RGS17 (RGS-Z2) and RGS20 (RGS-Z1) are two members of the RGS-Rz subfamily of GTPase-activating proteins (GAP) that efficiently deactivate GalphazGTP subunits, and thereby turn off the signaling pathway of G protein-coupled receptors (GPCRs), including opioid receptors. Considering their intimate interactions with opioid receptors (mainly the mu-receptor) and their gene locations (RGS17 is in the vicinity of OPRM1 and RGS20 is in the vicinity of OPRK1), the two genes (RGS17 and RGS20) encoding them are both positional and functional susceptibility candidate loci for SD. We found that multiple RGS17 SNPs were associated with multiple SD phenotypes in both AAs and EAs. However, none of the eight RGS20 SNPs were associated with any of the four dependence traits.
Although variation in RGS17 influences susceptibility to multiple dependence traits in both AAs and EAs, our results suggest that different mechanisms may be operative in some cases. SNPs rs596359 (in the promoter region) and rs6931160 (in intron 1) were associated with at least one of the four dependence traits in both populations. SNPs rs9397585, rs1933258 and rs9371276, which are all located in intron 1 and the same haplotype block (Block II) (Figure 2), were associated with one of the four dependence traits only in EAs. SNPs rs516557 (in intron 2) and rs545323 (in intron 4) were associated with one of the four dependence traits in only AAs (Table 3). Analyses of haplotypes harboring these SNPs supported the individual SNP findings. Haplotypes GATTC and GTTCT, containing alleles (underlined) of SNP rs545323 (Figure 2, haplotype Block I) were associated with OD and MjD, respectively, only in AAs, after permutation tests to correct for multiple comparisons. Haplotype CCC C, containing alleles (underlined) of the three SNPs (rs9397585, rs1933258 and rs9371276) located in intron 1 and haplotype Block II (Figure 2) was associated with OD only in EAs, after permutation tests to correct for multiple comparisons. These findings suggest that the population-specific associations were dependent on the location of variants in RGS17.
The most statistically significant result was obtained for SNP rs596359, which is located in the promoter region of RGS17. Chi-square tests, logistic regression analyses, and permutation tests all showed a positive association between SNP rs596359 and all four SD phenotypes in both populations (Table 3). Further, meta-analyses that combined data from both AAs and EAs showed that SNP rs596359 yielded an odds ratio from 1.28 to 1.33 for risk of all four SD traits (data not shown). Specifically, the G allele of SNP rs596359 was significantly more frequent in cases than in controls in both populations (Additional file 1: Tables S1 and S2). Thus, this promoter variant may increase the risk for SD by influencing RGS17 transcription. To validate the functional effect of this promoter variant on RGS17 transcription, we analyzed the correlation of rs596359 genotypes and RGS17 mRNA expression levels in lymphoblastoid cell lines from both CEU and YRI subjects recruited for the HapMap project (http://hapmap.ncbi.nlm.nih.gov/). The G allele of SNP rs596359 showed a dose-related decrease in RGS17 transcription by decreasing mRNA expression levels (Figure 3). Moreover, bioinformatic analyses indicated that substitution of the A allele for the G allele at rs596359 site generated a transcription binding site in the promoter region of RGS17 for transcription factor AML1a. This transcription factor has a higher affinity for DNA-binding than AML1b, but lacks the putative transcriptional activation domain that is possessed by AML1b. Thus, AML1a dominantly suppresses the transcriptional activity exerted by AML1b . Several other studies have demonstrated that AML1a inhibited erythroid or granulocytic differentiation [37, 38]. Based on these findings, we would speculate that rs596359 G allele carriers have lower RGS17 activity and thus greater synaptic neurotransmission and rewarding function mediated by GPCRs such as opioid receptors. To test this hypothesis, the influence of SNP rs596359 on RGS17 promoter activity should be measured using other approaches (e.g., luciferase reporter gene assays).
None of the eight RGS20 SNPs showed significant association with any of the four SD phenotypes in either AAs or EAs. There are three possible explanations for this lack of association. First, the RGS20 SNPs selected for this study may have a minor or undetectable effect on SD. Fine-mapping of this gene could identify variants showing a stronger association with SD traits. Second, RGS20 may have a weak effect on susceptibility to SD due to its being physically linked to OPRK1, which has a less important role than OPRM1 (which is physically linked to RGS17) in mediating the rewarding effects of alcohol or drugs [17, 19]. Third, similar to OPRK1, which mediates the psychotomimetic effects of some drugs , RGS20 may mainly regulate other biological activities than SD. Further studies are warranted to determine whether RGS20 is a susceptibility gene for SD.
The present study has several limitations. First, our finding is limited by the relatively small size of the control sample. Moreover, we did not control for prior genotyping performed on this sample in multiple testing corrections because we were concerned that overly conservative results might be obtained. Second, SD frequently co-occurs with Axis I disorders (e.g., depression and anxiety disorders) and Axis II disorders (i.e., personality disorders). Thus, our findings of an association between RGS17 variants and SD may be cofounded by comorbid disorders. Third, given the close relationship between the RGS-Rz (RGS17 and RGS20) and opioid receptor (OPRM1 and OPRK1, respectively) genes, gene-gene interaction analyses should be conducted. We would speculate that strong gene-gene interaction effects (e.g., of OPRM1 and RGS17) on SD would be detectable. Even though variation at RGS20 did not show significant association with SD in individual gene analysis, interaction effects of that gene with OPRM1 or OPRK1 on SD risk may exist. Fourth, in this study, we ignored polymorphisms in exonic regions because they are rare in the genes examined. There is only one known SNP rs2295230 (synonymous) in RGS17 exon 2 that had a minor allele frequency greater than 5% in AA and EA populations. Exonic SNP rs2295230 is in tight LD with intronic SNP rs9371276 (which was included in this study) (CEU: D’ = 0.96, r2 =0.92; YRI: D’ = 0.86, r2 = 0.54, using genotyping data from the 1000 Genomes Project). It is also situated close to SNP rs2295230. Thus, exonic SNP rs2295230 was not considered in the present study. As we know, rare variants in coding regions may have a larger impact on disease risk (in the few individuals who carry them) than common non-coding variants (which may have a greater impact at the population level). Recent genome-wide association studies using common genetic variants have identified specific loci and/or genomic regions that contribute to the etiology of certain disorders. However, only a small proportion of the heritability of complex disorders, such as SD, can be accounted for by common variants [40, 41]. Therefore, it is necessary to sequence exons of target genes (such as RGS17 and RGS20) or perform exomic sequencing using next-generation sequencing technology to identify new rare variants and analyze their association with SD. Fifth, given the incomplete penetrance of susceptibility genes for alcohol or drug dependence in monozygotic twins , epigenetic mechanisms should be studied to determine their contribution to SD risk. Altered DNA methylation levels in a number of genes (e.g., OPRM1) have been found in patients with alcohol or opioid dependence [43, 44]. Altered methylation of RGS17 and RGS20 (especially in their promoter regions) could increase the risk for SD. Therefore, epigenetic studies may provide further evidence about the role of RGS17 and RGS20 in the etiology of SD.