2.1 Participants

The current exploratory study is part of a large prospective research project, ‘Studies in Neuroendocrinology and Neurogenetics in Alcoholism (NENA)’. The local Ethics Committee of the University of Erlangen-Nuremberg approved the original study. The Ethics Committee at Hannover Medical School approved the described investigation (approval number 10399_BO_K_2022). The investigation followed the Declaration of Helsinki. Before enrolment, each participant gave written informed consent to participate in the study. All patients included suffered from alcohol dependence according to DSM-IV and ICD-10. The study and analyses were not pre-registered. Clinical assessment was performed by a trained physician following a standardized protocol following the diagnosis criteria of ICD-10 and DSM-IV. Patients with comorbidities concerning mental disorders, other substances of abuse apart from alcohol or nicotine, the existence of severe somatic pathologies, known HPA-axis, or thyroid diseases were excluded from this study. After enrolment, all patients underwent a detailed physical examination by a professional (clinician), routine laboratory testing, and urine drug screening. As a control group, we included males who were negative for alcohol abuse, alcohol dependence, or other mental or somatic disorders according to ICD-10 or DSM-IV. Fasting blood samples were drawn between 8:00 and 10:00 AM. Then, all blood samples were centrifuged and stored at −80°C immediately after collection. The methylation rate of the OT and OXTR receptor genes has been investigated during three time points, namely on Day 1, Day 7 and Day 14 of alcohol withdrawal. Blood alcohol concentration (BAC) was measured at admission by using an enzymatic assay (Roche/Hitachi Ethyl Alcohol, Mannheim/Germany, 904/911: ACN 270). Besides, the extent of alcohol craving was obtained at admission and during detoxification treatment using the Obsessive-Compulsive Drinking Scale (OCDS), which provides a well-validated and reliable assessment [20]. For the assessment of withdrawal severity, we used the Clinical Institute Withdrawal Assessment for Alcohol Scale (CIWA) [21]. Severity of alcohol dependence was assessed with the ‘Scale for the Assessment of Alcohol Dependence Severity’ (in German ‘Skala zur Erfassung der Schwere der Alkoholabhängigkeit’, SESA) [22]. Depressive symptoms were measured with the Beck Depression Inventory (BDI) [23] and State–Trait Anxiety Inventory (STAI) [24], respectively.

2.2 DNA Isolation and Bisulfite Conversion

Extraction and clean-up of genomic DNA from blood were done using the NucleoMag Blood 200-μL DNA Kit (Macherey-Nagel, Düren, Germany). The bisulfite conversion of the acquired DNA samples was done using the EpiTect 96 Bisulfite Kit (QIAGEN, Hilden, Germany) for OXTR and EZ96 Magprep Bisulfite Kit (Zymo Research Europe, Freiburg, Germany) for OT following the manufacturer's protocol.

2.3 Primer Design

All primers were designed manually to bisulfite-converted regions of the respective genes using the program Geneious (Biomatters, Auckland, New Zealand). All fragments covered the proximal promoter region of the respective gene. For our analysis of the OT gene, we designed and used the forward primer F1 TTTGTTTTATTTTAGTGGTTTAGGTTAT and reverse primer R1 CTCAACTCCTAAAATTCTCAAA. The total fragment size was 512 bp. For OXTR, we designed the forward primer F1 GGTATTTTATTTTTTGTGTTTAGATTAT and reverse primer ACTCACTACAAACTCTACCTCC. The total fragment size was 379 bp.

We used the primer analysis tool Netprimer (http://www.premierbiosoft.com, accessed on 01.8.2022) to check for the presence of secondary structures (e.g., hairpins and self-primer). Melting temperatures were also retrieved from Netprimer analysis. The resulting primers were ordered from Metabion (Metabion, Steinkirchen, Germany).

2.4 Amplification of the Bisulfite-Converted Target Sequences

Polymerase chain reactions (PCRs) were performed in a C1000 Thermal Cycler (BIO-RAD, Hercules, CA, USA). The target sequences of the purified bisulfite-converted DNA were amplified following a standard PCR protocol. The reaction components used were as follows: 0.4 μL (20 μmol) forward primer (F1 for OXTR, F1 for OT), 0.4 μL (20 μmol) reverse primer (R1 for OXTR, R1 for OT), 1 μL of DNA, 3.2 μL of H₂O and 5-μL HotStarTaq Master Mix Kit (Qiagen, Hilden, Germany). The bisulfite primer amplification temperature was set at 53°C (OXTR)/57°C (OT) for the polymerase chain reaction. Amplification products were purified using the Agencourt AMPure XP magnetic beads (Beckman Coulter).

2.5 Sequencing

Sequencing PCR for the target fragment was performed using a BigDye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems, Foster City, CA, USA) and an Applied Biosystems/HITACHI 3500xl Genetic Analyzer (Applied Biosystems) according to the manufacturer's instructions. The bisulfite primers F1_OXTR and R1_OT were used in 5-pM concentrations for the PCR reaction of the respective genes. The sequencing PCR products were purified using the Agencourt CleanSeq XP magnetic beads (Beckman Coulter) and then used for sequencing. Electropherograms and sequences detected by the Genetic Analyzer were processed using the specialized epigenetic sequencing methylation analysis software to determine the methylation rates for every CpG locus.

2.6 Analysis of Methylation Rates

We used the Epigenetic Sequencing Methylation Software (ESME) software package to determine methylation rates. After normalization to the highest fluorophore intensity, ESME aligns the generated and reference sequences for comparing methylation at each CpG site. CpG islands were then labelled in relation to the location from the first base pair of exon 1 (i.e., CpGp134 would be 134 bp in the 5′ direction from the exon 1 [p = plus], and CpGm112 would be 112 bp in the 3′ direction from the exon 1 [m = minus]). We calculated quantitative methylation for each site per subject (proportion of cytosine and thymine normalized peak values) [25].

2.7 Statistical Analysis

We used the Statistical Package for Social Sciences Versions 27, 28 and 29 (SPSS, IBM, Armonk, NY, USA) and GraphPad Prism version 7 and 9 (San Diego, CA, USA) for all statistical analyses. The latter was further used for data illustration.

As a standard procedure in our laboratory, we performed quality control of our methylation analysis. We included only those CpGs with less than 5% missing values for each gene. Then, we excluded each sample with more than 5% missing values. After applying these criteria, the total number of CpGs for OT were 26 and OXTR were 36.

From visual inspection, normality was present in the methylation data of the OXTR but not OT gene. For comparative reasons, we decided to perform nonparametric tests to assess any differences between healthy controls and patients and changes throughout withdrawal. In detail, to test for differences in mean methylation levels scores across the withdrawal period, we used the nonparametric Friedman's test for dependent samples. To test for differences between healthy controls and the different timepoints, we used the Kruskal–Wallis test or Whitney U test. The Bonferroni correction was applied to correct for multiple testing.

To assess changes in gene methylation throughout the cessation period and the effect of craving and withdrawal on methylation levels, we fitted a mixed linear model (MLM). In this model, mean methylation was set as the dependent variable, while CpG position as well as time points were set as factors while withdrawal severity and craving were set as covariates. The resulting estimated marginal means (EMMs) were used to compare methylation values corrected for the influence of the factors. We used post hoc tests to compare mean methylation at the different time points (Bonferroni's method) and Bonferroni's method to correct for multiple testing. To compare the effect of sex and timepoint on the EMMs we fitted a one-way ANOVA with post hoc tests and correction for multiple testing (Tukey's method). We also included age and amount of cigarettes smoked into the mixed modelling and correlated them with mean methylation to estimate the confounding effect of these influential epigenetic modifiers.

3.1 Demographics

The study population consists of 99 patients suffering from AUD and 31 healthy male controls (Table 1). The study population is already described in detail elsewhere [26].

3.2 Methylation in Healthy Controls vs. Subjects (Baseline)

After quality control, 26 CpGs for the OT gene and 36 for OXTR were analysed (Figure 1). Because there was no difference in mean methylation across the OT gene in controls compared with patients across all time points, we individually plotted the methylation values of each CpG in the OT promoter fragment. From visual inspection, we identified a cluster ranging from base 42 to 151 across 11 CpGs (Cluster 11) (Figure S1).

Because normality was absent in the methylation data of the OT gene, we used the Friedman test to compare mean methylation across all CpGs and the cluster. There was no change in mean methylation across withdrawal when using non-parametric tests. Cluster 11 mean methylation changed significantly during cessation (Friedman statistic = 7.327, p = 0.0256).

Regarding the OXTR gene, mean methylation was significantly higher in healthy controls compared with all-time points (t14: χ²(3) = 137.895, p < 0.001; t7: χ²(3) = 89.325, p < 0.001; t0: χ²(3) = 92.602, p < 0.001). Furthermore, mean methylation decreased significantly during withdrawal therapy (Friedman statistic = 17.02, p = 0.0002; Figure 1).

3.3 Effect of Timepoint, Withdrawal Severity and Craving on Gene Methylation

In our model, CpG position, timepoint, withdrawal severity score and the interaction between timepoint and craving, timepoint and withdrawal severity and timepoint, craving and withdrawal severity had a significant fixed effect on OT gene methylation (Table 2).

Post hoc tests (Tukey) showed that mean methylation (based on expected marginal means, EMM; Figure 2) was significantly lower at t14 (mean = 0.267) compared with t0 (mean = 0.286) and t7 (mean = 0.285) (mean difference (MD) = 0.019, p = 0.004, 95% CI = [0.005–0.33]; MD = 0.017, p = 0.004, 95% CI = [0.004–0.30], respectively).

We also fitted a mixed linear model to analyse the effect of psychometrics and time points on OXTR methylation levels. In this model, only CpG-position had a significant effect on mean methylation levels (Table 3). Post hoc tests (Tukey) showed that mean methylation (based on EMM; Figure 2) was significantly lower at t14 (mean = 0.111) compared with t0 (mean = 0.135) and t7 (mean = 0.139) (MD = 0.024, p ≤ 0.001, 95% CI = [0.015–0.33]; MD = 0.028, p ≤ 0.001, 95% CI = [0.020–0.36], respectively).

Controlling for confounders, we modelled psychometry against age and amount of smoked cigarettes for both genes. Here, in the OXTR receptor, it shows no singular relevance of age and cigarettes on methylation, but combinatorial effects of timepoint and age (F = 3.085, p = 0.046), CIWA (F = 3.762, p = 0.023) and OCDS (F = 8.482, p < 0.01) reveal subtle influences on overall methylation. OT methylation shows significance for the single fixed effects of age (F = 4.685, p = 0.03) and amount of cigarettes (F = 43.968, p < 0.001), as well as for most combinatorial effects, which are driven by this main confounder (see Supporting Information S2). Because we did not observe a significant difference in mean methylation of the OT promoter in controls compared with patients, we conducted a post hoc power analysis. When implementing mean values and the number of subjects at an alpha level of 0.05, the post hoc power for the oxytocin promoter was 27.5%, while for the receptor it was 99.2% (see Supporting Information S3).

4.1 Limitations

Naturally, the study has several limitations, some of which we have already discussed in another study on oxytocin and OXTR methylation in tobacco use disorder [50]. Epigenetic research is fraught with challenges, such as tissue specificity, reversibility of epigenetic marks and technical variability. While being matched for sex, we did not match patients and controls for other factors such as age [51] and smoking status [52], which undoubtedly affect gene methylation. In our study, smoking (as assessed with average cigarettes per day) and age affected oxytocin but not OXTR promoter methylation. Furthermore, we did not observe an association with CIWA scores and methylation but with OCDS. However, regarding OXTR promoter methylation, the inclusion of age and smoking revealed an association of methylation levels with the factors timepoint * OCDS as well as timepoint * CIWA. Of note, this was also found in the OT model with and without the mentioned factors smoking and age. This suggests an association between psychometrics and methylation depending on the timepoint measured. Furthermore, our post hoc power analysis suggests that a larger sample size and matching for factors such as age and smoking status would be needed to identify a potential difference in mean methylation of the OT gene. OXTR promoter methylation, on the other hand, seems to be less influenced by the mentioned factors age and smoking and more by the main differentiator between the groups, namely, alcohol consumption. However, the fact that we only observe an interaction between timepoint and psychometric variables in OXTR promoter after inclusion of age and smoking reveals that both variables play a major role in the methylation levels.

Regarding sex differences, an increasing number of studies have presented evidence for a sex-specific effect of intranasal oxytocin on human behaviour [14, 53] and sex-specific differences in the oxytocin system [33-36]. As highlighted in the introduction, sex-specific differences in OXTR function and addiction behaviours are well established. Thus, the absence of female participants limits the broader applicability of our findings. Regarding tissue specificity, we only measured gene methylation in peripheral whole blood; therefore, our results do not allow any conclusions about changes in methylation levels in the brain or other specific tissues. Of note, while some studies describe an association between OXTR promoter methylation and OXTR expression in brain tissue [54, 55], one thorough review discusses why it is unclear of and how peripheral measurements reliably mirror neural processes [56, 57]. As discussed above, studies on the effect of alcohol withdrawal on methylation levels need to consider the direct effect of alcohol and withdrawal severity on peripheral blood cells. In the present study, we did not control the alcohol level or the autonomic system activation when the blood was drawn. With respect to the reversibility of epigenetic marks, our results suggest a certain reversibility over time because methylation values changed significantly over withdrawal. However, complete reversibility cannot be concluded from our data due to the study design (such as the short observational period).

Regarding the experimental/technical methods used in this study, namely, Sanger sequencing, it is important to point out that modern sequencing methods, such as target enrichment sequencing and Oxford Nanopore Technology Sequencing, have methodic advantages compared with Sanger [58]. On the other hand, Sanger sequencing is the gold standard for comparison and investigation of specific genetic loci. Concerning its accuracy for measuring single promoter regions, our research group has shown that it achieves comparable results with ONT-S when analysing promoter methylation [59]. Nonetheless, future studies investigating the Oxytocin system in addiction and across the sexes should use more modern sequencing techniques such as ONT-S. One methodological limitation of the present study is the difference in group sizes between controls and patients. In detail, this group difference and nonparametric testing increase the probability of a type two error, that is, not rejecting the null hypothesis even though the alternative hypothesis is true. Future studies would need more balanced samples to minimize such effects on reliability.

4.2 Conclusion

In summary, our study is the first to report an association between AUD and OT and OXTR gene methylation. Methylation of the OXTR gene is reduced in AUD compared with healthy controls, with OT gene methylation being linked to craving and withdrawal severity, possibly via the interplay between the oxytocin system and the HPA-axis. Our results suggest that future studies on the use of oxytocin as a therapeutic agent need to consider epigenetic regulation of its receptor and gene as a mechanism that could influence oxytocin's effect on craving and withdrawal symptoms.