The promise of behavior genetics in the genomic era was that the molecular basis of the major psychiatric disorders could be identified in the absence of a detailed understanding of the biology of the brain. That is, the logic of genome-wide association studies (GWAS) is hypothesis free, requiring only that the target disorder display substantial heritability for statistical associations between specific polymorphisms and the target phenotype to be identified. This was a powerful promise in the face of the vast complexity of the brain and our rudimentary understanding of how molecules lead to mind. Moreover, genetic risk factors, once identified, would be powerful translational tools providing an easy manner of assessing an individual's risk for psychiatric disease and perhaps allowing for rational targeting of treatments to the individuals most likely to respond based on their genetic profile.
Unfortunately, now that large-scale, statistically powerful GWAS have been conducted for all of the major psychiatric disorders, it has become clear that common, large-effect genetic risk factors for behavioral disorders do not exist. Despite heritability estimates of 40–90% for psychiatric disorders including depression, autism, schizophrenia, and bipolar disorder (Burmeister et al. Reference Burmeister, McInnis and Zollner2008), only minor risk factors of minute effect size have been identified. Taken in aggregate, these risk factors only account for ~1% of the estimated genetic variance (Plomin & Davis Reference Plomin and Davis2009). The small effect size of identified risk factors and substantial missing heritability in behavior genetics have limited the value of this approach both in terms of the predictive value of genetic testing in clinical application and in understanding the molecular basis of behavior.
Charney describes a number of biological mechanisms by which the simple logic of behavior genetics is violated. The manner in which these factors violate the Mendelian assumptions of behavior genetics vary and have different implications for the design and success of behavior genetic studies. Phenomena such as somatic mutation, chimerism, and heteroplasmy of mitochondrial DNA all mean that monozygotic twins are more genetically discordant than has been assumed. This will lead to underestimates of the heritability of diseases or traits. In contrast, the systematic concordance of non-genetic factors in monozygotic versus dizygotic twins, such as prenatal stress in Charney's example, leads to exaggerated estimates of heritability. These deviations from Mendelian principles highlight that, rather than removing biology from behavior genetics, postgenomics requires that polymorphisms themselves become the focus of biological analysis. Statistics alone cannot bridge the gap from molecule to mind; instead, the biology of polymorphisms must be studied systematically in the context of non-genetic factors that might modify their biological properties to make meaningful and informative genotype-phenotype relationships.
How does a biologically focused analysis of a polymorphism work? First and foremost, it involves identifying the molecular effects of a polymorphism. This process is straightforward for polymorphisms that result in an amino acid substitution in the peptide sequence coded for by a gene. Such polymorphisms can be identified by applying the genetic code to the ancestral and variant codons, and the effects of the substitution on protein function can be assessed through in vitro assays. Taking the polymorphism out of the human context and studying it in molecular assays allow controlled experiments in which the alleles of the polymorphism are the only variable. Moreover, controlled experiments allow for the systematic study of physiologic factors that modify the effects of the polymorphism, and this information can be applied upwards into studies of more complex phenotypes. An example of this approach was applied to a common valine to methionine substitution in the human gene for brain-derived neurotrophic factor (BDNF). BDNF is a secreted signaling molecule that facilitates neuronal growth and synaptic plasticity (Autry & Monteggia Reference Autry and Monteggia2012). Expression of the recombinant ancestral and variant proteins in neuronal cultures revealed that the polymorphism altered activity-dependent but not constitutive secretion of BDNF (Chen et al. Reference Chen, Patel, Sant, Meng, Teng, Hempstead and Lee2004; Egan et al. Reference Egan, Kojima, Callicott, Goldberg, Kolachana, Bertolino, Zaitsev, Gold, Goldman, Dean, Lu and Weinberger2003). This result demonstrated that the polymorphism has a molecular phenotype and suggests specific domains of neuronal function that might be affected such as activity-dependent learning. This hypothesis was confirmed by creating a transgenic mouse in which the conserved ancestral valine was replaced with the variant methionine (Chen et al. Reference Chen, Jing, Bath, Ieraci, Khan, Siao, Herrera, Toth, Yang, McEwen, Hempstead and Lee2006). When the two strains were compared for their ability to extinguish a learned fear association, the methionine-expressing strain displayed an extinction-learning deficit (Soliman et al. Reference Soliman, Glatt, Bath, Levita, Jones, Pattwell, Jing, Tottenham, Amso, Somerville, Voss, Glover, Ballon, Liston, Teslovich, Van Kempen, Lee and Casey2010). This finding motivated the collection and assessment of human subjects for extinction learning in a parallel paradigm as a function of the BDNF polymorphism. Humans with the methionine allele of the BDNF polymorphism also displayed less efficient extinction learning (Soliman et al. Reference Soliman, Glatt, Bath, Levita, Jones, Pattwell, Jing, Tottenham, Amso, Somerville, Voss, Glover, Ballon, Liston, Teslovich, Van Kempen, Lee and Casey2010). This finding has clear implications for the genetic basis of anxiety, but anxiety is so imprecisely measured in humans that further characterization of the BDNF polymorphism is needed to meaningfully refine the anxiety phenotype. These refinements can be modeled in the transgenic mice and then applied to human studies. Ultimately, the goal of this approach is to identify biologically defined subtypes of behavioral disorders in which the effect size of individual polymorphisms are large and reliably assessed. At that point the polymorphism might be of practical predictive value as a genetic test.
There are many caveats to this biological approach. It is far from guaranteed that model systems will adequately recapitulate the complex biology of human brain. As Charney points out in the target article, abundant forms of genetic variability are unique to primates and even humans. Biological characterization of polymorphisms is also an extremely labor-intensive process that cannot be applied in a genome-wide manner, and thus choices must be made about which polymorphisms to study. One approach to integrating biometric and biomolecular genetics, to use Charney's terms, would be to perform biological characterization of variants that show greatest association in GWAS and then use the characterization of the variants' biology to refine target phenotypes for subsequent GWAS to increase the effect size of the variants to the point that they could be useful in clinical practice.
You have
Access
The promise of behavior genetics in the genomic era was that the molecular basis of the major psychiatric disorders could be identified in the absence of a detailed understanding of the biology of the brain. That is, the logic of genome-wide association studies (GWAS) is hypothesis free, requiring only that the target disorder display substantial heritability for statistical associations between specific polymorphisms and the target phenotype to be identified. This was a powerful promise in the face of the vast complexity of the brain and our rudimentary understanding of how molecules lead to mind. Moreover, genetic risk factors, once identified, would be powerful translational tools providing an easy manner of assessing an individual's risk for psychiatric disease and perhaps allowing for rational targeting of treatments to the individuals most likely to respond based on their genetic profile.
Unfortunately, now that large-scale, statistically powerful GWAS have been conducted for all of the major psychiatric disorders, it has become clear that common, large-effect genetic risk factors for behavioral disorders do not exist. Despite heritability estimates of 40–90% for psychiatric disorders including depression, autism, schizophrenia, and bipolar disorder (Burmeister et al. Reference Burmeister, McInnis and Zollner2008), only minor risk factors of minute effect size have been identified. Taken in aggregate, these risk factors only account for ~1% of the estimated genetic variance (Plomin & Davis Reference Plomin and Davis2009). The small effect size of identified risk factors and substantial missing heritability in behavior genetics have limited the value of this approach both in terms of the predictive value of genetic testing in clinical application and in understanding the molecular basis of behavior.
Charney describes a number of biological mechanisms by which the simple logic of behavior genetics is violated. The manner in which these factors violate the Mendelian assumptions of behavior genetics vary and have different implications for the design and success of behavior genetic studies. Phenomena such as somatic mutation, chimerism, and heteroplasmy of mitochondrial DNA all mean that monozygotic twins are more genetically discordant than has been assumed. This will lead to underestimates of the heritability of diseases or traits. In contrast, the systematic concordance of non-genetic factors in monozygotic versus dizygotic twins, such as prenatal stress in Charney's example, leads to exaggerated estimates of heritability. These deviations from Mendelian principles highlight that, rather than removing biology from behavior genetics, postgenomics requires that polymorphisms themselves become the focus of biological analysis. Statistics alone cannot bridge the gap from molecule to mind; instead, the biology of polymorphisms must be studied systematically in the context of non-genetic factors that might modify their biological properties to make meaningful and informative genotype-phenotype relationships.
How does a biologically focused analysis of a polymorphism work? First and foremost, it involves identifying the molecular effects of a polymorphism. This process is straightforward for polymorphisms that result in an amino acid substitution in the peptide sequence coded for by a gene. Such polymorphisms can be identified by applying the genetic code to the ancestral and variant codons, and the effects of the substitution on protein function can be assessed through in vitro assays. Taking the polymorphism out of the human context and studying it in molecular assays allow controlled experiments in which the alleles of the polymorphism are the only variable. Moreover, controlled experiments allow for the systematic study of physiologic factors that modify the effects of the polymorphism, and this information can be applied upwards into studies of more complex phenotypes. An example of this approach was applied to a common valine to methionine substitution in the human gene for brain-derived neurotrophic factor (BDNF). BDNF is a secreted signaling molecule that facilitates neuronal growth and synaptic plasticity (Autry & Monteggia Reference Autry and Monteggia2012). Expression of the recombinant ancestral and variant proteins in neuronal cultures revealed that the polymorphism altered activity-dependent but not constitutive secretion of BDNF (Chen et al. Reference Chen, Patel, Sant, Meng, Teng, Hempstead and Lee2004; Egan et al. Reference Egan, Kojima, Callicott, Goldberg, Kolachana, Bertolino, Zaitsev, Gold, Goldman, Dean, Lu and Weinberger2003). This result demonstrated that the polymorphism has a molecular phenotype and suggests specific domains of neuronal function that might be affected such as activity-dependent learning. This hypothesis was confirmed by creating a transgenic mouse in which the conserved ancestral valine was replaced with the variant methionine (Chen et al. Reference Chen, Jing, Bath, Ieraci, Khan, Siao, Herrera, Toth, Yang, McEwen, Hempstead and Lee2006). When the two strains were compared for their ability to extinguish a learned fear association, the methionine-expressing strain displayed an extinction-learning deficit (Soliman et al. Reference Soliman, Glatt, Bath, Levita, Jones, Pattwell, Jing, Tottenham, Amso, Somerville, Voss, Glover, Ballon, Liston, Teslovich, Van Kempen, Lee and Casey2010). This finding motivated the collection and assessment of human subjects for extinction learning in a parallel paradigm as a function of the BDNF polymorphism. Humans with the methionine allele of the BDNF polymorphism also displayed less efficient extinction learning (Soliman et al. Reference Soliman, Glatt, Bath, Levita, Jones, Pattwell, Jing, Tottenham, Amso, Somerville, Voss, Glover, Ballon, Liston, Teslovich, Van Kempen, Lee and Casey2010). This finding has clear implications for the genetic basis of anxiety, but anxiety is so imprecisely measured in humans that further characterization of the BDNF polymorphism is needed to meaningfully refine the anxiety phenotype. These refinements can be modeled in the transgenic mice and then applied to human studies. Ultimately, the goal of this approach is to identify biologically defined subtypes of behavioral disorders in which the effect size of individual polymorphisms are large and reliably assessed. At that point the polymorphism might be of practical predictive value as a genetic test.
There are many caveats to this biological approach. It is far from guaranteed that model systems will adequately recapitulate the complex biology of human brain. As Charney points out in the target article, abundant forms of genetic variability are unique to primates and even humans. Biological characterization of polymorphisms is also an extremely labor-intensive process that cannot be applied in a genome-wide manner, and thus choices must be made about which polymorphisms to study. One approach to integrating biometric and biomolecular genetics, to use Charney's terms, would be to perform biological characterization of variants that show greatest association in GWAS and then use the characterization of the variants' biology to refine target phenotypes for subsequent GWAS to increase the effect size of the variants to the point that they could be useful in clinical practice.