by Joseph T. Coyle, M.D.
Professor of Psychiatry, Harvard Medical School
As a neuroscientist, I would have thought that the rules of genetics were “settled law” and that dissecting the genetics of psychiatric disorders was simply a matter of investing the time and effort. To the contrary, this year has provided a metaphorical splash of cold water in my face, reminding me that science is never as simple as it seems. Specific noteworthy advances include the findings that methods used successfully to identify genes responsible for many heritable disorders were shown not to be applicable for identifying genes responsible for psychiatric disorders, that new mutations may be a frequent cause of psychiatric disorders and that early life experiences can alter gene function throughout life in ways that have substantial behavioral consequences.
Genetics and Brain Disorders
Many rare brain disorders exhibit “Mendelian” inheritance, so named after the monk Gregor Mendel, who first described dominant and recessive patterns of inheritance in the colors of pea flowers. For example, Tay-Sachs disease is inherited via an autosomal recessive gene, meaning that a sibling has a 25 percent risk of also having the disease. Familial amytrophic lateral sclerosis (ALS) is inherited via an autosomal dominant gene, meaning that a sibling or offspring has a 50 percent risk. In contrast, the heritability of psychiatric disorders is much less clear-cut.
Although by the late nineteenth century, psychiatrists were aware that psychotic disorders appeared to occur more frequently in families where a relative was affected, the study of the genetics of psychiatric disorders was neglected for nearly a century. This situation developed because the concept of heritability of psychiatric disorders fell out of favor as the popularity of psychoanalytic theory rose in American and then European psychiatry. The fact that serious mental illness seemed to be concentrated in certain families was entirely consistent with psychoanalytic theory, in which the origin of mental illness was thought to be based on adverse early (parenting) life experiences.
Then, fifty years ago, researchers rigorously examined the heritability of schizophrenia by exploiting the fact that identical twins have essentially the same genomes, whereas fraternal twins do not. The relative contribution of environment versus genetics can be inferred from how far the rate of concordance—that is, the presence of the same trait or disorder in twin siblings—deviates from that predicted if the disorder were completely based on genetics (i.e., 100 percent concordance for identical twins if one is affected). The evidence showed a concordance rate of approximately 50 percent in identical twins when one was affected with schizophrenia. The concordance rate was about 10 percent in fraternal twins, who share half their genes. Still, psychiatry resisted the obvious inference of the heritability of schizophrenia because the finding could be explained by the assumption that identical twins would be treated much more alike by their mother than fraternal twins, thereby increasing “concordance.”
To address these reservations, Seymour Kety and his colleagues carried out a heroic study nearly forty years ago assessing the risk for schizophrenia in Scandinavian adopted offspring: the adult psychiatric status of children who had one parent suffering from schizophrenia was compared with the risk for adopted children whose biological parents were free of mental illness but whose adoptive family had one parent who had been diagnosed with schizophrenia. This human “cross-fostering” experiment, possible only because of the detailed and lengthy records maintained by the Swedish Health Ministry, demonstrated that the risk for schizophrenia was significantly higher in the offspring with a schizophrenic biological parent. Children adopted into families with a schizophrenic adoptive parent showed no greater incidence of schizophrenia than the general population. This study did much to undermine the belief that schizophrenia was merely a psychologic maladaptation. Rather, it supported the notion that the disorder involved substantial genetic, i.e., inheritable, risk factors.
As the twin paradigm became more accepted, it was quite helpful in clarifying the role of genes in the risk for autism. Leo Kanner first described autism in 1943, when he also noted that the parents of some afflicted children appeared to be aloof and rigid. Psychoanalytic theorists identified the “refrigerator mother” as the cause of autism in the children. However, twin studies demonstrated a remarkably high heritability, with 90 percent concordance of autism in identical twins when one was affected, but only 4 percent concordance in fraternal twins. Here, too, the results of twin studies eliminated maternal-infant interactions as responsible for a severe psychiatric disorder.
Components of the autistic phenotype (observable characteristics) can be seen more frequently in the parents and sibs (first-degree relatives) of individuals with autism than in the general population. These component symptoms, called endophenotypes, are thought to reflect the impact of shared risk genes. For example, deficits in social language similar to those occurring in autism—although much less severe—are found more frequently in parents with an autistic child than in parents of children without autism. The high concordance rate in identical twins coupled with the rapid falloff in risk for first-degree relatives in autism (4 percent) and schizophrenia (10 percent) reveal a non-Mendelian pattern of genetic transmission (i.e., inconsistent with autosomal dominant, recessive or sex-linked genes). Thus, by the turn of this century, the genetics of psychiatric disorders began to look more like those of other commonmedical disorders, such as hypertension and diabetes mellitus, than like rare neurologic disorders. These common medical disorders involve complex genetics in which multiple risk genes of moderate effect interact with the environment to cause the disorder. Risk genes increase the likelihood of developing the disorder but do not cause it in the absence of other risk factors.
Candidate Genes Versus Whole Genome-wide Association Studies
The search for genes that cause psychiatric disorders began in earnest more than a decade ago. Researchers took advantage of their knowledge of the biochemical and cellular abnormalities that had been identified in the postmortem studies of the brains of individuals who died with specific psychiatric disorders. Plausible pathways of pathology in schizophrenia, for example, include abnormal neural development and aberrant neurotransmission involving dopamine, GABA and glutamate. Genes encoding proteins involved in these pathologic processes were considered “candidate genes,” which is to say, genes likely to be mutant in the disorder. As precedent, the candidate gene strategy was the basis for the successful identification of risk genes for Alzheimer’s disease because of their role in the disposition of the pathogenic peptide, beta-amyloid.
Scientists identified single nucleotide polymorphisms (SNPs)—places in the DNA where individuals differ with regard to a given nucleic acid—in potential candidate genes and looked for preferential transmission of those SNPs within families that had members affected by the disorder. These studies were conducted with dozens, then hundreds of participants. Well over tenty candidate genes were implicated by this strategy. However, a disconcerting lack of reproducibility among studies became apparent.
Given the estimates of the attributable genetic risk suggested by these positive results (much less than 25 percent), geneticists argued that the studies were statistically underpowered (i.e., had an insufficient number of subjects) to make legitimate comparisons and would therefore predictably yield false positive and false negative results. On the basis of studies of diabetes, another disorder of complex genetics, it was estimated that the number of affected subjects and controls by necessity would be in the range of several thousands to generate statistically meaningful results.
To counter the biases inherent in candidate gene studies, scientists proposed the genome-wide association study (GWAS), a method that scans the entire genome in an unbiased manner to identify any SNPs associated with the disorder—not just those located in potential candidate genes.1 Recently, GWAS studies have been published on schizophrenia and on autism that have used thousands of subjects.2,3 Results of the GWAS studies on schizophrenia were mixed. Despite the large number of subjects, researchers could confirm only a few of the putative risk genes identified during the last decade. However, the GWAS studies did identify new risk genes, and some of these seemed consistent with current hypotheses about the pathophysiology of schizophrenia. Two genes could affect brain development, and two other genes could affect the function of the neurotransmitter glutamate.
The limited success of these endeavors has convinced scientists that substantive advances in the genetics of psychiatric disorders will require considerable cooperation among investigators to assemble the large number of genotyped subjects required. To that end, diseasespecific data repositories have been developed with a compilation of genetic findings, such as that provided by www.szgene.org.4 In this way, it will be possible to accumulate genotypes on tens of thousands of subjects to achieve the requisite statistical rigor. Perhaps as the number of subjects increases, previously identified candidate genes will be confirmed with expanded GWAS.
Another source of concern about the candidate gene approach for psychiatric disorders comes from the recent discovery of a high concentration of copy number variations (CNVs) in the genomes of patients with psychiatric disorders. A CNV results from either deletions or duplications of stretches of DNA, which involve several genes. It has long been known that a rare, small deletion on chromosome 22q11, which causes velocardiofacial syndrome—a condition resulting in abnormal development of the parathyroid gland, thymus and heart—is associated with an increased risk for psychosis. Chromosome 15g11-13 deletions are associated with the developmental disorder Prader-Willi Syndrome; and genes such as neurexin 1, which are located within this deleted region, have been associated with risk for autism spectrum disorders (ASD). Given these findings, Jonathan Sebat and colleagues at the Cold Spring Harbor Laboratory wondered whether spontaneous CNVs in the genome might account for “sporadic” autism, a form in which no other family members are affected and thus one that is less likely to have a genetic basis.5 The researchers found that CNVs occurred in 10 percent of the sporadic autistic subjects, 3 percent of autistic subjects with an affected first-degree relative and only 1 percent of controls. A recent study involving nearly 2,500 autistic patients and controls supported these findings.6 Studies in individuals with schizophrenia have demonstrated that CNVs are three times more likely in those with the typical onset in their late teens or twenties than in people who do not have the disorder, but four times more likely in patients who have early-onset schizophrenia, which develops in young adolescence and sometimes sooner. These results suggest that new mutations are a much more significant factor in the causes of schizophrenia and autism than previously thought.
The French naturalist Lamarck proposed that acquired traits such as the long neck that a giraffe develops from stretching for leaves in tall trees could be transferred to its offspring. This theory was discarded with the ascendance of Darwin’s theory of natural selection. However, recent research is showing that the genome can be modified by environmental events, thereby altering gene expression. This process, known as epigenesis, refers to alterations in gene functioncaused by mechanisms other than changes in the gene’s DNA sequence. Epigenetic changes of a gene can result in persistent alterations in its expression and thereby the physical traits, or phenotype, associated with that gene. These epigenetic changes in the gene can even be inherited by future generations. Recent advances suggest that epigenetic mechanisms may have robust effects on the manifestation of psychiatric disorders.
An intriguing and compelling story has emerged on how epigenetics may mediate the persistent effects of adverse experiences early in life. From biblical times through the elaboration of psychoanalytical theory by Sigmund Freud, it has been recognized that abuse and neglect in early childhood are associated with an increased risk of anxiety disorders, depression and even suicide in adulthood. This vulnerability has been linked to a dysregulation of the hypothalamic-pituitary-adrenal axis (HPA), which mediates the body’s stress response. The hypothalamus, pituitary and adrenal glands orchestrate the “fight or flight” response to acute stress. In normal individuals, they release the stress hormones the body needs to face or flee the threat and then abate the response when the danger is past. Key to this process are glucocorticoid receptors in the brain. These monitor levels of the stress hormone corticosterone and enable the signals that turn the stress response up or down. Persistently high levels of stress, particularly during infancy and childhood, disrupt this homeostatic feedback loop such that the individual becomes hyper-responsive to a given level of stress, with excessive corticosterone secretion. In fact, this dysregulation appears to be a major contributing factor to the high risk for depression in adults who experienced abuse and neglect in childhood.
Michael Meaney and his colleagues at McGill University have been exploring the molecular mechanisms that underpin persistent dysregulation of the HPA axis due to stress early in development. Mother rats frequently groom and lick their pups during the first week after birth. Pups of mothers who engage in these behaviors rarely grow up to be hyper-responsive to stress. Meaney and his colleagues showed that the licking and grooming stimuli increased serotonin release in the hippocampus of the pups. Serotonin stimulates the expression of a transcription factor known as NGFI-A that binds to the promoter region of the gene for the glucocorticoid receptor (Nr3c1), thereby causing more receptors to be made. They found that the promoter region of the glucocorticoid receptor gene of adult rats that had been infrequently groomed and licked as pups was blocked by methylation, a chemical change of the nucleic acid cytosine. The blocked promoter was no longer responsive to NGFI-A, so fewer glucocorticoid receptors were made, causing persistent hyper-responsiveness to stress.
Meaney and his colleagues wondered whether similar results of early-life neglect played out in the human brain.7 To this end, they studied the status of the glucocorticoid receptor gene in the hippocampus of twelve individuals who committed suicide as adults and who had a history of serious child abuse, twelve who committed suicide without evidence of child abuse and twelve non-psychiatric controls who died suddenly. The expression of the glucocorticoid receptor was significantly reduced in suicides with the history of child abuse as compared with the two types of controls. They examined the glucocorticoid gene (Nr3c1) in the human homolog of the promoter region implicated in the rats and found that in the suicides with a history of child abuse there was significantly greater methylation and reduced expression of the glucocorticoid receptor, as compared to the two types of controls. Thus, the findings from a rodent model were successfully extended to humans. How the chemical changes of that particular brain region might lead to suicide, however, remains a subject of further study.
Recent research has identified another novel mechanism whereby small RNAs known as microRNAs (or miRNAs) regulate the synthesis of proteins by affecting mRNAs, which carry the blueprint from the DNA to the protein building site.8 miRNAs were first discovered fifteen years ago in the nematode c. elegans. The DNA sequences encoding these miRNAs are located in regions of the genome that do not contain genes and were therefore thought to serve no function. Their role in mammalian gene expression is just now emerging, with more than half the scientific reports on the subject appearing within the last eighteen months. The miRNA binds to a complementary site on a targeted mRNA, thereby blocking the mRNA’s ability to make protein or accelerating its degradation. In this way, the miRNA regulates the expression of the gene’s protein product. More than one thousand distinct miRNAs have been identified, half of which are expressed primarily or exclusively in the brain.
Because individual miRNAs may have hundreds of potential mRNA targets, it is hypothesized that a specific miRNA might be involved in coordinated regulation of protein expression in functional networks such as those regulating brain development or synaptic plasticity. Because of the abundance of miRNA genes, it is now appreciated that genetic “hot spots” such as 22ql1 for schizophrenia might be linked to miRNA genes and not just protein-coding genes. John Rubenstein, a child psychiatrist at the University of California, San Francisco, has noted that the chromosomal 8p region, which contains a number of genes associated with risk for psychiatric disorders, also encodes at least 8 miRNAs. Thus, non-coding regions of the genome, until recently thought to be functionally unimportant, actually harbor miRNA genes that can have substantial impact on brain function and disease processes.
What can we take away from these remarkable advances in our understanding of the molecular genetics of psychiatric disorders? First, it is apparent that the established research strategies for studying Mendelian disorders are simply inadequate for the complex genetics of psychiatric disorders such as schizophrenia, autism and depression. Individual laboratories working with a few hundred subjects are unlikely to succeed in identifying risk genes for these illnesses. Rather, we need large consortia that are able to collect thousands of subjects and study their genetics. Open databases must serve as repositories for the genotypes for specific disorders and should provide the most current summary of risk genes and CNVs.
Second, researchers must pay greater attention to the influence of epigenetics on behavior. Meaney’s corpus of research on the glucocorticoid receptor and anxiety is a wonderful example of how focusing on a particular behavioral phenotype can ultimately be highly informative. Nevertheless, his research addresses only one region of a single gene. The possibilities of environmental-gene interactions combined with gene-gene interactions that impact behavior seem immense.
Given these genetic advances, we have much to look forward to in the coming years. But no one knows what additional surprises lurk in the human genome, like the recently described CNVs and miRNAs, that will further disabuse us about simplistic hypotheses as to the causes of psychiatric disorders.