Abstract
Bipolar disorder (BD) features heterogenous clinical presentation and course of illness. It remains unclear how subphenotypes associate with genetic loadings of BD and related psychiatric disorders. We investigated associations between the subphenotypes and polygenic risk scores (PRS) for BD, schizophrenia, and major depressive disorder (MDD) in two BD cohorts from Sweden (N = 5180) and the UK (N = 2577). Participants were assessed through interviews and medical records for inter-episode remission, psychotic features during mood episodes, global assessment of functioning (GAF, function and symptom burden dimensions), and comorbid anxiety disorders. Meta-analyses based on both cohorts showed that inter-episode remission and GAF-function were positively correlated with BD-PRS but negatively correlated with schizophrenia-PRS (SCZ-PRS) and MDD-PRS. Moreover, BD-PRS was negatively, and MDD-PRS positively, associated with the risk of comorbid anxiety disorders. Finally, SCZ-PRS was positively associated with psychotic symptoms during mood episodes. Assuming a higher PRS of certain psychiatric disorders in cases with a positive family history, we further tested the associations between subphenotypes in index BD people and occurrence of BD, schizophrenia, or MDD in their relatives using Swedish national registries. BD patients with a relative diagnosed with BD had: (1) higher GAF and lower risk of comorbid anxiety than those with a relative diagnosed with schizophrenia or MDD, (2) lower risk of psychotic symptoms than those with a relative diagnosed with schizophrenia. Our findings shed light on the genetic underpinnings of the heterogeneity in clinical manifestations and course of illness in BD, which ultimately provide insights for developing personalized approaches to the diagnosis and treatment.
Similar content being viewed by others
Introduction
Ever since Kraepelin distinguished manic-depressive illness from dementia praecox, the hallmark of bipolar disorder (BD) has been disruptive episodes of mania and depression between which patients recover and regain function [1]. Persons with BD may experience psychotic symptoms during mood episodes but not during euthymic periods [2]. This course of illness contrasts the prototypical form of schizophrenia—which Kraepelin called dementia praecox—that follows a chronic, deteriorating course. Although psychotic symptoms may wax and wane in schizophrenia, negative symptoms persist and full recovery is rare. More recently, the distinguishing feature of hypomanic or manic episodes was used to separate recurrent unipolar depression from BD [3]. Comorbid anxiety disorder is more common in unipolar depression that in BD [4].
Although these characteristics distinguishing between BD, schizophrenia, and recurrent unipolar depression remain in modern diagnostic classification systems, the actual clinical presentation and natural course of BD varies considerably among individuals [5]. While many patients with BD do indeed regain full function between mood episodes (complete inter-episode remission), a significant portion suffer from residual mood symptoms or lingering cognitive impairment that prevent full functional recovery [6, 7]. Likewise, some but not all BD patients feature psychotic symptoms during mood episodes, or suffer from comorbid anxiety disorders.
Genome-wide association studies (GWAS) and post-GWAS analyses have found common risk variants shared between BD, schizophrenia, and major depressive disorder (MDD) that help explain the overlap in symptom presentations [8,9,10,11]. It has been less studied to what extent genetic factors differentially influence features within a diagnostic category. In recent years, polygenic risk scores (PRS) have been widely applied to study the association between complex genetic traits and clinical symptoms. Studies of BD have reported that the polygenic loading of schizophrenia, as measured by PRS, is negatively associated with lithium response and age of onset, but positively associated with psychotic features [12,13,14,15]. Moreover, BD patients with a higher genetic liability of MDD respond worse to lithium [16]. Finally, a recent study found that the MDD-PRS was strongly associated with the depression dimension, whereas the schizophrenia PRS (SCZ-PRS) was strongly associated with the psychosis dimension among BD patients [17].
To reconcile the categorical distinctions at the core of the current psychiatric diagnostic systems—such as the Kreapelinian dichotomy and the differentiation between BD and unipolar depression—with the evolving comprehension of these disorders being dimensional and having polygenic underpinnings, we investigated how polygenic liabilities of BD, schizophrenia, and MDD are associated with the following key subphenotypes of BD: inter-episode remission, global functioning and symptom burden, psychotic features during mood episodes, and comorbid anxiety disorders. We performed a BD case-only study and tested the associations between the three psychiatric disorders’ PRS and these BD subphenotypes in a Swedish cohort and a UK cohort, with subsequent meta-analyses. Assuming a higher PRS of certain psychiatric disorders in cases with a positive family history, we also examined the features of subphenotypes in a familial coaggregation design using Swedish national registries, where BD index persons who had a relative with BD were compared with BD index persons who had a relative with schizophrenia or MDD.
Methods
Subjects
BD cases were enrolled in the Swedish Bipolar Collection study (SWEBIC), to which most participants had been recruited via the Swedish National Quality Register for BD (BipoläR) [18]. Patients were diagnosed according to the DSM-IV-TR with BD type 1, BD type 2, BD not otherwise specified (BD-NOS), or schizoaffective disorder bipolar type. BipoläR captures basic demographic data for each individual along with interventions and outcomes. Outpatient clinics can register patients in BipoläR at any time point during the course of illness. A small number of cases and controls (N = 284) were enrolled through the St. Göran Bipolar Project—whose work-up procedures have been previously described elsewhere [19,20,21]—but were all also included in BipoläR. SWEBIC study participants were recruited up until December 2013 and more than 5000 patients volunteered to participate. All ascertainment procedures were approved by the Regional Ethical Review Board in Stockholm, Sweden, and all participants provided written informed consent.
For the familial coaggregation study, we established a study population of biological relatives to index BD persons by linking the quality register BipoläR to Swedish National Registries [22]. The first-degree (parent, offspring, and full sibling) and second-degree relatives (grandparents, grandchildren, aunt/uncle, nephew/niece, maternal and paternal half-sibling) were identified through the Multi-Generation Registry. The lifetime diagnoses of BD, schizophrenia, and MDD in relatives to BD index persons were identified from the National Patient Registry where records of psychiatric inpatient discharges are available since 1973 and outpatient visits in specialized psychiatric care since 2001. Diagnoses for BD, schizophrenia, and MDD were coded according to the International Classification of Diseases (ICD) with a hierarchical approach (Supplementary Table S1). Individuals’ sex and year of birth were obtained from the Total Population Registry. All registries were followed from their start until December 2013. The study was approved by the regional ethics committee in Stockholm.
Cases from the UK were obtained from the Bipolar Disorder Research Network (BDRN) study, an ongoing program of research into the genetic and environmental causes of BD and related mood disorders. Detailed description of the program can be found in prior publications [23, 24]. Briefly, participants were recruited via community mental health teams, advertisements in the media, and through patient support organizations across the UK. Participants were ≥ 18 years old, met DSM-IV diagnostic criteria for main lifetime diagnosis of BD subtype 1 and 2, schizoaffective disorder of bipolar type, and BD-NOS, and provided written informed consent. A total of 2577 BD cases remained after genotyping quality control in the genotyping wave we received in 2014. The BDRN study has approval from the West Midlands NHS Research Ethics Committee (MREC/97/7/01) and all participating NHS Trusts and Health Boards.
Subphenotype definitions
The BD subphenotypes studied were inter-episode remission, global assessment of functioning (GAF; function and symptom burden dimensions for Swedish cohort, and function dimension for the UK cohort), psychotic symptoms during mood episodes, and comorbid anxiety disorders. [25] In the Swedish cohort, subphenotypes were obtained from three sources: BipoläR (with patients’ information registered at outpatient clinics), telephone interviews conducted by trained research nurses, and the National Patients Registry. In the UK cohort, subphenotypes were derived from clinical assessments done retrospectively by SCAN interview (assessed when the participants’ moods were stable) and available case-notes review. [26] A summary of definitions and sample sizes for each subphenotype are shown in Tables 1–2.
Genotyping and quality control (QC)
DNA collection and genotyping procedures in SWEBIC has been previously described [27]. In brief, DNA was extracted from whole blood samples stored at the Karolinska Institutet Biobank. Genotyping was conducted at the Broad Institute of Harvard and MIT using Affymetrix 6.0 (wave 1, Affymetrix, Santa Clara, CA, USA), Illumina OmniExpress (wave 2, Illumina, San Diego, CA, USA), and Illumina PsychArray-24 v1.2 (wave 3, Illumina, San Diego, CA, USA). The data genotyped on different arrays were processed in the Psychiatric Genomics Consortium (PGC) RICOPILI pipeline for QC and imputation [28]. Ancestry outliers were identified using data from 1000 Genome Project (Phase 3 version 5) [29]. The final sample contained 5458 cases of which 5180 had at least one recorded phenotype of interest. The QC exclusionary parameters for subjects were: genotype missingness rate >5%, ancestry outliers identified via multidimensional scaling (MDS), suspected sample error or contamination (i.e., subject heterozygosity rate >10%), ambiguous genetic sex, and a randomly selected member of any pair of subjects identified as related (pairwise pi-hat >0.20). Exclusionary parameters for single nucleotide polymorphisms (SNPs) were: marked deviations from Hardy-Weinberg equilibrium (P < 1 × 10−6), SNP missingness rate >5%, minor allele frequency (MAF) < 1%, differential missingness based on affection status (P < 1 × 10−6), and differential missingness based on haplotype (P < 1 × 10−10). Following the QC steps, imputation was performed by first pre-phasing the data using SHAPEIT2 and then imputing using IMPUTE2 with the 1000 Genomes Project integrated variant set (Phase 1, released March 2012) as the reference panel [30,31,32].
In the UK sample, DNA was extracted from whole blood at the neuropsychiatric genetics laboratory at Cardiff University. DNA were genotyped using Illumina OmniExpress and Illumina ComboChip. Genotyping QC and imputation for UK samples is described in detail in previous publications. [27, 33] Briefly, the QC exclusionary parameters for the BDRN sample were: subject missingness rate >2%, ambiguous genetic sex, subject heterozygosity rate >15%, SNP missingness rate >2%, MAF < 1%, marked departure from Hardy-Weinberg equilibrium (P < 5 × 10−5), differential missingness based on haplotype (P < 1 × 10−10) between cases and controls and differential missingness for SNPs (P < 1 × 10−3), population outliers identified via MDS, and a random member of each pair of related subjects (pairwise pi-hat >0.10). Datasets after QC were then pre-phased using SHAPEIT and imputed with the 1000 Genomes Project integrated variant set (Phase 1, released March 2012) as the reference panel using IMPUTE2 [30,31,32].
Polygenic risk score profiling
The most recent and largest GWAS for schizophrenia and BD performed by PGC were used as discovery sets for SNP selection and risk allele weighting for generating PRS [13, 34, 35]. For MDD-PRS discovery set, we used the GWAS summary statistics of MDD in 2018 excluding subjects from 23andMe [36]. We did not use the most recent MDD GWAS [37] because the publicly available summary statistics lack the imputation INFO score, and we wanted to avoid possible sample overlap between the UK biobank and the BDRN sample. We used the summary statistics of meta-analyzed GWAS with European ancestry only and performed GWAS meta-analyses after removing SWEBIC or BDRN samples to resolve sample overlap between training and testing sets. To further evaluate the specificity and potential overlap across the genetic liabilities of mental-related traits, we also generated PRS for anxiety disorders, neuroticism, and educational attainment using the most recent GWAS data available [38,39,40]. This was done in consideration of their potential relationship to the BD subphenotypes under investigation.
For PRS calculation, we applied PRS-CS (version Aug 10, 2023) that generates posterior variant effects accounting for linkage disequilibrium (LD) structure and genetic architecture [41], which outperforms the standard pruning and thresholding (P + T) method. PRS-CS was run with default values for parameters and automatic estimation of the global shrinkage parameter values phi per chromosome, with a specified seed for random number generation for reproducibility. The total sample size of each discovery GWAS was used for the sample size parameter. We used the “best guess” genotype hardcall files filtered on imputation quality (INFO score >0.8) as the bim file for target dataset to provide a list of available variants. The GWAS meta-analyses above were used to train PRS-CS through the LD reference panel constructed using the 1000 Genomes Project phase 3, European reference [29]. The genetic markers were further restricted to those present in the HapMap3 reference sample [42]. PRS were then generated for each individual as the sum of the imputed SNP dosages weighted by the PRS-CS posterior allele effect using PLINK1.9 [43]. All PRS were standardized using z-score transformations to have mean zero and unit variance within each target set (i.e., within each wave in SWEBIC cohort and within BDRN cohort) to account for variation in SNP numbers used for PRS calculation.
Association tests between PRS of major psychiatric disorders and BD subphenotypes
We tested if PRS of BD, schizophrenia and MDD were associated with BD subphenotypes using regression models fit for the respective outcome. In SWEBIC cohort, we applied ordinal logistic regression for the inter-episode remission variable, logistic regressions for the psychotic symptoms and comorbid anxiety disorder variables, and linear regressions for GAF-function and GAF-symptom variables. In the BDRN cohort, the same models (as in SWEBIC) were applied for psychotic symptoms, comorbid anxiety disorders, and GAF-function. For inter-episode remission, we merged ‘partial remission’ and ‘no remission’—because only 12 individuals reported ‘no remission’—and used a logistic regression model. Moreover, we did not test for GAF-symptom since this phenotype was unavailable in the UK sample. PRS were used as continuous variables in all analyses with adjustment for sex, birth year, the first six ancestry principal components, and genotyping platforms. All three PRS were tested jointly in the model. Additionally, we performed association tests stratified by the BD subtypes 1 and 2. Psychotic symptoms during mood episodes was not tested for BD2 cohort because BD2 cases did not have this subphenotype. Finally, we performed random-effect meta-analyses combining the results from the two cohorts for all patients with BD, BD1 and BD2, respectively. As a sensitivity test to evaluate the specificity of PRS effects for the three major psychiatric disorders on BD subphenotypes, we conducted additional analyses by including the PRS of anxiety disorders, neuroticism, and educational attainment in the models for all BD cases. We used the random-effect model to obtain more conservative estimates as it accounts for uncertainties resulting from heterogeneity across the two cohorts. The Bonferroni method was used to correct for multiple testing, with respective number of tests corrected for (indicated under each result table). All association tests were performed using R (v4.0.3) [44] and the meta-analyses were performed using package metafor [45].
Familial coaggregation between BD subphenotypes and major psychiatric disorders in relatives
Assuming that individuals with a family history of BD, schizophrenia, or MDD would have a higher genetic loading of the corresponding psychiatric disorder, we performed familial coaggregation analyses to examine the magnitude of subphenotype occurrence (or levels of continuous subphenotype measures, i.e., GAF) in BD index person with relatives of BD, schizophrenia and MDD. We investigated GAF-function, GAF-symptom, as well as diagnoses of psychotic and anxiety disorders in index persons with BD (see Supplementary Table S1 for ICD codes). We restricted our analysis to cohorts of relatives diagnosed with either BD, schizophrenia, or MDD, and to index BD cases who had data on the respective subphenotype. We used a linear regression model to estimate the difference in GAF-function/symptom between index BD cases with schizophrenia relatives and index BD cases with BD relatives. Logistic regression models were used to estimate the odds ratios (ORs) of a comorbid anxiety disorder and psychotic symptoms in index BD cases who had relatives with schizophrenia compared with those whose relatives had BD. The same analyses were repeated in BD individuals who had relatives with MDD compared with those whose relatives had BD. To increase power, we estimated the coefficients by combining all first- and second-degree relatives. We adjusted for sex, year of birth category (before 1955, 1955–1962, 1963–1969, 1970–1977, 1978–1988, after 1988), and biological relatedness (first- and second-degree relatives) to the index individuals. We calculated the 95% confidence intervals (CIs) with robust standard errors to account for the non-independence between individuals due to familial clustering. The subphenotype ‘inter-episode remission’ was not tested because SWEBIC telephone interview data could not be linked to the Multi-Generation Registry due to ethical restraints. SAS 9.4 (SAS Institute, Cary, NC) was used for data management and analysis for this part [46].
Results
The descriptive characteristics for SWEBIC and BDRN study participants are shown in Table 2. The two cohorts differ in several aspects, including sex ratio, frequencies of BD subtypes and the prevalence of the four studied subphenotypes, which is expected considering different sample recruitments and different subphenotype assessments. Of note, we observed a higher percentage of BD1 in BDRN, which is likely because cases used in this study were mainly recruited via mental health services for the International Cohort Collection of Bipolar Disorder consortium [33]. Moreover, the few BD-NOS cases were removed due to the low frequency after genotyping QC in BDRN. The separate characteristics in BD1 and BD2 cases was available in Supplementary Table S2.
The associations between PRS for the three psychiatric disorders and each BD subphenotype are shown in Fig. 1 and Supplementary Table S3. For BD-PRS, we observed similar trends of associations with BD subphenotypes in SWEBIC and BDRN cohorts, except for psychotic symptoms during mood episodes that showed a positive correlation with BD-PRS in BDRN (OR 1.21, 95%CI 1.09–1.33, P = 2.12 × 10−4) but no correlation in the SWEBIC cohort (OR = 1.02, P = 0.53). Meta-analyses based on both cohorts yielded no significant heterogeneities for other subphenotypes. BD-PRS was positively associated with complete inter-episode remission (OR 1.16, 95%CI 1.10–1.23, P = 1.05 × 10−7). This means that the odds of complete vs. incomplete inter-episode remission increases by 16% (95%CI 10–23%) with each increasing unit of BD-PRS given the same SCZ-PRS and MDD-PRS. Further, BD-PRS associated positively with GAF-function (beta 0.78, 95%CI 0.38–1.17, P = 1.06 × 10−4), meaning that higher BD-PRS predicted higher global functioning. Finally, higher BD-PRS was associated with lower risk of comorbid anxiety disorders (OR 0.88, 95%CI 0.83–0.93, P = 1.60 × 10−5). Additionally, in the SWEBIC cohort, higher BD-PRS was also associated with lower global symptoms (beta for GAF-symptom 0.96, 95%CI 0.53–1.38, P = 1.14 × 10−5).
For SCZ-PRS, in both cohorts, we found a negative association with full inter-episode remission and a positive association with higher likelihood of psychotic symptoms during mood episodes, with meta-analyzed OR estimates of 0.91 (95%CI 0.86–0.96, P = 6.98 × 10−4) and 1.19 (95%CI 1.12–1.26, P = 2.07 × 10−8), respectively. The association for the GAF-function did not survive correction for multiple testing (beta −0.49, 95%CI −0.88 – −0.11, P = 0.01, Fig. 1 and Table S3). Notably, however, the estimates of correlations between SCZ-PRS and GAF were all negative in both cohorts (GAF-function and GAF-symptom in SWEBIC, GAF-function in BDRN, see Fig. 1 and Table S3).
Regarding MDD-PRS, we found negative associations with full inter-episode remission (meta OR 0.84, 95%CI 0.80–0.89, P = 2.78 × 10−11) and GAF-function (meta beta −0.70, 95%CI −1.00 – −0.40, P = 3.76 × 10−6) and a positive association with the risk of comorbid anxiety disorders (meta OR 1.15, 95%CI 1.09–1.21, P = 8.73 × 10−7), with no significant heterogeneity between the two study cohorts. Further, MDD-PRS was negatively associated with GAF-symptom in the SWEBIC cohort. While MDD-PRS was found to be negatively associated with psychotic symptoms in the BDRN cohort, no association was found in the SWEBIC cohort or in the meta-analysis, with a Heterogeneity test P = 0.02.
In the sensitivity analysis where we tested the PRS effects of six traits jointly (BD, schizophrenia, MDD, anxiety disorders, neuroticism, and educational attainment), we observed similar results for the PRS of the three major psychiatric disorders, except that the associations for MDD-PRS with comorbid anxiety and GAF-function were no longer significant after multiple testing correction (P = 0.02 and 0.002, respectively; see Supplementary Table S4). Additionally, we found that neuroticism-PRS was associated with a higher risk of comorbid anxiety, while PRS for educational attainment was positively associated with complete inter-episode remission and with psychotic symptoms during mood episodes.
The results for association tests stratified by the BD subtypes 1 and 2 are shown in Fig. 2 (meta-analyses results) and Supplementary Tables S5–6. Among BD1 cases, the results remained similar compared to the main analyses, although the associations between BD-PRS and anxiety as well as SCZ-PRS and psychotic symptoms were not statistically significant (P = 0.04 and 0.05 in the meta-analyses, respectively). Among BD2 cases, the magnitude of the associations attenuated, and only the association between MDD-PRS and inter-episode remission remained statistically significant (meta OR 0.87, 95%CI 0.79–0.94, P = 0.001).
The familial coaggregation analyses between psychiatric disorders in relatives and subphenotypes (GAF, psychotic symptoms, and anxiety disorders) in index BD cases are shown in Fig. 3 and Supplementary Table S7. First, index BD cases with a relative diagnosed with schizophrenia or MDD had lower GAF-function and GAF-symptom than index BD cases with a relative diagnosed with BD. Second, anxiety disorders were more prevalent in index BD cases who had a relative diagnosed with schizophrenia or MDD compared with BD index cases who had a relative diagnosed with BD. Third, compared with BD cases who had a relative diagnosed with BD, the risk of having been diagnosed with a psychotic disorder was higher in BD cases who had a relative diagnosed with schizophrenia.
Discussion
Combining two large BD cohorts in Sweden and the UK, we found distinct associations between key BD subphenotypes and genetic loading for BD, schizophrenia, and MDD. Most notably, full inter-episode remission and GAF were positively correlated with BD-PRS, but negatively correlated with SCZ-PRS and MDD-PRS. Moreover, BD-PRS was negatively, and MDD-PRS positively, associated with the risk of comorbid anxiety disorders. Finally, SCZ-PRS was positively associated with psychotic symptoms during mood episodes. These findings were similar when restricted to the BD1 subtype and were further supported by familial coaggregation analyses using Swedish national registries. Here, GAF scores were higher in BD cases with BD relatives than in BD cases who had schizophrenia or MDD relatives. Further, psychotic and anxiety disorders were less common in BD cases with BD relatives than in BD cases with schizophrenia relatives. Finally, anxiety disorders were less common in BD cases with BD relatives than in BD cases with MDD relatives. Taken together, these findings help explain how genetic factors contribute to the heterogeneous presentations of BD, enhancing our understanding of how polygenic vulnerabilities underpin its multifaceted manifestations.
Intriguingly, these results resonate with the century-old distinction between recurrent manic-depressive disorder (BD) and progressive dementia praecox (later schizophrenia): The finding that complete inter-episode remission was associated with higher BD-PRS and lower SCZ-PRS echoes the prototypical episodic course of illness of BD and the chronic course of schizophrenia. Likewise, the finding that higher global functioning (as measured with GAF) was associated with higher BD-PRS and lower SCZ-PRS (borderline significant, P = 0.01) aligns with the Kraepelinian notion that schizophrenia features a chronic deteriorating course with persistent impairment, whereas individuals with BD are typically symptom-free when recovered from mood episodes. The divergent influences of genetic liabilities on global psychosocial functioning were further supported by the familial co-occurrence analysis where BD cases with a family history of schizophrenia had lower GAF compared with those with a family history of BD. Taken together, modern polygenetic and epidemiological analyses reflect the historical landmark distinction between recurrent manic-depressive insanity and progressive dementia praecox as outlined by Emil Kraepelin.
The fact that a higher genetic loading of BD was associated with higher psychosocial functioning (i.e., higher GAF-function score)—and lower symptom burden measured by GAF-symptom within the Swedish group—might seem to be counterintuitive. A higher genetic loading of a disease would be assumed to predict a more severe disorder. Such findings might be explained in several perspectives. First, in our cohorts, BD caseness is defined based on the occurrence of mood episodes, without a requirement for long-term impaired functioning or the presence of anxiety outside mood episodes. Indeed, BD-PRS, by definition increases the risk for elated mood episodes, has been reported to confer benefits in other domains, for example creativity and educational attainment [35, 47]. Second, the assessment of GAF took place in an outpatient setting, where patients are more likely to be in a state of recovery with fewer symptoms. Higher BD-PRS is thus associated with better functioning and fewer symptoms when BD patients are not experiencing acute mood episodes. Notably, this does not mean that higher BD-PRS would be associated with less severe (or less frequent) mood episodes, which we did not assess in this study. In fact, both in the SWEBIC sample and in a recent study, BD-PRS was associated with higher number of hospitalizations. [48, 49]
Higher MDD-PRS associated with less likelihood of inter-episode remission, lower psychosocial function (lower GAF-function score) and higher symptom burden (lower GAF-symptom score, tested in the Swedish sample only), as well as a higher rate of comorbid anxiety disorders. By and large, BD patients with a higher genetic loading of MDD present with worse symptoms and outcomes. The reverse association between MDD-PRS and inter-episode remission was not driven by BD2 cases. Moreover, the associations with lower GAF and higher likelihood of anxiety were also supported by the familial co-aggregation analyses using Swedish registry data (i.e., those with MDD relatives had lower GAF and higher risk of anxiety than those with BD relatives). Schizophrenia and MDD are the two psychiatric disorders with the highest genetic correlations with BD [50], but in this regard the BD genetic component seems unique and not shared with the two other psychiatric disorders. PRS derived from schizophrenia and MDD GWAS may thus partly index a general liability for chronic psychopathology severity, as suggested in another study that demonstrates a PRS gradient across schizophrenia and BD subtypes [14].
The observed associations remained significant even after accounting for the PRS of anxiety disorders, neuroticism and educational attainment, though the associations for MDD-PRS with comorbid anxiety and GAF-function were attenuated and only borderline significant. This was expected given the high genetic overlap between MDD, anxiety disorders, and neuroticism. The absence of a significant association between PRS for anxiety disorders and comorbid anxiety in BD cases might be partially attributed to the limited sample size (N = 17,310) of the discovery GWAS, which did not fully capture the genetic loading. The consistent findings underscore the specificity of PRS for BD, schizophrenia, and MDD in relation to BD subphenotypes, even when considering other correlated PRS. This reinforces our conclusion that, although psychiatric disorders share genetic risk factors, their polygenic liabilities may exert distinct influences on the trajectory and outcome within each disorder. Considering the genetic loading of multiple disorders could provide valuable prognostic insights beyond categorical diagnoses, although it should be noted that the predictive power may be limited for clinical purposes.
Despite somewhat differing characteristics between the two cohorts with respect to sex ratio, subtype frequencies and subphenotype distributions, most results were similar across the SWEBIC and BDRN cohorts. However, we also noted some differences. First, psychotic symptoms during mood episodes were positively associated with BD-PRS and negatively associated with MDD-PRS in the UK sample but not in the Swedish sample, and significantly heterogenous in the meta-analyses. Notably, however, these across cohort differences were not significant when the analysis was restricted to BD1 cases. Second, although in the same direction and demonstrating no signs of heterogeneity in statistical tests, the magnitudes of associations between three PRS and GAF-function differed. Moreover, we observed a negative association between GAF-function and neuroticism-PRS, and a positive association with education-PRS, in the Swedish sample, while no associations were found in the UK sample. Our result of inter-episode remission is also inconsistent with a recent study that uses OPCRIT item 90 to measure the inter-episode remission among BD cases and reports no relationship with BD-PRS [48]. This heterogeneity across studies, which is likely in part due to the different phenotype assessments, limits the generalizability of our findings. For example, the mean of GAF was higher in the BDRN cohort than the SWEBIC cohort (Table 2). This might be because GAF was assessed the week before interview when the participants were well in BDRN, but could occur at any time in SWEBIC. Future investigations in larger cohorts with harmonized phenotype assessments, and comparisons with GAF values from routine clinical samples, are warranted.
The strengths of this study include detailed phenotyping in two large BD cohorts from different countries. Moreover, the PRS analyses using genotype data and familial coaggregation analyses using registry data yield converging evidence between the genetic liability of three psychiatric disorders and BD subphenotypes. The limitations to consider first include that data on inter-episode remission and history of psychotic symptoms were collected during a telephone interview with trained research nurses in the Swedish cohort, the validity of which has not been formally tested. SWEBIC diagnoses and GAF assessments were made by mental health professionals in regular clinical care and not in a controlled research setting. Hence, the subphenotypes in our study cohort, with a mixture of telephone interviews, register data from the quality register BipoläR, and registry records from the patient register (using ICD codes) might not be readily generalizable, which may affect attempts to replicate findings. Although we tried to harmonize the phenotype assessments across the Swedish and UK studies, heterogeneity remains and further investigation are warranted. Second, it is challenging to interpret the results of PRS for one disorder and subtypes of the same disorder because “the detailed interpretation depends on the proportion of these subtypes in the discovery sample” [51]. Despite that the interpretation of how SCZ- and MDD-PRS impact BD subphenotypes is less limited by this constraint, caution is still warranted when interpreting the results for subphenotypes that commonly co-occur in two disorders (e.g., anxiety disorders in MDD and BD). Third, the assessments of subphenotypes should ideally rely on longitudinal assessments, which was not possible for individuals recently diagnosed with BD. Additionally, the determination of GAF differed across cohorts: the Swedish GAF-rating reflects the average scoring gathered during annual follow-up within the BD quality register, while in the UK cohort, GAF was assessed as a one-time evaluation. Although the remaining subphenotypes were based on lifetime history up to the point of interview, study persons are interviewed at a random time point (at which they were relatively well) during their course of illness, and attempts to estimate manifestations of a lifelong illness are subject to uncertainties. To improve precision in subphenotypes, future investigations with extended observation periods are encouraged, as this will enable a more accurate characterization of subphenotypes over time. Fourth, the associations between BD subphenotypes in the index person and three major psychiatric disorders in the relatives are due to shared familial factors, which could be genetic and/or environmental factors. Hence, although the results from our familial analyses provide further support for our genetic findings, there are alternative explanations. Moreover, the definition of psychotic symptoms differs in the PRS and familial analyses (see Table 1). Finally, it should be emphasized that our findings do not suggest that BD patients with a higher genetic loading of BD have a less severe disorder. Rather, these cases are more likely to exhibit typical BD symptoms, including periods of remission and good functioning between mood episodes.
In summary, our study indicates that polygenic liabilities for BD, schizophrenia, and MDD are differently associated with subphenotypes of BD. These findings help explain how genetic factors contribute to the heterogeneous presentations of BD, which ultimately provide insights for developing personalized approaches to the diagnosis and treatment.
Data availability
Custom written R and SAS scripts used for statistical analyses can be provided upon request. The pipeline to compute polygenic scores can be found on GitHub (https://github.com/getian107/PRScs). Due to data protection regulations, we are not allowed to share individual-level register data with a third party. However, Swedish register data used in this study can be applied from Statistics Sweden (https://www.scb.se/en/), the Swedish National Board of Health and Welfare (Socialstyrelsen, https://www.socialstyrelsen.se/en/), and the Centre of Register Västra Götaland (https://registercentrum.se/in-english/centre-of-registers-vaestra-goetaland/p/H1nJ_MX3z). GWAS summary statistics are publicly available through their respective consortia (Psychiatric Genomics Consortium, https://pgc.unc.edu/, Social Science Genetic Association Consortium (SSGAC), http://www.thessgac.org/data, and https://ctg.cncr.nl/software/summary_statistics).
References
Kraepelin E. Manic-Depressive Insanity and Paranoia. Livingstone: Edinburgh; 1919.
Dunayevich E, Keck PE Jr. Prevalence and description of psychotic features in bipolar mania. Curr Psychiatry Rep. 2000;2:286–90.
American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders (5th ed.; DSM–5). Washington, DC: American Psychiatric Association; 2013.
Forty L, Smith D, Jones L, Jones I, Caesar S, Cooper C, et al. Clinical differences between bipolar and unipolar depression. Br J Psychiatry. 2008;192:388–9.
Smedler E, Bergen SE, Song J, Landen M. Genes, biomarkers, and clinical features associated with the course of bipolar disorder. Eur Neuropsychopharmacol. 2019;29:1152–60.
Abe C, Rolstad S, Petrovic P, Ekman CJ, Sparding T, Ingvar M, et al. Bipolar disorder type I and II show distinct relationships between cortical thickness and executive function. Acta Psychiatr Scand. 2018;138:325–35.
Bora E, Yucel M, Pantelis C. Cognitive endophenotypes of bipolar disorder: a meta-analysis of neuropsychological deficits in euthymic patients and their first-degree relatives. J Affect Disord. 2009;113:1–20.
Cross-Disorder Group of the Psychiatric Genomics Consortium. Identification of risk loci with shared effects on five major psychiatric disorders: a genome-wide analysis. Lancet. 2013;381:1371–9.
Andreassen OA, Thompson WK, Schork AJ, Ripke S, Mattingsdal M, Kelsoe JR, et al. Improved detection of common variants associated with schizophrenia and bipolar disorder using pleiotropy-informed conditional false discovery rate. PLoS Genet. 2013;9:e1003455.
Bulik-Sullivan B, Finucane HK, Anttila V, Gusev A, Day FR, Loh PR, et al. An atlas of genetic correlations across human diseases and traits. Nat Genet. 2015;47:1236–41.
Frei O, Holland D, Smeland OB, Shadrin AA, Fan CC, Maeland S, et al. Bivariate causal mixture model quantifies polygenic overlap between complex traits beyond genetic correlation. Nat Commun. 2019;10:2417.
Amare AT, Schubert KO, Hou L, Clark SR, Papiol S, Heilbronner U, et al. Association of polygenic score for schizophrenia and HLA antigen and inflammation genes with response to lithium in bipolar affective disorder: a genome-wide association study. JAMA Psychiatry. 2018;75:65–74.
Bipolar Disorder, Schizophrenia Working Group of the Psychiatric Genomics Consortium. Genomic dissection of bipolar disorder and schizophrenia, including 28 subphenotypes. Cell. 2018;173:1705–15.e1716.
Allardyce J, Leonenko G, Hamshere M, Pardinas AF, Forty L, Knott S, et al. Association between schizophrenia-related polygenic liability and the occurrence and level of mood-incongruent psychotic symptoms in bipolar disorder. JAMA Psychiatry. 2018;75:28–35.
Coombes BJ, Markota M, Mann JJ, Colby C, Stahl E, Talati A, et al. Dissecting clinical heterogeneity of bipolar disorder using multiple polygenic risk scores. Transl psychiatry. 2020;10:314.
Amare AT, Schubert KO, Hou L, Clark SR, Papiol S, Cearns M, et al. Association of polygenic score for major depression with response to lithium in patients with bipolar disorder. Mol Psychiatry. 2021;26:2457–70.
Allardyce J, Cardno AG, Gordon-Smith K, Jones L, Di Florio A, Walters JTR, et al. Specificity of polygenic signatures across symptom dimensions in bipolar disorder: an analysis of UK bipolar disorder research Network data. Lancet Psychiatry. 2023;10:623–31.
Palsson E, Melchior L, Lindwall Sundel K, Karanti A, Joas E, Nordenskjold A, et al. Cohort profile: the Swedish national quality register for bipolar disorder(BipolaR). BMJ open. 2022;12:e064385.
Jakobsson J, Zetterberg H, Blennow K, Johan Ekman C, Johansson AG, Landen M. Altered concentrations of amyloid precursor protein metabolites in the cerebrospinal fluid of patients with bipolar disorder. Neuropsychopharmacology. 2013;38:664–672.
Ryden E, Thase ME, Straht D, Aberg-Wistedt A, Bejerot S, Landen M. A history of childhood attention-deficit hyperactivity disorder (ADHD) impacts clinical outcome in adult bipolar patients regardless of current ADHD. Acta Psychiatr Scand. 2009;120:239–46.
Ekman CJ, Lind J, Ryden E, Ingvar M, Landen M. Manic episodes are associated with grey matter volume reduction - a voxel-based morphometry brain analysis. Acta Psychiatr Scand. 2010;122:507–15.
Ludvigsson JF, Almqvist C, Bonamy AK, Ljung R, Michaelsson K, Neovius M, et al. Registers of the Swedish total population and their use in medical research. Eur J Epidemiol. 2016;31:125–36.
Gordon-Smith K, Forty L, Chan C, Knott S, Jones I, Craddock N, et al. Rapid cycling as a feature of bipolar disorder and comorbid migraine. J Affect Disord. 2015;175:320–4.
Di Florio A, Mei Kay Yang J, Crawford K, Bergink V, Leonenko G, Pardinas AF, et al. Post-partum psychosis and its association with bipolar disorder in the UK: a case-control study using polygenic risk scores. Lancet Psychiatry. 2021;8:1045–52.
Pedersen G, Karterud S. The symptom and function dimensions of the Global Assessment of Functioning (GAF) scale. Compr Psychiatry. 2012;53:292–8.
Wing JK, Babor T, Brugha T, Burke J, Cooper JE, Giel R, et al. SCAN. schedules for clinical assessment in neuropsychiatry. Arch Gen Psychiatry. 1990;47:589–93.
Song J, Bergen SE, Di Florio A, Karlsson R, Charney A, Ruderfer DM, et al. Genome-wide association study identifies SESTD1 as a novel risk gene for lithium-responsive bipolar disorder. Mol Psychiatry. 2016;21:1290–7.
Lam M, Awasthi S, Watson HJ, Goldstein J, Panagiotaropoulou G, Trubetskoy V, et al. RICOPILI: rapid imputation for COnsortias PIpeLIne. Bioinformatics. 2020;36:930–3.
Genomes Project C, Auton A, Brooks LD, Durbin RM, Garrison EP, Kang HM, et al. A global reference for human genetic variation. Nature. 2015;526:68–74.
Genomes Project C, Abecasis GR, Auton A, Brooks LD, DePristo MA, Durbin RM, et al. An integrated map of genetic variation from 1,092 human genomes. Nature. 2012;491:56–65.
Delaneau O, Coulonges C, Zagury JF. Shape-IT: new rapid and accurate algorithm for haplotype inference. BMC Bioinforma. 2008;9:540.
Howie B, Marchini J, Stephens M. Genotype imputation with thousands of genomes. G3. 2011;1:457–470.
Charney AW, Ruderfer DM, Stahl EA, Moran JL, Chambert K, Belliveau RA, et al. Evidence for genetic heterogeneity between clinical subtypes of bipolar disorder. Transl Psychiatry. 2017;7:e993.
Trubetskoy V, Pardinas AF, Qi T, Panagiotaropoulou G, Awasthi S, Bigdeli TB, et al. Mapping genomic loci implicates genes and synaptic biology in schizophrenia. Nature. 2022;604:502–8.
Mullins N, Forstner AJ, O’Connell KS, Coombes B, Coleman JRI, Qiao Z, et al. Genome-wide association study of more than 40,000 bipolar disorder cases provides new insights into the underlying biology. Nat Genet. 2021;53:817–29.
Wray NR, Ripke S, Mattheisen M, Trzaskowski M, Byrne EM, Abdellaoui A, et al. Genome-wide association analyses identify 44 risk variants and refine the genetic architecture of major depression. Nat Genet. 2018;50:668–81.
Howard DM, Adams MJ, Clarke TK, Hafferty JD, Gibson J, Shirali M, et al. Genome-wide meta-analysis of depression identifies 102 independent variants and highlights the importance of the prefrontal brain regions. Nat Neurosci. 2019;22:343–52.
Otowa T, Hek K, Lee M, Byrne EM, Mirza SS, Nivard MG, et al. Meta-analysis of genome-wide association studies of anxiety disorders. Mol Psychiatry. 2016;21:1485.
Lee JJ, Wedow R, Okbay A, Kong E, Maghzian O, Zacher M, et al. Gene discovery and polygenic prediction from a genome-wide association study of educational attainment in 1.1 million individuals. Nat Genet. 2018;50:1112–21.
Nagel M, Jansen PR, Stringer S, Watanabe K, de Leeuw CA, Bryois J, et al. Meta-analysis of genome-wide association studies for neuroticism in 449,484 individuals identifies novel genetic loci and pathways. Nat Genet. 2018;50:920–7.
Ge T, Chen CY, Ni Y, Feng YA, Smoller JW. Polygenic prediction via Bayesian regression and continuous shrinkage priors. Nat Commun. 2019;10:1776.
International HapMap C, Altshuler DM, Gibbs RA, Peltonen L, Altshuler DM, Gibbs RA, et al. Integrating common and rare genetic variation in diverse human populations. Nature. 2010;467:52–8.
Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MA, Bender D, et al. PLINK: a tool set for whole-genome association and population-based linkage analyses. Am J Hum Genet. 2007;81:559–75.
Team RC. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria 2015.
Viechtbauer W. Conducting Meta-Analyses in R with the metafor Package. J Stat Softw. 2010;36:1–48.
Inc. SI. SAS Institute Inc. SAS® 9.4 Guide to Software Updates. Cary, NC: SAS Institute Inc 2013.
Power RA, Steinberg S, Bjornsdottir G, Rietveld CA, Abdellaoui A, Nivard MM, et al. Polygenic risk scores for schizophrenia and bipolar disorder predict creativity. Nat Neurosci. 2015;18:953–5.
Kalman JL, Papiol S, Grigoroiu-Serbanescu M, Adorjan K, Anderson-Schmidt H, Brosch K, et al. Genetic risk for psychiatric illness is associated with the number of hospitalizations of bipolar disorder patients. J Affect Disord. 2022;296:532–40.
Jonsson L, Hörbeck E, Primerano A, Song J, Karlsson R, Smedler E et al. Not all bipolar disorder outcomes are created equal: occupational dysfunction and hospital admissions associate with different polygenic profiles. Am. J. Psychiatry. 2023 (in press).
Song J, Bergen SE, Kuja-Halkola R, Larsson H, Landen M, Lichtenstein P. Bipolar disorder and its relation to major psychiatric disorders: a family-based study in the Swedish population. Bipolar Disord. 2015;17:184–193.
Wray NR, Lee SH, Mehta D, Vinkhuyzen AA, Dudbridge F, Middeldorp CM. Research review: polygenic methods and their application to psychiatric traits. J child Psychol Psychiatry,Allied Discip. 2014;55:1068–87.
Rucker J, Newman S, Gray J, Gunasinghe C, Broadbent M, Brittain P, et al. OPCRIT+: an electronic system for psychiatric diagnosis and data collection in clinical and research settings. Br J Psychiatry. 2011;199:151–5.
Acknowledgements
SWEBIC was supported by the Stanley Center for Psychiatric Research, Broad Institute from a grant from Stanley Medical Research Institute. ML was supported by the Swedish Research Council (2022-01643), Hjärnfonden/the Swedish Brain foundation (FO2022-0217), the Wenner-Gren foundation (SSv2019-0008), and the Swedish Government under the LUA/ALF agreement (ALFGBG-965444). JS was supported by the Brain & Behavior Research Foundation Young Investigator Grant (2022-31182) and by a research grant from the Karolinska Institutet (FS-2018:0007). ADF is funded by ERC Starting Grant number 947763 and UKRI grant number MR/w004658/1. LJ is supported by ALFGBG-966370 and Genomic Medicine Sweden (GMS K131050263). ES is supported by ALFGBG-933732, The Wallenberg Centre for Molecular and Translational Medicine and the Swedish Research Council (2022-01444). We thank the participation of all subjects contributing to this research and the collection team. BDRN wishes to acknowledge funding from by the Wellcome Trust and Stanley Medical Research Institute, and to thank all BDRN research participants for their ongoing support.
Funding
Open access funding provided by Karolinska Institute.
Author information
Authors and Affiliations
Contributions
JS contributed to the study conception and design, conducted the statistical analyses, drafted and revised the manuscript. RK, KGS, LJ (Lisa Jones), LJ (Lina Jonsson), NC, PL, IJ, ADF and ML contributed to the phenotypic and genetic data collection or preparation. JS, LJ (Lina Jonsson), YL, SEB, RK, ES, PFS, PL, LJ (Lisa Jones), ADF and ML contributed to the interpretation and discussion of the results. ML conceived the study idea, supervised the implementation, and revised the manuscript. All authors critically reviewed the manuscript, made important intellectual contributions, and approved the final version.
Corresponding author
Ethics declarations
Competing interests
ML declares that he has received lecture honoraria from Lundbeck pharmaceuticals outside the present work.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary information
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Song, J., Jonsson, L., Lu, Y. et al. Key subphenotypes of bipolar disorder are differentially associated with polygenic liabilities for bipolar disorder, schizophrenia, and major depressive disorder. Mol Psychiatry 29, 1941–1950 (2024). https://doi.org/10.1038/s41380-024-02448-1
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1038/s41380-024-02448-1
- Springer Nature Limited