Advertisement

The Genetic and Neural Substrates of Externalizing Behavior

Open AccessPublished:October 06, 2021DOI:https://doi.org/10.1016/j.bpsgos.2021.09.007

      Abstract

      Background

      To gain more insight into the biological factors that mediate vulnerability to display externalizing behaviors, we leveraged genome-wide association study summary statistics on 13 externalizing phenotypes.

      Methods

      After data classification based on genetic resemblance, we performed multivariate genome-wide association meta-analyses and conducted extensive bioinformatic analyses, including genetic correlation assessment with other traits, Mendelian randomization, and gene set and gene expression analyses.

      Results

      The genetic data could be categorized into disruptive behavior (DB) and risk-taking behavior (RTB) factors, and subsequent genome-wide association meta-analyses provided association statistics for DB and RTB (Neff = 523,150 and 1,506,537, respectively), yielding 50 and 257 independent genetic signals. The statistics of DB, much more than RTB, signaled genetic predisposition to adverse cognitive, mental health, and personality outcomes. We found evidence for bidirectional causal influences between DB and substance use behaviors. Gene set analyses implicated contributions of neuronal cell development (DB/RTB) and synapse formation and transcription (RTB) mechanisms. Gene-brain mapping confirmed involvement of the amygdala and hypothalamus and highlighted other candidate regions (cerebellar dentate, cuneiform nucleus, claustrum, paracentral cortex). At the cell-type level, we noted enrichment of glutamatergic neurons for DB and RTB.

      Conclusions

      This bottom-up, data-driven study provides new insights into the genetic signals of externalizing behaviors and indicates that commonalities in genetic architecture contribute to the frequent co-occurrence of different DBs and different RTBs, respectively. Bioinformatic analyses supported the DB versus RTB categorization and indicated relevant biological mechanisms. Generally similar gene-brain mappings indicate that neuroanatomical differences, if any, escaped the resolution of our methods.

      Keywords

      Externalizing psychopathologies are highly common in the general population (
      • Krueger R.F.
      • Kotov R.
      • Watson D.
      • Forbes M.K.
      • Eaton N.R.
      • Ruggero C.J.
      • et al.
      Progress in achieving quantitative classification of psychopathology.
      ,
      • Krueger R.F.
      • South S.C.
      Externalizing disorders: Cluster 5 of the proposed meta-structure for DSM-V and ICD-11.
      ,
      • Beauchaine T.P.
      • Zisner A.R.
      • Sauder C.L.
      Trait impulsivity and the externalizing spectrum.
      ,
      • Ahmad S.I.
      • Hinshaw S.P.
      Attention-deficit/hyperactivity disorder, trait impulsivity, and externalizing behavior in a longitudinal sample.
      ), and expressions of externalizing behavior including bullying, violence, delinquent activities, risk taking, and other related actions have a negative impact on individual prospects, local communities, and our society as a whole (
      GBD 2016 Alcohol Collaborators
      Alcohol use and burden for 195 countries and territories, 1990–2016: A systematic analysis for the Global Burden of Disease Study 2016.
      ). It is therefore vitally important to gain more insight into the underlying (neuro)biological processes to understand why some individuals are more susceptible to displaying externalizing behaviors than others.
      Externalizing symptoms generally co-occur substantially (
      • Bartels M.
      • Hendriks A.
      • Mauri M.
      • Krapohl E.
      • Whipp A.
      • Bolhuis K.
      • et al.
      Childhood aggression and the co-occurrence of behavioural and emotional problems: Results across ages 3–16 years from multiple raters in six cohorts in the EU-ACTION project.
      ), which suggests commonalities in their etiologies (
      • Beauchaine T.P.
      • Zisner A.R.
      • Sauder C.L.
      Trait impulsivity and the externalizing spectrum.
      ,
      • Mullins-Sweatt S.N.
      • DeShong H.L.
      • Lengel G.J.
      • Helle A.C.
      • Krueger R.F.
      Disinhibition as a unifying construct in understanding how personality dispositions undergird psychopathology.
      ). This is supported by indications for common genetic and environmental risk factors based on twin family studies (
      • Jang K.L.
      • Vernon P.A.
      • Livesley W.J.
      Personality disorder traits, family environment, and alcohol misuse: A multivariate behavioural genetic analysis.
      ,
      • Krueger R.F.
      • Hicks B.M.
      • Patrick C.J.
      • Carlson S.R.
      • Iacono W.G.
      • McGue M.
      Etiologic connections among substance dependence, antisocial behavior, and personality: Modeling the externalizing spectrum.
      ,
      • Venables N.C.
      • Hicks B.M.
      • Yancey J.R.
      • Kramer M.D.
      • Nelson L.D.
      • Strickland C.M.
      • et al.
      Evidence of a prominent genetic basis for associations between psychoneurometric traits and common mental disorders.
      ,
      • Young S.E.
      • Friedman N.P.
      • Miyake A.
      • Willcutt E.G.
      • Corley R.P.
      • Haberstick B.C.
      • Hewitt J.K.
      Behavioral disinhibition: Liability for externalizing spectrum disorders and its genetic and environmental relation to response inhibition across adolescence.
      ) and evidence for shared brain substrates from neurobiological studies (
      • Noordermeer S.D.S.
      • Luman M.
      • Oosterlaan J.
      A systematic review and meta-analysis of neuroimaging in oppositional defiant disorder (ODD) and conduct disorder (CD) taking attention-deficit hyperactivity disorder (ADHD) into account.
      ,
      • Yang Y.
      • Raine A.
      Prefrontal structural and functional brain imaging findings in antisocial, violent, and psychopathic individuals: A meta-analysis.
      ,
      • Ameis S.H.
      • Ducharme S.
      • Albaugh M.D.
      • Hudziak J.J.
      • Botteron K.N.
      • Lepage C.
      • et al.
      Cortical thickness, cortico-amygdalar networks, and externalizing behaviors in healthy children.
      ). There is also evidence for causal relationships, particularly between substance use and (other) externalizing symptoms (
      • Treur J.L.
      • Demontis D.
      • Smith G.D.
      • Sallis H.
      • Richardson T.G.
      • Wiers R.W.
      • et al.
      Investigating causality between liability to ADHD and substance use, and liability to substance use and ADHD risk, using Mendelian randomization.
      ), a topic of high relevance for public health.
      To broaden our understanding of common genetic and neurobiological backgrounds that underlie the co-occurrence of different externalizing behaviors, we leveraged publicly available genome-wide association study (GWAS) summary statistics on lifetime cannabis use (
      • Pasman J.A.
      • Verweij K.J.H.
      • Gerring Z.
      • Stringer S.
      • Sanchez-Roige S.
      • Treur J.L.
      • et al.
      GWAS of lifetime cannabis use reveals new risk loci, genetic overlap with psychiatric traits, and a causal influence of schizophrenia.
      ), antisocial behavior (
      • Tielbeek J.J.
      • Johansson A.
      • Polderman T.J.C.
      • Rautiainen M.R.
      • Jansen P.
      • Taylor M.
      • et al.
      Genome-wide association studies of a broad spectrum of antisocial behavior.
      ), aggressive behavior (
      • Pappa I.
      • St Pourcain B.
      • Benke K.
      • Cavadino A.
      • Hakulinen C.
      • Nivard M.G.
      • et al.
      A genome-wide approach to children’s aggressive behavior: The EAGLE consortium.
      ), four item indicators of angriness and irritability from UK Biobank (
      • Sudlow C.
      • Gallacher J.
      • Allen N.
      • Beral V.
      • Burton P.
      • Danesh J.
      • et al.
      UK Biobank: An open access resource for identifying the causes of a wide range of complex diseases of middle and old age.
      ), and attention-deficit/hyperactivity disorder (ADHD) diagnosis (
      • Demontis D.
      • Walters R.K.
      • Martin J.
      • Mattheisen M.
      • Als T.D.
      • Agerbo E.
      • et al.
      Discovery of the first genome-wide significant risk loci for attention deficit/hyperactivity disorder.
      ). Furthermore, we added five items from a recent study on the genetics of risk tolerance (
      • Karlsson Linnér R.
      • Biroli P.
      • Kong E.
      • Meddens S.F.W.
      • Wedow R.
      • Fontana M.A.
      • et al.
      Genome-wide association analyses of risk tolerance and risky behaviors in over 1 million individuals identify hundreds of loci and shared genetic influences.
      ). The 13 included datasets are summarized in Table 1.
      Table 1GWAS Summary Statistics Included in the Multivariate Genome-wide Association Meta-analysis
      PhenotypeSample SizeSource
      Aggression18,988Pappa et al. (
      • Pappa I.
      • St Pourcain B.
      • Benke K.
      • Cavadino A.
      • Hakulinen C.
      • Nivard M.G.
      • et al.
      A genome-wide approach to children’s aggressive behavior: The EAGLE consortium.
      )
      Angry Outbursts71,196Sudlow et al. (
      • Sudlow C.
      • Gallacher J.
      • Allen N.
      • Beral V.
      • Burton P.
      • Danesh J.
      • et al.
      UK Biobank: An open access resource for identifying the causes of a wide range of complex diseases of middle and old age.
      )
      Extreme Irritable157,357Sudlow et al. (
      • Sudlow C.
      • Gallacher J.
      • Allen N.
      • Beral V.
      • Burton P.
      • Danesh J.
      • et al.
      UK Biobank: An open access resource for identifying the causes of a wide range of complex diseases of middle and old age.
      )
      Irritability501,652Sudlow et al. (
      • Sudlow C.
      • Gallacher J.
      • Allen N.
      • Beral V.
      • Burton P.
      • Danesh J.
      • et al.
      UK Biobank: An open access resource for identifying the causes of a wide range of complex diseases of middle and old age.
      )
      Irritable for 2 Days202,883Sudlow et al. (
      • Sudlow C.
      • Gallacher J.
      • Allen N.
      • Beral V.
      • Burton P.
      • Danesh J.
      • et al.
      UK Biobank: An open access resource for identifying the causes of a wide range of complex diseases of middle and old age.
      )
      ADHD53,293Demontis et al. (
      • Demontis D.
      • Walters R.K.
      • Martin J.
      • Mattheisen M.
      • Als T.D.
      • Agerbo E.
      • et al.
      Discovery of the first genome-wide significant risk loci for attention deficit/hyperactivity disorder.
      )
      Antisocial Behavior25,781Tielbeek et al. (
      • Tielbeek J.J.
      • Johansson A.
      • Polderman T.J.C.
      • Rautiainen M.R.
      • Jansen P.
      • Taylor M.
      • et al.
      Genome-wide association studies of a broad spectrum of antisocial behavior.
      )
      General Risk Tolerance466,571Karlsson Linnér et al. (
      • Karlsson Linnér R.
      • Biroli P.
      • Kong E.
      • Meddens S.F.W.
      • Wedow R.
      • Fontana M.A.
      • et al.
      Genome-wide association analyses of risk tolerance and risky behaviors in over 1 million individuals identify hundreds of loci and shared genetic influences.
      )
      Drinks per Week414,343Karlsson Linnér et al. (
      • Karlsson Linnér R.
      • Biroli P.
      • Kong E.
      • Meddens S.F.W.
      • Wedow R.
      • Fontana M.A.
      • et al.
      Genome-wide association analyses of risk tolerance and risky behaviors in over 1 million individuals identify hundreds of loci and shared genetic influences.
      )
      Ever Smoker518,633Karlsson Linnér et al. (
      • Karlsson Linnér R.
      • Biroli P.
      • Kong E.
      • Meddens S.F.W.
      • Wedow R.
      • Fontana M.A.
      • et al.
      Genome-wide association analyses of risk tolerance and risky behaviors in over 1 million individuals identify hundreds of loci and shared genetic influences.
      )
      Number of Sexual Partners370,711Karlsson Linnér et al. (
      • Karlsson Linnér R.
      • Biroli P.
      • Kong E.
      • Meddens S.F.W.
      • Wedow R.
      • Fontana M.A.
      • et al.
      Genome-wide association analyses of risk tolerance and risky behaviors in over 1 million individuals identify hundreds of loci and shared genetic influences.
      )
      Automobile Speeding Propensity404,291Karlsson Linnér et al. (
      • Karlsson Linnér R.
      • Biroli P.
      • Kong E.
      • Meddens S.F.W.
      • Wedow R.
      • Fontana M.A.
      • et al.
      Genome-wide association analyses of risk tolerance and risky behaviors in over 1 million individuals identify hundreds of loci and shared genetic influences.
      )
      Lifetime Cannabis Use184,765Pasman et al. (
      • Pasman J.A.
      • Verweij K.J.H.
      • Gerring Z.
      • Stringer S.
      • Sanchez-Roige S.
      • Treur J.L.
      • et al.
      GWAS of lifetime cannabis use reveals new risk loci, genetic overlap with psychiatric traits, and a causal influence of schizophrenia.
      )
      ADHD, attention-deficit/hyperactivity disorder; GWAS, genome-wide association study.
      Using these data, we first identified clusters of traits with high genetic resemblance that together comprise the higher-order externalizing behavior dimension. Subsequently, the GWAS data of the clusters were meta-analyzed using N-weighted meta-analysis (
      • Baselmans B.M.L.
      • Jansen R.
      • Ip H.F.
      • van Dongen J.
      • Abdellaoui A.
      • van de Weijer M.P.
      • et al.
      Multivariate genome-wide analyses of the well-being spectrum.
      ). To characterize the genetic information in the meta-analyzed data, we evaluated the genetic relationships of the clusters with other phenotypes. In addition, for an identified cluster characterized by disruptive-type behaviors, we tested for causal relationship with smoking and alcohol consumption using bidirectional Mendelian randomization (MR). Previous studies have not been able to uncover the causal nature and direction of the relationship between disruptive behaviors (DBs) and smoking and alcohol use, two major risk factors for morbidity and mortality. Finally, biological annotation analyses were performed to identify associated brain regions and cell types.

      Methods and Materials

      Phenotype Selection

      Based on the disinhibited externalizing spectrum described by Krueger et al. (
      • Krueger R.F.
      • Kotov R.
      • Watson D.
      • Forbes M.K.
      • Eaton N.R.
      • Ruggero C.J.
      • et al.
      Progress in achieving quantitative classification of psychopathology.
      ), we collected GWAS summary statistics for 13 externalizing phenotypes (Table 1). Only samples with a European/North-American ancestry (
      • Gravel S.
      Population genetics models of local ancestry.
      ) were included (see Supplement 1).

      Identification of Genetic Factor Structures

      To identify genetic factor structures among the 13 included phenotypes, we first applied linkage disequilibrium score regression (LDSC) (
      • Bulik-Sullivan B.K.
      • Loh P.R.
      • Finucane H.K.
      • Ripke S.
      • Yang J.
      • Schizophrenia Working Group of the Psychiatric Genomics Consortium
      • et al.
      LD Score regression distinguishes confounding from polygenicity in genome-wide association studies.
      ,
      • Bulik-Sullivan B.
      • Finucane H.K.
      • Anttila V.
      • Gusev A.
      • Day F.R.
      • Loh P.R.
      • et al.
      An atlas of genetic correlations across human diseases and traits.
      ) to compute pairwise genetic correlations (rg) (see Supplement 1).
      Second, knowing the genetic correlations, we examined relationships using hierarchical clustering with (1 − rg) as genetic distance measures between phenotypes and with linkage based on Ward’s method (
      • Ward Jr., J.H.
      Hierarchical grouping to optimize an objective function.
      ). We followed the Calinski-Harabasz criterion to indicate the optimal number of clusters (
      • Caliński T.
      • Harabasz J.
      A dendrite method for cluster analysis.
      ).
      Third, in addition to hierarchical clustering, we applied factor analysis on the GWAS summary statistics using genomic structural equation modeling (
      • Grotzinger A.D.
      • Rhemtulla M.
      • de Vlaming R.
      • Ritchie S.J.
      • Mallard T.T.
      • Hill W.D.
      • et al.
      Genomic structural equation modelling provides insights into the multivariate genetic architecture of complex traits.
      ) (see Supplement 1).

      Multivariate N-Weighted Genome-wide Association Meta-analysis

      Univariate GWAS summary statistics were separately combined for identified factors using N-weighted genome-wide association meta-analysis (GWAMA) (
      • Baselmans B.M.L.
      • Jansen R.
      • Ip H.F.
      • van Dongen J.
      • Abdellaoui A.
      • van de Weijer M.P.
      • et al.
      Multivariate genome-wide analyses of the well-being spectrum.
      ), which is robust against sample overlap (see Supplement 1).

      Genetic Relationships With Other Traits

      We computed pairwise genetic correlations using LDSC (
      • Bulik-Sullivan B.K.
      • Loh P.R.
      • Finucane H.K.
      • Ripke S.
      • Yang J.
      • Schizophrenia Working Group of the Psychiatric Genomics Consortium
      • et al.
      LD Score regression distinguishes confounding from polygenicity in genome-wide association studies.
      ,
      • Bulik-Sullivan B.
      • Finucane H.K.
      • Anttila V.
      • Gusev A.
      • Day F.R.
      • Loh P.R.
      • et al.
      An atlas of genetic correlations across human diseases and traits.
      ) for our identified factors with 61 additional phenotypes in the following categories: mental health, cognition and socioeconomic status, personality, social, substance use, cardiovascular disease risk, physical health, anthropomorphic, and reproduction (see Table S6 in Supplement 2 and Supplement 1 for details).

      Mendelian Randomization

      For one of the identified clusters characterized by DBs, we applied MR to test for causal relationship with substance use behaviors (see Supplement 1).

      Gene Associations

      We performed gene-based analyses in MAGMA version 1.08 (http://ctg.cncr.nl/software/magma) (
      • de Leeuw C.A.
      • Mooij J.M.
      • Heskes T.
      • Posthuma D.
      MAGMA: Generalized gene-set analysis of GWAS data.
      ) with the N-weighted GWAMA summary statistics of the identified factors as input. The gene test statistics are defined as the mean single nucleotide polymorphism (SNP) association using the sum of −log (SNP p value).

      Tissue-Type Associations

      We used the gene associations of the two factors as input for a tissue-type analysis using MAGMA version 1.08 (
      • de Leeuw C.A.
      • Mooij J.M.
      • Heskes T.
      • Posthuma D.
      MAGMA: Generalized gene-set analysis of GWAS data.
      ). We investigated tissue-specific gene expression values as gene properties using 53 tissues from the Genotype-Tissue Expression project v.7 (
      GTEx Consortium
      Human genomics. The Genotype-Tissue Expression (GTEx) pilot analysis: Multitissue gene regulation in humans.
      ) (see Supplement 1).

      Gene Set Associations

      We continued with competitive gene set analyses using MAGMA version 1.08 (
      • de Leeuw C.A.
      • Mooij J.M.
      • Heskes T.
      • Posthuma D.
      MAGMA: Generalized gene-set analysis of GWAS data.
      ) to test whether the genes in a gene set are more strongly associated with the factor phenotypes than the other genes in the genome (see Supplement 1).

      Conditional Analyses for Tissue Types and Gene Sets

      For the significantly associated tissue types and gene sets, we performed conditional analyses using MAGMA version 1.08 (
      • de Leeuw C.A.
      • Mooij J.M.
      • Heskes T.
      • Posthuma D.
      MAGMA: Generalized gene-set analysis of GWAS data.
      ) to evaluate redundancy between associations (see Supplement 1).

      Stratified LDSC of Tissue Types

      We applied stratified LDSC to investigate which tissues and cell types are enriched for the identified factors (see Supplement 1).

      Stratified LDSC of Local Gene Expression Across the Human Brain

      To identify brain regions where genes relevant for externalizing behavior factors are differently expressed, we computed stratified LD scores based on differential gene expression, using data from 3707 gene expression measurements across 211 different brain regions (
      • Hawrylycz M.J.
      • Lein E.S.
      • Guillozet-Bongaarts A.L.
      • Shen E.H.
      • Ng L.
      • Miller J.A.
      • et al.
      An anatomically comprehensive atlas of the adult human brain transcriptome.
      ) (see Supplement 1).

      Stratified LDSC of Brain Cell Types

      We obtained a matrix of gene counts for single nuclei (n = 14,963) from the prefrontal cortex and hippocampus of multiple human donors studied by Habib et al. (
      • Habib N.
      • Avraham-Davidi I.
      • Basu A.
      • Burks T.
      • Shekhar K.
      • Hofree M.
      • et al.
      Massively parallel single-nucleus RNA-seq with DroNc-seq.
      ). We subsequently determined the differential expression of genes in seven types of brain cells: GABAergic (gamma-aminobutyric acidergic) interneurons, excitatory neurons, astrocytes, oligodendrocytes, microglia, endothelial cells, and neural stem cells (
      • Habib N.
      • Avraham-Davidi I.
      • Basu A.
      • Burks T.
      • Shekhar K.
      • Hofree M.
      • et al.
      Massively parallel single-nucleus RNA-seq with DroNc-seq.
      ) (see Supplement 1).

      Results

      Identification of Genetic Factor Structure

      LDSC (Figure 1A) indicated substantial genetic correlations, particularly among phenotypes characterized by disruptive-type behaviors (e.g., aggression and ADHD: rg = 0.72, SE = 0.18) and phenotypes characterized by risk-taking behaviors (RTBs) (e.g., cannabis use and number of sexual partners: rg = 0.69, SE = 0.02), with less genetic overlap between pairs of phenotypes characterized by the different behavior types (e.g., aggression and cannabis use: rg = 0.03, SE = 0.13). Hierarchical clustering on genetic resemblance (Figure 1B) confirmed this categorization into DBs (aggression, angry outbursts, different measures of irritability, and ADHD) and RTBs (antisocial behavior, general risk tolerance, drinks per week, ever smoker, number of sexual partners, automobile speeding propensity, and lifetime cannabis use). Exploratory factor analysis by genomic structural equation modeling supported a division into two phenotypic clusters by demonstrating a substantial increase in explained variance from 34% for a one-factor model to 53% for a two-factor model, but a much more moderate increase to 58% for a three-factor model (Figure 1C; Tables S1–S3 in Supplement 2). The two-factor exploratory factor analysis model assigned similar phenotypes to each factor as our hierarchical clustering approach. The loadings of the two-factor model, however, did not support a clear allocation for ADHD and antisocial behavior (Table S1 in Supplement 2). For these phenotypes, we followed the hierarchical clustering results, assigning ADHD to factor 1 (DB) and antisocial behavior to factor 2 (RTB).
      Figure thumbnail gr1
      Figure 1Genetic factor structure of externalizing behaviors. (A) Genetic correlations between the externalizing phenotypes calculated by linkage disequilibrium score regression. (B) Hierarchical clustering dendrogram based on genetic resemblance of the externalizing phenotypes. Blue represents the disruptive behavior cluster, and red represents the risk-taking behavior cluster. (C) Based on the results of an exploratory factor analysis of the genetic correlations presented in panel (A), a confirmatory factor model with two correlated genetic factors was specified using genomic structural equation modeling. In this model, the common factors account for the genetic covariation among the externalizing traits, i.e., each of the two common genetic factors represents variation in genetic liability that is shared across the phenotypes that load on it. Disruptive behavior represents shared genetic liability among disorders characterized by disruptive behavior, and risk-taking behavior represents the shared liability for risk-taking behavior. One-headed arrows connecting the common genetic factors to the individual traits represent standard loadings, which can be interpreted as coefficients from a regression of the true genetic liability for the trait on the common factor. Two-headed arrows connecting the genetic components represent their correlations. Two-headed arrows connecting the genetic components of the individual traits to themselves represent residual genetic variances and correspond to the proportion of heritable variation in liability to each individual trait that is unexplained by the two factors. ADHD, attention-deficit/hyperactivity disorder; Agg, aggression; Ao, angry outbursts; As, antisocial behavior; Asp, automobile speeding propensity; Can, lifetime cannabis use; Dr, drinks per week; Irxl, extreme irritable; Ir, irritability; Ir2d, irritable for 2 days; R, general risk tolerance; Sm, ever smoker; Sx, number of sexual partners.

      Multivariate GWAMA

      For both factors, we subsequently meta-analyzed the phenotype GWAS data using N-weighted GWAMA. For the DB factor, we identified 50 independent genome-wide significant SNPs at 42 loci (Neff = 523,150) (Figure 2A; Table S4 in Supplement 2). For RTB, we identified 257 independent genome-wide significant SNPs at 194 genomic risk loci (Neff = 1,506,537) (Figure 2B; Table S5 in Supplement 2). The LD score intercepts were close to 1 for both DB (intercept = 1.0167, SE = 0.0086; LDSC ratio = 0.0367, SE = 0.0156) and RTB (intercept = 1.0031, SE = 0.0132; LDSC ratio = 0.0242, SE = 0.0106), indicating that neither population stratification nor sample overlap, but rather an increase of polygenic signal, was driving the SNP associations. The SNP heritability as defined by LDSC was 0.0396 (SE = 0.0015) for DB and 0.022 (SE = 0.0007) for RTB. The genetic correlation between DB and RTB was 0.33 (SE = 0.02).
      Figure thumbnail gr2
      Figure 2Manhattan plots of the meta-analyzed phenotype genome-wide association study data. (A) Disruptive behavior factor. (B) Risk-taking behavior factor. The x-axis represents the chromosomal position, and the y-axis represents the significance on a −log10 scale. Each approximately independent genome-wide significant association (lead SNP) is marked by a triangle (p < 5 × 10−8). SNP, single nucleotide polymorphism.

      Genetic Relationships With Other Traits

      Figure 3 and Table S7 in Supplement 2 show pairwise genetic correlations for DB and RTB with 61 additional phenotypes. DB, much more than RTB, showed a pattern of genetic overlap pointing to adverse outcomes of cognition, socioeconomic status, several mental and physical health measures, and personality. For example, DB showed a negative genetic correlation with educational attainment (rg = −0.34, SE = 0.02) and income (rg = −0.44, SE = 0.04), whereas genetic correlations for RTB with these phenotypes were close to zero. In addition, DB was more positively correlated with depressive symptoms (rg = 0.73, SE = 0.03) and neuroticism (rg = 0.68, SE = 0.04) and more negatively correlated with agreeableness (rg = −0.59, SE = 0.05) compared with RTB (depressive symptoms: rg = 0.21, SE = 0.03; neuroticism: rg = −0.09, SE = 0.005; agreeableness: rg = −0.12, SE = 0.05).
      Figure thumbnail gr3
      Figure 3Genetic correlations with other phenotypes for disruptive behavior (blue) and risk-taking behavior (green) factors. Points represent the correlation estimates, and lines represent the 95% confidence intervals. Significant associations, after correction for multiple testing, are marked by orange stars. ADHD, attention-deficit/hyperactivity disorder; BMI, body mass index; HDL, high-density lipoprotein; LDL, low-density lipoprotein; MDD, major depressive disorder; SES, socioeconomic status.

      Causal Relationship of DB With Smoking and Alcohol Use

      We tested for bidirectional causal effects between DB (which does not include substance use phenotypes) and measures of smoking (
      • Liu M.
      • Jiang Y.
      • Wedow R.
      • Li Y.
      • Brazel D.M.
      • Chen F.
      • et al.
      Association studies of up to 1.2 million individuals yield new insights into the genetic etiology of tobacco and alcohol use.
      ) and alcohol use (
      • Walters R.K.
      • Polimanti R.
      • Johnson E.C.
      • McClintick J.N.
      • Adams M.J.
      • Adkins A.E.
      • et al.
      Transancestral GWAS of alcohol dependence reveals common genetic underpinnings with psychiatric disorders.
      ) using MR. We focused on these specific relationships because DBs and substance use are particularly strongly associated, and knowledge of (the direction of) potential causal effects has major public health implications. When DB was the exposure variable, inverse variance weighted analyses provided strong evidence for causal effects such that DB increases the odds of smoking initiation (β = 0.39, 95% CI = 0.24–0.54, p = 2.0 × 10−7) and decreases the odds of being able to successfully quit smoking (β = −0.20, 95% CI = −0.33 to −0.07, p = .002) (Table 2; Figure S1 in Supplement 1). Effect sizes and statistical evidence for a causal relationship between DB and smoking initiation were broadly consistent across the different MR methods. However, for DB to smoking cessation, the weighted mode and MR-Egger did not support a causal relationship (Table 2). Some evidence for pleiotropic effects was provided by a significant Cochran’s heterogeneity test (Cochran’s Q p value = 2.4 × 10−33) (Table S8 in Supplement 2) but not supported by the MR-Egger intercept (only DB → smoking initiation) (Table S9 in Supplement 2). Steiger filtering and MR-PRESSO did not affect the results and continued to support causal effects (Tables S10–S12 in Supplement 2). No consistent results were found for a causal relationship between DB and number of cigarettes smoked per day (Table 2; Table S9 in Supplement 2).
      Table 2Results of the Two-Sample Bidirectional Mendelian Randomization Analyses Between DB and Smoking and Alcohol Use Behavior
      ExposureOutcomeSNPs, nFIVWWeighted MedianWeighted ModeMR-Egger (SIMEX)GSMR
      β95% CIpβ95% CIpβ95% CIpβ95% CIpSNPs, nβ95% CIp
      DBSmoking initiation4037.410.390.24 to 0.542.0 × 10−70.250.13 to 0.373.2 × 10−50.15−0.06 to 0.36.1800.430.09 to 0.77.017330.3420.27 to 0.411.7 × 10−20
      DBCigarettes/day3837.610.140.01 to 0.27.029−0.01−0.12 to 0.10.895−0.06−0.16 to 0.04.284−0.31−0.60 to −0.02.041340.1780.09 to 0.263.3 × 10−5
      DBSmoking cessation4037.33−0.20−0.33 to −0.07.002−0.11−0.25 to 0.03.1180.02−0.24 to 0.28.8730.17−0.11 to 0.45.25437−0.190−0.29 to −0.099.9 × 10−5
      DBAlcohol/week4037.41−0.01−0.04 to 0.02.472−0.002−0.03 to 0.03.8970.01−0.02 to 0.04.545−0.04−0.11 to 0.03.34135−0.020−0.04 to 0.00.043
      DBAlcohol use disorder4637.000.260.08 to 0.44.0040.16−0.10 to 0.42.2450.14−0.12 to 0.40.288n.a.n.a.n.a.430.3100.10 to 0.52.003
      Smoking InitiationDB31727.530.170.14 to 0.209.9 × 10−310.150.12 to 0.188.0 × 10−210.150.05 to 0.25.004n.a.n.a.n.a.2960.1860.17 to 0.216.4 × 10−72
      Alcohol/weekDB8028.610.04−0.13 to 0.21.6320.052−0.12 to 0.22.543−0.17−2.07 to 1.73.859−0.24−0.54 to 0.06.133740.0719−0.05 to 0.20.261
      Alcohol Dis. 5 × 10−8DB726.35−0.02−0.06 to 0.02.299−0.02−0.06 to 0.02.480−0.01−0.08 to 0.06.833n.a.n.a.n.a.70.00016−0.04 to 0.04.993
      Alcohol Dis. 1 × 10−5DB2521.94−0.01−0.03 to 0.01.328−0.01−0.03 to 0.01.654−0.01−0.06 to 0.04.652n.a.n.a.n.a.24−0.0078−0.02 to 0.01.365
      n.a. indicates that MR-Egger results were not reported because of limited reliability based on the I2 measure being <0.60. F > 10 generally indicates the instrument is sufficiently strong.
      DB, disruptive behavior; Dis., disorder; F, F-statistic indicating instrument strength; GSMR, generalized summary data–based Mendelian randomization; IVW, inverse variance weighted regression; SIMEX, simulation extrapolation; SNP, single nucleotide polymorphism.
      From DB to alcohol use disorder, there was evidence for a causal increasing effect (inverse variance weighted analysis: β = 0.26, 95% CI = 0.08–0.44, p = .004), which was consistent (but weaker) in direction of effect across multiple MR methods (Table 2). MR-Egger was not available owing to low reliability; however, there was no heterogeneity based on Cochran’s Q (p = .457) (Table S8 in Supplement 2). Steiger filtering and MR-PRESSO did not change the outcome (Tables S10–S12 in Supplement 2).
      In the opposite direction, we found strong evidence for a causal, increasing effect of smoking initiation on DB (inverse variance weighted analysis: β = 0.17, 95% CI = 0.14–0.20, p = 9.9 × 10−31), which was consistent across weighted median, weighted mode, and generalized summary data–based MR (MR-Egger not available). Steiger filtered analyses and MR-PRESSO did not change the results (Table 2; Tables S10–S12 in Supplement 2). However, there was marked evidence for heterogeneity (Cochran’s Q p = 7.3 × 10−46) (Table S8 in Supplement 2). There was no clear evidence for causal effects of alcohol use on DB.

      Biological Annotations

      Gene Associations

      Genome-wide gene-based association analysis identified 81 genes significantly associated with DB (Table S13 in Supplement 2) and 318 genes significantly associated with RTB (Table S14 in Supplement 2) after Bonferroni correction. CADM2 showed an exceptionally strong association (2.10 × 10−29) with RTB.

      Tissue-Type Associations

      We used the gene-based test statistics as input for gene set analysis in MAGMA. Starting with a tissue-type analysis, we found significant associations at the Bonferroni level for nine brain regions for DB and 13 brain regions for RTB (Tables S15 and S16 in Supplement 2). Conditional analyses indicated that only the frontal cortex showed an independent association with DB, while independent associations were limited to the frontal cortex and cerebellum for RTB.

      Gene Set Associations

      Subsequent gene set analyses identified three gene sets for DB involved in cell development and 13 gene sets for RTB involved in cell development, synapse formation, and transcription (Tables S17 and S18 in Supplement 2). Conditional gene set analyses on the three gene sets associated with DB showed that the three gene set associations were highly related (Table S17 in Supplement 2). In addition, the conditional analyses of the 13 gene sets identified for RTB traits revealed redundancy for multiple gene sets related to synapse and cell development and transcription, leaving seven gene sets with independent association (Table S18 in Supplement 2).

      Stratified LDSC

      Complementary to gene set analysis, we applied stratified LDSC to study effects on 10 general tissue-type groups. Consistent with the gene set analyses, these analyses indicated significant enrichment after Bonferroni correction of the central nervous system (CNS) for both DB (ZCNS = 6.84) (Table S19 in Supplement 2) and RTB (ZCNS = 7.97) (Table S20 in Supplement 2). It has to be noted that other tissues were associated with DB and RTB as well (e.g., liver, kidney), but these associations were considerably weaker (Tables S19 and S20 in Supplement 2). Zooming in on associations on the cell-type level for DB revealed 45 significant cell type–specific annotations after Bonferroni correction, of which the top 15 all involved the CNS (Table S21 in Supplement 2). A similar pattern was observed for RTB, where we found 96 significant cell type–specific annotations, of which 32 indicated CNS involvement (Table S22 in Supplement 2).

      Differential Gene Expression

      To pinpoint relevant brain areas associated with DB and RTB more accurately, we proceeded with a local differential gene expression approach using stratified LDSC (
      • Hawrylycz M.J.
      • Lein E.S.
      • Guillozet-Bongaarts A.L.
      • Shen E.H.
      • Ng L.
      • Miller J.A.
      • et al.
      An anatomically comprehensive atlas of the adult human brain transcriptome.
      ). For DB, we identified significant enrichment after Bonferroni correction exclusively in the dentate nucleus (cerebellum, Z = 3.35, p = .0004) (Figure 4A; Table S23 in Supplement 2). For RTB, there was also enrichment in the dentate nucleus (cerebellum, Z = 3.22, p = .0006) and additional enrichment in the cuneiform nucleus (brain stem, Z = 4.74, p = 1.06 × 10−6) (Figure 4B), claustrum (cortex, Z = 4.08, p = 2.27 × 10−5) (Figure 4C), paracentral lobule (cortex, Z = 3.35, p = .0004) (Figure 4D), lateral amygdaloid nucleus (subcortex, Z = 3.12, p = .0009) (Figure 4E), and preoptic region (subcortex, Z = 3.34, p = .0004) (Figure 4F; Table S24 in Supplement 2).
      Figure thumbnail gr4
      Figure 4Brain regions with local differential gene expression enrichment for the disruptive behavior and risk-taking behavior factors in coronal, sagittal, and axial views. The locations of the samples of brain tissues that were used to measure gene expression by Hawrylycz et al. (
      • Hawrylycz M.J.
      • Lein E.S.
      • Guillozet-Bongaarts A.L.
      • Shen E.H.
      • Ng L.
      • Miller J.A.
      • et al.
      An anatomically comprehensive atlas of the adult human brain transcriptome.
      ) are projected to a standard Montreal Neurological Institute template brain (Colin27). For every annotation, the figure is centered on the averaged Montreal Neurological Institute coordinates of the brain samples. (A) Cerebellum: dentate nucleus. (B) Brainstem: cuneiform nucleus. (C) Cortex: claustrum. (D) Cortex: paracentral lobule. (E) Subcortex: lateral amygdaloid nucleus. (F) Subcortex: preoptic region.

      Single-Cell Analysis

      Finally, we investigated the involvement of specific brain cell types using LDSC. For DB, this revealed enrichment in excitatory neurons of the prefrontal cortex (Z = 2.88, p = .002) and the hippocampal CA3 region (Z = 2.85, p = .002) (Table S25 in Supplement 2). For RTB, we found enrichment in excitatory prefrontal cortex neurons (Z = 3.14, p = .0008) (Table S26 in Supplement 2).

      Discussion

      In a bottom-up approach to evaluate overlap in genetic and neurobiological backgrounds underlying externalizing symptoms, we collected 13 publicly available GWAS summary statistics for a range of externalizing phenotypes. Assessment of genetic resemblance indicated a categorization into two externalizing-related factors, one characterized by DB and another by RTB, that together explained 53% of the total genetic variance. Meta-analyzing the factor-specific phenotypes yielded 50 loci for DB and 257 loci for RTB.
      Clustering of aggression, angriness, and irritability items in the DB factor fits very well with the general finding that these are highly comorbid behaviors and also cluster together in diagnoses of conduct and oppositional defiant disorder (
      American Psychiatric Association
      Diagnostic and Statistical Manual of Mental Disorders.
      ). Addition of ADHD aligns with the fact that DBs coexist most commonly with ADHD (
      • Kunwar A.
      • Dewan M.
      • Faraone S.V.
      Treating common psychiatric disorders associated with attention-deficit/hyperactivity disorder.
      ) and with evidence of strong genetic resemblance between ADHD and conduct and oppositional defiant disorder (
      • Dick D.M.
      • Viken R.J.
      • Kaprio J.
      • Pulkkinen L.
      • Rose R.J.
      Understanding the covariation among childhood externalizing symptoms: Genetic and environmental influences on conduct disorder, attention deficit hyperactivity disorder, and oppositional defiant disorder symptoms.
      ). The phenotypes of the RTB factor related to pursuing risk (
      • Karlsson Linnér R.
      • Biroli P.
      • Kong E.
      • Meddens S.F.W.
      • Wedow R.
      • Fontana M.A.
      • et al.
      Genome-wide association analyses of risk tolerance and risky behaviors in over 1 million individuals identify hundreds of loci and shared genetic influences.
      ). Three RTB traits (general risk tolerance, number of sexual partners, lifetime cannabis use) similarly clustered together in another recent GWAMA (
      • Karlsson Linnér R.
      • Mallard T.T.
      • Barr P.B.
      • Sanchez-Roige S.
      • Madole J.W.
      • Driver M.N.
      • et al.
      Multivariate genomic analysis of 1.5 million people identifies genetic associations with traits related to self-regulation and addiction.
      ). Addition of antisocial behavior aligns with the fact that antisocial behavior frequently includes thoughtless, self-centered, and immediately rewarding acts, which are also typical for risk taking. The present RTB phenotypes related to substance use, speeding, and sexual promiscuity are closely linked to social misconduct (
      • Karlsson Linnér R.
      • Mallard T.T.
      • Barr P.B.
      • Sanchez-Roige S.
      • Madole J.W.
      • Driver M.N.
      • et al.
      Multivariate genomic analysis of 1.5 million people identifies genetic associations with traits related to self-regulation and addiction.
      ,
      • Mishra S.
      • Lalumiére M.L.
      Risk-taking, antisocial behavior, and life histories.
      ). In genomic structural equation modeling, antisocial behavior also loaded on DB, which is concordant with observations from the Hierarchical Taxonomy of Psychopathology consortium that antisocial behavior is a broad construct and also alludes to disruptive phenotypes (
      • Krueger R.F.
      • Kotov R.
      • Watson D.
      • Forbes M.K.
      • Eaton N.R.
      • Ruggero C.J.
      • et al.
      Progress in achieving quantitative classification of psychopathology.
      ,
      • Krueger R.F.
      • Hobbs K.A.
      • Conway C.C.
      • Dick D.M.
      • Dretsch M.N.
      • Eaton N.R.
      • et al.
      Validity and utility of Hierarchical Taxonomy of Psychopathology (HiTOP): II. Externalizing superspectrum.
      ).
      Assessment of genetic relationships with other traits indicated associations for DB with higher neuroticism and lower conscientiousness and agreeableness, corroborating previous findings (
      • Nigg J.T.
      • John O.P.
      • Blaskey L.G.
      • Huang-Pollock C.L.
      • Willcutt E.G.
      • Hinshaw S.P.
      • Pennington B.
      Big five dimensions and ADHD symptoms: Links between personality traits and clinical symptoms.
      ). For RTB, we found associations with higher extraversion, higher openness to experience, and lower conscientiousness, which is also in line with earlier reports (
      • McGhee R.L.
      • Ehrler D.J.
      • Buckhalt J.A.
      • Phillips C.
      The relation between five-factor personality traits and risk-taking behavior in preadolescents.
      ). Another notable result was that DB, much more than RTB, showed negative genetic associations with indices of mental and physical health. Previous studies indicated associations with mental and physical health problems for both disruptive-ADHD and risk-taking behaviors (
      • Barker E.D.
      • Ing A.
      • Biondo F.
      • Jia T.
      • Pingault J.B.
      • Du Rietz E.
      • et al.
      Do ADHD-impulsivity and BMI have shared polygenic and neural correlates?.
      ,
      • Loth A.K.
      • Drabick D.A.G.
      • Leibenluft E.
      • Hulvershorn L.A.
      Do childhood externalizing disorders predict adult depression? A meta-analysis [published correction appears in J Abnorm Child Psychol 2014; 42:1115–1116].
      ,
      • Pailing A.N.
      • Reniers R.L.E.P.
      Depressive and socially anxious symptoms, psychosocial maturity, and risk perception: Associations with risk-taking behaviour.
      ,
      • McCloughen A.
      • Foster K.
      • Marabong N.
      • Miu D.
      • Fethney J.
      Physical health risk behaviours in young people with mental illness.
      ). This finding of stronger genetic correlations for DB points to the possible distinction that DB phenotypes, characterized by affective instability/irritability and impaired cognitive control, are generally central to mental and physical health problems (
      • Loth A.K.
      • Drabick D.A.G.
      • Leibenluft E.
      • Hulvershorn L.A.
      Do childhood externalizing disorders predict adult depression? A meta-analysis [published correction appears in J Abnorm Child Psychol 2014; 42:1115–1116].
      ,
      • Derks I.P.M.
      • Bolhuis K.
      • Yalcin Z.
      • Gaillard R.
      • Hillegers M.H.J.
      • Larsson H.
      • et al.
      Testing bidirectional associations between childhood aggression and BMI: Results from three cohorts.
      ), whereas for RTB, mental health problems are more often at the core (
      • Pailing A.N.
      • Reniers R.L.E.P.
      Depressive and socially anxious symptoms, psychosocial maturity, and risk perception: Associations with risk-taking behaviour.
      ), resulting in risk behaviors, such as substance use, that ultimately affect physical health (
      • McCloughen A.
      • Foster K.
      • Marabong N.
      • Miu D.
      • Fethney J.
      Physical health risk behaviours in young people with mental illness.
      ).
      In addition, we applied MR to explore causal relationships between the DB factor and smoking and alcohol use. In full agreement with previous findings (
      • Treur J.L.
      • Demontis D.
      • Smith G.D.
      • Sallis H.
      • Richardson T.G.
      • Wiers R.W.
      • et al.
      Investigating causality between liability to ADHD and substance use, and liability to substance use and ADHD risk, using Mendelian randomization.
      ,
      • Fluharty M.E.
      • Sallis H.
      • Munafò M.R.
      Investigating possible causal effects of externalizing behaviors on tobacco initiation: A Mendelian randomization analysis.
      ), we obtained evidence that DB traits promote substance use (strong evidence for an increased odds of smoking initiation and decreased odds to successfully quit smoking and weak evidence for increased odds of alcohol use disorder). This causal direction is also consistent with expectations for behavioral development, in that DB traits are more likely to occur at younger ages than risk behaviors. Smoking and alcohol use are not yet appropriate and less feasible during early childhood (
      • Treur J.L.
      • Demontis D.
      • Smith G.D.
      • Sallis H.
      • Richardson T.G.
      • Wiers R.W.
      • et al.
      Investigating causality between liability to ADHD and substance use, and liability to substance use and ADHD risk, using Mendelian randomization.
      ,
      • Fluharty M.E.
      • Sallis H.
      • Munafò M.R.
      Investigating possible causal effects of externalizing behaviors on tobacco initiation: A Mendelian randomization analysis.
      ). However, we also replicated the observation that smoking might causally increase the risk for DB traits. This finding could have important consequences for intervention strategies but needs further inquiry. To distinguish a causal influence from horizontal pleiotropy, i.e., when a genetic variant influences the two traits through independent pathways, dose-response effects should be included in future MR analyses (taking cigarettes per day as the exposure and applying stratification on smoking status).
      Biological annotation by gene set analysis and stratified LDSC converged on genetic enrichment in the brain for both DB and RTB traits. An atlas of differential brain region gene expression pointed specifically to the dentate nucleus of the cerebellum, for both DB and RTB, and the cuneiform nucleus of the brainstem, preoptic region of the hypothalamus, lateral nucleus of the amygdala, and the claustrum and paracentral cortical lobule, which were significant for RTB.
      Involvement of these brain regions confirms current hypotheses on the neurobiological background of externalizing symptoms but also points to the importance of relatively underrepresented regions. Especially for the medial part of the preoptic hypothalamic region, animal studies indicate an important role in stimulating aggression to facilitate reproduction (
      • Aleyasin H.
      • Flanigan M.E.
      • Russo S.J.
      Neurocircuitry of aggression and aggression seeking behavior: Nose poking into brain circuitry controlling aggression.
      ), as well as in modulating mesolimbic activity involved in reward processing (
      • Tobiansky D.J.
      • Roma P.G.
      • Hattori T.
      • Will R.G.
      • Nutsch V.L.
      • Dominguez J.M.
      The medial preoptic area modulates cocaine-induced activity in female rats.
      ). The contribution of the amygdala to emotional processing is also generally recognized (
      • Phelps E.A.
      • LeDoux J.E.
      Contributions of the amygdala to emotion processing: From animal models to human behavior.
      ), and there is ample evidence that amygdala–frontal cortex network abnormalities predispose to externalizing behaviors (
      • Ameis S.H.
      • Ducharme S.
      • Albaugh M.D.
      • Hudziak J.J.
      • Botteron K.N.
      • Lepage C.
      • et al.
      Cortical thickness, cortico-amygdalar networks, and externalizing behaviors in healthy children.
      ), including aggression and risk taking (
      • Davidson R.J.
      • Putnam K.M.
      • Larson C.L.
      Dysfunction in the neural circuitry of emotion regulation—A possible prelude to violence.
      ,
      • O’Brien J.W.
      • Hill S.Y.
      Neural predictors of substance use disorders in young adulthood.
      ). In addition, amygdala volume reductions are commonly found in ADHD (
      • Tajima-Pozo K.
      • Ruiz-Manrique G.
      • Yus M.
      • Arrazola J.
      • Montañes-Rada F.
      Correlation between amygdala volume and impulsivity in adults with attention-deficit hyperactivity disorder.
      ,
      • Van Dessel J.
      • Sonuga-Barke E.
      • Moerkerke M.
      • Van der Oord S.
      • Lemiere J.
      • Morsink S.
      • Danckaerts M.
      The amygdala in adolescents with attention-deficit/hyperactivity disorder: Structural and functional correlates of delay aversion.
      ,
      • Hoogman M.
      • Bralten J.
      • Hibar D.P.
      • Mennes M.
      • Zwiers M.P.
      • Schweren L.S.J.
      • et al.
      Subcortical brain volume differences in participants with attention deficit hyperactivity disorder in children and adults: A cross-sectional mega-analysis [published correction appears in Lancet Psychiatry 2017; 4:436].
      ). Furthermore, a recent study showed that GABAergic neurons in the medial amygdala project to the medial preoptic area to regulate reward from social stimuli by controlling the release of dopamine in the nucleus accumbens (
      • Hu R.K.
      • Zuo Y.
      • Ly T.
      • Wang J.
      • Meera P.
      • Wu Y.E.
      • Hong W.
      An amygdala-to-hypothalamus circuit for social reward.
      ).
      The other brain regions have been implicated less frequently, despite previous evidence for associations. For the cerebellar dentate, a relevant role in the development of reinforcement learning relevant to addiction has been indicated (
      • Moulton E.A.
      • Elman I.
      • Becerra L.R.
      • Goldstein R.Z.
      • Borsook D.
      The cerebellum and addiction: Insights gained from neuroimaging research.
      ), and there is evidence that abnormal development of corticocerebellar connections contributes to ADHD (
      • Shaw P.
      • Ishii-Takahashi A.
      • Park M.T.
      • Devenyi G.A.
      • Zibman C.
      • Kasparek S.
      • et al.
      A multicohort, longitudinal study of cerebellar development in attention deficit hyperactivity disorder.
      ). The cuneiform nucleus is part of the mesencephalic locomotor control network (
      • Caggiano V.
      • Leiras R.
      • Goñi-Erro H.
      • Masini D.
      • Bellardita C.
      • Bouvier J.
      • et al.
      Midbrain circuits that set locomotor speed and gait selection.
      ) and related to autonomic fear and stress responses (
      • Silva B.A.
      • Gross C.T.
      • Gräff J.
      The neural circuits of innate fear: Detection, integration, action, and memorization.
      ,
      • Korte S.M.
      • Jaarsma D.
      • Luiten P.G.
      • Bohus B.
      Mesencephalic cuneiform nucleus and its ascending and descending projections serve stress-related cardiovascular responses in the rat.
      ). The claustrum is richly interconnected with almost all regions of the cerebral cortex and hypothesized to play a role in regulating attention and resilience to distraction (
      • Goll Y.
      • Atlan G.
      • Citri A.
      Attention: The claustrum.
      ,
      • Atlan G.
      • Terem A.
      • Peretz-Rivlin N.
      • Sehrawat K.
      • Gonzales B.J.
      • Pozner G.
      • et al.
      The claustrum supports resilience to distraction.
      ), which links this region to ADHD (
      • Dickstein S.G.
      • Bannon K.
      • Castellanos F.X.
      • Milham M.P.
      The neural correlates of attention deficit hyperactivity disorder: An ALE meta-analysis.
      ). The anterior section of the paracentral lobule includes motor control regions but has also been linked to executive control function and attention orienting (
      • Baselmans B.M.L.
      • Jansen R.
      • Ip H.F.
      • van Dongen J.
      • Abdellaoui A.
      • van de Weijer M.P.
      • et al.
      Multivariate genome-wide analyses of the well-being spectrum.
      ,
      • Gravel S.
      Population genetics models of local ancestry.
      ). Single-cell analysis further elucidated the important role of the CNS for both DB and RTB and pointed to glutamatergic prefrontal neurons for DB and RTB and pyramidal hippocampal neurons for DB. Glutamatergic neurotransmission has been associated with alcohol and tobacco use (
      • Liu M.
      • Jiang Y.
      • Wedow R.
      • Li Y.
      • Brazel D.M.
      • Chen F.
      • et al.
      Association studies of up to 1.2 million individuals yield new insights into the genetic etiology of tobacco and alcohol use.
      ) and RTB (
      • Karlsson Linnér R.
      • Biroli P.
      • Kong E.
      • Meddens S.F.W.
      • Wedow R.
      • Fontana M.A.
      • et al.
      Genome-wide association analyses of risk tolerance and risky behaviors in over 1 million individuals identify hundreds of loci and shared genetic influences.
      ) before, but more research is needed to address the role of these specific cell types and their functions on the externalizing phenotypes.
      Finally, testing for specific biological mechanisms using gene set analysis indicated associations for gene sets primarily related to neuronal development for both DB and RTB. In addition, synaptic functions and transcription regulation were identified for RTB. These three biological mechanisms have been related to a wide range of psychiatric disorders (
      Cross-Disorder Group of the Psychiatric Genomics Consortium
      Genomic relationships, novel loci, and pleiotropic mechanisms across eight psychiatric disorders.
      ) and may play a broad role in behaviors and brain-related traits.
      This study comes with a number of limitations. First, the two identified factors DB and RTB combine several distinct phenotypic measures, which inevitably leads to etiologic heterogeneity (reflected by the imperfect genetic correlations). In addition, phenotyping could be as poor as a single question (e.g., about regular use of a substance). Poor and heterogenous phenotyping, with the trade-off of having large sample sizes available, are a common phenomenon in GWAS meta-analyses of single traits (
      • Wray N.R.
      • Ripke S.
      • Mattheisen M.
      • Trzaskowski M.
      • Byrne E.M.
      • Abdellaoui A.
      • et al.
      Genome-wide association analyses identify 44 risk variants and refine the genetic architecture of major depression.
      ) or multiple traits (
      • Baselmans B.M.L.
      • Jansen R.
      • Ip H.F.
      • van Dongen J.
      • Abdellaoui A.
      • van de Weijer M.P.
      • et al.
      Multivariate genome-wide analyses of the well-being spectrum.
      ) and generally results in lower SNP heritability estimates compared with the individual traits, which was also the case here for DB (4%) and RTB (2%). Second, we incorporated three measures on irritability, which seems redundant. However, the mutual genetic correlation between these phenotypes was not perfect (rg < 1), and therefore, we note that including GWAS statistics of all three phenotypes does provide additional genetic information in GWAMAs. Third, in the MR analyses, the tested effects from smoking and alcohol use to DB did not fully comply with the expected temporal order. The aggression and ADHD variables of DB included children (which amounts to ∼5% of the total number of DB participants). This means that smoking and alcohol use were, for those variables, not valid exposures. In addition, we limited assessment of causal directions to the relationship between the DB factor and smoking and alcohol use. We also demonstrated interesting relationships of DB and RTB with several other traits, such as personality and health characteristics. Given the complexity of the included study cohorts forming the different summary statistics, it was beyond the scope of this study to also test for causal relationships with these traits. Fourth, except for ADHD, the GWAS data in our study related to behavioral extremes that did not exceed the diagnostic threshold. Given previous evidence of genetic heterogeneity between nondiagnosed and diagnosed individuals (
      • Johnson E.C.
      • Demontis D.
      • Thorgeirsson T.E.
      • Walters R.K.
      • Polimanti R.
      • Hatoum A.S.
      • et al.
      A large-scale genome-wide association study meta-analysis of cannabis use disorder.
      ,
      • Kranzler H.R.
      • Zhou H.
      • Kember R.L.
      • Vickers Smith R.
      • Justice A.C.
      • Damrauer S.
      • et al.
      Genome-wide association study of alcohol consumption and use disorder in 274,424 individuals from multiple populations [published correction appears in Nat Commun 2019; 10:2275 and Nat Commun 2019; 10:4050].
      ), we emphasize that the findings in this study apply in particular to general rather than excessive diagnosed externalizing behaviors. Finally, regarding the biological annotation analyses, we must consider limitations in drawing conclusions about different findings for DB and RTB. DB and RTB are moderately correlated (rg = 0.33) and therefore not completely independent. In addition, the effective sample size of RTB is ∼3 times larger than that of DB (1,506,537 vs. 523,150), so there is likely more power to detect significant associations for RTB. For example, differential gene expression indicated more brain regions with statistically significant enrichment for RTB, but we note that the same regions also showed relatively increased enrichment among the 211 brain regions tested for DB.
      In summary, we extend previous findings involving externalizing behavior and provide further evidence for common genetic architectures, particularly for different DBs and RTBs. Follow-up evaluations of the data in this study indicated genetic relationships with personality traits and mental and physical health behaviors in agreement with the DB versus RTB categorization and highlighted possible bidirectional causal relationships between DB and substance use traits. Biological annotation revealed generally similar gene-brain mappings that mediate the predisposition to comorbid externalizing phenotypes. Possible subtle differences in neurobiological backgrounds between DB and RTB traits may be resolved in future studies based on higher-resolution genetic association and gene expression data that are becoming increasingly available.

      Acknowledgments and Disclosures

      We thank the research participants and employees of 23andMe and UK Biobank for making this work possible. In addition, we thank the Social Science Genetic Consortium, the EAGLE consortium, and the International Cannabis Consortium for making their summary statistics publicly available. Without their data sharing policy, this study was not possible.
      This study used genome-wide association study summary statistics published by 23andMe and the UK Biobank. All input data used in the present study are from public sources (listed in Table 1). Output data are accessible after a data request application to [email protected]
      The code to perform the multivariate analyses can be found at: https://github.com/baselmans/multivariate_GWAMA
      BB is supported by an NWO-ZonMW Rubicon Grant (Grant No. 45219101). All other authors report no biomedical financial interests or potential conflicts of interest.

      Supplementary Material

      References

        • Krueger R.F.
        • Kotov R.
        • Watson D.
        • Forbes M.K.
        • Eaton N.R.
        • Ruggero C.J.
        • et al.
        Progress in achieving quantitative classification of psychopathology.
        World Psychiatry. 2018; 17: 282-293
        • Krueger R.F.
        • South S.C.
        Externalizing disorders: Cluster 5 of the proposed meta-structure for DSM-V and ICD-11.
        Psychol Med. 2009; 39: 2061-2070
        • Beauchaine T.P.
        • Zisner A.R.
        • Sauder C.L.
        Trait impulsivity and the externalizing spectrum.
        Annu Rev Clin Psychol. 2017; 13: 343-368
        • Ahmad S.I.
        • Hinshaw S.P.
        Attention-deficit/hyperactivity disorder, trait impulsivity, and externalizing behavior in a longitudinal sample.
        J Abnorm Child Psychol. 2017; 45: 1077-1089
        • GBD 2016 Alcohol Collaborators
        Alcohol use and burden for 195 countries and territories, 1990–2016: A systematic analysis for the Global Burden of Disease Study 2016.
        Lancet. 2018; 392: 1015-1035
        • Bartels M.
        • Hendriks A.
        • Mauri M.
        • Krapohl E.
        • Whipp A.
        • Bolhuis K.
        • et al.
        Childhood aggression and the co-occurrence of behavioural and emotional problems: Results across ages 3–16 years from multiple raters in six cohorts in the EU-ACTION project.
        Eur Child Adolesc Psychiatry. 2018; 27: 1105-1121
        • Mullins-Sweatt S.N.
        • DeShong H.L.
        • Lengel G.J.
        • Helle A.C.
        • Krueger R.F.
        Disinhibition as a unifying construct in understanding how personality dispositions undergird psychopathology.
        J Res Pers. 2019; 80: 55-61
        • Jang K.L.
        • Vernon P.A.
        • Livesley W.J.
        Personality disorder traits, family environment, and alcohol misuse: A multivariate behavioural genetic analysis.
        Addiction. 2000; 95: 873-888
        • Krueger R.F.
        • Hicks B.M.
        • Patrick C.J.
        • Carlson S.R.
        • Iacono W.G.
        • McGue M.
        Etiologic connections among substance dependence, antisocial behavior, and personality: Modeling the externalizing spectrum.
        J Abnorm Psychol. 2002; 111: 411-424
        • Venables N.C.
        • Hicks B.M.
        • Yancey J.R.
        • Kramer M.D.
        • Nelson L.D.
        • Strickland C.M.
        • et al.
        Evidence of a prominent genetic basis for associations between psychoneurometric traits and common mental disorders.
        Int J Psychophysiol. 2017; 115: 4-12
        • Young S.E.
        • Friedman N.P.
        • Miyake A.
        • Willcutt E.G.
        • Corley R.P.
        • Haberstick B.C.
        • Hewitt J.K.
        Behavioral disinhibition: Liability for externalizing spectrum disorders and its genetic and environmental relation to response inhibition across adolescence.
        J Abnorm Psychol. 2009; 118: 117-130
        • Noordermeer S.D.S.
        • Luman M.
        • Oosterlaan J.
        A systematic review and meta-analysis of neuroimaging in oppositional defiant disorder (ODD) and conduct disorder (CD) taking attention-deficit hyperactivity disorder (ADHD) into account.
        Neuropsychol Rev. 2016; 26: 44-72
        • Yang Y.
        • Raine A.
        Prefrontal structural and functional brain imaging findings in antisocial, violent, and psychopathic individuals: A meta-analysis.
        Psychiatry Res. 2009; 174: 81-88
        • Ameis S.H.
        • Ducharme S.
        • Albaugh M.D.
        • Hudziak J.J.
        • Botteron K.N.
        • Lepage C.
        • et al.
        Cortical thickness, cortico-amygdalar networks, and externalizing behaviors in healthy children.
        Biol Psychiatry. 2014; 75: 65-72
        • Treur J.L.
        • Demontis D.
        • Smith G.D.
        • Sallis H.
        • Richardson T.G.
        • Wiers R.W.
        • et al.
        Investigating causality between liability to ADHD and substance use, and liability to substance use and ADHD risk, using Mendelian randomization.
        Addict Biol. 2021; 26e12849
        • Pasman J.A.
        • Verweij K.J.H.
        • Gerring Z.
        • Stringer S.
        • Sanchez-Roige S.
        • Treur J.L.
        • et al.
        GWAS of lifetime cannabis use reveals new risk loci, genetic overlap with psychiatric traits, and a causal influence of schizophrenia.
        Nat Neurosci. 2018; 21: 1161-1170
        • Tielbeek J.J.
        • Johansson A.
        • Polderman T.J.C.
        • Rautiainen M.R.
        • Jansen P.
        • Taylor M.
        • et al.
        Genome-wide association studies of a broad spectrum of antisocial behavior.
        JAMA Psychiatry. 2017; 74: 1242-1250
        • Pappa I.
        • St Pourcain B.
        • Benke K.
        • Cavadino A.
        • Hakulinen C.
        • Nivard M.G.
        • et al.
        A genome-wide approach to children’s aggressive behavior: The EAGLE consortium.
        Am J Med Genet B Neuropsychiatr Genet. 2016; 171: 562-572
        • Sudlow C.
        • Gallacher J.
        • Allen N.
        • Beral V.
        • Burton P.
        • Danesh J.
        • et al.
        UK Biobank: An open access resource for identifying the causes of a wide range of complex diseases of middle and old age.
        PLoS Med. 2015; 12e1001779
        • Demontis D.
        • Walters R.K.
        • Martin J.
        • Mattheisen M.
        • Als T.D.
        • Agerbo E.
        • et al.
        Discovery of the first genome-wide significant risk loci for attention deficit/hyperactivity disorder.
        Nat Genet. 2019; 51: 63-75
        • Karlsson Linnér R.
        • Biroli P.
        • Kong E.
        • Meddens S.F.W.
        • Wedow R.
        • Fontana M.A.
        • et al.
        Genome-wide association analyses of risk tolerance and risky behaviors in over 1 million individuals identify hundreds of loci and shared genetic influences.
        Nat Genet. 2019; 51: 245-257
        • Baselmans B.M.L.
        • Jansen R.
        • Ip H.F.
        • van Dongen J.
        • Abdellaoui A.
        • van de Weijer M.P.
        • et al.
        Multivariate genome-wide analyses of the well-being spectrum.
        Nat Genet. 2019; 51: 445-451
        • Gravel S.
        Population genetics models of local ancestry.
        Genetics. 2012; 191: 607-619
        • Bulik-Sullivan B.K.
        • Loh P.R.
        • Finucane H.K.
        • Ripke S.
        • Yang J.
        • Schizophrenia Working Group of the Psychiatric Genomics Consortium
        • et al.
        LD Score regression distinguishes confounding from polygenicity in genome-wide association studies.
        Nat Genet. 2015; 47: 291-295
        • Bulik-Sullivan B.
        • Finucane H.K.
        • Anttila V.
        • Gusev A.
        • Day F.R.
        • Loh P.R.
        • et al.
        An atlas of genetic correlations across human diseases and traits.
        Nat Genet. 2015; 47: 1236-1241
        • Ward Jr., J.H.
        Hierarchical grouping to optimize an objective function.
        J Am Stat Assoc. 1963; 58: 236-244
        • Caliński T.
        • Harabasz J.
        A dendrite method for cluster analysis.
        Commun Stat. 1974; 3: 1-27
        • Grotzinger A.D.
        • Rhemtulla M.
        • de Vlaming R.
        • Ritchie S.J.
        • Mallard T.T.
        • Hill W.D.
        • et al.
        Genomic structural equation modelling provides insights into the multivariate genetic architecture of complex traits.
        Nat Hum Behav. 2019; 3: 513-525
        • de Leeuw C.A.
        • Mooij J.M.
        • Heskes T.
        • Posthuma D.
        MAGMA: Generalized gene-set analysis of GWAS data.
        PLoS Comput Biol. 2015; 11e1004219
        • GTEx Consortium
        Human genomics. The Genotype-Tissue Expression (GTEx) pilot analysis: Multitissue gene regulation in humans.
        Science. 2015; 348: 648-660
        • Hawrylycz M.J.
        • Lein E.S.
        • Guillozet-Bongaarts A.L.
        • Shen E.H.
        • Ng L.
        • Miller J.A.
        • et al.
        An anatomically comprehensive atlas of the adult human brain transcriptome.
        Nature. 2012; 489: 391-399
        • Habib N.
        • Avraham-Davidi I.
        • Basu A.
        • Burks T.
        • Shekhar K.
        • Hofree M.
        • et al.
        Massively parallel single-nucleus RNA-seq with DroNc-seq.
        Nat Methods. 2017; 14: 955-958
        • Liu M.
        • Jiang Y.
        • Wedow R.
        • Li Y.
        • Brazel D.M.
        • Chen F.
        • et al.
        Association studies of up to 1.2 million individuals yield new insights into the genetic etiology of tobacco and alcohol use.
        Nat Genet. 2019; 51: 237-244
        • Walters R.K.
        • Polimanti R.
        • Johnson E.C.
        • McClintick J.N.
        • Adams M.J.
        • Adkins A.E.
        • et al.
        Transancestral GWAS of alcohol dependence reveals common genetic underpinnings with psychiatric disorders.
        Nat Neurosci. 2018; 21: 1656-1669
        • American Psychiatric Association
        Diagnostic and Statistical Manual of Mental Disorders.
        Fifth Edition. American Psychiatric Association Publishing, Washington, DC2013
        • Kunwar A.
        • Dewan M.
        • Faraone S.V.
        Treating common psychiatric disorders associated with attention-deficit/hyperactivity disorder.
        Expert Opin Pharmacother. 2007; 8: 555-562
        • Dick D.M.
        • Viken R.J.
        • Kaprio J.
        • Pulkkinen L.
        • Rose R.J.
        Understanding the covariation among childhood externalizing symptoms: Genetic and environmental influences on conduct disorder, attention deficit hyperactivity disorder, and oppositional defiant disorder symptoms.
        J Abnorm Child Psychol. 2005; 33: 219-229
        • Karlsson Linnér R.
        • Mallard T.T.
        • Barr P.B.
        • Sanchez-Roige S.
        • Madole J.W.
        • Driver M.N.
        • et al.
        Multivariate genomic analysis of 1.5 million people identifies genetic associations with traits related to self-regulation and addiction.
        Nat Neurosci. 2021; 24: 1367-1376
        • Mishra S.
        • Lalumiére M.L.
        Risk-taking, antisocial behavior, and life histories.
        in: Duntley J.D. Shackelford T.K. Evolutionary Forensic Psychology: Darwinian Foundations of Crime and Law. Oxford University Press, New York, NY2008
        • Krueger R.F.
        • Hobbs K.A.
        • Conway C.C.
        • Dick D.M.
        • Dretsch M.N.
        • Eaton N.R.
        • et al.
        Validity and utility of Hierarchical Taxonomy of Psychopathology (HiTOP): II. Externalizing superspectrum.
        World Psychiatry. 2021; 20: 171-193
        • Nigg J.T.
        • John O.P.
        • Blaskey L.G.
        • Huang-Pollock C.L.
        • Willcutt E.G.
        • Hinshaw S.P.
        • Pennington B.
        Big five dimensions and ADHD symptoms: Links between personality traits and clinical symptoms.
        J Pers Soc Psychol. 2002; 83: 451-469
        • McGhee R.L.
        • Ehrler D.J.
        • Buckhalt J.A.
        • Phillips C.
        The relation between five-factor personality traits and risk-taking behavior in preadolescents.
        Psychology. 2012; 3: 558-561
        • Barker E.D.
        • Ing A.
        • Biondo F.
        • Jia T.
        • Pingault J.B.
        • Du Rietz E.
        • et al.
        Do ADHD-impulsivity and BMI have shared polygenic and neural correlates?.
        Mol Psychiatry. 2021; 26: 1019-1028
        • Loth A.K.
        • Drabick D.A.G.
        • Leibenluft E.
        • Hulvershorn L.A.
        Do childhood externalizing disorders predict adult depression? A meta-analysis [published correction appears in J Abnorm Child Psychol 2014; 42:1115–1116].
        J Abnorm Child Psychol. 2014; 42: 1103-1113
        • Pailing A.N.
        • Reniers R.L.E.P.
        Depressive and socially anxious symptoms, psychosocial maturity, and risk perception: Associations with risk-taking behaviour.
        PLoS One. 2018; 13e0202423
        • McCloughen A.
        • Foster K.
        • Marabong N.
        • Miu D.
        • Fethney J.
        Physical health risk behaviours in young people with mental illness.
        Issues Ment Health Nurs. 2015; 36: 781-790
        • Derks I.P.M.
        • Bolhuis K.
        • Yalcin Z.
        • Gaillard R.
        • Hillegers M.H.J.
        • Larsson H.
        • et al.
        Testing bidirectional associations between childhood aggression and BMI: Results from three cohorts.
        Obesity (Silver Spring). 2019; 27: 822-829
        • Fluharty M.E.
        • Sallis H.
        • Munafò M.R.
        Investigating possible causal effects of externalizing behaviors on tobacco initiation: A Mendelian randomization analysis.
        Drug Alcohol Depend. 2018; 191: 338-342
        • Aleyasin H.
        • Flanigan M.E.
        • Russo S.J.
        Neurocircuitry of aggression and aggression seeking behavior: Nose poking into brain circuitry controlling aggression.
        Curr Opin Neurobiol. 2018; 49: 184-191
        • Tobiansky D.J.
        • Roma P.G.
        • Hattori T.
        • Will R.G.
        • Nutsch V.L.
        • Dominguez J.M.
        The medial preoptic area modulates cocaine-induced activity in female rats.
        Behav Neurosci. 2013; 127: 293-302
        • Phelps E.A.
        • LeDoux J.E.
        Contributions of the amygdala to emotion processing: From animal models to human behavior.
        Neuron. 2005; 48: 175-187
        • Davidson R.J.
        • Putnam K.M.
        • Larson C.L.
        Dysfunction in the neural circuitry of emotion regulation—A possible prelude to violence.
        Science. 2000; 289: 591-594
        • O’Brien J.W.
        • Hill S.Y.
        Neural predictors of substance use disorders in young adulthood.
        Psychiatry Res Neuroimaging. 2017; 268: 22-26
        • Tajima-Pozo K.
        • Ruiz-Manrique G.
        • Yus M.
        • Arrazola J.
        • Montañes-Rada F.
        Correlation between amygdala volume and impulsivity in adults with attention-deficit hyperactivity disorder.
        Acta Neuropsychiatr. 2015; 27: 362-367
        • Van Dessel J.
        • Sonuga-Barke E.
        • Moerkerke M.
        • Van der Oord S.
        • Lemiere J.
        • Morsink S.
        • Danckaerts M.
        The amygdala in adolescents with attention-deficit/hyperactivity disorder: Structural and functional correlates of delay aversion.
        World J Biol Psychiatry. 2020; 21: 673-684
        • Hoogman M.
        • Bralten J.
        • Hibar D.P.
        • Mennes M.
        • Zwiers M.P.
        • Schweren L.S.J.
        • et al.
        Subcortical brain volume differences in participants with attention deficit hyperactivity disorder in children and adults: A cross-sectional mega-analysis [published correction appears in Lancet Psychiatry 2017; 4:436].
        Lancet Psychiatry. 2017; 4: 310-319
        • Hu R.K.
        • Zuo Y.
        • Ly T.
        • Wang J.
        • Meera P.
        • Wu Y.E.
        • Hong W.
        An amygdala-to-hypothalamus circuit for social reward.
        Nat Neurosci. 2021; 24: 831-842
        • Moulton E.A.
        • Elman I.
        • Becerra L.R.
        • Goldstein R.Z.
        • Borsook D.
        The cerebellum and addiction: Insights gained from neuroimaging research.
        Addict Biol. 2014; 19: 317-331
        • Shaw P.
        • Ishii-Takahashi A.
        • Park M.T.
        • Devenyi G.A.
        • Zibman C.
        • Kasparek S.
        • et al.
        A multicohort, longitudinal study of cerebellar development in attention deficit hyperactivity disorder.
        J Child Psychol Psychiatry. 2018; 59: 1114-1123
        • Caggiano V.
        • Leiras R.
        • Goñi-Erro H.
        • Masini D.
        • Bellardita C.
        • Bouvier J.
        • et al.
        Midbrain circuits that set locomotor speed and gait selection.
        Nature. 2018; 553: 455-460
        • Silva B.A.
        • Gross C.T.
        • Gräff J.
        The neural circuits of innate fear: Detection, integration, action, and memorization.
        Learn Mem. 2016; 23: 544-555
        • Korte S.M.
        • Jaarsma D.
        • Luiten P.G.
        • Bohus B.
        Mesencephalic cuneiform nucleus and its ascending and descending projections serve stress-related cardiovascular responses in the rat.
        J Auton Nerv Syst. 1992; 41: 157-176
        • Goll Y.
        • Atlan G.
        • Citri A.
        Attention: The claustrum.
        Trends Neurosci. 2015; 38: 486-495
        • Atlan G.
        • Terem A.
        • Peretz-Rivlin N.
        • Sehrawat K.
        • Gonzales B.J.
        • Pozner G.
        • et al.
        The claustrum supports resilience to distraction.
        Curr Biol. 2018; 28: 2752-2762.e7
        • Dickstein S.G.
        • Bannon K.
        • Castellanos F.X.
        • Milham M.P.
        The neural correlates of attention deficit hyperactivity disorder: An ALE meta-analysis.
        J Child Psychol Psychiatry. 2006; 47: 1051-1062
        • Cross-Disorder Group of the Psychiatric Genomics Consortium
        Genomic relationships, novel loci, and pleiotropic mechanisms across eight psychiatric disorders.
        Cell. 2019; 179: 1469-1482.e11
        • Wray N.R.
        • Ripke S.
        • Mattheisen M.
        • Trzaskowski M.
        • Byrne E.M.
        • Abdellaoui A.
        • et al.
        Genome-wide association analyses identify 44 risk variants and refine the genetic architecture of major depression.
        Nat Genet. 2018; 50: 668-681
        • Johnson E.C.
        • Demontis D.
        • Thorgeirsson T.E.
        • Walters R.K.
        • Polimanti R.
        • Hatoum A.S.
        • et al.
        A large-scale genome-wide association study meta-analysis of cannabis use disorder.
        Lancet Psychiatry. 2020; 7: 1032-1045
        • Kranzler H.R.
        • Zhou H.
        • Kember R.L.
        • Vickers Smith R.
        • Justice A.C.
        • Damrauer S.
        • et al.
        Genome-wide association study of alcohol consumption and use disorder in 274,424 individuals from multiple populations [published correction appears in Nat Commun 2019; 10:2275 and Nat Commun 2019; 10:4050].
        Nat Commun. 2019; 10: 1499