The main findings of our study are: (1) SNP rs1049353 in CNR1 gene is associated with CU; (2) the highest level of education, College graduate/Masters/Ph.D., reduced significantly the risk to be cannabis users; and (3) changes in DNA methylation, comparing cannabis users and control, were observed in two sites related to DRD2 and NCAM1 genes.

Our findings are consistent with the nominal association found between the G allele, SNP rs1049353, and cannabis dependence symptoms¹⁰. The SNP rs1049353 polymorphism appears to play a significant role in the function of cannabinoid receptor by influencing mRNA translation as well as secondary structure and stability of the protein^35,36. Thus, it is possible that individuals who carry this SNP might experience different marijuana-induced behavioral or physiological effects and may also show different phenotypes that are regulated by the cannabinoid system^10,37–40. These suggestions might explain why the G allele may serve as a risk factor for CU⁴¹.

To increase the power to detect genetic traits associated with CU, haplotype-based association analysis was also used. Different combinations of SNPs in CNR1 showed significant associations with CU at 2–4 SNPs levels for seven haplotypes in our study. These observations are consistent with another research group that had reported a significant association of CNR1 haplotype rs806368-rs1049353-rs2023239-rs6454674 with the level of cannabis exposure and decreased volume of the right anterior cingulum⁴².

For the variant rs2501431 in the other cannabinoid receptor CNR2 gene, differences were observed in genotype distribution, with the homozygous A/A genotype being more frequent in the cannabis group. When we excluded GG genotypes, the association did not reach significant values (p = 0.7), leading us to assume a very important role of the G allele. This statement is consistent with the report that the GG genotype is associated with alcoholism in a Japanese population⁴³. Together, these observations suggest the need for further study to elucidate the role of rs2501431 in CNR2 gene as possible risk factor of CU and other addiction diagnoses.

In addition, our data highlight the importance of gender in CU susceptibility. Compared to females, male subjects in the study presented a significantly higher risk to use cannabis. These findings underlie the importance of better exploring gender difference in addiction research. Gender differences in CU are indeed frequently reported, with male proneness to use the drug more frequently or at a higher rate than females^44–46. Gender-based differences in cannabinoid effects have been evidenced by preliminary studies investigating the role of gonadal hormones in the regulation of cannabis receptors density and affinity⁴⁷. Other studies highlighted that sex differences could also be influenced by drug disposition and body fat distribution⁴⁸. However, the fact that females are underrepresented in our study and others may also lead to incomplete and biased results.

Importantly, the evaluation of sociodemographic risk factors helped to confirm that higher education attainment might serve as a possible protective factor, considering the significant inverse relation between education levels and CU. A previous study had reported that users with higher education do not suffer from cognitive deficits with chronic abuse of cannabis⁴⁹. Interestingly, strong engagement in school and good school achievements have also been repeatedly reported to be important resilience factors against substance use disorders^50–54. Thus, poor school performance and dropping out of schools may not only be risk factors for substance use disorders but may be the consequences of shared genetic/epigenetic risks that predict both school failure and CU initiation^55,56. The potential impact of cannabis on school performance is also very important to consider^57–60. Logistic regression analysis demonstrated that both genetic (rs1049353, CNR1 gene) and sociodemographic risk factors are coincident, representing endophenotypes associated with CU risks. The role of endophenotypes for the rational development of therapeutic and preventive strategies has previously been assessed and suggests that these are important variables in treatment approaches⁶¹.

Our epigenetic findings also appear to suggest that not only inherited gene variants, but also changes in gene expression could be associated to CU. Hypermethylation was reported in cannabis users compared to controls at DRD2 exon 8 and in a CpG-rich region of the NCAM1 gene. The observed NCAM1 differentially methylated region (DMR) was located on the 5′ CpG island of the gene and promoter-associated DMR-CpGs are usually shown to cause transcriptional silencing⁶². Even if the functional significance of DNA methylation in introns, exons, and intergenic intervals is still less well-understood^63,64, the DRD2 DMR, located within the exon 8, suggests the importance of these gene body sites, as already shown in a study on parental THC exposure⁶⁵.

Both genes, DRD2 and NCAM1 have a central role in the dopaminergic pathway⁶⁶ and have been recently found significantly associated with lifetime CU¹⁶. They are part of the NCAM1–TTC12–ANKK1–DRD2 gene cluster (NTAD), suggested to be associated to nicotine dependence⁶⁷ and other substance use disorders in general⁶⁸.

Other behavioral conditions were reported to be associated with DNA methylation in these two genes. DRD2 promoter showed highest methylation levels in patients with gambling behavior, compared to non-gambling participants and, because DNA-hypermethylation is usually associated with transcriptional repression, the subsequently higher availability of D2 receptor was assumed to result in a more active working of the reward system⁶⁹.

Based on these evidences, we could hypothesize (Fig. 3) that increased DNA methylation observed in cannabis users' blood cells may in some way reflect a lower D2 mRNA expression, lower availability of D2 receptors underlying a reward deficit condition. This difference in the reward system function related to methylation may lead to a reduced need for reinforcing stimuli in non-cannabis users, and a higher necessity of pharmacological stimulation in cannabis-dependent subjects.

The observed hypermethylation in NCAM1 gene is consistent with the hypermethylated CpG regions of NCAM1 in alcohol-dependent patients compared with controls^70,71. This gene that was previously found significantly associated with lifetime CU¹⁶ encodes a neural cell adhesion molecule implicated in important functions during development and maintenance of the nervous system. A study in rodents proposed NCAM as a modulator of the dopaminergic pathway and a potential pharmacological target for dopamine-related psychiatric disorders⁷².

In the interpretation of our findings concerning hypermethylation in cannabis users, it is impossible to establish any causal relationship. Epigenetic changes affecting DRD2 and NCAM1 could be the result of long-term exposure to cannabis and not a pre-existing condition involved in CU neurobiological vulnerability.

The present study has certain limitations. First, because the number of cannabis users was relatively low, statistical power was not as strong as expected. Moreover, since polymorphisms may vary in frequency among different ethnic groups⁷³, it might have been more appropriate to split the samples based on ethnic groups. However, this was not possible due to the relatively small number of subjects per group. The results obtained by the present study could thus be attributed to the ethno-racial background of the participants. Thus, any extension to other ethnic groups should be done with caution and will need to be investigated further.

Another important limitation is that most of the changes in DNA methylation patterns were observed in peripheral cells⁷⁴ such as WBCs that may not reflect changes in reward brain regions. Moreover, the changes in DNA methylation patterns were investigated at a single time point, thus, precluding the possibility of examining time-dependent effects of marijuana exposure on DNA methylation. Replication studies using specific cell types and brain tissues should help to further clarify the role of DNA methylation in CUD. Another issue that needs to be investigated in future replication studies include assessment of the role of marijuana on other epigenetic markers. This will entail examining changes in histone modifications and DNA methylation during the clinical course of repeated marijuana exposure. Indeed, the use of genome-wide methylation or chromatin immunoprecipitation studies should help to identify other genes that might be involved as either risk factors or consequences of cannabis exposure.

Overall, our results suggest a significant role of genes encoding proteins involved in the endocannabinoid system in the susceptibility for CU. Our study also identified epigenetic modifications affecting dopamine system function as risk factors and/or consequences of marijuana exposure. Confirmation studies that address the shortcomings detailed above are necessary to clarify to what extent changes in peripheral DNA methylation could serve as accessible sources of biological markers and/or targets for pharmacological interventions that might reduce or prevent the effects of marijuana in the brain. Finally, because education attainment was identified as a protective factor for CU, these observations suggest the building and expansion of early childhood educational programs and strategies to prevent CUD in adolescence and later in life.