Methylomic Investigation of Problematic Adolescent Cannabis Use and Its Negative Mental Health Consequences

Abstract

Objective:

The impact of adolescent cannabis use is a pressing public health question due to the high rates of use and links to negative outcomes. Here we consider the association between problematic adolescent cannabis use and methylation.

Method:

Using an enrichment-based sequencing approach, we performed a methylome-wide association study (MWAS) of problematic adolescent cannabis use in 703 adolescent samples from the Great Smoky Mountain Study. Using epigenomic deconvolution, we performed MWASs for the main cell types in blood: granulocytes, T-cells, B-cells and monocytes. Enrichment testing was conducted to establish overlap between cannabis-associated methylation differences and variants associated with negative mental health effects of adolescent cannabis use.

Results:

Whole blood identified 45 significant CpGs, and cell-type specific analyses yielded 32 additional CpGs not identified in the whole blood MWAS. We observed significant overlap between the B-cell MWAS and genetic studies of education attainment and intelligence. Furthermore, the results from both T-cells and monocytes overlapped with findings from a MWAS of psychosis conducted in brain tissue.

Conclusion:

In one of the first methylome-wide studies of adolescent cannabis use, we identified several methylation sites located in genes of importance for potentially relevant brain functions. Our findings resulted in several testable hypotheses by which cannabis-associated methylation can impact neurological development, inflammation response, as well as potential mechanisms linking cannabis-associated methylation to potential downstream mental health effects.

Introduction

Cannabis is a commonly used psychoactive drug. It has a particularly high rates of use among adolescents with a recent survey showing a prevalence rate of 22.7% for past 30 day use in US high school seniors¹. Problematic cannabis use during adolescence is of concern as the brain is intrinsically more vulnerable during this developmental period. Indeed, a growing body of literature suggests that cannabis use during adolescence is associated with negative mental health outcomes, most notably deficits in cognition and memory, an increased prevalence of affective and psychotic disorders, and a higher risk of abuse of other drugs in adulthood².

DNA methylation involves the addition of a methyl group to the carbon 5 position of cytosine and, in blood, is often found in the sequence context CG. Methylation could be relevant to the mechanisms underlying substance use and addiction. For example, repeated exposure to some drugs of abuse leads to chronic cycles of consumption and withdrawal that evoke long-lasting neuroplasticity thought to contribute to abusive consumption, dependence, and sensitization³. These observations suggest lasting cellular and molecular modifications. DNA methylation, being the archetypal cellular information storage mechanism, has been heavily implicated as a chief mechanism for stable activity-dependent transcriptional alterations within the central nervous system. Some drugs of abuse are known to induce epigenetic changes that can persist over time and have lasting effects⁴, which may explain the observations above.

Little, however, is known about the impact of cannabis on the methylome. Previous candidate gene studies in peripheral blood have shown altered methylation of DRD2 and NCAM1 in adult cannabis users⁵, CB1R⁶ in cannabis dependent adults, DAT1 in cannabis dependent adult male participants⁷, and COMT⁸ in high-frequent adult cannabis users. A methylome-wide association study conducted in peripheral blood identified one methylation site located in CEMIP that was significantly associated with lifetime cannabis use ever in adult female participants⁹. Cannabis use has also been shown to impact DNA methylation on a genome-wide scale in the sperm of both human and rats¹⁰. Further evidence comes from model systems where adolescent mice chronically treated with a synthetic cannabinoid had hypermethylation at Rgs7 which predicted memory impairment in adulthood¹¹. Further, adolescent exposure to Δ⁹-tetrahydrocannabinol in rats and female mice changed the normal developmental trajectory of the transcriptome in the brain^12,13 and these transcriptional changes were enriched for genes involved in epigenetic processes¹³. Taken together, these studies suggest that cannabis can impact the methylome, and that cannabis use during adolescence can influence epigenetic processes leading to detrimental effects on gene expression and adult outcomes.

Methylation studies, including previous studies of cannabis, are typically conducted using bulk tissue (e.g., whole blood) that is composed of multiple cell-types, each with a potentially different methylation profile. In studies using bulk tissue, estimates of cell-type proportions are typically included as covariates to protect against false positive associations¹⁴. By focusing only on bulk tissue, however, many potentially important associations may be missed¹⁵. For example, when effects involve a single cell type or have opposite directions in different cell-types, they may be diluted or potentially nullify each other in bulk tissue. Further, as the most common cell-types will drive the results in bulk tissue, effects from low abundance cells may be missed altogether. Additionally, knowing which specific cells harbor effects is key for the biological interpretation and crucial for designing follow-up experiments.

Here, we conduct a cell-type specific human methylome-wide association study (MWAS) of problematic adolescent cannabis use. To study possible cell-type specific effects, we used an epigenomic deconvolution strategy to perform MWASs for cell populations of granulocytes, T-cells, B-cells and monocytes, the main cell types found in whole blood. We used a sequencing-based approach that provides nearly complete coverage of all 28 million autosomal CpGs in the human genome¹⁶ to avoid missing methylation sites of possible importance. We further explored if significant overlap existed between the MWAS results and association results from other studies of the negative mental health effects assumed to be associated with problematic adolescent cannabis use (i.e., cognition, depression, psychosis and substance use).

Method

Participants

The Great Smoky Mountain Study (GSMS) is an ongoing, prospective longitudinal, representative study of children in 11 predominantly-rural counties in the southeast United States, which began in 1993¹⁷. Three cohorts of children, age 9 to 13 years, were recruited resulting in N=1,420 participants. Details about the GSMS design, recruitment and data collection are published elsewhere¹⁷. As part of a larger methylation study on childhood and adolescent exposures, 525 participants were selected that were 9–21 years of age and had a bloodspot taken at the time of assessment available. The youngest participants reached age 21 in 2006, well before cannabis legalization reforms passed in parts of the United States. Of the 525 participants, 178 had bloodspots included from additional older time points (age 10–21 years). This resulted in 703 samples from 525 unique participants included in the current study. Parents and participants provided their informed consent and the current study was approved by Institutional Review Boards at Duke University and Virginia Commonwealth University.

Participant substance involvement was assessed during an in-person interview. At each interview assessment between ages 9 and 16, the participant and parent completed a full structured clinical interview about substance involvement using the structured Child and Adolescent Psychiatric Assessment (CAPA)¹⁸. After age 18, the Young Adult Psychiatric Assessment (YAPA)¹⁹, the upward extension of the CAPA, was used. The substance use module of the CAPA/YAPA included assessments of DSM-IV abuse and dependence. Although DSM-5 cannabis use disorder (CUD) symptoms of craving and withdrawal were not part of the DSM-IV abuse or dependence diagnostic criteria, these data have been collected since the start of GSMS is 1993. The timeframe for determining the presence of diagnostic items was the 3 months immediately before the interview to minimize recall bias. Blood spots were obtained during the same visit as the interview, and methylation measurements based on the blood spots should be viewed as concurrent with CUD symptom endorsement. For the current study, DSM-5 CUD symptom count is used as the measure of problematic adolescent cannabis use and was treated as a quantitative trait in subsequent analyses.

Assaying the Methylome

To assay the methylome, we used an optimized protocol for methyl-CG binding domain sequencing (MBD-seq)¹⁶ that achieves near-complete coverage of the 28 million possible methylation sites in blood at the cost of the commonly used methylation arrays that assay only 2–3% of all possible methylation sites¹⁶. Briefly, genomic DNA was sheared into 150 bp fragments using the Covaris^™ E220 focused ultrasonicator system. We performed enrichment with MethylMiner^™ (Invitrogen) to capture the methylated fraction of the genome. Next, dual-indexed sequencing library for each methylation capture was prepared using the Accel-NGS^® 2S Plus DNA Library Kit (Swift Biosciences) and sequenced on a NextSeq500 instrument (Illumina). The sequence reads were aligned to the human reference genome (hg19/GRCh37) using Bowtie2²⁰.

Data Processing and Methylation Score Calculation

Data quality control and analyses were performed in RaMWAS²¹. Details on data processing and quality control are provided Supplement 1, available online. Methylation scores were calculated by estimating the number of fragments covering the CpG using a non-parametric estimate of the fragment size distribution. These scores provide a quantitative methylation measure for each individual at that specific site.

Methylome-wide Association Study in Whole Blood

The methylome-wide association study (MWAS) in whole blood was performed using multiple regression analyses with four sets of covariates. First, we regressed out assay-related variables (i.e., potential technical artifacts) such as sample batches and peak²¹. Second, we regressed out age, age squared, race and regular cigarette use. Sex was also included as a covariate to account for potential sex differences in cannabis use. Third, to avoid false positives due to cell type heterogeneity in whole blood, we regressed out cell type proportions as estimated from the methylation data¹⁴. Fourth, principle component analysis (PCA) was used to capture any remaining unmeasured sources of variation. One principal component was selected based on the screen test. A false discovery rate (FDR) of 10% (q-value < 0.1) was used to classify sites as differentially methylated.

Cell-type Specific MWAS

Bulk tissue such as whole blood consists of several different cell types with potentially different methylation profiles. In MWAS of bulk tissue, this cell type heterogeneity can negatively impact the ability to detect and interpret associations with phenotypic and clinical outcomes¹⁵. Here we used an epigenomic deconvolution approach to perform cell-type specific MWAS for the major nucleated cell types found in blood: granulocytes (CD15+), T-cells (CD3+), B-cells (CD19+) and monocytes (CD14+). This approach applies statistical methods in combination with methylation profiles from a reference set of purified cells to deconvolute the cell-type specific effects from data generated in bulk tissue. In short, cell type proportion estimates for each sample were obtained using Houseman’s method¹⁴. These estimates were then used to test the null hypothesis that methylation of a given site is not correlated with cannabis use for each cell type¹⁵. This approach has been validated for methylation sequencing data²². Assay-related variables, demographic variables (sex, age, age squared, race, regular cigarette use), and one principal component were included as covariates in the cell-type specific MWAS. Cell-type proportions were also included as main effects in cell-type specific MWAS as the epigenomic deconvolution is essentially an interaction model¹⁵.

Pathway Analysis

To gain insight into the biological pathways affected by cannabis in blood, we used ConsensusPathDB (CPDB)²³ to test for overrepresentation of top MWAS findings (P<1×10⁻⁵) located within genes in the biological pathways in the KEGG and Reactome databases. At least three genes among the top MWAS findings had to be present in the pathway for it to be considered enriched.

Enrichment Testing

To perform enrichment tests of the overlap between the top cannabis MWAS results and top results from other enrichment datasets (e.g., brain chromatin states and health effect relevant association results), we used the R package shiftR. To perform these tests, shiftR uses circular permutations by shifting the mapping of the two data sets by a single random integer in each permutation. This approach generates the empirical test statistic distribution under the null hypothesis while preserving the correlational structure of the data. We used 1 million permutations for each test. Multiple thresholds (i.e., for our analyses we used the top 0.5%, 1% and 5%) were to define “top” findings for the cannabis MWAS and for the quantitative enrichment datasets. To account for this “multiple testing”, the same thresholds are used in the permutations where the test statistic distribution under the null hypothesis is generated from most significant combination of thresholds. For bivariate datasets, such as the chromatin states, the presence versus absence of a feature defined the “top”.

The top MWAS results were tested for enrichment against Roadmap Epigenomics Project chromHMM Core 15-state model chromatin tracks²⁴ in brain tissue. For these analyses, we created a “consensus” track by identifying regions that had the same histone state in 4 of 7 adult brain tissues while using Quiescent/Low as the reference state.

Further enrichment testing was performed to establish overlap between cannabis-associated methylation changes and loci associated with negative mental health effects of adolescent cannabis use. To study cognition we used a genome-wide association study (GWAS) in 78,308 individuals of human intelligence²⁵ and a GWAS of educational attainment that involved 293,723 individuals²⁶. For affective disorders, we used a GWAS of 75,607 participants diagnosed with major depressive disorder (MDD) and 231,747 controls²⁷. In addition, we used results from an MDD MWAS conducted in bulk tissue in a collection of human post-mortem brain samples²⁸. For psychosis, we used the results of a schizophrenia MWAS conducted in bulk brain tissue²⁹. Finally, for addiction we used a GWAS of cigarettes smoked per day in 74,053 participants³⁰.

Results

Sample Descriptives

Forty-two (~6%) participants endorsed at least one current CUD symptom with ~3.5% meeting criteria for a current CUD diagnosis (i.e., symptom count greater than or equal to 3). While non-problematic users may have initiated use, they had not progressed to regular use nor endorsed a CUD symptom at the concurrent or any preceding phenotypic measurement occasion. Problematic users tended to be older, male, and regular smokers when compared with non-problematic users (Table 1). There were no racial differences between problematic and non-problematic users. There were no significant differences in sex, race, regular smoking status or problematic cannabis use between individuals with multiple bloodspots and those with only one blood spot (see Table S1, available online).

Table 1:

Sample Descriptives

	Non-Problematic User (n=661)	Problematic Cannabis User (n=42)	p
	Freq (%)	Freq (%)
Male participant	329(49.7%)	31(73.9%)	0.027
Female participant	332(50.2%)	11(26.1%
White	433(65.5%)	28(66.6%)	0.714
African American	31(4.69%)	3(7.14%)
American Indian	197(29.8%)	11(26.2%)
Regular Smoker	98(14.8%)	30(71.4%)	3.94×10−06
	Mean(SD)	Mean(SD)
Age when bloodspot collected	14.1(2.53)	17.1(2.30)	2.75×10−10
CD03: T-cells	0.27(0.08)	0.29(0.06)	0.135
CD14: Monocytes	0.139(0.05)	0.14(0.05)	0.769
CD15: Granulocytes	0.504(0.10)	0.501(0.06)	0.961
CD19: B-cells	0.085(0.04)	0.07(0.04)	0.017

Note: The first 2 columns of the table show the sample descriptives for non-problematic users and problematic cannabis users, where a problematic cannabis user is defined as endorsing one or more DSM-5 cannabis dependence criteria. The top half of the table displays the frequency of the variable and the percentage of the sample in parentheses for non-problematic user and problematic users. The bottom half of the table displays the mean and standard deviation for non-problematic users and problematic users. A t or chi-square test was used to test for group differences in problematic cannabis use, the p value of which is reported in the last column.

Freq = frequency.

MWAS in Whole Blood

The Quantile-Quantile (QQ) plot (see Figure S1, available online) for whole blood suggests considerable signal with 45 CpGs reaching methylome-wide significance at FDR=0.1 (see Table S2, available online). The most significant CpG was located in CLMN (P=6.21×10⁻¹⁰, q=0.008), a gene previously associated with neuronal differentiation³¹. The second top genic finding was in SENP7 (P=3.29 ×10⁻¹⁰, q=0.018), a gene required for proper neuronal differentiation. The CPDB pathway results revealed 19 pathways (see Table S3, available online) that were significantly overrepresented among the top MWAS results in whole blood and included pathways related to cholinergic synapse (P=1.28×10⁻⁴, q=0.018) and serotonergic synapse (P=7.75×10⁻⁴, q=0.074).

For a subset of individuals, multiple measurements were included in the analysis. Therefore, we further investigated the potential effect of dependency of the observations, which may lead to biased standard error estimates and test statistics. First, a parallel MWAS of whole blood, using only one sample, selected randomly, from each of the 525 independent individuals was performed. This analysis generated similar results to the full dataset (see Figure S2, available online). Second, as an additional sensitivity analysis, we performed 500 MWASs where we permuted problematic cannabis use while preserving the correlational structure among methylation sites and retaining the multiple measurements. Figure S3, available online, shows that the average λ observed from the permutations was close to 1, and well within the 95% confidence interval values. These results show that the test statistic behaves as expected under the null when the complete study sample is investigated.

Cell-type Specific MWAS

Table 1 shows the correlation between problematic cannabis use and estimated cell type proportions. Problematic cannabis users had slightly decreased B-cell (CD19) levels. No significant correlation was observed for the other cell types.

Overall, the cell-type specific MWAS results showed similarly shaped QQ-plots (see Figure S1, available online) as whole blood. In granulocytes, 16 CpGs were methylome-wide significant (see Table S4, available online), for T-cells there were three significant CpGs (see Table S5, available online), 209 significant CpGs for monocytes (see Table S6, available online), and 25 significant CpGs for B-cells (see Table S7, available online). Table S8, available online, displays the overlap among the top cell-type specific MWAS findings. The top finding for granulocytes was the long intergenic non-protein coding (LINC) RNA gene LINC01090 (P=8.9×10⁻⁹, q=0.036). For B-cells, the top genic finding was located in TRRAP (P=1.4×10⁻⁸, q=0.028), a gene involved in DNA repair.

Pathway analyses yielded several significant over-represented pathways among the top cell-type specific MWAS results. For the granulocyte MWAS the most significant pathway was “Signaling by Robo receptor” (P=7.6×10⁻⁴, q=0.032) (see Table S9, available online), which is known to be modulated by endocannabinoids during cortical development³². Among the top pathway results from T-cells (see Table S10, available online), were “ErbB signaling” (P=6.94×10⁻⁴, q=0.041) and “CTLA4 inhibitory signaling” (P=8.20×10⁻⁴,q=0.041). For monocytes the top pathways (see Table S11, available online) were “axon guidance” (P=2.43×10⁻⁴,q=0.071) and “neuronal system” (P=4.6×10⁻⁴, q=0.071). Finally, for B-cells there were several pathways implicated related to DNA repair such as “resolution of D-loop structures” (P=8.5×10⁻³, q=0.008, see Table S12, available online).

Enrichment of loci with brain histone marks

MWAS results were tested for enrichment against Roadmap Epigenomics Project chromHMM Core 15-state model chromatin tracks²⁴ in brain tissue. Results show (Figure 1) that top MWAS findings for granulocytes and B-cells were in regions likely transcribed in brain as indicated by the points for strong and weak transcription exceeding the significance threshold (vertical dashed line). For T-cells, top MWAS findings were likely to be in regions that harbor zinc finger (ZNF) genes and repeats. We did not observe any enrichment of chromatin states among top results in monocytes or in whole blood.

Figure 1:

Enrichment Test Results For Whole Blood and Cell-type Specific Methylome-Wide Association Results Against Roadmap Epigenomics Project chromHMM Core 15-state Model Chromatin Tracks²⁴

Note: The x-axis shows the −log₁₀(p-value) for each enrichment test of whole blood, CD3, CD14, CD15, and CD19 methylome-wide association results against the corresponding brain histone state on the y-axis. P values falling to the right of the vertical dashed line (>1.3 −log₁₀(p-value)) indicate significance at p value < 0.05. CD3 = T-cells; CD14 = monocytes; CD15 = granulocytes; CD19 = B-cells; chromHMM – chromatin hidden markov model²⁴; TSS = transcription start site.

Enrichment of GWAS findings for Cannabis Associated Mental Health Risks

MWAS findings were further tested to examine if their top results were enriched for loci associated with the main health risks of adolescent cannabis use (Table 2). The B-cell MWAS showed significant enrichment for the GWASs of intelligence (P=0.042) and educational attainment (P= 0.008). Finally, T-cells, monocytes and whole blood showed significant enrichment with the psychosis MWAS in bulk post-mortem brain tissue (P=3.0×10⁻⁵, P<10⁻⁶, P<10⁻⁶, respectively). There was no significant enrichment with the granulocyte MWAS results. None of the MWASs were significantly enriched for loci in the GWAS or MWAS for MDD, nor the smoking GWAS.

Table 2:

Enrichment Tests of Methylome-Wide Association Results and Loci Associated With the Mental Health Risks of Adolescent Cannabis Use

MWAS	Association set	EOR	p
Whole blood
	Human intelligence GWAS	1.02	0.401
	Educational attainment GWAS	0.84	0.994
	MDD GWAS	1.26	0.146
	MDD MWAS	1.03	0.328
	Psychosis MWAS	1.06	<1×10–06
	Nicotine dependence GWAS	1.07	0.143
T-cells
	Human intelligence GWAS	1.01	0.225
	Educational attainment GWAS	0.87	0.983
	MDD GWAS	0.93	0.747
	MDD MWAS	1.03	0.286
	Psychosis MWAS	1.05	0.011
	Nicotine dependence GWAS	1.00	0.899
Monocyte
	Human intelligence GWAS	1.15	0.116
	Educational attainment GWAS	0.91	0.998
	MDD GWAS	0.93	0.748
	MDD MWAS	1.01	0.55
	Psychosis MWAS	1.06	<1×10–06
	Nicotine dependence GWAS	1.09	0.054
Granulocyte
	Human intelligence GWAS	1.03	0.456
	Educational attainment GWAS	0.95	0.958
	MDD GWAS	1.32	0.098
	MDD MWAS	0.97	0.999
	Psychosis MWAS	1.01	0.884
	Nicotine dependence GWAS	1.02	0.760
B-cells
	Human intelligence GWAS	1.15	0.042
	Educational attainment GWAS	1.14	0.008
	MDD GWAS	0.83	0.792
	MDD MWAS	0.98	0.985
	Psychosis MWAS	0.86	1.000
	Nicotine dependence GWAS	1.07	0.143

Note: EOR = enrichment odds ratio; GWAS = genome-wide association study; MDD = major depressive disorder; MWAS = methylome-wide association study.

Discussion

The impact of cannabis involvement during adolescence is a pressing public health question due to possible links to negative outcomes and increasing use rates. We used genome-wide methylation data to study the association between problematic adolescent cannabis use and methylation differences in blood. Associated methylation differences may help identify novel hypothesis for the downstream effects of problematic adolescent use.

Results from the MWAS in whole blood identified several associated methylation markers with many of the associated methylation marks located in genes of importance for potentially relevant brain functions. The most significant finding was located in CLMN, which encodes the protein CALMIN. Gene expression studies have shown that CLMN is expressed in adult human brain tissue and Clmn is expressed in neurons in the developing mouse brain where it regulates neurite outgrowth during neuronal differentiation³¹. While there have been no previous links between cannabis and this gene, our results indicate that problematic adolescent cannabis use may potentially impact CLMN expression through methylation thereby potentially disrupting neuronal development and function. Another top finding in whole blood was SENP7, which encodes a regulatory protease of the Small Ubiquitin-related Modifier (SUMO) proteins, and has been linked to neuronal function³³. Cannabis has been shown to induce expression of SUMO proteins in neurons³⁴, where SUMOylation plays important roles in synaptic organization and function³⁵. Together, this suggests that cannabis-associated methylation may influence neuronal maturation via protein SUMOylation pathways.

Cell-type Specific MWAS

As different blood cell types may have different methylation profiles, and knowing which specific cell types harbor effects being key for biological interpretation and designing follow-up experiments, we conducted a cell-type specific MWAS that yielded several interesting results not detected in the whole blood MWAS. We focus our discussion on the granulocyte and B-cell results, which were enriched for chromatin states related to transcription in brain. Pathway analysis for the granulocyte MWAS implicated Slit-Robo signaling and axon guidance pathways. Slit-Robo signaling is known regulate axonal tract formation, and thereby axon guidance, during development³⁶ and endocannabinoids are known to modulate axon guidance by coordinating Slit2/Robo1 signaling³². This modulation of Slit-Robo signaling by endocannabinoids continues into adolescence where perturbations to endocannabinoid signaling introduces unwanted modifications in neural transmission³², potentially inducing cognitive deficits and neuroinflammation. However, Slit2-Robo1 signaling is also known to inhibit the migration of neutrophils, the main cellular component of granulocytes (i.e., CD15+), during inflammatory responses³⁷. While no study has investigated the impact of cannabis on Slit2-Robo1 signaling on neutrophil migration, it may partly explain the anti-inflammatory effects of cannabis, but would require further investigation. Taken together, these findings imply that cannabis-associated methylation may alter Slit2-Robo1 signaling thereby affecting axon guidance and inflammatory responses.

Inspection of both the individual CpG and pathway results for the B-cell MWAS revealed a theme around DNA repair. Two of the three genic regions that were significant, TRRAP and ERRC2, are directly involved with DNA repair and are expressed in human brain tissue. TRRAP encodes for a protein common to histone aceltyltransferases (HATs) and appears to responsible for recruiting HAT complexes to chromatin during DNA repair³⁸. ERRC2 is involved in base excision repair which ensures the genomic stability of CpG sites by preventing damaging mutations and regulates demethylation³⁹. Further, the pathway results also indicated DNA repair specifically around the resolution of displacement loop (D-loop) structures. A D-loop is a DNA structure in which two strands of DNA are separated by a third strand of DNA and occurs during the process of repairing double-strand breaks (DSBs). DSBs are the most dangerous form of DNA damage, however, they are programmed to occur during B-cell development and maturation for cell differentiation⁴⁰. If these pre-programmed DSBs are not repaired correctly, they can lead to immune and neurological defects⁴⁰. Methylation is a critical feature of DSB repair⁴¹ and previous work has demonstrated other drugs of abuse can alter methylation in DNA repair genes⁴². Taken together these findings suggests that problematic cannabis use may alter methylation needed for DNA repair in B-cells, with potential downstream effects on immune and neurological functioning.

MWAS and Cannabis Associated Mental Health Risks

Encouraged by studies reporting shared genetic risk for multiple outcomes, we explored possible overlap between our identified methylation marks and previously reported associations for the negative mental health effects previously linked to adolescent cannabis involvement (e.g., cognition, depression, psychosis and substance use). Two of our overlapping findings were between the B-cell MWAS, and two independent GWAS meta-analyses of educational attainment²⁶ and intelligence²⁵. There is evidence linking adolescent cannabis use to deficits in cognition⁴³ where potential explanations include the possibility that cannabis-associated methylation may impact how genetic variants relate to cognitive function. A recent study proposed a similar hypothesis where cannabis-associated hypomethylation in DRD2 (dopamine receptor D₂) in blood may reflect defects in the brain reward pathway⁵. Our methylation results overlapping genetic variants involving cognitive deficits specifically involved B-cells (i.e., CD19+). Whereas B-cells have been implicated in cognitive deficits in stroke and Alzheimer’s patients⁴⁴, they have not yet been studied in healthy patients. These loci linked to both cognitive function and cannabis-associated methylation may also simply reflect loci affected downstream of cannabinoid signaling. Cannabinoid receptor 2 (CB₂) is the major cannabinoid receptor expressed by immune cells and is most highly expressed in B-cells⁴⁵. Therefore, the cognitive effects of problematic adolescent cannabis use may be mediated by methylation at cognition-linked loci as an effect of cannabinoid receptor activation. In such a model, methylation at these loci in B-cells could mirror similar effects in brain.

The other overlapping finding was between an MWAS for psychosis conducted in brain tissue, and the T-cell and monocyte (i.e., CD3+ and CD14+) MWASs. Studies have demonstrated that there is a relationship between adolescent cannabis use and psychosis and that this relationship is even stronger in heavy or frequent adolescent cannabis users⁴⁶. One hypothesis for this relationship is that cannabis primes the immune system towards psychosis susceptibility. Specifically, cannabinoids alter cytokine production in most immune cells, including T-cells and monocytes⁴⁷, and in the central nervous system (CNS). Cytokines are required for normal CNS development and are thought to play a role in the genesis of psychosis by modulating neurotransmission systems and neuroinflammation. A recent review suggested a possible role for methylation in this process where cannabis use induces methylation changes, which alters cytokine production thereby increasing the liability of psychosis⁴⁸. A second explanation for our overlapping findings is that the psychosis MWAS is detecting cannabis effects due to higher cannabis use levels in patients with psychosis, suggesting the need to control for cannabis use in future methylation studies of psychosis. However, this overlap provides preliminary evidence that our cannabis findings in blood may track methylation in brain.

Our findings must be interpreted in the context of potential limitations. First, problematic cannabis use was assessed through self-reports during a structured clinical interview, and participants may have innacurately reported their current cannabis use. Second, in order to assess cell-type specific methylation we utilized a deconvolution approach. Ideally, cell populations of interest would have been isolated prior to assaying methylation. However, it is not feasible to sort cells from dried blood spots. Even if we could have sorted the blood cells in all of our samples, performing the assays on all cell types considered for all subjects would be practically and economically challenging. Third, the mechanisms for the impact of problematic cannabis use on the methylome are likely complex and involve interplay with clinical correlates such as trauma as well as differences demographic variables. Important clinical correlates like trauma exposure and demographic differences were not considered due to the low power of such analyses that would necessarily involve interaction terms. Biological sex is specifically important to consider as there is evidence suggesting that sex hormones may impact cannabis sensitivity⁴⁹. We re-analyzed our data using only male participants, which comprised the majority of our sample, to confirm that the direction of effect seen in the overall sample was maintained (see Tables S2,S4–S7, available online). Fourth, while the enrichment testing demonstrated significant overlap between our MWAS results and external datasets, the observed overlap does not imply that specific pathways exist between problematic cannabis use and the other outcomes, nor does it allow for disentangling confounding from causal effects. Fifth, the results from our study did not match the results from previous cannabis methylation studies. Lastly, interpreting peripheral blood-based methylation differences as etiological should be done with caution as many of the health risks of adolescent cannabis use likely involve brain.

Future research directions include replicating our findings in independent samples. Additionally, longitudinal studies with methylation measurements occuring both before and after problematic use onset are needed to disentangle whether the associated site are susceptibility loci and therefore may have been present before problematic use onset, while others occurred after problematic use onset and may instead reflec the post-use state. After replication, functional studies such as using clustered regularly interspaced short palindromic repeats-associated protein 9 (CRISPR-Cas9)-based technologies would enable targeted alterations to test the downstream functional effects of replicated methylation sites in the associated cell type experimentally. Further, the biomarker potential and clinical utility of replicated methylation sites could be explored in future work using specific criteria and steps described by the Biomarkers Definition Working Group⁵⁰.

This work involves one of the first human methylation-wide problematic adolescent cannabis use studies performed to date. We identified novel methylation sites and detailed several testable hypotheses by which cannabis-associated methylation can impact neurological development, inflammation response, as well as potential mechanisms linking cannabis-associated methylation to the downstream mental health effects of problematic adolescent cannabis use. Although further studies are needed to validate our findings, our results suggest methylation as a promising new mechanism in adolescent cannabis use research.