However, since it relies on
LD estimates from a reference panel LDpred is not a panacea for polygenic prediction. Although an
improvement, it is inherently limited by how well that reference sample reflects the LD structure in the
GWAS sample. If instead LD information from individual cohorts is available, the prediction accuracy of
LDpred could improve even further.
Lea Davis1, Hae-Kyung Im2, Eric Gamazon2, Emily Kistner-Griffith3, Jim Sutcliffe4, Ed Cook 5, Lauren
McGrath6, International OCD Foundation Genetics Consortium, Tourette Syndrome International
Association Consortium for Genetics, Nancy Cox2
The University of Chicago, 2University of Chicago, 3University of South Carolina, 4Vanderbilt
University, 5University of Illinois, 6American University
Background There has been significant debate over the relative contributions of rare variants and
common polygenic risk to the etiology of psychiatric disorders. We hypothesize that if a liability
threshold can be crossed by either polygenic risk or rare variant risk, we should detect an inverse
correlation between polygenic liability to phenotype and rare variant burden among cases. We use the
estimated breeding value (EBV), which has been in use in animal breeding and plant genetics with
multiple applications for many years, as a quantitative measure of aggregate polygenic liability to
phenotype (Yang et al., 2010; de los Campos et al., 2013). Additionally, we use rare variant burden data
on the same samples, taken from previously published studies of copy number variation and exome
sequencing (Sanders et al., 2011; Sanders et al., 2012; McGrath et al., in press).
Methods In order to test our hypothesis, we calculated EBVs from a matrix that defines genetic
relationships between unrelated cases and controls based on risk allele sharing. Using the same genetic
relationship matrices that yielded trait heritability estimates, we calculated EBVs for Tourette Syndrome
(TS), obsessive-compulsive disorder (OCD), and autism spectrum disorder (ASD) and use these EBVs in
multiple analyses with rare variants available from previous analyses on each phenotype (Sanders et al.,
2011; Sanders et al., 2012, Davis et al., 2013, McGrath et al., in press).
Results In ASD, we show that there is a significant negative correlation between EBV and burden of
novel heterozygous variants (i.e., rare inherited or de novo variants never seen in dbSNP or 1,289 control
whole-exome samples) from exome sequencing (Sanders et al., 2012), r(151) = -.16, p = 0. 05, that
appears to be driven by individuals with below average IQ (<100), r(105) = -.27, p=0.004. Additionally,
we found that ASD probands harboring de novo sequence variants had significantly lower polygenic
scores than those without de novo variants (p=0.05). Moreover, we found an inverse correlation between
EBV and rare (<1%) CNV burden (Sanders et al., 2011) in ASD probands, r(769) = -.07; p=0.05, that
was not detected in unaffected siblings. Similarly, in TS, we find that probands who harbor large CNVs
(> 500 Kb) have significantly lower polygenic load than those who do not carry large CNV events
(p=0.02). We find no such significant differences between OCD probands with and without large CNVs.
Discussion The latter finding is consistent with previous work suggesting a limited role for rare variation
in OCD compared to TS (Davis et al., 2013). Taken together, our data suggests that both sources of
genomic risk are critically important and inversely related. The use of EBVs and the characterization of
an “inverse axis of risk” should facilitate novel strategies for the identification of both genetic and nongenetic risk factors related to disease and disease severity.
Andrew Jaffe1
Lieber Institute for Brain Development
Background The transcriptome of the human brain changes dramatically across development and
aging, with the largest gene expression changes occurring during fetal life, tapering into infancy
(Colantuoni 2011, Kang 2011). Previous transcriptome characterizations used primarily microarray
technologies based on pre-defined probe sequences that capture only a limited proportion of
transcriptome diversity.
The technological advances of RNA sequencing (RNAseq) now permit a flexible and potentially
unbiased characterization of the transcriptome at high resolution and coverage (Trapnell 2010).
Methods We have implemented a method for RNAseq analysis at single base resolution to more fully
characterize transcription dynamics. We performed deep coverage sequencing of the transcriptomes of 72
human dorsolateral prefrontal cortex (DLPFC) samples across 6 important life stages – fetal (2nd
trimester), infant, child, teen, adult and elderly (n=6 per group) – and implemented an annotationagnostic differential expression analysis called "derfinder" to leverage the power of RNAseq without the
difficulties of transcript assembly.
Results We identified 50,650 differentially expression regions (DERs) agnostic of annotation, with
significant and replicated expression changes across fetal and postnatal development. While many DERs
annotated to non-exonic sequence, they were validated in cytosolic mRNA, suggesting that they are not
nuclear pre-mRNAs. We found similar expression profiles of these DERs across 16 diverse human brain
regions and within the developing mouse cortex, and observed expression among subsets of non-exonic
DERs in diverse cell and tissue types. Lastly, we demonstrate that many expression changes are driven by
changing neuronal phenotype related to differentiation and maturation.
Discussion These data highlight conserved molecular signatures of transcriptional dynamics across brain
development, as well as the incomplete annotation of the human brain transcriptome.
Dale Nyholt1, Huiying Zhao1
Queensland Institute of Medical Research
Background Studies have indicated genetic overlap between the five disorders in the Psychiatric
Genomics Consortium (PGC): autism spectrum disorder (ASD), attention deficit-hyperactivity
disorder (ADHD), bipolar disorder (BPD), major depressive disorder (MDD), and schizophrenia
(SCZ). In this study, we aimed to identify specific genes overlapping the five psychiatric disorders
utilizing a novel gene-based approach.
Methods Optimized gene-based tests were performed utilizing genome-wide association (GWA) results
from the PGC analysis of single-nucleotide polymorphism (SNP) data for the five disorders in 33332
cases and 27888 controls of European ancestry. After accounting for correlation (i.e., non-independence)
of gene-based test results due to linkage disequilibrium we examined the significance of the proportion of
genes nominally associated across the five disorders. Pathway and network based analyses were
performed on the sets of genes significantly overlapping the disorders.
Results We found highly significant overlapping genes between SCZ and BPD, moderate overlap
between SCZ and MDD, SCZ and ASD, MDD and ASD, and ADHD and BPD. After combining disorders
as discovery sets, we found significant overlap across SCZ, BPD and MDD, across SCZ, BPD, MDD and
ASD, and across BPD, MDD and ASD/ADHD. No significant overlap was observed between the
individual adult-onset disorders and ADHD. Pathways implicated by the genes overlapping the adultonset disorders include MAPK signalling, calcium signalling, dorso ventral axis formation, chemokine
signaling, and melanogenesis. Pathways for BPD and ASD include glycosphingolipd biosynthesis globo
series, and pathogenic Escherichia coli infection. Pathways implicated by genes overlapping BPD and
ADHD include glutathione metabolism, and arachidonic acid metabolism. Additionally, combining genebased association results across disorders identified numerous genes surpassing our cutoff for genomewide significance.
Discussion Utilizing a novel approach, we identified numerous genes associated across multiple
psychiatric disorders. Our results extend previous findings from single SNP-based genetic overlap
analyses by providing important insight into the likely genes and biological mechanisms underlying the
observed genetic correlation and co-morbidity between these major psychiatric disorders.
Antonio PardiГ±as1, Peter Holmans1, Marian Hamshere1, James Walters1
Cardiff University
Background The genome-wide association study approach (GWAS) is now standard practice in
researching complex psychiatric disorders. One of its key quality control procedures involves the
assessment of population stratification, an important potential confounder of true association signals.
While this is straightforward to perform with Principal Component Analysis (PCA), routinely used in
psychiatric genetics, other alternatives for the inference of population structure exist. Some of these are
explicitly based on different postulates of population genetic theory, claiming to achieve higher accuracy
than PCA. Applied to a case-control GWAS framework, outputs from these techniques could be used to
successfully control for stratification. Furthermore, as these techniques reduce the structure patterns into
one or two variables, there could be an increase in power when using them as the covariates of a logistic
regression, as the number of significant PCs is higher than two in most genotype datasets.
Methods To test these premises, we have used several model-based estimators of population structure,
such as SPA (Yang et al. 2012), SNPWEIGHTS (Chen et al. 2013) or GAGA (Lao et al. 2014), to correct
two case-control association studies of schizophrenia. The first sample contains 3322 cases and 3587
controls from six different ancestries within Europe (ISC; International Schizophrenia Consortium 2009),
and thus contains a great deal of population stratification. The second sample consists of 5616 cases and
6380 controls sampled within the United Kingdom (CLOZUK; Rees et al. 2013), and thus would be
expected to show much less stratification. In both datasets, we have assessed the results of these methods
against other commonly used correction procedures (Stram 2014), which involve PCA using
EIGENSTRAT, Cochran–Mantel–Haenszel stratified analysis and two mixed linear models computed
with GCTA (Yang et al. 2011).
Results After using all the tested methods to correct for stratification, inflation of the regression statistic
was assessed by computing the lambda parameter (Devlin and Roeder 1999). In this assessment, linear
mixed models achieved the lowest inflation values, followed by PCA and the model-based methods.
However, regression results were very similar for all the procedures after using them to correct for
stratification in each of the datasets, with only SPA showing a marked increase on the significance of the
top association signals for the case of the most stratified sample.
Discussion The results of these analyses indicate that the probabilistic model included in SPA might be
useful in a GWAS framework, leading to increased power to detect true associations in some scenarios.
Nevertheless, as this model is best suited to detect gradients in allele frequency, it does not account for all
the possible sources of stratification in a sample. However, this concern is shared with non-model-based
alternatives, and in fact imposes a limit on how much any of them can improve association analyses.
While mixed models could overcome these limitations, the most stringent procedures available also seem
to suffer from overcorrection and power loss, as is consistent with recent discussions on their statistical
properties (Yang et al. 2014).
Daniel Howrigan1, Benjamin Neale1, Kaitlin Samocha1, Jennifer Moran2, Kimberly Chambert2, Sam
Rose2, Menachem Fromer3, Sharon Chandler4, Nan Laird5, Hai-Gwo Hwu6, Wei J. Chen6, Stephen
V. Faraone7, Stephen J. Glatt7, Ming Tsuang5, Steven McCarroll8
Massachusetts General Hospital, 2Broad Institute, 3Mount Sinai School of Medicine, 4University of
California, San Diego, 5Harvard School of Public Health, 6National Taiwan University, 7SUNY
Upstate Medical University, 8Harvard University
Background Increased rates of deleterious de-novo mutations, both across the genome and within
specific genes, have emerged as significant genetic risk factors among developmental disorders such as
autism, intellectual disability, and epilepsy. In contrast, only modest effects of de novo mutation have
been discovered so far among schizophrenia cohorts. Despite the weaker effect size, patterns of observed
de novo mutations are converging on gene networks highly expressed in the brain, and larger samples
will uncover specific genes as putative de novo risk factors for schizophrenia.
Methods Whole-exome sequencing has been performed on 1,141 complete trios from a Taiwanese
cohort, with 1,110 trios having high quality sequence reads sufficient for de novo analysis, thus making it
the largest cohort to date among schizophrenia de novo analyses. Exome sequencing data were generated
using the Illumina HiSeq sequencing with the Agilent SureSelect exome capture platform, achieving an
average coverage of 87% at 20X coverage. Validation of candidate de novo signals was analyzed using
targeted high-throughput genotyping on Illumina HiSeq and Illumina MiSeq platforms with a downsampled mean coverage of 373X. Confirmed de novo mutations were annotated using the NCBI RefSeq
Results Among the Taiwanese cohort, de novo mutation rates per trio and across the exome fall in line
with the expected mutation rate. Using models that incorporate gene size and site-specific mutation rates
into expectations of de novo mutation rates, we do not observe any single gene that surpasses exomewide correction for multiple testing (set at p = 1e-6). We do, however, see a significant enrichment of
genes with multiple non-synonymous de novo mutations (empirical p = 7e-4). When we combine our
results with published de novo studies of schizophrenia, we observe nine genes with multiple loss-offunction events (empirical p < 1e-4), and 87 genes with multiple missense events (empirical p = 0.01).
Gene set analyses also indicate that both loss-of-function and missense de novo mutations are enriched
among targets of the Fragile X mental retardation protein (p = 0.004 and p = 6e-5, respectively) and
among genes under evolutionary constraint (p = 0.001 and p = 2e-5, respectively).
Discussion The current findings do not identify any single gene as an unequivocal risk factor for
schizophrenia when disrupted by de novo mutation; however, aggregate analyses of genes hit with
multiple damaging mutations and among well characterized gene sets in the literature indicate that
significant patterns of de novo risk for schizophrenia are clearly emerging. While we firmly believe
larger cohorts and the accumulation of de novo mutations in the literature will soon lead to specific genes
being unequivocal risk factors, we are well aware that the increased liability toward schizophrenia due to
de novo mutations comprises only a modest fraction of the overall genetic liability to the disorder, and
stress this limitation throughout the presentation.
James Walters1, Stephan Ripke2, Elliot Rees3, George Kirov3, Peter Holmans3, Jennifer Moran2,
Kimberley Chambert2, Ben Neale2, Giulio Genovese4, Steven McCarroll2, Michael Owen3, Michael
Cardiff University, 2Stanley Center for Psychiatric Research, Broad Institute of MIT and Harvard,
Analytical and Translational Genetics Unit, Massachusetts General Hospital, Analytical and
Translational Genetics Unit, Massachusetts General Hospital, 3MRC Centre for Neuropsychiatric
Genetics and Genomics, Cardiff University,
Background A third of those with schizophrenia (Sz) fail to respond to antipsychotic medication and are
termed 'treatment-resistant'. These are the most severely unwell patients and experience marked
impairments in functioning. It is unclear whether treatment-resistance represents a distinct biological subtype of those with Sz or a more severe form of the condition. Similarly little is known about the genetic
architecture underpinning treatment resistant schizophrenia (TRS). In this study we use large-scale
GWAS and CNV data to answer questions relating to the genetic nature of TRS: 1. Are Sz polygenic
scores higher in TRS than in generic Sz (suggesting a more severe form of the condition). 2. Does GWAS
identify risk loci that are specific to TRS (i.e. not seen in generic Sz, suggesting unique biological
pathways to TRS). 3. Do the rates of schizophrenia-associated CNVs differ between TRS and generic Sz
(suggesting an alternative mechanism associated with TRS).
Methods Polygenic risk prediction scores were generated from PGC2 Sz samples excluding our TRS
samples. The TRS sample used for the primary GWAS and as the target set for the polygenic analysis
was the CLOZUK sample of 5600 cases with a clinical diagnosis of TRS and 6300 controls. The
polygenic overlap with CLOZUK was compared to results from generic schizophrenia samples within
PGC. The CLOZUK GWAS was performed as part of the PGC2 Sz GWAS analysis through the same
QC, imputation and analysis pipelines. For the GWAS we then sought replication of the initial CLOZUK
GWAS signals in 2500 independent samples of those taking clozapine. In order to identify genetic
variants specific to TRS we compared the SNPs to emerge from the CLOZUK GWAS in the replication
samples using clozapine users as cases and non-clozapine users as �controls’. CNV analysis was
performed according to standard calling and QC procedures. Rates of CNVs were compared with
available generic schizophrenia samples (ISC, MGS).
Results There was no evidence from the polygenic analysis that the TRS sample is associated with a
stronger polygenic signal than generic Sz samples. In the initial CLOZUK GWAS we identified three
genome-wide significant SNPs (p<5x10-8). Several of these SNPs replicated in the independent samples
of 2500 of those with Sz who have ever taken clozapine. In combining these with the CLOZUK samples
we identified eight genome-wide significant SNPs, three of which appear to be specific to TRS (i.e. no
association in generic schizophrenia samples). The specificity of these signals to TRS will be confirmed
by completion of the clozapine case versus non-clozapine schizophrenia case analysis. Rates of CNVs are
broadly equivalent in the TRS and generic schizophrenia samples, although at present these results are
preliminary given different genotyping platforms and SNP/CNV coverage between samples.
Discussion Our results suggest that there maybe specific polymorphic associations to TRS not seen in
generic schizophrenia and thus point toward the involvement of distinct molecular pathways for TRS. In
contrast there is little support from a polygenic or CNV analyses that TRS represents simply a more
severe form of schizophrenia in that polygenic risk scores and CNV frequencies are broadly equivalent
between TRS and generic schizophrenia samples.
Vanessa Gonçalves1, Andriy Derkach2, Stephanie Willians3, Jennie Peuget4, Andrew D. Paterson5,
Christina Hultman6, Pamela Sklar6, Patrick Sullivan3, Jo Knight4, James Kennedy4, Lei Sun7
CAMH, 2University of Toronto, 3University of North Carolina, 4Neuroscience Section, Centre for
Addiction and Mental Health, 5Program in Genetics and Genomic Biology, Hospital for Sick Children,
Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, 7Biostatistics Division,
Dalla Lana School of Public Health, University of Toronto
Background Schizophrenia (SCZ) is a severe psychiatric disorder with a strong genetic component and
high heritability. Mitochondria are the main sources of aerobic energy for neuronal functioning, and are
involved in many cellular activities disturbed in SCZ such as neurotransmission, synaptic plasticity, and
oxidative stress. We hypothesized that variants in nuclear-encoded mitochondrial genes may influence
SCZ risk. We further hypothesized that variants in the nuclear-encoded oxidative phosphorylation
(OXPHOS) genes could be characteristically different from the other nuclear-encoded mitochondrial
variants due to the direct interaction of their protein products with mitochondrial DNA proteins. We
applied hypothesis-driven analysis in genome-wide association summary results to identify novel
nuclear- encoded mitochondria susceptibility loci for SCZ.
Methods SCZ GWAS results were obtained as summary statistics from the Psychiatric Genomics
Consortium (PGC, Nature Genetics 2011) for a total of 1,252,901 SNPs in 9,394 cases and 12,462
controls of European ancestry. All variants were analyzed using stratified false discovery rate (sFDR).
Stratum 1 contained 1,744 OXPHOS SNPs, stratum 2 contained 20,414 other nuclear-encoded
mitochondrial SNPs, and stratum 3 contained the remaining 1,230,743 GWAS SNPs. Replication of our
association results was performed through meta-analysis with an independent sample (N=11,244).
Results Using a q-value criterion of 0.05, all loci previously reported by PGC remain significant, and
17 novel variants from 6 nuclear-encoded mitochondrial genes: C6orf136, VARS2, FTSJ2, AK3,
C10orf32 and AS3MT were identified. We then replicated the association at C6orf136, VARS2 and
FTSJ2 through meta-analysis. All replication results are based on the same reference allele used in PGC
GWAS and the direction of effect was always consistent.
Discussion Our study showed that novel loci could be discovered from existing GWAS summary data
by performing a hypothesis-driven analysis of the genome, provided a sound biological hypothesis being
conceived for the complex trait of interest. Identification of genetic predisposition to SCZ in the
mitochondrial system may help to reveal other circuits possibly disturbed in major psychosis disorders
and offer new research directions.
Semanti Mukherjee1, Stephan Ripke2, Ole Andreassen3, Aiden Corvin4, Paul deBakker5, Jo Knight6,
Yunpeng Wang3, Steve McCarroll2, Ben Neale2, Vishwajit Nimgaonkar7, Roel Ophoff8, PGC SCZ
Working Group, Jennie Pouget6, Patrick Sullivan9, Todd Lencz10
The Feinstein Institute for Medical Research, 2Broad Institute, 3Oslo University Hospital, 4Trinity
College, 5UMC Utrecht, 6CAMH, University of Toronto, 7University of Pittsburgh, 8UCLA, 9University
of North Carolina, 10Zucker Hillside Hospital
Background The major histocompatibility complex (MHC) has emerged as a region of major interest in
schizophrenia genetics. Large-scale genome-wide association studies (GWAS) in schizophrenia have
converged to demonstrate that the MHC contains the strongest association signal for illness susceptibility.
However, prior GWAS have been unable to precisely localize the source of this signal, due to the
extensive long-range linkage disequilibrium throughout the MHC; different studies have identified top
SNPs ranging across a nearly 10Mb extent (coordinates ranging from 25-35Mb on Chromosome 6). A
subcommittee within the PGC SCZ working group has been formed to parse the signal within the MHC
using HLA imputation and conditional analysis.
Methods Of the 52 PGC-SCZ cohorts, raw genotype data were available for 38 cohorts of European
ancestry, with a total n = 64,631 (29,148 cases and 35,483 controls). Imputation of classical HLA alleles
was performed using SNP2HLA (Jia et al. 2013) applied to raw genotype data from each cohort. A total
of 267 HLA alleles and 5698 SNP markers were successfully imputed. Regression and conditional
regression analyses, covarying for top 10 PCAs and study site, were performed in PLINK and metaanalysis was performed in METAL.
Results Six HLA alleles attained genomewide significance (10-15 < all p-values <10-9): HLA_B_0801,
HLA_DRB1_0301, HLA_DQB1_0201, HLA_A_0101, HLA_DQA1_0501, HLA_C_0701. In each
instance, these alleles were protective, with higher frequencies amongst controls relative to cases
(frequencies in cases ~10.5-11% vs ~12-12.5% for controls; OR~0.87). Frequencies for these alleles,
which form the so-called 8.1 ancestral haplotype (AH8.1) were strongly correlated across cohorts.
Notably, these associations were much less strong than those observed for individual SNPs across the
extended MHC, and conditional analyses covarying for AH8.1 components revealed genomewide
significant SNPs remaining throughout the region, with top signals at rs34661691 (an eQTL for
BTN3A2) and rs2523721 (an eQTL for HLA-A and VARS2). Additional conditional analyses, covarying
for top individual SNPs rather than HLA alleles, are ongoing and results will be presented at the meeting.
Discussion Initial results demonstrate a pattern that is markedly different from that observed for
autoimmune disorders. In most autoimmune disorders, AH8.1 is associated with risk; whereas in
schizophrenia, the opposite relationship is observed. Moreover, in autoimmune disorders, SNP effects in
the MHC are entirely accounted for by HLA alleles and corresponding amino acid changes. By contrast,
SNP effects in schizophrenia are likely to play a significant independent role, probably regulatory in
nature, and potentially implicating non-HLA genes.
Jack Euesden1, Gerome Breen1, Anne Farmer1, Peter McGuffin1, Cathryn Lewis1
King's College London
Background Unusual epidemiological patterns have historically provided a route towards understanding
disease aetiology. Rheumatoid arthritis (RA) is an autoimmune disorder characterized by inflamed,
swollen and ultimately fused joints. Schizophrenia (SCZ) is a psychiatric disorder characterized by
auditory hallucinations, disorganized thought and delusions. Both are highly heritable and both have been
targets of active research within the field of complex disease genetics. Furthermore, both disorders show
an involvement with autoimmune-related alleles leading to models for disease aetiology. A number of
authors have reported an unusual epidemiological relationship between the two disorders, with studies
dating back to 1936 reporting that RA is rarer in SCZ patients than would be expected by chance. Despite
this, many of these studies have been underpowered, lacked appropriate control populations, or found
conflicting results.
Methods We performed a systematic review of studies investigating the prevalence of RA amongst SCZ
patients with reference to an appropriate control population, a meta-analysis of RA prevalence in SCZ
patients versus controls, and used polygenic risk scoring (PRS) to investigate the presence of non-shared
genetic risk between RA and SCZ. We used the most recently publicly available SCZ Genome-Wide
Association Study provided by the Psychiatric Genomics Consortium. This used genotype data on 13,833
SCZ patients and 18,310 controls to report association between SCZ and 9,989,079 single nucleotide
polymorphisms. We calculated polygenic risk scores for SCZ – a measure of an individual’s genetic risk
of disease – in an independent RA case-control dataset. Our RA cases were taken from the WTCCC, and
our RA controls were taken from the RADIANT study. We tested the fit of models regressing RA status
on SCZ genetic risk after accounting for ancestry using principal components.
Results Our systematic review identified 10 studies reported in 9 papers meeting inclusion criteria.
There was heterogeneity amongst studies (p < 0.005), therefore we performed a random effects metaanalysis.
After meta-analysis, we found substantial evidence supporting an unusual epidemiological relationship
between RA and SCZ - unadjusted odds ratio for RA in SCZ patients versus controls = 0.48 (95% CI:
0.34 – 0.67). Polygenic risk scoring, however, revealed that genetic risk for SCZ had no predictive value
on RA status; the most predictive polygenic risk score explained under 0.1% of the variance in RA status
– calculated using Nagelkerke’s Pseudo R2 – and was non-significant (p = 0.085).
Discussion Despite finding robust evidence for a protective effect of SCZ on RA in previous literature,
we did not find evidence for a genetic component to this relationship. We suggest that the observed
relationship may be due to an environmental factor, such as an anti-inflammatory effect of antipsychotic
medication, for which there is substantial evidence. Furthermore, epidemiological studies may be
confounded by the effect of SCZ on life expectancy and the relatively late age at onset of RA. Although
we did not find evidence for a shared genetic architecture, studies utilizing polygenic risk scoring have
begun to stratify the genome based on functional properties of alleles – therefore this may be the next
step in exploring the genetic relationship between these two complex disorders.
Karolina Aberg1, Shaunna L. Clark1, Joseph L. McClay1, Linying Xie1, Gaurav Kumar1, Andrey
Shabalin1, Daniel E. Adkins1, Swedish Schizophrenia Consortium2, Vladimir Vladimirov2, Patrik KE.
Magnusson3, Edwin JCG. van den Oord1
Center for Biomarker Research and Personalized Medicine, School of Pharmacy, Virginia
Commonwealth University, 2Center for Biomarker Research and Personalized Medicine, School of
Pharmacy, and Virginia Institute for Psychiatric and Behavioral Genetics, Virginia Commonwealth
University, 3Department of Medical Epidemiology and Biostatistics, Karolinska Institutet,
Background The methylation of DNA cytosine residues at the carbon-5 position is a common epigenetic
modification that is most often, although not exclusively, found in the sequence context CpG. There are
approximately 1.4 million CpGs that are created or destroyed by common SNPs (SNP-CpGs).
Investigations of SNP-CpGs provide a promising complement to schizophrenia studies of DNA sequence
as these sites involves variation in both sequence and methylation status.
Methods We applied methyl-binding domain 2 protein enrichment to extract the methylated fraction of
the genome followed by sequencing (MBD-seq). Using GWAS genotyping in combination with
1000-genomes imputation we then investigated the methylation status of SNP-CpGs (MAF > 5%) in
DNA extracted from blood in 1408 schizophrenia cases and controls. Next, we performed the same
investigations in post-mortem brain tissue from individuals diagnosed with schizophrenia (N=26), bipolar
disorder (N=22) and controls (N=27). Our main outcome for the post-mortem samples was the presence
of psychotic feature, which allowed us to include all schizophrenia cases and also a subset (N=15) of the
bipolar cases. Findings from the methylome wide SNP-CpG analysis in blood and brain tissue were
further replicated using highly quantitative targeted bisulfite pyrosequencing, in blood and post-mortem
brain tissue from independent schizophrenia case-control samples.
Results We found that 68.3% and 67.7% of SNP-CpGs where likely methylated in blood and brain,
respectively. The overlap of methylated sites was high (94%). Using a false discovery rate (FDR) of 0.1,
the combined analysis of the datasets identified four methylome-wide significant sites (one-sided p
ranging 6.5E-09 to 1.5E-06 with q ranging from 0.002 to 0.08). Thus, sites that were associated with
schizophrenia in blood were also associated with the presence of psychotic feature in brain. Furthermore,
two of the top findings remained significant (one-sided p: 7.8E-09 and 5.9E-07; q: 0.002 and 0.07) when
the brain data was limited to include only schizophrenia cases but not when limited to include bipolar
cases. One of our top findings was located upstream of ENC1, which is of critical importance for
regulation of neuronal processes. This site has been further replicated with pyrosequencing in
independent samples (p-value = 1.6E-04, N=368). Replication of the other three SNP-CpGs is on going.
Discussion This study represent one of the first genome-wide analysis of SNP-CpGs that identifies
associated sites that are consistent between blood and brain and therefore may improve our understanding
of disease etiology and may play a role as biomarkers to improve treatment, diagnosis and prognosis for
schizophrenia patients.
Michael Talkowski1, Marta Biagioli1, Christelle Golzio2, Ian Blumenthal1, Serkan Erdin1, Poornima
Manavalan1, Ashok Ragavendran1, Diane Lucente1, Judith Miles3, Steven Sheridan1, Stephen Haggarty1,
Nicholas Katsanis2, James Gusella1, Michael Talkowski1
Massachusetts General Hospital, 2Duke University, 3University of Missouri
Background Inactivating mutations of the chromodomain helicase CHD8, and many other genes with
diverse functions, act as strong-effect risk factors for autism spectrum disorder (ASD), suggesting
multiple mechanisms of pathogenesis.
Methods We perturbed the transcriptional networks that CHD8 regulates early in neurodevelopment by
suppressing its expression in neural precursors using 5 independent shRNA hairpins, and performing the
experiments in duplicate. We integrated transcriptome sequencing with genome-wide sites of CHD8
binding to chromatin from three independent antibodies.
Results Suppressing CHD8 altered expression of 1,756 genes, most of which were up-regulated (~65%),
consistent with its putative function as a transcriptional repressor. Chromatin immunoprecipitationsequencing (ChIP-seq) revealed widespread pervasive binding of CHD8 throughout the genome, with as
7,324 replicated sites from all three antibodies markinged 5,658 genes, yet just 9.2% of these genes were
differentially expressed. These data suggest that a limited array of direct regulatory effects of CHD8
produces a much larger network of expression changes through secondary indirect regulatory
mechanisms. Interestingly, the networks associated with direction of CHD8 regulation are functionally
distinct. Genes indirectly down-regulated (i.e., without CHD8 binding sites) are strongly enriched for
genes associated with ASD (p = 1.01x10-9) and reflect pathways involved in brain development,
including synapse formation, neuron differentiation, cell adhesion, and axon guidance. In striking
contrast, g
Discussion These data indicate that heterozygous disruption of CHD8 precipitates a network of gene
expression changes that include indirect down-regulation of many other ASD risk genes, and that some
genes associated with ASD and neurodevelopmental disorders may converge on shared mechanism of
Susan Santangelo1, Richard Anney 2, Dan Arking3, Joseph Buxbaum4, Ed Cook5, Nick Craddock6, Mark
Daly7, Ken Kendler8, Phil H Lee7, Ben Neale7, John Nurnberger9, Stephan Ripke7, Jordan Smoller7, Pat
Sullivan10, Jim Sutcliffe11
Maine Medical Center, 2Trinity College, 3The Johns Hopkins University, 4Mount Sinai, Icahn School of
Medicine, 5University of Illinois, 6Cardiff University, 7Harvard Medical School, 8 Virginia
Commonwealth University, 9Washington University, 10University of North Carolina, 11Vanderbilt
Background In 2013, the cross-disorder group (CDG) of the PGC published the largest genetic study of
mental illness ever done, finding substantial evidence for common variants and shared genetic etiology
among all five disorders studied, including SNPs in two calcium channel genes. We also identified some
variants shared by some but not all of the disorders, such as those shared by autism spectrum disorder
(ASD) and schizophrenia. Findings from this and other studies raise questions about how unique and
separable the five psychiatric disorders studied by the PGC are and what the relative strengths of unique
and shared pathophysiologies might be across the disorders.
Methods This presentation will describe how the results of large-scale genomic studies can be used to
inform psychiatric nosology, with particular attention to the diagnosis of DSM-V autism spectrum
disorder. Evidence will be reviewed from genome-wide association studies of psychiatric diagnostic
categories (e.g. ASD, schizophrenia and others) and quantitative phenotypes, using various molecular and
statistical methods, investigations of common variants (SNPs), rare variants (CNVs, SNVs, LOF
mutations, etc.), regulatory elements (i.e., CHD8, miRNAs) and their targets, various pathway analyses
and emerging data from epigenetic studies in animal models and humans.
Results From studies such as the (as yet unpublished) PGC Autism Meta-Analysis, the recent Autism
Genome Project investigation of structural variants, the 2012 exome sequencing efforts, and the current
Autism Sequencing Consortium project, we now know that there are many hundreds, if not more, genes
involved in ASD, that both rare and common variation are important, and that little overlap is seen
between the rare and common variants. However, the hundreds of genes identified so far appear to
converge on a few common biological pathways involved in brain development, synapse function and
chromatin regulation.
Discussion The studies reviewed provide empirical evidence that genetics/genomics can help us move
beyond the design of psychiatric nosological systems based on purely descriptive clinical categories to
those informed by biological factors in disease causation. Further, genetic and genomic studies can
contribute to the prediction, prevention and treatment of psychiatric disorders, such as autism, and to the
identification of molecular targets for new generations of psychotropic drugs, some of which are likely to
cross arbitrarily assigned disease classification boundaries. Discussion will include how best to exploit the
fact that the many hundreds of genes identified for ASD and other psychiatric disorders appear to
converge on a few common biological pathways, and whether the pathways themselves might be treated
as drugable targets.
Kaitlin Samocha1, Silvia De Rubeis2, Xin He3, Arthur Goldberg2, Christopher Poultney2,
Lambertus Klei4, Benjamin Neale1, Stephen Scherer5, Jeffrey Barrett6, David Cutler7, Kathryn
Roeder3, Bernie Devlin4, Mark Daly1, Joseph Buxbaum2
Massachusetts General Hospital, 2Icahn School of Medicine at Mount Sinai, 3Carnegie Mellon
University, 4University of Pittsburgh School of Medicine, 5The Hospital for Sick Children, 6The
Wellcome Trust Sanger Institute, 7Emory University School of Medicine
Background The autism spectrum disorders (ASDs) are characterized by impaired speech and social
interactions. Both common and rare genetic variation contribute significantly to ASD risk. The Autism
Sequencing Consortium has been pursuing trio exome sequencing and using de novo point variants as
anchor for gene discovery. The ASDs also show a notable gender bias with four times as many male
cases as female. It is thought that there is a higher liability threshold for females to be considered autistic,
and that females require a greater genetic burden to be diagnosed with ASD (also known as the female
protective effect). A consequence of this model is that events with the same effect on liability – the same
odds ratio – in males and females will be seen at a higher frequency in female cases. We can therefore
use the frequency difference between male and female cases to estimate the odds ratio of a variant or set
of variants.
Methods Exome sequencing was performed on 3,781 autism cases – 2,297 of which were part of
sequenced trios – and 9,937 ancestry-matched or parent controls. In order to identify sets of genes with
strong associations to ASD risk, de novo variation was combined with transmission and case-control
information in a weighted, Bayesian model named TADA (“Transmission and De Novo Association”; He
et al. 2013).
Results The TADA model identified a set of 33 genes with a false discovery rate (FDR) of < 0.1 and
107 genes with FDR < 0.3. All de novo and transmitted loss-of-function variants in these gene sets were
then separated by the sex of the individual carrying the event. The set of genes with the greatest evidence
for association with ASD (FDR < 0.1) had a significant enrichment of de novo loss-of-function variants
within female cases (p = 0.005). Importantly, the ratio of female to male de novo loss-of-function events
was 2, which indicates an odds ratio of 20 or greater for such events in this set of 33 genes. By contrast,
the set of genes with FDR between 0.1 and 0.3 (n = 74) had much less of an enrichment of de novo lossof-function variants in female cases (ratio of ~1.3), indicating an odds ratio between 2-4 for these events.
The rate of transmitted loss-of-function variants showed no significant difference between sexes for
either gene set.
Discussion These two sets of genes, while both significantly associated with ASD in this large cohort,
appear to function in different ways. The set of genes with FDR < 0.1 represents a small group of genes
in which damaging events, specifically de novo loss-of-function variants, have a high odds ratio. These
genes likely represent key neurodevelopmental processes. This conclusion is supported by the observed
overlap between this set of genes and established risk factors for epilepsy and intellectual disability. The
genes with a smaller odds ratio, on the other hand, have less of an effect on phenotype and may
contribute more to the heritability of ASD.
Beate St. Pourcain1, Dheeraj Raj2, Skuse David3, William Mandy3, Angelica Ronald4, Robert
Plomin5, Jean Golding6, Sue Ring2, Wendy McArdle6, Nicholas Timpson2, John Kemp2, David
Evans2, George Davey Smith2
University of Bristol, 2MRC Integrative Epidemiology Unit, University of Bristol, 3Institute of Child
Health, University College London, , 4Department of Psychological Sciences, Birkbeck, University of
London, 5Social, Genetic & Developmental Psychiatry Centre, Institute of Psychiatry, King’s College
London, 6School of Social and Community Medicine, University of Bristol
Background Maternal smoking during pregnancy has been discussed as a potential risk factor for the
increase of autistic symptoms and autism though findings have been inconsistent. It is possible that
maternal smoking might only elevate the expression of autistic symptoms in genetically susceptible
individuals. Thus, GxE may conceal the true underlying relationship, and explain the inconsistencies in
the literature. Using a genome-wide gene-environment analysis approach, we aimed to identify evidence
for non-additive genetic influences in the architecture of social communication traits during childhood
and adolescence as well as single genetic variant effects, which are specifically involved in a GxE
interaction with respect to maternal smoking during pregnancy.
Methods We studied up to 5628 children (ALSPAC) with information on maternal tobacco use during
the first trimester of the pregnancy, data on social communication problems (Social Communication
Disorders Checklist: 8, 11, 14 and 17 years) and imputed genome-wide genotype information. Follow-up
analyses were performed in 1330 independent children (TEDS, 8 years) with information on maternal
smoking during pregnancy, autism-like traits (total Childhood Asperger Syndrome Test scores) and
imputed genotype information. Statistical analysis included a genome-wide screen for differences in
phenotypic variance per genotype (Levene’s test) to identify a set of variants, which are likely to be
involved in interactions. We then investigated all identified independent signals with respect to maternal
smoking during the first trimester. For comparison, we also carried out a genome-wide screen for logadditive effects at each age using Quasi-Poisson regression.
Results Levene-test screens in ALSPAC showed deviations from the expected distribution under the
null hypothesis for social-communication difficulties at 8 and 11 years of age, compared to QuasiPoisson regression screens, but not later during development. 182 independent SNPs at 8 and 11 years
(SNPs with Levene-P<10-5, and SNPs near ASD susceptibility loci with Levene-P<10-4) were
investigated for interactions with maternal smoking during pregnancy. The strongest interaction effect
(P(E)= 4.4x10-7, P(G)= 0.96, P(GxE)= 0.00016, P(E+GxE)=1.7x10-9) was found at a SNP within the
RBFOX1 gene on 16p13.2 at 8 years of age, though this effect was attenuated during later development.
Risk effects were only present in carriers of the homozygous common allele. There was no evidence for a
main genetic effect. So, far we observed no evidence for replication within a smaller sized follow-up
sample (TEDS) using similar measures (P(GxE)=0.34).
Discussion Our findings suggest that non-additive effects contribute to the genetic architecture of social
communication problems, especially during puberty, although the identification of single genetic variants
might be affected by limited power, temporal variation and the search for the correct interaction partner.
Danielle Posthuma1, Gwen Dieleman2, Christiaan de Leeuw3, Andrea Goudriaan4, Tinca Polderman5,
Matthijs Verhage5, Guus Smit4, Mark Verheijen4, Frank Verhulst6
VU University, 2Department of Child and Adolescent Psychiatry and Psychology, Erasmus University
Medical Center-Sophia Children’s Hospital, 3Department of Complex Trait Genetics, VU University
Medical Centre, Neuroscience Campus Amsterdam, the Netherlands. Institute for Computing and
Information Sciences, Radboud University, 4Department of Molecular and Cellular Neurobiology, Centre
for Neurogenomics and Cognitive Research, Neuroscience Campus Amsterdam, VU University,
Department of Complex Trait Genetics, VU University Medical Centre, Neuroscience Campus
Amsterdam, 6Department of Child and Adolescent Psychiatry and Psychology, Erasmus University
Medical Center-Sophia Children’s Hospital
Background Autism spectrum disorders (ASD) are a highly heritable and heterogeneous group of
disorders characterized by problems with communication, social interaction and repetitive and restricted
behavior. Despite its high heritability, little is known about the genetic pathways underlying ASD and the
contribution of common genetic variants. In this study we systematically investigated the contribution of
common genetic variation in expert-curated sets of genes expressed in the brain to the risk for ASD.
Methods We conducted gene-set analyses using data from healthy parents and their affected probands
(trios) from the Autism Genome Project and Autism Genetic Resource Exchange, and gene-sets that were
informative for specific synaptic and glial (oligodendrocyte, astrocyte, microglia) cell-type functions.
Trio data were available from 3989 ASD patients and their parents and expression data were available for
193 healthy individuals. For secondary functional interpretation, we applied a joint association test to
gene-expression data of cortical structures in healthy controls.
Results We found three significantly associated gene-sets consisting of microglial genes related to cell
proliferation, cell locomotion and taxis, and DNA metabolism. Genetic variation in a subset of microglial
genes from the three significant gene-sets significantly predicted expression levels of microglial genes in
post-mortem cortical samples of healthy subjects.
Discussion Our results indicate that genetic variation in genes related to specific microglial function
contribute to the pathophysiology in ASD. Given earlier observations of head overgrowth and
hyperconnectivity in the brains of autism patients, and recent evidence that microglia cells might
influence synaptic pruning and the number of neural cells in the developing brain, our findings suggest
that genetic variation in microglia partially underlie neural overgrowth.
Lauren Weiss1, Guillaume Desachy2, Anthony Torres3, Martin Kharrazi4, Gerald Delorenze5, Gayle
Windham4, Cathleen Yoshida5, Lisa Croen5
University of California San Francisco, 2UCSF, 3Utah State University, 4California Department of
Health Services, 5Kaiser Permanente Northern California
Background Recently, increased copy number variant (CNV) burden in mothers compared with fathers
of children with autism spectrum disorders (ASDs) and other neurodevelopmental disorders (NDDs) has
been observed. We thus set out to explore this phenomenon, and surprisingly, we observed a higher
autosomal burden of large, rare CNVs in females in the general population, reflected in, but not unique
to, autism families.
Methods We analyzed rare (<1%), large (>30kb) autosomal CNVs from published datasets including
three autism family datasets (Autism Genetic Resource Exchange, Autism Genome Project, Simons
Simplex Collection) and two large control datasets (HapMap, 1000 Genomes) to perform sex
comparisons. We additionally utilized the Early Markers for Autism (EMA) study to compare the CNV
profile between mothers of children with autism and mothers of controls to assess for the first time a sexmatched parental genetic contribution.
Results In autism family genetic datasets, we observed maternal compared with paternal CNV
enrichment. Similarly, in control datasets, we observed female compared with male enrichment for
autosomal large, rare CNVs. Meta-analysis across datasets confirms consistent female excess in CNV
number (P = 2.1 x 10-5) and deletion number (P = 1.1 x 10-3), as well as gene content within all CNVs
= 4.1 x 10-3) and duplications (P = 3.2 x 10-3). Further, the increased burden was primarily in females
with a high individual-level burden, e.g. multiple large, rare CNVs. To exclude a parent-ascertainment
effect (e.g. fertility) from using primarily family-based datasets, we show an increased burden in the
young female compared with male unaffected siblings of autism probands (P = 6.2 x 10-4). In a sexmatched maternal comparison, Wwe observed CNV enrichment in mothers of children with autism
compared to control mothers (P = 0.03), but not in autism probands compared to controls.
Discussion Overall, we found that non-psychiatric samples showed increased female CNV burden. Thus,
we speculate that our data reflect an early developmental decreased male tolerance for high CNV burden,
which might be consistent with increased male fetal loss in the population. Our sex-matched maternal
analysis suggests that autism-specific maternal CNV burden may contribute to high sibling recurrence in
autism. Our results emphasize the importance of sex-matched controls in all genetic studies and open a
novel avenue for studying sexual-dimorphism in the population
2:45 PM - 4:45 PM
Concurrent Symposia Sessions
Chair: Annick Vogels, University Hospitals Leuven, Department of Human Genetics, KU Leuven,
Overall Abstract Details The development of microarray based technologies for comparative genomic
hybridisation (array-CGH) analysis has enabled the detection of submicroscopic microdeletions or
microduplications also referred as copy number variations (CNVs). Over the past 5 years a number of
papers have reported an enrichment of CNVs in neurodevelopmental disorders such as intellectual
disability, autism, attention deficit disorders and schizophrenia. The first speaker will present data of an
international collaborative effort to provide an estimate of the frequency of CNV’s in 600 adults with a
dual diagnosis of intellectual disability and one or more additional psychiatric disorder. Neuropsychiatric
phenotypes including DSM-IV based diagnoses and cognitive functioning were investigated. Microarray
analysis was performed in all adults to uncover the genetic risk factors that may contribute to the
neurodevelopmental phenotypes. A diagnostic yield of 30 % was obtained. Some CNV’s such as
Neurexin1 (NRXN1) are large and have multiple functions in both foetal and adult brain. The fourth
speaker will present data on three unrelated patients affected by dysmorphy, mild intellectual disability
and psychiatric disorder with a 2p16.3 deletion including NRXN1. We suggest that NRXN1 deletion is
not only a risk factor for intellectual disability but is responsible of psychiatric disease with a variable
expressivity. There was a particular high prevalence of 16p11.2 duplications in the group of mildly
retarded adults while 22q11 deletions were relatively common in those with moderate to severe
intellectual disability. The psychiatric phenotype of these CNVs are variable and will be discussed in the
third and fourth presentations respectively.
Griet Van Buggenhout1, Miriam Guitart2, Nick Bass3, Annick Vogels4, Ramon Novell5, Andrew
McQuillin3, Eddy Weyts6, Richard Cayenberghs6, Marina ViГ±as7, Kate Wolfe3, Joris Vermeesch4,
Susanna Esteba-Castillo8, AndrГ© Strydom3
University Hospital Leuven, 2Genetic Laboratory UDIAT-CD, CorporaciГі SanitГ ria UniversitГ ria Parc
TaulГ­, Sabadell, 3University College London4 University Hospitals of Leuven, KU Leuven - University of
Leuven Department of Human Genetics, 5Mental Health & Intellectual Disability Specialized Service,
Psychiatric Hospital Sint-Kamillus, 7Genetic Laboratory, UDIAT-CD, CorporaciГі SanitГ ria
UniversitГ ria Parc TaulГ­, Sabadell, Universitat AutГІnoma de Barcelona, 8Neuropsychology, Mental
Health & Intellectual Disability Specialized Service. IAS, Girona
Individual Abstract The development of microarray based technologies for comparative genomic
hybridisation (array-CGH) analysis has enabled the detection of submicroscopic microdeletions or
microduplications referred to as copy number variations (CNVs). Over the past 5 years a number of
papers have reported an enrichment of CNVs in neurodevelopmental disorders such as intellectual
disability, autism, attention deficit disorders and schizophrenia. The goal of this study is to provide an
estimate of the frequency of CNV’s in adults with a dual diagnosis of intellectual disability and one or
more additional psychiatric diagnosis. This study is the result of an international collaborative effort of
three institutions and combines genomic with phenotypic data to advance the understanding of the
pathogenesis of neurodevelopmental disorders particularly intellectual disability, autism and psychosis.
The consortium provides the largest available sample to date (n = 600) of genetic and phenotypically
characteristics of adults with a dual diagnosis of intellectual disability and psychiatric disorders.
Neuropsychiatric phenotypes including DSM-IV based diagnoses and cognitive functioning were
investigated. Microarray analysis was performed in all 600 adults to uncover the genetic risk factors that
may contribute to the neurodevelopmental phenotypes. A diagnostic yield of 30% was obtained. The
CNV’s at loci 7q11.23, 15q11.2, 15q13, 16p11.2, 16p13.1, 22q11.21, 17p11.2, 3q29 were associated
most frequently with the dual diagnosis of intellectual disability and neuropsychiatric disorders. We will
present an overview of cognitive functioning, the neuropsychiatric phenotypes and CNV’s found in these
600 adults.
Ramon Novell1, Marina ViГ±as3, Susanna Esteba-Castillo2, Elisabet Gabau4, Neus Baena4, NГєria Ribas2,
Miriam Gitart4
Institut Assistencia Sanitaria, 2Ps Mental Health & Intellectual Disability Specialized Service. IAS,
Girona, 3Universitat AutГІnoma de Barcelona, 4CorporaciГі SanitГ ria UniversitГ ria Parc TaulГ­, Sabadell
Individual Abstract Deletions in 2p16.3 region including the neurexin (NRXN1) gene are involved
with intellectual disability and different psychiatric disorders, in particular autism and schizophrenia. We
present three unrelated patients, two adults and one child, affected with dual diagnoses of mild
intellectual disability and psychiatric disorder in whom we identify an intragenic 2p16.3 deletion within
NRXN1 gene using an oligonucleotide array CGH. All three patients had common cognitive features and
a dysmorphic phenotype characterized by long face, deep set eyes and premaxilary prominence. Genetic
analysis of family members showed one de novo and two inherited deletions. An exhaustive
neuropshycological examination of the 2p16.3 deletion carriers revealed borderline intelligence, anxiety
disorder and dysexecutive syndrome. The cognitive profile of dysexecutive syndrome that includes
difficulties in working memory, switch attention, mental flexibility and verbal fluency was the same than
the one observed in the adult probands. We propose that NRXN1 deletion is not only a risk factor for
intellectual disability but is also responsible for the psychiatric disease. The phenotype found in the
2p16.3 deletion carriers suggests the notion that a 2p16.3 deletion has a variable expressivity instead of
an incomplete penetrance.
16P11.2 BP4-5 LOCUS
SГ©bastien Jacquemont1, Anne Maillard2, Loyse Hippolyte2, SГ©bastien Lebon3, AurГ©lien MacГ©4, Sandra
Martin2, AurГ©lie Pain2, Katrin Mannik5, Carina Ferrari6, Eugenia Migliavacca5, Zoltan Kutalik4, Philippe
Conus6, Jacques S Beckmann7, Alexandre Reymond5, SГ©bastien Jacquemont2
CHUV, University of Lausanne, 2University Hospital of Lausanne, CHUV, 3Department of Pediatrics,
CHUV, Lausanne, 4 University of Lausanne, and Swiss Institute of Bioinformatics Lausanne, 5Center for
Integrative Genomics, University of Lausanne, Lausanne, Switzerland, 6Department of Psychiatry,
CHUV, Lausanne, Switzerland, 7Swiss Institute of Bioinformatics
Individual Abstract The 16p11.2 BP4-BP5 deletion (29.5-30.1), one of the most frequent known
genetic etiologies of autism spectrum disorder (ASD), is associated with increased head circumference,
body mass index and language impairment, while the reciprocal duplication is related to schizophrenia
(SZ), decrease in head circumference, and underweight. As observed with other genomic disorders, the
16p11.2 Copy number variants (CNVs) show remarkably variable expressivity, which has hampered
genetic counseling and patient management. The objective of the study is to i) define the medical,
cognitive, and behavioral phenotypes in carriers of the BP4-5 deletion and reciprocal duplication and ii)
characterize clinical manifestations and traits correlated to gene dosage at the 16p11.2 locus. Clinical data
was collected on 310 deletion and 232 duplication carriers and performed detailed neuropsychological
and psychiatric evaluations on 50 deletion and 41 duplication carriers as well as 38 intrafamilial controls.
Effects of the CNVs were estimated using mixed models to account for the effect of kinship. The deletion
demonstrates a consistent impact on neurodevelopment regardless of ascertainment. It lowers full scale
IQ (FSIQ) by two standard deviation, verbal IQ being lower than non-verbal IQ, with a majority of
carriers requiring speech therapy. Epileptic seizures are present in 24% of them. We also identified
increased growth velocity of head circumference during infancy, which recapitulates the welldocumented pattern seen in ASD. The duplication shows a complex effect influenced by ascertainment
methods. It is significantly associated with higher rates of both high- (FSIQ>100) and low-functioning
(FSIQ<50) carriers when compared with deletion carriers (OR=4 and OR=8 respectively, p< 0.01). This
may suggests that the extreme clinical spectrum is the consequence of an interaction between the
duplication and other genetic or environmental factors. The frequency of a psychiatric diagnosis is high
(approximately 80%) in both groups but psychosis was lower than expected in duplication carriers,
possibly due our ascertainment methods. In the deletion and the duplication group, a second copy number
variant was identified in a subset of carriers and the clinical presentation of these individuals sheds light
on the interaction between the 16p11.2 locus and additional CNVs. Reciprocal rearrangements at the
16p11.2 locus represent powerful paradigms to investigate how genetic variants lead to complex
neuropsychiatric phenotypes.
Annick Vogels1, Ann Swillen2, Eddy Weyts3, Richard Cayenberghs3, Griet Van Buggenhout2
University Hospitals Leuven, Department of Human Genetics, KU Leuven, Belgium, 2University
Hospitals of Leuven, Kuleuven-University of Leuven Department of Human Genetics, Leuven, Belgium
Individual Abstract The study described in the first presentation of this symposium tried to estimate the
frequency of copy number variations in adults with neurodevelopmental disorder. Out of a group of 600
adults with intellectual disability and one or more psychiatric disorders, 250 were recruited through a
psychiatric inpatient unit for moderate to severely intellectually disabled. In this population, seven adults
were found to have a previously undiagnosed 22q11deletion. Apart from the intellectual disability, none
of them showed any of the clinical screening criteria for 22q11 Deletion Syndrome (22q11DS). They
showed a wide variety of psychiatric diagnoses including psychotic disorders, mood disorders and severe
aggression. The psychiatric phenotype will be described in the presentation. The average at diagnosis was
high (51 years) and it is unlikely that they would have been diagnosed outside this screening program.
These results raised the possibility that 22q11DS in adults without physical symptoms is more common
than previously reported and easily missed. We therefore performed a retrospective study with the aim to
describe presenting symptoms and age at diagnosis in a large 22q11DS population. A retrospective study
was performed on 65 individuals diagnosed with 22q11DS at adult age. Data were collected on patients
referred to the genetic clinic or actively recruited through systematic diagnostic examination in both
institutions and a psychiatric unit for intellectually disabled. Presenting symptoms were categorized into
seven groups: familial occurrence, intellectual disability, cardiac anomalies, palatal anomalies, facial
dysmorphic features, psychiatric problems and 'other' (comprising all other features associated with
22q11DS). Age at diagnosis was defined as the age at which the 22q11.2 deletion was detected by
fluorescence in situ hybridization or comparative genomic hybridization. Ascertainment subgroups were
different in presenting symptoms and age at diagnosis. Adults were referred to the genetic clinic mainly
because of familial occurrence, cardiac defects and psychiatric disorders whereas adults diagnosed in
institutions for intellectually disabled presented mainly with moderate to severe intellectual disability and
psychotic disorders. Adults diagnosed at the psychiatric unit for intellectually disabled had a variety of
psychiatric disorders but none of them had additional physical features. This emphasizes the need to stay
alert for presenting symptoms such as conotruncal heart defects or moderate to severe intellectual
disability in combination with a history of psychiatric disorders, even in the absence of obvious physical
Chair: Jennie Pouget, Centre for Addiction and Mental Health
Overall Abstract Details Interest in an immunological cause of schizophrenia has been renewed in
recent years due to accumulating evidence from epidemiological, clinical, and genetic studies suggesting
that the immune system is involved in the pathogenesis of schizophrenia. Most notable are the robust
observations that early life infections, autoimmune diseases, antibodies against neuronal proteins,
increased levels of circulating inflammatory cytokine, and genetic variation in the major
histocompatibility complex (MHC) region are associated with schizophrenia. Taken together with recent
neurobiological studies illustrating the important role of immune components (such as MHC class I, IL-6,
and complement) in brain development, the immune disturbances observed in schizophrenia suggest an
underlying immunological cause of the disease in some subset of patients. The immune hypothesis of
schizophrenia is exciting clinically because it may provide an opportunity to leverage biologic
therapeutics currently being developed to treat autoimmune diseases for use in schizophrenia. Previously
it has been difficult to test this hypothesis rigorously, and the majority of evidence has relied on
epidemiological data. The generation of large-scale genome-wide association study (GWAS) and
sequencing data has provided an opportunity to begin testing the immune hypothesis of schizophrenia
rigorously using genetic approaches. This session is designed to illustrate the genetic approaches that can
be used to further our understanding of the role of the immune system in schizophrenia. It will give a first
glimpse of the results of such investigations, and highlight the need for future research in this area.
Jennie Pouget1, Vanessa Gonçalves1, Lei Sun2, James Kennedy1
Centre for Addiction and Mental Health, 2Dalla
Individual Abstract Background: Converging evidence from epidemiological and animal studies
suggests that early-life infections increase the risk of schizophrenia. Genetic susceptibility modulates the
risk of developing schizophrenia following exposure to early-life infection, leading us to hypothesize that
risk variants in immune genes may constitute a necessary "first hit" in order for a "second hit", such as
infection, to cause schizophrenia. To evaluate this hypothesis, we investigated the contribution of 953
known immune genes to schizophrenia using a hypothesis-driven genome-wide association study
(GWAS) approach. Methods: The Psychiatric Genomics Consortium schizophrenia GWAS (N=9,394
cases and 12,462 controls) was analyzed using a univariate approach with all single-nucleotide
polymorphisms (SNPs) weighted equally. Hypothesis-driven analysis of the GWAS data was then
performed using the stratified false-discovery rate (sFDR) method to upweight 15,070 SNPs in 953 genes
with a demonstrated function in immune response based on public annotation databases. Results: None
of the immune SNPs achieved genome-wide significance in the univariate analysis (p>5x10-8). After
upweighting immune gene SNPs using the sFDR method, SNPs in UBD (rs404240, p=1.1x10-6,
qSFDR=0.02), CFB (rs1270942, p=5.7x10-6, qSFDR=0.02), HLA-DQA1 (rs2187668, p=4.4x10-6,
qSFDR=0.02), and HLA-DQB1 (rs2854275, p=4.8x10-6, qSFDR=0.02) were significantly associated
with schizophrenia. Discussion: Incorporating prior biological evidence improved the ability to identify
immune genes important in schizophrenia, with four SNPs in the extended major histocompatibility
complex (xMHC) identified in the immune hypothesis-driven analysis. The xMHC is an 8Mb region of
chromosome 6p previously associated with schizophrenia. Our results highlight the potential importance
of ubiquitin D (UBD) and complement factor B (CFB), which are involved in innate immunity and
complement system activation, respectively. Interestingly, HLA-DQA1 and -DQB1 have previously been
associated with autoimmune diseases. All four of these genes are expressed by microglia, the resident
immune cells of the central nervous system. Our results further highlight the importance of the xMHC in
schizophrenia, and point to microglia as a cell type of particular interest in future functional genetic
Naomi Wray1, S. Hong Lee 2, Enda M. Byrne2, Stephan Ripke 3, Xinli Hu 4, Yukinori Okada5, Eli A.
Stahl 6, Thomas Frissell 7, PGC-SCZ Consortium, RACI Consortium, Swedish SCZ Consortium, Bryan
F. Mowry 2, Soumya Raychaudhuri4
The University Of Queensland, 2Queensland Brain Institute, The University of Queensland, 3Analytic
and Translational Genetics Unit, Massachusetts General Hospital, 4Brigham and Women's Hospital,
Harvard Medical School, 5Graduate School of Medical and Dental Sciences, Tokyo Medical and Dental
University, Tokyo, Japan & Laboratory for Statistical Analysis, RIKEN Center for Integrative Medical
Sciences, Yokohama, Japan, 6 The Department of Psychiatry at Mount Sinai School of Medicine,
Karolinska Institutet
Individual Abstract Background: A long-standing epidemiological puzzle is the reduced rate of
rheumatoid arthritis (RA) in those with schizophrenia (SCZ) and vice versa, made even more puzzling
because smoking is a major risk factor for RA and smoking rates are high in those with SCZ. Traditional
epidemiological approaches to determine if this negative association is underpinned by genetic factors
would test for reduced rates of one disorder in relatives of the other. However, since both disorders affect
only ~1% of the population very large samples of families with multiple family members measured for
both disorders are needed, which are difficult to achieve. The genomics era presents an alternative
paradigm for investigating the genetic relationship between two uncommon disorders using samples of
cases and controls that are unrelated in the classical sense. Methods: We use data from genome-wide
association studies comprising 8064 seropositive cases and 26737 controls for RA and 12,793 cases and
15,912 controls for SCZ. We used the single nucleotide polymorphism (SNP) genotypes to estimate the
genetic similarity between all pairs of individuals, both across the whole genome and for regions
annotated by function. We tested the hypothesis that SCZ cases are genetically different to RA cases.
Results: We estimated a small but significant negative genetic correlation between RA and SCZ of -0.05 (s.e. 0.03,
p=0.04) across the whole genome. The negative correlation increased to -0.17 (s.e.0.071, p=0.007) when only
coding and regulatory regions were considered. Previous analyses of RA have highlighted the importance of genes
expressed in CD4+ effector memory T cells. The correlations were more negative and more significant when the
MHC region was excluded. Discussion: We provide evidence that some genetic risk factors for RA are protective for
SCZ and vice versa and that these risk factors are clustered according to functional annotation. We hypothesize that
these pathways are of particular relevance in the context of environmental immunological challenges. The existence
of antagonistically pleiotropic alleles may explain why common risk variants are maintained in the population.
Steven McCarroll1
Stanley Center for Psychiatric Research, Broad Institute of MIT and Harvard
Individual Abstract We describe an extreme form of structural variation that is present in the human
genome’s HLA locus. We show that parts of the HLA locus segregate in dozens of different structural
forms in human populations, appearing in widely different copy numbers and larger-scale structures. This
form of genetic variation has not been previously ascertained in genome data resources (such as
HapMap), in imputation resources (such as 1000 Genomes), in genome-wide association studies, or in
exome sequencing. We show that this series of structural alleles consists of common, low-frequency, and
rare alleles. These structures have complex linkage disequilibrium relationships with other markers in the
HLA. Of course an exciting possibility is that such variation might be contributing to schizophrenia's
complex and historically confusing pattern of association to markers across the HLA locus. To analyze
the relationship of this novel form of genome variation to human phenotypes, we developed a way to
infer these complex genome structures by imputation from dense SNP data. This has allowed us to test
long allelic series of functional alleles for relationships to both gene expression and clinical phenotypes.
We will describe the results of analyses relating these genome structures to schizophrenia, autoimmune,
and auto inflammatory diseases.
Robert Yolken1, Faith Dickerson2
Johns Hopkins, 2Sheppard Pratt Hospital
Individual Abstract Recent studies indicate that individuals with schizophrenia have evidence of
immune activation that may contribute to disease pathogenesis. The source of this immune activation has
not been identified but is likely to be related to both genetic and environmental components. Recently it
has become apparent that the composition of microbes on mucosal surfaces, termed the microbiome,
represents an important modulator of the immune response in humans and in experimental animals. The
microbiome has been linked to the generation of an aberrant immune response and also been shown to
modulate brain development and behaviour in animal model systems. We employed high throughput
sequencing to characterize the complete oro-pharyngeal microbiome of 41 individuals with schizophrenia
and 32 controls without a psychiatric disorder. We also examined the role of probiotics in modulating the
microbiome. Interim analysis indicates that there are large differences between case and control
individuals in terms of bacterial, viral, and fungal composition. Individuals with schizophrenia had
increased levels of lactic acid bacteria including Lactobacillus casei, Lactobacillus salivarias,
Lactobacillus lactis, and Streptococcus thermophilius as well as several other species of streptococci
including S mitis and S mutans. Several of these bacteria have been associated with altered Th2 immune
responses, an immunological change also noted in schizophrenia. On the other hand, individuals with
schizophrenia had decreased levels of many non-pathogenic bacteria such as strains of Neisseria,
Haemophilus, Prochlorococcus, and Shwanella. Within the group of individuals with schizophrenia,
altered levels of microorganisms were associated with an increased prevalence of the deficit syndrome as
well as increased levels of intestinal immune activation as indicated by antibodies to food and intestinal
antigens. In terms of fungi, individuals with schizophrenia had higher levels of pathogenic yeasts such as
Candida glabrata and Candida tropicalis, but lower levels of the relatively less pathogenic Candida
albicans. We also characterized a number of known human viruses such as Herpesviruses and
Papillomaviruses, as well as bacteriophages and novel viruses. The microbiome was significantly altered
by probiotic therapy, with a tendency towards normalization following treatment. Furthermore, many of
the species which are increased in the oral microbiome of individuals with schizophrenia, such as
streptococci, are modifiable by the administration of antibiotic medications. These studies indicate that
the oral microbiome is altered in individuals with schizophrenia and that the microbiome is a potential
target for novel therapies.
Chair: Kasper Lage, Harvard Medical School, Massachusetts General Hospital, Broad Institute
Overall Abstract Details The recent explosion in genome-wide association studies, exome-sequencing
projects, and epigenetic data sets, have revealed many genetic variants likely to be involved in psychiatric
disease processes, but the composition and function of the molecular systems they affect remain largely
obscure. This limits our progress towards biological understanding and therapeutic intervention. To
deduce function from genetic variation there is a need for systematic approaches that can harness the
power of model systems, functional genomics, medical records, biological networks, patterns of
comorbidity and computation to identify unknown or unexpected pathways perturbed in disease. This
session will highlight methods and experiments being developed in this area and broadly exemplify how
systems biology has been used to analyze and inform genetic variation and phenotype-genotype
relationships. We will introduce specific algorithms, portals, and experimental techniques being applied
in the field, which will enable the audience to apply systems-based approaches and mentality to their
work on psychiatric disorders moving forward.
Kasper Lage1
Harvard Medical School, Massachusetts General Hospital, Broad Institute
Individual Abstract Computational analyses that integrate biological networks (e.g., based on proteinprotein interaction data, gene expression correlations, synthetic lethality relationships, or text mining)
with genetic data have emerged as a powerful approach to functionally interpret large genomic data sets
scalable to the rapid production of data. With a particular focus on psychiatric disorders, this talk will
highlight algorithms, statistics, and web portals being developed in this area and will exemplify how draft
molecular systems involved in many different diseases have been reverse engineered from genomic data.
Moreover, the talk will illustrate how network-based analyses of genetic variants associated with
psychiatric disorders could be used as the starting point for targeted and cost-efficient experiments to
deduce high-resolution networks involved brain signaling.
Shaun Purcell1
Icahn School of Medicine at Mount Sinai
Individual Abstract Large-scale whole-exome sequencing studies of common, complex disease, now an
achievable reality for many groups, hold great promise for connecting genetic risk to the specific,
molecular mechanisms of disease. However, as we will discuss, it is becoming increasingly clear that
such studies will often be unlikely to succeed if each gene is analyzed individually and independently of
its broader genetic and biological context. To address this, network-based representations of exome data
provide a means to capture the dependencies between genes and to test jointly multiple genes for
increased rare variant burden. Applied to a large (N>5000) exome-sequencing study of schizophrenia, in
contrast to geneset-enrichment approaches, we evaluate the performance of several network-based
association methods (using protein-protein interaction data) to a) test for increased connectivity of topranked genes, b) test dynamically-generated sets of neighboring genes jointly, and c) combine genetic
data from different assays (CNV, GWAS and exome data) within this context. We will also present
software that implements the analyses discussed.
SГёren Brunak1
Technical University of Denmark
Individual Abstract Electronic patient records remain a rather unexplored, but potentially rich data
source for discovering correlations between diseases, drugs and genetic information in individual patients
– correlations which may point at the underlying network biology. Such data makes it possible to
compute fine-grained disease co-occurrence statistics, and to link the comorbidities to the treatment
history of the patients. A fundamental issue is to resolve whether specific adverse drug reaction stem
from variation in the individual genome of a patient, from drug/environment cocktail effects, or both.
Temporal analysis of the records can be used to identify ADRs directly from the free text narratives
describing patient disease trajectories over time. ADR profiles of approved drugs can then be constructed
using drug-ADR networks, or alternatively patients can be stratified from their ADR profiles and
compared. Given the availability of longitudinal data covering long periods of time we can extend the
temporal analysis to become more life-course oriented. We also describe how the use of an unbiased,
national registry covering 6.2 million people from Denmark can be used to construct disease trajectories
which describe the relative risk of diseases following one another over time. We show how one can
“condense” millions of trajectories into a smaller set which reflect the most frequent and most populated
ones. References Using electronic patient records to discover disease correlations and stratify patient
cohorts. Roque FS et al., PLoS Comput Biol. 2011 Aug;7(8):e1002141. Mining electronic health records:
towards better research applications and clinical care. Jensen PB, Jensen LJ, and Brunak S, Nature
Reviews Genetics, 13, 395-405, 2012. A nondegenerate code of deleterious variants in mendelian Loci
contributes to complex disease risk. Blair DR, Lyttle CS, Mortensen JM, Bearden CF, Jensen AB,
Khiabanian H, Melamed R, Rabadan R, Bernstam EV, Brunak S, Jensen LJ, Nicolae D, Shah NH,
Grossman RL, Cox NJ, White KP, Rzhetsky A. Cell. 155, 70-80, 2013. Dose-specific adverse drug
reaction identification in electronic patient records, Robert Eriksson R, Werge T, Jensen LJ, Brunak S.
Drug Safety, 37, 237-247, 2014. Temporal disease trajectories condensed from population-wide registry
data covering 6.2 million patients Jensen AB, Moseley PL, Oprea TI, EllesГёe SG, Eriksson R, Schmock
H, Jensen PB, Jensen LJ, Brunak S. Nature Comm, to appear 2014.
Iiris Hovatta1
University of Helsinki
Individual Abstract Anxiety and fear are normal emotional responses to threatening situations.
However, these responses are excessive and prolonged in anxiety disorders, which include panic disorder,
obsessive-compulsive disorder, post-traumatic stress disorder, social and specific phobias, and
generalized anxiety disorder. Anxiety disorders were the most common mental disorders in the EU in
2010. The major challenges in the field are to identify the molecular events that initiate and maintain
pathological anxiety, and determine how to normalize this pathology. Accordingly, there is a need to find
novel, well-defined and clinically relevant drug targets. We have used mouse and human genetic
approaches to identify genes that regulate anxiety. The use of mouse models is supported by two strong
lines of evidence. First, neuroevolutionary studies show that anxiety is an adaptive response conserved in
evolution. Second, several well-validated mouse models are considered appropriate for human anxiety.
We have used RNAseq and miRNAseq of known brain anxiety circuits, followed by bioinformatic
analysis to identify gene pathways involved in anxiety-like behavior. We are using two different mouse
models. The first is a well-established model of innate anxiety, consisting of 6 inbred strains. The
second, called social defeat, is widely used in psychosocial stress studies because such stress is the major
environmental risk factor for anxiety disorders. This model increases anxiety-like behavior, causing
long- term molecular changes in brain. By applying this model to four different genetic backgrounds
allows us to investigate gene-environment interactions behind psychosocial stress-induced anxiety. To
translate results from the mouse models to human anxiety disorders, we rely on DNA-based methods
because of the impracticalities of studying human brain tissue, even post-mortem, from wellcharacterized anxiety disorder patients. We use publicly available mouse genome sequence data to select
preferentially cis- regulated differentially expressed genes from our mouse models to be investigated for
association to anxiety disorders and anxiety symptoms in Finnish epidemiological cohorts.
Chair: Anil Malhotra, The Zucker Hillside Hospital
Overall Abstract Details The proposed symposium will focus on large scale efforts targeted toward
understanding the effects of psychosis risk genes using phenotypes of gene expression (from postmortem
brain), structural and functional neuroimaging, and cognitive performance. Dr. Thomas Hyde will use a
lifespan developmental approach focused upon the identification of abnormalities in the transcriptome by
studying full length and alternative transcripts in large postmortem brain datasets in a host of genes
identified through clinical genome-wide association studies. Such an approach is particularly important
since many risk variants may exert their deleterious effects long before the range of ages typically
associated with the onset of illness. Dr. Ole Andreassen will describe transcriptomic approaches in blood,
and enriched polygenic approaches, which have recently been developed using a Bayesian statistical
framework. He will present findings showing a polygenic pleiotropy between recent schizophrenia
GWAS hits from the PGC, and prefrontal cortical GWAS, strongly indicating a common genetic
mechanisms between schizophrenia disease development and prefrontal cortical abnormalities, the brain
region most often implicated in schizophrenia. Dr. Anil Malhotra will then address the relationship
between genes implicated in schizophrenia by large scale GWAS and normal human cognitive ability.
Moreover, in the context of a large collaborative enterprise, he will discuss the overlap between genes
that influence cognitive ability, as well as specific domains of cognitive function, and genes that
influence schizophrenia risk. Finally, Dr. Aristotle Voineskos will discuss the use of additive genetic
approaches in a study examining relationships among several schizophrenia risk genes, brain structural
connectivity, and cognitive performance. He will show that an additive genetic risk model can predict a
substantial percentage of the variance in brain structural connectivity and cognitive performance among
people with schizophrenia, and will discuss how such an approach is relevant in the clinical domain,
namely via heterogeneity dissection and biological subtyping of disease. Overall, this collection of
presentations brings together the effects of psychosis risk genes on gene expression, neuroimaging, and
cognitive data, which when taken together outline schizophrenia risk pathways. In addition, when these
approaches are applied in patients with schizophrenia, they can be used to predict disease severity. In
turn, these pathways provide converging evidence for specific targets for therapeutic development.
Thomas Hyde1, Joel Kleinman1, Daniel Weinberger1, Andrew Jaffe1, Ran Tao1, Joo Heon Shin1,
Gianluca Ursini1, Paul Harrison2, Helena Cousijns2, Brady Maher1, Sharon Eastwood2, Michelle
Lieber Institute for Brain Development, 2Oxford University
Individual Abstract The neurodevelopmental hypothesis of schizophrenia posits that the pathological
basis of this disorder originates from anomalies in the normal trajectory of brain maturation. Recent
large- scale genome-wide association studies have identified multiple sites of allelic variation associated
with increased risk for schizophrenia. Reconciling the neurodevelopmental hypothesis with recently
identified genetic variations requires a careful molecular interrogation of the developing nervous system.
Many of these genetic variations may exert their deleterious effect on the brain decades before the onset
of illness. For a number of genes, the risk allele for schizophrenia does not appear to alter the expression
of the full- length transcript, either in the fetus or in the adult. The key to understanding the molecular
mechanisms of genetic risk relies upon defining the family of transcripts derived from each risk gene.
High throughput analyses of gene expression using RNAseq and related technologies have identified
transcripts that are preferentially expressed in fetal development. Moreover, risk alleles in multiple genes
are associated with alterations in the expression pattern of these fetal-predominant alternative transcripts.
Schizophrenia- associated alleles in ZNF804a and GAD1 are examples of genetic variations that are
associated with changes in transcripts highly expressed early in brain development. For both genes,
studying the expression of the full-length transcript does not fully explain how the risk allele might alter
gene expression. Instead, the risk allele is associated with significant alterations in the expression of a
truncated transcript that is expressed at high levels early in brain development. The molecular mechanism
of genetic risk only can be explained by defining the library of transcripts derived from each risk gene,
delineating the pattern of the expression of each transcript across the lifespan in non-neurologic nonpsychiatric controls, and then looking for a relationship between genetic variation and transcript
Ole Andreassen1
University of Oslo
Individual Abstract The functional consequences of newly discovered schizophrenia risk genes can be
investigated in vivo in patients, using brain imaging technology and deep phenotyping (cognition,
symptom characteristics). This approach has been fruitful for some individual risk variants, providing
insight of gene effects in clinical samples. Variants in the Major Histocompatibility Complex (MHC) are
associated with larger ventricle size - a MRI characteristic found in many schizophrenia studies.
CACNA1C risk variants have repeatedly been found associated with functional brain abnormalities,
using fMRI technique. The phenotypic picture associated with other variants, such as ZNF804A and
DISC1 seem to be less clear, while TCF4 is associated with a more neurodevelopmental deficit pattern,
associated with earlier age at onset, negative symptoms as well as cognitive dysfunction in schizophrenia.
From a clinical perspective, blood mRNA levels could be a fruitful avenue for biomarker identification,
and recent evidence suggest a series of expression changes related to risk variants of TCF4, ANK3 and
NOTCH4 in severe mental disorders. However, all these variants each confer a small increase in disease
risk. Thus, maybe the most interesting line of results come from enriched polygenic approaches,
investigating the complex effect of a series of risk variants for schizophrenia. We have recently
developed novel Bayesian statistical framework leveraging the polygenic architecture of schizophrenia
and other
complex disorders. We will present recent findings showing a polygenic pleiotropy between recent
schizophrenia GWAS hits from the Psychiatric Genomics Consortium, and prefrontal cortical brain MRI
GWAS. These results strongly indicate a common genetic mechanisms between schizophrenia disease
development and prefrontal cortical abnormalities, the brain region most often implicated in
schizophrenia development.
Anil Malhotra1, COGENT Consortium
The Zucker Hillside Hospital,
Individual Abstract It has long been recognized that generalized deficits in cognitive ability represent a
core component of schizophrenia (SCZ), evident before full illness onset and independent of medication.
The possibility of genetic overlap between risk for SCZ and cognitive phenotypes has been suggested by
the presence of cognitive deficits in first-degree relatives of patients with SCZ; however, until recently,
molecular genetic approaches to test this overlap have been lacking. Within the last few years, large-scale
genome-wide association studies (GWAS) of SCZ have demonstrated that a substantial proportion of the
heritability of the disorder is explained by a polygenic component consisting of many common singlenucleotide polymorphisms (SNPs) of extremely small effect. Similar results have been reported in
GWAS of general cognitive ability. The primary aim of the present study is to provide the first molecular
genetic test of the classic endophenotype hypothesis, which states that alleles associated with reduced
cognitive ability should also serve to increase risk for SCZ. We tested the endophenotype hypothesis by
applying polygenic SNP scores derived from a large-scale cognitive GWAS meta-analysis (~5000
individuals from nine nonclinical cohorts comprising the Cognitive Genomics Consortium (COGENT))
to four SCZ case- control cohorts. As predicted, cases had significantly lower cognitive polygenic scores
compared to controls. In parallel, polygenic risk scores for SCZ were associated with lower general
cognitive ability.
In addition, using our large cognitive meta-analytic data set, we identified nominally significant cognitive
associations for several SNPs that have previously been robustly associated with SCZ susceptibility.
Moreover, we also tested SNPS linked to schizophrenia risk in the COGENT data set, and found modest
evidence that that schizophrenia risk alleles predicted lower performance on tests of generalized cognitive
ability. Additional, domain-based analyses are now being conducted in a new significantly larger data set
and should provide more specific data on the genetic overlap between SCZ and general cognitive ability,
and may provide additional insight into pathophysiology of the disorder.
Aristotle Voineskos1, Tristram Lett2, James Kennedy2, Daniel Felsky2, Benoit Mulsant2, Tarek Rajji2,
Mallar Chakravarty2, Jo Knight2
University of Toronto, 2Centre for Addiction and Mental Health, University of Toronto
Individual Abstract There is growing theoretical and empirical evidence that additive genetic variation
accounts for a considerable percentage of the variance in complex traits. The emerging results of the PGC
coupled with known effects of other genetic variants on brain structure and function provides an
opportunity to use additive risk modeling to obtain a more comprehensive neurobiological understanding
of phenotypic variability among people with schizophrenia. In healthy controls and schizophrenia
patients (N=198), we examined the association between an additive genetic model and brain structure via
brain- wide analysis of cortical thickness (vertex-wise analysis), and white matter FA (tract-based spatial
statistics), as well as cognitive performance. Our additive model included risk alleles with genome-wide
association evidence, namely MIR137 (rs1622579), CACNA1C (rs1006737), ZNF804A (rs1344706);
and risk alleles from genes with well-established effects on brain structure and function, namely GAD1
(rs3749034), and BDNF (rs6265). Voxel-wise white matter FA mediation analysis was performed on
cognitive domains significant associated with additive genetic risk. We found that additive schizophrenia
risk score predicted white matter integrity throughout the brain (pcorrected<0.001), and there was a
significant model-by-diagnosis interaction predominately in the corpus callosum. There was also a
significant vertex-wise interaction between our additive risk score and diagnosis in cortical thickness.
High genetic risk loading predicted poor cognitive performance and the effect was greater among
schizophrenia patients for verbal fluency (F1,64=9.8, p=0.003; interaction, F1,64=4.7, p=0.031) and
motor functioning (F1,64=5.4, p=0.020; interaction, F1,64=10.1, p=0.002)). Voxel-wise FA mediation
analyses showed that genetic risk loading on verbal fluency was caused by white matter changes
predominately in the corpus callosum (Pcorrected[Sobel] < 0.001). Our findings suggest that the additive
genetic risk model that we tested predicts changes in brain structure and cognitive function, and provides
a direct link from genetic variation to white matter FA to cognitive performance.
7:00 PM - 9:00 PM
Concurrent Symposia Sessions
Chair: Ingrid Melle, Division of Mental Health and Addiction, Institute of Clinical Medicine, University
of Oslo
Overall Abstract Details Early stressful life events increase the risk of developing severe mental
disorders and influence post-onset severity indicators, course and outcome. The biological mechanisms
behind these effects including when- and how- the effects of early trauma interact with genetic risk
factors are not fully known. The current symposium focuses on the effect of early trauma spanning the
range from animal studies to clinical studies. The first presentation describes the use of a rat prenatal
stress (PNS) model to investigate molecular and functional changes that could contribute to development
or maintenance of a phenotype based in early life adversity, comprising changes in the expression of
neurotrohpins and epigenetic regulators, changes in prefrontal methylation of gene promotors relevant for
neuronal function and psychiatric disorders including schizophrenia, in addition to a cross-species
approach, as this may allow us to prioritize the list of relevant genes affected by early life adversities.
The second presentation focus on blood transcriptomics; studying mRNA levels of inflammatory
biomarkers and cytokine related pathways in groups exposed to versus not exposed to childhood trauma
across healthy control subjects and patients with depressive disorders; to capture gene expression
changes related to both genetic variation and environmental effects. The third presentation investigates
how different types of childhood trauma influence the clinical expression and severity indicators in 308
patients with bipolar disorders; in particular the interaction between early trauma and known variants of
the serotonin transporter gene. Emotional and sexual abuse was associated with a more severe expression
of the disorder such as an earlier age at onset, increase in suicide attempts, more rapid cycling and greater
proneness to depression. There was an additional effect of 5HTTLPR genotype on time to onset of BD.
The fourth presentation looks at how childhood trauma and a known variant in the gene for the
neurotrophin BDNF (val66met) influence the volume of hippocampal subfields in psychotic disorders. It
has previously been shown changes in hippocampal subfields in patients with schizophrenia and bipolar
disorder relative to controls and that early trauma and BDNF genetic variation influence hippocampal
volume. The study indicates an additive effect of trauma and the BDNF(val66met) on blood BDNF RNA
levels and on hippocampal subfield volume. Taken together; the presentations imply that early adversity
have long term effects on gene expression, affecting neurotrophins, stress response systems and
inflammatory biomarkers with potential influence on brain morphology and severity indicators in severe
mental disorders (bipolar disorder and schizophrenia).
Marco Riva1
University of Milan
Individual Abstract Perinatal life is a period of high plasticity and vulnerability to adverse life
conditions, which may enhance the susceptibility to chronic diseases, including psychiatric disorders. In
particular, exposure to stress during gestation produces complex alterations, including depressive-like
behavior and cognitive defects. With this respect, the use of animal models is instrumental for the
identification of the systems that may be responsible for the occurrence of a pathologic phenotype. On
these bases, we used the rat prenatal stress (PNS) model to investigate molecular and functional
alterations that may contribute to the development or maintenance of the phenotype that originate from
the exposure to early life adversity. At molecular level, PNS rats show a region- and time-specific
reduction in the expression of the neurotrophin BDNF, a marker of neuronal plasticity that has an
important role in mood and cognitive function. BDNF changes are sustained by the modulation of
specific neurotrophin transcripts with the contribution of epigenetic mechanism. We also found that
exposure to PNS produces significant changes in the expression of several epigenetic regulators,
including DNMT1, Gadd45Гџ as well as HDACs. In order to characterize in more details the epigenetic
changes produced in response to PNS, we performed an epigenome-wide analysis in the prefrontal cortex
and hippocampus of male and female rats using a 400K promoter tiling array. A high number of gene
promoters were differentially methylated in PNS rats when compared to control animals, with a highly
significant association for neuronal functions and psychiatric disorders, in particular schizophrenia. We
next employed a convergent cross-species approach to compare the list of genes differentially methylated
in PNS rats with methylation changes identified in a cohort of monkeys exposed to maternal separation as
well as with changes found in CD34+ stem cells derived from cord blood in human neonates whose
mother were grouped on the basis of early life stress exposure. Such analyses allowed us to prioritize the
list of genes that are affected by early life adversities and that may therefore play a relevant role for
psychopathology and disease susceptibility. Our data provide further support to the notion that early life
stress leads to permanent functional and molecular changes in the offspring and highlight the importance
of the identification of methylation signatures that could serve as predictive and diagnostic markers. This
will eventually lead to the identification of novel genes and pathways that contribute to long-term
susceptibility for mental illness and may be a suitable target for pharmacological intervention.
Annamaria Cattaneo1, Alessia Luoni2, Giona Plazzotta2, Marco A. Riva2, Valeria Mondelli3, Patricia
Zunszain3, Carmine M. Pariante3
King's College London, Institute of Psychiatry; IRCCS Fatebenefratelli Brescia; Department of
Pharmacological and Biomolecular Sciences., 2Department of Pharmacological and Biomolecular
Sciences, University of Milan, 3King's College London, Institute of Psychiatry
Individual Abstract It is well known that a history of early life stressful events increases the
vulnerability in the adulthood to develop psychiatric disorders; however, the biological mechanisms
underlying this association require further investigation. This talk will focus on the role of childhood
trauma in causing changes in specific molecular pathways, which persist over time, and thus, are
responsible of increasing the vulnerability to develop depression or other psychopathologies in the
adulthood. Blood transcriptomics captures not only gene expression changes due to genetic variability,
but also those related to the effect of the environment. Thus, it closer reflects the individual phenotype,
and may better represent a promising approach to identify biomarkers associated with increased
vulnerability for psychiatric disorders, and novel targets for pharmacological interventions. By using a
transcriptomic approach we found that control subjects, which were exposed to childhood trauma events,
have higher blood mRNA levels of several inflammatory biomarkers, including: pro-inflammatory
cytokines IL-6, MIF and TNF-a, as well as alterations in cytokines-related pathways. Moreover, we have
found alterations in the mRNA levels of genes involved in the stress response, and in particular of the
mineralcorticoid receptor, of the glucocorticoid receptor and of the serum glucocorticoid kinase (SGK1),
a kinase specifically activated by glucocorticoids. SGK1 mRNA levels are higher in control subjects with
a history of childhood trauma events as compared with subjects without such experiences; higher SGK1
mRNA levels are also observed in depressed patients without early life stressful events and a further
increase can be observed in depressed patients with childhood trauma. This suggests an additive effect of
the two components, illness and trauma, in the modulation of SGK1 mRNA levels. Similarly, SGK1
mRNA levels are increased also in the hippocampus of adult rats, which have been exposed to prenatal
stress, whereas no alterations can be observed during the previous ages. This supports a long lasting
effect of the prenatal stress on SGK1 levels. In order to explain the effect of an early stress on molecular
alterations observed later in life, putative mechanisms involving miRNAs and methylation changes will
therefore be discussed.
Bruno Etain1, Monica Aas2, Frank Bellivier3, Ingrid Melle2, Chantal Henry4, Ole Andreassen5, Marion
INSERM U955, 2Institute of Clinical Medicine, University of Oslo, 3Service de Psychiatrie et
d'addictologie, APHP - Hopital Fernand Widal - LariboisiГЁre, 4Pole de Psychiatrie and Inserm U955, APHP, Groupe Hospitalier Henri Mondor, 5 Institute of Clinical Medicine, University of Oslo; Division of
Mental Health and Addiction, Oslo University Hospital
Individual Abstract The pathophysiology of bipolar disorders (BD) is likely to be partly determined by
environmental susceptibility factors that interact with genetic risk variants. Among them, childhood
trauma has been proposed a relevant environmental factor for BD. However, case-controls studies are
lacking and most studies focused only on physical and sexual abuse (thus neglecting emotional abuse).
Furthermore, the influence of trauma on the clinical expression of the disorder remains to be clarified in
terms of severity of the course, psychopathological dimensions and interaction with genetic
moderators. To investigate these issues, we used a four steps approach. First, we have assessed 206
patients with BD and 94 controls with the Childhood Trauma Questionnaire to perform a case/control
study. Second, 587 patients with BD were consecutively recruited from France and Norway, assessed
using the Childhood Trauma Questionnaire, and characterized for various clinical features. Third, we
studied the interaction between childhood trauma and serotonin transporter gene on the age at onset of
BD in 308 patients.
Finally, we used the Affective Lability Scale and the Affect Intensity Measure to correlate childhood
trauma and adulthood affective instability. Multiple trauma were frequent in patients as compared to
controls (63% versus 33%) and among trauma subtypes only emotional abuse was associated with BD
with a suggestive dose-effect. We then found that emotional and sexual abuses were associated with a
more severe expression of the disorder, as characterized by an earlier age at onset, increased suicide
attempts, more rapid cycling and greater proneness to depression. Emotional and sexual abuses were the
strongest predictors of increased suicide attempts (OR=1.60[1.07-2.39] and OR=1.80[1.14-2.86]
respectively), whilst sexual abuse was the strongest predictor for rapid cycling (OR=1.92[1.14-3.24]).
We then used Cox regression analysis to model the effects of emotional trauma and 5HTTLPR (serotonin
transporter-linked polymorphism) genotypes on time to onset of BD. This model showed that there was a
significant difference in the probability of developing BD between the patients with no emotional neglect
and ll/ls genotype and those with emotional neglect and ss genotype (p=0.003). Finally, we demonstrated
that the higher the exposure to trauma was, the higher the level of affective instability as measured using
the Affective Lability Scale and the Affect Intensity Measure. Our results demonstrate the importance of
childhood trauma, not only as a risk factor for bipolar disorders per se, but also for a more severe clinical
and dimensional profile of expression of the disorder. We also demonstrated for the first time that an
effect of childhood trauma on AAO of BD was observed only in patients who carry a specific stress
responsiveness-related 5HTTLPR genotype.
Monica Aas1, Unn K Haukvik 2, Srdjan Djurovic3, Martin S. Tesli4, Lavinia Athanasiu5, Thomas Bjella5,
Annamaria Cattaeno6, Ole A. Andreassen7, Ingrid Agartz2, Ingrid Melle4
NORMENT, KG Jebsen Centre for Psychosis Research, Division of Mental Health and Addiction, Oslo
University Hospital and Institute of Clinical Medicine, University of Oslo, 2NORMENT, Institute of
Clinical Medicine, University of Oslo, Norway; Department of Psychiatric Research, Diakonhjemmet
Hospital, Oslo, Norway, 3NORMENT, Institute of Clinical Medicine, University of Oslo, Norway;
Department of Medical Genetics, Oslo University Hospital, Oslo, Norway, 4NORMENT, Institute of
Clinical Medicine, University of Oslo, Norway; NORMENT, Psychosis Research Unit, Division of
Mental Health and Addiction, Oslo, 5NORMENT, Psychosis Research Unit, Division of Mental Health
and Addiction, Oslo, 6Institute of Psychiatry, Kings College London, UK; University of Milan, Italy, 7
NORMENT, Institute of Clinical Medicine, University of Oslo, Norway; NORMENT, Psychosis
Research Unit, Division of Mental Health and Addiction, Oslo University Hospital, Oslo, Norway
Individual Abstract Objective: Both childhood trauma and BDNF val66met have been linked to
changes in volume of the hippocampus. Here we investigated additive effects of BDNF val66met-met
and childhood trauma and volume of hippocampal subfields in patients with psychoses. The role of
BDNF RNA on this the association was also investigated. Method: 323 patients with a broad DSM-IV
schizophrenia spectrum disorder or bipolar disorder (meanВ±age: 30.40В±10.76; gender: 54% males;
diagnosis: 56% schizophrenia spectrum were consecutively recruited to the NORMENT, TOP research
study. History of childhood trauma was obtained using the Childhood Trauma Questionnaire. BDNF
RNA was analyzed using standardized procedures. A subsample of n=99, all Caucasians with a broad
DSM-IV schizophrenia spectrum disorder and bipolar disorder (meanВ±age: 32.10В±11.46; gender: 49%
males; diagnosis: 49% schizophrenia spectrum) had data on sMRI. 1.5 T T1-weighted MRI scans were
acquired, and the FreeSurfer software (v 5.2.0) was used to automatically obtain measures of interest
(hippocampal subfields). All sMRI data were corrected for age and gender. BDNF val66met was
genotyped using standardized procedures. Correction of multiple testing was performed. Results: Reports
of childhood trauma, as well as being a carrier of the met variant of the BDNF val66met was
independently, as well as additively, significantly associated with reduced BDNF RNA levels. Moreover,
BDNF val66met- met carriers reporting high levels of childhood trauma demonstrated reduced volume of
the hippocampal subfields CA1, CA2-3 and CA4 DG, which was most prominent in females. Lastly,
BDNF RNA levels were positively associated with hippocampal volume; however statistical predictor
values were low, and only significant in females. Conclusion: Reduced BDNF RNA levels as a result of
both childhood trauma and genetic factors may be part of the complex pathophysiological mechanisms
behind reduced hippocampal subfields in severe mental disorders. This should be investigated further in
larger independent samples.
Chair: Carlos Lopez, University of Antioquia
Overall Abstract Details Endophenotypes in bipolar disorder type I, in a genetically isolated population
There are still doubts about the genetic bases of BDI; one of the ways that can guide the genetic studies is
the distinction between clinical heritable phenotype characteristics and those that are associated with a
disease, as in this way it is possible to propose possible endophenotypes and guide further genetic studies.
This presentation intends to show the results of a study (Fears SC, Service SK, Kremeyer B, et al.
Multisystem Component Phenotypes of Bipolar Disorder for Genetic Investigations of Extended
Pedigrees. JAMA Psychiatry. 2014;():. doi:10.1001/jamapsychiatry.2013.4100.) made in collaboration
by our group (Psychiatry Investigation group, University of Antioquia), UCLA and Costa Rica university
done in two genetically isolated populations(Medellin, Antioquia-Colombia and Central Valley of Costa
Rica) with high prevalence of BDI, with 738 subjects (181 BDI patients), were assessed for
neurocognitive characteristics related with temperament and neuroanatomical changes measured with
neuroimages. This presentation will review the main phenotypes identified and will make a difference
between those who were heritable and those that were associated with the disorder. The results that will
be discussed will provide in the close future the identification of specific clusters for genetic sampling in
Henriette RaventГіs1
Universidad de Costa Rica
Individual Abstract Genetic research in Costa Rica has a long history, starting in the 70s, mostly on
Mendelian inheritance disorders. For the last 25 years, we have been working on complex inheritance
neuropsychiatric disorders such as schizophrenia, bipolar disorder, Alzheimer disease, migraine, and
alcohol abuse, in collaboration with international partners. We have recruited over 5000 subjects, used
different diagnostic instruments and assessments, conducted whole genome scans on these subjects and
identified genetic variants associated to some of these disorders, some of which are now being further
characterized at a functional level. In this presentation, I will review some of these findings, starting with
the mapping and identification of a gene for a dominant form of deafness and moving to our studies on
schizophrenia, bipolar disorders and psychosis. I will describe the population genetic structure studies,
assessment of the phenotype, some of the positive genotype findings and further characterization of the
polymorphisms on candidate positional genes found on schizophrenia, and functional analysis of NRG1
as an example of one of the genes studied, using bioinformatics tools, genome, transcriptome and
proteome analysis, and animal models. Ongoing studies on an extended multigenerational family with
psychosis will also be described.
Homero Vallada1
University of Sao Paulo Medical School
Individual Abstract Brazil is one of the most heterogeneous populations in the world, formed mainly by
the admixture between European, African and Native American populations, and Brazilian studies on
psychiatric genetics has been observed since in the 1970's. A review of the literature, including the
advantages and difficulties of psychiatric genetic studies in the Brazilian population will be presented.
Humberto Nicolini1, Nuria Lanzagorta2, Alma Genis3, Ricardo Aguilar4, Jose Moreno5, Mirna Morales6,
Humberto GarcГ­a6, Lorena Orozco6, Michael Escamilla7
National Institute of Genomic Medicine INMEGEN, 2Carracci Medical Group, 3SAP, INMEGEN,
UACM, 4UACM, 5La Salle University, 6 INMEGEN, 7Center of Excellence for Neurosciences TTUHSC
Paul L. Foster School of Medicine
Individual Abstract Bipolar disorder (BP) has been estimated between 0.8 % to 1% by several studies
(Angst et al., 2004, Simon et al., 2004, Tukel R et al., 2007). Furthermore, comorbid patients have greater
suicidal attempts, substance abuse and lower lithium treatment responses. We now realize that a large
number of genes, each with a small contribution, explain the heritability of most psychiatric disorders,
including BP. In addition, several genes of the immune system have consistently been associated with
mental disorders (MHC, TNF, IL4, IL6). Genome profiling using specific tools directed to the immune
system may provide some additional inside into the common genetic pathways of patients with comorbid
diagnosis. In this study, we will use a genotype array (Golden Gate Custom Array from Illumina) that
covers 484 genes, which participate in the immune system, genotyping 1440 polymorphisms along with
96 ancestry makers. We will genotype 92 patients with BP disorder from Mexico City along with 250
controls from the blood bank in Mexico City with no medical disease.
Statistical Analysis will be performed by the Immunogenetics department in the National Institute of
Genomic Medicine (HG and LO) comparing allele frequencies between cases and controls.
Carlos Lopez1
University of Antioquia
Individual Abstract There is still doubts about the genetics of BDI. A way in which we can guide future
studies (genetic ones) is the finding of special phenotypical characteristics that not only are heritable but
that are associated with BDI, as this can be used to create the so called endophenotypes the ones that can
guide future studies to identify precise genetic abnormalities. The goal of this presentation is to show and
explain the results of a study (Fears SC, Service SK, Kremeyer B, et al. Multisystem Component
Phenotypes of Bipolar Disorder for Genetic Investigations of Extended Pedigrees. JAMA
Psychiatry.2014;71(4):375-387. doi:10.1001/jamapsychiatry.2013.4100.) that analyzed two genetically
secluded subject populations (Costa Rica central Valley and Antioquia-Colombia), with 738 subjects
(151 BDI), in which neurocognitive characteristics related to temperament, and neuroanatomical changes
measured with MRI were assessed. I will present the main findings, and will make emphasis between the
ones that were only heritable and those that were heritable and associated with BDI. These results will
allow us to guide further studies for early diagnosis on certain populations with high risk.
Chair: Bru Cormand, University of Barcelona
Overall Abstract Details Aggression is a basic physiological trait with important roles throughout
evolution, both in defense and predation. However, when expressed in humans in the wrong context,
aggression leads to maladjustment, social impairment and crime. Despite this, knowledge about
aggression aetiology is limited and current treatment strategies are insufficient. The aim of the
"Aggressotype" project, recently funded as part of the EU FP7 Program, is to investigate the biological
basis of both the reactive (emotional, impulsive) aggression and the proactive (instrumental, predator)
presentations of aggression, working in human subjects and in animal models. This involves different
levels of scrutiny, including genetics, brain imaging, epigenetics, work on neuron cell lines derived from
stem cells or cognitive and behavioral assessments. But most important, we are committed to translate
our preclinical findings into predictive, preventive and eventually therapeutic strategies, e.g. by using
animal (mice, zebrafish) and cellular models to identify novel leads for treatment. The Symposia speakers
are members of the Aggressotype consortium, and their presentations will focus on different aspects of
this collaborative effort. Barbara Franke will provide an overview of the aims of the project, with
emphasis on characterizing the differences and communalities of reactive impulsive and low emotional,
instrumental subtypes of aggression. The possible role of specific genes, such as NOS1 and MAOA, in
brain structure and function will be discussed. William Norton will present the zebrafish as a model
organism for studying the aetiology of aggression. Aggression can be reliably measured by recording
stereotypic agonistic postures elicited when the animal is shown its own mirror image. A mediumthroughput screening of more than one hundred compounds will be performed by administering the drugs
by immersion in the tank water, and their effects on fish behavior will be examined. Promising drugs will
be validated in mouse aggression paradigms. Jeffrey Glennon will present work on mouse models of
aggression, including the TPH2 -/- knockout and the BALB/Cj inbred strain. The resident intruder task, a
well-established aggression paradigm, will be used. The aim is to identify neural markers and also
epigenetic moderators of impulsive aggressive behavior, which have remained elusive so far. State-ofthe- art MRI acquisition and analysis methods and microRNA sequencing will be applied. Finally,
Tetyana Zayats will focus on parent of origin effects (POE), an expression of genomic imprinting that is a
source of genetic complexity in common neuropsychiatric conditions. Application of multinomial
modelling may serve to better understand POE -and to establish a distinction from maternal effects- and
will be used to examine over 3,000 trios with ADHD-affected offspring, where aggression can be a
particular problem.
Barbara Franke1
Departments of Human Genetics and Psychiatry, Donder’s Institute for Brain, Cognition and Behaviour,
Radboud University Medical Center
Individual Abstract Aggression, overt and covert behavior with the intention of inflicting physical and
psychological damage, is a physiological trait with important roles throughout evolution, both in defense
and predation. When expressed in humans in the wrong context, however, aggression leads to social
maladjustment and crime. Maladaptive aggression is commonly observed across childhood disruptive
behavioral disorders, in particular in attention-deficit/hyperactivity disorder (ADHD) and conduct
disorder (CD). The aetiology and mechanisms underlying juvenile aggression are still largely elusive.
Clearly, genetic factors in interaction with the environment are at play, but how those change cell
function, neural organization and brain function leading to aggressive behavioral outcomes needs to be
investigated before more effective treatment strategies can be developed. As part of the EU FP7 Program,
the international consortium Aggressotype was funded in 2013 to study the mechanisms underlying
pathological aggression. With its 21 partners, the consortium focuses on characterizing the differences
and communalities of labile, reactive impulsive and low emotional, instrumental subtypes of aggression
in ADHD and CD as a route towards facilitating and improving treatment. The neural substrates of
impulsive and instrumental aggression have been suggested to involve different parts of the prefrontal
cortex and several limbic structures like the striatum and the amygdala. In impulsive aggression the
activation of the amygdala seems to be increased, whereas in instrumental aggression the activity of the
reward system including the striatum is larger than expected. In this presentation, we present work
showing the role of aggression genes in regulating brain structure and activity. We show that NOS1
increases striatal activity during reward anticipation and present data on the role of MAOA in the
connectivity between the amygdala and prefrontal cortex. Data from Aggressotype partners suggest that
both genes are related to impulsive rather than instrumental aggression.
William Norton1, Lauren Jones1
University of Leicester
Individual Abstract Aggression is common side effect of psychiatric disorders including attentiondeficit/hyperactivity disorder (ADHD) and conduct disorder. However, our knowledge about aggression
aetiology is limited and current treatment strategies are insufficient. The zebrafish is an ideal model
organism to address this issue by developing high-throughput drug screens. Aggression can be reliably
measured by recording the stereotypic agonistic postures elicited when a fish is shown its own mirror
image. Fish display aggression from around one month of age onwards allowing large numbers of
animals to be generated in a relatively short time period. Aggression appears to be controlled by similar
genes and neurotransmitters in zebrafish and other vertebrates allowing the translation of data to other
species. Finally, drugs can be administered by immersion (dissolving the drug in the tank water) thus
speeding up application. As part of the Aggressotype project we are undertaking a medium-throughput
screen to identify novel drugs that can modulate aggression levels. We will start by investigating the
ability of existing aggression therapeutics (methylphenidate, risperidone, valproate and lithium) to reduce
aggression levels in fish. We will then choose novel drugs with similar chemical properties and examine
their behavioural function. Our screen will screen a minimum of one hundred novel drugs in the first
year, and will test further compounds time permitting. Promising drugs will be validated in mouse
aggression paradigms, in order to demonstrate a conserved behavioral function across species. This
approach represents an excellent opportunity to identify novel aggression therapeutics, with the global
aim of improving treatments options for patients suffering from psychiatric disorders.
Jeffrey Glennon1, Amanda Jager1, Houshang Amiri1, Armaz Aschrafi2, Arend Heerschap1, Jan Buitelaar1
Radboud University Medical Center, 2Radboud University
Individual Abstract The neural correlates of impulsive aggressive behavior and their epigenetic
moderators have to date remained elusive. Mechanisms such as the neuron specific tyrosine hydroxylase
TPH2 isoform and its yin haplotype have been implicated in the inefficient functional engagement of
cortical areas involved in impulsive control and alterations in the mode of functional connectivity of
dorsal anterior cingulate cortex pathways (Kennedy et al. 2012). Equally, inbred mouse strains such as
the BALB/Cj mouse which display aggressive behavior and reduced sociability have been suggested to
express 20% less TPH2 mRNA and possess 28% fewer TPH2 immunolabeled neurons particularly in the
raphe nuclei and cerebral cortex than control C57Bl/6J mice (Bach et al. 2011). Taken together, this
suggests altered serotonergic transmission may play an important role in impulsive aggression. Our
current work applies state-of-the-art MRI acquisition and analysis methods that enable serial assessment
of whole-brain structural and functional neuronal networks; beginning with the TPH2 -/- and BALB/Cj
mouse models following sub-chronic exposure to the resident intruder task (a well-established aggression
paradigm). Current MR data acquisition of diffusion tensor imaging (DTI) and diffusional kurtosis
imaging (DKI) as well as resting state functional MRI (rs-fMRI) is ongoing to establish neural markers of
impulsive aggression in frontostriatal circuits in both BALB/Cj and TPH2 -/- mice. Others have
demonstrated significant positive regression (p?<?0.001) between social behavior in the BALB/Cj mouse
and fractional anisotropy in the thalamic nuclei, zona incerta/substantia nigra,
visual/orbital/somatosensory cortices and entorhinal cortex (Kim et al., 2011). Furthermore, the same
group reported a significant negative regression (p?<?0.001) between social behavior and mean
diffusivity in the sensory cortex, motor cortex, external capsule and amygdala. Whether aggressive
behavior in the resident intruder task is correlated with the same regions is under investigation.
MicroRNA sequencing of the same frontostriatal pathways which may underlie aggressive behavioral
changes in both models is underway. Preliminary evidence suggests that alterations in microRNAs are
correlated with alterations in frontostriatal connectivity. Whether alteration of the expression of these
microRNAs using RNAi technology results in functional changes in resident intruder task performance is
a topic of active research and will be discussed.
Tetyana Zayats1, Tor-Arne Hegvik1, Johansson Stefan1, Haavik Jan1
University of Bergen
Individual Abstract Epigenetics is an increasingly expanding field, examining alterations in gene
expression caused by mechanisms other than changes in DNA sequence. One of important derivations of
such epigenetic influences are parent of origin effects (POE), that are a recognized source of genetic
ramification in a number of common complex disorders, including those of neurodevelopmental origin.
There is strong evidence for POE in rare genetic syndromes, such as Prader-Willi and Angelman
syndromes, where imprinting is known to occur. However, POE has also been noted in bipolar disorder,
schizophrenia, autism, Alzheimer’s and attention deficit hyperactivity disorder (ADHD). Several of
these neuropsychiatric conditions share an element of dysregulated mood and maladaptive aggression.
Aggression can be a particular problem in ADHD that often coexists with conduct disorder. The
fundament behind POE - genomic imprinting - has generally been examined by? 2 interrogation of
paternal versus maternal transmissions to an affected offspring. However, such approach has been shown
to be insufficient in certain combinations of parental genotypes as well as in detection of maternal effects
(impact of a genetic locus expressed by mothers, but not their offspring), often confound with POE.
Application of multinomial modeling serves further understanding of POE and aids its distinction from
maternal effects that refer to entirely different patterns of gene expression than POE. We will discuss
multinomial modeling of POE and its application to examination of POE in over 3,000 trios with
ADHD- affected offspring.
Chair: Po-Hsiu Kuo, Institute of Epidemiology and Preventive Medicine, NTU
Overall Abstract Details The process of translational sciences involves many steps. One of the major
early steps is to foster discovery, including basic research and proof-of-principle human studies, and then
to pave the way from new discoveries to clinical applications. In the field of psychiatry, the progress of
translational research has been limited so far, hindered by lacking in reproducibly findings of biomarkers
to point a way for improving the accuracy of diagnosis and providing new therapeutic targets. With the
joint efforts from several large consortia in a variety of psychiatric disorders in the past few years, few
common genetic variants are reliably identified for major psychiatric disorders. However, the effect size
of these genetic variants is small, and phenotypic variation explained by these findings is not substantial.
The search for the causes of complex traits, historically has mainly focused on coding genes that are
mutated or genetically altered. Due to recent progress in advanced technologies and detailed
characterization of functional genomic elements, there is a paradigm shift to study transcriptome, which
is in response to both the inheriting genetic sequence and environmental stimuli, as well as to study the
regulation mechanisms for the gene expression levels that are associated with disease status.
Additionally, adoption of novel statistical methods to utilize genomic findings for accurate prediction of
treatment response is desired, which can bridge the gap between knowledge discovery phase and
clinically application phase. This symposium attempts to address a series of topics in relation to different
stages of translational psychiatry, covering a wide range of areas of medical research in humans as well
as using mouse models, including molecular biology, gene expression, epigenetics (in particular
microRNAs mechanism), imaging, and prediction model. More specifically, microRNAs expression and
regulation are studied for psychotic symptoms and schizophrenia. Knock-in and knock-down
experiments are conducted in mouse hippocampus to explore molecular mechanism underlying the
therapeutic effects of electroconvulsive therapy for psychoses. Specific microRNAs expressions are
found to correlate with gray matters structures in schizophrenia. In addition, expression levels of genes
and transcripts, and microRNAs are examined for the acute and remission status of schizophrenia and
bipolar disorder.
Finally, artificial neural network model is applied to identify predictors of antidepressant treatment
response in patients with major depressive disorder.
Yu-Lin Chao1, Chia-Hsiang Chen2, Hwei-Hsien Chen3
Tzu Chi General Hospital, 2Department of Psychiatry, Chang-gung Memorial Hospital, 3Center
for Neuropsychiatric Research, National Health Research Institutes, Taiwan
Individual Abstract Micro-RNAs (miRNAs) are small regulatory RNAs that individually regulate
hundreds of genes. Recent evidence supports a role for miRNA dysregulation in psychiatric disorders,
including schizophrenia, bipolar disorder and autism. MiRNAs may also mediate some of the effects of
psychiatric drug therapies. Methamphetamine (MAP) is a psychotomimetic drug, which can induce
abnormal behaviors in mice. On the other hand, to treat major psychoses, electroconvulsive therapy
(ECT) has been proved to be a highly effective and safe treatment option. However, the underlying
mechanism of ECT action remains largely unknown, and there is no research directly addressing the
molecular mechanisms of miRNAs with the adverse behavioral effects causing by MAP and the
therapeutic effects of ECT. We hypothesize that there are specific miRNAs in brain mediating the
changes of behaviors and brain functions after chronic MAP administration and/or repeated
electroconvulsive shock (ECS). The main goal of this study was to uncover the miRNA-mediated
molecular mechanisms underlying psychotic symptoms. The differentially expressed miRNAs in the
hippocampus of mice pre-treated with MAP and/or ECS were identified via the genome-wide mature
miRNA PCR array quantification. Our results showed that miR-138, miR-328, miR-339-5p and miR-652
were up-regulated by chronic use of MAP and down-regulated after ECS, while the changes of direction
of miR-126-5p and miR-203 were the opposite. These six miRNAs in mouse hippocampus were
significantly correlated with the changes of animal behaviors, such as prepulse inhibition during ECS
interventions. Using in silico prediction for the target genes of the six differentially expressed miRNAs
and pathway analysis, we found several significantly enriched biological pathways that involve with
neuronal synapse and axonal guidance. Moreover, we applied the lenti-viral expression vectors to
perform the transduction in the mouse hippocampus. Preliminary results demonstrated that
overexpression of
miR-328 in bilateral hippocampi of mouse could further impair the deficit of prepulse inhibition and
behavioral sensitization caused by pre-treated MAP. On the contrary, knocking-down of miR-328 could
partially rescue the PPI deficit and decrease the behavioral sensitization. We also identified that both?secretase coding gene BACE1 and post-density 95 protein coding gene DLG4 were targets of miR-328.
They were reported to be associated with the regulation of expression of AMPA receptors in
postsynaptic neurons. Further studies addressing on the underlying molecular mechanisms of behavioral
effects and neuronal function mediated by miR-328 are on-going. Elucidation of the roles of miRNA in
the therapeutic mechanism of ECT would bring new insights into the pathogenesis of psychotic
disorders, and shed some light on the development of new therapeutic agents.
Wei J. Chen1, Chi-Yu Lai2, Su-Yin Lee3, Chun-Chieh Fan4, Ya-Hui Yu5, Chih-Min Liu6, Wen-Yih
Tseng7, Hai-Gwo Hwu6
College of Public Health, National Taiwan University, 2Institute of Epidemiology and Preventive
Medicine, National Taiwan University,; Molecular Psychiatry Laboratory, Florey Institute of
Neuroscience and Mental Health, Parkville, 3Institute of Epidemiology and Preventive Medicine,
National Taiwan University, 4University of California, San Diego, 5Genetic Epidemiology Core, Center
for Genomic Medicine, National Taiwan University, 6College of Medicine and National Taiwan
University Hospital, National Taiwan University; and Graduate Institute
of Brain and Mind Sciences, College of Medicine, National Taiwan University, 7Center for
Optoelectronic Medicine, College of Medicine, National Taiwan University; and Graduate Institute of
Brain and Mind Sciences, College of Medicine, National Taiwan University
Individual Abstract Background: Previously we have identified a seven-miRNA signature (hsamiR-34a, miR-449a, miR-564, miR-432, miR-548d, miR-572 and miR-652) that could
patients with schizophrenia from healthy controls. This study aimed to investigate whether the expression
levels of the seven miRNAs: (1) changed from acute admission to partial remission in inpatients with
schizophrenia; and (2) were associated with gray matter volume, thickness, surface area, as well as
subcortical volume in schizophrenia patients. Methods: Two independent samples were recruited for this
study. For aim 1, a total of 48 patients with schizophrenia were recruited and their peripheral blood
mononuclear cells (PBMC) were collected both at acute admission and at discharge with partial
remission at least two months later. Age- and gender-matched healthy controls (n = 37) were recruited
from university staff and students and similar blood sample collections were performed at two time points
with two months apart. For aim 2, 35 patients with schizophrenia and 12 healthy controls were recruited.
Quantitative real-time PCR was used to quantify the expression level of each miRNAs. T1 weighted
images were obtained through 3T magnetic resonance imaging, while imaging processing was done via
Freesurfer. The Pearson correlation coefficient (with Benjamini and Hochberg false discovery rate
correction at 0.1 level) of each miRNA and gray matter structure in each brain region were calculated for
cases and controls, respectively, as well as the pooled sample of both groups. Results: There were no
significant changes in the peripheral blood levels of the seven miRNAs from baseline to 2-month followup for both schizophrenia inpatients and healthy controls, except miR-548d in healthy controls. Four
miRNAs (miR-34a, miR-449a, miR-548d and miR-572) out of the original 7-miRNA signature were
replicated in showing up-regulation in schizophrenia inpatients in comparison with healthy controls. On
the relationship between cortical gray matter structures, hsa-miR-449a showed moderate positive
association with the volume of right posterior cingulate cortex in cases. Hsa-miR-572 and hsa-miR-652
showed moderate negative association with the thickness of left caudal middle frontal gyrus and left
cuneus cortex. Whereas in healthy controls, only hsa-miR-34a was negatively associated with left
parahimppocamal gyrus and right pars orbitalis in volume/surface area structures. The reduction of most
regions of thickness structures as well as PBMC-miRNA expressions were correlated with increasing
duration of illness in schizophrenia patients. Conclusions: These findings indicate that the aberrant
expressions of the seven miRNAs in PBMC of patients with schizophrenia persist from acute admission
to partial remission, and the miRNA expression alteration in PBMC may be related to the cortical
structural changes that occur with disease progression in patients with schizophrenia.
Ya-Chin Lee1, Ming-Chyi Huang3, Chung-Feng Kao2, Hsi-Chung Chen4, Po-Hsiu Kuo2
Institute of Epidemiology and Preventive Medicine, College of Public Health, National Taiwan
University, 2Institute of Epidemiology and Preventive Medicine, College of Public Health, National
Taiwan University, 3Taipei City Psychiatric Center, Taipei City Hospital, 4Department of Psychiatry &
Center of Sleep Disorders, National Taiwan University Hospital
Individual Abstract More than three-thirds of patients with bipolar disorder (BPD) have recurrent
episodes throughout lifetime, which demonstrating a chronic course with poor prognosis. To explore the
underlying biological mechanisms of bipolar disorder, investigation of the genomic influences for its
episodic features is critical. RNA transcriptome is influenced by both the inheriting genetic sequence and
environmental stimuli, and it has been found that many non-coding RNAs play important roles in the
regulation of human genes and transcripts. We aim to study the change of transcriptome in response to
mania episodic status (acute vs. remission) at the genome-wide level to identify potential biomarkers for
BPD. A discovery sample consists of six BPD patients who have repeatedly measures in symptom
severity and RNA transcriptome at acute manic episode (Young Mania Rating Scale, YMRS score> 20)
and remission at two-month time (YMRS score< 12). We used Affymetrix Human Gene 2.0 ST array to
obtain transcriptome expressions at each time point. This array includes more than 1.35 million probes to
discover about 30,000 transcripts and 11,000 long intergenic non-coding transcripts. We perform quality
control processes, quantile normalization and Lowess normalization on expression data, and build
volcano plot to select candidate genes and transcripts. Potentially differentially expressed
genes/transcripts are validated by quantitative real-time PCR and replicated in an independent set of BPD
patients. We apply hierarchical clustering method to show the patterns of gene expressions in
corresponding to episodic status. Pathway analysis is performed to construct pathway cross-talk for
differentially expressed candidates. Currently, data analysis is still ongoing, and results in this study are
anticipated to shedding light on our understanding about the possible mechanisms underlying episodic
feature of bipolar disorder and providing novel targets for future therapeutic research.
Po-See Chen1
National Cheng Kung University Medical College
Individual Abstract Background: Predicting the treatment response of antidepressants by pre-treatment
features would be of great usefulness for clinical practice since up to 50% of the patients with major
depressive disorder (MDD) do not have response as expected. Here, we demonstrated artificial neural
network (ANN) and linear regression models to identify predictors of antidepressant treatment response
in patients with MDD. Methods: The sample consisted of a reanalysis of 149 MDD outpatients. The use
of back-propagation network (BPN) of ANN was investigated. Randomly, 80% of the subjects were used
to train the ANN model, 20% of whom were used to validate the algorism model. In the structure of the
ANN, inputs contain the information about genetic factors (GNB3 rs5443, BDNF rs6265, 2C19*3
rs4986893, and C2677A rs2032582), biomarker (hs-CRP), and environmental factors (social support
scales: available number of social support in crisis status, and perceived number of social supports in
routine status), and output contains the information of the percentage reduction in 21-item Hamilton
Rating Scale for Depression (HAM-D) scores at week 2. Results: The regression for treatment response
in the training and testing groups were 0.94 and 0.86, respectively. From the sensitivity analysis, BDNF
rs6265, hs-CRP, and available social support in crisis status were three influences on antidepressant
treatment response in chief. However, none of these factors were significantly correlated with treatment
response by linear regression. Conclusions: The complex interactions modeled through ANN may be
more useful to investigate this complexity than linear regression techniques at the clinical level for
predicting individualized response of antidepressants. In addition, further clinical studies will be needed
to validate the accuracy of prediction.
Thursday, October 16, 2014
8:30 AM - 10:30 AM
Concurrent Symposia Sessions
Chair: Naomi Wray, The University Of Queensland
Overall Abstract Details One of the most vexing questions in psychiatric genetics is elucidating the
pathogenesis and genetic architecture of major depression (MDD). In contrast to schizophrenia and
bipolar disorder, efforts to identify the genetic basis of MDD have been complicated by etiological
heterogeneity. A focus on genetic studies of postpartum depression (PPD), a potentially more
homogenous MDD subtype involving exposure to a similar biopsychosocial stressor, may offer
significant advantages. PPD is common, affecting at least 1 in 8 women, and is associated with serious
adverse consequences for the mother, child and family. This symposium will evaluate the validity of
focusing on PPD as a genetically more homogenous MDD subtype by discussing novel research
approaches being applied to this type of investigation. It will also discuss how the genetic findings in
PPD can provide valuable insight for future genetic and biomarker studies of MDD. Speaker 1,
Alexander Viktorin will describe new evidence from the Swedish Twin Study and Swedish National
Patient Registers to analyze the heritability of PPD as compared to MDD outside of the perinatal period.
The heritability of PPD was first estimated in 2321 parous twins using the classical twin-model, and then
was followed by an extended multivariate sibling design including over 1 million parous female siblings:
PPD is more heritable than MDD. Speaker 2, Samantha Meltzer-Brody will discuss the application of
latent class analysis (LCA) to assess the empirical validity of heterogeneity and subtypes of PPD using
data aggregated from a large-scale international perinatal psychiatry consortium (PACT, Postpartum
Depression: Action Towards Causes & Treatment). Using phenotypic data, a 3 class solution yielded the
best fit and the most striking characteristics were severity, timing of onset, comorbid anxiety, and
suicidal ideation in women with PPD. The importance of precise phenotypes as a critical first step toward
future large-scale biological and genetic investigations of PPD will be explored. Speaker 3, Naomi Wray
will describe the use of risk profile scores (RPS), as the best measure of genetic liability as applied to
MDD GWAS datasets with well phenotyped cases of PPD. Using results for MDD and bipolar disorder
from the Psychiatric Genomics Consortium, RPS were created for all individuals and results suggest that
focusing on PPD as a more homogeneous subset of MDD may be an effective strategy for genetic
studies. Finally, the search for a prospective biomarker that predicts PPD will be discussed as well as the
clinical/translational application and impact of this type of finding. Speaker 4, Divya Mehta will discuss a
study of the predictive value of gene expression profiles from peripheral blood samples collected from
women in the third trimester of pregnancy to uncover robust and reproducible biomarkers for PPD. Trine
Munk-Olsen will moderate and Patrick Sullivan will serve as discussant for the symposium.
Alexander Viktorin1, Samantha Meltzer-Brody2, Ralf Kuja-Halkola3, Mikael Landen3, Paul Lichtenstein3,
Patrik Magnusson3
Karolinska Institutet, 2Department of Psychiatry University of North Carolina School of Medicine,
Department of Medical Epidemiology and Biostatistics, Karolinska Institutet
Individual Abstract Perinatal depression (PND) is defined as a major depressive episode that takes
place either during pregnancy or within the first 6 months postpartum. Estimated prevalence is in the
order of 10-15% and the disorder can have devastating effects both for the mother and the child.
Although some research has challenged the view that PND is a separate clinical diagnosis from major
depressive disorder (MDD), there may be a distinct biological basis to PND that may be due to any of the
major changes in hormonal levels associated with pregnancy and parturition. In the current study,
heritability of PND was first estimated in 2321 parous twin women utilizing the classical twin-model,
where PND was defined using a lifetime version of the 10-item Edinburgh Postnatal Depression Scale.
The narrow sense heritability of PND was estimated to be 54% (95% CI, 35-70%), with the remaining
variance attributable to unique environment. This was followed by an extended multivariate sibling
design including over 1 million parous female siblings. Here, hospital discharge diagnoses in the Swedish
National Patient Register were used to define depression, and the timing of the start of the depression
(defined as a) within pregnancy or 6 months postpartum, or b) any other time) was used to separate the
two disorders. The heritability of PND was estimated to be 49% (95% CI, 39-60%), with the remaining
variance attributable to unique environment. The heritability of MDD was estimated to 28% (95% CI, 1838%), with the remaining variance attributable to shared environment (7%; 95% CI 2-11%) and unique
environment (65%; 95% CI, 59-71%). Further, bivariate analysis revealed that the variance in PND was
explained by 14% common genetic factors with MDD and 27% unique genetic factors for PND, lending
evidence to a partially different genetic etiology of PND and MDD.
Samantha Meltzer-Brody1, the PACT Consortium,
University of North Carolina at Chapel Hill
Individual Abstract Background/Objective: Postpartum Depression (PPD) confers substantial
morbidity and mortality but the definition of PPD is a matter of some controversy. PPD is categorized as
a subtype of major depression (MDD) in DSM-5, and, as such, a diagnosis of PPD requires that DSM
criteria are fulfilled during the specified perinatal period. Additionally, the phenotypic presentation of
PPD may have distinguishing characteristics compared to an episode of MDD occurring outside of the
perinatal period.
Therefore, this study is an empirical investigation of PPD heterogeneity to identify clinical subtypes.
Data were aggregated from the international perinatal psychiatry consortium, PACT (Postpartum
Depression: Action Towards Causes and Treatment). PACT members are from 24 institutions in 7
countries, and had 27,776 subject records submitted with phenotypic data. Methods: We applied latent
class analyses (LCA) in a 2-tiered approach to assess the empirical validity of heterogeneity and subtypes
of PPD. Tier 1 examined PPD heterogeneity in subjects with complete data on the Edinburgh Postnatal
Depression Scale (EPDS) (N=6556), including PPD cases and controls. Tier 2 subjects included only
PPD cases (N=4245). In the Tier 2 analyses, indicator variables were hypothesized based on
distinguishing clinical features of PPD having commonality among sites. These indicator variables
included depression severity, EPDS total, EPDS anxiety subscale, timing of onset, pregnancy
complications, obstetric complications, suicidality, and psychiatric history/comorbidity of anxiety and
depression. Results: A 3 class solution yielded the best fit in both Tier 1 and Tier 2. In both Tiers, the
most striking characteristics were severity, timing of onset, comorbid anxiety, and suicidal ideation. The
class with the most severe PPD symptoms had significantly worse mood (mean EPDS=20.3), greater
anxiety, symptom onset that began during pregnancy, more obstetrical complications and endorsed
suicidal ideation. The other PPD class (mean EPDS=12.3) had less severe symptoms; the majority (54%)
endorsed symptom onset in the first month postpartum and had more pregnancy complications.
Conclusion: PACT represents an important next step toward large scale collaborative research efforts
needed to disentangle the pathophysiology of PPD. Examination of PPD heterogeneity to identify more
precise phenotypes is a critical first step toward future biological and genetic investigations.
Naomi Wray1, Enda Byrne1, Tania Carrillo-Roa2, Samantha Meltzer-Brody3, Brenda Penninx4, Hannah
Sallis5, Alexander Viktorin6, Psychiatric Genomic Consortium Major Depressive Disorder, Patrick
Sullivan4, Paul Lichtenstein6, Patrik Magnusson8, David Evans3, Grant Montgomery3, Dorret Boomsma7,
Nicholas Martin3
The University Of Queensland, 2Queensland Institute of Medical Research, 3University of North
Carolina at Chapel Hill, 4VU University Medical Center, 5School of Social and Community Medicine,
University of Bristol, 6Karolinska Institute, 7Vrije Universiteit
Individual Abstract The etiology of major depressive disorder (MDD) is likely to be heterogeneous and
researchers are faced with the dilemma of balancing power through increased sample size or sacrificing
sample size to achieve greater homogeneity of the case sample. Which strategy is optimal depends on the
underlying genetic architecture, which is unknown. Here, using currently available GWAS data for
MDD, we explore the validity of focusing on postpartum depression (PPD) as a genetically more
homogeneous MDD subtype. We used MDD GWAS data sets from Australia (1450 MDD cases, 1703
controls) and the Netherlands (1699 cases, 1765 controls). From these we identified PPD cases (484
Australian cases and 208 Dutch cases). We used SNP association results from the Psychiatric Genomics
Consortia (PGC) of Bipolar Disorder (BPD) and MDD to create polygenic scores for all individuals. The
R2 from a logistic regression of PPD case-control status on a SNP profile score weighted by PGC-BPD
association results was highly significant for both Australian and Dutch cohorts (R2 > 1.1%, p < 0.008).
Interestingly, the BPD profile score explained less variance in the much larger samples of MDD cases
and controls (R2
=0.06%, p= 0.08) in both data sets. Our results provide empirical genetic evidence for a more important
shared genetic etiology between BPD and PPD than between BPD and MDD. Further, they suggest that
focusing on PPD as a defined and more homogeneous subset of MDD may be a fruitful strategy for
genetic studies.
Divya Mehta1
University of Queensland
Individual Abstract Postpartum depression (PPD) is a significant public health problem with
approximately 13% incidence affecting not only women and their partners but has widespread cascading
consequences on the infant, family and friends and subsequently on the community (Meltzer-Brody &
Stuebe 2014). Despite harmful outcomes of PPD such as maternal suicide and infanticide, up to 80% of
PPD cases remain undetected (Yonkers 2003). Given the potentially serious consequences of PPD on
maternal and infant health and wellbeing, identification of reliable biological tests for early detection is
imperative. In a recent study (Mehta et al. 2014) we aimed to identify biomarkers for PPD by global
assessment of peripheral blood gene expression (N = 225 samples from 86 women at different timepoints including 29 PPD cases and 40 controls, recruited in Emory University, Atlanta, USA). Using gene
expression profiles in the third trimester where all women had no significant depressive symptoms, 116
transcripts were significantly differentially expressed (corrected p-value <0.01) between PPD cases and
controls, with functional annotation indicating a role of estrogen signaling. These transcripts
differentiated cases and controls with 88% accuracy in the discovery and replication sample. To the best
of our knowledge this is the earliest prospective gene expression predictor for PPD. The robustness of
these predictors needs to be tested in larger, independent and ethnically diverse cohorts. Also, due to the
overlapping etiology of depression and other mood disorders, these results might also be relevant to other
mood disorders. Power calculations depicting the number of samples required to test the predictive
biomarkers with adequate sensitivity and specificity at a 95% Confidence interval will be described
(unpublished). These calculations show that large numbers of samples are required to be able to detect the
small and subtle biological changes observed in postpartum depression. For instance, assuming the
average postpartum depression disease prevalence rate of 13%, we calculate that even for a predictor with
specificity and sensitivity as high as 93%, a total of over 2300 samples will be required. Given the
prospective nature of the studies and accounting for variation in rates of women who go on to develop
PPD, it is clear that a large international interface is required to pool biological resources to reach these
sample sizes. We will discuss the importance of a current initiative associated with PACT to establish a
PPD biological repository for gene expression profiling to identify robust early predictive biomarkers for
PPD. This talk will demonstrate the scale of data required to uncover robust and reproducible biomarkers
for postpartum depression and these results will provide a valuable insight for future ongoing biomarker
studies in Major depressive disorders.
Chair: John Kelsoe, University of California San Diego
Overall Abstract Details Lithium is the oldest and still the best mood stabilizing medication for bipolar
disorder. There is a wide range in response to lithium, and a subset of patients have a very robust
response with almost complete elimination of symptoms. Lithium response has been shown to be
familial, and it has been argued that lithium responsive bipolar disorder may constitute a mechanistically
distinct form of illness. If this is true, it suggests that genetic variants associated with good lithium
response, may also be susceptibility genes and predispose to a form of illness that involves pathways that
are responsive to lithium. The identification of genes for lithium response may therefore have a double
yield of a possible clinically useful predictor, and a way to dissect bipolar disorder into less
heterogeneous parts.
Urs Heilbronner will present the latest results of the ConLiGen consortium GWAS in 2500 subjects.
Martin Alda will present genome sequencing results from lithium responsive bipolar patients. Sarah
Bergen will describe GWAS results from a large sample of subjects treated with lithium. Lastly, John
Kelsoe will describe cellular models in lymphoblasts and iPS derived neurons that may help guide the
search for genomic variants by identifying physiologically meaning full targets. These studies suggest a
dramatic difference in induction of gene expression in lithium responders as compared to non-responders,
as well as, differences in electrophysiological and calcium signaling.
Urs Heilbronner1, Liping Hou2, Marcella Rietschel3, Francis McMahon2, Thomas Schulze4, The
ConLiGen Consortium
University Medical Center Goettingen, 2Human Genetics Branch, NIMH, NIH, 3
Department of Genetic Epidemiology in Psychiatry, Central Institute of Mental Health, Mannheim,
Germany, 4Section on Psychiatric Genetics, Department of Psychiatry and Psychotherapy, University
Medical Center, Georg-August-University, Göttingen, Germany
Individual Abstract Background: Lithium remains a mainstay in the long-term treatment of bipolar
disorder (BD). The individual response to lithium is variable. About 30% of patients treated with lithium
have fewer illness episodes over time, while about 20% have no response. Data from pharmacogenetic
studies of lithium are comparatively sparse, and these studies have generally employed small sample
sizes and varying definitions of response. Genetic markers of lithium response would be valuable for
treatment planning and could provide insights into the biological mechanism of lithium action. To put
that idea into practice, the international Consortium on Lithium Genetics (www.ConLiGen.org) was
Methods In a first GWAS, ConLiGen studied 1080 European and European American lithium-treated BD
patients. All patients were characterized for lithium response with an 11-point treatment response scale
(“Alda Scale”, Grof et al. 2002). The Alda Scale assesses clinical improvement attributable to lithium,
taking into account various confounding variables. Phenotype definitions were developed by consensus
within ConLiGen. The whole sample was genotyped using Illumina arrays to perform a genome-wide
association study (GWAS) of lithium response. Results: Inter-rater reliability of lithium response
assessment was good, with kappa values >0.7. GWAS genotyping was completed at excellent call rates
(>99% of samples had a call rate >98%). While no genome-wide significant finding at the p<5*10-8
level was observed, the top hit SNP rs17728078 in the gene SLC4A10 (solute carrier family 4, sodium
bicarbonate transporter, member 10; p=9.59*10-6) yielded an odds ratio of 1.58, which is quite
uncommon for complex phenotypes, and represents a common allele at a minor allele frequency of ~0.4,
increasing the chances of replication in an independent sample. Discussion: Our finding in the SLC4A10
gene is promising as this gene belongs to a small family of sodium-coupled bicarbonate transporters
(NCBTs) that regulate the intracellular pH of neurons and the pH of the brain extracellular fluid.
However, replication of this finding in additional samples will be crucial to establish it as true
susceptibility factor for lithium response. Within the framework of ConLiGen, 1571 new samples from
Europe, North America and Australia are currently being analyzed for that purpose. Results of this
replication GWAS will be presented at the meeting. Additionally, ConLiGen has access to 222 lithiumtreated BD patients from East Asia. Reference Grof P, Duffy A, Cavazzoni P, Grof E, Garnham J,
MacDougall M, O’Donovan C, Alda M: Is response to prophylactic lithium a familial trait? J Clin
Psychiatry 2002; 63:942–947.
Martin Alda1, Cristiana Cruceanu2, Gustavo Turecki2, Guy A. Rouleau2
Dalhousie University, 2McGill University
Individual Abstract Phenotypic and genetic heterogeneity complicates the genetic and neurobiological
research in psychiatry. A promising strategy to reduce heterogeneity is a study of validated subtypes of
illness. Robins and Guze (Am J Psychiatry, 1970) proposed five criteria of diagnostic validity: (1)
clinical description, (2) follow-up study, (3) delimitation from other disorders, (4) family study, and (5)
laboratory studies. We applied these criteria to bipolar disorder responsive to long-term lithium
treatment. In a series of investigations we examined characteristics of patients with bipolar disorder
responsive to lithium and compared them with lithium non-responders as well as responders to other
mood stabilizers. Patients responsive to lithium showed (1) a typical clinical picture with euphoric
manias, melancholic depressions, low rates of co-morbid conditions, and recurrent episodic clinical
course. The response to lithium was (2) longitudinally stable even after 20 years of follow-up. In
comparison, (3) responders to lamotrigine or carbamazepine showed more often atypical features such
mood liability, comorbid anxiety, or mood-incongruent psychosis. In family studies (4) lithium
responders had higher prevalence of bipolar disorder and lower rates of schizophrenia among their
relatives and the relatives suffering from bipolar disorder were about 3.7 times more likely to respond to
lithium compared to patients unselected for family history. Finally, several neurobiological studies (5)
supported the view of lithium responders as a distinct group. For instance in a positron emission
tomography study lithium responders showed a distinct pattern of regional cerebral blood flow changes
in response to induced sadness – similar to their unaffected relatives, but significantly different from
responders to valproate – with main differences in the rostral anterior cingulate and dorsolateral
prefrontal cortex. These results support the view of lithium responsive bipolar disorder as a distinct
subtype of bipolar disorder. They also make lithium responders a promising group for genetic and
neurobiological studies. In the last 15 years, we established a large sample of patients and relatives
characterized for their response to long term treatment. The samples is the basis for a whole exome
sequencing study in a combined sample of moderately-sized families (41 families, 234 exomes) and
unrelated probands (117 subjects). The results of the study will be presented at the symposium.
Preliminary analyses indicate a higher rate of damaging mutations in genes that have been previously
proposed as related to pharmacological effects of lithium.
Sarah Bergen1, Jie Song2, Christina Hultman2, Paul Lichtenstein2, Mikael Landen3
Karolinska Institute, 2Medical Epidemiology and Biostatistics, Karolinska Institutet, 3Institute of
Neuroscience and Physiology, The Sahlgrenska Academy at Gothenburg University; Medical
Epidemiology and Biostatistics, Karolinska Institutet
Individual Abstract Background: Lithium is one of the oldest and most common treatments for bipolar
disorder, but it is only effective in a subgroup of patients. Currently, the effectiveness of lithium is
assessed through trial and error, but understanding the factors predicting lithium response could help to
prevent wasted time and needless suffering faced by non-responders. This genome-wide association
study sought to identify common genetic variation influencing response to lithium. Methods: In a sample
of 940 Swedish patients with bipolar disorder, a self-reported measure of lithium response was tested for
association against genome-wide genotype data. The 64% of patients reporting full response were
contrasted with the 36% with partial response or no benefit. Additionally, the lithium responsive
subgroup of patients was tested against 1215 healthy controls, predicated on the idea that they may define
a more etiologically homogeneous patient population. Logistic regression was performed in PLINK
incorporating four MDS covariates to account for population substructure. Results: For the lithium
responder versus non-responder analyses, there were no genome-wide significant results. The top hit was
a marker in the zinc-finger gene ZNF83 (p = 7.45 x 10-6). The analyses of lithium responders versus
controls, however, yielded one marker approaching genome-wide significance in the DLGAP1 gene (p =
9.26 x 10-8). Discussion: No single common genetic variant with a strong effect on lithium response was
detected, but analyses to detect risk markers for bipolar disorder in the subgroup of lithium responsive
cases tentatively implicate DLGAP1, a postsynaptic scaffolding protein with several binding partners
previously associated with other neuropsychiatric disorders. Additionally, genotyping for 1600 more
Swedish subjects with lithium response information is in progress, and collaborations to incorporate 2600
cases from the UK will substantially enhance power to detect associations.
Kangguang Lin1, Jun Yao2, Kristen Brennand3, John Kelsoe , Michael McCarthy4, Abesh Bhattacharjee4,
Susan Leckband5, Cory White4, Wei-Wei Matsuda4, Christopher Woelk4, Fred Gage6,
Pharmacogenomics of Bipolar Disorder Study Investigators
University of Hong Kong, 2University of Wisconsin, 3Mt. Sinai School of Medicine, 4University of
California San Diego, 5VA San Diego Healthcare System, 6Salk Institute
Individual Abstract Pharmacogenetics has great potential to both guide clinical treatment and to help
unravel the genetic and etiological heterogeneity of bipolar disorder. However, attempts to identify genes
associated with response are hampered by the difficulty, as well as, labor and expense of determining
drug response in psychiatric subjects. Whereas, GWAS is successfully identifying small effect variants
using very large sample sizes, it is likely cost prohibitive at this time to phenotype very large samples for
drug response. For this reason, an efficient strategy may be to employ biological information to identify a
set of candidate genes that are more likely to be involved in drug response. In this way, a much smaller
set of hypotheses can be tested with resulting substantial gain in statistical power. Here, we describe two
cellular phenotypes that distinguish bipolar lithium responders from non-responders. Blood for
immortalized lymphoblasts and skin biopsies for iPS cells were obtained from bipolar I subjects
participating in a multi-site prospective lithium trial, the Pharmacogenomics of Bipolar Disorder Study,
or a similar study of veterans. In each of these studies, bipolar I subjects were stabilized on lithium
monotherapy over a 16 week period, then those that reached remission were followed for up to 2 years in
order to determine time to relapse as a measure of response. Lymphoblasts from 8 prospectively
documented lithium responders and 8 non-responders were treated with 1mM lithium for one week at
which time RNA was harvested and gene expression determined by RNAseq. Analysis was conducted
using BioconductoR and R routines, and gene expression was compared within each subject with and
without lithium. While 1557 genes underwent a statistically significant change in expression in the
responder group, only 75 were significantly changed in the non-responder group. This suggests, simply,
that the responders had a clinical response to lithium because their cells underwent a much greater
change in physiological state as compared to non-responders. Most notable was the gene CRIP2 which
underwent an 18 fold increase in expression in responder and only a 3 fold increase in non-responders.
Induced pluripotent stem (iPS) cells offer a route to model response in derived neurons. Skin biopsies
were obtained from 3 responders, 3 matched non-responders and 4 matched controls. These were
reprogrammed to iPS cells using Sendai virus and pluripotency validated. They were then differentiated
into Prox1 positive hippocampal dentate gyrus/glutamatergic neurons. Patch clamp studies revealed
extended trains of action potentials in 2 of 3 responder lines, that were not observed in any of the control
lines. The non-responders displayed similar electrophysiology as the responders. A corresponding
alteration in calcium flux was also observed, and both phenotypes were rescued by lithium treatment.
RNAseq studies in these samples are in progress.
Chair: Thalia Eley, Institute of Psychiatry, Kings College London
Overall Abstract Details Background: Anxiety and depressive disorders are highly prevalent and
debilitating conditions. Psychological interventions are commonly the treatment of choice for anxiety
disorders, and are also widely used in depression. However, as with pharmacological treatments, not all
individuals respond to psychological interventions. The small but growing field of therapy genomics,
parallel to pharmacogenomics, explores genetic, genomic and epigenetic factors as both predictors of
treatment response and as potential mechanisms. Our first speaker will present a GWAS of response to
cognitive-behavior therapy (CBT) in child anxiety disorders. Although none of the individual SNPs
reached genome-wide significance, there were several at a suggestive level, and a polygenic risk index
created from the top ~20 hits, showed a strong and significant association with treatment response. Our
second speaker will look at change in (a) genome-wide gene expression and (b) DNA methylation in
selected candidate genes, across exposure-based therapy for individuals with panic disorder or specific
phobias. RNA and DNA methylation levels are analyzed at three time-points from pre-treatment to posttreatment and follow-up allowing the exploration of both gene expression DNA methylation as potential
mechanisms of change. Our third speaker will present a study examining the role of epigenetics in
response to psychotherapy for Post-Traumatic Stress Disorder (PTSD). Analyses explore the role of pretreatment methylation levels as a predictor of outcome, and increase/decrease in methylation levels as a
potential mechanism of change. The session will close with findings from a study in which adults with
depressive disorders have been randomized to CBT versus medication. Analyses include imaging data as
well as genotype, gene expression and DNA methylation. Preliminary data of a GWAS for PET imaging
predictors of CBT vs MED response in a subsample will be shown, with the top hit being replicated in an
independent sample. Conclusion: Genetic factors are likely to become a useful predictor of psychological
treatment response. Whilst GWAS analyses are likely to be underpowered, a combination of using
polygenic risk scores and working towards multiple datasets on which meta-analyses can be conducted
may prove fruitful. It is particularly interesting that genetic findings to date appear to be independent of
clinical and demographic predictors of psychological treatment response. However, in order to
understand mechanisms of change we need to look beyond genotype alone. Findings relating to gene
expression and DNA methylation changes during psychological interventions may provide useful
indicators of potential underlying mechanisms and thus new treatment development.
Jonathan Coleman1, Kathryn Lester2, Susanna Roberts2, Robert Keers2, Chloe Wong2, Jennifer Hudson3,
Gerome Breen2, Thalia Eley2, Team GxT
Institute of Psychiatry, King's College London, 2King’s College London, MRC Social, Genetic and
Developmental Psychiatry Centre, Institute of Psychiatry, 3Centre for Emotional Health, Department of
Psychology, Macquarie University, Sydney
Individual Abstract Background: Anxiety disorders are common, with lifetime prevalence in adults of
approximately 30%. Psychosocial treatments, including cognitive behavioral therapy (CBT), are the
primary treatment modality for anxiety disorders in the United Kingdom. Remission of the disorder
following CBT is estimated at approximately 60% post-treatment, and is likely to increase in the period
after treatment. A modest candidate gene literature suggests individual differences in treatment response
may have a genetic basis; however, such studies are limited by methodological issues. Early successes
and failures from the candidate gene literature motivated the presented genome-wide association study
(GWAS), examining the association of common single nucleotide polymorphisms (SNPs) with
differential response to CBT. Methods: Genetic samples were gathered from 1596 children (aged 5-18)
diagnosed with anxiety disorders, undergoing CBT. The cohort originated from eleven different sites in
the UK, USA, Australia, and Western Europe. The Anxiety Disorder Interview Schedule was used to
provide diagnoses and clinical severity ratings at baseline, upon completion of treatment, and at followup. Buccal swabs were used to obtain DNA, which was extracted using a standard protocol, concentrated
by filtration and resuspension, and genotyped using the Illumina HumanCoreExome array. SNPs were
then imputed to the latest release from the 1000 Genomes project, providing data on more than 6 x 106
variants for analysis. The worldwide ascertainment of this sample confounds the ability of PCA to
control for population stratification, and so mixed linear model association analyses were performed in
EMMAX. In addition, an estimate of SNP-chip pseudoheritability was calculated by EMMAX and this is
compared to that found using GCTA. Results: The initial outcome variable was change in clinical
severity rating across treatment. Findings above suggestive significance (p < 5 x 10-6) were tested
against other phenotypes, including change in clinical severity ratings between baseline and follow-up.
Pseudoheritabilty and predictions of polygenic effect were calculated. The results and implications of
these analyses are discussed. Discussion: We present a GWAS of response to CBT in a cohort of children
with anxiety disorder ascertained at sites across the globe. This is, to our knowledge, the first GWAS of
response to psychosocial treatment, and the first treatment response GWAS to be performed in anxiety
disorder. Although the sample size is relatively large for a study of response to therapy, it is small
compared to successful psychiatric GWAS, and is likely underpowered to detect the small effect sizes
expected. Power analyses estimate that the sample has 80% power to detect variants at a genotypic
relative risk of 1.36 and frequency of 0.4. However, estimates of pseudoheritability suggest common
SNPs can explain a portion of the variance in treatment response.
Susanna Roberts1, Kathryn J. Lester1, JГјrgen Margraf2, Silvia Schneider2, Tobias Teismann2, Jonathan
Coleman1, Gerome Breen1, Chloe C.Y. Wong1, Thalia C. Eley1
King's College London, 2Ruhr University Bochumm, Germany
Individual Abstract Exposure-based CBT is a psychological therapy which involves exposing the
individual to the anxiety-provoking stimuli, and is an effective treatment option in anxiety disorders such
as phobias, panic disorder and agoraphobia. However, there is still substantial variability in response;
around 30% of adults are not diagnosis-free following treatment, and many retain a high level of fear.
Recent research suggests that response to psychological therapies such as manualised CBT and exposure
therapy are associated with differences in DNA methylation change across the course of treatment.
Furthermore, investigation of gene expression levels across the course of treatment may provide insight
into the potential molecular mechanisms involved in response to therapy, as they can be influenced by
both genetic and environmental factors and are useful indicators of gene function and activity.
Participants consisted of 100 adults receiving exposure-based CBT for a primary diagnosis of panic
disorder, agoraphobia or specific phobia. Anxiety disorder diagnoses were made by trained clinicians
according to DSM criteria. DNA and RNA were extracted from whole blood samples collected at pretreatment, post-treatment and follow-up time points. DNA methylation at 4 candidate genes (SERT,
FKBP5, NR3C1 and CRHR1) was quantitatively measured using the Sequenom EpiTYPER. Genomewide expression was assessed using the Illumina HumanHT-12 v4 Expression BeadChip. Methylation
status and expression levels at all three time-points were investigated for association with both
dichotomous clinical outcome and changes in symptom severity. The findings of these studies will be
reported, and their relevance and implications within clinical and biological frameworks will be
Rachel Yehuda1, Linda Bierer1, Amy Lehrner1, Nikos Daskalakis1, Iouri Makotkine1, Frank Desarnaud1,
Janine Flory1
Mount Sinai School of Medicine, James J. Peters VA Medical Center
Individual Abstract Epigenetic alterations offer promise as prognostic or diagnostic markers, but it is
not known whether these measures associate with or predict clinical state. These questions were
addressed in a pilot study with combat veterans with PTSD to determine whether cytosine methylation in
promoter regions of the glucocorticoid-related NR3C1 and FKBP51 genes would predict or associate
with treatment outcome. Veterans with PTSD were treated with prolonged exposure (PE) psychotherapy,
yielding responders, defined by no longer meeting diagnostic criteria for PTSD, and non-responders.
Blood samples were obtained at pre-treatment, after 12 weeks of psychotherapy (post-treatment), and
after a 3 month follow-up. Methylation was examined in DNA extracted from lymphocytes. Measures
reflecting glucocorticoid receptor (GR) activity were also obtained from lymphocytes; plasma and 24-hr
urine cortisol and plasma neuropetide-Y levels were also measured. Methylation of the GR gene
(NR3C1) promoter assessed at pre-treatment predicted treatment outcome, but was not significantly
altered in responders or non-responders at post-treatment or follow-up. In contrast, methylation of the
FKBP5 gene (FKBP51) promoter region did not predict treatment response, but decreased in association
with recovery. In a smaller subset, a corresponding group difference in FKBP5 gene expression was
observed, with responders showing higher gene expression at post-treatment than non-responders.
Endocrine markers also changed in association with symptom change. These preliminary observations
require replication and validation. However, the results support research indicating that some
glucocorticoid related genes are subject to environmental regulation throughout life. Moreover,
psychotherapy resulting in substantial symptom change constitutes a form of �environmental regulation’
that may alter epigenetic state. Finally, the results further suggest that different genes may be associated
with prognosis and symptom state, respectively.
Elisabeth Binder1, Tania Carrillo-Roa1, Caleb A. Lareau 2, Callie L. McGrath3, Boadie W. Dunlop3, Mary
E. Kelley3, Helen S. Mayberg3
Max-Planck Institute of Psychiatry, 2University of Tulsa, 3Emory University
Individual Abstract Currently the choice of antidepressant treatment strategy is not based on the
underlying pathophysiology. Neuroimaging results suggest that patients preferentially responding to
either psychotherapy or antidepressant drugs have different resting state neural activation pattern
(McGrath et al. 2013). These different activations on the neural circuit level maybe related to different
genomic risk factors, both on the DNA sequence and the epigenetic level. This presentation will highlight
approaches combining imaging and genetic and epigenetic data to predict differential response to
cognitive behavioral therapy (CBT) or antidepressant drugs in patients with major depression. Patients
were recruited at Emory University and randomized at baseline to 12 weeks sCIT, or 16 sessions of CBT.
Genome-wide genotypes (Illumina OmniExpress) and DNA methylation (Illumina HM-450K) were
measured in peripheral blood DNA drawn at baseline. Genome-wide SNPs and CpGs methylation
univariate and multivariate association analyses were initially conducted in 76 patients with major
depression to test for association with activation pattern in brain regions predicting differential response
to CBT vs. drug assessed by Brain glucose metabolism with positron emission tomography prior to
treatment randomization. We observed genome-wide significant association of rs34383296 (p =
9.4x10-9) in a multivariate analysis that included three of the 6 tested brain regions. The associated
variant lies in a gene dense region on chromosome 9 within the NDOR1 gene and it is an eQTL for
ARRDC1, a gene ~400kb downstream, related to arrestin-mediated internalization of cell surface
receptors. This SNP was genotyped in an independent larger sample (N = 260) with major depression and
predicted differential response to CBT vs. drug in the expected direction. New data of on-going analyses
of DNA methylation pattern and GWAS in a larger sample of 310 patients treated with either CBT or
escitalopram or duloxetine will be reported. Genetic and genomic biomarkers allowing a prediction of
preferential treatment response to psychotherapy vs. antidepressant drugs could optimize the treatment
for the individual patient and decrease time to clinical response. McGrath CL, Kelley ME, Holtzheimer
PE, Dunlop BW, Craighead WE, Franco AR, Craddock RC, Mayberg HS. Toward a neuroimaging
treatment selection biomarker for major depressive disorder. JAMA Psychiatry. 2013 Aug;70(8):821-9.
Chair: Pippa Thomson, Institute of Genetics and Molecular Medicine
Overall Abstract Details Low frequency, high penetrance alleles in a background of high locus and/or
allelic heterogeneity are immune to discovery by genome-wide association study (GWAS) analysis in a
conventional case-control or cross-sectional study design. In this situation next generation sequencing is
an alternative approach to the identification of causative variants. The high frequency of rare variants in
the human genome makes it difficult to identify causative variants from the background level of
variation. In this context, utilizing samples for which additional family members are available both
increases the filtering power to identify variants shared by cases and provides an in-built quality control
for these relatively new sequencing technologies. The increasing move to rare variation, estimation of
heritability and the search for endophenotypes has seen a return to the analyses of familial information.
We will describe the recent results including exome sequencing data joint analyses of quantitative and
binary traits, and aggregate gene- and pathway-centric testing. Using examples from schizophrenia,
bipolar disorder and depression, this symposia will highlight the value of such studies and the progress
Lan Xiong1, Cristiana Cruceanu2, Pingxing Xie2, Qin He1, Mina Ohadi3, Narge Moghimi4, John Vincent4,
Martin Alda5, Gustavo Turecki2, Guy Rouleau2
University of Montreal, 2McGill University, 3Genetics Research Center, University of Social Welfare
and Rehabilitation Services, 4Centre for Addiction and Mental Health, 5Department of Psychiatry,
Dalhousie University
Individual Abstract In the past decade the main study design for genetic studies of complex traits has
addressed the common disease/common variant hypothesis using case-control cohorts. This was possible
because of the availability of cheap high-throughput DNA chip technology and the recruitment of large
numbers of unrelated cases and controls. However, the development of NGS technologies has led to a
refocus onto the contribution of rare variants to disease risk. Most existing case-control cohorts do not
have the power to define the role of such rare variants in disease susceptibility. However, familial cohorts
provide significant advantages in genetic studies for studying the role of rare variants in complex traits.
We are recruiting families to participate in various genetic studies, at first for gene identification of
different Mendelian disorders; now to study more complex diseases, such as bipolar disorder,
schizophrenia, restless legs syndrome, essential tremor etc., in which familial cases have significantly
reduced the clinical and genetic heterogeneity and have led to successful gene discoveries. We have also
used a family study design to investigate the role of de novo high-penetrant variants in autism and
schizophrenia. We will give examples of how we are using special family cohorts for genetic discovery
in psychiatric disorders: (1) Large extended consanguineous pedigrees from Pakistan with schizophrenia,
in which we have performed high density SNP genome scan and exome sequencing on all affected
individuals in each pedigree; (2) Inbred pedigrees from Iran with bipolar disorders, in which we have
performed exome sequencing of all affected individuals; (3) Large pedigrees from Canada with bipolar
disorder responsive to lithium treatment, in which we have focused on a more homogenous subphenotype
of bipolar disorder and its related candidate genes and pathways. In each of these projects/families we are
currently generating whole exome data on every affected individual with reliable diagnosis, then
prioritizing on highly penetrant (e.g. protein-truncating, missense, or frameshift) or functionally relevant
variants (e.g. 3’UTR, 5’UTR, splicing) shared among all or most affected individuals within the family.
We are also testing alternative hypotheses (e.g. different diagnostic schemes and modes of inheritance)
and further validating the potential candidate variants through additional genetic testing and functional
assays. Family studies may provide answers for more profound genetic questions, such as origin of
mutations, heritability of diseases, complex mode of inheritance (such as digenic, oligogenic or
multigenic inheritance, or epigenetic inheritance), modifier genes and gene-gene interactions. Family
studies are also by default longitudinal and prospective studies, providing significant biological insights
into the contribution of genetic/genomic variants to human disease phenotypes.
Simone de Jong1, Mateus Diniz2, Shaza Alsabban3, Gadelha Ary2, Andiara Rodrigues2, de Jong Simone3,
McGuffin Peter3, Bressan Rodrigo2
MRC SGDP Centre, Institute of Psychiatry, King's College London, 2Federal University of Sao Paulo, 3
King's College London
Individual Abstract We present a phenotypic and molecular study of one of the largest families ever
found with multiple cases of bipolar disorder (BP, n=39) and major depressive disorder (MDD, n=59).
The family come from a rural area of Brazil and also contains many mood disorder cases with comorbid
autoimmune thyroid disease (n=24) and type 1 diabetes (n=31) as well as Parkinson's disease (n=8), selfreported as mature onset insulin dependence, with 7 cases reporting both autoimmune comorbidities.
There are multiple child/teenage family members who exhibit some form of mood or psychiatric
disturbance with anticipation in age of onset indicated. We have conducted an initial ascertainment of the
family in 2009/10 with basic psychiatric phenotyping and self-report of physical comorbidities. In all,
333 of the family members consented and gave blood in the first wave of the study. We performed a
whole genome linkage scan of the BBF-A family to find regions of chromosomes that are segregating
with mood disorders in the family (Diniz/Al-Sabban et al., in prep). Our data analysis so far has primarily
used the 269 family members representing the most densely affected part of the pedigree, using
autosomal single nucleotide polymorphisms (SNPs) genotyped using the Affymetrix 10K genotyping
array, with approximately 67% of information content genome wide. Multipoint parametric (HLOD) and
non- parametric linkage (NPL) analyses were performed using MERLIN splitting BBF-A into twelve
subfamilies and breaking loops (9 from 267 meioses in the first wave of data). In addition, Multipoint
parametric linkage analyses were performed using MCLINKAGE, where the family could be analyzed
with their structure and loops intact. Parametric Linkage analyses were conducted under dominant and
recessive modes of disease transmission and non-parametric linkage (NPL) analyses was performed.
Genomewide significant linkage, allowing for multiple phenotype definitions, was identified for
2p23.1-p22.3 (LOD=3.83) for all mood disorders, 3p23-p24.1 (LOD=4.18) for narrow bipolar 1, and
both 11p14 (LOD=4.49) and 12q24.22-q24.32 (LOD=4.74) for depression. In addition, 22q11.21-q12.1
had a suggestive/trend lod score of 3.76 for a broad Bipolar disorder definition. Exome sequencing has
been carried out on a limited number of cases and will be presented.
John Blangero1, Laura Almasy1, David Glahn2
Texas Biomedical Research Institute, 2Yale University
Individual Abstract Although the number of psychiatric disease-related QTL localizations has risen
rapidly during the GWA era, causal gene discoveries have been few. The accumulating data now suggest
that common variants (such as those found in traditional GWA studies) have small biological effects that
are extremely difficult to assess functionally and, hence, are unlikely to be easily associated with the
underlying causal genes. The advent of economically reasonable whole genome sequencing (WGS) now
allows us to turn attention to rare sequence variants that overwhelmingly comprise the majority of
human genetic variation. Rare functional variants tend to have substantially larger biological effect sizes
that should be much easier to molecularly characterize. However, the rarity of these variants requires a
different study design than that of unrelated cases and controls. Pedigree-based studies are optimal for
testing the rarest of genetic variants (specifically private variants). In this paper, I will show how large
pedigrees can be assessed for their power to detect and test private functional variants using WGS data
from the Genetics of Brain Structure and Function Study. A vast amount of private non-synonymous
variation is observable for major biological pathways of psychiatric relevance. Many of these private
variants are captured in sufficient numbers due to Mendelian transmission to allow direct statistical
testing while those that are observed in too few copies can still be utilized for aggregate gene- and
pathway-centric testing. Combining the analysis of WGS data with new statistical methods for detecting
disease-related endophenotypes in randomly ascertained pedigrees, we identify rare functional variants
that implicate several novel likely genes relevant for schizophrenia and mood disorders.
Pippa Thomson1, Jennifer E. Huffman1, James Prendergast3, Victoria Svinti1, Generation Scotland1,
Caroline Hayward1, Martin Taylor1, Malcolm Dunlop1, David Porteous1, Andrew McIntosh1, Alan
Wright1, Nick Hastie1
Institute of Genetics and Molecular Medicine, 2Roslin Institute
Individual Abstract The high frequency of rare variants in the human genome makes it difficult to
identify causative variants from the background level of variation. In order to identify rare variants
influencing recurrent major depressive disorder, we have generated whole exome sequence data from 277
members of 44 families and 148 unrelated individuals with familial early onset, drawn from the
Generation Scotland: Scottish Family Health Study. Exome variants, that are enriched within affected
individuals relative to almost a thousand control exomes have been analyzed for predicted functionality,
co-segregation within families, association with endophenotypes of depression, and clustering within
genes or pathways. Linkage analyses identify two genome-wide significant regions: a peak at the tip of
chromosome 12 under a recessive model (12p13.33, hLOD 6.4, mLOD 3.7) within a micro-deletion
region associated with neurodevelopmental delay, and a peak overlapping the centromere at chromosome
18 under an additive model (18p11.2-18q11.2, hLOD 4.12, mLOD 2.5). Quantitative linkage analyses of
depression endophenotypes support the involvement of the chromosome 18pq region in age of onset and
cognitive phenotypes including logical memory (LOD > 1). Multiple loss-of-function variants have been
identified in these regions and burden analyses are underway to identify the gene/genes underlying the
linkage peak