Changes in DNA methylation up to 4 months after COVID-19 infection from four population studies

Changes in DNA methylation (DNAm) levels have been widely investigated in epigenome-wide association studies (EWAS) during the acute phase of SARS-CoV-2 infection, particularly in critically ill patients requiring hospitalization. Studies have described DNAm signatures correlating with disease progression or distinguishing between cases and controls [1–7]. These studies have demonstrated significant shifts in white blood cell composition and associated epigenetic and transcriptomic changes that drive COVID-19 disease progression [1, 2, 8]. Some epigenome-wide studies have focused on identifying DNAm changes predictive of COVID-19 outcomes. Castro de Moura et al. identified a 44-CpG epigenetic signature predictive of disease severity in mild and severe COVID-19 patients without comorbidities [3]. Balnis et al. identified a COVID-19-specific signature with 77 differentially methylated regions (DMR) predictive of disease severity in hospitalized patients [4]. Konigsberg and colleagues reported 13,033 CpGs associated with case status and DNAm signatures linked to disease progression and hospitalization [5]. Similarly, Calzari et al. identified DNAm markers predictive of severe infection in comparison to mild disease in individuals with comorbidities [7]. Additionally, several studies have sought to distinguish COVID-19 specific DNAm signatures by comparing them with DNAm profiles from individuals with other respiratory and infectious diseases [4, 6]. Despite differences in study designs – such as sample sizes, patient populations, and analytical methods –, common findings across these studies indicate that widespread hypomethylation in severe cases [1, 2], as well as changes in genes and pathways related to interferon response to viral infection [1–3, 5, 6].

In contrast, considerably fewer studies have investigated epigenetic changes associated with milder forms of COVID-19 [8, 9] and/or with longer follow-up periods post-infection [9–13]. A longitudinal study examining asymptomatic and mildly symptomatic infection in 133 young adults found that differentially methylated sites persisted for several months [8]. The study with the longest follow-up assessed epigenetic changes in 15 patients one year after hospital discharge identifying 71 persistently dysregulated differentially methylated regions (DMRs) out of 1,505 DMRs associated with acute illness at baseline [4, 10]. Two additional EWAS have identified changes three to six months post-infection [11, 13]: one study, analysing blood samples collected 8–12 weeks after SARS-CoV-2 infection in 109 patients and 73 controls, identified three hypomethylated CpGs in cases [11]. Another, examining 96 post-COVID individuals six months after infection alongside 191 pre-pandemic controls, found 42 differentially methylated CpGs [13].

These studies provide insights into the long-term epigenetic changes associated with severe COVID-19 in comparison to mild disease or pre-pandemic controls. However, most lack the statistical power to detect smaller effects linked to mild disease. Consequently, evidence on the epigenetic signature of mild COVID-19 remains scarce, particularly when compared to exposed controls. Furthermore, it remains unclear how these changes correlate with those observed during the acute phase of infection and whether they persist for months post-infection. This study aims to thus identify differentially methylated CpGs in individuals from population-based cohorts in the largest EWAS to date in order to characterize the epigenetic landscape up to four months following mild SARS-CoV-2 infection.

We conducted the largest EWAS to date assessing DNAm changes up to four months after SARS-CoV-2 infection in 675 individuals from four population-based European cohorts within the ORCHESTRA Consortium [14]. DNAm changes were identified in 16 DMP and 21 DMRs, most of which were located in promoter regions or body of protein-coding genes; most DMPs and all DMRs were hypomethylated in cases vs. controls. Three DMPs reached genome-wide significance after Bonferroni-correction (p < 7.48E-08): the strongest association was observed for cg03607951 (IFI44L), followed by cg24678928 (DDX60) and cg19397320 (BTBD3). The top two DMPs were also identified in DMR sensitivity analyses, supporting the robustness and complementarity of the DMR analyses. We also demonstrate that SARS-CoV-2–associated DNAm alterations at both the DMP and DMR levels are linked to transcriptional differences, with downstream analyses indicating these regions are involved in immune response pathways and pro-inflammatory processes.

Cell deconvolution methods were applied to estimate WBC proportions across cohorts; no significant differences in their distribution were observed between cases and controls, except for basophil levels in NAKO. Two prior studies on DNAm changes 3–6 months post-infection also found no evidence of altered immune system composition in individuals with mild infection [11, 13], in contrast to significant changes correlated with COVID-19 severity observed during the acute phase of infection [1, 2]. In line with this and our WBC-adjusted analyses, most genes mapped to the top DMP/DMR associations did not show cell-type-specific expression in the HPA datasets. Adjusting for smoking and prevalent chronic disease attenuated the estimated effects, but the associations remained robust. This was expected, as both smoking and chronic disease are known risk factors for severe COVID-19 [46]. Furthermore, recent evidence suggests that smoking has short- and long-term effects mediated by DNAm on immune regulation that persist after smoking cessation [47]. Despite the high heterogeneity observed in two of the three strongest CpG-case status associations, the direction of effect was consistent across studies, and the associations remained robust in the leave-one-out and serology-based sensitivity analyses.

Our DMP and DMR results are in line with epigenetic changes seen in severe COVID-19 during the acute phase of infection [1–7]. The majority of probes associated with case status in this study exhibited lower DNA methylation levels, consistent with hypomethylation associated with severe COVID-19 during early disease [1, 4, 5] and three to twelve months post-infection [9, 11, 48]. Several interferon-stimulated genes (ISG) identified in the DMP analyses –namely IFI44L, MX1 and DDX60 – have been consistently identified in severe [1–3, 5] and early mild COVID-19 [6]. IFI44L is involved in response to type I interferon (IFN), a key pathway in COVID-19 disease development and severity [49–51]. IFN are signalling cytokines involved in the regulation of the immune response to both viral and bacterial infections [52] and autoimmune disease [53]. MX1 (MX Dynamin Like GTPase 1) is an interferon-induced gene (ISG) encoding for a GTPase that inhibits viral replication; genetic variation in MX1 is associated with differential risk to developing severe COVID-19 [54] and its differential methylation is linked to disease progression [55]. Our DMP and eQTM results are in line with previous reports of hypomethylation in IFI44L both during acute infection and persisting for months thereafter [3, 8, 11]. Hypomethylation of MX1 was identified across COVID-19 severity groups [2, 6], was specific to COVID-19 patients in comparison to patients with other respiratory diseases [4, 5]) and was still present at 12 months after infection [8, 10]. In addition to cg03607951-IFI44L and cg22862003-MX1, cg24678928 (DDX60) and cg25429619 (RABGAP1L) were hypomethylated in COVID-19 patients during the acute-phase of disease [5]; likewise, altered DNAm in cg24678928-DDX60 and cg22862003-MX1 was observed up to 3 months post-infection [11]. DDX60 (DExD/H-Box Helicase 60), encodes for an RNA helicase acting as an antiviral factor involved in the interferon inducible response to viral infection, and it was included in DMP signatures in COVID-19 cases [5, 56]. Interestingly, hypomethylation at cg24678928-DDX60 was correlated with worse quality of life in convalescent COVID-19 patients with Post-Acute COVID-19 Syndrome (PASC) [57].

These DMPs partially overlap DNAm changes in autoimmune and inflammatory disease [8, 11] – e.g. systemic lupus erythematosus [58], primary Sjögren’s Syndrome [59] and rheumatoid arthritis [60] – but also feature in interferon response to COVID-19 [61]. Importantly, individuals with autoimmune disease were excluded from analyses in cohorts where this information was available (see Supplemental Note 1). This points to shared dysregulation of the innate immune system [5, 6], a cross-talk between acute COVID-19 pathogenesis and autoimmune disease [62, 63], and further supports DNAm of ISGs to be involved in long-term dysregulation [8].

DMR analyses identified ten additional protein-coding genes associated with case status. Of these, DMRs in PARP9 and GNA12 show the strongest and most robust association with case status, as seen across sensitivity analyses, replication in at least one cohort, and supportive eQTM data. Although expressed in immune cells, their expression is not cell-type-specific. The location of DMR4-GNA12 (in proximity to TSS) and DMR10-PARP9 (5’ UTR) suggests these changes might regulate gene expression. In line with this, our eQTM results show hypomethylation across CpGs within DMR10 correlates with higher expression of several genes (PARP9, PARP14, DTX3L), and hypomethylation at DMR4 with expression of GNA12. PARP9-DT3XL is an interferon-induced module crucial to regulating PARP14 activity in a cascade of antiviral signaling [64], and GNA12 encodes a G-protein subunit in signaling hub involved in cancer progression, vascular dysfunction and inflammation [65]. Our PARP9 DMR and eQTM results replicate observations in COVID patients versus non-COVID respiratory controls [4]; likewise, DMR4 is in proximity to a DMP identified across COVID severity groups [6] and to two DMRs persistently hypomethylated one year post-infection [48]. Interestingly, this DMR almost perfectly replicates the COVID-19-specific DMR reported in [56]. Hypomethylation of GNA12 and its corresponding increased gene expression have been reported one year post-infection in hospitalized patients [48]. DMR18 (THRB) was also replicated in at least one cohort and across all sensitivity analyses but SA1 (most strict seed value). THRB encodes for a receptor involved in thyroid-hormone signaling in metabolism and endocrine homeostasis, while its expression in immune cells points to endocrine-immune cross-talk [66]. Additional DMRs mapped to genes involved in endocrine-metabolic dysregulation and stress (UCN3, SLC35F3), as well as in inflammatory signaling (FSCN12, GNA12, PPP1R13L) and platelet activation and coagulation (PF4, F3, STX2) (Suppl. Table 22). Similar pathways were identified in a study on differential DNAm in mucosal nasal tissue of COVID-19 patients during the acute infection phase [67]. This study is the first to report persistent differential DNAm in these genes months after a SARS-CoV-2 infection.

We identified biologically plausible changes at the DMP and DMR level in whole blood of mild cases up to four months post-infection, with independent eQTM data supporting the potential functional impact in gene regulation of these subtle DNAm changes. Our findings suggest that subtle but persistent DNAm changes contribute to long-term epigenetic remodeling following a SARS-CoV-2 infection that overlaps pathways involved in metabolic, inflammatory and autoimmune disease. Other changes we identified in intergenic regions and pseudogenes are less directly related to specific molecular pathways, but potentially reflect widespread changes in the epigenome [68]. Despite multiple levels of heterogeneity in studies assessing long-term epigenetic changes (populations assessed, follow-up time and disease severity groups), DNAm changes at the DMP- and DMR-level partially replicate those observed in multiple independent studies. Our findings suggest some epigenetic injuries acquired during infection are persistent months after infection, although the epigenetic signature reported here is associated with mild infection in the general population and is therefore more subtle than the large effects observed during severe infections in hospitalized patients. A recent review on DNAm in persistent COVID-19 suggests infection alters the epigenetic landscape [69], with these changes potentially contributing to the molecular signature of long COVID [70].

This study is the largest multi-cohort cross-sectional EWAS in the months following a SARS-CoV-2 infection in population-based cohorts. Associations were consistent across multiple European cohorts, and the inclusion of non-vaccinated individuals allowed to control for confounding by vaccination-induced changes. Unlike earlier studies detecting large effects in small sample sizes, this study was designed to detect small sized effects (Cohen’s d = 0.2), which enabled the identification of less pronounced DNAm changes months after infection. These findings suggest that subtle but persistent epigenetic changes in blood mirror alterations in well-described immune and inflammatory pathways, help further understand the multi-systemic nature of long-term biological dysregulation following SARS-CoV-2 infection and offer initial insights into potential biomarkers for risk stratification or therapeutic intervention.

This study has several limitations. Time between infection and blood sampling (DNAm) was variable within and across cohorts, which contributed to case heterogeneity and is an important limitation inherent to cohorts’ study designs. Case definition in each cohort relied on differing SARS-CoV-2 detection methods (test-based or self-reported), so it may have been affected by recall and misclassification bias; however, sensitivity analyses using a serology-based case definition in a subset of the participants produced same conclusions. NAKO and Lifelines were identified as likely influential studies, potentially due to larger sample sizes and varying case definitions, yet leave-one-out sensitivity analyses produced similar results. Although inflation at the study-level was not significant, residual confounding is inherent to observational studies like this one. Beyond cohort heterogeneity, underlying inter-individual variation in disease severity/symptomatology and recovery cloud the interpretation of the results in terms of molecular mechanisms and pathophysiology. Additional limitations include limited coverage of the epigenome by using array-based methods, assessing autosomes only and profiling mixed white blood cell populations instead of other potentially more informative tissues [67, 71]. Study designs across cohorts did not allow for longitudinal analyses, assessment of concomitant gene expression or genetic variation, or the examination of epigenetic changes considering long-COVID and related outcomes; these aspects should be considered by future studies. External validation of our findings in independent cohorts will be important to confirm the robustness and generalizability of the DNAm changes identified, as this study only included European cohorts representative from the early pandemic.

This study represents the first large-scale EWAS examining the long-term epigenetic landscape of mild SARS-CoV-2 infection. Our findings indicate moderate but persistent changes in whole blood up to 4 months post-infection. This study adds to the growing evidence on the long-term epigenetic signatures of COVID-19 infection and its overlap with inflammation, metabolic and immune dysregulation. We confirmed associations previously reported for IFI44L, MX1, DDX60, PARP9 and GNA12, and identify additional regions warranting further research. This study identifies persistent epigenetic injuries acquired during acute phase of disease and presents initial evidence suggesting a few of these changes to be correlated with quality of life following an infection. Future longitudinal studies having concomitant DNAm and omics data should investigate the downstream effects of epigenetic changes and their change over time; the study setup of future studies should also consider appropriate collection of data on clinical symptoms long after an infection to contribute to molecular understanding of epigenetic changes associated with post-acute sequelae of SARS-CoV-2 infection.

We would like to extend our appreciation to the members of the ORCHESTRA Advisory Panel: Prof. Marc Lallemant, Prof. Leonard Leibovici, Prof. Jaques Scheres, Prof. Ranjeeta Thomas, Prof. Maria Vehreschild and Dr. Catherine Cohet (EMA) and Dr. Anastasia Pharris (ECDC) as well as to the Ethics Board: Prof. GreetIeven, Prof. Nikolaus Forgó and Prof. Satinder Aneja for their support and guidance throughout.Methylation analysis and initial quality control was performed at the Core Facility Genotyping (Helmholtz Munich) by Nadine Lindemann under the supervision of Dr. Melanie Waldenberger. Further methylation data quality control and preprocessing was performed by Dr. Thomas Delerue, Research Unit Molecular Epidemiology, Institute of Epidemiology, Helmholtz Munich. Special thanks to Dr. Bärbel Fösel and Dr. Sapna Sharma, from the Research Unit Molecular Epidemiology, Institute of Epidemiology, Helmholtz Munich, for providing mentoring and scripts to conduct quality control of GSA data.We thank all participants who took part in the NAKO study and the staff in this research program. The authors also wish to acknowledge the services of the Lifelines Cohort Study, the contributing research centres delivering data to Lifelines, and all the study participants.We would like to give special thanks to all participants of the CON-VINCE study and to each volunteer that helped in sample processing or sample kit preparation. We acknowledge the joint effort of the CON-VINCE team involved in sample and data collection. We also would like to thank the Ministry of Health of Luxembourg as well as the Directorate of Health of Luxembourg for their support. We would like to thank the Research Luxembourg COVID-19 Task Force (Frank Glod, Paul Wilmes, Lars Geffers, Jasmin Schulz, Henry-Michel Cauchie, Ulf Nehrbass, Rudi Balling) for their overall support of the CON-VINCE study. A special thanks also to the data protection officers Sandrine Munoz and Laurent Prévotat. Furthermore, we would like to acknowledge the whole Communication teams involved within CON-VINCE, especially Sabine Schmitz, Arnaud D’Agostini, Didier Gossens, Hélène Jacuszin for their excellent work and support during the implementation and execution of CON-VINCE. We would like to thank Philippe Lamesch for important and successful fundraising for research on COVID-19 in Luxembourg.

Authors’ list with affiliations and contributions were compiled using the online tool tenzing [72, 73]. Conceptualization: P.R.M.-G., M.W., A.P., I.M.H., J.O., R.K., M.P.O.S., R. W., L.F., J.S.-M., N.M., K.G., B.B., K.B., J.F., and P.A.Data curation: P.R.M.-G., I.M.H., J.O., A.R., M.P.O.S., A.K., R. W., J.S.-M., N.M., K.G., B.B., K.B., J.F., and P.A.Formal analysis: P.R.M.-G., L.L., T.D., and I.M.H.Funding acquisition: A.P., K.Ü., R. W., R.K., and L.F.Investigation: P.R.M.-G., O.G., R. W., L.F., B.B., K.B., J.F., and P.A.Methodology: P.R.M.-G., M.W., L.L., T.D., A.P., I.M.H., K.J.S., O.G., K.Ü., R. W., J.O., L.F., J.S.-M., N.M., K.G., B.B., K.B., J.F., and P.A.Project administration: P.R.M.-G., M.W., A.P., K.J.S., O.G., K.Ü., R. W., J.O., R.K., R. Warmerdam, L.F., J.S.-M., N.M., K.G., B.B., K.B., J.F., and P.A.Resources: M.W., A.P., I.M.H., K.J.S., O.G., K.Ü., R. W., J.O., R.K., L.F., J.S.-M., N.M., and K.G. Supervision: P.R.M.-G., M.W., A.P., I.M.H., K.J.S., O.G., K.Ü., R. W., R.K., and L.F.Validation: P.R.M.-G., L.L., and T.D.Visualization: P.R.M.-G., L.L., and T.D.Writing - original draft: P.R.M.-G., M.W., and A.P.Writing - review & editing: P.R.M.-G., M.W., L.L., T.D., A.P., I.M.H., K.J.S., O.G., K.Ü., R. Wagner, J.O., R.K., A.R., M.P.O.S., A.K., R. W., L.F., J.S.-M., N.M., K.G., B.B., K.B., J.F., and P.A.

Open Access funding enabled and organized by Projekt DEAL. The present study was carried out in the context of the ORCHESTRA Project (www.orchestra-cohort.eu↗). ORCHESTRA has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No. 101016167. The views expressed in this abstract are the sole responsibility of the authors and the Commission is not responsible for any use that may be made of the information it contains. This project was conducted with data from the German National Cohort (NAKO, www.nako.de↗). The NAKO is funded by the Federal Ministry of Education and Research (BMBF, project funding reference numbers: 01ER1301A/B/C and 01ER1511D), federal states and the Helmholtz Association with additional financial support by the participating universities and the institutes of the Leibniz Association. The TiKoCo study was supported by the Bavarian Ministry of Science and Arts as well as by the National Research Network of the University Medicine. The Lifelines initiative has been made possible by subsidy from the Dutch Ministry of Health, Welfare and Sport, the Dutch Ministry of Economic Affairs, the University Medical Center Groningen (UMCG), Groningen University and the Provinces in the North of the Netherlands (Drenthe, Friesland, Groningen). FES (Fonds Economische Structuurversterking), SNN (Samenwerkingsverband Noord Nederland) and REP (Ruimtelijk Economisch Programma). Furthermore, this project has received funding from the European Union’s Horizon Europe research and innovation programme under grant agreement No 101057553. The CON-VINCE Study is funded by the National Research Fund Luxembourg (14716281/CONVINCE/Kruger) and the André Losch Foundation (Luxembourg).

NAKO: Data of the NAKO are generally not available to the public due to strict data protection regulations. However, scientists can apply for data use according to the official usage regulation specifications. Please refer to https://transfer.nako.de↗ for further information.Lifelines: Data may be obtained from a third party and are not publicly available. Researchers can apply to use the Lifelines data used in this study. More information about how to request Lifelines data and the conditions of use can be found on their website.CON-VINCE: The dataset for this manuscript is not publicly available as it is linked to the CON-VINCE Study and its internal regulations. Any requests for accessing the dataset can be directed at con-vince@lih.lu. All data of the manuscript will be provided upon reasonable request and approval by the ethics committee.TiKoCo: The dataset for this manuscript is not publicly available as it is linked to the TiKOCo Study and its internal regulations. Any requests for accessing the dataset can be directed to the Study PIs, Prof. Dr. Ralf Wanger (ralf.wagner@ukr.de) and Prof. Dr. med. Klaus Überla (klaus.ueberla@uk-erlangen.de).DNAm datasets: DNAm datasets are stored independently by each cohort and therefore protected by the same data use and access policies mentioned above.

Pamela R. Matías-García, Email: pamela.matiasgarcia@helmholtz-munich.de

Melanie Waldenberger, Email: melanie.waldenberger@helmholtz-munich.de.