Epimutations in migraine and in medication overuse headache: a longitudinal study after acute medication withdrawal

Abstract

Background

Headache disorders are among the leading causes of neurological disability worldwide, affecting over half of the global population. Recent data indicate that headache prevalence among young people increased by 23.7% between 1990 and 2021. Migraine, a highly disabling primary headache, significantly impacts daily life, yet timely and accurate diagnosis remains challenging. Many patients self-manage with analgesics, contributing to medication-overuse headache (MOH), which affects 1–2% of the general population. Emerging evidence suggests that epigenetic mechanisms, particularly DNA methylation (DNAm), play a key role in the pathophysiology of migraine and MOH.

Methods

We investigated epigenetic entropy by calculating Stochastic Epigenetic Mutations (SEMs) and epivariations in 57 participants of the longitudinal Epimode project. Patients were enrolled in three groups: Chronic Migraine with MOH (assessed before and at various timepoints after clinical treatment), Episodic Migraine (EM; assessed at baseline and follow-up), and Healthy Controls (HC; assessed at baseline and follow-up). Peripheral blood DNA methylation was analysed genome-wide, and SEMs were annotated by genomic context and functional category. Statistical analyses assessed group differences and temporal dynamics of SEMs.

Results

At baseline (T0), both EM and MOH patients exhibited significantly higher SEMs counts compared with HC, with no differences between EM and MOH groups, either in total counts or in hypo-/hypermethylated CpGs. Differences persisted across functional categories and chromosomes. Overall, no significant changes in SEMs counts were observed in MOH patients pre- and post-treatment overall. However, MOH non-responders displayed high SEMs counts at T0 and the greatest reshuffling rate between lost, maintained and gained SEMs over time. Enrichment analyses identified SEMs in genes associated with metabolic, synaptic, and inflammatory function.

Conclusion

These findings indicate that migraine promotes epigenetic entropy and suggest that epigenetic instability contributes to headache chronification and treatment resistance. Moreover, the dynamics of SEMs over time provide more informative insights than absolute counts alone in evaluating clinical outcomes. This study reinforces the relevance of epigenetic mechanisms in the clinical management of headache disorders.

Supplementary Information

The online version contains supplementary material available at10.1186/s10194-026-02343-w.

Background

Impact of headache disorders

Headache disorders are among the leading causes of neurological disability worldwide [1]. A systematic review revealed that 52% of the global population is affected by active headache disorders, although these statistics are subject to methodological factors [2]. Among individuals aged 5–24 years, prevalence reached approximately 811.7 million cases by 2021, a 23.72% increase since 1990 [3], underscoring a growing public health burden.

A European survey revealed that migraine, a disabling primary headache disorder, substantially impacts daily life, interfering with professional and social activities. Delayed diagnosis and limited access to specialist care often lead patients to self-manage with analgesics, increasing the risk of Medication overuse headache (MOH) [4]. MOH is defined as headache occurring on ≥15 days per month with regular overuse of acute medications and affects 1–2% of the general population [5]. It is frequently associated with progression or chronification of headache [6].

Migraine, MOH and epigenetics

Epigenetic mechanisms, particularly DNA methylation (DNAm), are increasingly implicated in migraine and MOH. DNAm regulates gene expression and is involved in synaptic plasticity and neuroinflammatory processes [7]. In MOH, longitudinal studies showed that DNAm changes are associated with a reduction in monthly headache days after detoxification treatment [8]. Although not demonstrating causality, these findings support a role for epigenetic regulation in headache biology. DNAm studies may help clarify disease mechanisms and identify biomarkers for diagnosis and prognosis [9].

Stochastic epigenetic mutations (SEMs) as a novel approach

Stochastic Epigenetic Mutations (SEMs) represent individual-level epigenetic deviations from population reference intervals, capturing a condition of epigenetic instability. Originally described in aging, where they increase exponentially over time [10], SEMs have since been associated with cancer risk [11], autism spectrum disorders [12], major depressive disorder [13], and Parkinson’s disease progression and mortality [14].

It is noteworthy that in the above-mentioned study on epigenetic profiling of MOH detoxification SEMs were not analysed. Our study therefore represents the first investigation on DNAm entropy in migraine with perspective to uncover potential mechanisms underlying the chronification of headache disorders and the toxic effect of medication overuse.

Methods

Cohort Description

The study was conducted on patients attending the Headache Centre of IRCCS of Neurological Sciences of Bologna, Italy. The study was approved by the Local Ethic Committee of the local services of Bologna, and the participation was subscribed by patients in an informed consent.

All patients were considered eligible if they were 18 years of age or older and capable of providing informed consent, both verbally and in writing. Exclusion criteria included pregnancy or breastfeeding, diagnosis of secondary headache, a personal history of substance abuse (such as alcohol, sedatives, cannabis, or other psychoactive drugs), and the presence of any serious ongoing medical or psychiatric condition. Secondary headaches were ruled out through clinical evaluation, supported by laboratory tests and neuroimaging when necessary.

Patients were enrolled in three groups:

1. Episodic Migraineurs (EM) - Episodic migraine patients included in the study met the diagnostic criteria for migraine with or without aura, as defined by the International Classification of Headache Disorders 3rd edition (ICHD-3) beta version [15]. Only individuals experiencing a low frequency of attacks—no more than seven episodes per month—were considered eligible. Patients undergoing prophylactic migraine treatment at baseline were excluded from enrollment.

2. Chronic migraine with medication overuse headache (MOH) - Patients with chronic migraine and medication overuse headache were included if they fulfilled the diagnostic criteria outlined by the ICHD-3. Specifically, eligible individuals experienced headaches on 15 or more days per month and had a prior history of migraine. Additionally, they had a pattern of regular overuse-lasting more than three months-of one or more medications typically used for acute or symptomatic headache relief, defined as follows:

   1. Use of ergotamine, triptans, or combination analgesics on 10 or more days per month
   2. Use of simple analgesics on 15 or more days per month
   3. Use of any combination of acute or symptomatic medications on 10 or more days per month, even without exceeding the threshold for any single drug class.

3. Healthy Control (HC)- Healthy controls were defined as individuals without a history of headache disorders. Screening was performed using the HARDSHIP questionnaire (Headache-Attributed Restriction, Disability, Social Handicap and Impaired Participation) [16], a validated tool that includes demographic data, diagnostic items aligned with ICHD-3. To determine eligibility, two specific questions from the questionnaire were considered: question 12, “Have you ever had a headache in your lifetime?” and question 13, “Have you had a headache during the last 12 months?”. Only participants who answered “no” to question 13 were included in the study.

Patients with opioid overuse were excluded due to their distinct clinical and behavioural profile, including a higher risk of substance misuse and specific management implications [17]. In our cohort, only two patients were excluded for tramadol overuse, confirming the limited representation of this subgroup.

Study design

The present analysis is based on data generated within the “Epimode project”: a pilot, longitudinal, prospective, observational study designed to investigate DNA methylation changes in headache disorders. In this study, an epigenome-wide association study (EWAS) approach was applied to characterise Stochastic Epigenetic Mutations in patients with Medication Overuse Headache, Episodic Migraine and Healthy Controls. Figure 1 provides an overview of the study design.

Fig. 1.

Study design. Schematic representation of the study design across groups. Assessments were scheduled at baseline (enrolment) for all groups. Healthy controls (HC) and episodic migraine (EM) participants underwent a follow-up evaluation at T1. In the medication overuse headache (MOH) group, baseline assessment was followed by in-hospital withdrawal, with subsequent follow-up evaluations at T2 and T3

MOH were patients who had been unresponsive to prophylactic treatments for at least three months, as determined by clinical history and headache diaries. These patients were enrolled at baseline (T0). At T0, patients received education regarding their MOH diagnosis, were advised to discontinue overused medications, and asked to stop any prophylactic therapy. The educational session involved a concise explanation of the disease’s nature and the risks associated with frequent use of headache medications.

Patients who continued to meet MOH criteria at the three-month follow-up (T1) underwent a supervised five-day inpatient withdrawal program. During this pharmacological detoxification, patients received paracetamol at doses of 1000 mg at 8:00 a.m., 1000 mg at 2:00 p.m., and 1000 mg at 8:00 p.m. Rescue medications permitted during withdrawal included intramuscular metoclopramide (10 mg) and lorazepam (1 mg or 2.5 mg orally). Subsequent follow-up visits took place at two months (T2) and six months (T3) post-withdrawal (Fig. 1).

EM and HC patients were evaluated at baseline (T0) and six months later (T1). All participants were provided with a headache diary during the initial visit to record headache frequency, intensity (rated on a scale from 1 = mild to 10 = severe), type and number of painkillers used, and days treated. Diaries were reviewed at every follow-up visit. Peripheral blood samples for DNA methylation analysis were collected at all timepoints (T0, T1, T2, T3) for MOH patients and at T0 and T1 for EM and HC participants.

Clinical and questionnaire-based assessments followed the procedures described in detail in the original Epimode cohort report by Terlizzi et al. [18]. These evaluations were performed at all timepoints (T0, T1, T2, T3) for MOH patients and at T0 and T1 for EM and HC participants. Neurological examinations and interviews were conducted by headache specialists.

In this work we focused on investigation of Stochastic Epigenetic Mutations and Epilesion in MOH, EM and HC of Epimode cohort.

Sample handling and data preprocessing

Genomic DNA was isolated from peripheral whole blood using QIAamp DNA Mini Kit (QIAGEN) following the manufacturer’s instructions, and bisulfite conversion was performed with EZ DNA Methylation-Gold Kit (Zymo Research). Genome-wide DNA methylation was assessed using the Infinium HumanMethylationEPIC BeadChip (Illumina). To minimize potential batch effects, all samples from all groups and timepoints were processed within the same laboratory workflow. Samples were randomized across arrays so that different groups and timepoints were evenly distributed on the chips.

Raw IDAT files were processed in R version 4.3.1 using the minfi Bioconductor package (version 1.48.0) and were normalized using functional normalization (preprocessFunnorm). Probes failing detection in any sample, located on sex chromosomes, containing SNPs, or previously reported as cross-reactive or multi-mapping were removed. Exploratory analyses, including density plots, multidimensional scaling (MDS), and inter-sample distance heatmaps, were performed to detect potential batch-driven structure or outliers. No evidence of systematic batch effects was observed, supporting the reliability of downstream analyses.

DNAm-based cell count estimation

Cell type composition was estimated from beta values using the IDOL algorithm, as implemented in the peripheral blood immune profiling deconvolution method [19]. This approach allowed quantification of 12 distinct immune cell populations, including neutrophils, eosinophils, basophils, monocytes, naïve and memory B cells, naïve and memory CD4⁺ and CD8⁺ T cells, natural killer cells, and T regulatory cells.

Stochastic epigenetic mutations identification and annotation

For each subject at all timepoints, stochastic epigenetic mutations (SEMs) were calculated following the method described by Gentilini et al. [10]. The β-values of healthy controls (HC) at baseline (T0) were used to calculate a reference interval for each CpG locus based on the interquartile range (IQR), defined as:

Lower threshold: Q1-(3xIQR)

Upper threshold: Q3+ (3xIQR)

where Q1 and Q3 correspond to the 25th and 75th percentiles, respectively, and IQR = Q3 – Q1.

β-value of every CpG site was compared with the corresponding reference interval and values below the lower threshold were classified as hypomethylated SEMs, and values above the upper threshold as hypermethylated SEMs. Outlier subjects with total number of SEMs exceeding ±3 standard deviations from the mean calculated within phenotype-timepoint group were excluded from the analysis.

Obtained SEMs were annotated with additional mapping details (if located within genic/non-genic region, if positioned within CpG island/non-island region and chromosomal coordinates) using the ChAMP R package (version 2.34.0), which provides probe annotation based on the Illumina Infinium HumanMethylationEPIC BeadChip manifest. Based on this information we grouped SEMs in four classes: Island_Genic, NonIsland_Genic, Island_NonGenic and NonIsland_NonGenic.

Epilesions

Epilesions were identified for each subject and timepoint using a validated approach adapted from Gentilini et al. [20, 21]. Briefly, a sliding-window analysis was applied across the annotated genome, and SEMs enrichment within each window was assessed using a hypergeometric test. Adjacent windows with significant enrichment were merged to define genomic regions exhibiting elevated SEMs frequency.

Normalization and statistical tests

We performed Kruskal–Wallis test to check for statistical differences in immune cell proportions between the groups. Generalized linear models adjusted for sex, age, and estimated immune cell proportions were used to assess group differences in log-transformed SEM counts. Benjamini–Hochberg approach was applied to correct for multiple tests. Analysis was performed in R (version 4.3.1).

Analysis framework

Epimode dataset allowed us to perform multidirectional analysis. Therefore, we performed baseline group comparison, longitudinal MOH comparison and treatment outcome-based comparison.

Baseline (T0) comparison

We compared SEMs counts across episodic migraine, medication overuse headache and healthy control groups, considering overall SEM burden (i.e., total number of epimutations) as well as CpG classification, chromosomal distribution, and methylation status (hypo- vs. hypermethylated).

Longitudinal MOH comparison (T0 vs T2)

In MOH subjects, SEMs were compared between baseline and follow-up (T2) at global, class-wise, and chromosome-wise levels.

Baseline methylation profiling according to clinical outcome at T2

MOH patients with methylation data at both T0 and T2 and available clinical outcome at T2 were stratified into EM_noOveruse, Chronic_noOveruse, and Chronic_Overuse groups. Reclassification was based on the clinical response after detoxification: patients with fewer than 15 headache days per month at T2 were considered episodic responders (EM_noOveruse), whereas those with ≥15 headache days per month were classified as chronic. Among chronic patients, those using acute medications on more than 15 days per month were further categorized as Chronic_Overuse. Healthy controls with methylation data at T0 and T1 served as a reference (T1 corresponding to MOH T2). SEMs fluctuations were evaluated in terms of intra-subject difference (ΔSEMs = SEMs_T2 – SEMs_T0) and persistence, classifying SEMs as maintained (present at both timepoints), lost (present only at baseline), or new (present only at follow-up).

Results

Quality control and cohort overview

As a first step, we characterized the distribution of participants across groups and time points, documented in Table 1. Overall, the cohort included 57 unique individuals (47 females). The mean age was comparable between sexes (p = 0.43, n.s.), being 48.6 ± 10.0 years for females and 52.9 ± 10.6 years for males. The groups were comparable in age at baseline (p = 0.15, n.s.), with overlapping age distributions (median ages: HC = 52.8 years, EM = 45.8 years, MOH = 49.8 years).

Table 1.

Distribution of participants across groups and time points in the Epimode cohort

Group	Count	F	M
HC_T0	13	8	5
HC_T1	11	7	4
EM_T0	18	17	1
EM_T1	20	18	2
MOH_T0	23	20	3
MOH_T1	20	17	3
MOH_T2	19	16	3
MOH_T3	18	15	3

Numbers indicate total participant count per group, stratified by sex (F, female; M, male). HC, healthy controls; EM, episodic migraine; MOH, medication-overuse headache; T0–T3, longitudinal assessment time points

All 142 samples passed initial quality control, with a mean detection p-value < 0.05, and after filtering steps, 660,284 high-quality probes were retained for downstream analyses.

Immune cell composition analysis

DNA methylation data were used to infer the immune cell composition of each subject. This type of analysis allows the detection of potential immunological heterogeneity within the study samples and helps to evaluate whether cell type–specific differences could influence methylation patterns. In our cohort, no significant differences in immune cell composition were observed between groups, either at baseline, in the longitudinal MOH comparison, or in the outcome-based stratification at T2.

SEMs calculation

Stochastic Epigenetic Mutations were calculated per subject in every timepoint basing on beta values. The number of SEMs per subject ranged from a minimum of 2412 (subject HC12 at T0) to a maximum of 30,715 (subject MOH8 at T3). Two subjects (EM11_T1 and MOH24_T0) were excluded from the analyses as they were identified as outliers. In Supplementary Fig. 1 (Additional file 1), the available data for all subjects included in the analysis are reported.

Baseline comparison

As a preliminary investigation, a comparative analysis was conducted between the three groups at baseline (n = 53) to detect any potential differences at the time of enrollment.

Epimutation burden in EM and MOH

At baseline, both EM and MOH participants had a significantly higher total number of SEMs compared to healthy controls (HC) (EM vs HC p < 0.001; MOH vs HC p < 0.001), whereas no significant difference was observed between EM and MOH (Fig. 2). This pattern was consistent for both hypomethylated and hypermethylated SEMs (Fig. 3). It is generally observed that individuals suffering from migraine, both with and without medication overuse, have a similar number of SEMs but both significantly higher number of epimutations in comparison to healthy controls at baseline.

Fig. 2.

Total number of SEMs at baseline (T0). Boxplots show log-transformed counts [log(sems tot)] in healthy controls (HC), episodic migraine (EM), and medication-overuse headache (MOH) patients. Significance levels: * p < 0.05, ** p < 0.01, *** p < 0.001

Fig. 3.

Distribution of hypomethylated (a) and hypermethylated (b) SEMs at baseline (T0). Boxplots show log-transformed counts [log(hypo SEMs) and log(hyper SEMs)] across healthy controls (HC), episodic migraine (EM), and medication-overuse headache (MOH) participants. Significance levels: * p < 0.05, ** p < 0.01, *** p < 0.001

SEMs analysis by CpG functional category and chromosome

The biological function of CpG methylation is largely determined by genomic location, with sites residing in CpG islands (genic or non-genic), gene bodies, or gene-desert regions. To explore whether SEM enrichment was driven by specific CpG functional subcategories, analyses were stratified by genomic location (genic vs non-genic) and CpG island context (island vs non-island), as well as by chromosome.

Both EM and MOH exhibited higher SEM burden than healthy controls across all CpG contexts, with no differences between the two patient groups. Hyper- and hypomethylated SEMs showed similar patterns. These results are illustrated and described in detail in the captions of Supplementary Fig. 2–5 (Additional files 2–5).

Analysis of SEM frequency across chromosomes showed consistent differences between HC and MOH, and between HC and EM. Full details are provided in Supplementary Tables 1–2 and Supplementary Fig. 6–7 (Additional files 6–9).

Longitudinal analysis in MOH (T0 vs. T2)

Methylation data at both T0 and T2 were available for 18 MOH subjects. These samples were analysed using the same framework applied at baseline, including total SEMs count and stratification by genomic region, chromosome, and methylation status (hypo- vs. hypermethylation). No statistically significant differences were observed between T0 and T2 (Supplementary Fig. 8, Additional File 10).

Epimutations and clinical outcome

We next investigated whether baseline SEMs (T0) could discriminate MOH patients according to their clinical outcome at T2. To this end, MOH subjects were reclassified at T2 based on treatment outcome, considering both medication overuse and migraine phenotype. Among the 17 MOH patients with available clinical outcome data, three groups were defined:

• EM_noOveruse: episodic headache, no medication overuse (n = 5)

• Chronic_noOveruse: chronic headache, no medication overuse (n = 5)

• Chronic_Overuse: chronic headache with continued medication overuse (n = 7)

For comparative purposes, 11 HC subjects with available methylation data at T0 and T1 were also included in the analysis.

Epimutation burden

SEMs were analysed only as total counts; no further stratification by CpG functional category, chromosome, or methylation status was performed.

When comparing HC to Chronic_Overuse, a statistically significant difference (p = 0.004) in total SEMs count was observed after false discovery rate correction (BH). Individuals who were classified as Chronic_Overuse at T2 already exhibited nearly three times more SEMs at T0 than HC, after adjusting for age, sex, and cell counts (Fig. 4). This indicates that the subgroup of patients who ultimately did not benefit from treatment was also the one with the highest SEM burden at baseline.

Fig. 4.

Baseline SEMs in HC and MOH participants stratified by T2 outcome. Boxplots show baseline log-transformed SEM counts [log(epimuttot)] in healthy controls (HC) and medication-overuse headache (MOH) patients classified according to their T2 clinical outcome. Three MOH groups are shown: EM_noOveruse (episodic headache, no medication overuse), Chronic_noOveruse (chronic headache, no medication overuse), and Chronic_Overuse (chronic headache with continued medication overuse). The Chronic_Overuse group showed significantly higher baseline SEM numbers compared to HC (** p = 0.004)

Longitudinal SEMs fluctuation

To further explore the clinical aspects, we performed a longitudinal analysis of epimutations, focusing on the fluctuation of SEMs, considered as the difference of the total number of epimutations between T0 and T2. No significant differences were observed in the change in total SEMs count across groups. However, a trend emerged: the Chronic_Overuse group showed a loss in SEMs over time (Fig. 5).

Fig. 5.

Longitudinal change in SEMs (T2 – T0) across groups. Boxplots show the difference in total SEM counts [log(sems) at T2 minus log(sems) at T0] in healthy controls (HC) and medication-overuse headache (MOH) participants, stratified by clinical outcome at T2. MOH groups are: EM_noOveruse (episodic headache, no medication overuse), Chronic_noOveruse (chronic headache, no medication overuse), and Chronic_Overuse (chronic headache with continued medication overuse). No statistically significant differences were observed between groups, although a trend toward a reduction in SEMs over time was noted in the Chronic_Overuse group

SEMs persistence

We then examined SEMs persistence across timepoints as a measure of epigenetic stability. SEMs were classified as new (present only at follow-up), lost (present only at baseline), and maintained (present at both timepoints) to evaluate whether SEMs tend to persist as stable molecular features or instead show a dynamic and entropic pattern across time and according to disease status. The evidence is that MOH patients in the Chronic_Overuse group were characterized by significantly greater epigenomic instability. Specifically, they exhibited a significantly higher number of new SEMs compared to both HC (p = 0.006) and EM_noOveruse (p = 0.024), and a significantly higher number of lost SEMs compared to HC (p < 0.001), EM_noOveruse (p < 0.001), and Chronic_noOveruse (p < 0.001). While differences among HC, and EM_noOveruse, HC and Chronic_noOveruse did not reach statistical significance, a consistent trend of increasing numbers of new SEMs from HC to EM_noOveruse to Chronic_noOveruse was observed (both p = 0.058).

Finally, maintained SEMs were significantly higher in the Chronic_Overuse group compared to both HC (p = 0.002) and Chronic_noOveruse (p = 0.027). Subjects in the EM_noOveruse group also showed a significant increase of maintained SEMs compared to HC in the simple linear model (p = 0.043). However, after multiple testing correction, this comparison did not reach statistical significance, showing only a trend toward significance (p = 0.085). Figure 6.

Fig. 6.

Persistence of SEMs across groups stratified by T2 clinical outcome. Boxplots show the number of epimutations classified as (a) new (present only at follow-up), (b) lost (present only at baseline), and (c) maintained (present at both timepoints) in healthy controls (HC) and medication-overuse headache (MOH) participants stratified by T2 clinical outcome. MOH groups are: EM_noOveruse (episodic headache, no medication overuse), Chronic_noOveruse (chronic headache, no medication overuse), and Chronic_Overuse (chronic headache with continued medication overuse). Significance levels: * p < 0.05, ** p < 0.01, *** p < 0.001

Epilesions

The analysis of epilesions was performed to identify genes with a particularly high burden of epimutations that might be implicated in disease mechanisms. SEMs enrichment analysis identified a total of 85 genes carrying epilesions—12 hypomethylated and 73 hypermethylated—in at least one of the three diagnostic groups (HC, EM, MOH).

Epilesions at baseline

At timepoint T0, 85 unique epilesions were detected across the cohort. Restricting the analysis to genes with at least one epilesion exclusively in MOH_T0 subjects, 16 genes remained, including NDUFS7, IRAK4, LRRC4C, and KCNQ1.

Epilesions at clinical outcome

We next examined the distribution of epilesions in HC and MOH subjects at T0, stratified according to their clinical outcome at T2. This analysis revealed a total of 42 epilesions. Among these, six genes—LRRC4C, IRAK4, PUS7L, ZNF833P, VTRNA2-1, and KCNQ1—presented epilesions uniquely in individuals classified as Chronic_Overuse at T2.

Discussion

SEMs as a measure of epigenetic instability

Gentilini et al. introduced and promoted stochastic epigenetic mutations (SEMs) as an emerging metric of epigenetic entropy associated with ageing and cumulative stress [10]. The present study applies SEMs for the first time to headache disorders and specifically to MOH patients.

Interesting, EM and MOH have a similar baseline accrual of SEMs, and both present a significant and increased number of SEMs when compared with healthy control. The higher SEMs count in EM and MOH is consistent across the whole genome, in all the autosomal chromosomes, and in both coding and non-coding regions.

It is noteworthy that these results suggest that increased epigenetic variability may represent a feature of the headache condition itself rather than being exclusively attributable to medication overuse and could reflect the cumulative biological stress associated with recurrent pain and associated environmental exposures.

However, these findings uncover correlations and do not imply causality. Moreover, SEMs were assessed in peripheral blood, which does not necessarily mirror epigenetic patterns in the central nervous system. Therefore, SEM burden in blood should be interpreted as a result of the systemic impact of migraine in this specific tissue. Any mechanistic inference regarding central processes remains speculative.

Epigenetic entropy and susceptibility to chronification

The most interesting results of our study came from the longitudinal analysis.

Indeed, there is little knowledge regarding epimutations half-life, if they are stable over time and the rate of the variability. Thus, we profited of our longitudinal sampling to explore these neglected dynamics in relation to the therapy and its outcome. Given the limited sample size, particularly in the outcome-based stratification, these analyses should be considered exploratory. We classified our MOH outcomes in EM_noOveruse, Chronic_noOveruse, and Chronic_Overuse. Based on this classification, we first examined SEMs at T0 to determine whether baseline factors could predispose patients to respond—or not respond—to treatment, we then focused on the dynamic changes of SEMs over time. When we compared the three groups, we found that subjects classified as Chronic_Overuse at T2 had a significantly higher number of epimutations at baseline than the other groups, especially compared to HC (p = 0.004). To further investigate this trait, we categorised epimutations as maintained, lost or newly acquired between the two time points.

By “instability,” we refer here to increased intra-individual variability in the SEM profile over time, operationalized as the turnover of SEM loci. In this context, “gain” and “loss” do not imply directed epigenetic programming but rather changes in the composition of loci classified as SEMs at different time points.

Notably, the subjects in the Chronic_Overuse group exhibit the greatest reshuffling of epimutations, as they concomitantly lose, gain and maintain the highest number of epimutations compared to the other groups. This suggests that the longitudinal analysis of epimutation dynamics is likely much more informative than the absolute number of epimutations. However, such reshuffling should not be interpreted as direct evidence of intrinsic epigenomic instability at the single-cell level.

Indeed, from our data we can get that in the evolution of specific clinical phenotype, such as MOH, is the epigenetic instability that play a role and, the absolute number of epimutations is limited in describing such instability.

To our knowledge, this is the first study to examine SEMs in a longitudinal neuroscience framework, revealing extensive reshuffling of gained and lost epimutations. In peripheral blood, these dynamics may plausibly reflect reshuffling of circulating cell populations, including population turnover and varying contributions from distinct clonal sources over time. The occurrence of different clonal contributions could result in measurable shifts in the overall SEM burden and SEM turnover. Therefore, SEM reshuffling likely represents a composite signal influenced by systemic biological processes rather than a direct marker of neural instability.

Overall, our findings indicate that greater longitudinal SEM variability is associated with persistent chronic headache and continued medication overuse. However, the observational design does not allow conclusions regarding directionality—whether epigenetic variability predisposes to chronification or arises as a consequence of it.

SEMs enrichment

The analysis of epivariations, i.e. the search of epimutations mapping in genes, reveal in particular signals of epigenetic damage in MOH patients mapping in these genes: NDUFS7, KCNQ1, IRAK4, and LRRC4C.

Specifically, NDUFS7 is involved in migraine-related mitochondrial dysfunction, KCNQ1 is linked to cortical spreading depression and trigeminovascular activation, and IRAK4 is associated to neuroinflammatory processes [22–24].

More interestingly, LRRC4C, which encodes a synaptic adhesion molecule involved in glutamatergic and endocannabinoid signalling and in shaping synaptic plasticity, may influence nociceptive modulation and contribute to variability in clinical response to anti-CGRP therapies [25].

These observations are intriguing, as they involve pathways of biological relevance to migraine. However, the identification of epilesions in these genes should be considered hypothesis-generating. The present data do not demonstrate functional consequences, nor do they establish a causal role in migraine chronification or susceptibility to medication-overuse headache. Rather, they point to candidate pathways that merit investigation in larger and mechanistically oriented studies.

While these results highlight the potential role of epigenetic entropy in MOH, our study suffer a major limitation that is the cohort size. The whole patients and controls number is rather small, however the differences in SEMs count and dynamics are evident and consistent in the longitudinal path. Moreover, although comorbidities were recorded, the limited sample size and heterogeneity of conditions did not allow meaningful stratified analyses, and residual confounding cannot be excluded. In addition, although blood samples at T2 were collected after the detoxification phase, it cannot be excluded that medication withdrawal itself may induce transient epigenetic changes, which could have partially influenced the observed SEM dynamics. Accordingly, the present study represents a first small but significant step to integrate epimutations in the context of migraine and medication overuse, to further exploit the role of epigenetics in this clinical setting.

Conclusion

In this longitudinal study, we applied stochastic epigenetic mutations (SEMs) measurement in nucleated blood cells to migraine and medication-overuse headache (MOH), showing that both episodic migraine and MOH patients exhibit a higher SEM burden than healthy controls. This finding suggests that increased epigenetic variability may represent a broader feature of the headache condition rather than being solely attributable to medication overuse.

Greater longitudinal SEM turnover was observed in patients that do not get any benefit by the therapeutic intervention and remain in MOH condition, indicating that the high resilience of their condition is associated with a measurable increased epigenetic entropy.

The identification of recurrent epimutations in biologically relevant pathways is hypothesis-generating and requires validation. Larger and tissue-specific studies will be necessary to clarify the biological and clinical significance of SEM dynamics in headache disorders.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

DNAm

DNA methylation

EM

Episodic migraine

HARDSHIP

Headache-Attributed Restriction, Disability, Social Handicap and Impaired Participation

HC

Healthy Control

ICHD-3

International Classification of Headache Disorders 3rd edition

IQR

Interquartile range

MOH

Medication Overuse Headache

SEMs

Stochastic epigenetic mutations

Author contributions

Conceptualization: S.C., P.G., V.F., and F.F. Methodology: F.F., P.G., C.P., L.C., and D.G. Formal analysis: F.F., K.M.K., F.R., M.G.B., D.D’O., C.S., G.C., and D.G. Investigation: F.F, V.F., K.M.K., F.R., C.P., L.C., and M.G.B. Resources: S.C., P.G., G.P., and P.C. Data curation: F.F., V.F., D.M., K.M.K., L.C., and R.T. Writing – original draft: F.F. and V.F. Writing – review & editing: All authors. Visualization: F.F., K.M.K. and C.P. Supervision: S.C., G.P., P.C, and P.G. Project administration: S.C. Funding acquisition: P.C.

Funding

The publication of this article was supported by the “Ricerca Corrente” funding from the Italian Ministry of Health.

Data availability

All data is available on GEO NCBI repository with accession number GSE318177 [number will be provided after submission of dataset]

Declarations

Ethics approval and consent to participate

The study was conducted in accordance with the principles of good clinical practice and the study protocol was approved by the Local Ethics Committee of the local health service of Bologna, Italy (n. 20,185/CE). All patients gave their written informed consent to study participation. Written informed consent was obtained from all enrolled patients, both for study participation and data publication. All procedures were conducted according to the latest version of the Declaration of Helsinki.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.