Journal of Affective Disorders 183 (2015) 310–314
Contents lists available at ScienceDirect
Journal of Affective Disorders journal homepage: www.elsevier.com/locate/jad
Research report
Polygenic risk scores in bipolar disorder subgroups Sofie Ragnhild Aminoff a,b,n, Martin Tesli b, Francesco Bettella b, Monica Aas b, Trine Vik Lagerberg b, Srdjan Djurovic b, Ole A. Andreassen b, Ingrid Melle b a b
Department of Specialized Inpatient Treatment, Division of Mental Health Services, Akershus University Hospital, Norway NORMENT, KG Jebsen Centre for Psychosis Research, Oslo University Hospital and Institute of Clinical Medicine, University of Oslo, Oslo, Norway
art ic l e i nf o
a b s t r a c t
Article history: Accepted 10 May 2015 Available online 22 May 2015
Background: Bipolar disorder (BD) is a genetically and clinically heterogeneous disorder. Current classifications of BD rely on clinical presentations without any validating biomarkers, making homogenous and valid subtypes warranted. This study aims at investigating whether a BD polygenic risk score (PGRS) can validate BD subtypes including diagnostic sub-categories (BD-I versus BD-II), patients with and without psychotic symptoms, polarity of first presenting episode and age at onset based groups. We also wanted to investigate whether illness severity indicators were associated with a higher polygenic risk for BD. Methods: Analyze differences in BD PGRS scores between suggested subtypes of BD and between healthy controls (CTR) and BD in a sample of N ¼669 (255 BD and 414 CTR). Results: The BD PGRS significantly discriminates between BD and CTR (po 0.001). There were no differences in BD PGRS between groups defined by diagnostic sub-categories, presenting polarity and age at onset. Patients with psychotic BD had nominally significantly higher BD PGRS than patients with nonpsychotic BD after controlling for diagnostic sub-category (p¼ 0.019). These findings remained trend level significant after Bonferroni corrections (p¼ 0.079). Limitations: The low explained variance of the current PGRS method could lead to type II errors. Conclusions: There are nominally significant differences in BD PGRS scores between patients with and without psychotic symptoms, indicating that these two forms of BD might represent distinct subtypes of BD based in its polygenic architecture and a division between BD with and without psychotic symptoms could represent a more valid subclassification of BD than current diagnostic sub-categories. If replicated, this finding could affect future research, diagnostics and clinical practice. & 2015 Elsevier B.V. All rights reserved.
Keywords: Bipolar disorder Polygenic risk score Psychosis Depression Mania Hypomania
1. Introduction Bipolar disorder (BD) has a complex phenotype characterized by manic or hypomanic episodes with the addition of depressive episodes in 80–95% of cases (Baek et al., 2014), psychotic features in 20–60% of cases (Goodwin and Jamison, 2007; Mazzarini et al., 2010) and with significant individual differences in illness course and severity (Ketter, 2010; Sanchez-Moreno et al., 2009; Treuer and Tohen, 2010). One of the aims of BD research has thus been the identification of valid subgroups to aid treatment planning, course prediction and the search for the genetic underpinnings of BD (Aminoff et al., 2013; Craddock et al., 2005; Craddock and Owen, 2010). Current classifications of BD subtypes however rely on clinical presentations without any validating biomarkers n Corresponding author at: NORMENT, KG Jebsen Centre for Psychosis Research – TOP Study, Building 49, Oslo University Hospital, Ullevål, Kirkeveien 166, PO Box 4956 Nydalen, 0424 Oslo, Norway. Tel.: þ47 23 02 73 50; fax: þ 47 23 02 73 33. E-mail address: [email protected] (S.R. Aminoff).
http://dx.doi.org/10.1016/j.jad.2015.05.021 0165-0327/& 2015 Elsevier B.V. All rights reserved.
(O'Donovan et al., 2009; Phillips and Kupfer, 2013). This includes the DSM sub-categorization into Bipolar I (BDI) versus Bipolar II (BDII) disorders based on the presence of mania (BDI) versus depression and hypomania (BDII). Other subtypes have also been suggested; based on the presence of psychotic symptoms (Simonsen et al., 2011), differences in age at onset (Azorin et al., 2013) or the polarity of the first (presenting) mood episode (Perugi et al., 2001). BD is a multifactorial disorder with a strong genetic component (Geoffroy et al., 2013). Genome-wide association studies (GWAS) indicate the involvement of several common single nucleotide variants of small effects (Craddock and Sklar, 2013; PGC, 2011; Purcell et al., 2009), and have identified genes involved in neuronal growth and cell adhesion, calcium signaling, hormone regulation, second messenger and glutamate systems (Nurnberger et al., 2014; Sklar et al., 2011) with considerable genetic overlap between BD and schizophrenia (SCZ) and unipolar depression (Andreassen et al., 2013; Cardno and Owen, 2014; Craddock et al., 2009, 2005; Lee et al., 2013; Nurnberger et al., 2014; PGC, 2013;
S.R. Aminoff et al. / Journal of Affective Disorders 183 (2015) 310–314
van Os and Tamminga, 2007). Recent approaches take into consideration the multiplicity of genes involved in psychiatric disorders by providing information on the cumulative genomic risk (Lee et al., 2013; Purcell et al., 2009; Ripke et al., 2013) accounting for a larger proportion of the phenotypic variance (Walton et al., 2013). By using summary statistics representing differences in genetic variants between cases and controls in a discovery study, one can assign a polygenic risk score (PGRS) to individuals in the independent study sample. The PGRS method has been shown to predict case-control status at high significance levels for both BD and schizophrenia (SCZ) (Purcell et al., 2009; Ripke et al., 2013). At clinical level, associations have been found between BD PGRS and mania in patients with SCZ (Ruderfer et al., 2013). At proposed intermediate phenotype level, BD PGRS has been shown to correlate with increased brain activation in limbic regions during a language paradigm in CTR and individuals with a family load of BD (Whalley et al., 2012). The effect of a major depressive disorder PGRS was shown to be increased in cases with additional childhood trauma (Peyrot et al., 2014). Recently, the PGRS method also provided support for the psychosis continuum model in a sample with BD spectrum disorders, SCZ spectrum disorders and healthy controls (Tesli et al., 2014). To our knowledge, the PGRS method has not previously been used to validate subtypes of BD within a BD sample. The aim of the current study is to investigate whether BD PGRS can validate previously suggested BD subtypes (i.e. BDI versus BDII, psychotic- versus non-psychotic BD, subtypes based on age-atonset groups or the polarity of the presenting episode). We also explored whether the BD PGRS was associated with indicators of illness severity in BD.
2. Methods 2.1. Participants All participants were consecutively recruited to the on-going Thematically Organized Psychosis study (TOP). Patients were recruited from outpatient and inpatient units of the three major hospitals in Oslo, Norway, between 2003 and 2011. Inclusion criteria for this particular study were age between 17 and 65 years and having a DSM-IV diagnosis of BD type I or II (total N ¼ 255; BDI n ¼181 and BDII n ¼ 74). 134 patients (74%) from the BDI sample and 14 patients (19%) from the BDII sample had a history of psychosis. 90 (35%) patients had a first, or presenting, episode characterized by elevated mood (mania, mixed or hypomania) and 153 (60%) had a depressive presenting episode (information uncertain or missing in n ¼12 (4%)). 105 (41%) patients were employed and 88 (34.5%) married or cohabitant. Mean duration of treatment was 6 79 years. Family history was based on a semi-structured interview asking patients about the presence of a probable or sure diagnosis of BD or schizophrenia in first-degree relatives (parents, siblings or children). 51 patients (21%) had first degree relatives with diagnosis of BD and 10 patients (4%) had first degree relatives with a diagnosis of SCZ. For other demographic details see Table 1. The CTR sample (N ¼414) was randomly drawn from the statistical population registers of the same areas of Oslo as the patients and was contacted by mail. The entire sample consisted of Northern European Caucasians, mainly Norwegians, previously demonstrated to be genetically homogenous (Athanasiu et al., 2010; Djurovic et al., 2010). General exclusion criteria were hospitalized head injury, neurological disorder, unstable or uncontrolled medical condition that interferes with brain function and/or an IQ below 70. The CTR sample was additionally screened for physical and mental disorders, a history of head injury, ongoing drug abuse, and a family history of severe mental disorders. The Regional
311
Table 1 Demographics. Healthy controls Bipolar disorder Sample size, N Gender, N (%) Male Female Age (mean 7SD) Age at onset (mean 7SD) Duration of illness (mean7 SD) Bipolar disorder in first degree relatives, N (%)
414
255
208 (50.2) 206 (49.8) 357 10
101 (39.6) 154 (60.4) 367 12 237 9 13 710 51 (20)
Missing data in N patients ¼7 (3%).
Committee for Medical Research Ethics and the Norwegian Data Inspectorate approved the study, and participants' written, informed consents according to the Declaration of Helsinki were obtained. 2.2. Clinical assessment Patients were clinically characterized based on a personal interview by trained assessment staff that had completed the TOP Study's assessment training and reliability program, either medical doctors or clinical psychologists. Diagnosis was based on the Structured Clinical Interview for DSM-IV Axis I disorders (SCID-I) (First et al., 1995) and information from medical charts. A good inter-rater reliability for diagnoses was achieved with an overall kappa score of 0.77 (95% confidence interval ¼0.60–0.94) (Ringen et al., 2008). A history of psychosis was defined as having one or more lifetime psychotic episodes. We defined polarity of presenting episode as the polarity of the first SCID verified mood episode, divided into depressive or elevated onset (including mixed episodes). Age at onset was defined as the age of the first SCID-verified mood episode. Age was collapsed into three groups based on results from previous admixture analysis in large samples finding relatively stable age-at-onset groups in different cultures and birth cohorts: (1) early onset (first episode r22 years), (2) intermediate onset (first episode 23–34) and (3) late onset (first episode Z35 years) (Bellivier et al., 2001, 2003; Hamshere et al., 2009; Leboyer et al., 2005; Manchia et al., 2008). 2.3. Genotyping All participants were genotyped at Expression Analysis Inc. (Durham, NC, USA) using the Affymetrix Genome-Wide Human SNP Array 6.0 (Affymetrix Inc., Santa Clara, CA, USA). Quality control was performed using PLINK (version 1.07; http://pngu. mgh.harvard.edu/purcell/plink/) (Purcell et al., 2007). As a quality control, exclusions of individuals based on genotyping were made of (I) one of two duplicates, (II) one of two relatives (identity by descent (IBD) 40.1875), (III) individuals with a recorded gender differing from that determined by X chromosome marker homozygosity, (IV) mixup-samples (calculated by pairwise genomewide identity by state (IBS)), (V) individuals with non-European ancestry (calculated with HapMap3 and PLINK's first two MDS components) and (VI) individuals with more than 5% missing genotype data. SNPs were excluded based on (I) severe deviation from Hardy–Weinberg equilibrium, (II) minor allele frequency below 1% and (III) low yield ( o95% in controls), as described in earlier reports (Athanasiu et al., 2010; Djurovic et al., 2010). PLINK's multidimensional scaling (MDS) components were used as ancestry covariates in the analyses. The ancestry-based exclusion affected outliers in the space generated by the first two MDS components. It is plausible that once the outliers are removed, the
312
S.R. Aminoff et al. / Journal of Affective Disorders 183 (2015) 310–314
third MDS component acquires a higher discriminating power. We therefore included the first three MDS components to account for ancestry in the subsequent analyses.
episodes adjusted for duration of illness was investigated with multivariate hierarchical regression analysis.
2.4. Imputation of SNPs
3. Results
Following the above mentioned quality control, SNPs were imputed with MaCH (Li et al., 2010) (http://www.sph.umich.edu/ csg/abecasis/MACH/download/1000G-PhaseI-Interim.html) using the European samples in the Phase I release of the 1000 Genomes project. SNPs not present in the 1000 Genomes reference, and SNPs with ambiguous strand alignments (A/T and G/C SNPs), were removed from the sample data sets. Imputation was a three stage process, involving (I) ChunkChromosome where the data set was broken into 2500 SNP pieces with 500 SNP overlap (http://gen ome.sph.umich.edu/wiki/ChunkChromosome), (II) MaCH where each piece was phased (40 rounds and 400 states) (http://www. sph.umich.edu/csg/abecasis/MaCH/download/), and (III) Minimac where each phased piece was imputed to the 1000 Genomes European reference panel (20 rounds and 400 states) (http://gen ome.sph.umich.edu/wiki/Minimac). In the third stage, all imputed SNPs were provided with an estimated r2 score as quality metric. Exclusions were made of SNPs with MAF o0.01 or an r2 score o0.2, leaving 8,331,417 SNPs.
Taken ancestry into account the BD PGRS for the BD sample was 0.153 (70.63) (standard error), significantly higher than for the CTR group (mean BD PGRS score 0.179 70.50, po 0.001). In the BD sample, there were no statistically significant differences between males and females, F(1,250) ¼0.188, p ¼0.665), or between patients with or without first degree relatives with BD, F (1,243)¼0.281, p ¼ 0.596. Patients with a history of psychosis had nominally higher BD PGRS scores than those without when controlling for ancestry (p ¼0.049) (Table 2). When also controlling for diagnostic subcategory the findings were slightly strengthened (p ¼0.019). However these results remained only trend level significant after Bonferroni-corrections for four group comparisons (p ¼0.079). There were no significant differences in BD PGRS between the diagnostic subcategories (BDI versus BDII), between those with a depressive presenting polarity versus those with an elevated presenting polarity or across different age at onset groups (Table 2). In the follow up analysis, there was also no significant interaction effect between history of psychosis and diagnostic subcategory, F(1,251) ¼0.490, p ¼0.485. The BD PGRS was not associated with any severity indicators, including number of mood episodes and number of mood episodes adjusted for duration of illness (depressive episodes, F(4,233) ¼1.060, p ¼0.377; elevated episodes, F(4,234) ¼1.069, p ¼0.372; total episodes, F(4,238) ¼ 1.121, p¼ 0.347).
2.5. Polygenic risk score BD PGRS were computed based on imputed SNPs following the method developed by Purcell et al. (2009). Using PLINK version 1.07 (http://pngu.mgh.harvard.edu/purcell/plink/) (Purcell et al., 2007), we performed a meta-analysis including all Psychiatric Genomics Consortium (PGC) substudies (PGC, 2011) except the one based on the sample used for the current study (TOP3 in reference publication) (N ¼ 7278 BD cases and 8901 controls) and obtained risk allele effect sizes (ln(OR)) for all imputed SNPs. In order to avoid any inflation due to LD patterns across the genome, the SNPs were subsequently pruned using PLINK's-clump option (r2 o0.25, 500 kb windows) to select representatives with lowest p-Values from all LD blocks (209088 SNPs). Using the clumped SNP set, 10 PGRS were computed including SNPs at different p-Value thresholds: p ¼1, p ¼0.5, p ¼0.4, p¼ 0.3, p ¼0.2, p ¼0.1, p ¼0.05, p ¼0.01, p ¼0.001, and p ¼0.0001. For the subsequent analyses, we selected the PGRS resulting from the p-Value threshold of 0.05 (23062 SNPs), since this was explaining most phenotypic variance (Nagelkerke pseudo-r2). For more details see Tesli et al. (2014). The selected BD PGRS was transformed into standard scores before proceeding with the subsequent analyses. 2.6. Statistical analyses All analyses from this point onward were performed using the IBM SPSS, version 20. The BD PGRS was normally distributed across all diagnostic subgroups (Kolmogorov Smirnov¼0.200). For analyses of PGRS group differences we used ANCOVAs with three MDS components to control for ancestry. Bivariate associations were measured through Pearson's correlations (r). All tests were two sided with a p o0.05. We made Bonferroni corrections based on conducting four separate group comparisons (diagnostic subcategories, presenting polarity, age at onset groups and psychotic BD). Since there is a considerable overlap between having a BDI diagnosis and the risk for psychotic symptoms, we included a follow-up, two-way ANOVA to investigate interaction effects on BD PGRS between sub-diagnosis and history of psychosis despite minor differences in the bivariate analyses. The relationship between BD PGRS and number of total, depressive and elevated
4. Discussion The main finding of the study was that there were no statistically significant differences in the BD PGRS for groups defined by diagnostic sub-categories (BD-I versus BD-II), presenting polarity or age at onset subgroups. These findings do not support the notion that these are genetically different subgroups of BD, at least as based in the PGRS. There was however a nominally statistically Table 2 Between group differences in bipolar polygenic risk score in a bipolar sample. Classification
Total sample: N ¼ 255 Diagnostic subcategory: BDI BDII History of psychosis: Yes No Polarity of first presenting episode: Depressive Elevated Missing data Age at onset: Early Mid Late Missing data
N (%)
BD PGRS mean 7S. D.
ANCOVA with 3 MDS as covariates
0.1557 1.00 F(1,250)¼ 0.012, p ¼0.913 181 (71) 74 (29)
0.167 0.99 0.157 1.05 F(1,250) ¼ 3.915, p¼ 0.049
148 (58) 107 (42)
0.277 0.96 0.004 7 1.04 F(1,249)¼ 0.566, p ¼0.568
152 (60) 91 (36) 12 (4)
0.147 0.99 0.20 7 1.05 F(2,247) ¼ 0.469, p ¼0.628
150 71 32 2
(59) (28) (12) (1)
0.107 1.02 0.277 1.00 0.177 0.95
S.R. Aminoff et al. / Journal of Affective Disorders 183 (2015) 310–314
313
Fig. 1. Polygenic risk scores in bipolar disorder subtypes (mean z scores and standard errors). Abbreviations: CTR ¼ Healthy controls, BDI ¼ Bipolar type I disorder and BD2 ¼Bipolar type 2 disorder.
significant difference in BD PGRS between patients with BD with and without psychotic features. In a subsample of the current, we have previously shown that BD diagnostic sub-category, presenting polarity and history of psychosis, was associated with underlying neurocognitive differences, suggesting that they might represent distinct BD subtypes (Aminoff et al., 2013). The link to neurocognitive dysfunctions is in line with studies indicating that BD and SCZ share genetic basis with cognitive factors (Fernandes et al., 2013; Reilly and Sweeney, 2014). If replicated, the finding of different BD PGRS between patients with and without a history of psychosis would suggest that the distinction between psychotic BD and non-psychotic BD could be the more valid categorization, reflecting underlying genetic differences to a greater extent than the BDI–BDII distinction (Fig. 1). This might seem at odds with one study comprising a broad spectrum of psychotic disorders (BD- and SCZ spectrums plus psychosis not otherwise specified) that indicated a relationship between lifetime psychosis and SCZ PGRS but not BD PGRS (Tesli et al., 2014). However, the main part of the patients with lifetime psychosis in the previous study was part of the SCZ- and not the BD spectrum. Another study did not find significant differences between BD individuals with and without psychotic features using a SCZ PGRS (Hamshere et al., 2011). A potential limitation of the current study is related to the construction of the PGRS. The BD PGRS is not an absolute measure of the genetic risk for BD but a relative measure of congruence with the complex genetic structure of cases and controls in the initial discovery sample and thus dependent of its sample characteristics. The present results, with higher BD PGRS in psychotic BD could be taken to imply that this discovery sample has a predominance of patients with psychotic BD. Since more than half of patients with BD experience psychotic symptoms (Goodwin and Jamison, 2007) this appears as a probable explanation. An important limitation is the explained variance of the current PGRS method, which is low (2.4%). This may lead to type II errors, which suggest the current findings cannot rule out that differences in polygenic risk related to other BD subgroups and clinical characteristics could be detected investigating larger samples.
data, in the writing of the manuscript or in the decision to submit the article for publication. Conflict of interest All authors declare no conflict of interest.
Contributors Sofie R. Aminoff has written the manuscript and contributed to the study design, data collection, data analyses and the discussion of findings. Martin Tesli has written parts of the manuscript and contributed to study design, data collection, data analyses and discussion of findings. Francesco Bettella has contributed to study design, data analysis, manuscript and discussion of findings. Monica Aas has contributed to the manuscript and discussion of findings. Trine Vik Lagerberg has contributed to the study design, data collection, data analyses, manuscript and the discussion of findings. Srdjan Djurovic has contributed to the manuscript and discussion of findings. Ole A. Andreassen has contributed to the manuscript and discussion of findings. Ingrid Melle has contributed to the study design, the data analyses, the manuscript and the discussion of findings. All authors have approved the final article.
Acknowledgments We would particularly like to thank the study participants for their contribution and the clinicians and research staff who contributed to patient recruitment and management of data. This work was supported by Grants from the Norwegian Research Council (#421716 and #223273), the Regional Health Authority SouthEastern Norway (#N1, #2011085 and #2013123) and from the K.G. Jebsen Foundation.
References Role of the funding source This work was supported by Grants from the Norwegian Research Council (#421716 and #223273), the Regional Health Authority South-Eastern Norway (#N1, #2011085 and #2013123) and from the K.G. Jebsen Foundation. Neither had any involvement in study design, collection of data, analysis and interpretation of
Aminoff, S.R., Hellvin, T., Lagerberg, T.V., Berg, A.O., Andreassen, O.A., Melle, I., 2013. Neurocognitive features in subgroups of bipolar disorder. Bipolar Disord. 15 (3), 272–283. http://dx.doi.org/10.1111/bdi.12061. Andreassen, O.A., Thompson, W.K., Schork, A.J., Ripke, S., Mattingsdal, M., Kelsoe, J.R., 2013. Bipolar Disorder and Schizophrenia Working Groups, 2013. Improved detection of common variants associated with schizophrenia and bipolar disorder using pleiotropy-informed conditional false discovery rate. PLoS Genet. 9 (4), e1003455. http://dx.doi.org/10.1371/journal.pgen.1003455.
314
S.R. Aminoff et al. / Journal of Affective Disorders 183 (2015) 310–314
Athanasiu, L., Mattingsdal, M., Kähler, A.K., Brown, A., Gustafsson, O., Agartz, I., Andreassen, O.A., 2010. Gene variants associated with schizophrenia in a Norwegian genome-wide study are replicated in a large European cohort. J. Psychiatry Res. 44 (12), 748–753. http://dx.doi.org/10.1016/j. jpsychires.2010.02.002. Azorin, J.M., Bellivier, F., Kaladjian, A., Adida, M., Belzeaux, R., Fakra, E., Golmard, J.L., 2013. Characteristics and profiles of bipolar I patients according to age-atonset: findings from an admixture analysis. J. Affect. Disord. 150 (3), 993–1000. http://dx.doi.org/10.1016/j.jad.2013.05.026. Baek, J.H., Eisner, L.R., Nierenberg, A.A., 2014. Epidemiology and course of unipolar mania: results from the national epidemiologic survey on alcohol and related conditions (NESARC). Depress. Anxiety 0, 1–10. http://dx.doi.org/10.1002/ da.22261. Bellivier, F., Golmard, J.L., Henry, C., Leboyer, M., Schurhoff, F., 2001. Admixture analysis of age at onset in bipolar I affective disorder. Arch. Gen. Psychiatry 58 (5), 510–512. Bellivier, F., Golmard, J.L., Rietschel, M., Schulze, T.G., Malafosse, A., Preisig, M., Leboyer, M., 2003. Age at onset in bipolar I affective disorder: further evidence for three subgroups. Am. J. Psychiatry 160 (5), 999–1001. Cardno, A.G., Owen, M.J., 2014. Genetic relationships between schizophrenia, bipolar disorder, and schizoaffective disorder. Schizophr. Bull. 40 (3), 504–515. http://dx.doi.org/10.1095043/schbul/sbu016. Craddock, N., O’Donovan, M.C., Owen, M.J., 2009. Psychosis genetics: modeling the relationship between schizophrenia, bipolar disorder, and mixed (or “schizoaffective”) psychoses. Schizophr. Bull. 35 (3), 482–490. http://dx.doi.org/ 10.1093/schbul/sbp020, sbp020. Craddock, N., O’Donovan, M.C., Owen, M.J., 2005. The genetics of schizophrenia and bipolar disorder: dissecting psychosis. J. Med. Genet. 42 (3), 193–204. http://dx. doi.org/10.1136/jmg.2005.030718. Craddock, N., Owen, M.J., 2010. The Kraepelinian dichotomy – going, going… but still not gone. Br. J. Psychiatry 196 (2), 92–95. http://dx.doi.org/10.1192/bjp. bp.109.073429 196/2/92. Craddock, N., Sklar, P., 2013. Genetics of bipolar disorder. Lancet 381 (9878), 1654–1662. http://dx.doi.org/10.1016/S0140-6736(13)60855-7. Djurovic, S., Gustafsson, O., Mattingsdal, M., Athanasiu, L., Bjella, T., Tesli, M., Andreassen, O.A., 2010. A genome-wide association study of bipolar disorder in Norwegian individuals, followed by replication in Icelandic sample. J. Affect. Disord. 126 (1–2), 312–316. http://dx.doi.org/10.1016/j.jad.2010.04.007. Fernandes, C.P., Christoforou, A., Giddaluru, S., Ersland, K.M., Djurovic, S., Mattheisen, M., Le Hellard, S., 2013. A genetic deconstruction of neurocognitive traits in schizophrenia and bipolar disorder. PLoS One 8 (12), e81052. http://dx.doi.org/10.1371/ journal.pone.0081052. First, M., Spitzer, R., Gibbon, M., Williams, J.B.W., 1995. Structured Clinical Interview for DSM-IV Axis I Disorders: Patient Edition (SCIP-P), Version 2. New York State Psychiatric Institute, Biometrics Research, New York. Geoffroy, P.A., Etain, B., Houenou, J., 2013. Gene X environment interactions in schizophrenia and bipolar disorder: evidence from neuroimaging. Front. Psychiatry 4, 136. http://dx.doi.org/10.3389/fpsyt.2013.00136. Goodwin, F.K., Jamison, K.R., 2007. Manic-Depressive Illness. Bipolar Disorders and Recurrent Depression, second ed. Oxford University Press, New York. Hamshere, M.L., Gordon-Smith, K., Forty, L., Jones, L., Caesar, S., Fraser, C., Smith, D.J., 2009. Age-at-onset in bipolar-I disorder: mixture analysis of 1369 cases identifies three distinct clinical sub-groups. J. Affect. Disord. 116 (1–2), 23–29. http://dx.doi. org/10.1016/j.jad.2008.10.021, S0165-0327(08)00431-X. Hamshere, M.L., O’Donovan, M.C., Jones, I.R., Jones, L., Kirov, G., Green, E.K., Craddock, N., 2011. Polygenic dissection of the bipolar phenotype. Br. J. Psychiatry 198 (4), 284–288. http://dx.doi.org/10.1192/bjp.bp.110.087866. Ketter, T.A., 2010. Diagnostic features, prevalence, and impact of bipolar disorder. J. Clin. Psychiatry 71 (6), 14. http://dx.doi.org/10.4088/JCP.8125tx11c. Leboyer, M., Henry, C., Paillere-Martinot, M.L., Bellivier, F., 2005. Age at onset in bipolar affective disorders: a review. Bipolar Disord. 7 (2), 111–118. http://dx. doi.org/10.1111/j.1399-5618.2005.00181.x, BDI181. Lee, S.H., Ripke, S., Neale, B.M., Faraone, S.V., Purcell, S.M., Perlis, R.H., Wray, N.R., 2013. Genetic relationship between five psychiatric disorders estimated from genome-wide SNPs. Nat. Genet. 45 (9), 984–994. http://dx.doi.org/10.1038/ ng.2711. Li, Y., Willer, C.J., Ding, J., Scheet, P., Abecasis, G.R., 2010. MaCH: using sequence and genotype data to estimate haplotypes and unobserved genotypes. Genet. Epidemiol. 34 (8), 816–834. http://dx.doi.org/10.1002/gepi.20533. Manchia, M., Lampus, S., Chillotti, C., Sardu, C., Ardau, R., Severino, G., Del, Z.M., 2008. Age at onset in Sardinian bipolar I patients: evidence for three subgroups. Bipolar Disord. 10 (3), 443–446. Mazzarini, L., Colom, F., Pacchiarotti, I., Nivoli, A.M., Murru, A., Bonnin, C.M., Vieta, E., 2010. Psychotic versus non-psychotic bipolar II disorder. J. Affect. Disord. 126 (1–2), 55–60. http://dx.doi.org/10.1016/j.jad.2010.03.028, S0165-0327(10) 00321-6.
Nurnberger .Jr, J.I., Koller, D.L., Jung, J., et al., 2014. Identification of pathways for bipolar disorder: a meta-analysis. JAMA Psychiatry (6), 657–664. http://dx.doi. org/10.1001/jamapsychiatry.2014.176. O’Donovan, M., Craddock, N., Owen, M., 2009. Genetics of psychosis; insights from views across the genome. Hum. Genet. 126 (1), 3–12. http://dx.doi.org/10.1007/ s00439-009-0703-0. Perugi, G., Micheli, C., Akiskal, H.S., Madaro, D., Socci, C., Quilici, C., Musetti, L., 2001. Polarity of the first episode, clinical characteristics, and course of manic depressive illness: a systematic retrospective investigation of 320 bipolar I patients. Compr. Psychiatry 41 (1), 13–18. http://dx.doi.org/10.1016/S0010440X(00)90125-1. Peyrot, W.J., Milaneschi, Y., Abdellaoui, A., Sullivan, P.F., Hottenga, J.J., Boomsma, D.I., Penninx, B.W.J.H., 2014. Effect of polygenic risk scores on depression in childhood trauma. Br. J. Psychiatry 205 (2), 113–119. http://dx.doi.org/10.1192/bjp. bp.113.143081. PGC, 2011. Large-scale genome-wide association analysis of bipolar disorder identifies a new susceptibility locus near ODZ4. Nat. Genet. 43 (10), 977–983. http://dx.doi.org/10.1038/ng.943. PGC, 2013. Identification of risk loci with shared effects on five major psychiatric disorders: a genome-wide analysis. Lancet 381, 1371–1379. Phillips, M.L., Kupfer, D.J., 2013. Bipolar disorder diagnosis: challenges and future directions. Lancet 381 (9878), 1663–1671. http://dx.doi.org/10.1016/S01406736(13)60989-7. Purcell, S., Neale, B., Todd-Brown, K., Thomas, L., Ferreira, M.A.R., Bender, D., Sham, P.C., 2007. PLINK: a tool set for whole-genome association and population-based linkage analyses. Am. J. Hum. Genet. 81 (3), 559–575. http://dx.doi. org/10.1086/519795. 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 (7256), 748–752. http://dx.doi.org/10.1038/ nature08185, nature08185. Reilly, J.L., Sweeney, J.A., 2014. Generalized and specific neurocognitive deficits in psychotic disorders: utility for evaluating pharmacological treatment effects and as intermediate phenotypes for gene discovery. Schizophr. Bull. 40 (3), 516–522. http://dx.doi.org/10.1093/schbul/sbu013. Ringen, P.A., Lagerberg, T.V., Birkenaes, A.B., Engh, J., Faerden, A., Jonsdottir, H., Andreassen, O.A., 2008. Differences in prevalence and patterns of substance use in schizophrenia and bipolar disorder. Psychol. Med. 38 (9), 1241–1249. http: //dx.doi.org/10.1017/S003329170700236X, S003329170700236X. Ripke, S., O'Dushlaine, C., Chambert, K., Moran, J.L., Kahler, A.K., Akterin, S., Sullivan, P.F., 2013. Genome-wide association analysis identifies 13 new risk loci for schizophrenia. Nat. Genet. 45 (10), 1150–1159. http://dx.doi.org/10.1038/ ng.2742. Ruderfer, D.M., Fanous, A.H., Ripke, S., McQuillin, A., Amdur, R.L., Gejman, P.V., Kendler, K.S., 2013. Polygenic dissection of diagnosis and clinical dimensions of bipolar disorder and schizophrenia. Mol. Psychiatry (9), 1017–1024. http://dx. doi.org/10.1038/mp.2013.138. Sanchez-Moreno, J., Martinez-Aran, A., Tabares-Seisdedos, R., Torrent, C., Vieta, E., Ayuso-Mateos, J.L., 2009. Functioning and disability in bipolar disorder: an extensive review. Psychother. Psychosom. 78 (5), 285–297. http://dx.doi.org/ 10.1159/000228249 000228249. Simonsen, C., Sundet, K., Vaskinn, A., Birkenaes, A.B., Engh, J.A., Faerden, A., Andreassen, O.A., 2011. Neurocognitive dysfunction in bipolar and schizophrenia spectrum disorders depends on history of psychosis rather than diagnostic group. Schizophr. Bull. 37 (1), 78–83. http://dx.doi.org/10.1093/ schbul/sbp034, sbp034. Sklar, P., Ripke, S., Scott, L.J., Andreassen, O.A., Cichon, S., Craddock, N., Purcell, S.M., 2011. Large-scale genome-wide association analysis of bipolar disorder identifies a new susceptibility locus near ODZ4. Nat. Genet. 43, 977–983. http://dx. doi.org/10.1038/ng.943, ng.943. Tesli, M., Espeseth, T., Bettella, F., Mattingsdal, M., Aas, M., Melle, I., Andreassen, O.A., 2014. Polygenic risk score and the psychosis continuum model. Acta Psychiatr. Scand. (4), 311–317. Treuer, T., Tohen, M., 2010. Predicting the course and outcome of bipolar disorder: a review. Eur. Psychiatry 25 (6), 328–333. http://dx.doi.org/10.1016/j. eurpsy.2009.11.012, S0924-9338(10)00053-2. van Os, J., Tamminga, C., 2007. Deconstructing psychosis. Schizophr. Bull. 33 (4), 861–862. Walton, E., Geisler, D., Lee, P.H., Hass, J., Turner, J.A., Liu, J., Ehrlich, S., 2013. Prefrontal inefficiency is associated with polygenic risk for schizophrenia. Schizophr. Bull. (6), 1263–1271. http://dx.doi.org/10.1093/schbul/sbt174. Whalley, H.C., Papmeyer, M., Sprooten, E., Romaniuk, L., Blackwood, D.H., Glahn, D.C., McIntosh, A.M., 2012. The influence of polygenic risk for bipolar disorder on neural activation assessed using fMRI. Transl. Psychiatry 2, e130. http://dx.doi.org/ 10.1038/tp.2012.60.
Contents lists available at ScienceDirect
Journal of Affective Disorders journal homepage: www.elsevier.com/locate/jad
Research report
Polygenic risk scores in bipolar disorder subgroups Sofie Ragnhild Aminoff a,b,n, Martin Tesli b, Francesco Bettella b, Monica Aas b, Trine Vik Lagerberg b, Srdjan Djurovic b, Ole A. Andreassen b, Ingrid Melle b a b
Department of Specialized Inpatient Treatment, Division of Mental Health Services, Akershus University Hospital, Norway NORMENT, KG Jebsen Centre for Psychosis Research, Oslo University Hospital and Institute of Clinical Medicine, University of Oslo, Oslo, Norway
art ic l e i nf o
a b s t r a c t
Article history: Accepted 10 May 2015 Available online 22 May 2015
Background: Bipolar disorder (BD) is a genetically and clinically heterogeneous disorder. Current classifications of BD rely on clinical presentations without any validating biomarkers, making homogenous and valid subtypes warranted. This study aims at investigating whether a BD polygenic risk score (PGRS) can validate BD subtypes including diagnostic sub-categories (BD-I versus BD-II), patients with and without psychotic symptoms, polarity of first presenting episode and age at onset based groups. We also wanted to investigate whether illness severity indicators were associated with a higher polygenic risk for BD. Methods: Analyze differences in BD PGRS scores between suggested subtypes of BD and between healthy controls (CTR) and BD in a sample of N ¼669 (255 BD and 414 CTR). Results: The BD PGRS significantly discriminates between BD and CTR (po 0.001). There were no differences in BD PGRS between groups defined by diagnostic sub-categories, presenting polarity and age at onset. Patients with psychotic BD had nominally significantly higher BD PGRS than patients with nonpsychotic BD after controlling for diagnostic sub-category (p¼ 0.019). These findings remained trend level significant after Bonferroni corrections (p¼ 0.079). Limitations: The low explained variance of the current PGRS method could lead to type II errors. Conclusions: There are nominally significant differences in BD PGRS scores between patients with and without psychotic symptoms, indicating that these two forms of BD might represent distinct subtypes of BD based in its polygenic architecture and a division between BD with and without psychotic symptoms could represent a more valid subclassification of BD than current diagnostic sub-categories. If replicated, this finding could affect future research, diagnostics and clinical practice. & 2015 Elsevier B.V. All rights reserved.
Keywords: Bipolar disorder Polygenic risk score Psychosis Depression Mania Hypomania
1. Introduction Bipolar disorder (BD) has a complex phenotype characterized by manic or hypomanic episodes with the addition of depressive episodes in 80–95% of cases (Baek et al., 2014), psychotic features in 20–60% of cases (Goodwin and Jamison, 2007; Mazzarini et al., 2010) and with significant individual differences in illness course and severity (Ketter, 2010; Sanchez-Moreno et al., 2009; Treuer and Tohen, 2010). One of the aims of BD research has thus been the identification of valid subgroups to aid treatment planning, course prediction and the search for the genetic underpinnings of BD (Aminoff et al., 2013; Craddock et al., 2005; Craddock and Owen, 2010). Current classifications of BD subtypes however rely on clinical presentations without any validating biomarkers n Corresponding author at: NORMENT, KG Jebsen Centre for Psychosis Research – TOP Study, Building 49, Oslo University Hospital, Ullevål, Kirkeveien 166, PO Box 4956 Nydalen, 0424 Oslo, Norway. Tel.: þ47 23 02 73 50; fax: þ 47 23 02 73 33. E-mail address: [email protected] (S.R. Aminoff).
http://dx.doi.org/10.1016/j.jad.2015.05.021 0165-0327/& 2015 Elsevier B.V. All rights reserved.
(O'Donovan et al., 2009; Phillips and Kupfer, 2013). This includes the DSM sub-categorization into Bipolar I (BDI) versus Bipolar II (BDII) disorders based on the presence of mania (BDI) versus depression and hypomania (BDII). Other subtypes have also been suggested; based on the presence of psychotic symptoms (Simonsen et al., 2011), differences in age at onset (Azorin et al., 2013) or the polarity of the first (presenting) mood episode (Perugi et al., 2001). BD is a multifactorial disorder with a strong genetic component (Geoffroy et al., 2013). Genome-wide association studies (GWAS) indicate the involvement of several common single nucleotide variants of small effects (Craddock and Sklar, 2013; PGC, 2011; Purcell et al., 2009), and have identified genes involved in neuronal growth and cell adhesion, calcium signaling, hormone regulation, second messenger and glutamate systems (Nurnberger et al., 2014; Sklar et al., 2011) with considerable genetic overlap between BD and schizophrenia (SCZ) and unipolar depression (Andreassen et al., 2013; Cardno and Owen, 2014; Craddock et al., 2009, 2005; Lee et al., 2013; Nurnberger et al., 2014; PGC, 2013;
S.R. Aminoff et al. / Journal of Affective Disorders 183 (2015) 310–314
van Os and Tamminga, 2007). Recent approaches take into consideration the multiplicity of genes involved in psychiatric disorders by providing information on the cumulative genomic risk (Lee et al., 2013; Purcell et al., 2009; Ripke et al., 2013) accounting for a larger proportion of the phenotypic variance (Walton et al., 2013). By using summary statistics representing differences in genetic variants between cases and controls in a discovery study, one can assign a polygenic risk score (PGRS) to individuals in the independent study sample. The PGRS method has been shown to predict case-control status at high significance levels for both BD and schizophrenia (SCZ) (Purcell et al., 2009; Ripke et al., 2013). At clinical level, associations have been found between BD PGRS and mania in patients with SCZ (Ruderfer et al., 2013). At proposed intermediate phenotype level, BD PGRS has been shown to correlate with increased brain activation in limbic regions during a language paradigm in CTR and individuals with a family load of BD (Whalley et al., 2012). The effect of a major depressive disorder PGRS was shown to be increased in cases with additional childhood trauma (Peyrot et al., 2014). Recently, the PGRS method also provided support for the psychosis continuum model in a sample with BD spectrum disorders, SCZ spectrum disorders and healthy controls (Tesli et al., 2014). To our knowledge, the PGRS method has not previously been used to validate subtypes of BD within a BD sample. The aim of the current study is to investigate whether BD PGRS can validate previously suggested BD subtypes (i.e. BDI versus BDII, psychotic- versus non-psychotic BD, subtypes based on age-atonset groups or the polarity of the presenting episode). We also explored whether the BD PGRS was associated with indicators of illness severity in BD.
2. Methods 2.1. Participants All participants were consecutively recruited to the on-going Thematically Organized Psychosis study (TOP). Patients were recruited from outpatient and inpatient units of the three major hospitals in Oslo, Norway, between 2003 and 2011. Inclusion criteria for this particular study were age between 17 and 65 years and having a DSM-IV diagnosis of BD type I or II (total N ¼ 255; BDI n ¼181 and BDII n ¼ 74). 134 patients (74%) from the BDI sample and 14 patients (19%) from the BDII sample had a history of psychosis. 90 (35%) patients had a first, or presenting, episode characterized by elevated mood (mania, mixed or hypomania) and 153 (60%) had a depressive presenting episode (information uncertain or missing in n ¼12 (4%)). 105 (41%) patients were employed and 88 (34.5%) married or cohabitant. Mean duration of treatment was 6 79 years. Family history was based on a semi-structured interview asking patients about the presence of a probable or sure diagnosis of BD or schizophrenia in first-degree relatives (parents, siblings or children). 51 patients (21%) had first degree relatives with diagnosis of BD and 10 patients (4%) had first degree relatives with a diagnosis of SCZ. For other demographic details see Table 1. The CTR sample (N ¼414) was randomly drawn from the statistical population registers of the same areas of Oslo as the patients and was contacted by mail. The entire sample consisted of Northern European Caucasians, mainly Norwegians, previously demonstrated to be genetically homogenous (Athanasiu et al., 2010; Djurovic et al., 2010). General exclusion criteria were hospitalized head injury, neurological disorder, unstable or uncontrolled medical condition that interferes with brain function and/or an IQ below 70. The CTR sample was additionally screened for physical and mental disorders, a history of head injury, ongoing drug abuse, and a family history of severe mental disorders. The Regional
311
Table 1 Demographics. Healthy controls Bipolar disorder Sample size, N Gender, N (%) Male Female Age (mean 7SD) Age at onset (mean 7SD) Duration of illness (mean7 SD) Bipolar disorder in first degree relatives, N (%)
414
255
208 (50.2) 206 (49.8) 357 10
101 (39.6) 154 (60.4) 367 12 237 9 13 710 51 (20)
Missing data in N patients ¼7 (3%).
Committee for Medical Research Ethics and the Norwegian Data Inspectorate approved the study, and participants' written, informed consents according to the Declaration of Helsinki were obtained. 2.2. Clinical assessment Patients were clinically characterized based on a personal interview by trained assessment staff that had completed the TOP Study's assessment training and reliability program, either medical doctors or clinical psychologists. Diagnosis was based on the Structured Clinical Interview for DSM-IV Axis I disorders (SCID-I) (First et al., 1995) and information from medical charts. A good inter-rater reliability for diagnoses was achieved with an overall kappa score of 0.77 (95% confidence interval ¼0.60–0.94) (Ringen et al., 2008). A history of psychosis was defined as having one or more lifetime psychotic episodes. We defined polarity of presenting episode as the polarity of the first SCID verified mood episode, divided into depressive or elevated onset (including mixed episodes). Age at onset was defined as the age of the first SCID-verified mood episode. Age was collapsed into three groups based on results from previous admixture analysis in large samples finding relatively stable age-at-onset groups in different cultures and birth cohorts: (1) early onset (first episode r22 years), (2) intermediate onset (first episode 23–34) and (3) late onset (first episode Z35 years) (Bellivier et al., 2001, 2003; Hamshere et al., 2009; Leboyer et al., 2005; Manchia et al., 2008). 2.3. Genotyping All participants were genotyped at Expression Analysis Inc. (Durham, NC, USA) using the Affymetrix Genome-Wide Human SNP Array 6.0 (Affymetrix Inc., Santa Clara, CA, USA). Quality control was performed using PLINK (version 1.07; http://pngu. mgh.harvard.edu/purcell/plink/) (Purcell et al., 2007). As a quality control, exclusions of individuals based on genotyping were made of (I) one of two duplicates, (II) one of two relatives (identity by descent (IBD) 40.1875), (III) individuals with a recorded gender differing from that determined by X chromosome marker homozygosity, (IV) mixup-samples (calculated by pairwise genomewide identity by state (IBS)), (V) individuals with non-European ancestry (calculated with HapMap3 and PLINK's first two MDS components) and (VI) individuals with more than 5% missing genotype data. SNPs were excluded based on (I) severe deviation from Hardy–Weinberg equilibrium, (II) minor allele frequency below 1% and (III) low yield ( o95% in controls), as described in earlier reports (Athanasiu et al., 2010; Djurovic et al., 2010). PLINK's multidimensional scaling (MDS) components were used as ancestry covariates in the analyses. The ancestry-based exclusion affected outliers in the space generated by the first two MDS components. It is plausible that once the outliers are removed, the
312
S.R. Aminoff et al. / Journal of Affective Disorders 183 (2015) 310–314
third MDS component acquires a higher discriminating power. We therefore included the first three MDS components to account for ancestry in the subsequent analyses.
episodes adjusted for duration of illness was investigated with multivariate hierarchical regression analysis.
2.4. Imputation of SNPs
3. Results
Following the above mentioned quality control, SNPs were imputed with MaCH (Li et al., 2010) (http://www.sph.umich.edu/ csg/abecasis/MACH/download/1000G-PhaseI-Interim.html) using the European samples in the Phase I release of the 1000 Genomes project. SNPs not present in the 1000 Genomes reference, and SNPs with ambiguous strand alignments (A/T and G/C SNPs), were removed from the sample data sets. Imputation was a three stage process, involving (I) ChunkChromosome where the data set was broken into 2500 SNP pieces with 500 SNP overlap (http://gen ome.sph.umich.edu/wiki/ChunkChromosome), (II) MaCH where each piece was phased (40 rounds and 400 states) (http://www. sph.umich.edu/csg/abecasis/MaCH/download/), and (III) Minimac where each phased piece was imputed to the 1000 Genomes European reference panel (20 rounds and 400 states) (http://gen ome.sph.umich.edu/wiki/Minimac). In the third stage, all imputed SNPs were provided with an estimated r2 score as quality metric. Exclusions were made of SNPs with MAF o0.01 or an r2 score o0.2, leaving 8,331,417 SNPs.
Taken ancestry into account the BD PGRS for the BD sample was 0.153 (70.63) (standard error), significantly higher than for the CTR group (mean BD PGRS score 0.179 70.50, po 0.001). In the BD sample, there were no statistically significant differences between males and females, F(1,250) ¼0.188, p ¼0.665), or between patients with or without first degree relatives with BD, F (1,243)¼0.281, p ¼ 0.596. Patients with a history of psychosis had nominally higher BD PGRS scores than those without when controlling for ancestry (p ¼0.049) (Table 2). When also controlling for diagnostic subcategory the findings were slightly strengthened (p ¼0.019). However these results remained only trend level significant after Bonferroni-corrections for four group comparisons (p ¼0.079). There were no significant differences in BD PGRS between the diagnostic subcategories (BDI versus BDII), between those with a depressive presenting polarity versus those with an elevated presenting polarity or across different age at onset groups (Table 2). In the follow up analysis, there was also no significant interaction effect between history of psychosis and diagnostic subcategory, F(1,251) ¼0.490, p ¼0.485. The BD PGRS was not associated with any severity indicators, including number of mood episodes and number of mood episodes adjusted for duration of illness (depressive episodes, F(4,233) ¼1.060, p ¼0.377; elevated episodes, F(4,234) ¼1.069, p ¼0.372; total episodes, F(4,238) ¼ 1.121, p¼ 0.347).
2.5. Polygenic risk score BD PGRS were computed based on imputed SNPs following the method developed by Purcell et al. (2009). Using PLINK version 1.07 (http://pngu.mgh.harvard.edu/purcell/plink/) (Purcell et al., 2007), we performed a meta-analysis including all Psychiatric Genomics Consortium (PGC) substudies (PGC, 2011) except the one based on the sample used for the current study (TOP3 in reference publication) (N ¼ 7278 BD cases and 8901 controls) and obtained risk allele effect sizes (ln(OR)) for all imputed SNPs. In order to avoid any inflation due to LD patterns across the genome, the SNPs were subsequently pruned using PLINK's-clump option (r2 o0.25, 500 kb windows) to select representatives with lowest p-Values from all LD blocks (209088 SNPs). Using the clumped SNP set, 10 PGRS were computed including SNPs at different p-Value thresholds: p ¼1, p ¼0.5, p ¼0.4, p¼ 0.3, p ¼0.2, p ¼0.1, p ¼0.05, p ¼0.01, p ¼0.001, and p ¼0.0001. For the subsequent analyses, we selected the PGRS resulting from the p-Value threshold of 0.05 (23062 SNPs), since this was explaining most phenotypic variance (Nagelkerke pseudo-r2). For more details see Tesli et al. (2014). The selected BD PGRS was transformed into standard scores before proceeding with the subsequent analyses. 2.6. Statistical analyses All analyses from this point onward were performed using the IBM SPSS, version 20. The BD PGRS was normally distributed across all diagnostic subgroups (Kolmogorov Smirnov¼0.200). For analyses of PGRS group differences we used ANCOVAs with three MDS components to control for ancestry. Bivariate associations were measured through Pearson's correlations (r). All tests were two sided with a p o0.05. We made Bonferroni corrections based on conducting four separate group comparisons (diagnostic subcategories, presenting polarity, age at onset groups and psychotic BD). Since there is a considerable overlap between having a BDI diagnosis and the risk for psychotic symptoms, we included a follow-up, two-way ANOVA to investigate interaction effects on BD PGRS between sub-diagnosis and history of psychosis despite minor differences in the bivariate analyses. The relationship between BD PGRS and number of total, depressive and elevated
4. Discussion The main finding of the study was that there were no statistically significant differences in the BD PGRS for groups defined by diagnostic sub-categories (BD-I versus BD-II), presenting polarity or age at onset subgroups. These findings do not support the notion that these are genetically different subgroups of BD, at least as based in the PGRS. There was however a nominally statistically Table 2 Between group differences in bipolar polygenic risk score in a bipolar sample. Classification
Total sample: N ¼ 255 Diagnostic subcategory: BDI BDII History of psychosis: Yes No Polarity of first presenting episode: Depressive Elevated Missing data Age at onset: Early Mid Late Missing data
N (%)
BD PGRS mean 7S. D.
ANCOVA with 3 MDS as covariates
0.1557 1.00 F(1,250)¼ 0.012, p ¼0.913 181 (71) 74 (29)
0.167 0.99 0.157 1.05 F(1,250) ¼ 3.915, p¼ 0.049
148 (58) 107 (42)
0.277 0.96 0.004 7 1.04 F(1,249)¼ 0.566, p ¼0.568
152 (60) 91 (36) 12 (4)
0.147 0.99 0.20 7 1.05 F(2,247) ¼ 0.469, p ¼0.628
150 71 32 2
(59) (28) (12) (1)
0.107 1.02 0.277 1.00 0.177 0.95
S.R. Aminoff et al. / Journal of Affective Disorders 183 (2015) 310–314
313
Fig. 1. Polygenic risk scores in bipolar disorder subtypes (mean z scores and standard errors). Abbreviations: CTR ¼ Healthy controls, BDI ¼ Bipolar type I disorder and BD2 ¼Bipolar type 2 disorder.
significant difference in BD PGRS between patients with BD with and without psychotic features. In a subsample of the current, we have previously shown that BD diagnostic sub-category, presenting polarity and history of psychosis, was associated with underlying neurocognitive differences, suggesting that they might represent distinct BD subtypes (Aminoff et al., 2013). The link to neurocognitive dysfunctions is in line with studies indicating that BD and SCZ share genetic basis with cognitive factors (Fernandes et al., 2013; Reilly and Sweeney, 2014). If replicated, the finding of different BD PGRS between patients with and without a history of psychosis would suggest that the distinction between psychotic BD and non-psychotic BD could be the more valid categorization, reflecting underlying genetic differences to a greater extent than the BDI–BDII distinction (Fig. 1). This might seem at odds with one study comprising a broad spectrum of psychotic disorders (BD- and SCZ spectrums plus psychosis not otherwise specified) that indicated a relationship between lifetime psychosis and SCZ PGRS but not BD PGRS (Tesli et al., 2014). However, the main part of the patients with lifetime psychosis in the previous study was part of the SCZ- and not the BD spectrum. Another study did not find significant differences between BD individuals with and without psychotic features using a SCZ PGRS (Hamshere et al., 2011). A potential limitation of the current study is related to the construction of the PGRS. The BD PGRS is not an absolute measure of the genetic risk for BD but a relative measure of congruence with the complex genetic structure of cases and controls in the initial discovery sample and thus dependent of its sample characteristics. The present results, with higher BD PGRS in psychotic BD could be taken to imply that this discovery sample has a predominance of patients with psychotic BD. Since more than half of patients with BD experience psychotic symptoms (Goodwin and Jamison, 2007) this appears as a probable explanation. An important limitation is the explained variance of the current PGRS method, which is low (2.4%). This may lead to type II errors, which suggest the current findings cannot rule out that differences in polygenic risk related to other BD subgroups and clinical characteristics could be detected investigating larger samples.
data, in the writing of the manuscript or in the decision to submit the article for publication. Conflict of interest All authors declare no conflict of interest.
Contributors Sofie R. Aminoff has written the manuscript and contributed to the study design, data collection, data analyses and the discussion of findings. Martin Tesli has written parts of the manuscript and contributed to study design, data collection, data analyses and discussion of findings. Francesco Bettella has contributed to study design, data analysis, manuscript and discussion of findings. Monica Aas has contributed to the manuscript and discussion of findings. Trine Vik Lagerberg has contributed to the study design, data collection, data analyses, manuscript and the discussion of findings. Srdjan Djurovic has contributed to the manuscript and discussion of findings. Ole A. Andreassen has contributed to the manuscript and discussion of findings. Ingrid Melle has contributed to the study design, the data analyses, the manuscript and the discussion of findings. All authors have approved the final article.
Acknowledgments We would particularly like to thank the study participants for their contribution and the clinicians and research staff who contributed to patient recruitment and management of data. This work was supported by Grants from the Norwegian Research Council (#421716 and #223273), the Regional Health Authority SouthEastern Norway (#N1, #2011085 and #2013123) and from the K.G. Jebsen Foundation.
References Role of the funding source This work was supported by Grants from the Norwegian Research Council (#421716 and #223273), the Regional Health Authority South-Eastern Norway (#N1, #2011085 and #2013123) and from the K.G. Jebsen Foundation. Neither had any involvement in study design, collection of data, analysis and interpretation of
Aminoff, S.R., Hellvin, T., Lagerberg, T.V., Berg, A.O., Andreassen, O.A., Melle, I., 2013. Neurocognitive features in subgroups of bipolar disorder. Bipolar Disord. 15 (3), 272–283. http://dx.doi.org/10.1111/bdi.12061. Andreassen, O.A., Thompson, W.K., Schork, A.J., Ripke, S., Mattingsdal, M., Kelsoe, J.R., 2013. Bipolar Disorder and Schizophrenia Working Groups, 2013. Improved detection of common variants associated with schizophrenia and bipolar disorder using pleiotropy-informed conditional false discovery rate. PLoS Genet. 9 (4), e1003455. http://dx.doi.org/10.1371/journal.pgen.1003455.
314
S.R. Aminoff et al. / Journal of Affective Disorders 183 (2015) 310–314
Athanasiu, L., Mattingsdal, M., Kähler, A.K., Brown, A., Gustafsson, O., Agartz, I., Andreassen, O.A., 2010. Gene variants associated with schizophrenia in a Norwegian genome-wide study are replicated in a large European cohort. J. Psychiatry Res. 44 (12), 748–753. http://dx.doi.org/10.1016/j. jpsychires.2010.02.002. Azorin, J.M., Bellivier, F., Kaladjian, A., Adida, M., Belzeaux, R., Fakra, E., Golmard, J.L., 2013. Characteristics and profiles of bipolar I patients according to age-atonset: findings from an admixture analysis. J. Affect. Disord. 150 (3), 993–1000. http://dx.doi.org/10.1016/j.jad.2013.05.026. Baek, J.H., Eisner, L.R., Nierenberg, A.A., 2014. Epidemiology and course of unipolar mania: results from the national epidemiologic survey on alcohol and related conditions (NESARC). Depress. Anxiety 0, 1–10. http://dx.doi.org/10.1002/ da.22261. Bellivier, F., Golmard, J.L., Henry, C., Leboyer, M., Schurhoff, F., 2001. Admixture analysis of age at onset in bipolar I affective disorder. Arch. Gen. Psychiatry 58 (5), 510–512. Bellivier, F., Golmard, J.L., Rietschel, M., Schulze, T.G., Malafosse, A., Preisig, M., Leboyer, M., 2003. Age at onset in bipolar I affective disorder: further evidence for three subgroups. Am. J. Psychiatry 160 (5), 999–1001. Cardno, A.G., Owen, M.J., 2014. Genetic relationships between schizophrenia, bipolar disorder, and schizoaffective disorder. Schizophr. Bull. 40 (3), 504–515. http://dx.doi.org/10.1095043/schbul/sbu016. Craddock, N., O’Donovan, M.C., Owen, M.J., 2009. Psychosis genetics: modeling the relationship between schizophrenia, bipolar disorder, and mixed (or “schizoaffective”) psychoses. Schizophr. Bull. 35 (3), 482–490. http://dx.doi.org/ 10.1093/schbul/sbp020, sbp020. Craddock, N., O’Donovan, M.C., Owen, M.J., 2005. The genetics of schizophrenia and bipolar disorder: dissecting psychosis. J. Med. Genet. 42 (3), 193–204. http://dx. doi.org/10.1136/jmg.2005.030718. Craddock, N., Owen, M.J., 2010. The Kraepelinian dichotomy – going, going… but still not gone. Br. J. Psychiatry 196 (2), 92–95. http://dx.doi.org/10.1192/bjp. bp.109.073429 196/2/92. Craddock, N., Sklar, P., 2013. Genetics of bipolar disorder. Lancet 381 (9878), 1654–1662. http://dx.doi.org/10.1016/S0140-6736(13)60855-7. Djurovic, S., Gustafsson, O., Mattingsdal, M., Athanasiu, L., Bjella, T., Tesli, M., Andreassen, O.A., 2010. A genome-wide association study of bipolar disorder in Norwegian individuals, followed by replication in Icelandic sample. J. Affect. Disord. 126 (1–2), 312–316. http://dx.doi.org/10.1016/j.jad.2010.04.007. Fernandes, C.P., Christoforou, A., Giddaluru, S., Ersland, K.M., Djurovic, S., Mattheisen, M., Le Hellard, S., 2013. A genetic deconstruction of neurocognitive traits in schizophrenia and bipolar disorder. PLoS One 8 (12), e81052. http://dx.doi.org/10.1371/ journal.pone.0081052. First, M., Spitzer, R., Gibbon, M., Williams, J.B.W., 1995. Structured Clinical Interview for DSM-IV Axis I Disorders: Patient Edition (SCIP-P), Version 2. New York State Psychiatric Institute, Biometrics Research, New York. Geoffroy, P.A., Etain, B., Houenou, J., 2013. Gene X environment interactions in schizophrenia and bipolar disorder: evidence from neuroimaging. Front. Psychiatry 4, 136. http://dx.doi.org/10.3389/fpsyt.2013.00136. Goodwin, F.K., Jamison, K.R., 2007. Manic-Depressive Illness. Bipolar Disorders and Recurrent Depression, second ed. Oxford University Press, New York. Hamshere, M.L., Gordon-Smith, K., Forty, L., Jones, L., Caesar, S., Fraser, C., Smith, D.J., 2009. Age-at-onset in bipolar-I disorder: mixture analysis of 1369 cases identifies three distinct clinical sub-groups. J. Affect. Disord. 116 (1–2), 23–29. http://dx.doi. org/10.1016/j.jad.2008.10.021, S0165-0327(08)00431-X. Hamshere, M.L., O’Donovan, M.C., Jones, I.R., Jones, L., Kirov, G., Green, E.K., Craddock, N., 2011. Polygenic dissection of the bipolar phenotype. Br. J. Psychiatry 198 (4), 284–288. http://dx.doi.org/10.1192/bjp.bp.110.087866. Ketter, T.A., 2010. Diagnostic features, prevalence, and impact of bipolar disorder. J. Clin. Psychiatry 71 (6), 14. http://dx.doi.org/10.4088/JCP.8125tx11c. Leboyer, M., Henry, C., Paillere-Martinot, M.L., Bellivier, F., 2005. Age at onset in bipolar affective disorders: a review. Bipolar Disord. 7 (2), 111–118. http://dx. doi.org/10.1111/j.1399-5618.2005.00181.x, BDI181. Lee, S.H., Ripke, S., Neale, B.M., Faraone, S.V., Purcell, S.M., Perlis, R.H., Wray, N.R., 2013. Genetic relationship between five psychiatric disorders estimated from genome-wide SNPs. Nat. Genet. 45 (9), 984–994. http://dx.doi.org/10.1038/ ng.2711. Li, Y., Willer, C.J., Ding, J., Scheet, P., Abecasis, G.R., 2010. MaCH: using sequence and genotype data to estimate haplotypes and unobserved genotypes. Genet. Epidemiol. 34 (8), 816–834. http://dx.doi.org/10.1002/gepi.20533. Manchia, M., Lampus, S., Chillotti, C., Sardu, C., Ardau, R., Severino, G., Del, Z.M., 2008. Age at onset in Sardinian bipolar I patients: evidence for three subgroups. Bipolar Disord. 10 (3), 443–446. Mazzarini, L., Colom, F., Pacchiarotti, I., Nivoli, A.M., Murru, A., Bonnin, C.M., Vieta, E., 2010. Psychotic versus non-psychotic bipolar II disorder. J. Affect. Disord. 126 (1–2), 55–60. http://dx.doi.org/10.1016/j.jad.2010.03.028, S0165-0327(10) 00321-6.
Nurnberger .Jr, J.I., Koller, D.L., Jung, J., et al., 2014. Identification of pathways for bipolar disorder: a meta-analysis. JAMA Psychiatry (6), 657–664. http://dx.doi. org/10.1001/jamapsychiatry.2014.176. O’Donovan, M., Craddock, N., Owen, M., 2009. Genetics of psychosis; insights from views across the genome. Hum. Genet. 126 (1), 3–12. http://dx.doi.org/10.1007/ s00439-009-0703-0. Perugi, G., Micheli, C., Akiskal, H.S., Madaro, D., Socci, C., Quilici, C., Musetti, L., 2001. Polarity of the first episode, clinical characteristics, and course of manic depressive illness: a systematic retrospective investigation of 320 bipolar I patients. Compr. Psychiatry 41 (1), 13–18. http://dx.doi.org/10.1016/S0010440X(00)90125-1. Peyrot, W.J., Milaneschi, Y., Abdellaoui, A., Sullivan, P.F., Hottenga, J.J., Boomsma, D.I., Penninx, B.W.J.H., 2014. Effect of polygenic risk scores on depression in childhood trauma. Br. J. Psychiatry 205 (2), 113–119. http://dx.doi.org/10.1192/bjp. bp.113.143081. PGC, 2011. Large-scale genome-wide association analysis of bipolar disorder identifies a new susceptibility locus near ODZ4. Nat. Genet. 43 (10), 977–983. http://dx.doi.org/10.1038/ng.943. PGC, 2013. Identification of risk loci with shared effects on five major psychiatric disorders: a genome-wide analysis. Lancet 381, 1371–1379. Phillips, M.L., Kupfer, D.J., 2013. Bipolar disorder diagnosis: challenges and future directions. Lancet 381 (9878), 1663–1671. http://dx.doi.org/10.1016/S01406736(13)60989-7. Purcell, S., Neale, B., Todd-Brown, K., Thomas, L., Ferreira, M.A.R., Bender, D., Sham, P.C., 2007. PLINK: a tool set for whole-genome association and population-based linkage analyses. Am. J. Hum. Genet. 81 (3), 559–575. http://dx.doi. org/10.1086/519795. 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 (7256), 748–752. http://dx.doi.org/10.1038/ nature08185, nature08185. Reilly, J.L., Sweeney, J.A., 2014. Generalized and specific neurocognitive deficits in psychotic disorders: utility for evaluating pharmacological treatment effects and as intermediate phenotypes for gene discovery. Schizophr. Bull. 40 (3), 516–522. http://dx.doi.org/10.1093/schbul/sbu013. Ringen, P.A., Lagerberg, T.V., Birkenaes, A.B., Engh, J., Faerden, A., Jonsdottir, H., Andreassen, O.A., 2008. Differences in prevalence and patterns of substance use in schizophrenia and bipolar disorder. Psychol. Med. 38 (9), 1241–1249. http: //dx.doi.org/10.1017/S003329170700236X, S003329170700236X. Ripke, S., O'Dushlaine, C., Chambert, K., Moran, J.L., Kahler, A.K., Akterin, S., Sullivan, P.F., 2013. Genome-wide association analysis identifies 13 new risk loci for schizophrenia. Nat. Genet. 45 (10), 1150–1159. http://dx.doi.org/10.1038/ ng.2742. Ruderfer, D.M., Fanous, A.H., Ripke, S., McQuillin, A., Amdur, R.L., Gejman, P.V., Kendler, K.S., 2013. Polygenic dissection of diagnosis and clinical dimensions of bipolar disorder and schizophrenia. Mol. Psychiatry (9), 1017–1024. http://dx. doi.org/10.1038/mp.2013.138. Sanchez-Moreno, J., Martinez-Aran, A., Tabares-Seisdedos, R., Torrent, C., Vieta, E., Ayuso-Mateos, J.L., 2009. Functioning and disability in bipolar disorder: an extensive review. Psychother. Psychosom. 78 (5), 285–297. http://dx.doi.org/ 10.1159/000228249 000228249. Simonsen, C., Sundet, K., Vaskinn, A., Birkenaes, A.B., Engh, J.A., Faerden, A., Andreassen, O.A., 2011. Neurocognitive dysfunction in bipolar and schizophrenia spectrum disorders depends on history of psychosis rather than diagnostic group. Schizophr. Bull. 37 (1), 78–83. http://dx.doi.org/10.1093/ schbul/sbp034, sbp034. Sklar, P., Ripke, S., Scott, L.J., Andreassen, O.A., Cichon, S., Craddock, N., Purcell, S.M., 2011. Large-scale genome-wide association analysis of bipolar disorder identifies a new susceptibility locus near ODZ4. Nat. Genet. 43, 977–983. http://dx. doi.org/10.1038/ng.943, ng.943. Tesli, M., Espeseth, T., Bettella, F., Mattingsdal, M., Aas, M., Melle, I., Andreassen, O.A., 2014. Polygenic risk score and the psychosis continuum model. Acta Psychiatr. Scand. (4), 311–317. Treuer, T., Tohen, M., 2010. Predicting the course and outcome of bipolar disorder: a review. Eur. Psychiatry 25 (6), 328–333. http://dx.doi.org/10.1016/j. eurpsy.2009.11.012, S0924-9338(10)00053-2. van Os, J., Tamminga, C., 2007. Deconstructing psychosis. Schizophr. Bull. 33 (4), 861–862. Walton, E., Geisler, D., Lee, P.H., Hass, J., Turner, J.A., Liu, J., Ehrlich, S., 2013. Prefrontal inefficiency is associated with polygenic risk for schizophrenia. Schizophr. Bull. (6), 1263–1271. http://dx.doi.org/10.1093/schbul/sbt174. Whalley, H.C., Papmeyer, M., Sprooten, E., Romaniuk, L., Blackwood, D.H., Glahn, D.C., McIntosh, A.M., 2012. The influence of polygenic risk for bipolar disorder on neural activation assessed using fMRI. Transl. Psychiatry 2, e130. http://dx.doi.org/ 10.1038/tp.2012.60.
