In this multiple timepoint epigenome-wide association study of middle-aged adults, we observed 201 methylation markers associated with recent and cumulative marijuana use across time. We replicated 8 previously reported methylation markers associated with marijuana use. We also observed 638 cis-meQTLs associated with several marijuana-methylation markers, as well as 198 differentially methylated regions. During pathway and disease analyses, marijuana-associated genes were statistically overrepresented in numerous pathways and diseases. While replication of these findings in independent cohorts is warranted, our results provide novel insights into the association between recent and cumulative marijuana use and the epigenome and related biological processes, which may serve as a mechanism of early-stage disease associated with marijuana use.

We identified numerous methylation markers associated with recent and cumulative marijuana use. Of these, cg05575921 in AHRR was associated with recent and cumulative marijuana use at both timepoints, including the single most-associated methylation marker for two of the four analyses. This methylation marker has previously been associated with heavy cannabis use among tobacco users [22], tobacco use [45–47], and is 1 of 172 CpGs included in the estimation of a DNA methylation surrogate for pack-years of smoking (DNAmPACKYRS) for GrimAge, a measure of biological age associated with lifespan [48]. The association of this epigenetic marker with both tobacco and marijuana use may suggest common modulating effects on DNA methylation and may represent a nondiscriminatory smoke related biomarker, irrespective of tobacco or marijuana use. Additionally, cg05575921 has been associated with psychiatric disorders [49, 50]. The top methylation marker associated with recent marijuana use at Y15, cg18110140, is located on chromosome 15 in an ‘open sea’ region of the epigenome. This marker was recently found to be associated with smoking status [51–53]. Several top epigenomic loci have also previously been associated with tobacco smoking, including BMF and MYO1B [52], and may provide additional measurable biomarkers for tobacco and marijuana exposure. Moreover, numerous epigenomic loci have been reported to have potential therapeutic benefits via the endocannabinoid system. NOX4 is a member of the NADPH oxidase family and an enzyme that synthesizes reactive oxygen species (ROS) and cannabidiol (CBD), one of the most common cannabinoids, has been reported to attenuate ROS formation and enhance expression of NOX4 [54]. Similarly, TFEB is associated with the autophagy-lysosomal pathway and may aid in reducing inflammation and cognitive impairment via the cannabinoid receptor type II [55]. Although the effect estimates for the observed associations are relatively small, the magnitude of the beta coefficients are consistent with previous EWAS studies [22, 28] and further studies investigating the cumulative effect of these individual CpGs (e.g., polyepigenetic risk score) may yield greater biological, and potentially clinical, relevance. We also replicated several previously reported marijuana loci, i.e., AHRR, ALPG, F2RL3, and MYO1G [22], in this mixed sex and self-reported race study sample, although additional studies in more diverse populations are needed to further evaluate previously associated epigenetic markers. Additionally, we observed differential DNA methylation levels by self-reported race and tobacco smoking status. While regression coefficients were highly correlated during stratified analyses, these findings provide insight into the interactive roles of self-reported race and tobacco smoking on marijuana associated methylation markers. For example, recent and cumulative marijuana use tended to exhibit greater hypomethylation of cg05575921 among Black participants and nonsmokers compared to White participants and former and current smokers, respectively. For the latter finding, the hypomethylation of cg05575921 during pooled and stratified analyses by tobacco smoking status suggests marijuana’s association with methylation may be consistent and independent of tobacco smoking. Our results highlight the interactive influences of biological and environmental factors on methylation signatures and provide insight into the differing impact of marijuana on the epigenome by population strata. These findings may serve as potential biomarkers to identify recent and long-term marijuana use and molecular targets for further investigation.

The epigenome is dynamic and responsive to environmental and lifestyle factors throughout the lifespan. Due to the ever-changing nature of the epigenome, evaluating differences in methylation patterns across time not only enables the temporal (and, potentially, causal) assessment of a phenotype and epigenetic changes in the context of the natural history of a disease, but also permits examination of intra- and inter-individual variability and trajectories in methylation patterns over time [56]. Additionally, longitudinal epigenetic studies will allow for the examination of the impact of interventions on epigenetic changes. For example, longitudinal examination of smoking-induced DNA methylation patterns identified dynamic and stable markers across time and also observed reversal of smoking induced methylation changes after smoking cessation [57, 58]. Using repeated measures of DNA methylation and marijuana use, we cross-sectionally identified numerous marijuana associated epigenetic markers associated at one time point but not the other (i.e., dynamic), including 6 (e.g., BEND3 and GNG12) and 10 (e.g., PHF2 and PSMD2) loci associated with both recent and cumulative marijuana use at Y15 and Y20, respectively. Additionally, one stable epigenetic marker, cg05575921, was associated with both marijuana variables across the examination years with consistent effect estimates (recent use: β_Y15 = −3.37E−02 vs β_Y20 = −3.41E−02; cumulative use: β_Y15 = −8.51E−02 vs β_Y20 = −6.94E−02). We also performed longitudinal analyses to investigate changes in methylation and marijuana use across the examinations and identified 12 CpGs that varied with change in marijuana use, including markers in AHRR, COL11A2, and TFEB. Together, these results suggest a majority of the observed marijuana associated epigenetic associations are dynamic, although stable epigenetic patterns maybe observed with marijuana use. Furthermore, the identification of dynamic markers across time suggests both recent and cumulative marijuana use may modulate epigenetic changes differently during the aging process. A possible explanation for the observation of different CpGs, as well as biological pathways and diseases, across the timepoints may relate to the pharmacokinetic properties influenced by age. For example, reductions in hepatic and renal clearance can increase the bioavailability of marijuana metabolites with prolongation of its half-life and subsequently, may impact molecular and cellular processes differently by age [59]. Consistent with our findings, dynamic epigenetic markers are more likely to be identified compared to stable markers during longitudinal analyses [60]. However, further studies investigating the modulatory effects of marijuana on the epigenome on different age groups may provide additional insight. Moreover, changes in marijuana use may alter DNA methylation signatures, which may serve as biomarkers to evaluate continued or ceased marijuana use. Although additional studies are needed to evaluate these markers, our findings demonstrate marijuana may induce dynamic and stable epigenetic signatures that may have utility as biomarkers for recent and cumulative marijuana use across time.

The impact of lifestyle factors and behaviors on health is complex and often involves an integrative approach to elucidate the underlying biological processes. By investigating genetic contributions to methylation markers associated with marijuana use, we identified 650 cis-meQTLs, including 56 cis-meQTLs that mapped to traits in the NHGRI-EBI GWAS Catalog. Among the mapped traits, we observed consistent terms related to immunological factors, cardiovascular traits, and brain measurements. Marijuana use has been associated with alterations in white blood cell counts [61], blood pressure [62], and brain structures [63]. We found the most significant mapped cis-meQTL has previously been associated with coffee consumption. Caffeine is the most consumed psychoactive substance in the world and induces dopamine release in the nucleus accumbens, a brain structure mediating pleasure and reward processing [64]. Analogously, marijuana exerts similar effects on the nucleus accumbens via the endocannabinoid system [65], suggesting the pleasure and reward of caffeine and marijuana use share the same reward center. Additionally, we identified 198 DMRs associated with recent and cumulative use of marijuana at Y15 and Y20. Among the top DMRs, several regions have previously been associated with cognitive function, psychiatric disorders, and immune function. RNF39 was the most significant DMR in two of the four analyses and has previously been associated with general cognitive function [66] and bipolar and major depressive disorders [67]. TRIOBP is the most significant DMR associated with recent marijuana use at Y20 and has been associated with general cognitive function [66, 68], schizophrenia [69], and basophil count [70]. Similarly, SH3RF3 has been associated with general cognitive ability [66], schizophrenia [71], and eosinophilia [72]. Lastly ZFP57 has been associated with general cognitive ability [66, 68], schizophrenia [73], autism [74], and rheumatoid arthritis [75]. In sum, these findings suggest marijuana use shares common genetic and epigenetic pathways associated with immunological factors, cognitive function, and brain structures and may regulate similar molecular mechanisms and biological processes. These insights could help lead to the development of new preventive and predictive tools for marijuana-associated health outcomes.

As a psychoactive substance, marijuana may modulate pathways and diseases associated with homeostasis and health outcomes. Our pathway analysis revealed differentially methylated markers overrepresented in pathways associated with cellular proliferation, hormone signaling, and infection. The MAPK signaling cascades are signaling pathways that regulate cellular proliferation, differentiation, and apoptosis; studies have suggested potential therapeutic benefits of CBD on cancer treatment via these pathways [76, 77]. With regard to hormones, the endocannabinoid system modulates dopaminergic neurons and acute use of tetrahydrocannabinol (THC) increases dopamine release and neuron activity, whereas long-term use has been associated with diminishing of the dopamine system [78]. THC has also been shown to modulate oxytocin and areas of the brain associated with reward and addiction behaviors [79]. Moreover, cannabinoids have been reported to promote progression of human papillomavirus positive head and neck squamous cell carcinoma, primarily through MAPK activation [80]. Notably, a previous genome-wide DNA methylation study of marijuana identified the latter two pathways during pathway analysis [22]. In addition to these biological pathways, differentially methylated genes associated with marijuana use were overrepresented in psychiatric diseases and spasticity. Marijuana use has been associated with several psychotic disorders including schizophrenia [81, 82], bipolar disorder [83, 84], autism [85], and psychosis [86], as well as substance-related disorders [87, 88]. Additionally THC [89] and smoked marijuana [90] have been shown to reduce spasticity among patients with multiple sclerosis and spinal cord injuries. Additionally, connections between the top marijuana associated pathways and diseases have been previously reported. For example, abnormalities in the MAPK signaling [91] and dopamine pathways [92] have been associated with schizophrenia, as well as the use of oxytocin for treatment of substance related disorders [93]. Collectively, we identified pathways and diseases overrepresented with marijuana-associated methylation markers, suggesting common epigenetic regulations which could serve as potential diagnostic and therapeutic targets for these related traits.

The current CARDIA study leveraged repeated methylation levels and marijuana data to examine the association of marijuana use on DNA methylation. The availability of genetic data enabled the examination of potential genetic modulation of methylation markers associated with marijuana via meQTL analyses. Moreover, compared to other countries where residents use a mixture of marijuana and tobacco, CARDIA is a US-based cohort where mixing of marijuana and tobacco is less prevalent, allowing for a more complete examination of the independent associations of marijuana and tobacco smoking on DNA methylation [94]. This study, however, is not without limitations. Although we identified biologically relevant epigenetic loci and replicated previously reported methylation markers, we were unable to replicate our findings in an independent study, and as such, the findings presented warrant validation. Residual confounding from additional factors, e.g., use of other or co-drug use and social support, may partially explain the observed associations. As marijuana use was considered illegal for most yearly examinations in CARDIA, use may have been underreported. However, at each examination, marijuana use was self-reported (as opposed to interviewer obtained), collected at a research site (rather than an employer), and participants’ responses were confidential [95]. The route of administration of marijuana can also affect the onset, intensity, and duration of the psychoactive effects, as well as organ systems [96]. Investigations into marijuana use via other routes of administration (e.g., edibles, pills, vaping) may provide novel additional insights, including the latter, which was not present during the timepoints in the current study but is becoming more widely used. Additionally, this study examined acute exposure to marijuana (within the last 30 days), compared to hyperacute exposure (within hours) and investigations into DNA methylation changes due to hyperacute exposure may provide further insight into the acuity of exposure on epigenetic factors. And lastly, although CARDIA is a diverse cohort, Black and White participants were sampled from four centers across the US. As such, additional studies from more diverse populations across different geographical locations will enable for better generalizability of the findings presented here.