Analyses for ANYDEP were performed using the Mantel–Haenszel χ2 test of association assuming an additive genetic model. Analyses for QUANTDEP were performed using analysis of covariance employing substance, genotype, and the substance × genotype interaction to test for differences in genotype by substance. Gender and birth cohort were included as covariates.Independent replication of SNPs demonstrating evidence of significant association in the COGA sample was evaluated in the Study of Addiction: Genetics and Environment sample. SAGE is a case-control sample comprised of three complementary studies: COGA, the Family Study of Cocaine Dependence and the Collaborative Genetics Study of Nicotine Dependence . There were 129 individuals from the 118 COGA families in the current study that overlapped with the SAGE sample, and were removed from the SAGE replication dataset. The remaining independent SAGE sample used for replication was limited to 2647 individuals of European-American descent. Factor analysis scores from Mplus were independently estimated for this study, as described above, based on DSM-IV dependence criteria for the four substances. Analyses were implemented in Plink and included age at interview and gender as covariates.The number of individuals utilized for the categorical phenotype ANYDEP was 1770 .Nearly half the sample met the DSM-IV criteria for at least one substance . The COGA sample was ascertained through an alcohol-dependent individual in treatment and families were selected for the highest density of alcohol dependent members; therefore,mobile grow system it was expected that there would be many individuals meeting criteria for alcohol dependence . In addition, 19% met criteria for cannabis dependence . The rates for cocaine and opioid dependence were lower .

There were 832 individuals that met criteria for at least one substance dependence diagnosis; of those, 312 endorsed at least two diagnoses. Alcohol and cannabis dependence were the most common .The number of individuals included in the analysis of the quantitative factor score QUANTDEP was 2,183 . The confirmatory one-factor model fits the data well in COGA [comparative fit index = 0.96; root mean square error of approximation = 0.07], and in the replication sample, SAGE , supporting our proposed unidimensional conceptualization of dependence criteria for alcohol, cannabis, cocaine and opioids. Factor loadings in the COGA sample ranged from 0.67 to 0.99 and were highly consistent across COGA and SAGE. In general, factor loadings for alcohol and cannabis dependence criteria were lower, and ranged between 0.67 and 0.85, while those for cocaine and opioids were uniformly high . Data across drug classes and across criteria loadings for all seven criteria for the four substances are available in Supporting Information Fig. S1.This is one of the first GWAS to test for the association of overall substance dependence phenotypes, defined both categorically and quantitatively . This approach implicitly tested the hypothesis that there are genes with pleiotropic effects contributing to dependence on alcohol, cannabis, cocaine and opioids. Using these multi-substance phenotypes, we detected genome-wide significant results with SNPs in two different genes. This finding is consistent with an extensive twin literature that provides demonstrable support for common genetic liability underlying addiction to multiple substances . Furthermore, a previous study in a slightly different COGA sample demonstrated aggregation of drug dependence in relatives of alcohol-dependent probands, even after controlling for co-morbidity in the probands . Genome-wide significant association for ANYDEP was observed with a SNP in an uncharacterized gene, LOC151121 . Further evidence of association was corroborated by surrounding SNPs, both genotyped and imputed. Nominal replication was found in the SAGE sample with the same phenotype. This SNP was moderately associated with QUANTDEP and also with the number of DSM-IV alcohol dependence criteria endorsed in another related study with data from the same sample . Similar to the replication results here, this SNP was nominally associated with the alcohol dependence symptom count in the SAGE sample as well . Significant association was also detected with QUANTDEP for the SNP rs2567261 in ARHGAP28 . Further evidence of association was observed with both genotyped and imputed SNPs within the gene. ARHGAP28 is also known as Rho GTPase activating protein 28.

GTPase-activating proteins target GTPases, and are mediated by exposure to alcohol, cannabis, cocaine and opioids. For example, Rho1 and Rac moderate the stimulating and sedative effects of acute ethanol intoxication in Drosophila . Thus, there is strong biological rationale for this gene as a potential candidate for substance dependence. Of note, this SNP was modestly associated with ANYDEP in this sample and with previously published alcohol symptom count in the COGA family sample . However, rs2567261 was not significantly associated with alcohol symptom count in the SAGE sample, although there was a trend in that direction . Although the association did not replicate in the SAGE sample for this phenotype, there was a trend toward association with the other phenotype, ANYDEP . This weak replication for a different phenotype may be due to the fact that the majority of the SAGE sample was ascertained on nicotine and cocaine dependence, whereas the COGA sample was recruited based on an alcohol-dependent proband and expanded to include the maximum number of alcohol-dependent family members. The exclusion of nicotine dependence criteria may have attenuated the likelihood of replication, given ascertainment for nicotine dependence in SAGE. Since the SNP association was not primarily due to alcohol dependence, it is possible that high rates of co-morbidity in the COGA sample as compared with the SAGE sample contributed to the finding. In addition, the family-based test in the large COGA families with co-morbidity may have had greater power to detect the association. QUANTDEP seems to represent an underlying severity of addiction. As seen in Fig. 3a, the higher the number of co-morbid diagnoses, the higher the QUANTDEP scores. Overall, for the QUANTDEP measure, the loadings for the alcohol and cannabis criteria appear generalizable to general population studies ; in contrast, those for cocaine and opioid dependence are higher . In COGA, cocaine and opioid dependence criteria were less commonly endorsed than those for alcohol and cannabis , with the former also showing less range in endorsement rates . In SAGE, cocaine dependence criteria were somewhat more commonly endorsed than cannabis dependence criteria, yet the cocaine criteria had higher loadings than the cannabis criteria, identical to COGA. For both cocaine and opioids, the range of prevalence of individual criteria was highly restricted . Thus, in both COGA and SAGE, the likelihood of endorsement of each of the seven dependence criteria for cocaine and opioids was similar while certain alcohol and cannabis criteria were endorsed more often than others . This may be related to the ascertainment strategy and over-representation of family history for alcoholism in both samples. Nonetheless, all factor loadings were high, indicating that QUANTDEP reflects a general liability to dependence across multiple substances. In particular, QUANTDEP captures the liability to cocaine and opioid dependence criteria in these two studies. Therefore, in addition to being an index of severity and a measure of general liability to addiction across alcohol and drugs, QUANTDEP likely also reflects variation in prevalence and the expected pattern of co-morbid relationships and co-aggregation across alcohol and drug dependence criteria in these subjects ascertained on specific substance dependence. While the heritability of the binary phenotype of dependence on any substance was similar in this sample to the most common heritability estimate reported for any of the four substances from twin studies , the heritability for the quantitative phenotype was much higher .

This is consistent with one prior twin study of a latent genetic factor underlying alcohol and drug problems as well as measures of impulsivity and conduct problems but significantly higher than some others . Although this high heritability should not be over-interpreted , it is possible that the use of a multi-variable quantitative phenotype, utilizing the pattern of endorsement of the seven DSM-IV criteria across all four substances, captured valuable genetic information across the vulnerability spectrum. The two phenotypes used in this study were both aggregate measures of overall dependence. Although their top genetic signals did not overlap , there was evidence of association for the other phenotype for the two SNPs that attained genome-wide significance . The difference in magnitude of P-value is not surprising given the arguable validity of the diagnostic cutoff implemented in DSM-IV, which likely excluded from affected status ,mobile vertial rack a number of individuals who may have met criteria for abuse or endorsed 1–2 dependence symptoms across one or even multiple drugs, and thus did not qualify for dependence. Viewed alternatively, the unaffected individuals for ANYDEP represent a heterogeneous group varying in severity. Such variability was better captured by QUANTDEP, which while not taking abuse criteria into account, was a better approximation of the range of vulnerability to substance-related problems. Thus, it is likely that ANYDEP reflects the more severe of the QUANTDEP scores. Finally, the possibility that our findings reflect false positives cannot be excluded. There have been multiple prior GWAS that have utilized symptom counts and factor scores of alcohol dependence criteria but only one attempted to combine indices of alcohol , nicotine and drug misuse using hierarchical factorial analyses for GWAS. In that study, McGue and colleagues reported on four SNPs associated with multiple first-order and higher order externalizing factors. One of these SNPs, rs10037670 in GALNT10, with the highest association for illicit drug dependence factor was modestly associated in this study with both ANYDEP and QUANTDEP . There are several strengths of this study design, the first being the use of families densely affected with alcohol-dependent individuals. Family and twin studies suggest familial co-aggregation and heritable overlap across alcohol, cannabis, cocaine and opioids. Thus, this family-based COGA sample, enriched for dependence on multiple substances and shared genetic risk, allowed us to test for the association of common variants with risk for dependence across multiple substances. A second strength of this study was the use of a family-based association design. This allowed us to examine association within a family consisting of members who endorsed criteria for dependence on different substances. Third, family-based analysis is robust to population substructures such as nuanced differences in ethnicity, which might occur with marry-in individuals of a different race, and in turn, affects the genetic diversity of the offspring. Three caveats are worth considering.

First, only a small subset of individuals met DSM-IV criteria for opioid and cocaine dependence. Thus, it is possible that results pertain more closely to lower liabilities to these substances. Second, we did not include nicotine dependence criteria in this analysis. As we were interested in a confirmatory model of unidimensional genetic risk, we elected to exclude nicotine symptoms based on published evidence for a preponderance of non-overlapping genetic influences on these criteria. Finally, we elected not to utilize abuse criteria , despite DSM-5-related changes. Previous findings in COGA families demonstrated that abuse did not aggregate in relatives of alcohol-dependent probands . In addition, the extant psychometric literature suggests that with the exception of hazardous use, which is frequently endorsed to the exclusion of other abuse or dependence criteria, the remaining abuse criteria and craving are infrequently endorsed in the absence of co-occurring dependence criteria. This is particularly true in samples ascertained for substance use disorders, such as SAGE and COGA. For instance, in SAGE, of those who reported no alcohol dependence criteria, only 12–26 individuals endorsed at least one abuse criterion other than hazardous use . Hence, it is unlikely that the exclusion of abuse criteria resulted in our inability to capture a relevant portion of the liability continuum. In summary, this study provides evidence that there are common variants that contribute to the risk for a general liability to substance dependence, defined qualitatively and quantitatively. The results of this study require replication in independent samples to further explore whether overall dependence on multiple or individual substances is associated with the SNPs in these regions.This national Collaborative Study is supported by NIH Grant U10AA008401 from the National Institute on Alcohol Abuse and Alcoholism and the National Institute on Drug Abuse . Funding support for GWAS genotyping, which was performed at the Johns Hopkins University Center for Inherited Disease Research, was provided by the National Institute on Alcohol Abuse and Alcoholism, the NIH GEI and the NIH contract ‘High throughput genotyping for studying the genetic contributions to human disease’ . A.A. is supported by K02DA32573 and AAR21235 and J.E.S. by F32AA22269. Funding support for the Study of Addiction: Genetics and Environment was provided through the NIH Genes, Environment and Health Initiative [GEI] .