|Additional related information may be found at:|
|Neuropsychopharmacology: The Fifth Generation of Progress|
Joel Gelernter and David Goldman
"May you live in interesting times" is an ancient Chinese curse. These are interesting times in the field of psychiatric genetics. We are presently experiencing the most important series of methodological breakthroughs in the short history of genetics research. The result in medical genetics overall has been dramatic progress in mapping of genes influencing complex traits such as diabetes and hypertension. The result in genetics of psychiatric phenotype has been more ambiguous, but in general, we have seen considerable progress in mapping genes influencing risk for the major psychiatric disorders.
This gene mapping progress is attributable to many factors. Any list of these factors would have to include the following: recognition that most psychiatric traits must be treated as complex from a genetic standpoint; general acceptance of standard diagnostic procedures; progress in the sequencing and genomic mapping of genes; widespread availability of ever-increasing numbers of genetic markers; advances in gene variant detection and genotyping technologies; innovation in statistical genetics and strategies for collecting datasets; and advances in our understanding of gene function and evolution.
In this chapter, our aim will be to review the meaning of each of these factors; their specific application in research in psychiatric genetics; some important recent findings in psychiatric genetics; and possible future directions in the post-Genome Project era.
The basic biological fact underlying psychiatric genetic research is the existence of genetic polymorphism which alters molecular function and, ultimately, influences behavior. The existence of DNA variation inherited from parents to offspring accounts for the heritability of traits, where heritability is usually measured as the proportion of the population variance in the trait which is attributable to additive genetic factors. DNA sequence variation can produce differences in both protein sequence and gene regulation, leading to interindividual differences and to differences between definable populations. Variation has been identified in the great majority of genes where it has been sought. Thus, although molecular biologists might study a particular gene, whatever gene is studied is likely to be a representative of a set of different versions of the gene present in a population (unless unlike the human the animal has been inbred, eliminating variation). The different versions of a gene in the population might represent normal and pathological, two equally functional versions, or anything in between. There are numerous cases in which nonfunctional versions of genes are known to be found in the human population, creating the potential for human "knockouts" for those loci in persons homozygous for nonfunctional alleles; for example, there are nonfunctional alleles of the D4 dopamine receptor (Nöthen et al., 1994) present in many populations.
Polymorphisms which affect function may be directly related to phenotype. Polymorphisms, whether or not they affect function themselves, provide a pathway to locating other polymorphisms by identifying specific chromosomal regions. The process of identifying genes related to illness or phenotype usually involves statistical analysis of frequencies of polymorphic variants (and patterns of transmission, when family material is available) in certain specialized samples including affected individuals.
We will review some of the basic methods used in locating genes related to phenotype, including genetic linkage; methods used in identifying genes related to phenotype, including genetic association; methods used in identifying variants (or, less properly, mutations) in candidate genes; and methods used in establishing that a particular variant is related to a phenotype. We will then review selected results in genetics of the major psychoses, substance dependence, and other psychiatric phenotypes illustrative of some of these methods.
Genetic linkage refers to cotransmission of polymorphic genetic markers or traits, one of which might be a phenotype such as a disease state, in families. When statistical evidence for cotransmission is sufficient, marker and trait (or marker and gene, or marker and marker) are said to be linked, and the location of a gene influencing the disease trait may be inferred to be close to the marker. The gene influencing the disease is then said to be mapped, i.e. located on a chromosome. Linkage in families using observations of identity by descent (i.b.d.) has been the preferred initial approach when appropriate family materials are available. It provides a general solution to the problem of localization of genetic susceptibility loci, a solution that does not depend on any prior knowledge other than that the phenotype studied has at least a partially genetic basis (for nonparametric analyses) and sometimes an idea of how the disorder is transmitted (for parametric analyses). The i.b.d. linkage approach enables one to screen the entire genome for susceptibility loci with the use of a relatively small panel of genetic markers (usually around 400) (Botstein et al., 1980). This is because each marker locus is capable of detecting an i.b.d. linkage signal in a large flanking region (see below).
Once a gene is mapped, it can usually be identified eventually, if the available sample of linked families is large enough. This task, called "positional cloning," is usually laborious after an i.b.d. linkage finding, for the same reason: the region of the genome potentially implicated is often very large, molecularly speaking. It is even more difficult to identify genes which have been imprecisely localized through affected sib pair strategies, and the problem of fuzzy localization sometimes has made it difficult even to establish if a result has actually been replicated (Risch and Botstein, 1996).
Genetic Association; Linkage Disequilibrium
Association studies are a popular and critically important alternative to i.b.d. linkage, and provide an approach with requirements for less specialized clinical material. Association studies can also be regarded as identity by state (i.b.s.) linkage. Markers that are i.b.d., shared between family members, are copies of the same exact allele, neglecting the effects of new mutations (which are usually not important in humans). The baseline rate of recombination, which is, roughly, one to several recombinations per chromosome arm per meiosis, guarantees that i.b.d. regions in families are quite long, measured in centimorgans (with one centimorgan equal to a 1% chance of recombination between markers). That is, markers that are i.b.d. in families are likely to be flanked by many other markers i.b.d., which makes the task of identifying a small stretch of DNA related to a shared phenotype difficult. Markers that are i.b.s. may derive from a common ancestral allele, but the common relative between individuals who share an allele i.b.s. may be many thousands of years in the past. Since individuals who share i.b.s. alleles may have hundreds of recombinations between them (so to speak), regions of identity by state (between unrelateds) are always much less extensive than regions of identity by descent (within families). (Regions with some remaining probability of a contiguous i.b.s. relationship are characterized by linkage disequilibrium relationships.) Thus, i.b.s. methods have the potential to locate genes influencing traits at a much finer level of resolution than i.b.d. methods.
In the classic i.b.s. study, allele frequencies at candidate loci are compared in sets of unrelated individuals differing for the phenotype of interest. Frequently, an identity-by-state association analysis is the capstone of a successful identity-by-descent linkage analysis, becoming the method by which the effect of a variant gene on physiology and phenotype is defined. Candidate gene studies are usually premised on hypotheses about relationships between specific known loci and particular phenotypes, sometimes bolstered by data about a locus's approximate map location.
Although i.b.s. association is a powerful method, especially when the functional variant (candidate allele) is tested directly, it is widely recognized that such case-control genetic association studies are susceptible to various kinds of artifact. The most significant problem has proven to be population stratification, the situation where differences in allele frequency seen in different groups arises from broadly observable population differences rather than phenotypic differences. Such differences have been demonstrated to exist for a range of loci of neuropsychiatric interest (e.g., Gelernter et al., 1998). The problem of stratification has led to the development of i.b.s. linkage association approaches that are not vulnerable to population stratification, as described below.
I.b.s. association is also critically useful as an initial study design. It would be very surprising if variation in some of these genes known to play a major role in neurotransmission, and in the action of psychotropic drugs, did not affect psychiatric phenotype in some way. For example, a functional variant mapping to the promoter of SLC6A4 (the gene encoding the serotonin transporter protein) was reported to be associated with anxiety-related traits (Lesch et al., 1996). To the extent that identification of such genetic effects allows for valid subtyping and parsing out of variance, it may simplify future genetic studies by enabling the construction of more homogeneous samples suitable for analysis of the variance (genetic and nongenetic) which remains to be assigned. An alternative initial approach is the simultaneous analysis of multiple loci for effect on phenotype; however, this has the potential to lower power to detect gene effects if there is appropriate correction for the additional degrees of freedom inherent in using different combinations of markers.
Linkage disequilibrium: TDT and related methods
The problem of population stratification for genetic association studies can be overcome with use of family association designs. The haplotype relative risk (HRR) method (Falk and Rubinstein, 1987; Ott, 1989; Terwilliger and Ott, 1992), a family-based genetic association method using subjects and their parents, compares allele frequency in a set of ill subjects with the set of nontransmitted parental alleles. With the HRR method, genotypes of affected individuals and their biological parents are identified. Of the four parental alleles, those transmitted to the affected offspring are compared with the nontransmitted alleles; if the alleles segregate independently from disease status, the transmitted and nontransmitted groups will be equivalent. This eliminates potential problems with population stratification: the transmitted and nontransmitted alleles derive from identical populations. The transmission disequilibrium test (TDT) (Spielman et al., 1993; Spielman and Ewens, 1996) can also be used with an HRR sample. TDT uses parents heterozygous for the disease-associated marker and considers observed transmission of the marker to affected offspring compared to expected (i.e. 50%) transmission. An extension of the TDT, the sib-TDT (s-TDT) (Spielman and Ewens, 1998), provides a TDT-like method for comparing genotypes in affected-unaffected sibling pairs; this method is likely to see wide use in the future.
Sequence Variant Detection
The two limiting factors for detection of sequence variants have been the availability of gene sequence and methods for high throughput, low cost detection. Although partial sequence (expressed sequence tags, or ESTs) is available for most human genes, complete sequence including intron and flanking sequences necessary for a comprehensive evaluation of sequence variation is not. Fortunately, publicly and commercially supported efforts to define the sequence of the human genome should complete the task within the next few years.
At the present time, many gene families of relevance in neuropsychiatry are well-represented or even nearly complete in their cloning (although intron and flanking sequences are frequently unavailable) Such gene families include the D1-5 dopamine receptors, serotonin receptors, monoamine biosynthetic and metabolic enzymes, certain peptide precursor proteins and receptor families, transporters, and some of the ligand-gated ion channels, for example, GABAA receptors.
Conservatively, sequence variants with rare allele frequency ³0.05 are found at approximately 1/1000 DNA bases overall and are less abundant in coding sequence and more abundant in regions which do not encode amino acid sequence. Thus, the human genome contains several million sequence variants of relatively high abundance. The Human Genome Project will produce genome-wide sequence on a limited number of subjects, perhaps one to five; however, tens of thousands of sequence variants should be detected even in this relatively small pool of chromosomes. A technical challenge which is now partially resolved is the development of methods suitable for the screening of hundreds of thousands of chromosomes, and enabling the discovery of all of the less common and uncommon variants which predispose to human disease. Variants which strongly predispose to diseases showing dominant inheritance will be uncommon for precisely that reason, although examples of common alleles which strongly predispose to recessively inherited diseases are known for example, the most common allele causing cystic fibrosis has an allele frequency of about 3% in European Americans.
Fortunately, recent advances in DNA sequencing and variant screening technologies have greatly facilitated this process, so that it is now feasible to screen large populations of subjects for variation. Several of the screening technologies are based on the principle that DNA sequence containing a mismatch (heteroduplexes) melt at a lower temperature than homoduplex DNA. A second group of screening methods is based on mismatch recognition. The third group is based on sequence-dependent variation in the structures formed by single-stranded DNA. Notably, some of the new methods (for example, dHPLC) are adaptable to sample pooling, and several are suitable for high throughput implementation with the assistance of robotics. (Table 1 summarizes features of variant screening methods.)
On the Horizon: New Technology
In the early 1980's, the use of Southern blotting and restriction fragment length polymorphisms made the study of molecular DNA polymorphism widely accessible and led to a wave of linkage studies and mapping of Mendelian traits. In the 1990's, PCR technology made generating the same kind of data much more efficient, perhaps by an order of magnitude or more; partly because of this development, most common Mendelian diseases have been mapped and progress has been made in mapping many complex traits. Today, most high-throughput genotyping uses sequencing instruments such as the ABI 377 to resolve differing alleles at short tandem repeat (STR) polymorphic sites. Now, we are in the early stages of the next development in DNA technologies that promise to make high-throughput genotyping cheaper and more efficient. These include the use of microarrays of DNA probes fixed to solid supports ("DNA chips") suitable for use in hybridization reactions which will allow a range of techniques, including sequencing, studies of RNA expression, and rapid genotyping of polymorphisms, primarily single nucleotide polymorphisms (SNPs) (Lander, 1999). DNA microarrays have been used for SNP discovery, and a prototype chip has been developed that can genotype 500 SNP polymorphisms simultaneously (Wang et al., 1998). Several commercial enterprises are in the forefront of the effort to develop these technologies. SNP-type polymorphisms are also amenable to analysis by other highly-automated but non-chip based technologies. These include solution-hybridization-based methods in which a fluorescent signal is revealed only if allele-specific hybridization occurs. Such genotyping can be performed on a large scale, and as an immediate fluorescent plate read following DNA amplification. When these techniques are widely available, the increase in genotyping capacity will provide a tool for understanding genetic variation and linkage disequilibrium relationships across the human genome in different populations, and it has been argued by Chakravarti (1999) that this will be a critical development for the understanding of the genetics of complex traits. We may thus be on the threshold of practicable applicability for whole-genome association (linkage disequilibirum) studies (per Risch and Merikangas, 1996), and such studies should substantially advance our understanding of the genetics of psychiatric disorders and behavioral traits.
III. Specific Results
Bipolar affective disorder
Bipolar affective disorder was one of the first major psychiatric disorders approached through a linkage strategy. It is clearly familial, and has long been recognized as such; support for a genetic contribution comes from twin (e.g. Bertelsen et al., 1977), adoption (e.g. Mendelwicz and Rainer, 1977) and family (e.g., Gershon et al., 1982) studies. At least in some families, inheritance appeared consistent with a single major locus and autosomal dominance. Although a common disorder from a genetic standpoint, bipolar disorder is less obviously biologically and genetically heterogeneous than unipolar depression, which has a lifetime prevalence of 5-25%. A gene, or several genes, for bipolar affective disorder thus should be amenable to gene localization by i.b.d. linkage analysis.
Genetic linkage has been employed for bipolar disorder since the pre-molecular era, using such classical markers as colorblindness (e.g. Mendelwicz et a., 1979). With the introduction of restriction fragment length polymorphism (RFLP) molecular markers, research accelerated, and there was an early report of linkage of bipolar affective disorder to chromosome 11 markers in a large Old Order Amish (OOA) kindred (Egeland et al., 1987). This linkage reached a reasonably high degree of statistical significance (lod score 4.9; a value >3 is considered significant by conventional standards). Nearby candidate genes included tyrosine hydroxylase (TH) (itself a linked marker) (Egeland et al., 1987), and the subsequently mapped D4 dopamine receptor (DRD4). However, with extensions of the Amish pedigree and updated diagnoses for a small number of individuals added in a new set of analyses, the chromosome 11 linkage in this family was no longer supported (Kelsoe et al., 1989). In addition, the finding was not replicated elsewhere. A similar fate befell the bipolar disorder linkage to markers near the colorblindness locus, on distal chromosome Xq. Although a lod score of >9 was reported, support for the positive result was diminished when more polymorphic molecular markers were employed (Baron et al. 1987; Baron et al., 1993).
It was not until 1994 that new positive linkage results for bipolar disorder appeared (Berrettini et al., 1994), and there have been a series of relevant results since. Berrettini et al. (1994), while not entirely abandoning the parametric linkage approach, added nonparametric sib pair analyses to their study, and it was these analyses that resulted in the first linkage finding for BPD that has, arguably, been replicated. Berrettini et al. studied chromosome 18 markers in a set of 22 pedigrees segregating BPD (comprising 159 ill individuals out of 365 total) and found evidence for linkage to pericentric markers using affected sib pair analyses and the affected pedigree member (APM) method. Seven markers had allele sharing corresponding to p£0.05. The authors considered these results consistent with either complex inheritance, or a gene of small effect.
Stine et al. (1995) studied 31 chromosome 18 markers in a set of 28 nuclear families (11 "paternal" and 16 "maternal," comprising 243 individuals) segregating BPD. They also reported increased allele sharing at some chromosome 18 loci. Their results were strongest when interpreted in the context of a parent-of-origin effect, with the evidence for linkage strongest in "paternal" kindreds and significant evidence for heterogeneity between "paternal" and "maternal" families. This parent-of-origin effect was consistent with observed clinical data (McMahon et al., 1995) and would suggest that, among subtypes of bipolar disorder, there exists a "paternal" chromosome 18 type, and a "maternal" type (as well as others; see below). Although the markers showing statistically significant evidence of linkage overlap between the studies of Berrettini et al. (1994) and Stine et al. (1995), they are not coincident; Berrettini et al. (1994) reported strongest support for linkage to markers on 18p, Stine et al. to markers on 18q. Gershon et al. (1996) have since obtained consistent results regarding an excess of paternal pedigrees among those showing evidence of chromosome 18 linkage in their series (overlapping pedigrees used by Berrettini et al. (1994)). Several other investigators have failed to find statistically significant support for chromosome 18 linkage in BPD.
Freimer et al. (1996) studied severe BPD in a large genetically isolated Costa Rican population using a genome screen (473 markers) in two pedigrees, and described evidence for linkage and association to chromosome 18q22-q23 markers. Particular, relatively long chromosome 18q haplotypes segregated with bipolar disorder in both families.
The three studies discussed above presenting evidence for genetic loci predisposing to BPD on chromosome 18 are inconsistent in the exact map locations implicated; Risch and Botstein (1996) point out that the implicated regions span a total of 136 centimorgans of chromosome 18. Thus, these studies may, or may not, have identified the same predisposing loci. Risch and Botstein (1996) suggest that a more-complex-than-expected mode of inheritance for BPD has led to reduced power (compared to Mendelian illnesses) and contributed to the problem of identifying replicable linkages. Proper interpretation of chromosome 18 linkage results for BPD remains controversial (Rice, 1997). The uncertainty regarding map location for some of the reported linkages may reflect inherent difficulties with ASP-type methods for gene localization, but could also reflect effects from more than one locus.
There have been several reports of linkage of BPD to markers on other chromosomes. Straub et al. (1994) reported linkage to the PFKL locus on chromosome 21q22.3 in one family of 47 studied, early in a genome scan project. APM analyses, this time using all families, also supported linkage. A chromosome 21 BPD locus was also supported by Detera-Wadleigh et al. (1996), using affected sib pair analysis in their set of 22 multiplex families (essentially the same clinical sample as for Berrettini et al. (1994)), but, although the regions implicated in the two studies overlap, in this data set PFKL did not give statistically significant evidence for linkage. Several other groups (Smyth et al., 1997; Aita et al., 1999) have since reported results consistent with a chromosome 21 locus for BPD, the latter in the context of investigation of a series of 40 families with use of affecteds-only analysis. Chromosomes 18 and 21 may thus be considered the most promising leads for genes predisposing to BPD.
The history of linkage studies for schizophrenia is especially illustrative of the difficulties involved in trying to map complex diseases using parametric analyses in extended pedigrees. The first few findings of linkage for schizophrenia that used molecular methods could not be confirmed, and it was not until nearly a decade later that more reliable results began to appear. In retrospect, it appears that when the first molecular linkage studies of schizophrenia (and bipolar affective disorder) began to appear, the scrutiny they received concentrated more on the promise (and actual robustness) of the molecular methods used than on the potential weakness of the studies, which involved definition of phenotype and methods of statistical analysis. Recent schizophrenia linkage studies have, largely, corrected these deficiencies.
Several authors have reported results consistent with linkage to markers on chromosome 6 (Wang et al (1995); Straub et al (1995); Schwab et al., 1995) using both parametric and nonparametric (extended affected sib pair) analyses, but without settling firmly on a particular chromosomal region. Numerous other chromosomal regions have also been implicated in schizophrenia by recent linkage findings. Perhaps the most interesting of these is a region on chromosome 22. Some linkage studies have provided suggestive evidence of a locus linked to schizophrenia on chromosome 22 (Gill et al., 1996). Velocardiofacial syndrome (VCFS) is a chromosome 22q11 deletion syndrome. Recent studies have established a degree of overlap between VCFS and schizophrenia (Karayiorgou et al., 1995). Groups of schizophrenic subjects have increased rates of VCFS, compared to a population sample (Bassett et al., 1998); and groups of subjects with VCFS show increased rates of schizophrenia and other psychotic disorders (Pulver et al., 1994). The set of subjects with 22q deletions may thus represent the first true genetic subgroup of schizophrenia.
A genetic locus on chromosome 15 close to the α7-nicotinic receptor gene (and possibly that gene itself) has been reported to contribute to abnormal P50 inhibition in families of subjects with schizophrenia (Freedman et al., 1997). At this point there are numerous putative schizophrenia susceptibility loci mapped (Blouin et al., 1998) but there are few specific candidates for loci responsible for these effects.
Genetic factors are important for the development of alcohol dependence, as established by twin, family, and adoption studies (reviewed in Gelernter, 1995). The largest twin studies have settled on heritability estimates in the range of 50-60% (e.g. Kendler et al., 1992; Kendler et al., 1997). A particularly noteworthy recent study (Kendler et al., 1997) considered the intersection between the Swedish Twin Registry, which logged almost all twins born from 1902 to 1949 (almost 9000 male pairs); and Swedish temperance board registrations from 1929 to 1974 (about 2500 twins). Subjects came to the attention of temperance boards mostly from physicians and law enforcement agencies because of alcoholism or crimes related to alcohol intake. (Temperance board registration is thus a reasonable, though obviously somewhat inexact, proxy for the diagnosis of alcohol dependence.) Kendler et al. were thus able to study not only the heritability of alcohol dependence per se, but the stability of that heritability over time, in an epidemiological sample. These authors found that although prevalence was similar in monozygotic (MZ) and dizygotic (DZ) twins, the concordance rate was significantly higher in MZ (47.9%) than in DZ (32.8%) twins. Moreover, heritability was stable over time. Calculation of tetrachoric correlations and model fitting yielded a best model attributing twin resemblance to genetic factors (54%) and family-environment factors (14%). It is particularly remarkable that such a complex and detailed analysis yielded results consistent with previous studies, and consistent over time within the study using essentially ad hoc diagnostic criteria. This provides an additional degree of confirmation that the diagnostic constructs used for alcohol dependence for which genetic liability can be estimated have validity and meaning in terms of consequences and outcomes in a person's life.
As discussed above, genome-scan studies and candidate gene studies are two applicable methods that may be used for identifying the location of genes influencing risk for the phenotype. The schism between candidate gene-motivated (genetic association) and linkage-motivated genetics research, as reflected by published research, has been particularly marked for alcohol dependence; however, at the time of this writing, there have been notable successes achieved by both methodologies. The clear successes by the former set of techniques provide a piquant reminder of the greater consistency expected when candidate gene hypotheses are moored to known, well-understood physiological mechanisms. We thus contrast current knowledge for two well-studied sets of candidate loci that have been studied by association: the alcohol metabolizing enzymes, and the D2 dopamine receptor locus.
Influence of genetic polymorphism at the loci encoding acetaldehyde dehydrogenase and alcohol dehydrogenase on risk of alcohol dependence in some populations is well established, and the mechanism is very clear. Ethanol is metabolized to acetaldehyde by alcohol dehydrogenases, for which the most important loci for our purposes are ADH2 and ADH3, both of which have been cloned and mapped to chromosome 4q. Acetaldehyde is metabolized primarily by acetaldehyde dehydrogenase, for which the relevant locus is mitochondrial ALDH2, which has been cloned and mapped to chromosome 12q. Acetaldehyde is toxic and produces a "flushing reaction" characterized by a set of uncomfortable symptoms including flushing, lightheadedness, palpitations, and nausea. Thus, any deviation from the normal metabolism of ethanol that resulted in increased exposure to acetaldehyde would create an aversive aspect to ethanol use, which might decrease risk of alcohol dependence (Goedde et al., 1979). Indeed, a variant (Lys487 for Glu487) that greatly reduces or eliminates ALDH function is common in Asian populations and is protective against alcohol dependence. Similarly, ADH variants that increase function (i.e., increase the rate of acetaldehyde synthesis) are also protective (see, e.g., Thomasson et al. (1991)). These are multiply replicated findings; polymorphic genetic variation is associated with phenotypic variation in a clear and reproducible way.
Probably the genetic locus most studied with respect to alcohol dependence is the D2 dopamine receptor locus, DRD2. In this case, there is a hypothesis for a mechanism of action and a proposed genetic association, but whether variation in this particular gene actually contributes to risk of alcoholism remains controversial. The polymorphic locus that has been studied most widely, the TaqI "A" system, has not yet been demonstrated to have any physiological effect, and its map position 3' to the gene makes such an effect relatively unlikely a priori. Of the other known DRD2 polymorphisms, two reported to have direct physiological effects, ser311cys (located in the coding region) and -141CIns/Del (in the promoter). However, these functional variants have been reported to have no measurable effect on risk for alcohol dependence (Goldman et al., 1997; Gelernter and Kranzler, in press). All published family-controlled association studies at this locus, including a sib-pair based linkage study in an American Indian population (Goldman et al., 1997) and a large haplotype relative risk study from the COGA collaboration (Edenberg et al., 1998), have failed to demonstrate significant linkage. The possible existence of a physiologically important association at this locus is still considered an important research issue, however. The continued publication of positive findings can be explained in terms of a small effect of polymorphic variation at this locus on risk, but it could also be explained in other ways, e.g. by population stratification (Gelernter, Goldman, and Risch, 1993) or publication bias (Gelernter and Kranzler, in press).
Recent linkage studies of alcohol dependence published by the COGA collaboration (Reich et al., 1998) and the NIAAA group (Long et al., 1998) have provided several chromosomal locations with promising lod scores that may harbor loci influencing risk of alcohol dependence. Interestingly, both groups reported data consistent with loci influencing risk for alcohol dependence mapping close to ADH2 and ADH3. The genetically complex nature of alcohol dependence will make it very difficult to identify specific genes influencing risk without recourse to other methods, depending on linkage disequilibrium scans or candidate gene hypotheses, or both. Taking all that is known about genetic risk for alcoholism and what can be accounted for by known mechanisms, most of the genetic factors remain unelucidated.
Studies of genetics of drug dependence are in their infancy relative to other major psychiatric disorders, owing to incomplete understanding of the genetic spectrum to which these disorders belong and questions about degree of heritability and mode of inheritance. At least some of these basic issues were resolved for bipolar disorder and schizophrenia decades ago. Understanding genetic overlap between alcohol and drug dependence will be critical for the determination of the genetic structure of each; recently, several well-designed studies addressing this issue have been published, and the recent results have bolstered the view that the genetic factors leading to alcohol and drug dependence are largely separate; moreover, distinct genetic factors affect liability to abuse particular substances (Goldman and Bergen, 1998). Tsuang et al. (1996) collected data on more than 3,000 twin pairs from the Vietnam Era Twin Registry, among whom drug abuse was defined as at least weekly use of any of a variety of drugs. Comparisons of pairwise concordance rates and tetrachoric correlations were used to demonstrate a familial basis for different forms of substance dependence, cumulatively and individually. For stimulant abuse, MZ twin tetrachoric correlation was 0.53 and DZ 0.24, with estimated heritability of 0.44, with unique environment accounting for most of the remaining variance. MZ concordance was 14.1% and DZ concordance 5.3%. For opioid dependence specifically, MZ twin tetrachoric correlation was 0.67; DZ, 0.29, with estimated heritability of 0.43, unique environment effect 0.31, and "non-additive" genetic components estimated to account for 26% of the variance. MZ concordance was 13.3% and DZ concordance 2.9%. The difference in pairwise concordance rates for MZ and DZ twins was significant for abuse of marijuana, stimulants, cocaine and for all drugs combined. Tsuang et al. (1998) reported further data from the Vietnam Era Registry, focusing on issues relating to comorbidity for different forms of substance abuse. They determined that there exist genetic factors both specific to certain individual drugs of abuse, including stimulants, and general to multiple forms of abuse.
Data from two recent family studies addressed familial aggregation of drug dependence (Merikangas et al., 1998; Bierut et al., 1998). Merikangas et al. (1998) found specifically increased risk for a range of substance abuse categories (opioids, cocaine, cannabis, and alcohol) in relatives of subjects with the disorder. In particular, these authors found their data consistent with independent aggregation of alcoholism and drug disorders. Bierut et al. (1998) studied alcohol, marijuana, cocaine, and nicotine dependence in the COGA sample. They reported evidence for both common liability factors, affecting risk for alcohol dependence as well as other substances; and substance-specific factors.
Thus, it has been demonstrated individually for several specific drugs of abuse that genetic factors play an important role in determining an individual's risk. These findings have clear implications in terms of research that must be completed to identify the specific genetic risk factors involved. Although genetic studies of alcohol dependence and substance dependence may converge on some common loci influencing aspects of phenotype such as personality structure, other factors will be identified only with collection of suitable clinical samples segregating each disorder of interest. It is easy to picture what some of these specific genetic factors might be; some could be analogous to ALDH2 in alcohol dependence, i.e. a lesion in a metabolic enzyme might render a certain drug especially aversive to some subjects, and this could protect against abuse of that substance. Others could affect ligand binding for a brain protein, such as a receptor, of particular importance for a certain class of abusable substances; e.g. variation at the μ opioid receptor locus (OPRM1) could specifically affect affinity for (say) heroin (and not endogenous ligands for this receptor), and therefore could affect risk of abuse for heroin only. Of course, others will certainly affect genes whose function we cannot yet imagine. So far, no large-scale genetic mapping studies, or family association studies, have been published, for drug dependence phenotypes.
Other phenotypes: OCD; ADHD; Homosexuality
The first linkage study of male homosexuality (Hamer et al, 1993) used 40 sib pairs of homosexual brothers (mostly European-Americans), and provided statistically significant support for a locus predisposing to male homosexuality on Xq28; the X chromosome was studied due to the authors observation albeit based on family history and not direct interview data of increased male homosexuality in maternal relatives and consequent inference of possible X-linked inheritance. Thirty-three of 40 sib pairs were concordant for Xq28 markers. This study was critiqued by Risch et al. (1993), who highlighted possible inconsistencies in previous knowledge about genetic contributions to homosexuality, and suggest that decreased male-to-male transmission for homosexuality is at least partially trait related (rather than indicative of sex-linked inheritance). However, Hu et al. (1995) replicated their finding in a new sample of 33 pairs of homosexual brothers: 22 of 32 informative sib pairs shared their Xq28 markers (p<0.05). These analyses have not as yet converged on a candidate gene from the Xq28 region. There was no linkage to this region for female homosexuality.
While psychiatric genetics is still in a relatively early stage relative to other subspecialties in medical genetics, progress is being made, and certain trends are evident. First, virtually any psychiatric or behavioral phenotype that can be named is subject to some amount of genetic influence, and this influence is often substantial. Second, the common psychiatric disorders and behavioral phenotypes are genetic complex and multifactorial. This is entirely consistent with expectations for common phenotypes. However, common phenotypes are caused by common genes, and such genes must therefore be scattered not only among those with the phenotype but those without it, and genes of this sort are hard to identify. Third, variability in expression of the common phenotypes has made it particularly difficult to recognize specific syndromes that might have been mapped separately. That is, while in medical genetics, biochemical and other phenotypic characteristics have made it possible to genetically dissect and identify genetically homogeneous subgroups with clear mode of inheritance of (for example) hypertension (Hansson et al., 1995; Mansfield et al., 1997), no analogous process has taken place in psychiatry, with rare exceptions such as Brunner's syndrome (Brunner et al., 1993). This has made the task immeasurably more difficult. Fourth, although candidate gene hypotheses have been of value in some cases, they have been of the greatest value when the polymorphism studied has a measurable effect on protein structure or regulation, such as the alcohol metabolism genes; and they have produced the most difficult-to-replicate results when nonfunctional polymorphisms were mated with case-control designs subjects to population stratification. Fifth, while genome scans have produced some replicable linkage findings for the major psychoses, the very nature of the methods of design and analysis that made such findings possible have also made it difficult to proceed from identified linkages to specific genes.
We therefore conclude that future advances in identifying specific genes with specific effects in the common psychiatric phenotypes will depend either on application of newer techniques, such as techniques relying on combining genome scan linkage with linkage disequilibrium methods; or on application of existing techniques, but using samples with considerably greater power than those that have been available to date.
This work was supported in part by funds from the U.S. Department of Veterans Affairs (the VA Medical Research Program), NIMH grant K02-MH01387, and NIAAA grant R01-AA11330.