Genetic Influences in Drug Abuse

George R. Uhl, Gregory I. Elmer, Michele C. LaBuda, and Roy W. Pickens

Individuals are differentially vulnerable to substance abuse. Everyone has access to addictive substances. Not everyone who has an opportunity to use an addictive substance does so, and not everyone who uses an addictive substance becomes addicted. Sixty-three percent of individuals 12 years of age or older report never using illicit drugs or psychotherapeutics, 29% never use cigarettes, and 17% never use alcohol (1992 National Household Survey; Substance Abuse and Mental Health Services Administration, 1993). Sixty-five percent of individuals reporting access to marijuana use the substance, while only 16% of those having access to heroin report use (National Institute on Drug Abuse, 1991).

The likelihood of continuing drug use also varies from individual to individual. Illicit drug use often begins in early teen years, peaks in the late teens and early twenties, and can decline substantially thereafter ( see ref. 39). However, some individuals continue to use drugs into later adulthood. Ninety-three percent of individuals who used alcohol, 60% of cigarette smokers, 19% of heroin users, and 8% of hallucinogen users continued to use drugs at the end of their third decades of life (see ref. 57).

The frequency and consequences of illicit drug use, factors that underlie most definitions of drug addiction, also vary from user to user. Forty-five percent of individuals using marijuana report using it 12 times or more; half of these report using it once a week or more (National Institute on Drug Abuse, 1992). Variability can also be seen in reports of symptoms of dependence, even among regular users. Forty-two percent of regular cocaine users did not report any symptom of cocaine dependence (National Institute on Drug Abuse, 1991).

Observations such as these suggest that genetic and environmental conditions that differentially predispose individuals to drug-taking behavior and to the transition from drug-taking behavior to established and maintained drug abuse might be found.

Processes involved in substance abuse are likely to be behaviorally complex; genetic mechanisms contributing to interindividual differences in substance abuse vulnerability are thus likely to be equally complex. Genetic influences on drug use and dependence might operate at a variety of levels. Genetic influences that contribute to the initiation of drug use may differ from those that contribute to heavier drug use or drug dependence (see ref. 30). Drug use vulnerability might also be modified by protective factors that could contribute to drug abstinence or protect from development of regular use patterns or drug dependence (see ref. 30). Allelic variants of specific genes could mediate differential drug reinforcing properties, alter drug pharmacodynamics or pharmacokinetics, influence "sensation-seeking" personality traits that may facilitate exposure to drugs, exacerbate drug toxicities, or minimize "protective" factors such as hangovers. Individuals also differ in their specific drugs of choice. Ideal investigations of the genetic components of addictions might therefore separately investigate the neurobiological substrates of vulnerability to drug use acquisition, maintenance, or resistance to extinction for each abused substance. In practice, however, individuals meeting substance abuse criteria which largely center on the extent and duration of consumption of addictive substances are the focus of most attention in studies attempting to identify components of the genetic bases of substance abuse vulnerability.

Familial and population genetic studies are beginning to reveal possible genetic bases for some of the interindividual differences in vulnerability to substance abuse. Studies of related individuals can provide evidence of the proportionate contributions of genes and environment to a behavioral disorder such as substance abuse. Data from these approaches can also be used to suggest whether a hypothesized simple genetic pattern of familial transmission fits with observed data to suggest mendelizing inheritance. Several differing methodologies are employed. Each is based upon a set of assumptions, and the results of each must therefore be interpreted in light of specific methodological limitations (see ref. 51); see Table 1). Twin and adoption studies have determined that both genetic and environmental influences are involved in drug use or dependence and have indicated the extent of their involvement (see below). Perhaps the most convincing evidence for the nature and degree of genetic bases for substance abuse comes from the convergence of results obtained through methodologically-distinct approaches.

Application of molecular genetic approaches to studies of substance abuse is relatively recent. Nevertheless, recent studies with polymorphic genetic markers at several candidate gene loci have initiated the search for specific genes whose alleles could contribute to genetic differences in substance abuse vulnerability (see below). Indeed, substantial collective work on markers at the dopamine D2 receptor gene locus (DRD2) has continued to suggest differences between substance abusers and controls in work from several, but not all, groups (see ref. 66, 67; but see ref. 23). Because molecular genetic studies in this area are likely to assume increasing prominence, the second portion of this chapter describes possible approaches to identifying candidate or anonymous genes that could contribute to substance abuse vulnerability. It also details evidence at the DRD2 locus, the single site most explored as a candidate for contributions to substance abuse vulnerability.

Animal studies have shown that a variety of species can be bred for drug-accepting preferences, and they have attempted to identify likely candidate gene loci for symptoms of drug dependence. Such studies cannot provide direct evidence concerning human genetic polymorphisms relevant to substance abuse. However, these studies can reveal some of the behavioral parameters susceptible to genetic variability. Some of the genetic possibilities for human substance abuse vulnerability, derived from animal studies, are thus detailed in the concluding section of this chapter.

Several reviews of the large body of classical genetic studies of alcoholism have recently appeared (e.g., see refs. 15, 44, and 61; also see Interactions Between the Nervous System and the Immune System: Implications for Psychopharmacology), but little has been done to summarize data that focus on drug abusers. This chapter will therefore review the genetics of drug abuse. Because few drug abusers abstain from alcohol, and because alcoholism is present in almost half of drug abusers in population-based surveys (see ref. 58), these approaches in fact often focus on polysubstance abusers.

This review's focus on genetic influences should not obscure the significant environmental nature of many risk factors for drug abuse. Drugs must be available for drug abuse to develop. The type of drug available can also be important, because drugs can differ considerably in their reinforcing effects and abuse liabilities. Risk factors reported to contribute to drug use include laws, social norms, drug availability, economic deprivation, neighborhood disorganization, family drug-related behavior, family management practices, family conflict, low family bonding, early and persistent problem behaviors, academic failure, low commitment to school, peer rejection in elementary grades, association with drug-using peers, alienation and rebelliousness, attitudes favorable to drug use, and early onset of drug use (see ref. 32). One major fruit of the labors of elucidating genetic influences of drug abuse behaviors will be that the environmental components of vulnerability will be more readily identified so that their independent and interactive components can be more easily assessed (see ref. 40). Appropriate recognition of genetic and environmental vulnerabilities can lead to improvements and better targeting of treatment and prevention strategies.

Family studies analyze transmission of substance abuse disorders from generation to generation through families (Table 1). The basic approach determines if family members of substance abusers are at increased risk for substance abuse. A variant, segregation analysis, uses the diagnostic status of family members to determine the relative likelihoods of different modes of genetic transmission. These studies display limitations noted in Table 1. Noting these limitations is especially important in substance abuse disorders in which marked generational and secular trends in drug of choice, nonpaternity, and assortative mating can all provide confounding factors. Family studies do not conclusively separate genetic from family environmental bases for these disorders. Adoption and twin study data are thus of additional importance.

Adoption studies aim to separate the effects of genes and environment by studying the similarity between adopted-away children, their biological parents, and their adoptive parents. Although prenatal environment is confounded with genetic influences, this approach also displays substantial power in differentiating between environmental and genetic contributions to substance abuse disorders (Table 1).

Twin studies determine if within-pair similarity for substance abuse is greater in genetically identical monozygotic (MZ) twins than in less genetically similar fraternal dizygotic (DZ) twins (Table 1). Careful studies of these relatively rare subjects have allowed the first tentative estimates of the magnitude of the genetic components of drug abuse vulnerability (e.g., see ref. 56).

There are now family, twin, and adoption studies of drug abuse. However, it is important to keep in mind the extensive alcoholism comorbidity found in drug abusers (see ref. 58). Alcoholism and drug abuse display substantial comorbidity both in clinical samples and in the general population. Abuse of cocaine, sedatives, opiates, hallucinogens, and amphetamines, for example, was found to be 10 times higher in alcoholics than in nonalcoholics in the epidemiological catchment area (ECA) survey (see ref. 35). Almost half of drug abusers attained DSM III-R criteria for alcohol abuse or dependence in this ECA sample (see ref. 58).

Extensive family study work now supports enhanced frequencies of drug abuse in families of drug abuser probands when compared to general population base rates. Most of these workers have utilized family history and family structure data obtained from the proband, and varying diagnostic criteria have been employed.

Several studies of the families of drug abusers now strongly suggest familial influences on drug abuse, although different diagnostic criteria and ascertainment methodologies have been used in different time periods (see references in refs. 13, 41, 42, 43, 45, and 59) (Table 2). These data are supported by results of family studies that cannot be summarized in the same fashion (e.g., see refs. 20 and 28). The more recent work of Mirin et al. (see ref. 45), who developed DSM III diagnoses in relatives of 350 substance abusers, provides features typical of many of these studies. These workers found that male relatives were almost twice as likely as female relatives to display substance abuse. Furthermore, this increased risk for male relatives was observed whether the proband preferred opiates or cocaine, although female relatives of sedative–hypnotic abusers were more likely than male relatives to abuse drugs. Unfortunately, this work also resembles most of the other reports in that appropriate control populations interviewed in the same fashion as the relatives of drug abusers were not included. ECA data suggest prevalences of 7% and 4% for drug abuse or dependence in males and females, respectively (see ref. 58). Therefore data from studies with control populations assessed using similar instruments are of special value in helping to interpret these frequency estimates.

Rounsaville and co-workers (41, 42, 59) studied relatives of 201 opiate addicts and 82 controls obtained in a similar population area. Between 18% and 23% of first-degree relatives of addicts, as well as 3% of first-degree relatives of normal controls, manifested drug abuse. These data suggested an odds ratio of almost 14-fold for drug abuse in first-degree relatives of drug abusers, compared to the normal controls examined (59). Rates of drug abuse in siblings were almost three times higher than those in parents, and gender differences were significant (59). Luthar and Rounsaville (43) used similar approaches to study first-degree relatives of 298 cocaine abuser probands. They found that 40% of male and 22% of female sibs of cocaine users, as well as 5% of parents, displayed drug abuse or dependence. These values were thus significantly different from ECA data and from previously studied control populations.

Several conclusions concerning the familial nature of drug abuse now seem reasonably well supported by this accumulated family study data:

1. A substantial degree of familiarity of substance abuse vulnerability fits best with the accumulated data. Family studies, standing alone, cannot separate family environment from genetic contributions to familial resemblance, however.

2. The marked secular trends in abused substances of choice and male/female differences in substance abuse muddy these data. Family studies of "vertical," transgenerational transmission of substance abuse can thus be significantly limited due to differences in the pattern of availability of drugs across generations, and due to gender differences in substance use. Anecdotal data obtained from studying pedigrees of families of drug abusers reveals frequent multigenerational patterns of abuse of alcohol and drugs, suggesting that certain underlying genetic and/or environmental determinants might predispose to abuse of both categories of substances (but see below). Control for these generational and gender factors in studies of the familial aggregation of drug abuse may be exerted by restricting analyses to individuals within the same temporal and gender cohort, such as siblings or twins of the same sex.

Cadoret et al. (8) studied almost 450 adopted individuals, finding 40 with drug abuse, 75 with alcohol abuse, and 46 with antisocial personality problems. There was a significant correlation between drug abuse in the adoptee and alcohol problems in the biological parent (odds ratio = 4.3), but there was no increased risk if only the adoptive parent drank (8). This evidence for a genetic basis for substance abuse heritability was accompanied by an enhanced risk for alcoholism if the biological parent was alcoholic (odds ratio = 5.9). Interestingly, while antisocial personality in the biological parent enhanced the relative risk of antisocial personality, it did not significantly increase the risk of drug abuse. This work, as well as a recent replication of the results in a different group of subjects (R. Cadoret, personal communication, 1993), suggests a significant genetic component to heritability of substance abuse.

Pickens et al. (56) examined twin concordance for drug abuse in a sample of 50 monozygotic and 64 dizygotic twins in which the proband was identified in drug and alcoholism rehabilitation programs. These workers found a significantly greater concordance for substance abuse in monozygotic males than in dizygotic males. These workers then used ECA data concerning population prevalences to estimate components of liability variance. These analyses provide, for the first time, an estimate of the proportion of genetic contribution to substance abuse in these individuals. When substance abuse and/or dependence in men was considered, 31% of the variance could be attributed to genetic components. The corresponding figures for alcohol dependence were 60%. Trends in the same direction in a smaller female population did not reach statistical significance (56).

A study of over 2600 Vietnam Era twin-pair registrants also indicated genetic components to drug abuse (31). Significant heritability (h2 values ranging from 0.4 to 0.6) for abuse of hallucinogens, stimulants, opiates, sedatives, and marijuana was found. Only cannabis users displayed shared environmental components. This large twin study thus provides the first evidence that vulnerability to each of these classes of addictive substances may display a genetic component. Equally striking is the conclusion that vulnerability to no substance could be attributed to shared environmental features alone. There is thus little evidence from either twin study to suggest that environmental factors shared within a twin pair, such as neighborhood characteristics, religious affiliation or common friends are significant causes of twin pair resemblance for drug usage.

Similar data is found in studies of nicotine abuse, although relative risk estimates from genetic twin studies are substantially more modest. Studies in the United States and Scandinavia suggest modest genetic contributions to relative risks of smoking (e.g., see ref. 9). Interestingly, stronger genetic components were identified in individuals who had never smoked and in those who successfully stopped (9). These data concerning a substance with wide environmental availability provide some of the only information concerning the possibility that genetic influences on resistance to substance use and on substance abuse cessation might be as robust as those impacting on more typical measures of quantity and frequency of substance use.

In summary, twin data now provide significant support for the idea that drug abuse vulnerability displays significant genetic components. Genetic components appear to be more prominent in the more severe abusers. However, environmental influences on drug abuse susceptibility are underscored by the failure of concordance for drug abuse in more than one-third of identical male twins and more than two-thirds of identical female twins (e.g., see ref. 56).

Alcoholism and drug abuse display substantial comorbidity both in clinical samples and in the general population. Abuse of cocaine, sedatives, opiates, hallucinogens, and amphetamines, for example, was found to be 10 times higher in alcoholics than in nonalcoholics in the ECA survey (35). Many family pedigrees contain both alcohol and drug abusers (see above). However, tentative evidence from two studies suggests caution in assuming that possible genetic bases for predisposition to alcoholism are identical to genetic bases for predisposition to substance abuse.

Hill et al. (36) found that alcoholism clustered in families to a greater extent than did opiate addiction, although secular trends with more striking opiate use in sibs than in parents may have obscured the analyses. Rounsaville et al. (59) also identified higher rates of alcoholism in relatives of opiate addict probands who had comorbid alcoholism than in relatives of opiate addicts lacking this comorbidity. Such data would mitigate against a single, identical genetic determinant underlying both alcoholism and substance abuse. Anecdotal pedigree studies do suggest that both of these disorders occur in many of the same families and thus probably share some genetic components. Identification of the specific genetic determinants shared by drug abuse and alcoholism and those that might be specific to each group of agents remains an area of current interest.

Familial patterns of polysubstance abuse are documented by the family studies undertaken to date, as noted above. Unfortunately, the striking secular trends in the substances used have rendered attempts to identify the familial patterns of transmission difficult. Indeed, even with alcoholism, two attempts to characterize the transmission of the underlying genetic vulnerability to alcoholism have reached dissimilar conclusions (3, 29, 9). Gilligan et al. (29) found evidence to support a recessive major locus in families of male alcoholics. However, Aston and Hill (3) concluded that the major effect transmitted within the families of male alcoholics was either nonmendelian or a major environmental effect with a polygenic background. No evidence for a major effect of either genes or environment was found in families of female alcoholics (29). Multifactorial transmission, combining effects at several gene loci with environmental influences, may be the most likely mode of transmission in these families. Although such a pattern appears most likely to also explain drug abuse, direct segregation study data documenting this conclusion is currently lacking. The lack of solid information concerning inheritance pattern provides difficulties for genetic linkage analyses, which require assumptions concerning mode of inheritance.

There are many points, beginning with initial exposure to drugs and leading up to substance abuse or dependence, upon which genetic research could focus. As genetic underpinnings of substance abuse in toto become better established, genetic contributions to specific features of substance abuse will be increasingly studied. Features of interest include the following:

1. Subjective and objective effects of substances. Studies in this area investigate whether genetic vulnerability is related to either subjective or objective effects of substances. These aspects have been approached in alcoholism, where results are contradictory. Individuals with positive family histories for alcoholism can report less intense feelings of intoxication at moderate doses of alcohol as well as other behavioral and biochemical effects (46).

2. Factors affecting exposure. Factors which lead to initial experimentation with drugs provide another area of research. The impact of genetic and environmental effects on teenage alcohol use have been estimated in a study of Australian twins (33), which found that both genetic and environmental factors shared within twin pairs are important in determining teenage abstinence. Additionally, genetic factors were found to be a major contributor to the observed variance in age of onset of alcohol use in females, whereas shared environmental factors were much more influential in males. Contrastingly, genetic factors were significant contributors to average weekly consumption in both sexes. Just as environmental factors important in exposure to alcohol and drugs may differ from environmental factors important in first use or in escalating use to abuse, the genetic factors important in the degree of alcohol consumption are important only once alcohol use has started but have little to do with the determinants of age of onset. Because this study involved nonalcoholic individuals, there is no evidence to determine whether the genetic factors affecting alcohol use are the same or different from those genetic factors which are involved in alcohol dependence. Studies in drug abusers can also help to tease out such components.

3. Factors influencing drug metabolism and distribution. The pathways by which most drugs are metabolized are known. Genetic variants at some loci, such as the cholinesterase that is responsible for much of cocaine's metabolism, are also known. Studies relating these variants to substance abuse phenotypes are just beginning. However, work on the alcohol metabolism genetics of alcohol dehydrogenase and aldehyde dehydrogenase gene variants has revealed (a) variants occurring more frequently in alcoholics than in controls and (b) variants associated with the characteristic "flushing" response noted in Asians (see ref. 63). Adult, nonalcoholic monozygotic twins are also more similar than dizygotic twins in measures of alcohol absorption, degradation, and elimination (34). Conceivably, similar genetic contributions to drug metabolism could help to explain components of the genetic contribution to substance abuse vulnerability.

4. Comorbid behavioral factors. Families with drug abuse and alcoholism are frequently enriched in individuals with other psychiatric diagnoses, especially antisocial personality. Indeed, Cadoret et al. (8) have suggested that drug abuse may occur on three distinct backgrounds in his adoptees: those of biological familial antisocial personality, those of biological familial background of substance abuse, and those with adoptive families providing an environment of disruption and psychiatric disturbance. Twin data also suggest that cross-concordances for alcoholism and antisocial personality in male twins are likely to display significant genetic components (Pickens et al., personal communication).

Knowledge about the cellular and molecular bases of acute drug action has exploded over the last several years. Information about the genes expressed by brain systems on which drugs act has allowed testing of the possibility that interindividual differences in genes encoding proteins expressed in these systems could contribute to interindividual differences in drug abuse vulnerability (reviewed in ref. 65).

Significant data now support the idea that virtually every abused drug can induce behavioral reinforcing properties by altering function in brain dopamine circuits arising from the ventral midbrain (16). Genes important in the mesolimbic/mesocortical dopaminergic pathways are thus strong candidate genes for possible contributions to interindividual differences in substance abuse vulnerability. The dopamine D2 receptor gene (see below) represents one such candidate gene.

Recent molecular cloning studies have also identified the genes encoding many drug receptors, including the dopamine transporter that is the pharmacologically defined cocaine receptor, G-protein-linked opiate receptors that are the heroin/morphine receptors, the G-protein-linked cannabinoid receptor that mediates marijuana action, the N-methyl-D-aspartate (NMDA)-glutamate receptor ligand-gated ion channels that mediate phencyclidine actions, the nicotinic acetylcholine receptor ligand-gated ion channel that is the site of action of nicotine, and the gamma-aminobutyric acid (GABA) receptor ligand-gated ion channels that mediate actions of barbiturates and benzodiazepines (see ref. 33 for review). GABA and NMDA receptors are also strong candidate loci for acute ethanol effects.

We know how addictive processes start: Drug occupies a brain receptor. Processes directly modulating receptor function, however, have not been able to account for significant fractions of the biochemical bases of addiction. Other candidate molecular mechanisms of information storage can be tentatively postulated as possibly involved in addiction (65). However, direct genetic approaches not dependent on biochemical hypotheses may be more likely to identify genes that could contribute to interindividual differences in substance abuse vulnerability than candidate gene approaches that assume greater knowledge of addiction process biochemistry than we may actually possess.

A large number of richly polymorphic genetic markers are now available for use. These markers include variable number polymorphisms such as simple sequence repeats and restriction fragment length polymorphisms (RFLPs). The increasing number of such markers, distributed on each segment of each human chromosome, provides increasing potential power for genetic studies scanning the genome for loci that are associated with substance abuse vulnerability. As noted below, this power may be essential for any approach that aims to detect functional genes with characteristics likely to contribute to substance abuse vulnerabilities.

Linkage analyses use family data to statistically link a trait to a genetic marker on a particular chromosome. Genetic markers at a single gene locus can be identified as linked with a disease if, in different generations of families displaying a genetic familial disorder, one form of a genetic polymorphic marker at a gene locus is coinherited with the disease. One gene marker form is thus present in family members with the disease, but not in those displaying normal phenotypes, in the simplest case. Genetic linkage studies thus involve analysis of the ways in which the disease phenotype and each of a number of genotypic markers cosegregate or appear together in different family members.

Association studies compare genetic variation at a specific site to behavioral variation within affected and unaffected individuals. Genetic markers at a gene locus can also be associated with a disease if they are present more often in unrelated individuals displaying the disease than in unaffected individuals.

The probable polygenic mode of inheritance of substance abuse vulnerabilities and the large environmental impact on substance abuse outcomes are both likely to dramatically weaken the power of the classical familial molecular genetic linkage approaches, in which tests are made to determine how genotypes at each of many genetic loci cosegregate with substance abuse phenotypes. If substance abuse in any individual could be caused by several different genes, or by strictly environmental influences, then numerous "phenocopy" substance abusers will share the same clinical characteristics but differ in genotype. Mathematical modeling studies indicate dramatic loss of power of genetic linkage studies as more different genes and more diverse environmental influences yield more frequent phenocopies. Thus, true presence of a genetic marker in only substance abusers in one family due to the marker's true linkage with a nearby gene allele causing this substance abuse will be masked by studies in a second phenocopy family in which other genes lead to substance abuse or in a third phenocopy family in which environmental factors lead to nongenetic but still familial patterns of substance abuse.

Each of these genetic and environmental phenocopies also weakens association studies, in which genetic marker presence in populations of substance abusers is compared to marker presence in nonabusers. However, modeling work suggests that the loss of power due to phenocopies may be less severe in association studies than in classical linkage studies. Thus, if a set of genetic markers adequately tags each of the genes potentially implicated in drug abuse vulnerability, association studies should be better able to elucidate the genes that contribute to the disorder. The abilities of genetic markers to provide such informative data about functional gene alleles, however, rests on specific considerations based on recombination between the polymorphic markers and the functional alleles of interest.

The secular trends in abused substances also provide substantial difficulties for linkage studies, which often rely on phenotype characterizations of individuals in different generations maturing in substantially different drug environments (see above). Linkage studies, which study affected and unaffected individuals in the same cohort (e.g., sib-pairs), are less susceptible to these problems.

The abilities of either candidate gene or anonymous polymorphic genetic markers to adequately indicate either genetic linkage or allelic association is based on considerations of chromosomal recombination. The meiotic events forming the chromosomes for each human generation result in "crossing over" recombinant events splicing sequences from one chromosomal copy with those of the other chromosomal copy. The average rate of this process allows 1% of chromosomal loci separated by 1 million base pairs of DNA to recombine in each generation. However, this average rate of recombination can vary substantially across different chromosomal loci (see ref. 66).

For a polymorphic marker to indicate genetic linkage, chromosomal recombination between that marker and a postulated gene leading to familial substance abuse must occur infrequently within the several generations assessed in the family. Markers located within several million base pairs of the functional gene defect causing substance abuse would thus be likely to detect genetic linkage if substance abuse were a mendelizing disorder. However, for allelic association to be detected, chromosomal recombination between a polymorphic marker and a postulated gene contributing to substance abuse in the population must occur infrequently within the many generations separating the inheritance of the population sampled. A polymorphic marker such as a specific RFLP successfully used in association studies could thus lie very close to the functional gene defect contributing to substance abuse vulnerability. Alternatively, such a marker could be in linkage disequilibrium with the functional gene defect. Linkage disequilibrium is a term used to define a process with poorly understood mechanisms that results in the much-lower-than-average rates of recombination observed between some chromosomal loci (see ref. 66). Linkage disequilibrium can thus allow a genetic marker used in an allelic association study to provide information not only about whether closely adjacent DNA contains a functional gene defect, but also about the possibility that DNA many thousands of bases removed from the polymorphic genetic marker but in linkage disequilibrium with it could also contain a functional gene defect. As we shall see below, the linkage disequilibrium that has now been well documented at the dopamine D2 receptor gene locus provides a plausible rationale for allelic association between RFLP markers at this locus and substance abuse vulnerability.

The DRD2 gene encodes a G-protein-linked, seven-transmembrane region receptor protein expressed abundantly in dopaminergic circuits important for behavioral reward (66). Genetic polymorphic markers have been identified at several DRD2 loci. A TaqI A RFLP is located in the 3¢ flanking region of the DRD2 gene, a TaqI B RFLP lies more 5¢, and a TaqI C RFLP provides a polymorphic marker for a site lying between TaqI A and TaqI B (see ref. 66).

Blum et al. (6) provided the first evidence that the DRD2 gene might display population variants influencing susceptibility to alcoholism. These workers found a striking allelic association between the TaqI A1 RFLP form and alcoholism. However, these results were viewed with caution for several reasons (25, 67). Thousands of different genes are expressed in the human brain; the a priori probability of identifying a vulnerability-enhancing allele of one of these genes was low. Genetic linkage of this marker in several families in which alcohol abuse appeared to pass from generation to generation in nearly mendelizing fashion was not supported (7, 52). The TaqI A RFLPs used in this study were demonstrated to be inhomogeneously distributed in different human populations (50), with high A1 allele frequencies in black, Asian, and American Indian populations (23, 50, 67). Spurious associations not indicative of true causal links between DRD2 allelic status and substance abuse could thus result from sample stratification—that is, disproportionate sampling of abusers or controls from population subgroups displaying atypical RFLP frequencies.

Despite these cautions, we and other investigators have examined DRD2 gene markers in drug abusers (Table 3). These data are now joined by results obtained by three other laboratories that have now provided data that allow assessments of DRD2 gene marker frequencies in drug abusers and nonabuser control populations. Several conclusions seem to be supported by the current status of these data:

1. The TaqI A1 and B1 DRD2 RFLPs are interesting reporters for events in significant portions of the DRD2 gene locus in Caucasians. Although the TaqI A and B RFLP sites are separated by chromosomal distances at which recombination events occurring randomly in the genome would have been expected to render them randomly associated with each other, or with the structural or regulatory regions of the dopamine receptor gene, substantial linkage disequilibrium does preserve the chromosomal connection between TaqI A and B RFLPs and, presumably, certain surrounding sites that could include DRD2 coding or regulatory sequences. Quantitative studies indicate that more than 95% of the possible linkage disequilibrium (D¢/Dmax) is maintained between these loci (50). Another polymorphic TaqI C RFLP locus, intermediate on the physical map of this region, nevertheless displays less linkage disequilibrium with the TaqI A and B sites that flank it. TaqI A and B genotypes could thus reliably mark a structural or functional gene variant at the DRD2 locus that could be directly involved in altering behavior.

2. A1 and B1 markers appear more frequently in drug abusers than in control populations in each of four currently available studies (10, 24, 48, 50, 64). Meta-analyses of these data suggest that differences between drug abuser and control populations are highly significant for both the four studies examining A1 and the two studies examining B1 frequencies.

3. The most severe abusers of addictive substances may manifest higher A1 and B1 DRD2 gene marker frequencies, while "control" comparison groups that are studied carefully to eliminate individuals with significant use of any addictive substance appear to display lower A1 and B1 frequencies than unscreened control populations (see ref. 66).

4. No data available to date derive from true population based sampling techniques. Although the theoretical possibility of false-positive error based on sample stratification thus remains, combining available data from nonpopulation based studies of Caucasians sampled in several centers appears to render unplanned stratification less likely.

Current meta-analyses based on the three studies reported in full suggest that drug abusers may display (a) a 2.4:1 odds ratio of drug abuse likelihood for individuals possessing an A1 allele and (b) a 3.3:1 odds ratio for those having a B1 allele [p < 0.001 in both cases; calculated per Smith et al. (66)].

Other genetic and environmental influences remain likely to determine the majority of variance in individual vulnerability to substance abuse.

Association studies for polymorphic markers at loci for other dopaminergic genes have also been carried out. Markers at the dopamine transporter, synaptic vesicular transporter, and tyrosine hydroxylase loci have not yielded positive allelic association results (6, 49, 53); C. Surratt, A. Persico, and G. R. Uhl, in preparation). However, none of these loci have been documented to display the striking linkage disequilibrium that could conceivably render the TaqI A and B DRD2 gene markers accurately reporters for hypothesized DRD2 gene variants at regulatory or coding sequences.

Workers investigating neurobiological factors in human addiction have limited ability to intervene in the genetics, personal environment, or neurobiology of drug abusers. In a complex disorder such as addiction, experimental approaches that can either manipulate or hold constant environmental and biological variables important for drug-taking behavior provide an opportunity to systematically examine the addiction process. Animal models can mimic human drug-taking behavior, perhaps the most basic aspect of addiction, to a remarkable degree. Virtually all drugs abused by humans will serve as positive reinforcers in animals under operant self-administration paradigms. Drug-related phenotypes such as dependence, tolerance, sensitization, and conditioned drug effects can also be experimentally analyzed separately in animal models with precision unavailable in clinical settings. Animal research can help to elucidate possible roles of genetic and environmental constituents in the addiction process that might otherwise be difficult to untangle.

Animal models of addiction are clearly limited in their ability to precisely mimic all aspects of the human condition. However, such studies provide experimental control of genetic and neurobiological manipulations essential to thoroughly explore mechanisms that contribute to vulnerability. Furthermore, they can allow environmental manipulations that would be ethically proscribed in human experimentation. Animal studies can thus help to define the limits of possible genetic contributions to human substance-abusing behaviors. Data from such studies can suggest areas that are likely to prove fruitful for future explorations in humans (see also Basic Concepts and Techniques of Molecular Genitics and Interactions Between the Nervous System and the Immune System: Implications for Psychopharmacology).

Several behavioral and genetic approaches have been used to determine genetic contributions to drug-related effects in experimental animals (Table 4). Most studies have been performed in rodents, due to the rich repertoire of genetically defined strains currently available and the facility with which transgenic mice can be created.

1. Correlations across inbred strains and outbred individuals assess the expression of each of several drug-related phenotypes in several genetically distinct strains and examine how the phenotypic traits covary across genotype. Alternatively, the same phenotypes can be examined in different individuals within genetically heterogeneous animal stocks, and the extent to which two or more phenotypes covary can be assessed (19).

2. Classic genetic analysis. Systematic cross-breeding experiments utilizing inbred strains and first filial "F1" generation individuals can be used to (a) determine the mode of inheritance of specific traits that may vary from parental strain to parental strain and (b) study gene–environment interaction. Congenic strain methodologies place gene(s) of interest from a donor strain into an acceptor strain while successively eliminating background donor genes by repeated backcrossing.

3. Selective breeding for a drug-related trait selects individuals of extreme phenotype for breeding over a number of generations to selectively isolate the genes responsible for the trait while randomizing irrelevant genes (19). Selective breeding can provide estimates of heritability (h2) and even biochemical and behavioral covariates of the selected trait.

4. Quantitative trait loci (QTL) analyses compare the strain distribution patterns of a drug-related response to the strain distribution patterns of molecular genetic markers. Multivariate analyses can determine the fraction of strain-to-strain differences in a trait that can be assigned to a particular locus. They also provide information about the location of the affected chromosomal locus. Substantial syntony is now documented between many mouse and human chromosomal regions (11). Although the presence of a behaviorally significant allelic variant in a mouse gene does not automatically predict a comparable behaviorally significant allelic variant in the corresponding human gene, data from such animal studies can provide testable hypotheses about possible roles for allelic variants at these loci that can be tested in human allelic association or other genetic studies.

5. Transgenic animals provide strains that overexpress or underexpress one or more introduced genes, so that behavioral consequences of over- or underexpression can be explored (38). As more and more candidate genes for involvement in substance abuse are identified, effects of altered gene structures or altered gene expression levels can thus be tested in vivo. In most transgenic animals, the chromosomal location of the introduced genetic material is random, so that examination of a single line of offspring of a single transgenic animal could provide confounded effects due to disruptions at the site of insertion (38). Homologous recombinant/embryonic stem cell technologies aid in targeting introduced genomic material to known sites whose disruption can thus be planned.

Studies investigating vulnerability to drugs in animals have now identified genetic components in several behavioral paradigms for most abused substances.

1. Opioids. Inbred mouse strains and selected lines have revealed genotype-dependent differences in virtually every opiate response including analgesia, locomotor stimulation, physical dependence, tolerance, respiratory depression, hypothermia, diuresis, and gastrointestinal motility (see ref. 5). Measures of the motivational effects of opioids such as conditioned place preference can also vary among different rodent strains; Cunningham et al. (14) demonstrated significant morphine-induced place preference that was greater in DBA/2J than in C57BL/6J mice. Quantitative trait locus analyses of inbred and recombinant inbred strains demonstrate association between morphine preference and polymorphic markers at the Es-1 region of murine chromosome 8 (4). Each of these studies clearly demonstrate genotype-dependent opioid preferences. Although each measure of drug-taking or drug-reinforced behavior differs in its interpretation, the overall results clearly indicate that many aspects of both opioid addiction and opioid side effects are influenced by genetic factors. The degree of genetic covariance among these opiate phenotypes can vary significantly, however, suggesting independent inheritance of at least several opiate phenotypes.

The combined data from each of these experimental approaches now overwhelmingly supports genotype-dependence of a variety of opioid effects in experimental animals, suggests probable involvement of several genes including those on murine chromosome 8, supports significant environmental influences even in individuals reared in similar fashions, and hint that the genetic bases of vulnerability to opioid reinforcement displays features that are not identical to the genetic bases for reinforcement due to other addictive substances.

2. Cocaine and amphetamine. Many aspects of cocaine and amphetamine responses, including locomotor stimulant, sensitizing, hyperthermic, hepatotoxic, and cardiovascular alterations can be influenced by genotype (for review see ref. 62). Inbred strain analyses have demonstrated significant effects of genotype on cocaine preference (1, 14, 26). Quantitative trait locus studies of amphetamine hyperthermia have localized genes responsible for strain-to-strain differences in this acute drug effect to the chromosome 1 region adjacent to the lamb2 locus (4). Psychostimulant abuse thus displays many signs of genetic predisposition in animal studies, with animal data supporting the same sorts of patterns of polygenic inheritance noted for opiates.

3. Other drugs. Inbred mouse strains and animal lines derived from selection studies have revealed genotype-dependent differences in drug abuse phenotypes. Genotype-dependent differences in locomotion induced by phencyclidine in recombinant inbred strains (21) demonstrate varying responses to NMDA or sigma receptor blockade. The cannabanoid D9-THC altered thresholds for intracranial self-stimulation in Lewis but not in F344, Sprague–Dawley, or Wistar rats (22). Lewis rats were also reinforced by diazepam, whereas F344 rats were not reinforced by the drug (37).

Although the reinforcing properties of abused drugs can vary as a function of genotype, evidence for common genetic bases for drug-taking behavior across drug classes and especially across behavioral paradigms is more mixed. C57BL/6J mice and Lewis rats self-administer ethanol, cocaine, and opioids, whereas rodents of several other genotypes will derive conditioned reinforcement from only a single drug class (reviewed by ref. 62).

Many differing risk or protective factors may be involved in the acquisition of drug-taking behavior, and these factors may display different mechanisms than those implicated in the maintenance, extinction, or relapse of drug-taking behavior. Genetic approaches in animals can help to suggest possible specific inherited behavioral, neurochemical, neuroendocrinological, or other features that could provide mechanisms for each of these drug related behaviors.

1. Behavioral correlates. Several inherited behaviors and stress responses manifest in animals prior to drug administration have now been correlated with the same strains' drug responses. If animals of different genotypes are segregated into two groups based upon a non-drug phenotypic response and if each group is then tested for drug-seeking behavior, the ability of the primary phenotype to predict subsequent drug-seeking behavior can be assessed (54). Animals segregated from populations of randomly inbred rats based on their high locomotor responses to novel environments acquire amphetamine self-administration behavior more rapidly than those with low locomotor responses to novel situations (54). Rats selectively bred for high rates of intracranial self-stimulation or for high "emotional reactivity" in open-field situations displayed stronger opiate-induced conditioned place preferences and two-bottle preference than did rats bred for the opposing characteristics (e.g., see ref. 60). Significant differences in operant self-administration of opioids that can be positively correlated with baseline locomotor activities have now been identified in several mouse and rat strains (2, 18).

2. Neurochemical correlates. Belknap and Crabbe (4) have used recombinant inbred strains to examine relationships between opioid-induced behaviors and brain opiate receptor densities or activities. In addition, lines of mice selectively bred for opiate-induced analgesia demonstrate significant differences in mu-opioid receptor binding and distribution (see ref. 5). Recombinant inbred mouse strains displaying fewer m1-opiate receptors also manifest less potent opiate responses than do other mouse strains including self-administration, while mice with supernormal opiate receptor binding densities are also more sensitive to several opiate drug effects (18 see references in ref. 5).

3. Neuroendocrinological correlates. Rats with high locomotor responses in novel environments display both prolonged corticosterone release and enhanced self-administration of amphetamine (55). Exogenous corticosterone administration can also induce more rapid acquisition of amphetamine self-administration (55). Because genotype can significantly affect individual stress responses in animals and humans, these data may thus have significant implications for human drug abuse vulnerabilities. However, inbred rat strains with exaggerated and blunted hormonal responses to stress do not demonstrate the predicted pattern of opioid, ethanol, cocaine, or diazepam self-administration behavior (e.g., see refs. 2, 27 and 37).

4. Chronic drug effects. Tolerance to, and repeated withdrawal from, drugs of abuse are important aspects of many human addictions. The chronic effects of a drug involve behaviors present in the maintenance and extinction phase of drug use that may be highly relevant for drug use in humans. However, mechanisms underlying chronic drug effects are likely to differ substantially from those underlying the acute effects of even the same drugs, although some common mechanisms might be identified as well.

Studies using selectively bred murine lines suggest that the genes which determine ethanol withdrawal severity can be involved in the ability of ethanol to sustain conditioned place preferences (12). Behavioral sensitization, one possible animal model of certain components of drug addiction, can also be studied in relationship to genotype and to the acquisition of drug-reinforced behaviors. Repeated administration of cocaine results in various degrees of sensitization that depend on the genotype (e.g., see ref. 17). Interestingly, the strain that most readily displays a sensitization response to cocaine also shows significant drug-reinforced behaviors mediated by cocaine, ethanol, and opioids (e.g., see ref. 18). The degree of sensitization following repeated administration of amphetamine can predict the rate of acquisition of amphetamine self-administration behavior (54). Repeated amphetamine administration sensitizes subsequent amphetamine responses in low-activity rats that do not readily acquire self-administration behavior but not in high-activity rats that readily acquire this behavior (54). Interestingly, the sensitization process enhances the rate of acquisition of amphetamine self-administration behavior in low-reactivity rats to the level demonstrated in high-reactivity rats (54, 55).

5. Environmental factors which can contribute to variance in drug-related behaviors can also be appropriately isolated and studied in animals. Such features as drug availability, schedule of drug reinforcement, and drug history can be conveniently manipulated in animals in a fashion not possible in humans (see ref. 47).

Animal studies suggest that interindividual differences at several gene loci, together with substantial environmental contributions, can mediate quantitative and qualitative differences in the behavioral effects of almost all drugs of abuse. Shared genetic components, and some differentiable genetic components, could underlie the acquisition, maintenance, and resistance-to-extinction characteristic of addictions to a number of abused substances in humans.

Studies in humans support a substantial role for genetic differences in vulnerability to human substance abuse. No data obtained to date are inconsistent with the picture emerging from animal studies: Actions of alleles at several gene loci are likely to interact with environmental factors to yield substance abuse vulnerabilities. Some genes' alleles may predispose to abuse of multiple substances, whereas others may yield preferential vulnerability to more specific classes of drugs. While molecular genetic studies of family pedigrees may be able to identify genes producing shared vulnerabilities to both alcoholism and drug abuse, secular trends in abuse substance fashion and availability will be more likely to make allelic association studies increasingly prominent in identifying any gene alleles that manifest greater impact on drug abuse than on alcoholism. Data from both animal and human studies will probably be necessary to unravel the complex interactions between multiple genes and environment likely to contribute to the pathogenesis of common and disabling human addictive disorders.

We are indebted to collaborators in this work, including Bruce O'Hara, Antonio Persico, Brian Suarez, and Stevens Smith; to Antonio Persico, Lucinda Miner, Charles Schuster, and Rodney Marley for helpful comments; and to Stevens Smith for help with meta-analyses. Support for the work from the authors' laboratories derives from the intramural program of the National Institute on Drug Abuse.

published 2000