tive coping strategies and produce lifelong behavioral substance abuse patterns.

In humans, twin models have been used to explore the relative contribution of genetic and environmental factors to substance use and dependence. This is accomplished by comparing concordance rates for a particular trait in monozygotic twins who share all of their genes in common to those for dizygotic twins who share roughly 50% of their genes (Kendler, 2001). As described in greater detail below, this methodology has been used to study the role of heritable factors for smoking, alcohol use, and use of illegal drugs.

Once a particular behavioral trait (also referred to as a “phenotype”) has been established as heritable, molecular genetic approaches are used to identify the specific genetic variants that may be responsible. One such approach identifies candidate genes based on neurobiological or biochemical pathways (e.g., dopamine or serotonin genes) and uses a case–control study design to compare the frequency of genetic variants (alleles) in these pathways among persons with and without the phenotype (e.g., nicotine-dependent persons vs. nondependent persons) (Sullivan, Jiang, Neale, Kendler, & Straub, 2001). Several studies employing the candidate gene approach to investigate substance abuse genes are described below. The role of specific genetic variants can be also investigated through family-based designs that examine allele sharing or allele transmission for candidate genes within families (Spielman et al., 1996). This latter approach controls for potential bias due to ethnic admixture, but has less statistical power and is more costly to implement.

In contrast to these hypothesis-driven approaches, genetic linkage analysis can be used to search for as yet unidentified genetic variants that may be linked with substance use phenotypes. In this approach families or relative pairs (e.g., sibling pairs) are used to look for linkage with anonymous markers across the genome. Because the effect sizes of any individual gene conferring susceptibility to a behavioral trait are expected to be small (Comings et al., 2001), this approach requires a large number of family members. As described below, the results of such studies are beginning to reveal regions of interest in the genome; however, it is likely to take several years before specific loci are identified and validated as being important in substance use and dependence.

Below, we summarize the literature on the heritability and specific genetic effects for tobacco use, alcohol use, and use of illegal drugs. Although most of these studies used adult populations, we highlight investigations that included adolescent participants. The results of both adolescent and adult studies provide insights into the biobehavioral basis of substance use and its relevance to prevention in high-risk youth.

Abundant data from twin studies provide evidence for the heritability of cigarette smoking. Using the Australian twin registry, Heath and Martin (1993) found that inherited factors accounted for 53% of the variance in smoking initiation. More recent data suggest that the heritability of a diagnosis of nicotine dependence is even higher (Kendler et al., 1999; True et al., 1999). Sullivan and Kendler (1999) summarized data from a large number of twin studies indicating that additive genetic effects account for 56% of the variance in smoking initiation and 67% of the variance in nicotine dependence. Significant genetic influences have also been documented for age at smoking onset (True et al., 1999) and for smoking persistence (Madden et al., 1999).

Genes in the dopamine pathway have been studied most extensively with respect to tobacco use and addiction. It is speculated that individuals with low-activity genetic variants may experience greater reinforcement from nicotine because of its dopamine-stimulating effects. In support of this hypothesis are three studies showing a higher prevalence of the more rare A1 or B1 allele of the dopamine 2 receptor (DRD2) gene among smokers than among nonsmokers (Comings et al., 1996; Noble et al., 1994; Spitz et al., 1998). However, a small, family-based analysis did not provide evidence for significant linkage of smoking to the DRD2 locus (Bierut et al., 2000).

In a case–control study of smokers and nonsmokers, Lerman and colleagues (1999) found that DRD2 interacted with the dopamine transporter (DAT) gene in its effects on smoking behavior. The DAT polymorphism is of particular interest because the 9-repeat allele has been associated with a 22% reduction in dopamine transporter protein (Heinz et al., 2000). Since a reduction in dopamine transporter level would result in less clearance and greater bioavailability of dopamine, it is speculated that individuals who have the 9-repeat may have less need to use nicotine to stimulate dopamine activity. The association of the DAT gene with smoking behavior has been supported in one study (Sabol et al., 1999), but not replicated in two other studies (Jorm et al., 2000; Vandenbergh, 2002). Thus, the role of DAT in smoking behavior remains unclear.

The serotonin pathway is also under investigation in genetic studies of smoking behavior. Candidate polymorphisms (genetic variants) include those in genes that are involved in serotonin biosynthesis (e.g., tryptophan hydroxylase, TPH) and serotonin reuptake (serotonin transporter, 5HTTLPR). Two recent studies have shown that individuals who are homozygous for the more rare A allele of TPH are more likely to initiate smoking and to start smoking at an earlier age (Lerman et al., 2001; Sullivan et al., 2001). Although 5HTTLPR was not associated with smoking status (Lerman et al., 1998), there is evidence from two studies that this polymorphism modifies the effect of anxiety-related traits on smoking behavior (Hu et al., 2000; Lerman et al., 2000).

While genes in the dopamine and serotonin pathways may have generalized effects on risk for substance abuse, genes that regulate nicotine metabolism should be specifically relevant to smoking behavior. One hypothesis is that slower metabolizers of nicotine may be less prone to initiate smoking because they may experience more aversive effects (Pianezza, Sellers, & Tyndale, 1998). Once smoking is initiated, slower metabolizers may require fewer cigarettes to maintain nicotine titers at an optimal level (Benowitz Perez-Stable, Herrera, & Jacobs, 2002). Initial support for this premise was provided in a study of the P450 CYP2A6 gene, which encodes the key enzyme involved in metabolism of nicotine to inactive cotinine (Pianezza et al., 1998). Unfortunately, however, later studies did not support this finding and suggested that the CYP2A6 variant is much more rare than originally reported (London, Idle, Daly, & Coetzee, 1999; Oscarson et al., 1998; Sabol et al., 1999).

Although genes regulating nicotine receptor function would be prime candidates for smoking risk, data on functional genetic variation in humans are not yet available. In two recent studies of the B2 nicotinic receptor, several single neucleotide polymorphisms (of unknown functional significance) were identified, but none were associated with smoking behavior (Lueders et al., 2002; Silverman et al., 2001).

As mentioned above, linkage analysis can also be used to scan the genome for regions that may harbor nicotine dependence susceptibility genes. There are, however, a limited number of reports using this approach. Straub and colleagues (1999) performed a complete genomic scan to search for loci that may confer susceptibility to nicotine dependence. Using a sample of affected sibling pairs, linkage analysis provided preliminary evidence for linkage to regions on chromosomes 2, 4, 10, 16, 17, and 18. But these results were not statistically significant, and the sample size in this study (130 families) may not have been large enough to identify genes with small effects. Two other studies used families from the Collaborative Study on the Genetics of Alcoholism (COGA) and reported evidence for linkage of smoking behavior to chromosomes 5 (Duggirala et al., 1999), 6, and 9 (Bergen et al., 1999). Notably, the regions identified in the different studies do not overlap. This may be attributable to the fact that regions identified in the COGA sample may harbor loci predisposing to addiction to both alcohol and smoking.

As with tobacco, twin studies of alcohol use provide consistent evidence for significant genetic effects. Estimates for the proportion of variance accounted for by genetic factors range from about 30% to 70%, depending on whether the studies used population-based or treatment sam

ples (Kendler, 2001) and on the specific phenotype examined (van den Bree, Johnson, Neale, & Pickens, 1998). One study of over 1,500 twin pairs, ages 20 to 30 years old, reported that 47% of the variance in use (vs. abstinence) in males was attributable to genetic factors with 48% of the variance being due to shared environment (and the remainder due to individual environmental effects; Heath & Martin, 1988). The comparable figures for females were 35% and 32%, respectively. Heritability estimates in other studies ranged from about 50% for alcohol dependence to 73% for early age of onset for alcohol problems (McGue, Pickens, & Svikis, 1992; Pickens, & Svikis, 1991; Prescott & Kendler, 1999). Physical symptoms of alcohol dependence also appear to have a significant heritable component (e.g., binge drinking, withdrawal), although the potential behavioral consequences appear to be less heritable (e.g., job trouble, arrests; Slutske et al., 1999). Of particular relevance to the biobehavioral model of substance abuse, there is evidence for shared genetic influences for tobacco and alcohol consumption (Swan, Carmelli, & Cardon, 1996).

The search for specific genetic effects on alcohol use has led to the discovery of genes in key neurotransmitter pathways and genes that influence the metabolism of alcohol. Once again, the dopamine pathway has been a central focus of this research. An initial study relating the DRD2 A1 allele to alcoholism attracted a great deal of attention (Blum et al., 1990); however, several studies failed to replicate this initial result (Bolos et al., 1990). Noble (1993) reviewed nine independent studies including 491 alcoholics and 495 controls. Across these studies, the more rare A1 allele of the DRD2 gene was carried by 43% of alcoholics, compared with 25% of nonalcoholic controls. When only severe alcoholics were examined, the prevalence of the A1 allele was 56%. Hill, Zezza, Wipprecht, Locke and Neiswanger (1999) used the more conservative family-based approach to test for linkage between DRD2 and alcoholism. Although an overall association with alcoholism was not supported, there was evidence for linkage when only severe cases were examined. Studies examining other genes within the dopamine pathway for association with alcoholism have yielded mostly negative results (Parsian, Chakraverty, Fishler, & Cloninger, 1997).

Genes in the serotonin pathway are also plausible candidates for alcohol dependence because of the effects of alcohol on brain serotonin levels (Lesch & Merschdorf, 2000). The low activity S allele of the serotonin transporter gene (5HTTLPR) has been linked with alcoholism in one family-based study (Lichtermann et al., 2000). Although the prevalence of this variant has not been found to differ significantly in case–control studies comparing alcoholics and nonalcoholics, there is evidence that it increases risk for particular alcoholism subtypes, including binge drinking (Matsushita et al., 2001) and early-onset alcoholism with violent features (Hallikainen et al., 1999). Similarly, the TPH gene has been linked with alcoholism with comorbid impulse control problems, such as antisocial behavior or suicidal tendencies (Ishiguro et al., 1999; Nielsen et al., 1998).

The most consistent evidence for genetic effects on alcoholism has been generated from studies of genes that regulate the metabolism of alcohol. Alcohol is converted to its major metabolite acetaldehyde by the enzyme alcohol dehydrogenase (ADH). Decreased metabolism results in more aversive effects of alcohol consumption, such as flushing and toxicity. A reduced-activity allele of the ADH2 gene (ADH2*2) is found more commonly in Asian populations and has been shown to be protective for alcohol dependence in Chinese (Chen et al., 1999) and European (Borras et al., 2000) populations. There is some evidence that the genetic effect is stronger for males than for females (Whitfield et al., 1998). The reduced-activity allele of ADH2 is also found more commonly in Ashkenazi Jewish populations and has been associated with reduced alcohol consumption among Jewish college students (Shea, Wall, Carr, & Li, 2001).

The opioid system has also been implicated in the reinforcing effects of alcohol as well as other drugs of abuse (see below). With respect to alcoholism, the results of initial studies have been mixed. Two studies have suggested that variants of the μ-opioid receptor gene may be associated with a general liability to substance dependence, including alcohol (Kranzler et al., 1998; Schinka

et al., 2002). However, another larger study did not find significant differences in allele frequencies in dependent and nondependent individuals (Gelernter, Kranzler, & Cubells, 1999).

The Harvard Twin study is one of the most extensive investigations of the role of heritable factors in drug use (Tsuang, Bar, Harley, & Lyons, 2001; Tsuang et al., 1999). Summarizing the results from 8,000 twin pairs, Tsuang and colleagues (2001) reported heritability estimates ranging from .38 for sedative drugs to .44 for stimulant drugs. Interestingly, the variance in illicit drug use attributed to shared environmental influences tended to be much smaller than that due to individual environmental effects. Somewhat higher estimates for heritability were generated from a study of twins ascertained through alcohol and drug programs and thus exhibiting more severe forms of substance abuse disorders (van den Bree et al., 1998). Among males, heritability estimates for substance dependence were 58% for sedatives, 57% for opiates, 74% for cocaine, 78% for stimulants, and 68% for marijuana. With the exception of dependence on stimulants, estimates were significantly lower for females. In general, the genetic variance appeared to be greater for heavy use or abuse than for ever using (Kendler, Gardner, & Gardner, 1998).

Of particular relevance to youth substance abuse is the finding that the transitions in survey drug use categories (never used to ever used to regular use) have a significant heritable component. For example, genetic variance for the transition from never to ever using was reported to be 44% for marijuana, 61% for amphetamine, and 54% for cocaine (Tsuang et al., 1999). For the transition to regular use, the comparable figures were 30%, 39%, and 34%. As was shown for tobacco and alcohol, family studies showed evidence for common genetic variance underlying dependence on illegal drugs (Pickens, Svikis, McGue, & LaBuda, 1995; Tsuang et al., 2001).

Because many drugs of abuse increase levels of dopamine (Dackis & O'Brien, 2001; Shimada et al., 1991), initial genetic investigations have fo cused on this pathway. Genetic variations affecting mesocorticolimbic function might affect drug-induced reward and thereby contribute to addiction vulnerability. Uhl, Blum, Noble, and Smith (1995) summarized data from nine studies of mixed groups of substance abusers and reported a 2-fold increase in risk in individuals who have at least one copy of the DRD2 A1 allele. The risk ratio was nearly 3-fold for more severe substance abuse. A high activity allele of the catechol-O-methyltransferase gene, which codes for a dopamine metabolizing enzyme, has also been associated with polysubstance abuse (Vandenbergh, Rodriguez, Miller, Uhl, & Lachman, 1997).

Several twin, family, and adoption studies have concluded that the vulnerability to develop heroin dependence is partially inherited. Twin studies have reported significantly higher concordance rates for identical twins than for nonidentical twins (Tsuang et al., 1996), and an estimated heritability of .34 has been published for heroin-dependent males (van den Bree et al., 1998). One study of male and female subjects who were adopted away from their natural parents found that opioid dependence correlated with genetic loading for antisocial personality and alcoholism, and with environmental factors such as divorce and turmoil in the adoptive family (Cadoret et al., 1986). Another study reported that subjects with opioid dependence had an 8-fold increase in addiction prevalence among their first-degree relatives, independent of alcoholism and antisocial personality disorder, with evidence of specificity for familial opioid dependence (Merikangas et al., 1998). Therefore, a family history of addiction appears to be a potent risk factor for the development of opioid dependence. Specific genes associated with increased vulnerability for heroin dependence have not been identified, although animal models demonstrate that genes encoding the μ-opioid receptor might influence the animal's opioid preference, as evidenced by their willingness to self-administer morphine (Berrettini, Alexander, Ferraro, & Vogel, 1994). However, human studies of the gene (OPRM1) that encodes the human μ-opioid receptor are mixed with regard to opioid dependence vulnerability (Crowley et al., 2003; Hoehe et al., 2000).