While numerous studies have examined the genetics of alcohol, nicotine, cannabis, opioid, and cocaine use disorders, relatively few studies have assessed the genetics of methamphetamine dependence . A genome-wide association study of methamphetamine dependence in a sample from Asia found significant associations between a diagnosis of methamphetamine dependence and single nucleotide polymorphisms clustered in genes for cell adhesion molecules including CDH13 and CSMD1 . A GWAS of amphetamine-response in healthy volunteers also identified SNPs in CDH13 as the most significant SNPs associated with subjective response to amphetamine . In addition, a recent GWAS found several SNPs near CREB1 were significantly associated with opioid response as well as lower risk of polydrug use in volunteers with methamphetamine dependence and altered CREB1 expression . Studies examining genetic associations with phenotypes of relevance to treatment for substance use disorders may identify new targets for treatments for addiction. Higher pre treatment methamphetamine use frequency is associated with greater severity of methamphetamine use disorder, worse clinical outcomes for outpatient treatment, and differential pharmacotherapy response . Urine drug screens detect recent drug use, are used ubiquitously as a treatment outcome measure in addiction treatment and clinical trials,indoor grow trays and are associated with long term outcomes following outpatient treatment for stimulant use disorders . We performed a candidate gene study of pre-treatment methamphetamine use frequency and urine drug screen results during treatment among methamphetamine dependent Hispanic and Non-Hispanic White participants of several outpatient methamphetamine dependence clinical trials in Los Angeles.
We selected SNPs in CDH13 given the two GWAS identifying variants in CDH13 associated with methamphetamine dependence and subjective response to amphetamine as well as SNPs associated with opioid response in a recent GWAS . Given the small number of methamphetamine genetic studies to date, we also included SNPs associated in previous studies with other phenotypes with relevance to methamphetamine dependence such as dependence on nicotine, cocaine, or alcohol, functioning of dopaminergic systems, brain structure, and other psychiatric diseases. A detailed rationale for each SNP is provided in Table S1. Data for the current study were taken from several methamphetamine dependence outpatient clinical trials at UCLA. Each trial had a similar design and inclusion/exclusion criteria and recruited volunteers seeking treatment for methamphetamine problems via print, radio, and internet ads. Participants visited a UCLA outpatient research clinic and completed the informed consent process, including separate consent for genotyping. Participants then underwent a battery of clinical assessments including the Structured Clinical Interview for DSM-IV , assessment of substance use, including the self-reported number of days with methamphetamine, marijuana, alcohol, and tobacco use during the past 30 days prior to entering the trial, and collection of blood for genotyping. Those participants meeting trial eligibility criteria then underwent outpatient treatment, including weekly cognitive behavioral therapy sessions and study medication for 8 to 12 weeks. During treatment, participants visited the clinic thrice weekly for urine drug screens for methamphetamine. Participants included in the current analysis met the following criteria: aged 18 and older, seeking treatment for methamphetamine problems, methamphetamine dependent per DSM-IV-TR criteria as assessed by the SCID, completed baseline substance use frequency assessments, provided consent and blood for genotyping, and Hispanic or Non-Hispanic White ancestry based on results of genotyping a panel of ancestry-informative markers . Demographics of the sample included in the current analysis are shown in Table S2.
The study was approved by the UCLA IRB and the clinical trials from which data is obtained were each registered with clinicaltrials.gov . Sixty four candidate SNPs hypothesized to be associated with methamphetamine use frequency were selected for genotyping . SNPs were selected on the basis of previous research associating the SNP with methamphetamine dependence or a related phenotype such as response to amphetamine in healthy volunteers, other psychiatric conditions such as ADHD, depression, schizophrenia, dependence on other substances such as cocaine, alcohol, or nicotine, dopaminergic functioning, and functional or structural brain imaging phenotypes. When available, preference was given to SNPs identified in previous GWAS studies over those from previous candidate gene studies. One candidate SNP of interest, rs2952768, which was associated with opioid sensitivity and severity of methamphetamine dependence in a Japanese GWAS was not able to be genotyped on the genotyping platform used and was replaced two nearby SNP also associated with opioid sensitivity in the GWAS: rs7591784 and rs2709386. Details of the SNPs and the rationale for their selection is provided in Supplemental Table S1. In addition, a panel of 128 ancestry-informative markers were genotyped in order to assess for and control population stratification by ancestry . Whole blood was collected from participants via venipuncture and DNA was extracted via Gentra Autopure LS nucleic acid purification instrument and then frozen and stored at −20° C for genotyping later. SNPs were ge notyped using Fluidigm SNP Type™ assays with the Fluidigm Biomark™ HD system at the UCLA genotyping core facility. SNPtype™ assays and reagents for each of the SNPs were purchased from Fluidigm. Genotype calls were made using the Fluidigm SNP Genotying Analysis Software and genotype cluster plots for each SNP were examined manually for quality control.
Of the 64 candidate SNPs, 6 SNPs failed genotyping quality control and were removed, leaving 58 candidate SNPs genotyped and available for analysis. Two of the AIM SNPs also failed genotyping leaving 126 AIMs for analysis. Of the 58 SNPs genotyped, 6 SNPs were in very high LD with other genotyped SNPs and were eliminated from further analyses leaving 52 SNPs for the candidate gene association analysis. After initial quality control, seventeen genotype values were missing and were imputed by sampling the missing genotype from the empirical distribution over all other individual’ genotype at that SNP. Ancestry was evaluated using the 126 genotyped AIMs. A reference population was obtained from the HGDP-CEPH Human Genome Diversity Cell Line Panel , containing genotype information for over 1,043 individuals. Using only the 126 AIMs common to both the reference data and the present study, the Bayesian clustering algorithms implemented in STRUCTURE v2.3 were used to estimate population admixture proportions. In order to determine the optimal number of ancestry-specific clusters, the log-likelihood of the data was evaluated as a function of cluster size. The choice to use a total of four separate clusters was made since the increase in the log-likelihood after adding the fifth group was minimal. Moreover, no individual had predominant ancestry from the fifth group when a total of five groups were used. After setting the number of distinct ancestry-specific groups to four, ancestry of the individuals in the current study was determined using the reference population over 25 runs in STRUCTURE. A total of 20,000 burn-ins and 50,000 iterations were performed in each run. CLUMPP v1.1 software was then used to adjust for permutations between the 25 runs and to align all four population clusters. The ancestry corresponding to each cluster was determined by aligning the ancestries in the reference group to the individuals in the current study.A total of 265 Hispanic White and Non-Hispanic White participants were included in the candidate gene analyses. Initial analyses showed that sex and proportion of Native American Ancestry determined by AIMs were significantly associated with pretreatment methamphetamine use frequency and therefore methamphetamine use frequency analyses were performed stratifying by sex and controlling for proportion of Native American Ancestry. Separate linear regression models were run for each of the SNPs predicting pretreatment methamphetamine use frequency,2 tier grow rack controlling for age, proportion Native American ancestry and study, in men and women assuming an additive, dominant, and recessive genetic model. A Bonferroni corrected p < 0.001 was used as the threshold for statistical significance accounting for the 52 SNPs included in the analyses. None of the SNPs deviated significantly from expected Hardy-Weinberg equilibrium among Hispanic or Non-Hispanic Whites. A linkage disequilibrium plot for the region surrounding the most significant SNPs, rs7591784 and rs2709386, was created using HaploView, version 4.2 and genotype data from HapMap population CEU. The SNP with the strongest association with pretreatment methamphetamine use frequency, rs7591784, was then assessed for association with methamphetamine urine drug screen results during outpatient treatment. This analysis was limited to male participants with treatment outcome data available from two clinical trials with identical 12 week outpatient treatment periods .
Generalized estimating equations using a first order auto-regressive correlation structure were fit to longitudinal data for methamphetamine urine drug screen results collected 3 times a week over a 12 week outpatient treatment period. Separate models were run for the additive, recessive, and dominant genetic models, controlling for pretreatment methamphetamine use, study, smoking status, and proportion Native American ancestry. The method of multiple imputations was used to deal with missing treatment outcomes, where logistic regression was used to impute intermittent missing values. A total of 50 imputed datasets were created and the results were combined using Rubin’s rules . The SNP most strongly associated with pretreatment methamphetamine use frequency, rs7591784, was then tested for association with methamphetamine use during treatment controlling for pretreatment methamphetamine use. As rs7591784 was associated with pretreatment methamphetamine use among males only, this analysis was limited to males. Among male participants with treatment outcome data available , rs7591784 was significantly associated with the probability of testing positive for methamphetamine via urine drug screens during a 12 week treatment period assuming a dominant genetic model and controlling for pretreatment past 30 day methamphetamine use frequency, study, cigarette smoker status, and proportion Native American ancestry. Participants homozygous for the minor G allele were significantly less likely to provide urine specimens positive for methamphetamine during treatment = 0.0002, where p̄ is the average p-value over 50 imputed datasets compared to participants with at least one A allele . Results using an additive or a recessive genetic model and without imputation of missing data yielded similar results, although with a larger p value, and the addition of covariates for active versus placebo conditions from the clinical trials did not change the results . We performed a candidate gene study of methamphetamine treatment among methamphetamine dependent Hispanic and Non-Hispanic Whites participating in several methamphetamine clinical trials and found one SNP, rs7591784, was significantly associated with methamphetamine use both before and during outpatient treatment in males but not females. Higher pretreatment methamphetamine use frequency is a marker of greater severity of methamphetamine use disorder and is a strong predictor of continued methamphetamine use and poor treatment outcomes during outpatient treatment for methamphetamine use disorder . The identification of an association between rs7591784 and pretreatment methamphetamine use frequency provides insight into the biological mechanisms influencing severity of methamphetamine use disorders and may also identify targets for new treatments for the group with the highest pretreatment use frequency, who respond poorly to existing behavioral therapies. Given the strong association between higher pretreatment methamphetamine use frequency and poor treatment outcomes, it is not surprising that rs7591784 was associated both with pretreatment frequency of methamphetamine use and methamphetamine use assessed via urine drug screens during subsequent outpatient treatment. But the association between rs7591784 and methamphetamine urine drug screen results during treatment was strongly significant after controlling for pretreatment methamphetamine use frequency suggesting that rs7591784 is associated with treatment outcomes independent of pretreatment use frequency. SNP rs7591784 is on chromosome 2 in the intergeneic region near CREB1 and METTL21A . CREB is a transcription factor that mediates changes in gene expression resulting from chronic exposure to a variety of drugs of abuse including methamphetamine and has been shown to influence drug reward, self-administration, and relapse in multiple animal models of addiction . Methamphetamine increases phosphorylated CREB, the active form of the transcription factor, via striatal dopamine receptor-mediated activation of adenylate cyclase resulting in increased cAMP and activation of protein kinase A . Phosphorylated-CREB then binds to the promoters of genes implicated in methamphetamine-induced epigenetic changes and neuroplasticity that are thought to underlie the persistent risk of relapse characteristic of addiction, such as c-fos, fosB, and BDNF, increasing expression of these genes in the striatum . CREB also mediates methamphetamine-induced astrocyte activation and increased expression of sigma-1 receptors which may contribute to neuroinflammatory changes observed in methamphetamine addiction . A SNP in CREB1, rs10932201, was associated with sensitivity to reward and activation of brain regions important in addiction including the nucleus accumbens during a reward-related decision making task among healthy young adults .