European Neuropsychopharmacology (2014) 24, 1567–1577

www.elsevier.com/locate/euroneuro

REVIEW

Genome-wide association studies of suicidal behaviors: A review Marcus Sokolowskin, Jerzy Wasserman, Danuta Wasserman National Centre for Suicide Research and Prevention of Mental Ill-Health (NASP), Karolinska Institute (KI), S-171 77 Stockholm, Sweden Received 21 March 2014; received in revised form 24 July 2014; accepted 10 August 2014

KEYWORDS

Abstract

Suicide attempt; GWAS; SNP; Gene; Pathway; Network

Suicidal behaviors represent a fatal dimension of mental ill-health, involving both environmental and heritable (genetic) influences. The putative genetic components of suicidal behaviors have until recent years been mainly investigated by hypothesis-driven research (of “candidate genes”). But technological progress in genotyping has opened the possibilities towards (hypothesis-generating) genomic screens and novel opportunities to explore polygenetic perspectives, now spanning a wide array of possible analyses falling under the term Genome-Wide Association Study (GWAS). Here we introduce and discuss broadly some apparent limitations but also certain developing opportunities of GWAS. We summarize the results from all the eight GWAS conducted up to date focused on suicidality outcomes; treatment emergent suicidal ideation (3 studies), suicide attempts (4 studies) and completed suicides (1 study). Clearly, there are few (if any) genome-wide significant and reproducible findings yet to be demonstrated. We then discuss and pinpoint certain future considerations in relation to sample sizes, the units of genetic associations used, study designs and outcome definitions, psychiatric diagnoses or biological measures, as well as the use of genomic sequencing. We conclude that GWAS should have a lot more potential to show in the case of suicidal outcomes, than what has yet been realized. & 2014 Elsevier B.V. and ECNP. All rights reserved.

1.

n Corresponding author. Tel.: +46 8 52 48 69 38; fax: +46 8 30 64 39. E-mail address: [email protected] (M. Sokolowski).

http://dx.doi.org/10.1016/j.euroneuro.2014.08.006 0924-977X/& 2014 Elsevier B.V. and ECNP. All rights reserved.

Introduction

A general genetic diathesis of suicidality has been shown to exist by family, adoption and twin studies, with heritability in the range between 30–55% (Voracek and Loibl, 2007). Such a general genetic diathesis mainly refers to the behavioral manifestations of suicidality (i.e. suicide attempts or completed

1568 suicides), rather than suicidal thoughts or ideation alone. However, this is not clear-cut, as ideation precede a proportion of suicide attempts or completed suicides (Baca-Garcia et al., 2011). Specific genetic components (i.e. particular candidate genes) are therefore often being studied, i.e. genes which are implicated in suicidality or its associated endophenotypes (Mann et al., 2009), selected on the basis of prior biological and/or pharmacological knowledge and observations (Currier and Mann, 2008; Wasserman et al., 2009). Selected genetic variants in serotonergic system genes have been the major historic focus of such investigations during the past decades, but the polygenetic perspective is nowadays also being addressed by e.g. the study of candidate genes in other neurosystems (Ben-Efraim et al., 2011, 2012, 2013; Ernst et al., 2009; Rujescu and Giegling, 2010; Sokolowski et al., 2012, 2010; Wasserman et al., 2005, 2007, 2008, 2009). A relatively novel approach for further elucidating the polygenetic perspective, which has become available in the last years, is to conduct a Genome-Wide Association Study (GWAS). GWAS is performed using many genetic markers across the whole genome to analyze for association with a trait (Ikegawa, 2012). It has been termed as a hypothesis-free approach (with regard to that any of the markers/genes are being hypothesized to show association), which is in contrast to the candidate gene approach (Goldstein et al., 2003). One main goal of GWAS is to suggest novel candidate genes which could not be hypothesized a priori by current knowledge. The majority of markers analyzed in GWAS are Single Nucleotide Polymorphisms (SNPs), in addition to certain Copy Number Variants (CNVs). Technological progress of genotyping chips (with Illumina Inc leading the market) combined with the efforts to map all human SNPs and their patterns of linkage disequilibrium (LD) in different populations (see e.g. http://www.hapmap.org and http:// www.1000genomes.org), have resulted in a continuous increase in the number of SNPs analyzed for each year, leading up to the present 5 million low and high allele-frequency SNPs assayed on one GWAS chip (e.g. the Illumina HumanOmni5-Quad chip). Given the multiple association tests subsequently conducted, one today usually requires a P-value in the range of 10 8 to declare a certain SNP significant at alpha 0.05 after accounting for LD (Li et al., 2012b). In addition, samples in the sizes of tens of thousands are the goal to obtain sufficient power to detect the usually small SNP-by-SNP effect sizes, as have been observed for many complex traits. The initial enthusiasms about GWAS were somewhat damped (Need and Goldstein, 2010; Rowe and Tenesa, 2012) as “genome-wide significant” GWAS SNP-associations were found to be of small effect, difficult to generalize and thus not directly applicable to real life, e.g. in the clinic or in public health prevention (Eichler et al., 2010). But GWAS data nevertheless remain a rich resource for continuing to elucidate polygenetic etiology. Its success (Teslovich et al., 2010) or failure (Sebastiani et al., 2010, 2011) to generate new hypotheses and insights seem to depend on the trait under study, the study design, analysis approaches and avoidance of previous mistakes (Califano et al., 2012; Lambert and Black, 2012; Need and Goldstein, 2010; Rowe and Tenesa, 2012; Terwilliger and Goring, 2009). Promis ing novel ways to analyze GWAS data on poly-SNP, gene-, pathway- or network-levels are being developed and applied (complementing the classical SNP-by-SNP approach), which are (i) less penalized by multiple comparison corrections and (ii) attempt to better capture a polygenetic biology more effi

M. Sokolowski et al. ciently. Therefore, there are now different analytical approaches to GWAS data available and their utility is ulti mately decided by the mainly unknown genetic architecture one tries to map.

1.1.

Classical SNP-by-SNP GWAS

The wide majority of GWAS conducted on many complex traits up to date, including suicidality (Table 1), has mainly focused on the identification of significant, single SNP-associations. The premise is that a subset of (usually higher frequency, “common”) SNPs capture the overall genetic involvement in a phenotypic effect, an effect which deviates from the rest of the genome. But complex traits are believed to involve the interplay of many different genetic perturbations. For psychiatric disorders, results from single SNP-analyses have yet resolved only a subset of consistent findings (see e.g. (Bhat et al., 2012; Hamshere et al., 2013)) among the presumably thousands of SNP-associations of small effect that are hypothesized to be involved (Binder, 2012; Moore et al., 2011), and which has as indeed also been demonstrated for e.g. schizophrenia (Purcell et al., 2009). Consequently, poly-SNP associations represent a valuable evolving approach, giving insight about the overall heritability mediated by many thousands SNPs at once, enabling comparisons of genetic overlaps between traits and being more powerful than single-SNP associations (Dudbridge, 2013).

1.2.

Gene-wide GWAS

For the case of genetic heterogeneity in a single gene, different random mutations can cause the same phenotypic (biological) effect, as simplistically exemplified by e.g. the mendelian phenylketonuria disorder (Scriver, 2007). Genetic heterogeneity may cause each SNP per se to have a lower power to be detected in a classical SNP-by-SNP GWAS as the relative importance of each SNP is diluted. Different strategies are continuously being published about how to best compile a gene-wide association signals from multiple SNP-signals. No golden standard exists in this domain, but the general procedure is to select the most significant SNP located within, or in LD with a gene, or to combine the signal from many different SNPs (Lehne et al., 2011; Li et al., 2012a). By doing so, the focus is shifted to association of known genes with the suicidality phenotype in question, facilitating biological interpretations and reducing the multiple comparisons burden (down to 20–30,000 tests).

1.3.

Pathway / network GWAS

In the case of complex polygenetic traits such as suicidality phenotypes, genetic heterogeneity can become problematic also at the gene-level. Namely, if different genes can drive the same phenotypic effect (pleiotrophy), or if multiple genes, each of small marginal effects (but with large, compounded epistatic effects) underlie the phenotype, then also each gene-wide effect will be small per se. Therefore, novel methods are being developed and applied to (re)analyze the associations from a SNP-GWAS in relation to (poly-)gene groupings based on prior knowledge, e.g. pathways (Holmans, 2010) and/or protein-protein interaction (PPI)

Summary of GWAS reviewed.

Ref

Design (sample name)

Sample size (cases)

Suicidality main outcome

 SNP odds ratio at power 0.8 d

Psychiatric diagnoses (treatment drug)

No. of SNPs (incl imputed) analyzed in GWAS

Association/interpretation levels Single SNPs (implicated gene or region)

Poly- Gene wide Pathway SNP network

TESI

4.2

MDD (citalopram)

109,365 ( )

rs11628713nn (PAPLN) rs10903034n (IL28RA)

–

–

–

Laje et al. (2009)

c–c, Matched 180 (90) (STARnD)

Perroud et al. (2012)

c–c (GENDEP) 706 (244) TWOSI (incl TESI)

2.0

MDD (escitalopram 539,199 nortriptylin) ( )

rs11143230n (GDA) rs358592n (KCNIP4) rs4732812n (ELP3)

–

–

–

Menke et al. (2012)

c-c (MARS)

5.3

Dep / BPD mix (mixed anti-dep)

rs1037448n – (TMEM138) rs10997044n (CTNNA3) rs1109089n (RHEB)

–

–

( 1.9 x106)

MDD: rs2576377nnn (ABI3BP) BPD or MDD: rs4918918n (SORBS1) rs10854398n (B3GALT5) rs1360550n (PRKCE)

–

–

–

397 (32)

TESI

371,335 ( )

Perlis et al. (2010) c–c (manya)

3117 (1295) 1273 (176) 8737 (2805)

SA

1.4

BPD

SA

2.2

MDD

SA

1.2

BPD or MDD

1.9 (SA)

MDD

532,774

Suicidality score: rs4751955n (GFRA1) SA: rs203136n (KIAA1244)

–

–

–

Wnt signaling and lithiumn

Schosser et al. (2011)

c–c (manyb)

Willour et al. (2012)

c–c see Perlis 5815 et al. (2010) (2496)

SA

1.3

BPD

 730,000 /516 024 ( 2.4 x106)

rs300774nnn (2p25; ACP1, SH3YL1, FAM150B)

–

–

Galfalvy et al. (2013)

c–c

99 (68)

SC

7.0

Mixed, with no diagnoses in controls

37,344 ( )

58n SNPs (19n genes)

–

CD44nand 4n pathwayse 6n genes e

Mullins et al. (2014)

c–c (manyc)

3270 (426)

SA

1.7

BPD or MDD

532,774

rs17173608 (RARRES2) Yes rs17387100 (PROM1) rs3781878 (NCAM1)

2023 (251 Suicidality SA) score or SA

–

GWAS on suicidal behaviors

Table 1

–

n

1569

reported or discussed as a (suggestive) association/relationship of interest for follow-up, e.g. with SNPs at Po10–5. Examples are shown, not complete listings. Genome-wide significant alpha 0.05 by experiment-wide correction. nnn Po5  10 8, genome-wide significant at alpha 0.05. nn

a Systematic Treatment Enhancement Program for Bipolar Disorder cohort (STEP-BP), Wellcome Trust Case Control Consortium bipolar cohort and University College London cohort for the GWAS-discovery in BPD subjects, with subsequent replication in the initial Genetic Association Information Network (GAIN) bipolar disorder project and the Translational Genomics Research Institute samples (both from the NIMH Bipolar Disorder Genetics Initiative; NIHM-BP) as well as a sample of German individuals collected at the Universities of Bonn and Heidelberg. The STARnD cohort was used for GWAS-discovery in MDD subjects, with replication in the GAIN depression cohort. b RADIANT Depression Case Control and Depression Network (using 3 samples; DeCC from U.K., DeNt from across of Europe and Bonn/Lausanne cases) for GWAS-discovery, with subsequent replication in GSK-Munich and Munich Antidepressant Response Signature (MARS) cohorts. c RADIANT, GSK-Munich and MARS cohorts as in Schosser et al. (2011), as well as Participants from a Bipolar Affective Disorder Case Control Study (BACCs) (n=470), London, United Kingdom. d The SNP-association odds ratio (OR) at 80% statistical power for the discovery analysis, given a minor allele frequency of 0.25 (log-additive model), population risk 0.01 and type I error rate set to the experiment-wide significance level (but no lower than 5  10 8). e 7 (out of 19) genes showed combined association with completed suicides and altered RNA expression (CYP19A1, MBNL2, KTBBD2, CD44, FOXN3, DSC2, CD300LB), with greatest confidence presented for CD44. Concurrently, pathways were also suggested for 8 (out of 19) genes suggestively associated with completed suicides: CNS development (MBNL2), homophilic cell adhesion (CD44, DSC2 and MARCH1), regulation of cell proliferation (SFRS11, TUBGCP3) and transmission of nerve impulse (LSAMP, SPTLC1).

1570

M. Sokolowski et al. networks (Akula et al., 2011; Hannum et al., 2009; Jia et al., 2011b; Rossin et al., 2011). Network analyses have been applied to screen e.g. GWAS data on Crohn's disease (Akula et al., 2011), as well as to analyze e.g. sets of candidate genes for schizophrenia (Jia et al., 2012; Sun et al., 2011) and major depressive disorder (Deo et al., 2013; Jia et al., 2011a). Networks can associate many different genes with the same biological functionality by “guilt-by-association” (Lee et al., 2011), and represent an interesting future avenue to develop and apply further for analysis of GWAS data on e.g. psychiatric conditions (Binder, 2012; Moore et al., 2011) as well as on suicidality. The aim herein was to review the current results of GWAS in the field of suicidality, describing also the different analytical approaches and study designs used.

2.

Experimental procedures

We searched PubMed (http://www.ncbi.nlm.nih.gov/) for original articles describing GWAS (as defined above) on suicidality outcomes, reading the 42 abstracts obtained after using the “ suicidn GWAS ” search string. We aimed to cover all GWAS on suicidality outcomes published in PubMed up to July 14th 2014 in the English language. Only GWAS which assayed SNPs were included. Linkage studies were excluded and have been reviewed elsewhere (Butler et al., 2010). Power for single SNP-associations (Table 1) was calculated by using software Quanto v.1.2.4 (Gauderman and Morrison, 2014).

3.

Review of GWAS in relation to suicidality

There are to date eight original GWAS published from different research groups/constellations in relation to different measures of suicidality (refer also to Table 1 for review Sections 3.1 and 3.2). As cases and controls were usually of the same psychiatric diagnosis, these GWAS assess association with suicidality per se (with their diagnosis controlled for), and if not being the case this was stated as a limitation.

3.1.

Suicidal ideation during treatment

Laje et al. (2009) conducted a GWAS on treatment emergent suicidal ideation (TESI), being the first ever GWAS conducted in relation to a suicidality measure. The subjects were non-psychotic major depressive disorders enrolled in the Sequenced Treatment Alternatives to Relieve (STARnD) cohort. Suicide ideation cases were defined by the emergence of any “Thoughts of Death or Suicide”, assessed using one question 12 on a 16-item Quick Inventory of Depressive Symptomatology self-report by the end of a 14 weeks of standard treatment routine with citalopram. One SNP (in PAPLN) was genome-wide significant (corrected P = 0.01, odds ratio 4.9) and another SNP (in IL28RA) was reported as suggestive (Table 1), in this subsample of STARnD. A fourSNPs multivariate model, including also previously nonGWAS identified SNPs in GRIA3 and GRIK2 (Laje et al., 2007), showed “reasonable fit” to observed data in both genders. The following limitations are also noteworthy: the instruments used were not primarily designed to assess ideation, the lack of placebo controls, the low genomic coverage of the SNP-array, only testing SNP-by-SNP GWAS

GWAS on suicidal behaviors and sufficient power only for large effect sizes of SNPs (odds ratios 44.2, due to the small sample sizes). Perroud et al. (2012) more recently investigated treatment worsening suicidal ideation (TWOSI) (which includes TESI as well as pre-treatment ideation cases), now during 12-week treatment with escitalopram or nortriptylin. TWOSI cases were defined as an increase on a computed score from three different items from three different depression-questionnaires, anytime during treatment. One SNP (rs11143230) was reported to have the strongest support for association with TWOSI, as also background haplotypes as well as other posthoc imputed SNPs showed association in the 88 kb LD-block containing the guanine deaminase (GDA) gene. In addition, treatment-specific SNPs (rs358592 – in KCNIP4, rs4732812 – near ELP3), drug x treatment interaction SNPs (rs1368607, rs2707159 – in APOO gene, rs2846685 – near p53AIP1 / RICS genes) as well as several genes (in NTRK2, CCK, YWHAE, SCN8A and CRHR2) from a supplemental candidate gene analysis (now using gene-wide statistics) were all reported as suggestive associations. No genotype x gender effect was observed. The study could not confirm the previous associations by Laje et al. (2007) on TESI, but there was a difference in suicide ideation measures used here and also a low numbers of cases (n=48) with TESI alone. The following limitations are also noteworthy: the instruments used were not primarily designed to assess ideation, the lack of placebo controls, only testing SNP-by-SNP GWAS, the lack of power to detect small effect sizes of SNPs (ORo2) with even lower power when dividing into each respective treatment, and that no associations were genome-wide significant. Finally, Menke et al. (2012) also performed a GWAS on TESI, with ideation measured by item 3 on the Hamilton Depression Rating Scale, during the first 12-weeks following treatment with a variety of different SSRIs. No significant associations were observed in the discovery sample, whereby 79 suggestive SNP associations with lowest Pvalues were instead used for testing in an independent replication sample. Fourteen SNPs were reported as suggestive associations after this two-stage analysis, of which 6 SNPs were in high LD (thus nine independent SNP associations remained; Table 1). Six of those SNPs could be annotated to five different genes (TMEM138, CTNNA3, RHEB, CYBASC3 and AIMI) and the other three SNPs had an intergenic location (Table 1). Finally, the SNP associations reported by Laje et al. could not be confirmed, but one must then consider the difference in suicide ideation measures used, the low numbers of cases and the mixed drug treatment used here. One SNP in GDA (gene suggested in GWAS by Perroud et al.) was suggestively associated with TESI, as well as were other candidate genes of neuropsychiatric disorders (FKBP5, ABCB1). The following limitations are also noteworthy: the instruments used were not primarily designed to assess ideation, the mixed drug treatments used, lack of placebo controls, sufficient power only for large effect sizes of SNPs (odds ratios 45.3), and that no associations were genome-wide significant.

3.2.

Suicidal behavior outcomes

The other five published GWAS focused on suicidality in behavioral domains, i.e. attempted or completed suicides.

1571 Perlis et al. (2010) was the first in this category to investigate lifetime suicide attempts in mood disorder patients from multiple cohorts, in a two-stage design, with analysis stratified by bipolar or major depression disorders. Controls were bipolar or major depression subjects without suicide attempts. However, no consistent findings were observed in such diagnosis-adjusted suicidality. One region was observed to be genome-wide significant in the major depression discovery STARnD cohort (with multiple SNPs annotated to gene ABI3BP), but this finding failed replication in the GAIN depression cohort (albeit having only 133 cases for replication). Finally, analysis was combined into one giant meta-analysis for the lifetime suicide attempt across either bipolar or depressed (n = 8737 in total, with n= 2805 suicide attempts, mostly bipolar), with four loci reported as suggestively associated (Table 1). Additionally, a gene-wide association analysis was conducted for 19 candidate genes, with observed suggestive association for FKBP5 and NGFR. The authors concluded that “inherited risk for suicide among mood disorder patients is unlikely to be the result of individual common variants of large effect”. The following limitations are also noteworthy: the samples were not primarily designed to study suicide attempts, the definitions of suicide attempt may have differed between and within cohorts, information about the severity/repetition / intent of the suicide attempts was missing, the lack of power to detect small effect sizes of SNPs (ORo2) in the depression cohorts, and only testing SNP-by-SNP GWAS. Next, Schosser et al. (2011) conducted a GWAS in relation to a suicidality score in MDD based on 2 items from the depression section in the SCAN interview, which reflected an increasing continuum from suicide ideation, self-harm to suicide attempts, as well as in relation to “serious” suicide attempts which scored highest on this scale (n=251; Table 1). However, no genome-wide significant results were observed and seven SNPs from three regions showed suggestive association (main examples in Table 1). Additionally, a gene-wide association analysis was conducted for 33 candidate genes, with varying suggestive associations for HTR1A, CCK, RGS18, NTRK2 and SCN8A across the different cohorts. The authors concluded “a genetic architecture of multiple genes with small effects” for suicidal behavior in depression. The following limitations are also noteworthy: the samples were not primarily designed to study suicide attempts, the definitions of suicide attempt may have differed between and within cohorts, the lack of power to detect small effect sizes of SNPs (ORo2), only testing SNP-by-SNP GWAS and that no associations were genome-wide significant. Willour et al. (2012) also conducted a GWAS on suicidal behavior in relation to lifetime suicide attempts in bipolar subjects, similar to that of Perlis et al. (2010). In fact, the same samples were used here as by Perlis et al. (2010), but inverting which ones were used for discovery vs replication; main results resulted from combining the discovery and replication samples herein. A two-stage analysis failed to replicate any of the 2507 SNP having Po0.001 in the discovery sample, after correction for multiple testing. However, a combined meta-analysis analysis of both the discovery and replication samples yielded a genome-wide significant association for SNP rs300774 (which also showed the lowest p-value in the discovery sample). LD-analysis of SNP rs300774 showed it was located in a large LD-block containing three genes

1572 (SH3YL1, ACP1 and FAM150B), as well as being near a consistently demonstrated linkage peak (at 2p11-12) for suicide attempts (Butler et al., 2010). Strength of the study is that they then assessed gene-expression data and found RNA expression changes in the prefrontal cortex of bipolar suicides for ACP1 (but not SH3YL1) and thus proposed the SNP association to be involved in ACP1 functions. But these results are also limited by that ACP1 was not the nearest gene of the three genes in the LD block (ACP1 is located 4100 kb away from rs300774) and with no further evidence presented as to why this SNP would be involved selectively in ACP1 function. Nevertheless, given also the functional involvement of ACP1 in the intracellular Wnt signaling cascade, a pathway regulated by lithium, this finding was reported as being the main one. Together, the authors discuss that their results gave support for “the existence of gene(s) influencing risk for suicidal behavior on the short arm of chromosome 2, an observation that is consistent with the linkage and association findings using pedigrees with BP, major depression and alcoholism”. The following limitations are also noteworthy: the samples were not primarily designed to study suicide attempts, the definitions of suicide attempt may have differed between and within cohorts, information about the severity / repetition / intent of the suicide attempts was missing, only testing SNPby-SNP GWAS and that the main finding only bordered to genome-wide significance. Galfalvy et al. (2013 but with e-pub already 2011 November 7) published what is the first and yet only available GWAS on completed suicides, using an alternative approach termed as a “pilot”. Controls were sudden deaths by other causes than suicide. A low-coverage array was used for association testing with completed suicides, and then the top-associated SNPs (nominal Po0.001) were then used to suggest genes of interest for follow-up by RNA-expression analysis of prefrontal and/or anterior cingulate cortex. Out of the 58 suggestively associated SNPs (indicated to be independent diagnoses or lifetime aggression scores), 22 SNPs were located in or near 19 known genes (Table 1). Nine of those 19 genes also showed altered RNA expression in suicides. Thus, the final analysis was to combine evidence from SNP-associations and gene-centric expression data (Table 1). The findings presented were both the 58 suggestively associated SNPs, 7 genes showing combined RNA gene expression alterations and suggestive SNP associations, as well as 4 different pathways for the subset of suggestive SNP associations which were annotated in or near 19 genes (Table 1). The main suggestive finding discussed was the CD44 gene, as it showed results on all the investigated parameters, as well as for its altered RNA expression in two previous studies. The authors concluded that the overall results “highlight a role for neuroimmunological effects in suicidal behavior”. The following limitations are also noteworthy: psychiatric diagnoses were not really controlled for, the exceptionally low genomic coverage of the SNP-array, problems with genotyping quality, sufficient power only for very large effect sizes of SNPs (odds ratios 47), only testing SNP-by-SNP GWAS and that no associations were genomewide significant. Finally, Mullins et al. (2014) is the most recent GWAS in this category. The samples for suicide attempts were the same as in Schosser et al. (2011) supplemented with another sample, and for TWOSI/TESI outcomes it was the sample of

M. Sokolowski et al. Perroud et al. (2012). Perhaps not surprisingly, genome-wide significant findings were again not observed in any of the SNP-by-SNP analyses, including by a meta-analysis on suicide attempts (Table 1). More interestingly, the authors then turned to poly-SNP analysis similar to those originally conducted in studies of Schizophrenia (Purcell et al., 2009), however to be regarded as preliminary due to powerreasons (Dudbridge, 2013). Poly-SNP scores very derived from each one of the four cohorts and then cross-tested for their predictive abilities on suicide attempts or TWOSI/TESI in any of the other cohorts. The GENDEP poly-SNP scores could predict suicide attempts in the RADIENT cohort. While no poly-SNP scores could predict suicide attempts in the GSK-Munich cohort, but the GSK-Munich poly-SNP scores could predict suicide attempts in the BACC cohort. Finally, no poly-SNP scores from any of the three suicide attempt cohorts (RADIENT, BACC or GSK-Munich) could predict TWOSI/TESI in GENDEP, whereas poly-SNP scores derived from the Psychiatric Genomics Consortium (PGC) cohort for MDD or for schizophrenia could. The latter results suggest that the genetics of psychiatric disorders are more related to TWOSI/TESI, rather than suicide attempts. However, PGC poly-SNP scores for MDD (but not other psychiatric diagnoses) could predict suicide attempts in all three (RADIENT, BACC or GSK-Munich) cohorts, also consistent with a genetic overlap between major depression and SA in particular. The authors conclude by that “the genetic architecture of SA and ideation is likely to be highly polygenic, with risk variants of small effect sizes which these GWAS studies were underpowered to detect, in a SNP-by-SNP analysis”, in line with the main theme of this review. The following limitations are also noteworthy: the samples were not primarily designed to study suicide attempts, the definitions of suicide attempt may have differed between and within cohorts, the lack of power to detect smaller effect sizes of SNPs (ORo1.7) and the limited power to derive accurate poly-SNP scores (Dudbridge, 2013).

4.

Conclusions and future considerations

Clearly, there are few (if any) genome-wide significant and reproducible findings yet to be demonstrated for suicidal outcomes (Table 1). But classical SNP-by-SNP GWAS approaches have managed to identify novel candidate genes for several complex diseases (Visscher et al., 2012), and follow-up re-sequencing have subsequently also revealed rare variants with higher effect sizes in such GWAS-associated genes (Butali et al., 2013; Diogo et al., 2013). Thus, classical SNP-by-SNP GWAS can point out a certain proportion of the heritability of complex traits (typically 5–10%). But the problem for suicidality research is those GWAS-successes have usually used samples in the sizes of tens- or even hundreds of thousands, which seems not obtainable in any foreseeable future for suicidality phenotypes. Is there some other way forward for success? GWAS statistical analyses up to date have broadly suffered from two problems which causes low power in the association testing; (i) the low effect sizes of “common” SNPs (those mainly analyzed in GWAS, in contrast to rare variants) and (ii) that classical SNP-by-SNP GWAS analysis do not capture complex polygenetic effects very well. One way to counteract this is to increase sample sizes i.e.

GWAS on suicidal behaviors by meta-analysis of several cohorts until even very small SNPeffects can theoretically be detected (typically in the thousands; see Table 1), but this power-gain strategy may not work as intended in the case of including heterogeneous populations or phenotypes (Liu et al., 2013). Another way to increase power is to adapt the statistical analysis (in smaller but more homogenous samples) and test a model which might better reflect a polygenetic architecture than single-SNP analyses, e.g. by incorporating prior knowledge about pathways or PPI networks (Binder, 2012; Califano et al., 2012). Novel computer simulations of evolution furthermore also shed more light on the nature of polygenetic etiology, namely that it is likely to reside in the suggestively significant P-value range of a classical SNP-by-SNP GWAS, and that any genome-wide significant SNP-associations might be “evolutionary outliers” which only reveal the tip of the genome-wide iceberg (Thornton et al., 2013). This importance of “suggestively significant” SNPs is also supported by poly-SNP analyses of liability which included thousands of such SNPs, showing potential to explain up to 30–40% of the expected heritability (Purcell et al., 2009; So et al., 2011; Vinkhuyzen et al., 2012). The recommendation for future GWAS on suicide outcomes is thus to not solely focus on increased sample sizes, but to also consider other statistical analytical approaches to complement the classical SNP-by-SNP GWAS, e.g. as was done in the most recent GWAS with the poly-SNP analysis. Furthermore, none of the samples/cohorts used for GWAS reviewed here were initially designed to primarily study suicidal outcomes, with one exception (Galfalvy et al., 2013). Therefore, the suicidal phenotypes could have been better ascertained from the start as well as more homogenous in their definition between different studies, to minimize e.g. misclassification biases between cases and controls and problems in replication attempts. We also see the need for conducting GWAS with family-based designs, as these designs do not suffer from concerns about biases in the control sample and enables better quality control checks. Our own Genetic Investigation of Suicide and Suicide attempts (GISS) sample was designed to primarily study suicidal outcomes per se as well as being family-based (Ben-Efraim et al., 2011, 2012, 2013; Sokolowski et al., 2012, 2010; Wasserman et al., 2007), and we are presently conducting GWAS analyses to complement the GWAS reviewed herein. The recommendation for future GWAS is mainly to ascertain the occurrence of suicidal outcomes better by collecting more detailed descriptions, at minimum using objective information about the level of selfinjury and actual circumstances of suicide attempts (Beck et al., 1975, 1976) or by use of established suicide ideation scales (Beck et al., 1979). On the other hand, the samples / cohorts used for GWAS on suicidality outcomes reviewed here were better ascertained and homogenous in relation to psychiatric diagnoses (mainly MDD and BPD), with one exception (Galfalvy et al., 2013). This meant that the case-control comparisons did not reflect SNP-association with those psychiatric disorders (as being controlled for by its presence in both case and controls), but with the suicidality outcomes per se. This is an advantage, as there is controversy surrounding the importance of psychiatric diagnoses per se in suicidal behavior (not ideation) outcomes, e.g. in relation to family co-transmission and aggregation (Brent et al., 1996; Johnson et al., 1998). Furthermore, trans-diagnostic

1573 approaches are also being suggested for psychiatry etiology research (Schumann et al., 2014) and recent GWAS studies suggest that many genetic susceptibility loci are not specific for distinct psychiatric diagnoses but rather shared across diagnostic boundaries (Schulze et al., 2012), exemplified by the recent results from the Cross-Disorder Group of the Psychiatric Genomics Consortium concerning calcium channel genes (using 5–10,000 cases per diagnostic category) (Smoller et al., 2013). GWAS on future samples might help to similarly unravel the relevance of diagnostic boundaries for the genome-wide genetics of suicidality outcomes. Given this uncertainty concerning psychiatric diagnoses in context of genetics research, investigations of suicidality behavior might perhaps instead be better refined by invoking intermediate phenotypes (Meyer-Lindenberg and Weinberger, 2006) or some form of biomarker (Le-Niculescu et al., 2013). Furthermore, biological state measures are required for improving individualized predictions by genetic data (Lehner, 2013). Studies on CACNA1C demonstrate these aspects of (cross-diagnostic) intermediate phenotypes as a prototypic example for genetics underlying the pathophysiology of psychiatric disease (Bhat et al., 2012; Smoller et al., 2013), and are perhaps relevant for suicidality phenotypes as well. The recommendation for future GWAS is to be aware about the controversy concerning the role of psychiatric diagnoses in the association between genetics and suicidal outcomes, and to consider the potential of biological measures in the associations with suicidal outcomes. Finally, there are many “rare” or segmental (e.g. CNVs, microsatellites) genetic variants, some of which are de novo (i.e. not inherited), which are not assayed in GWAS, but instead require e.g whole exome / genome sequencing (WES/WGS) to be revealed (a.k.a. “next generation” of GWAS). Here we should find more of the “missing heritability” than by only using “common” GWAS-SNPs, a heritability which is the range of 30–55% for suicidal behaviors (Brent and Melhem, 2008; Voracek and Loibl, 2007), although “rare” is not by default more useful than “common” variation (Lee et al., 2012; Muers, 2013). Indeed, WES/WGS genome screens conducted mostly “post-GWAS” are showing results for complex traits such as cholesterol, insulin and type 2 diabetes (Huyghe et al., 2013; Morrison et al., 2013; Steinthorsdottir et al., 2014), and has not yet been reported for suicidal outcomes. However, to couple large amounts of rare or even private gene-mutations to a phenotype by using solely WES/WGS remain challenging for complex disorders (MacArthur et al., 2014; Moutsianas and Morris, 2014), in addition to ongoing issues with the new technology (Dewey et al., 2014), and it is not to be regarded as a workaround for the other issues discussed herein for GWAS. It is recommended that sequencing can be used as a complement to GWAS, to obtain a better map of the genetic variants in GWAS-significant genes (Butali et al., 2013; Diogo et al., 2013) or well-established candidate genes, or as an alternative to GWAS for conducting “pilot” WDS / WES screens in smaller samples of well-defined extreme cases (sometimes using only one “discovery”-case), but then to also proceed with caution to minimize reporting of “false positives” (MacArthur et al., 2014). After addressing these issues discussed, one can then also move on to consider the contributions of genome-wide

1574 gene–environment interactions (Kendler, 2010), epigenetics (Kofink et al., 2013) and epistasis (Sun et al., 2014). In the end, there are no simple answers to the inherently complex question about the genetics of suicidality. But it is clear that GWAS remains to have lots of potential to help and provide some of the answers, which might ultimately be translated into better understanding and improvement of prevention, intervention and treatment options for suicidal behaviors.

Role of funding source The study was funded by the Knut and Alice Wallenberg Foundation. The funding organizations had no further role in study design; in the collection, analysis and interpretation of data; in the writing of the report; and in the decision to submit the paper for publication.

Contributors M.S. conducted the literature search and wrote the first draft, as well as the revised version of the review. D.W. and J.W. provided comments and suggestions during the writing process.

Conflict of interest The authors declare no conflict of interest.

Acknowledgment The Network on Suicide Research and Prevention of the European College of Neuropsychopharmacology (ECNP) commissioned this manuscript and contributed by acting as Reference Group and by providing comments and critical review to the manuscript. The article is a considerably revised reproduction of a recently published book chapter: Sokolowski, M., Wasserman, J., Wasserman, D., 2014. Genome-wide association studies of suicidal behaviors, in: Koslow, S., Ruiz, P., Nemeroff, C. (Eds.), A Concise Guide to Understanding Suicide: Epidemiology, Pathophysiology and Prevention. Cambridge University Press, Cambridge.

References Akula, N., Baranova, A., Seto, D., Solka, J., Nalls, M.A., Singleton, A., Ferrucci, L., Tanaka, T., Bandinelli, S., Cho, Y.S., Kim, Y.J., Lee, J.Y., Han, B.G., McMahon, F.J., 2011. A network-based approach to prioritize results from genome-wide association studies. PLoS One 6, e24220. Baca-Garcia, E., Perez-Rodriguez, M.M., Oquendo, M.A., Keyes, K.M., Hasin, D.S., Grant, B.F., Blanco, C., 2011. Estimating risk for suicide attempt: are we asking the right questions? Passive suicidal ideation as a marker for suicidal behavior. J. Affect. Disord. 134, 327–332. Beck, A.T., Beck, R., Kovacs, M., 1975. Classification of suicidal behaviors: I. Quantifying intent and medical lethality. Am. J. Psychiatry 132, 285–287. Beck, A.T., Weissman, A., Lester, D., Trexler, L., 1976. Classification of suicidal behaviors. II. Dimensions of suicidal intent. Arch. Gen. psychiatry 33, 835–837. Beck, A.T., Kovacs, M., Weissman, A., 1979. Assessment of suicidal intention: the Scale for Suicide Ideation. J. Consult. Clin. psychol. 47, 343–352. Ben-Efraim, Y.J., Wasserman, D., Wasserman, J., Sokolowski, M., 2011. Gene-environment interactions between CRHR1 variants

M. Sokolowski et al. and physical assault in suicide attempts. Genes Brain Behav. 10, 663–672. Ben-Efraim, Y.J., Wasserman, D., Wasserman, J., Sokolowski, M., 2012. Family-based study of HTR2A in suicide attempts: observed gene, gene  environment and parent-of-origin associations. Mol. Psychiatry. Ben-Efraim, Y.J., Wasserman, D., Wasserman, J., Sokolowski, M., 2013. Family-based study of AVPR1B association and interaction with stressful life events on depression and anxiety in suicide attempts. Neuropsychopharmacology. Bhat, S., Dao, D.T., Terrillion, C.E., Arad, M., Smith, R.J., Soldatov, N.M., Gould, T.D., 2012. CACNA1C (Ca(v)1.2) in the pathophysiology of psychiatric disease. Prog. Neurobiol. 99, 1–14. Binder, E.B., 2012. The genetic basis of mood and anxiety disorders – changing paradigms. Biol. Mood Anxiety Disord. 2, 17. Brent, D.A., Melhem, N., 2008. Familial transmission of suicidal behavior. Psychiatr Clin. N. Am. 31 (157-+). Brent, D.A., Bridge, J., Johnson, B.A., Connolly, J., 1996. Suicidal behavior runs in families. A controlled family study of adolescent suicide victims. Arch. Gen. Psychiatry 53, 1145–1152. Butali, A., Suzuki, S., Cooper, M.E., Mansilla, A.M., Cuenco, K., Leslie, E.J., Suzuki, Y., Niimi, T., Yamamoto, M., Ayanga, G., Erkhembaatar, T., Furukawa, H., Fujiwawa, K., Imura, H., Petrin, A.L., Natsume, N., Beaty, T.H., Marazita, M.L., Murray, J.C., 2013. Replication of genome wide association identified candidate genes confirm the role of common and rare variants in PAX7 and VAX1 in the etiology of nonsyndromic CL(P). Am. J. Med. Genet. A. Butler, A.W., Breen, G., Tozzi, F., Craddock, N., Gill, M., Korszun, A., Maier, W., Middleton, L.T., Mors, O., Owen, M.J., Perry, J., Preisig, M., Rice, J.P., Rietschel, M., Jones, L., Farmer, A.E., Lewis, C.M., McGuffin, P., 2010. A genomewide linkage study on suicidality in major depressive disorder confirms evidence for linkage to 2p12. Am. J. Med. Genet. B: Neuropsychiatr. Genet. 153B, 1465–1473. Califano, A., Butte, A.J., Friend, S., Ideker, T., Schadt, E., 2012. Leveraging models of cell regulation and GWAS data in integrative network-based association studies. Nat. Genet. 44, 841–847. Currier, D., Mann, J.J., 2008. Stress, genes and the biology of suicidal behavior. Psychiatr. Clin. N. Am. 31, 247–269. Deo, A.J., Huang, Y.Y., Hodgkinson, C.A., Xin, Y., Oquendo, M.A., Dwork, A.J., Arango, V., Brent, D.A., Goldman, D., Mann, J.J., Haghighi, F., 2013. A large-scale candidate gene analysis of mood disorders: evidence of neurotrophic tyrosine kinase receptor and opioid receptor signaling dysfunction. Psychiatr. Genet. 23, 47–55. Dewey, F.E., Grove, M.E., Pan, C., Goldstein, B.A., Bernstein, J.A., Chaib, H., Merker, J.D., Goldfeder, R.L., Enns, G.M., David, S.P., Pakdaman, N., Ormond, K.E., Caleshu, C., Kingham, K., Klein, T.E., Whirl-Carrillo, M., Sakamoto, K., Wheeler, M.T., Butte, A.J., Ford, J.M., Boxer, L., Ioannidis, J.P., Yeung, A.C., Altman, R.B., Assimes, T.L., Snyder, M., Ashley, E.A., Quertermous, T., 2014. Clinical interpretation and implications of whole-genome sequencing. JAMA: J. Am. Med. Assoc. 311, 1035–1045. Diogo, D., Kurreeman, F., Stahl, E.A., Liao, K.P., Gupta, N., Greenberg, J.D., Rivas, M.A., Hickey, B., Flannick, J., Thomson, B., Guiducci, C., Ripke, S., Adzhubey, I., Barton, A., Kremer, J.M., Alfredsson, L., Sunyaev, S., Martin, J., Zhernakova, A., Bowes, J., Eyre, S., Siminovitch, K.A., Gregersen, P.K., Worthington, J., Klareskog, L., Padyukov, L., Raychaudhuri, S., Plenge, R.M., 2013. Rare, low-frequency, and common variants in the protein-coding sequence of biological candidate genes from GWASs contribute to risk of rheumatoid arthritis. Am. J. Hum. Genet. 92, 15–27. Dudbridge, F., 2013. Power and predictive accuracy of polygenic risk scores. PLoS Genet. 9, e1003348.

GWAS on suicidal behaviors Eichler, E.E., Flint, J., Gibson, G., Kong, A., Leal, S.M., Moore, J.H., Nadeau, J.H., 2010. Missing heritability and strategies for finding the underlying causes of complex disease. Nat. Rev. Genet. 11, 446–450. Ernst, C., Mechawar, N., Turecki, G., 2009. Suicide neurobiology. Prog. Neurobiol. 89, 315–333. Galfalvy, H., Zalsman, G., Huang, Y.Y., Murphy, L., Rosoklija, G., Dwork, A.J., Haghighi, F., Arango, V., Mann, J.J., 2013. A pilot genome wide association and gene expression array study of suicide with and without major depression. World J. Biol. Psychiatry 14, 574–582. Gauderman, W., Morrison, J., 2014. Sample Size and Power Calculations for Genetic-epidemiology Studies. 〈http://bio stats.usc.edu/software〉. Goldstein, D.B., Ahmadi, K.R., Weale, M.E., Wood, N.W., 2003. Genome scans and candidate gene approaches in the study of common diseases and variable drug responses. Trends Genet. 19, 615–622. Hamshere, M.L., Walters, J.T., Smith, R., Richards, A.L., Green, E., Grozeva, D., Jones, I., Forty, L., Jones, L., Gordon-Smith, K., Riley, B., O'Neill, T., Kendler, K.S., Sklar, P., Purcell, S., Kranz, J., Morris, D., Gill, M., Holmans, P., Craddock, N., Corvin, A., Owen, M.J., O'Donovan, M.C., 2013. Genome-wide significant associations in schizophrenia to ITIH3/4, CACNA1C and SDCCAG8, and extensive replication of associations reported by the Schizophrenia PGC. Mol. Psychiatry 18, 708–712. Hannum, G., Srivas, R., Guenole, A., van Attikum, H., Krogan, N.J., Karp, R.M., Ideker, T., 2009. Genome-wide association data reveal a global map of genetic interactions among protein complexes. PLoS Genet. 5, e1000782. Holmans, P., 2010. Statistical methods for pathway analysis of genome-wide data for association with complex genetic traits. Adv Genet. 72, 141–179. Huyghe, J.R., Jackson, A.U., Fogarty, M.P., Buchkovich, M.L., Stancakova, A., Stringham, H.M., Sim, X., Yang, L., Fuchsberger, C., Cederberg, H., Chines, P.S., Teslovich, T.M., Romm, J.M., Ling, H., McMullen, I., Ingersoll, R., Pugh, E.W., Doheny, K.F., Neale, B.M., Daly, M.J., Kuusisto, J., Scott, L.J., Kang, H.M., Collins, F.S., Abecasis, G.R., Watanabe, R.M., Boehnke, M., Laakso, M., Mohlke, K.L., 2013. Exome array analysis identifies new loci and low-frequency variants influencing insulin processing and secretion. Nat. Genet. 45, 197–201. Ikegawa, S., 2012. A short history of the genome-wide association study: where we were and where we are going. Genomics Inform. 10, 220–225. Jia, P., Kao, C.F., Kuo, P.H., Zhao, Z., 2011a. A comprehensive network and pathway analysis of candidate genes in major depressive disorder. BMC Syst. Biol. 5 (Suppl. 3), S12. Jia, P., Zheng, S., Long, J., Zheng, W., Zhao, Z., 2011b. dmGWAS: dense module searching for genome-wide association studies in protein–protein interaction networks. Bioinformatics 27, 95–102. Jia, P., Wang, L., Fanous, A.H., Pato, C.N., Edwards, T.L., Zhao, Z., 2012. Network-assisted investigation of combined causal signals from genome-wide association studies in schizophrenia. PLoS Comput. Biol. 8, e1002587. Johnson, B.A., Brent, D.A., Bridge, J., Connolly, J., 1998. The familial aggregation of adolescent suicide attempts. Acta psychiatrica Scandinavica 97, 18–24. Kendler, K.S., 2010. Genetic and environmental pathways to suicidal behavior: reflections of a genetic epidemiologist. Eur. Psychiatry 25, 300–303. Kofink, D., Boks, M.P., Timmers, H.T., Kas, M.J., 2013. Epigenetic dynamics in psychiatric disorders: environmental programming of neurodevelopmental processes. Neurosci. Biobehav. Rev. 37, 831–845. Laje, G., Paddock, S., Manji, H., Rush, A.J., Wilson, A.F., Charney, D., McMahon, F.J., 2007. Genetic markers of suicidal ideation

1575 emerging during citalopram treatment of major depression. Am. J. Psychiatry 164, 1530–1538. Laje, G., Allen, A.S., Akula, N., Manji, H., JohnRush, A., McMahon, F.J., 2009. Genome-wide association study of suicidal ideation emerging during citalopram treatment of depressed outpatients. Pharmacogenet Genomics 19, 666–674. Lambert, C.G., Black, L.J., 2012. Learning from our GWAS mistakes: from experimental design to scientific method. Biostatistics 13, 195–203. Le-Niculescu, H., Levey, D.F., Ayalew, M., Palmer, L., Gavrin, L.M., Jain, N., Winiger, E., Bhosrekar, S., Shankar, G., Radel, M., Bellanger, E., Duckworth, H., Olesek, K., Vergo, J., Schweitzer, R., Yard, M., Ballew, A., Shekhar, A., Sandusky, G.E., Schork, N.J., Kurian, S.M., Salomon, D.R., Niculescu 3rd, A.B., 2013. Discovery and validation of blood biomarkers for suicidality. Mol. Psychiatry 18, 1249–1264. Lee, I., Blom, U.M., Wang, P.I., Shim, J.E., Marcotte, E.M., 2011. Prioritizing candidate disease genes by network-based boosting of genome-wide association data. Genome Res. 21, 1109–1121. Lee, S.H., DeCandia, T.R., Ripke, S., Yang, J., Schizophrenia Psychiatric Genome-Wide Association Study C., International Schizophrenia C., Molecular Genetics of Schizophrenia C., Sullivan, P.F., Goddard, M.E., Keller, M.C., Visscher, P.M., Wray, N.R., 2012. Estimating the proportion of variation in susceptibility to schizophrenia captured by common SNPs. Nat. Genet. 44, 247–250. Lehne, B., Lewis, C.M., Schlitt, T., 2011. From SNPs to genes: disease association at the gene level. PLoS One 6, e20133. Lehner, B., 2013. Genotype to phenotype: lessons from model organisms for human genetics. Nat. Rev. Genet. 14, 168–178. Li, M.X., Kwan, J.S., Sham, P.C., 2012a. HYST: a hybrid set-based test for genome-wide association studies, with application to protein-protein interaction-based association analysis. Am. J. Hum. Genet. 91, 478–488. Li, M.X., Yeung, J.M., Cherny, S.S., Sham, P.C., 2012b. Evaluating the effective numbers of independent tests and significant pvalue thresholds in commercial genotyping arrays and public imputation reference datasets. Hum. Genet. 131, 747–756. Liu, Y.J., Zhang, L., Pei, Y., Papasian, C.J., Deng, H.W., 2013. On genome-wide association studies and their meta-analyses: lessons learned from osteoporosis studies. J. Clin. Endocrinol. Metab. 98, E1278–1282. MacArthur, D.G., Manolio, T.A., Dimmock, D.P., Rehm, H.L., Shendure, J., Abecasis, G.R., Adams, D.R., Altman, R.B., Antonarakis, S.E., Ashley, E.A., Barrett, J.C., Biesecker, L.G., Conrad, D.F., Cooper, G.M., Cox, N.J., Daly, M.J., Gerstein, M.B., Goldstein, D.B., Hirschhorn, J.N., Leal, S.M., Pennacchio, L.A., Stamatoyannopoulos, J.A., Sunyaev, S.R., Valle, D., Voight, B.F., Winckler, W., Gunter, C., 2014. Guidelines for investigating causality of sequence variants in human disease. Nature 508, 469–476. Mann, J.J., Arango, V.A., Avenevoli, S., Brent, D.A., Champagne, F.A., Clayton, P., Currier, D., Dougherty, D.M., Haghighi, F., Hodge, S.E., Kleinman, J., Lehner, T., McMahon, F., Moscicki, E. K., Oquendo, M.A., Pandey, G.N., Pearson, J., Stanley, B., Terwilliger, J., Wenzel, A., 2009. Candidate endophenotypes for genetic studies of suicidal behavior. Biol. Psychiatry 65, 556–563. Menke, A., Domschke, K., Czamara, D., Klengel, T., Hennings, J., Lucae, S., Baune, B.T., Arolt, V., Muller-Myhsok, B., Holsboer, F., Binder, E.B., 2012. Genome-wide association study of antidepressant treatment-emergent suicidal ideation. Neuropsychopharmacology 37, 797–807. Meyer-Lindenberg, A., Weinberger, D.R., 2006. Intermediate phenotypes and genetic mechanisms of psychiatric disorders. Nat. Rev. Neurosci. 7, 818–827. Moore, S., Kelleher, E., Corvin, A., 2011. The shock of the new: progress in schizophrenia genomics. Curr. Genomics 12, 516–524.

1576 Morrison, A.C., Voorman, A., Johnson, A.D., Liu, X., Yu, J., Li, A., Muzny, D., Yu, F., Rice, K., Zhu, C., Bis, J., Heiss, G., O'Donnell, C. J., Psaty, B.M., Cupples, L.A., Gibbs, R., Boerwinkle, E., Cohorts for Heart and Aging Research in Genetic Epidemiology (CHARGE) Consortium, 2013. Whole-genome sequence-based analysis of highdensity lipoprotein cholesterol. Nat. Genet. 45, 899–901. Moutsianas, L., Morris, A.P., 2014. Methodology for the analysis of rare genetic variation in genome-wide association and resequencing studies of complex human traits. Brief. Funct. Genomics Jun 10. pii: elu012. [Epub ahead of print]. Muers, M., 2013. Complex traits: sequencing for disease architecture. Nat. Rev. Genet. 14, 518. Mullins, N., Perroud, N., Uher, R., Butler, A.W., Cohen-Woods, S., Rivera, M., Malki, K., Euesden, J., Power, R.A., Tansey, K.E., Jones, L., Jones, I., Craddock, N., Owen, M.J., Korszun, A., Gill, M., Mors, O., Preisig, M., Maier, W., Rietschel, M., Rice, J.P., Muller-Myhsok, B., Binder, E.B., Lucae, S., Ising, M., Craig, I.W., Farmer, A.E., McGuffin, P., Breen, G., Lewis, C.M., 2014. Genetic relationships between suicide attempts, suicidal ideation and major psychiatric disorders: a genome-wide association and polygenic scoring study. Am J Med Genet B Neuropsychiatr Genet 165B, 428–437. Need, A.C., Goldstein, D.B., 2010. Whole genome association studies in complex diseases: where do we stand? Dialogues Clin. Neurosci. 12, 37–46. Perlis, R.H., Huang, J., Purcell, S., Fava, M., Rush, A.J., Sullivan, P. F., Hamilton, S.P., McMahon, F.J., Schulze, T.G., Potash, J.B., Zandi, P.P., Willour, V.L., Penninx, B.W., Boomsma, D.I., Vogelzangs, N., Middeldorp, C.M., Rietschel, M., Nothen, M., Cichon, S., Gurling, H., Bass, N., McQuillin, A., Hamshere, M., Craddock, N., Sklar, P., Smoller, J.W., 2010. Genome-wide association study of suicide attempts in mood disorder patients. Am J Psychiatry 167, 1499–1507. Perroud, N., Uher, R., Ng, M.Y., Guipponi, M., Hauser, J., Henigsberg, N., Maier, W., Mors, O., Gennarelli, M., Rietschel, M., Souery, D., Dernovsek, M.Z., Stamp, A.S., Lathrop, M., Farmer, A., Breen, G., Aitchison, K.J., Lewis, C.M., Craig, I.W., McGuffin, P., 2012. Genome-wide association study of increasing suicidal ideation during antidepressant treatment in the GENDEP project. Pharmacogenomics J. 12, 68–77. Purcell, S.M., Wray, N.R., Stone, J.L., Visscher, P.M., O'Donovan, M. C., Sullivan, P.F., Sklar, P., 2009. Common polygenic variation contributes to risk of schizophrenia and bipolar disorder. Nature 460, 748–752. Rossin, E.J., Lage, K., Raychaudhuri, S., Xavier, R.J., Tatar, D., Benita, Y., Cotsapas, C., Daly, M.J., 2011. Proteins encoded in genomic regions associated with immune-mediated disease physically interact and suggest underlying biology. PLoS Genet. 7, e1001273. Rowe, S.J., Tenesa, A., 2012. Human complex trait genetics: lifting the lid of the genomics toolbox – from pathways to prediction. Curr. Genomics 13, 213–224. Rujescu, D., Giegling, I., 2010. The genetics of neurosystems in mental ill-health and suicidality: beyond serotonin. Eur. Psychiatry 25, 272–274. Schosser, A., Butler, A.W., Ising, M., Perroud, N., Uher, R., Ng, M. Y., Cohen-Woods, S., Craddock, N., Owen, M.J., Korszun, A., Jones, L., Jones, I., Gill, M., Rice, J.P., Maier, W., Mors, O., Rietschel, M., Lucae, S., Binder, E.B., Preisig, M., Perry, J., Tozzi, F., Muglia, P., Aitchison, K.J., Breen, G., Craig, I.W., Farmer, A.E., Muller-Myhsok, B., McGuffin, P., Lewis, C.M., 2011. Genomewide association scan of suicidal thoughts and behaviour in major depression. PLoS One 6, e20690. Schulze, T.G., Akula, N., Breuer, R., Steele, J., Nalls, M.A., Singleton, A.B., Degenhardt, F.A., Nothen, M.M., Cichon, S., Rietschel, M., McMahon, F.J., 2012. Molecular genetic overlap in bipolar disorder, schizophrenia, and major depressive disorder. World J. Biol. Psychiatry 15, 200–208.

M. Sokolowski et al. Schumann, G., Binder, E.B., Holte, A., de Kloet, E.R., Oedegaard, K.J., Robbins, T.W., Walker-Tilley, T.R., Bitter, I., Brown, V.J., Buitelaar, J., Ciccocioppo, R., Cools, R., Escera, C., Fleischhacker, W., Flor, H., Frith, C.D., Heinz, A., Johnsen, E., Kirschbaum, C., Klingberg, T., Lesch, K.P., Lewis, S., Maier, W., Mann, K., Martinot, J.L., Meyer-Lindenberg, A., Muller, C.P., Muller, W.E., Nutt, D.J., Persico, A., Perugi, G., Pessiglione, M., Preuss, U.W., Roiser, J.P., Rossini, P.M., Rybakowski, J.K., Sandi, C., Stephan, K.E., Undurraga, J., Vieta, E., van der Wee, N., Wykes, T., Haro, J.M., Wittchen, H.U., 2014. Stratified medicine for mental disorders. Eur. Neuropsychopharmacol. 24, 5–50. Scriver, C.R., 2007. The PAH gene, phenylketonuria, and a paradigm shift. Hum. Mutat. 28, 831–845. Sebastiani, P., Solovieff, N., Puca, A., Hartley, S.W., Melista, E., Andersen, S., Dworkis, D.A., Wilk, J.B., Myers, R.H., Steinberg, M.H., Montano, M., Baldwin, C.T., Perls, T.T., 2010. Genetic signatures of exceptional longevity in humans. Science, 2010. Sebastiani, P., Solovieff, N., Puca, A., Hartley, S.W., Melista, E., Andersen, S., Dworkis, D.A., Wilk, J.B., Myers, R.H., Steinberg, M.H., Montano, M., Baldwin, C.T., Perls, T.T., 2011. Retraction. Science 333, 404. Smoller, J.W., Craddock, N., Kendler, K., Lee, P.H., Neale, B.M., Nurnberger, J.I., Ripke, S., Santangelo, S., Sullivan, P.F., 2013. Identification of risk loci with shared effects on five major psychiatric disorders: a genome-wide analysis. Lancet 381, 1371–1379. So, H.C., Li, M., Sham, P.C., 2011. Uncovering the total heritability explained by all true susceptibility variants in a genome-wide association study. Genet. Epidemiol. 35, 447–456. Sokolowski, M., Wasserman, J., Wasserman, D., 2010. Association of polymorphisms in the SLIT2 axonal guidance gene with anger in suicide attempters. Mol. Psychiatry 15, 10–11. Sokolowski, M., Ben-Efraim, Y.J., Wasserman, J., Wasserman, D., 2012. Glutamatergic GRIN2B and polyaminergic ODC1 genes in suicide attempts: associations and gene–environment interactions with childhood/adolescent physical assault. Mol. Psychiatry, 985–992. Steinthorsdottir, V., Thorleifsson, G., Sulem, P., Helgason, H., Grarup, N., Sigurdsson, A., Helgadottir, H.T., Johannsdottir, H., Magnusson, O.T., Gudjonsson, S.A., Justesen, J.M., Harder, M.N., Jorgensen, M.E., Christensen, C., Brandslund, I., Sandbaek, A., Lauritzen, T., Vestergaard, H., Linneberg, A., Jorgensen, T., Hansen, T., Daneshpour, M.S., Fallah, M.S., Hreidarsson, A.B., Sigurdsson, G., Azizi, F., Benediktsson, R., Masson, G., Helgason, A., Kong, A., Gudbjartsson, D.F., Pedersen, O., Thorsteinsdottir, U., Stefansson, K., 2014. Identification of low-frequency and rare sequence variants associated with elevated or reduced risk of type 2 diabetes. Nat. Genet. 46, 294–298. Sun, J., Wan, C., Jia, P., Fanous, A.H., Kendler, K.S., Riley, B.P., Zhao, Z., 2011. Application of systems biology approach identifies and validates GRB2 as a risk gene for schizophrenia in the Irish Case Control Study of Schizophrenia (ICCSS) sample. Schizophr. Res. 125, 201–208. Teslovich, T.M., Musunuru, K., Smith, A.V., Edmondson, A.C., Stylianou, I.M., Koseki, M., Pirruccello, J.P., Ripatti, S., Chasman, D.I., Willer, C.J., Johansen, C.T., Fouchier, S.W., Isaacs, A., Peloso, G.M., Barbalic, M., Ricketts, S.L., Bis, J.C., Aulchenko, Y.S., Thorleifsson, G., Feitosa, M.F., Chambers, J., Orho-Melander, M., Melander, O., Johnson, T., Li, X., Guo, X., Li, M., Shin Cho, Y., Jin Go, M., Jin Kim, Y., Lee, J.Y., Park, T., Kim, K., Sim, X., Twee-Hee Ong, R., Croteau-Chonka, D.C., Lange, L.A., Smith, J.D., Song, K., Hua Zhao, J., Yuan, X., Luan, J., Lamina, C., Ziegler, A., Zhang, W., Zee, R.Y., Wright, A.F., Witteman, J.C., Wilson, J.F., Willemsen, G., Wichmann, H.E., Whitfield, J.B., Waterworth, D.M., Wareham, N.J., Waeber, G., Vollenweider, P., Voight, B.F., Vitart, V., Uitterlinden, A.G., Uda, M., Tuomilehto, J., Thompson, J.R., Tanaka, T., Surakka, I., Stringham, H.M., Spector, T.D., Soranzo, N., Smit, J.H.,

GWAS on suicidal behaviors Sinisalo, J.B., Silander, K., Sijbrands, E.J., Scuteri, A., Scott, J., Schlessinger, D., Sanna, S., Salomaa, V., Saharinen, J., Sabatti, C., Ruokonen, A., Rudan, I., Rose, L.M., Roberts, R., Rieder, M., Psaty, B.M., Pramstaller, P.P., Pichler, I., Perola, M., Penninx, B.W., Pedersen, N.L., Pattaro, C., Parker, A.N., Pare, G., Oostra, B.A., O'Donnell, C.J., Nieminen, M.S., Nickerson, D.A., Montgomery, G.W., Meitinger, T., McPherson, R., McCarthy, M.I., McArdle, W., Masson, D., Martin, N.G., Marroni, F., Mangino, M., Magnusson, P.K., Lucas, G., Luben, R., Loos, R.J., Lokki, M.L., Lettre, G., Langenberg, C., Launer, L.J., Lakatta, E.G., Laaksonen, R., Kyvik, K.O., Kronenberg, F., Konig, I.R., Khaw, K.T., Kaprio, J., Kaplan, L.M., Johansson, A., Jarvelin, M.R., Janssens, A.C., Ingelsson, E., Igl, W., Kees Hovingh, G., Hottenga, J.J., Hofman, A., Hicks, A.A., Hengstenberg, C., Heid, I.M., Hayward, C., Havulinna, A.S., Hastie, N.D., Harris, T.B., Haritunians, T., Hall, A.S., Gyllensten, U., Guiducci, C., Groop, L.C., Gonzalez, E., Gieger, C., Freimer, N.B., Ferrucci, L., Erdmann, J., Elliott, P., Ejebe, K.G., Doring, A., Dominiczak, A.F., Demissie, S., Deloukas, P., de Geus, E.J., de Faire, U., Crawford, G., Collins, F.S., Chen, Y.D., Caulfield, M.J., Campbell, H., Burtt, N.P., Bonnycastle, L.L., Boomsma, D.I., Boekholdt, S.M., Bergman, R.N., Barroso, I., Bandinelli, S., Ballantyne, C.M., Assimes, T.L., Quertermous, T., Altshuler, D., Seielstad, M., Wong, T.Y., Tai, E.S., Feranil, A.B., Kuzawa, C.W., Adair, L.S., Taylor Jr., H.A., Borecki, I.B., Gabriel, S.B., Wilson, J.G., Holm, H., Thorsteinsdottir, U., Gudnason, V., Krauss, R.M., Mohlke, K.L., Ordovas, J.M., Munroe, P.B., Kooner, J.S., Tall, A. R., Hegele, R.A., Kastelein, J.J., Schadt, E.E., Rotter, J.I., Boerwinkle, E., Strachan, D.P., Mooser, V., Stefansson, K., Reilly, M.P., Samani, N.J., Schunkert, H., Cupples, L.A., Sandhu, M.S., Ridker, P.M., Rader, D.J., van Duijn, C.M., Peltonen, L., Abecasis, G.R., Boehnke, M., Kathiresan, S., 2010. Biological, clinical and population relevance of 95 loci for blood lipids. Nature 466, 707–713. Sun, X., Lu, Q., Mukheerjee, S., Crane, P.K., Elston, R., Ritchie, M.D., 2014. Analysis pipeline for the epistasis search – statistical versus biological filtering. Front. Genet. 5, 106. Terwilliger, J.D., Goring, H.H., 2009. Update to Terwilliger and Goring's “Gene mapping in the 20th and 21st centuries“ (2000): gene mapping when rare variants are common and common variants are rare. Hum. Bio.l 81, 729–733.

1577 Thornton, K.R., Foran, A.J., Long, A.D., 2013. Properties and modeling of GWAS when complex disease risk is due to noncomplementing, deleterious mutations in genes of large effect. PLoS Genet. 9, e1003258. Vinkhuyzen, A.A., Pedersen, N.L., Yang, J., Lee, S.H., Magnusson, P.K., Iacono, W.G., McGue, M., Madden, P.A., Heath, A.C., Luciano, M., Payton, A., Horan, M., Ollier, W., Pendleton, N., Deary, I.J., Montgomery, G.W., Martin, N.G., Visscher, P.M., Wray, N.R., 2012. Common SNPs explain some of the variation in the personality dimensions of neuroticism and extraversion. Transl. Psychiatry 2, e102. Visscher, P.M., Brown, M.A., McCarthy, M.I., Yang, J., 2012. Five years of GWAS discovery. Am. J. Hum. Genet. 90, 7–24. Voracek, M., Loibl, L.M., 2007. Genetics of suicide: a systematic review of twin studies. Wien Klin Wochenschr. 119, 463–475. Wasserman, D., Geijer, T., Rozanov, V., Wasserman, J., 2005. Suicide attempt and basic mechanisms in neural conduction: relationships to the SCN8A and VAMP4 genes. Am. J. Med. Genet. B: Neuropsychiatr. Genet. 133B, 116–119. Wasserman, D., Geijer, T., Sokolowski, M., Rozanov, V., Wasserman, J., 2007. Genetic variation in the hypothalamic–pituitary–adrenocortical axis regulatory factor, T-box 19, and the angry/ hostility personality trait. Genes Brain Behav. 6, 321–328. Wasserman, D., Sokolowski, M., Rozanov, V., Wasserman, J., 2008. The CRHR1 gene: a marker for suicidality in depressed males exposed to low stress. Genes Brain Behav. 7, 14–19. Wasserman, D., Sokolowski, M., Wasserman, J., Rujescu, D., 2009. Neurobiology and the genetics of suicide. In: Wasserman, D., Wasserman, C. (Eds.), Oxford Textbook of Suicidology – The Five Continents Perspective. Oxford University Press, Oxford, U.K, pp. 165–182. Willour, V.L., Seifuddin, F., Mahon, P.B., Jancic, D., Pirooznia, M., Steele, J., Schweizer, B., Goes, F.S., Mondimore, F.M., Mackinnon, D.F., Perlis, R.H., Lee, P.H., Huang, J., Kelsoe, J.R., Shilling, P.D., Rietschel, M., Nothen, M., Cichon, S., Gurling, H., Purcell, S., Smoller, J.W., Craddock, N., DePaulo, J.R., Jr. Schulze, T.G., McMahon, F.J., Zandi, P.P., Potash, J.B., 2012. A genome-wide association study of attempted suicide. Mol. Psychiatry 17, 433–444.