This longitudinal randomised study (n=37) in detoxified patients with alcohol use disorder examined DNA methylation changes after psilocybin 25 mg versus placebo and found one psilocybin-linked methylation site, plus broader changes related to depression and drinking measures. The changes involved genes and pathways linked to neuroplasticity and immune function.
The serotonergic hallucinogen psilocybin has shown potential as a treatment for psychiatric conditions like alcohol use disorder (AUD) and depression in clinical studies. Epigenetic mechanisms, including DNA methylation, are hypothesized to contribute to its lasting therapeutic benefits. In this exploratory study, we present the first methylome-wide analysis of psilocybin-induced changes in a cohort of detoxified patients with AUD. The longitudinal study design included three assessment days in 37 patients with blood sampling and acquisition of psychometrics – at baseline, 24 h after administration of psilocybin (25 mg) or placebo (mannitol), and one month after treatment. As the primary endpoints (duration of abstinence and mean alcohol use) in this trial were not reached, our investigation included secondary psychometrics that differed significantly between groups: Beck’s Depression Inventory and Beck’s Hopelessness Scale. The epigenome-wide association study (EWAS) identified one CpG site in TLE4 ( p = 1.1e-7) associated with psilocybin treatment. Screening for differentially methylated regions, we observed altered methylation in the gene RASGRP4 ( pFDR = 3.2e-4). Network analysis revealed co-methylation modules related to psilocybin treatment, as well as modules associated with the reduction of depressive symptoms and drinking behavior. Gene ontology analysis indicated involvement of these modules in neuroplasticity and immune functions, suggesting that they may reflect abstinence-related recovery processes. Investigating candidate genes at nominal significance ( p < 0.05) uncovered promoter-associated methylation changes in HTR2A and TNF . Interestingly, several of the reported analyses point to immunomodulatory actions of psilocybin. While the findings of this pilot study are limited by the modest sample size, they align well with previous literature and might provide starting points for further, large-scale investigations or hypothesis-driven experiments.
Alcohol use disorder (AUD) remains a major cause of morbidity and mortality, and currently approved medications such as acamprosate, disulfiram, and naltrexone have only modest effectiveness and often require regular dosing, which can limit adherence. The introduction argues that psilocybin may produce more durable reductions in alcohol use after one or a few administrations, but its biological mechanisms are still uncertain. Previous work has suggested that psychedelic effects may involve changes in gene expression, neuronal plasticity, network connectivity, and possibly epigenetic regulation such as DNA methylation, yet human evidence on methylome changes has been limited. Urban and colleagues therefore set out to conduct the first epigenome-wide association study (EWAS) of psilocybin in patients with AUD. Using longitudinal blood samples from detoxified participants in a randomised clinical trial, the researchers aimed to identify psilocybin-associated methylation changes, explore whether such changes related to reductions in depressive symptoms, and examine whether baseline methylation differed between treatment responders and non-responders. The study is presented as exploratory and intended to generate hypotheses about potential epigenetic and immunological mechanisms underlying psilocybin’s effects.
The study analysed data from a randomised, placebo-controlled, double-blinded, parallel-group trial conducted at the Psychiatric University Hospital in Zürich, Switzerland. The trial enrolled detoxified adults with DSM-5 AUD, excluding those with major psychotic disorders, suicidality, other substance use disorders apart from alcohol and nicotine, or unstable medical conditions. Although power calculations had indicated a target sample of 60, only 37 participants completed the trial, and not all provided blood at every time point. The methylation analyses therefore used blood samples from 37 participants at baseline, 37 at 24 h after dosing, and 34 at about 4 weeks after dosing. The sample included 13 women and 24 men, with a mean age of 37.35 years. Participants received either 25 mg oral psilocybin or an inactive placebo (mannitol) on the dosing day. The trial included preparatory and integrative psychotherapeutic sessions, but the methylation paper focuses on the drug comparison and related clinical measures. The main clinical outcomes were mean daily alcohol use over 4 weeks and time to relapse, while secondary outcomes included Beck’s Depression Inventory (BDI) and Beck Hopelessness Scale (BHS). For downstream methylation analyses, the researchers calculated within-person change scores from baseline to 4 weeks for BDI and BHS. DNA was extracted from EDTA-treated whole blood, and methylation was measured on the Illumina Infinium MethylationEPIC BeadChip. After preprocessing with the CPACOR pipeline, 817,247 CpG sites were retained for analysis. The main EWAS used mixed linear models on M-values with treatment group, time, and their interaction, while adjusting for patient ID, estimated cell composition, control-probe principal components, sex, age, smoking, and pre-withdrawal alcohol intake. The primary longitudinal effects of interest were the time-by-group interactions at 24 h and 4 weeks. The researchers also performed differential methylated region analysis with dmrff, weighted correlation network analysis (WGCNA) on the most variable CpGs, gene ontology overrepresentation analysis, candidate-gene analyses, and an exploratory mediation analysis examining whether methylation changes mediated effects on depressive symptoms. Because the study was small, the authors report a suggestive EWAS threshold of p = 1e-5 and note that false discovery rate (FDR) correction was also used where relevant.
The clinical outcomes reported in this paper were mostly not significant for alcohol-related measures. Mean duration of abstinence was 11.24 days after placebo and 17.77 days after psilocybin (p = 0.095), and mean daily alcohol intake was 1.39 units after placebo and 0.84 units after psilocybin (p = 0.331). In contrast, the secondary mood-related outcomes improved significantly in the psilocybin group: change in BDI from baseline to 4 weeks was -0.41 in the placebo group versus -6.18 with psilocybin (p = 0.017), and change in BHS was 0.65 versus -1.59, respectively (p = 0.017). In the EWAS, one CpG site showed a significant psilocybin-by-time interaction at the suggestive threshold. This site, cg23107740, was annotated to TLE4 and showed a baseline group difference that was reversed by 4 weeks (t = -4.4, p = 1.1e-4, Δβ = 0.02, conditional R2 = 0.48). A second nearby site was annotated to LINC01250, but the main report emphasises the TLE4 finding. Differentially methylated region analysis identified two psilocybin-related regions at 24 h and two at 4 weeks. At 24 h, one intergenic DMR and one DMR spanning RASGRP4 were detected; the RASGRP4 region remained significant after FDR correction (pFDR = 3.2e-4). At 4 weeks, one intergenic DMR and one DMR in LOC101805491 were identified. Within the RASGRP4 region, only cg14565721 also showed a significant cross-sectional difference, with hypermethylation 24 h after psilocybin (p = 0.041, Δβ = 0.02). WGCNA yielded 34 co-methylation modules. Several module eigengenes correlated with the variables of interest: the pink module correlated with treatment group and with change in hopelessness (group r = -0.24, p = 0.044; ΔBHS r = 0.56, p = 5.2e-7), the lightgreen module with change in depression (r = 0.48, p = 2e-5), the lightcyan module with mean alcohol use (r = 0.47, p = 3.5e-5), and the green module with duration of abstinence (r = 0.27, p = 0.021). Gene ontology analysis suggested biological themes including neurodevelopment, immune function, cell cycle regulation, synaptic transmission, intracellular protein regulation, and calcium signalling, but none of these enrichments survived FDR correction. Among 330 candidate CpG sites in genes linked to serotonin signalling, plasticity, inflammation, glucocorticoid signalling, and epigenetic regulation, 19 were nominally significant at 24 h and 16 at 4 weeks for time-by-group effects, although none remained significant after FDR correction. Three of these also met the cross-sectional effect-size criterion. In a highly exploratory responder analysis within the psilocybin group, 12 candidate CpGs differed nominally at baseline between those who later remained abstinent and those who did not. The candidate sites implicated genes including BDNF, NTRK2, EGR2, FOSB, DRD1, DRD2, GRM1, HTR1A, SLC6A4, and NR3C1. The mediation analysis did not find significant effects of methylation on the outcome models for either BDI or BHS, so the authors did not pursue mediation further.
Urban and colleagues interpret the study as the first methylome-wide examination of psilocybin-induced blood DNA methylation changes in a clinical AUD sample. They argue that the findings identify CpG sites and co-methylation modules that may be relevant to psilocybin’s effects, but they present these signals as preliminary and hypothesis-generating rather than definitive biomarkers. The authors place the TLE4 EWAS signal in the context of transcriptional regulation, development, immune processes, and earlier preclinical work on addiction-related behaviour. They suggest that the RASGRP4 DMR may point towards immunomodulatory effects, noting that psychedelics have been associated with reduced neuroinflammation and that inflammatory processes are upregulated in AUD. However, they also emphasise that the negative mediation analysis argues against a direct mediating role for the identified methylation changes in symptom reduction. For the WGCNA results, they distinguish between modules that correlated with psilocybin treatment and depressive symptom change, versus modules associated with abstinence or alcohol use that may reflect recovery processes rather than drug effects. In particular, the pink module is presented as the most plausible candidate for a putative mediation role because it correlated with both treatment group and change in depression. The candidate gene findings are interpreted as consistent with psilocybin-related effects on serotonin and immune signalling. The authors highlight promoter hypomethylation in HTR2A, the main molecular target of psychedelics, and transient hypomethylation in TNF, which they relate to short-term changes in immune signalling. They also note nominal baseline differences between responders and non-responders in genes involved in neuronal plasticity and neurotransmission, but frame these as only a possible starting point for future biomarker research. The limitations are substantial. The primary alcohol-related endpoints of the parent trial were not significantly improved, so the dataset cannot support a definitive biomarker for AUD treatment. The sample was small and underpowered for conservative EWAS thresholds, meaning that most findings did not survive multiple-testing correction and effect sizes were generally small. The authors also state that single-dose psilocybin is unlikely to produce strong or persistent methylation effects. In addition, the use of blood rather than neuronal tissue limits mechanistic interpretation, although the authors argue that blood is pragmatic and translationally useful for biomarker discovery. They conclude that future work should replicate these findings in larger trials and incorporate molecular endpoints, with additional studies in blood, cell culture, or animal brain tissue to test mechanisms more directly.
The authors conclude that this exploratory study suggests novel epigenetic associations with psilocybin in AUD, including changes in genes involved in serotonin and immune signalling and possible methylomic markers of treatment responsivity. They propose that future psilocybin trials should include molecular endpoints to allow replication and cross-study comparison, with the longer-term aim of identifying clinically useful biomarkers. If confirmed, they suggest these findings would support an immunomodulatory component to psilocybin’s anti-depressive and possibly anti-addictive effects.
This research is based on the study Clinical and Mechanistic Effects of Psilocybin in Alcohol Addicted Patients (clinicaltrials.gov identifier: NCT04141501; Kofam identifier: SNCTP000003445) conducted at the Psychiatric University Hospital in Zürich, Switzerland, by Rieser et al.,. The study was a randomized, placebo-controlled, double-blinded, parallelgroups trial, which was completed by 37 AUD patients. Randomization accounted for age, sex, and AUD severity. Randomization and blinding are described in greater detail in the original publication. Note that a sample size of 60 participants was determined by power analysis, but could not be reached due to delays related to the COVID-19 pandemic. Out of the 37 patients that completed the study, three refused to give blood samples. For additional three patients that dropped out before the last blood sampling (two from the placebo group, one from the psilocybin group), blood methylation data for the first two sampling time points was available. Non-completion of the study was associated with the drug consumption in all three cases (two patients relapsed on alcohol, one consumed cocaine after the second blood sampling). Main inclusion criteria for the trial included an AUD diagnosis according to DSM-5 criteria, as well as detoxification from alcohol 6 weeks prior to enrolment in the study. Patients were excluded in case of major psychiatric comorbidities (schizophrenia, schizoaffective disorder, or psychosis) or a family history thereof, suicidality, substance use disorders other than alcohol and nicotine, or unstable medical conditions. We included the blood samples from the drop-out participants in the EWAS described below, but not for the analyses that related methylation levels to psychometric data, as this data was not available for these patients. Within the 37 participants (female: 13, male: 24) that made up our sample, the average number of fulfilled DSM-V criteria was 7.51 (SD: 2.79), body mass index (BMI) 24.71 (SD: 3.66), and 21 were smokers. The age within the cohort ranged from 21 to 58 years (mean: 37.35; SD: 12.49). The timeline of the trial (Fig.) included the acquisition of three blood samples: T1 (n = 37; baseline, two weeks before dosing visit), T2 (n = 37; one day after dosing visit), and T3 (n = 34; around four weeks after dosing visit). During the dosing visit, the treatment group (n = 18; after drop-out: n = 17) was administered 25 mg of psilocybin (acquired from Usona Institute, Madison, Wisconsin) orally while the placebo group (n = 19; after drop-out: n = 17) received an inactive placebo (mannitol). Blinding in clinical trials with psychedelics is challenging, given their subjective effects. Mannitol, a sugar, was chosen as placebo because it does not produce subjective effects, allowing us to investigate the mechanisms of action of psilocybin. Primary outcomes of interest were daily mean alcohol use in the four weeks after the dosing visit, as well as the time to relapse (≥1 standard unit of alcohol per day). Besides the primary outcomes, we included two secondary psychometric scores: Beck's Depression Inventory (BDI)and Beck Hopelessness Scale (BHS). For these variables, we calculated subject-wise Δ-values between T3 and T1 and used them for downstream analyses of the methylation data. More detailed information on study design, participants, inclusion and exclusion criteria, as well as psychometrics can be found in the original publication.
Create a free account to open full-text PDFs.
Papers cited by this study that are also in Blossom
Bogenschutz, M. P., Forcehimes, A. A., Pommy, J. A. et al. · Journal of Psychopharmacology (2015)
Calder, A. E., Hasler, G. · Neuropsychopharmacology (2022)
Rieser, N. M., Herdener, M. ;., Preller, K. H. · Current Topics in Behavioral Neurosciences (2021)
De La, M., Revenga, F., Zhu, B. et al. · Cell Reports (2021)
Ruffell, S. G. D., Netzband, N., Tsang, W. et al. · Frontiers in Psychiatry (2021)
Rieser, N. M., Bitar, R., Halm, S. et al. · EClinicalMedicine (2025)
Dinis-Oliveira, R. J. · Drug Metabolism Reviews (2017)
Ray, T. S. · PLOS ONE (2010)
Rickli, A., Moning, O. D., Hoener, M. C. et al. · European Neuropsychopharmacology (2016)
Moliner, R., Girych, M., Brunello, C. A. et al. · Nature Neuroscience (2023)
Mason, N. L., Szabo, A., Kuypers, K. P. C. et al. · Brain Behavior and Immunity - Health (2023)
Vohryzek, J., Cabral, J., Lord, L. D. et al. · Brain Communications (2024)
Dougherty, R. F., Clarke, P., Kuc, J. et al. · Psychopharmacology (2023)
The clinical part of this study was approved by Swiss legal agencies (Cantonal Ethics Committee, Swiss Agency for Therapeutic Products, Federal Office of Public Health [BAG]), and adhered to the revised declaration of Helsinki from 2000, as well as guidelines for Good Clinical Practice (GCP). Data sharing and processing adhered to the data protection laws outlined in the General Data Protection Regulation (GDPR) of the European Union. Patients gave informed consent to all experimental and data processing procedures.
DNA extraction and methylation analysis were carried out at Life&Brain GmbH in Bonn. 10 ml EDTA-treated whole blood samples were used for DNA extraction via Chemagen Chemagic Systems technology. Extracted DNA was screened for genotypic variants on Illumina's Infinium Global Screening Array-24 (GSA) v3.0 to include single-nucleotide polymorphism (SNP) outlier analysis in the pre-processing of methylation data. After bisulfite conversion of the DNA, CpG methylation was assessed on Illumina's Infinium MethylationEPIC BeadChip v2.0, yielding raw data with or placebo (mannitol) was administered on day 0, the dosing visit, embedded within a broader psychotherapeutic framework incorporating preparatory and integrative sessions. Follow-up assessments were conducted after T3 to monitor longer-term outcomes. unmethylated and methylated signal intensities for each of the ~950,000 probes stored in idat files, which were then processed as described below.
All statistical analyses were performed in the R statistical environment (version 4.2.1 and 4.3.0;).
Preprocessing was based on the CPACOR pipelineand included filtering for sample call rate, sex mismatches, genetic outliers, and crossreactive probes, as described previously. For estimation of genetic outliers, genotype data were preprocessed as previously described, reduced to 20 dimensions by principal component analysis (PCA), and samples for which a component's loading coefficient differed by 4.5 standard deviations from the mean would have been removed. No such outliers were found. We estimated cell type compositionand performed PCA on cell type data and internal control probes to extract covariates for statistical modeling. Duplicated CpG sites on the EPIC array were excluded at random. Eventually, 817 247 CpG sites were included in statistical analyses.
Using the R package lme4, we calculated mixed linear models on the M-values, with group (psilocybin vs. placebo) as a between and time as a within factor. We included a random effect for patient ID to account for inter-subject variability, one cell type PC and two control probe PCs as covariates, as well as sex, age, smoking, and daily alcohol intake before withdrawal (in units per day). For variance decompositionand multicollinearity analysis, see Sup. Fig.and, respectively. With its small size (n = 37), this study is statistically underpowered to reach conservative EWAS thresholds (α ≈ 10⁻ 6 -10⁻⁸). Therefore, we report genomewide significance at a suggestive threshold of p = 1e-5, as commonly done in such scenarios. For the sake of completeness, p-values corrected for false discovery rate (FDR) are given in the Supplementary Tables. M-values were used to improve normality and variance stability relative to β-values, which is advantageous for linear modeling. Diagnostic plots (QQ-plots and residuals vs fitted values) confirmed normality and homoscedasticity for model fits with longitudinal p < 0.001 and |Δβ| > 0.02. Primary effects of interest in the linear model were the interaction effects time2*group and time3*group, indicating significant differences in change from baseline methylation between treatment and placebo groups, 24 h and 28 days after the intervention, respectively. Significant longitudinal effects were post hoc tested by running cross-sectional t-tests for the relevant time points (T2 or T3, respectively) on the beta values adjusted for the covariates (using R's predict function). We report goodness-of-fit for models with significant effects as variance explained by all regressors in the model(conditional R 2 calculated with performance library). Model estimates resulting in singular fits were excluded from subsequent analyses, leaving 649 975 CpG sites in the dataset. Results were annotated using the manufacturer's manifest (beadchip-kit/downloads.html). Non-annotated CpGs that appeared as relevant in the post hoc tests, i.e., significant effects with a magnitude of |Δβ| > 0.02 for the cross-sectional difference, were also screened onfor associated genes.
A sensitivity analysis using G*Power 3.1to estimate the effect size required to detect a deviation from zero of the total explained variance R 2 in a linear multiple regression model was calculated. Parameters used here were α = 1e-5, n = 37, Power = 0.8, and number of predictors = 9.
Sample sizes for the different downstream analysis are detailed in Fig.. Differentially methylated regions (DMR) were detected using the dmrff algorithm(maximal gap size: 1 000 bp; p-cutoff: 0.05). Visualization was based on the qqman packageand used the p-values for longitudinal effects. We considered DMRs with a minimum coverage of two CpG sites and report the results of cross-sectional post hoc tests for the affected CpGs. Weighted Correlation Network Analysis (WGCNA)was performed on the 5% most variable CpG sites (45 962 CpGs) from the two postintervention timepoints to derive co-methylation modules that capture potential treatment effects. Networks were constructed using the following parameters: soft power threshold = 3 (defined by the criterion of approximate scale-free topology: truncated R 2 > 0.90), minimum module size = 50, mergeC FigureutHeight = 0.25, and maxBlockSize = 46 000. In WGCNA, modules are labeled by colors. The module's eigen-CpGs (analogous to eigengenes), representing a weighted average of the module's expression profile, were calculated and correlated with phenotypic variables of interest: group, duration of abstinence, and mean alcohol intake during the 4-week post-treatment period, as well as Δvalues (T3-T1) for BDI and BHS. Eigen-CpG methylation values were winsorized to two standard deviations. For each variable, we report the module with the strongest correlation, including the relation between module membership (correlation between methylation of CpG site and module's eigen-CpG) and CpG significance (-log(p) of correlation between CpG methylation and trait of interest). Furthermore, we performed Gene Ontology Overrepresentation Analysis (GO ORA) using missMethyl. This was done on the CpG sites included in the co-methylation modules we report.
We also examined methylation changes at CpG sites in a selection of candidate genes chosen based on their proposed involvement in AUD and/or the effects of psychedelics. This selection comprised (i) receptors presumably involved in the effects of or targeted by psilocin, the active metabolite of psilocybin, namely HTR2A, HTR1A, SLC6A4, NTRK2, GRM2, DRD1, and DRD2; (ii) immediate early genes (IEGs) and plasticityrelated genes associated with addictive disorders and/or psychedelic drug action: EGR1, EGR2, FOSB, JUND, BDNF and SV2A; and a more heterogeneous group (iii) consisting of genes related to inflammation (TNF, IL6, CXCL8), glutamate (GRIN2B) and glucocorticoid (NR3C1) signaling, as well as epigenetic regulation (HDAC2) that are implied to play a role in AUD. 330 CpG sites annotated to candidate genes were retrieved and screened for nominally significant longitudinal effects (p < 0.05). Significant CpGs were post hoc tested for cross-sectional differences at T2 or T3, respectively, using t-tests on β-values adjusted for the covariates from the linear model. We report CpGs with significant (p < 0.05) cross-sectional differences that exceeded |Δβ| = 0.02. Furthermore, we screened for baseline differences in the candidate CpGs between treatment responders (<1 standard unit of alcohol during 4-week follow-up; n = 6) and non-responders (≥1 standard unit of alcohol during follow-up; n = 11). Due to the small sample sizes for this comparison (n = 17), non-parametric Wilcoxon tests were chosen, and results for p < 0.05 uncorrected are reported.
We conducted a mediation analysis to explore whether changes in methylation mediated the effects of psilocybin treatment on depression scores (ΔBDI/ΔBHS), focusing on eight CpG sites with prior significance. These
T2 T3 to focus the analysis on potential treatment effects. The responder analysis was conducted on participants of the psilocybin groups that completed the study, leading to n samples = 17 at T3, although the samples were taken from baseline T1. CpGs sites were the ones with cross-sectional differences of |Δβ| > 0.02, identified in the EWAS, the DMR, and the candidate analysis. Methylation changes (ΔT2-T1 or ΔT3-T1) were modeled by group (mediator model), and ΔBDI/ΔBHS was modeled by methylation and group (outcome model). Methylation values were adjusted using prior mixed model predictions. As methylation showed no significant effect in the outcome models, mediation analysis was not pursued further.
A detailed description of the clinical results can be found in Rieser et al.. Drinking-related outcome metrics did not show significant differences: duration of abstinence had a mean of 11.24 days after placebo versus 17.77 days after psilocybin (t = -1.7, df = 31.6, p = 0.095, Cohen's d = -0.59); mean daily alcohol intake was 1.39 units after placebo and 0.84 units after psilocybin (t = 1, df = 26.9, p = 0.331, Cohen's d = 0.34). We also analyzed group differences in ΔBDI and ΔBHS (Δ: T3-T1) and found significant effects: ΔBDI = -0.41 in the placebo group vs. ΔBDI = -6.18 after psilocybin (t = 2.5, df = 31.2, p = 0.017, Cohen's d = 0.87), as well as ΔBHS = 0.65 in the placebo group and ΔBHS = -1.59 after psilocybin for ΔBHS (t = 2.5, df = 31.9, p = 0.017, Cohen's d = 0.86).
Linear modeling identified one intergenic CpG site with a significant interaction effect (p < 1e-5; Fig.None of these CpG sites were annotated to genes in the Illumina manifest. Using, however, cg23107740 was annotated to transducin-like enhancer of split 4 (TLE4) and cg01405499 to the non-coding RNA LINC01250. Of note, cg23107740 displayed a significant group difference at baseline, which was inverted at time point 3 (t = -4.4, df = 35, p = 1.1e-4, Δβ = 0.02; R 2 = 0.48; Sup. Fig.).
Sensitivity analysis revealed a minimal effect size of f 2 = 2.34 for our models. Based on , a model fit needs to explain a proportion of at least R 2 = 0.7 in the methylation values to produce accurate findings with a power of 0.8 in our study.
Two DMRs showed psilocybin-dependent longitudinal effects at T2 and T3, respectively. DMRs are highlighted in Fig.and. One DMR associated with psilocybin-dependent changes at T2 was intergenic (n = 2 CpGs; z = 5.49; p FDR = 0.026), the other one covered CpGs in the gene RAS (rat sarcoma) guanyl nucleotide-releasing protein 4 (RASGRP4) (n = 4 CpGs; z = 6.23; p FDR = 3.2e-4). Longitudinal psilocybin-dependent effects at T3 also covered two regions: one intergenic DMR (n = 7 CpGs; z = -6.47; p FDR = 6.6e-5) and one in the non-coding RNA LOC101805491 (n = 2 CpGs; z = -6.08; p FDR = 8e-4; annotated using). Cross-sectional post hoc tests of the covered CpGs revealed that only one affected methylation site displayed a significant cross-sectional effect of |Δβ| > 0.02. This was cg14565721 in RASGRP4 gene showing hypermethylation 24 h after psilocybin (t = -2.1, df = 35, p = 0.041; Δβ = 0.02; R 2 = 0.88). The four CpG sites in this DMR lie within a 2 000 bp distance from the transcription start site of RASGRP4, suggesting potential involvement in transcription regulation.
WGCNA identified 34 co-methylation modules (median size: n = 206; range: n = 73-17612). Significant correlations between module eigengenes and the variables of interest occurred in the following modules: pink (group: r = -0.24, df = 33, p = 0.044; ΔBHS: r = 0.56, df = 33, p = 5.2e-7), lightgreen (ΔBDI: r = 0.48, df = 33, p = 2e-5), lightcyan (mean alcohol use: r = 0.47, df = 33, p = 3.5e-5), green (duration of abstinence: r = 0.27, df = 33, p = 0.021). For most modules, the relationship between modules and the respective variables was confirmed by strong correlations between module membership and gene significance for the CpG sites within the modules(Fig. Gene ontology overrepresentation analysis (GO ORA) GO term analyses on the co-methylation modules revealed enrichment of terms broadly related to neurodevelopment and (lightgreen module, Sup. Tab. 2.1), immune function and cell cycle regulation (pink module, Sup. Tab. 2.2), synaptic transmission and intracellular protein regulation (lightcyan module, Sup. Tab. 2.3), as well as calcium signaling and gene/protein regulation (green module, Sup. Tab. 2.4), among other functions. However, no enrichment survived correction for false discovery rate (FDR).
Among the 330 target CpG sites investigated, 19 reached nominal significance (p < 0.05) for the interaction effect of time and group at T2 and 16 for the interaction effect at T3 (see Sup. Tab. 3.1 & 3.2). None of these effects remained significant after FDR correction. Among the significant CpGs, three sites showed significant cross-sectional effects exceeding |Δβ| = 0.02 (see Fig. Next, we compared methylation levels of the candidate CpGs between responders (abstinent until 4-week follow-up) and nonresponders in the psilocybin group. This comparison needs to be pointed out as highly exploratory given the small sample size of n = 17. We identified 12 CpGs that displayed nominally significant differences before the psilocybin treatment (Table).
Upon testing the eight statistically and biologically relevant CpG sites identified in our analyses on psilocybin treatment, we observed no significant effects for methylation in the outcome models for either ΔBDI or ΔBHS, rendering mediation analysis obsolete.
We present the first methylome-wide exploration of psilocybininduced changes in blood DNA methylation in a clinical population (n = 37). In this analysis, we identified a number of CpG sites and co-methylation modules with potential relevance for psilocybin's effects that may support future hypothesis-driven research. In our EWAS, four CpG sites showed significant methylation changes after psilocybin. One of these was annotated to a gene, TLE4, where it is located in the gene body. TLE4 is a transcriptional co-regulator involved in developmentaland immunoregulatoryprocesses. Furthermore, TLE4 regulates maturation and maintenance of corticothalamic projection neuron identity, Schwann cell differentiation, and post-synaptic gene transcription at neuromuscular junctions. TLE4 has also been implicated in addictive behavior in a preclinical study on oxycodone self-administration. Given this context and the suggested role of structural plasticity in psilocybin's therapeutic effects, there may be a relationship between psilocybin-induced alterations in TLE4 methylation and potential neuroplastic effects of psilocybin in AUD. Furthermore, we discovered four DMRs associated with psilocybin-induced methylation changes. One DMR implicated in effects at T2 covered a gene, RASGRP4. This signaling molecule contributes to the development of mast cellsas well as the regulation of immune responses. Psychedelics, including psilocybin, possess immunomodulatory capacities, and reductions in neuroinflammation may contribute to their lasting psychological benefits. In AUD, on the other hand, (neuro) inflammatory processes are upregulated, seemingly exacerbating cognitive symptoms associated with this condition. Methylation changes in RASGRP4 may reflect, at least in part, psilocybin's immunomodulatory effects. While the RASGRP4 methylation change, as well as the TLE4 effect, represent psilocybin-associated effects co-occurring with reduced depression symptoms at the group level, a direct mediating role is unlikely for both, as indicated by the negative mediation analysis. WGCNA revealed several co-methylation modules associated with either treatment group or the drinking-/depression-related psychometrics, making the distinction of effect classes especially important in this analysis. Only the pink module showed significant correlation with the treatment, as well as with one of the psychometric measures (ΔBDI), suggesting that methylation of the involved loci might fulfill a mediating role in psilocybininduced relief of depressive symptoms. The involvement of genes relevant to immune function and cell cycle regulation in this module, again, indicates a relationship between the potential effects of psilocybin on the immune system and its anti-depressive capacities, as implied by previous research. Interestingly, neuroinflammation has been suggested as a link between AUD and major depression beforeand might represent a common target for psilocybin's effects across these disorders. The other modules correlating with ΔBHS, duration of abstinence, and mean alcohol use during the 4-week follow-up, on the other hand, did not relate to the psilocybin treatment and seem to fall in the category of effects related to abstinence or reduced symptom load. It is known that DNA methylation patterns in AUD change during prolonged abstinence, for instance, involving gene loci related to immune functionand neuroplasticity. As the modules covered biological processes related to synaptic transmission and gene transcription, they possibly describe such abstinence-related methylation changes that are independent of psilocybin treatment. The candidate gene analysis revealed evidence for psilocybininduced hypomethylation in two CpG sites within the promoter of HTR2A, which codes for the primary molecular target of psychedelics, the 5HT2a receptor. Aberrant methylation of HTR2A has been associated with psychiatric symptoms such as impulsivity in cocaine use disorder, or depressive rumination in people suffering from adverse childhood experiences. Normalization of such HTR2A methylation might thus lead to symptom relief in patients suffering from conditions like depression or AUD. We also observed a transient hypomethylation in a CpG site in the TNF promoter after psilocybin. This may represent a temporary influence on immune signaling. Interestingly, psilocybin decreases TNF blood levels in the short term. As with the EWAS and DMR results, these druginduced effects accompany reductions in depressive symptoms at the group level, without clear evidence for a mediating role. Lastly, the descriptive examination of baseline differences between responders to psilocybin treatment (abstinent at 4-week follow-up) and non-responders revealed nominally significant differences in several CpG sites related to neuronal plasticity (BDNF, NTRK2, EGR2, FOSB) and various neurotransmitter systems (DRD1, DRD2, GRM1, HTR1A, SLC6A4, NR3C1). The search for predictors of psychedelic treatment responsivity is ongoing and currently focuses on phenomena like the acute effects of psilocybin on brain activity and phenomenology, or changes in language patterns shortly after substance intervention. Less is known about molecular factors that could predict treatment responses before psychedelics are administered, and the genes identified here might provide a starting point for future research focusing on biomarkers of treatment responsivity.
This study is subject to some limitations. Firstly, as the primary endpoints of the RCT were not significantly improved, the present dataset cannot provide a definitive biomarker for AUD treatment. Nonetheless, the reductions in depressive symptomswhich frequently co-occur with AUDtogether with a lack of previous methylome-wide screenings of psychedelic effects in clinical populations, justify this exploratory investigation. Secondly, the results presented here need to be seen as hypothesisgenerating, since the sample size was modest and the analyses lacked sufficient statistical power to satisfy conservative α-thresholds commonly used in EWAS. Accordingly, most findings did not survive multiple comparison corrections, and most model fits remain below our R 2 threshold for sufficient power. Furthermore, on average, the effect sizes we observed are small. Consequently, a single administration of psilocybin appears unlikely to cause strong and persistent effects on DNA methylation. The power issues here affect the initial linear modeling as well as downstream analyses like the WGCNA or GO ORA and the investigation of baseline differences. Lastly, our methylome analysis is based on blood samples instead of neuronal tissue. Given the variability of methylomic signatures between tissue types, mechanistically interpreting the effects of psychopharmaceuticals based on blood methylome data remains speculative. A blood-brain concordance analysisof the significant results from our candidate analysis illustrates this issue: correlations for the same CpG site differed between cortical regions, while within a single brain region CpG sites from the same promoter region showed inconsistent patterns (Sup. Fig.). Despite this caveat, blood remains the most pragmatic tissue for clinical biomarker discovery which lends blood-based EWAS a high translational value.
This exploratory analysis presents potential novel epigenetic associations with psilocybin treatment for AUD patients, indicators for methylation changes in genes involved in serotonin and immune signaling, as well as possible methylomic predictors of treatment responsivity. Future RCTs on psilocybin should incorporate molecular endpoints to enable cross-study data integration with the prospect of identifying reliable biomarkers for clinical usage. If replicated, our findings suggest that immunomodulatory mechanisms contribute to psilocybin's anti-depressiveand possibly anti-addictiveeffects, thereby presenting a potential therapeutic avenue for comorbid populations. Apart from that, further research to confirm our findings could comprise probing the methylomic effects of psilocybin in blood and neuronal cell cultures derived from AUD patients or in brain tissue from rodent models of AUD.