DNA hypomethylation leads to cGAS-induced autoinflammation in the epidermis.

DNA methylation is a fundamental epigenetic modification, important across biological processes. The maintenance methyltransferase DNMT1 is essential for lineage differentiation during development, but its functions in tissue homeostasis are incompletely understood. We show that epidermis-specific DNMT1 deletion severely disrupts epidermal structure and homeostasis, initiating a massive innate immune response and infiltration of immune cells. Mechanistically, DNA hypomethylation in keratinocytes triggered transposon derepression, mitotic defects, and formation of micronuclei. DNA release into the cytosol of DNMT1-deficient keratinocytes activated signaling through cGAS and STING, thus triggering inflammation. Our findings show that disruption of a key epigenetic mark directly impacts immune and tissue homeostasis, and potentially impacts our understanding of autoinflammatory diseases and cancer immunotherapy.

Introduction

As the outermost organ of the body, the skin is continuously exposed to pathogens and plays a critical role in the defense against the environment and infectious agents. The most superficial layer of the skin, the epidermis, functions as a critical barrier against viruses, bacteria, and fungi (Alonso & Fuchs, 2003; Fuchs & Horsley, 2008). Various immune cell types reside in the skin or are recruited to maintain skin homeostasis upon exposure to pathogens, including innate immune cells (Kabashima et al, 2019; Nguyen & Soulika, 2019). Pathogen‐derived nucleic acids are an important signal for innate immunity and detected through specific RNA‐ and DNA‐sensing mechanisms. These mammalian systems distinguish self from non‐self through attributes including intracellular localization, secondary structure, post‐transcriptional modifications, and abundance of specific nucleic acids (Roers et al, 2016; Schlee & Hartmann, 2016; Bartok & Hartmann, 2020). A major mechanism for cytosolic DNA detection is the cGAS/STING pathway (Ablasser & Hur, 2020). The cGAS protein is a central receptor for cytosolic double‐stranded DNA that mediates the upregulation of type I interferons and other inflammatory cytokines and chemokines. This occurs via STING activation by the cyclic dinucleotide cGAMP and results in phosphorylation and nuclear translocation of IRF3. The cGAS/STING pathway also plays an important role in the recognition of malignant cells by the immune system (Flood et al, 2019; Hoong et al, 2020; Kwon & Bakhoum, 2020; Yum et al, 2020).

The epidermis is a key context for the operation of the innate immune system and is a self‐renewing tissue that requires precise spatiotemporal control of gene expression (Alonso & Fuchs, 2003; Mikkola, 2007). This is achieved through a well‐orchestrated interplay between transcription factors and epigenetic regulators mediating the regulation of gene expression (Miroshnikova et al, 2019). Heritable CpG methylation is probably the best‐studied epigenetic mechanism in mammalian organisms. DNA methylation is mediated by DNA methyltransferases (DNMTs) and plays an important role during development by repressing gamete‐specific genes or entire chromosomes such as mammalian X chromosome inactivation (Schubeler, 2015; Ambrosi et al, 2017). DNMTs are also involved in the control of allele‐specific gene expression and silencing of repetitive elements, thereby maintaining genome integrity (Ambrosi et al, 2017). Among the five known members of the DNMT family, DNMT1, DNMT3A, and DNMT3B play an important role during mouse development (Smith & Meissner, 2013). DNMT3A and DNMT3B are known as the de novo DNA methyltransferases showing high activity toward unmethylated DNA (reviewed in Espada & Esteller, 2010; Karemaker & Baubec, 2020). DNMT1 has substrate specificity for hemimethylated DNA, but can also transfer the methyl group to unmethylated DNA templates and is therefore considered as the maintenance methyltransferase (reviewed in Mohan and Chaillet (2013) and Karemaker and Baubec (2020)). Of note, DNMTs are relevant targets for small molecule inhibitors in cancer treatment and specific DNMT inhibitors are FDA‐approved for the treatment of acute myeloid leukemia (Jones et al, 2016).

Several severe chronic inflammatory skin diseases, including psoriasis and atopic dermatitis, show abnormal DNA methylation at genes regulating epidermal differentiation and innate immune response in the epidermis of patients as well as in their T cells (Gudjonsson & Krueger, 2012; Han et al, 2012; Rodriguez et al, 2014). In these diseases, the defective DNA methylation is accompanied by disruptions in both keratinocyte differentiation and the barrier function of the skin. Several studies suggest that the appearance of various inflammatory skin diseases and autoimmune disorders, as well as skin cancer, stems from altered epigenetic mechanisms (Javierre et al, 2012).

A function for DNA methylation in epidermal self‐renewal and differentiation was previously described using human DNMT1 knockdown keratinocytes in immunodeficient mice (Sen et al, 2010). This study linked DNMT1 to the maintenance of progenitor function in the epidermis and its deletion resulted in premature differentiation and tissue loss. In another study, epidermal DNMT1 disruption in aging mice resulted in progressive alopecia due to decreased stem cell activation (Li et al, 2012). However, variable and incomplete deletion of Dnmt1 was seen in this conditional mouse model. We therefore used a different approach, epidermal DNMT1 ablation in immune competent mice using the Keratin 5‐Cre (K5‐Cre) system (Winter et al, 2013), to examine the effects of this loss on the skin.

Here, we show that DNMT1 ablation triggers a strong pathological innate immune response in the skin, resulting in immune cell infiltration and subsequent destruction of normal skin architecture. DNA hypomethylation in keratinocytes led to formation of micronuclei and accumulation of cytosolic DNA due to mitotic defects and genomic instability. This in turn caused cGAS‐dependent activation of the innate immune system inducing an autoinflammatory phenotype. Intriguingly, additional deletion of cGAS in Dnmt1‐deficient mice significantly diminished the innate immune response and significantly ameliorated the skin destruction phenotype. Our study reveals a novel role of DNA methylation in restraint of the innate immune response to self‐nucleic acids.

Results

DNA hypomethylation results in disruption of epidermal homeostasis

To study the role of DNA methylation in epidermal morphogenesis and homeostasis, we deleted Dnmt1 in mouse epidermis using Keratin 5‐Cre (K5‐Cre), which is active from embryonic day 9.5 (E9.5) (Ramirez et al, 2004). Mice examined had the following genotypes: K5‐Cre+ Dnmt1 / (Dnmt1 / ), K5‐Cre + Dnmt1 /+ (Dnmt1 /+ ), Dnmt1 / (control), and Dnmt1 /+ (control). Immunolabeling of skin sections showed that Cre activation in Dnmt1 / mice resulted in loss of DNMT1 expression in all layers of the epidermis and in hair follicles, as well as greatly reduced 5‐methylcytosine (5‐mC) in these cells (Fig 1A and D). Strong DNMT1 staining remained evident in non‐epithelial cells in the dermal layer (see also below). Analysis of mRNA and protein levels demonstrated that DNMT1 expression was strongly reduced in the epidermis of Dnmt1 / mice (Fig 1B and C). In accordance with data obtained with DNMT1‐deficient embryonic stem cells (Li et al, 1992), deletion of Dnmt1 in the epidermis resulted in a ˜ 60% reduction in genome‐wide DNA methylation in epidermal cells, as determined by quantitative HPLC (Fig 1E). Residual DNA methylation levels could be explained by DNMT1‐positive immune cells migrating into the epidermis as well as by compensating activities of DNMT3A and DNMT3B (Tsumura et al, 2006).

Figure 1

Epidermal DNA hypomethylation results in a severe skin phenotype

Immunofluorescence labeling of wild‐type mouse dorsal skin samples (Dnmt1 / , skin from P3) shows expression of DNMT1 throughout epidermal layers whereas keratinocyte‐specific knockout (Dnmt1 / ) results in the almost absence of a specific staining in the epidermis and in hair follicles. Dashed lines indicate the dermal–epidermal border, and dotted lines indicate hair follicles.

Deletion of DNMT1 was also confirmed in isolated epidermis by analysis of RNA expression ((B), and data are mean ± SEM, two‐tailed t‐test, ****P ≤ 0.001, n = 3‐4 mice) and Western blotting (C).

Knockout of Dnmt1 results in lower levels of DNA methylation in the epidermis and in hair follicles as determined by anti 5‐mC labeling of P3 skin. Dashed lines indicate the dermal–epidermal border, and dotted lines indicate hair follicles.

HPLC analysis of cytosine methylation in the epidermis of Dnmt1 / mice compared with control littermates at different time points after birth. Values are percentages of methylated CpGs compared with total number of CpGs in the genome. P3: n = 2, P7: n = 2, P8/9: n = 3. Data are shown as mean ± SEM.

Images of control (Dnmt1 / , Dnmt1 /+) and Dnmt1 / littermates at P3, P5, P7, and P9. An enlargement of the skin surface is shown on the right (G).

Hematoxylin and eosin (H&E) staining, immunolabeling for KERATIN 10 (KRT10), LORICRIN (LOR), and KERATIN 6 (KRT6) expression of dorsal skin sections (P5) of control and Dnmt1 / mice.

Data information: Scale bars, 50 µm.

Dnmt1 / mice were born at normal Mendelian ratios and initially displayed normal epidermal development, with regular barrier formation and unaffected differentiation (Appendix Fig S1A–C). However, from postnatal day 3 (P3) onward, Dnmt1 / mice developed a severe skin phenotype characterized by dry, reddish, scaly, and atrophied skin as well as by alopecia (Figs 1F and G, and EV1C and D). At later stages (P7‐9), wounds and lesions appeared on mechanically challenged areas, such as the neck and the extremities (Fig 1F and G). Compared to control littermates, Dnmt1 / mice gained less weight until P6, lost weight thereafter, and died between P8 and P9 (Fig EV1A and B). The full milk belly of those pups indicates that abandonment or starvation was not the cause of death.

Figure EV1

Keratinocyte‐specific deletion of Dnmt1 in mice results in early postnatal lethality associated with severe dermatopathologic alterations

Weight curve of control (Dnmt1 / ) and Dnmt1 / mice from P0‐P10. Control: P0 n = 2; P1 n = 25; P2 n = 44; P3 n = 46; P4 n = 17; P5 n = 23; P6 n = 12; P7 n = 19; P8 n = 14; P9 n = 18. Dnmt1 / mice: P0 n = 3; P1 n = 9; P2 n = 13; P3 n = 14; P4 n = 7; P5 n = 6; P6 n = 7; P7 n = 4; P8 n = 3; P9 all died. Data are mean ± SEM and compared using two‐tailed t‐test. *P ≤ 0.05, **P ≤ 0.01.

Kaplan–Meier plot comparing control (Dnmt1 /  = 13, Dnmt1 /+ n = 9, Dnmt1 /+  = 14) and Dnmt1 / (n = 11) mice using log‐rank test (Mantel–Cox). P < 0.0001.

H&E staining of dorsal skin sections from control and Dnmt1 / mice at P3, P5, and P7. Enlarged sections are indicated on the right. Dermatopathologic alterations found in Dnmt1 / mice at P7 are indicated, such as acanthosis (bracket), hyperkeratosis (hashtag), loss of keratohyalin granules (double arrow), spongiosis (arrowheads), necrotic cells (arrow), and exocytosis (asterisk).

Azan staining of dorsal skin section at P7. Enlarged sections are indicated on the right.

Data information: Scale bars, 50 µm (D), 100 µm (C).

Histological examination revealed that these phenotypes in Dnmt1 / pups were underpinned by a prominent alteration in tissue architecture, characterized by dermal atrophy and massive infiltration of immune cells (Figs 1H and EV1C and D). The dermis showed mild to marked fibrosis, characterized by proliferating fibroblasts, increased numbers of fibrocytes and deposition of collagen, as determined by Azan staining (Fig EV1D). Within the epidermal compartment, we found acanthosis, hyperkeratosis, loss of keratohyalin granules, parakeratosis, and impaired keratinocyte differentiation, as shown by reduced levels of keratin 10 and loricrin (Figs 1H and EV1C).

The basal layer of the epidermis contains highly proliferative cells, which regularly exit the cell cycle and undergo terminal differentiation. Strikingly, cell proliferation was markedly enhanced in the epidermis of Dnmt1 / mice, with significantly more keratinocytes immunopositive for Ki67 and PCNA compared with control samples (Appendix Fig S2A–C). This was accompanied by dramatic upregulation of keratin 6, a marker for hyperproliferative, activated keratinocytes, found in wound healing (Fig 1H). In addition, we observed infiltration of inflammatory cells into the epidermis (exocytosis) with formation of micro‐abscesses and crusts. Immunolabeling for p53 and for cleaved‐caspase 3 revealed enhanced apoptosis in cells of the interfollicular epidermis and in hair follicles of P7 Dnmt1 / mice (Appendix Fig S2D and E). Furthermore, immunoblot analysis of epidermal isolates revealed induced DNA damage, indicated by the presence of gamma‐H2A.X. (Appendix Fig S2F). Together, this analysis indicates that epidermal DNA hypomethylation induces a severe skin pathology.

Elevated innate immune response in the Dnmt1 / epidermis

Pathological assessment of skin histology suggests that disrupted epidermal homeostasis in Dnmt1 / mice is potentially the consequence of inflammatory disease. Thus, paraffin skin sections were immunolabeled for the presence of the common leukocyte antigen CD45. At P7, a massive increase in CD45‐positive cells was observed in both the epidermis and the dermis of Dnmt1 / mice (Fig 2A). Although more leukocytes were found in all skin layers compared with control animals, immune cell infiltration was most obvious within the papillary dermis, with the highest density of immune cells found toward the dermal/epidermal border (Fig EV2A). To determine the dynamics of immune cell infiltration, flow cytometric analysis was performed from skin specimens at different postnatal days. The proportion of CD45‐positive cells in the epidermis and dermis increased over time in Dnmt1 Δ/Δep mice compared with control samples (Figs 2B and EV2B). The increase in immune cells is mainly due to the CD11b+ myeloid population. Further analysis of markers showed that these cells represent neutrophils (CD11b+ Ly6G+) and macrophages (CD11b+ F4/80+), whereas the percentage of antigen‐presenting cells (LCs and DCs) remain largely unchanged (CD11b+ MHC‐II+) (Figs 2C and EV2B). Consistent with an increase in CD11b+ cells, qRT–PCR analysis of selected chemokines revealed that Ifng and interferon‐stimulated genes involved in innate immune responses, such as Isg15, Cxcl1, Cxcl10, IL1b, and IL6, were strongly induced in the epidermis, while Tnfa was moderately but significantly upregulated and Ifna was unchanged (Fig 2D). Furthermore, systemic effects such as reduced blood sugar levels, proteinuria, increased organ mass (liver and kidney), and reduced barrier function accompanied the immune phenotype and the early lethality of Dnmt1 / mice (Fig EV2, EV3, EV4, EV5). These data suggest that Dnmt1 / mice mount a strong innate immune response in the absence of infection.

Figure 2

Epidermal DNA hypomethylation results in induced immune response

Representative immunostaining for CD45 of dorsal skin sections from control and Dnmt1 / mice at P7. Dashed lines indicate the dermal–epidermal border.

Flow cytometric analysis of CD45‐positive cells isolated from the epidermis of control and Dnmt1 Δ/Δep mice of different postnatal ages. Cells are shown as percent of viable cells. Data are mean ± SD, two‐tailed t‐test, **P ≤ 0.01, ***P ≤ 0.001, n ≥ 4 mice.

Flow cytometric analysis of the respective immune cell populations from the epidermis of control and Dnmt1 Δ/Δep mice of different postnatal ages. Cells are shown as percent of viable cells. Data are mean ± SEM, two‐tailed t‐test. ns, not significant, *P ≤ 0.05, **P ≤ 0.01, ****P ≤ 0.0001, n = 3 (P3), 4 (P5), and 4 (P6 + P7).

Relative mRNA expression levels of immune‐related genes in control and Dnmt1 / mice at P3 (n = 3), P5 (n = 5), and P7 (n ≥ 3). Data are mean ± SEM. P‐values are calculated using one‐way ANOVA with post hoc Tukey multiple comparison test, *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001.

Gene ontology analysis of significantly up‐ and downregulated genes (p‐value < 0.05) obtained from epidermal keratinocytes of control and Dnmt1 / mice (n = 3 mice). Keratinocytes were isolated from 3‐day‐old mice and cultured for 72 h.

Relative mRNA expression of representative immune‐related genes found to be deregulated by RNA sequencing in keratinocytes cultured for 72 h and isolated from 3‐day‐old control and Dnmt1 / mice (n = 3 per group). Data are mean ± SEM, two‐tailed t‐test, **P ≤ 0.01; ***P ≤ 0.001.

Data information: Scale bars, 50 µm.

Figure EV2

Epidermal DNA hypomethylation results in induced immune response and in systemic effects

Immunolabeling of dorsal skin sections from control and Dnmt1 / mice for CD45. Dashed lines indicate dermal‐epidermal border or the edge of hair follicles.

Flow cytometric analysis of CD45‐positive cells and the respective immune cell populations in the dermis of control and Dnmt1 Δ/Δep mice of different postnatal ages. Cells are shown as percent of viable cells. Data are mean ± SD, two‐tailed t‐test. ns, not significant, *P ≤ 0.05, **P ≤ 0.01, CD45, P3‐P7, n ≥ 4 mice. All other markers: n = 3 (P3), 4 (P5), and 4 (P6 + P7).

Comparison of Dnmt1 / and control mice regarding blood sugar level (C, n = 4), protein concentration of urine (D, n = 6), and organ weight (E, n ≥ 4).

The inside out barrier of the skin of 3‐ to 7‐day‐old Dnmt1 / and control pubs was tested by a dehydration assay, n = 5 (P3), n = 11 (P5, control), n = 6 (P5, Dnmt1 / ), n = 3 (P7, control), and n = 4 (P7, Dnmt1 / ). Extent of fluids loss was calculated by measuring the decrease of the weight during a certain time period and calculated as percent of body weight.

Data information: For analysis of (C–F), two‐tailed Student's t‐test was used. Data are mean ± SEM. ns, not significant, *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001. Scale bars, 200 µm.

Figure EV3

Keratinocytes of Dnmt1 / mice exhibit DNA hypomethylation and derepression of transposon/repeat families

DNA methylation profile of wild‐type keratinocytes at selected upregulated genes upon knockout of Dnmt1. DNA Methylation data were obtained from Chatterjee et al, 2014 and visualized in the IGV browser. The proximal promotor regions show lower methylation levels than surrounding CpGs.

Expression of deregulated genes found by RNA sequencing of cultured primary keratinocytes that were isolated at P3 of control and Dnmt1 / p mice (n = 5‐7). Statistical analyses were done using two‐tailed Student's t‐test. Data are mean ± SEM. ***P ≤ 0.001, ****P ≤ 0.0001.

Changes in expression of transposon/repeat families were plotted in a scatter plot of log fold change versus log10(adjusted P‐values).

Relative mRNA expression of transposable elements in the epidermis of control and Dnmt1 / mice at P3 (n = 3), P5 (n = 5), and P7 (n = 3). Statistical analyses were done using one‐way ANOVA (Kruskal–Wallis test of multiple comparisons). Data are mean ± SEM. ns, not significant, **P ≤ 0.01.

DNA methylation of CD45‐depleted epidermal keratinocytes isolated from newborn wild‐type (n = 3) and Dnmt1 / p (n = 3) mice. Methylation status of repetitive elements was analyzed using deep amplicon bisulfite sequencing (IAP‐LTR1a) or deep hairpin‐bisulfite sequencing (major Satellites, mSat1). Mean values of DNA methylation of wild‐type and Dnmt1 / p mice are shown in (F), and statistical analysis was done using two‐tailed Student's t‐test (see also Supp. Tab. S4). Data are mean ± SEM. **P ≤ 0.01, ***P ≤ 0.001.

Figure EV4

DNA hypomethylation results in defective mitosis and increase in H3S10ph‐positive G2/M cells in suprabasal epidermal layers

Immunofluorescence double‐labeling of G2/M cells (H3S10ph) and proliferating cells (Ki67) of dorsal skin sections from day 7 old control and Dnmt1 / mice. Dashed lines indicate the dermal–epidermal border and arrows point to suprabasally located H3S10ph‐positive nuclei. (B) Quantification of suprabasal H3S10ph‐positive nuclei shown in (A) from day 7 old control (n = 4) and Dnmt1 / mice (n = 4). Values were compared using two‐tailed Student's t‐test. Data are mean ± SEM. *P ≤ 0.05.

Quantification of basal (left) and suprabasal (right) cells from day 7 old control (n = 4) and Dnmt1 / mice (n = 4) positive for H3S10ph and cells double positive for H3S10ph and Ki67 (as shown in A). Statistical analyses were done using two‐tailed Student's t‐test. Data are mean ± SEM. ns not significant, *P ≤ 0.05.

Effect of DAC treatment on expression of DNMTs and DNA methylation. (D) Cells were treated with 5‐aza‐2′‐deoxycytidine (DAC) or vehicle (PBS) for 72 h, and protein extracts were analyzed for expression of DNMT1, DNMT3A, and DNMT3B. ß‐Actin served as loading control. Signals were quantified relative to the ß‐Actin signal and indicated as percentage of the respective PBS control value. (E) 5‐meC quantification was performed by quantitative dot blot analysis. The PBS control value was set to 100%. The experiment was done in triplicates, and values were compared using two‐tailed Student`s t‐test. Data are mean ± SEM. *P ≤ 0.05. (f) Representative images from aberrant mitotic and post‐mitotic nuclei (H3S10ph‐positive) observed in human keratinocytes treated with DAC. Arrows indicate chromosomal fragments and nucleoplasmic bridges.

Human keratinocytes were treated with PBS (control), 1 µM DAC, or 5 µM DAC for 24 or 48 h and immunostained for H3S10ph and DNA (PicoGreen). Slides were randomized and analyzed in a blinded manner for mitotic and post‐mitotic nuclear aberrations as shown in panel (F). About 100 H3S10ph‐positive mitotic and post‐mitotic nuclei per condition and replicate were analyzed for nuclear defects and calculated as percentage of the total number of H3S10ph‐positive mitotic and post‐mitotic nuclei. Statistical analyses were done using one‐way ANOVA with post hoc Tukey multiple comparison test. Data are mean ± SEM. *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001.

Graphical description of the abnormal nuclei during each cell cycle phase. Data are mean ± SEM from 4 biological replicates per condition.

Data information: Scale bars, 10 µm (F), 100 µm (A).

Figure EV5

Ablation of the cGAS/STING pathway results in decreased expression of immune genes and increased lifespan of Dnmt1 / mice

Cultivated neonatal wild‐type primary mouse keratinocytes were treated with PBS (control) or 2'‐deoxy‐5‐azacytidine (DAC) for in total 72 h. After 48 h additionally, a specific STING inhibitor (C‐178) or DMSO was added, and 24 h later, cells were harvested. Expression of immune‐related genes was compared using one‐way ANOVA with post hoc Tukey multiple comparison test. Data are mean ± SEM from 4 biological replicates per condition. ns, not significant; *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001.

Immunoblot analysis for expression of DNMT1 and cGAS in lysates of isolated mouse epidermis of control, Dnmt1 / , and Dnmt1 / −/− mice. Additionally, mouse embryonic stem cells (mESC) with (DNMT1−/−) or without (DNMT1+/+) genetic deletion of Dnmt1 were analyzed. Beta‐actin was used as loading control.

Kaplan–Meier plot of control (Cgas −/−, n = 7), Dnmt1 / −/− (n = 32), and Dnmt1 / −/− (n = 34) mice using log‐rank test (Mantel–Cox). P < 0.0015.

To investigate whether the effects of DNMT1 ablation in keratinocytes are cell‐autonomous in the context of the innate immune response, RNA sequencing was performed using keratinocytes isolated from P3 Dnmt1 / and control mice. We found 1,339 genes up‐ and 114 genes downregulated in the absence of DNMT1 (abs(log2FC) < 1; P ≤ 0.05) (Dataset EV1). Using gene ontology analysis of the deregulated genes, we found enrichment of several pathways that are highly connected to inflammation, including immune response, defense response, or chemotaxis (Fig 2E). Upregulation of several immune genes such as Ccr2 and Oasl1 was confirmed by qRT–PCR (Fig 2F). Notably, the promoter regions of most of these genes, except for Ccr2, are hypomethylated in control keratinocytes, suggesting that their induction in Dnmt1 / keratinocytes is not a direct effect of local demethylation of regulatory regions in the absence of DNMT1 (Chatterjee et al, 2014) (Fig EV3A). With respect to proliferation, cultivated primary keratinocytes from Dnmt1 Δ/Δep mice behaved differently when compared to keratinocytes in vivo. In vitro, Dnmt1 / keratinocytes exhibited induction of the Cdk inhibitors Cdkn2a (p16) and Cdkn2b (p15) (Fig EV3B), while keratinocytes within the epidermis of Dnmt1 / mice showed hyperproliferation (Appendix Fig S2A–C). This difference emphasizes the importance of examining the role of DNA methylation in the in vivo context of a given tissue. In summary, these findings show that DNA hypomethylation in the epidermis induces a strong immune response in the skin.

The inflammatory phenotype of Dnmt1 / mice does not depend on MDA5/MAVS signaling

Previous studies have demonstrated that DNA‐demethylating agents activate the innate immune system via the MDA5 (melanoma differentiation‐associated protein 5) /MAVS (mitochondrial antiviral‐signaling) RNA recognition pathway in tumor cells (Chiappinelli et al, 2015; Roulois et al, 2015). This RNA‐sensing mechanism is induced by dsRNAs, which are in part derived from derepressed endogenous retroviral elements and its activation by viral mimicry was linked to the toxic effect of DNMT inhibitors on tumor cells (Licht, 2015). Analysis of transposon transcripts extracted from RNA sequencing data from Dnmt1‐deficient keratinocytes revealed that different classes and families of transposons, in particular LTR retrotransposons, are upregulated (Dataset EV2 and Fig EV3C). As expected (Gaudet et al, 2004), strong, time‐dependent activation of IAP (intracisternal A‐type particle) LTR retrotransposons as well as of MLV retroviruses and LV30 transposons was observed by qRT–PCR analysis of epidermal samples from P0‐P7 Dnmt1 / mice (Figs 3A and EV3C and D) and immunoblot analysis of the IAP Gag protein (Fig 3B). A similar increase in IAP transcript levels was also detected in cultivated keratinocytes from Dnmt1 / mice when compared to control samples (Fig EV3B). The increase in IAP transcripts was linked to reduced DNA methylation at IAP elements (Fig EV3E and F and Appendix Table S2). These data demonstrate that loss of DNMT1 in the epidermis results in DNA hypomethylation and derepression of transposons. Therefore, we examined whether activation of the MDA5/MAVS RNA recognition pathway is critical for the immune response and contributes to the skin phenotype observed in Dnmt1 / mice. We generated double knockout mice lacking DNMT1 and functional MAVS, the essential mediator of dsRNA recognition by MDA5. However, the histopathological phenotype including immune cell infiltration, as well as survival of the mice, was found to be similar to DNMT1 single knockout (Appendix Figs S3A–E and S4). To analyze whether additional ablation of MAVS has an influence on transcript levels of immune‐related genes, qRT–PCR was performed from epidermal isolates of P7 mice. Most immune genes analyzed showed a tendency toward reduced levels upon additional deletion of Mavs in the Dnmt1 / epidermis. However, with the exception of the chemokine Ccr2, none of these changes were statistically significant (Fig 3C). Taken together, these data indicate that induction of the innate immune system in the epidermis of Dnmt1 / mice is independent of transposon transcript recognition by the MDA5/MAVS pathway.

Figure 3

Epidermal deletion of DNMT1 results in derepression of transposable elements and activation of immune genes with minor impact of the MAVS pathway

Relative mRNA expression levels of epidermal Intracisternal A‐type particles (IAPs) during different postnatal time points (P0 n ≥ 3, P3 n = 3, P5 n = 5, P7 n ≥ 3 mice), which were compared using one‐way ANOVA with post hoc Tukey multiple comparison test. Data are mean ± SEM. ns not significant, *P ≤ 0.05, **P ≤ 0.01.

IAP protein was determined from epidermal isolates of two control and two Dnmt1 / mice. A bracket indicates the size of group‐specific antigen (gag) proteins.

Relative expression of immune‐related genes of the epidermis from P7 of control (n ≥ 7), Dnmt1 / , (n ≥ 8), MAVS knockout (Mavs −/−, n = 7), and Dnmt1 / −/− (n = 6) mice was compared using one‐way ANOVA with post hoc Holm–Sidak multiple comparison test. Data are mean ± SEM. ns not significant, *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, ****P ≤ 0.0001.

DNA hypomethylation induces the cGAS/STING pathway via cytoplasmic DNA

The Dnmt1 / epidermis displays increased DNA damage and apoptosis. Previous studies demonstrated that DNA hypomethylation in human colon cancer cells induced genomic instability and the formation of micronuclei (Stopper et al, 1993; Costa et al, 2016). To investigate whether epidermal cells of Dnmt1 / mice show similar nuclear aberrations, paraffin sections of the Dnmt1 / epidermis were stained with the ultra‐sensitive DNA dye PicoGreen and microscopically evaluated. Epidermal Dnmt1 deletion induced the formation of DNA blebs and micronuclei (Fig 4A). Importantly, these structures were not detected in the epidermis of control mice. Cytoplasmic double‐stranded DNA originating from viruses or chromosomal instability in senescent or malignant cells is sensed by the cGAS/STING pathway, which in turn induces interferon‐stimulated gene (ISG) expression (Dou et al, 2017; Gluck et al, 2017; Mackenzie et al, 2017; Bakhoum et al, 2018). In addition, the promoter region of the Cgas gene contains two adjacent IFN‐sensitive response elements and different pattern recognition receptor ligands induce cGAS expression in an IFN‐I‐dependent manner, indicating that cGAS itself can be regarded as an ISG (Ma et al, 2015). Therefore, we analyzed the expression of cGAS protein in the epidermis and found substantially increased levels of cGAS in the Dnmt1 Δ/Δep epidermis compared with wild‐type control both by immunolabeling and immunoblot analysis (Figs 4B and EV5B).

Figure 4

DNA hypomethylation results in release of cytoplasmic DNA and cGAS activation

Paraffin skin sections (P5) of control and Dnmt1 / mice were stained for DNA using PicoGreen. Enlarged sections are indicated on the right where arrows point to DNA‐positive nuclear protrusions. Dashed lines indicate the dermal–epidermal border.

Labeling of dorsal skin sections of control and Dnmt1 / mice for cGAS and PicoGreen. Enlarged sections are indicated on the right. Dashed lines indicate the dermal–epidermal border.

Immortalized human keratinocytes (NHEK SV‐Tert3‐5) were treated with 5‐aza‐2′‐deoxycytidine (DAC) or vehicle (PBS). (C, D) PicoGreen staining of PBS‐ and DAC‐treated keratinocytes for 48 h detected micronuclei (arrow), which were analyzed in a blinded manner, and compared using two‐tailed t‐test. Data are mean ± SEM, *P ≤ 0.05 (n = 3 biological replicates) (D). (E) Co‐labeling of DNA by PicoGreen and by cGAS in PBS‐ and DAC‐treated keratinocytes. Within enlarged sections on the right arrows point at DNA/cGAS double‐positive blebs and micronuclei. (F) Relative expression of CCL5 and IFT2 in keratinocytes treated with PBS or DAC for 45 h or 72 h using one‐way ANOVA with post hoc Dunnett multiple comparison test. Data are mean ± SEM. ns not significant, *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001 (n = 4 biological replicates).

Data information: Scale bars, 20 µm (A, D), 50 µm (B, E).

Next, we studied the relationship between DNMT1 inactivation and formation of micronuclei in in vitro cultured immortalized human keratinocytes. The drug 5‐aza‐2′‐deoxycytidine (DAC) leads to DNMT inhibition and thus DNA hypomethylation (Stresemann & Lyko, 2008). Accordingly, DAC treatment of human keratinocytes for 72 h led to a reduction in protein levels of DNMT1 and the de novo methyltransferases DNMT3A and DNMT3B and reduced DNA methylation (Fig EV4D and E). Compared to control cells, DAC‐treated keratinocytes generated significantly more micronuclei as shown by DNA labeling using PicoGreen (Fig 4C and D). Moreover, the majority of detected micronuclei recruited cGAS, as shown by double staining of DNA and anti‐cGAS antibody (Fig 4E). Consistent with induction of an innate immune response, increased expression levels of the immune‐related CCL5 and IFT2 were observed in DAC‐treated keratinocytes (Fig 4F). DNA hypomethylation at pericentric chromosomal regions has been linked to defective mitosis, chromosome segregation errors, and chromosomal instability resulting in the formation of micronuclei, nuclear blebs, and nucleoplasmic bridges (Fenech et al, 2011; Costa et al, 2016). Indeed, visualization of mitotic chromosomes with an antibody specific for histone H3S10ph showed increased numbers of incomplete and disorganized mitosis in DAC‐treated keratinocytes (Fig EV4, EV5). In particular, aberrant mitotic figures during telophase and nucleoplasmic bridges in post‐mitotic cells were significantly elevated upon DAC treatment. Similarly, loss of Dnmt1 led to DNA hypomethylation at pericentric repeats in murine keratinocytes (Fig EV3D and E). Analysis of H3S10ph‐positive G2/M cells in the Dnmt1 / epidermis revealed a significant increase in G2/M cells in the suprabasal layers, which are normally devoid of G2/M cells as seen in control mice (Fig EV4, EV5). Interestingly, the H3S10ph‐positive G2/M cells in the suprabasal layers are negative for the proliferation marker Ki67, indicating that they are arrested in mitosis. Taken together, these findings suggest that loss of DNMT activity in keratinocytes induces chromosomal instability and disrupts cell cycle progression resulting in the formation of DNA blebs and micronuclei. The detection of cytosolic DNA by cGAS induces a strong innate immune response in Dnmt1 Δ/Δep mice.

Deletion of cGAS ameliorates the autoimmune phenotype of Dnmt1 Δ/Δep mice

To examine whether cGAS/STING activation contributes to the autoinflammatory phenotype in response to DNA hypomethylation, we treated primary keratinocytes with DAC and a novel STING inhibitor C‐178 (Haag et al, 2018). DAC treatment of primary keratinocytes alone led to induction of immune genes including Isg15, Oasl1, and Ccl5 (Fig EV5A). Importantly, simultaneous treatment with the STING inhibitor C‐178 resulted in significantly reduced activation of innate response genes. To further test the hypothesis that cGAS represents a critical mediator of pro‐inflammatory signals in the hypomethylated epidermis, we generated Dnmt1 Δ/Δep Cgas −/− double knockout mice (Fig EV5B). The additional knockout of Cgas resulted in a significantly higher survival rate of mice when compared to Dnmt1 Δ/Δep mice (Fig EV5C). The severely impaired epidermal homeostasis observed in the Dnmt1 Δ/Δep mice before the full onset of an immune response suggests that DNMT1‐dependent effects that are not linked to the innate immune response contribute to the lethal phenotype. This is also supported by the study of Sen et al. (Sen et al, 2010) using a xenograft model with SCID mice. These immunocompromised mice show in the absence of DNMT1 severe epidermal defects independent of an activated immune system.

Importantly, histoscoring of H&E‐labeled paraffin skin sections revealed significant improvement of several histopathological parameters (Fig 5A and B, and Appendix Fig S4). Likewise, immune cell infiltration into the dermis and epidermis was strongly reduced upon additional deletion of Cgas, as shown by immunolabeling for CD45 (Fig 5C and D). To test, whether expression of immune‐related genes reflected improvement of skin pathology in Dnmt1 Δ/Δep Cgas −/−, epidermal isolates from P7 mice were analyzed. qRT–PCR revealed significant downregulation of interferon‐stimulated genes (Isg15, Ifit1, Oasl1) and of chemokines and cytokines (Ccr2, Ccl5, Cxcl10) (Fig 5E). In agreement with previously published data (de Koning et al, 2012), loss of epidermal DNMT1 resulted in a robust induction of AIM2 indicating a role of the inflammasome in the activation of the innate immune response. Interestingly, the AIM2 induction was largely dependent on the presence of cGAS supporting the idea of a cross‐talk between the cGAS and the inflammasome as previously described (Swanson et al, 2017). Taken together, these results indicate that activation of the innate immune system via the cGAS/STING pathway is a critical component of the inflammatory skin disease seen in Dnmt1 Δ/Δep mice.

Figure 5

Additional deletion of cGAS attenuates the autoinflammatory Dnmt1 / skin phenotype

Representative images of P7 H&E‐labeled dorsal skin sections of control (Dnmt1 / , Dnmt1 / , and Dnmt1 / −/− mice.

The histopathology of at least 9 animals per genotype was assessed in a blinded manner, and an overall score was determined from individual parameters (see Appendix Fig 7 and Appendix Table S3). Comparison was performed using one‐way ANOVA with post hoc Holm–Sidak multiple comparison test. Data are mean ± SEM. ****P ≤ 0.0001.

Representative images of CD45 immunofluorescence labeling of P7 dorsal skin sections from Dnmt1 / , Dnmt1 / , and Dnmt1 / −/− mice. Dashed lines indicate the dermal‐epidermal border.

Quantification of CD45‐positive cells. Tissue sections approximately 1 cm in length were scanned, and the area of CD45‐positive cells per area of nuclei was quantified using ImageJ. Sections obtained from five mice per genotype were analyzed, and calculated values were compared using one‐way ANOVA with post hoc Holm–Sidak multiple comparison test. Data are mean ± SEM. **P ≤ 0.01, ***P ≤ 0.001.

Epidermis obtained from P7 control, Dnmt1 / , Cgas −/−, and Dnmt1 / −/− mice was analyzed for relative mRNA expression levels of immune‐related genes using one‐way ANOVA with post hoc Holm–Sidak multiple comparison test. Data are mean ± SEM. ns not significant, *P ≤ 0.05, ****P ≤ 0.0001. At least 7 mice per genotype were used.

Data information: Scale bars, 50 µm (A), 100 µm (C).

Discussion

Our study reveals a crucial role for DNA methylation in epidermal homeostasis through both maintenance of genomic stability and prevention of inappropriate innate immune system activation. Mice with a Keratin 5‐Cre‐mediated deletion of Dnmt1 show a normal epidermal development until birth. From P3 onwards, mice develop a severe skin phenotype characterized by disturbed epidermal homeostasis, and immune cell infiltration reflecting an excessive immune response. Sen et al. (Sen et al, 2010) demonstrated that loss of DNMT1 in human keratinocytes resulted in impaired self‐renewal and premature differentiation of epidermal progenitors. Indeed, in vitro cultivated murine Dnmt1‐deficient keratinocytes express increased levels of the senescence marker p15 and p16, reflecting in part the behavior of human DNMT1 knockdown keratinocytes (Sen et al, 2010). In contrast, we observed hyperproliferation of keratinocytes in the epidermis of Dnmt1 / mice, which might be caused by the presence of immune cells and resident dermal cells, leading to mutual stimulation in vivo (Zheng et al, 2007; Briso et al, 2013 ). Another study using Keratin 14‐Cre to delete Dnmt1 in the epidermis reported reduced epidermal stem cell proliferation and hair germ formation in the absence of DNMT1 (Li et al, 2012). These mice are viable, and their epidermis shows uneven thickness due to hyperproliferative regions. The changes in epidermal proliferation were interpreted as arising from a compensatory mechanism (Li et al, 2012). Divergent results observed in these studies can be explained by differences in the origin of keratinocytes, the mouse models used, and the absence of immune cells in the xenograft model. The usage of an immunocompetent, conditional K5‐cre mouse model allowed us to conduct a robust analysis of Dnmt1 deletion in the skin.

Several reports have implicated DNMT1 in the regulation of mammalian cell fate decisions in different tissues and organs. Dnmt1 deletion during embryogenesis caused an acute phenotype of the intestine, characterized by weight loss, global DNA hypomethylation, genome instability, apoptosis, and loss of nascent villi (Elliott et al, 2015) (mitotic genes were top hits among downregulated genes). In the mature intestinal epithelium, DNMT1 controls cellular differentiation, but is dispensable for organ maintenance and organismal survival in adult mice (Sheaffer et al, 2014). In the nervous system, DNMT1 has been shown to be essential for survival of fetal mitotic neuroblasts. Mice with Nestin‐Cre‐mediated deletion of Dnmt1 die after birth due to respiratory defects (Fan et al, 2001). In the pancreas, ablation of DNMT1 results in a decrease in differentiated pancreatic cells with a concomitant increase in p53 levels, cell cycle arrest, and progenitor cell apoptosis (Georgia et al, 2013). Taken together, all these studies indicate a regulatory role of DNMT1 in the control of progenitor self‐renewal and differentiation. The present study links tissue‐specific loss of DNMT1 to innate immune system activation and systemic defects, providing a holistic view of the impact of deregulated DNA methylation at the organismal level.

We here report for the first time that autoinflammation of the skin is a direct consequence of DNMT1 ablation in vivo. DNA hypomethylation in the epidermis results in the activation of the innate immune system, likely linked to the skin’s unique function as an immune organ consisting of highly proliferating basal keratinocytes and skin‐resident and infiltrating immune cells (Kabashima et al, 2019; Nguyen & Soulika, 2019). We find that Dnmt1 / mice show an elevated innate immune response in the epidermis, immune cell infiltration in the skin, and systemic effects in response to inflammation. In vitro experiments with keratinocytes demonstrated that the elevated expression of immune genes in response to DNA hypomethylation is at least in part a cell‐intrinsic effect. Our data demonstrate that DNMT1 loss in the epidermis induces aberrant changes in cellular nucleic acid abundance and localization, triggering a strong innate immune response.

In accordance with previous reports on other cell types (Fan et al, 2001; Gaudet et al, 2004; Georgia et al, 2013), the Dnmt1 / epidermis displays a dramatic increase in the expression of IAPs and other transposons. An RNA‐dependent immune response to DNA hypomethylation was previously reported for human tumor cells upon treatment with DNMT inhibitor (Chiappinelli et al, 2015; Roulois et al, 2015). In this setting, the MDA5/MAVS pathway is activated in response to retrotransposon‐derived double‐stranded RNA. In a follow‐up study, it was shown that upon inhibition of DNA methylation, intronic and intergenic SINEs activate MDA5 in cancer cells in an ADAR1‐dependent manner (Mehdipour et al, 2020). Accordingly, loss of the RNA‐editing enzyme ADAR1 causes an enhanced interferon response and embryonic lethality, which is rescued by additional Mavs knockout (Mannion et al, 2014). In the context of epidermal DNA hypomethylation, we did not detect a significant contribution of MDA5/MAVS signaling to the observed phenotypes. It remains possible that transposon activation contributes to the immune phenotype of Dnmt1 / mice. LINE1 retrotransposon activation in response to depletion of the histone deacetylase SIRT6 leads to accumulation of cytoplasmic L1 cDNA, which triggers a strong type I interferon response via cGAS activation (Simon et al, 2019). Future work will test whether other nucleic acid‐sensing mechanisms are activated by (reverse) transcription of mobile elements.

A hallmark of Dnmt1‐deficient or 5‐aza‐2′‐deoxycytidine‐treated keratinocytes in our study is incomplete or defective mitosis resulting in the presence of cytosolic DNA, detected as blebs, micronuclei and nucleoplasmic bridges. DNA hypomethylation has been previously shown to induce genomic instability, DNA damage, and mitotic arrest in human cancer cells (Chen et al, 2007; Costa et al, 2016), and in combination with Poly(ADP ribose) polymerase inhibition, it causes inflammation and reduced DNA repair (McLaughlin et al, 2020). In macrophages, DNMT3A induces HDAC9 to deacetylate TBK1 for activation of antiviral innate immunity (Li et al, 2016). Mutations in the human DNMT3B gene result in ICF (immunodeficiency, centromeric region instability, and facial anomalies syndrome) (Xu et al, 1999). ICF patient cells show genomic instability highlighted by the presence of micronuclei and nuclear blebs. Our findings further highlight a crucial function for DNA methylation in the maintenance of chromosomal stability. Cytosolic DNA in the Dnmt1 / and 5‐aza‐C‐treated keratinocytes was recognized by cGAS, resulting in upregulation of the DNA‐sensor protein by feedback. Ablation of cGAS in Dnmt1 / mice significantly attenuated the activation of immune genes, infiltration of immune cells, and destruction of the skin architecture suggesting a major contribution of the cGAS/STING pathway to the autoinflammatory phenotype of Dnmt1 / mice. Mice deficient in the DNA exonuclease Trex1 develop systemic inflammation (Morita et al, 2004) ameliorated by simultaneous deficiency in STING or cGAS (Gray et al, 2015). Similarly, deletion of Cgas or Sting in mice lacking DNase II, a lysosomal enzyme that digests DNA, rescued the lethal autoimmune phenotypes of the DNase II knockout mice (Ahn et al, 2012; Gao et al, 2015). Taken together, DNA‐modifying and processing enzymes prevent the accumulation of cytosolic DNA and thereby the unwanted activation of the innate immune system by self DNA.

Additional loss of cGAS does not fully abolish the autoinflammatory phenotype of Dnmt1 / mice. It is conceivable that other pathways of the innate immune system such as TLR3 (dsRNA), TLR2/TLR4 (protein), the inflammasome (DNA, RNA, protein), TLR9 (hypomethylated ssDNA, RNA/DNA hybrids), or ZBP1 (Z‐DNA, Z‐RNA) are also induced in the epidermis of these mice (Kawai & Akira, 2011; Rigby et al, 2014; Bartok & Hartmann, 2020; Devos et al, 2020; Jiao et al, 2020). Gene mutations resulting in defective DNA/RNA modification and turnover pathways are associated with autoinflammatory diseases (Uggenti et al, 2019; Bartok & Hartmann, 2020; Melki & Fremond, 2020). In addition to genetic factors, also epigenetic dysregulation contributes to the development of pathogenesis (Alvarez‐Errico et al, 2017). Therefore, it will be interesting to investigate whether changes in epigenetic patterns and nucleic acid‐sensing mechanisms in the skin also contribute to autoinflammatory disorders.

The data presented here have potential implications for cancer treatment. cGAS/STING agonists have been considered as promising therapeutics in combination with immune checkpoint inhibitors in particular (Flood et al, 2019; Hoong et al, 2020). Current efforts to advance immunotherapy focus on testing drug combinations involving epigenetic drugs, which stimulate both innate and adaptive immune responses (Jones et al, 2019; Hogg et al, 2020) (Daver et al, 2018; Lai et al, 2021). Our study indicates that in addition to activation of RNA‐sensing pathways via derepression of mobile elements, DNMT inhibitors could be used to trigger the controlled activation of the immune system through DNA‐sensing pathways. Both processes might contribute to an increased efficacy of a combination of epigenetic and immune checkpoint therapies in tumor treatment.

Materials and Methods

Animal care, transgenic mouse lines, and mouse experiments

All mouse lines were bred to a mixed genetic background of C57BL/ 6J 129SV. Keratin 5‐Cre (K5‐Cre) mice (Ramirez et al, 2004) were mated to Dnmt1 / (Jackson‐Grusby et al, 2001; Lee et al, 2001), Mavs −/− (Sun et al, 2006), and Cgas −/− (Schoggins et al, 2014) mice to generate mice with deletions of Dnmt1 in the epidermis or epidermal Dnmt1, Mavs, and Cgas, respectively. Mice were kept in the animal facilities of the Medical University of Vienna in accordance with institutional policies and federal guidelines. All animal experiments were performed in accordance with a protocol authorized by the Austrian Ministry of Science and Research, following the approval by the National Ethical Committee for Animal Experimentation. The Austrian Ministry for Science and Research approved the corresponding proposals GZ BMWFW‐66.009/0138‐WF/V/3b/2017 and GZ BMWFW‐66.009/0276‐WF/V/3b/2017.

Genotyping primer

K5‐Cre forward and reverse (TAA TCG CCA TCT TCC AGC AG and CAA TTT ACT GAC CGT ACA C), Dnmt1 flox forward and reverse (AAC CGT TGG CTT TTT GAG TGA G and AGA AAT AAA AAG CCA GTT GTG T), Cgas forward, reverse 1, reverse 2 (ATA TTT CCC CCT GTG TTG GA, GTG CCA GGT GAC ACA ACA TC, CGG ATG GAT GAA CAA ACA GA), Mavs 1, 2, 3 (AGC CAA GAT TCT AGA AGC TGA GAA, TAG CTG TGA GGC AGG ACA GGT AAG G and GTG GAA TGT GTG CGA GGC CAG AGG C).

For the dye exclusion assay, embryos (E19.5) were killed and incubated for 1 min in 25, 50, and 75% methanol in PBS, followed by a 1‐min incubation in 100% methanol, and a descending series of incubations in 75, 50, and 25% methanol in PBS for 1 min. Pups were washed in PBS and stained with 0.1% toluidine blue O (Sigma) for 1 min. Destaining was done in PBS. Blue‐stained navel served as positive control for successful staining.

Dehydration assay was performed to measure the extent of fluid loss during a certain time frame and reflects the dysfunction of the skin barrier. Pups were separated from their parents on the indicated postnatal day (P3‐P9) and kept at 37°C in the incubator for 6 h. Their weight was measured every 30 min, and weight loss was calculated as a function of time in relation to the starting weight of the pubs (Leyvraz et al, 2005).

Blood sugar was monitored with a glucose meter. Spontaneous urine was collected and analyzed with the pyrogallol red‐molybdate method, according to Fujita et al. performed by the company In Vitro Laboratory for veterinary Diagnostics and Hygiene GmbH Vienna, Austria.

Epidermis isolation and keratinocyte culture

Murine skin was isolated and incubated with the dermis side in PBS containing 5 mg/ml Dispase (Roche) for 2 h, at 4°C for protein and RNA isolation and at 37°C for keratinocyte isolation. Epidermis was removed from dermis, washed in PBS, and snap‐frozen for protein and RNA analysis. In case of RNA isolation, solutions were supplemented with RNAse Inhibitor (Winter et al, 2013). Mouse epidermal keratinocytes were isolated as previously described (Sibilia et al, 2000) and cultured in keratinocyte growth medium (KGM Gold BulletKit, Lonza) containing 8% chelated fetal calf serum and 0.05 mM CaCl2 on collagen‐fibronectin‐coated dishes. Immortalized human epidermal keratinocytes NHEK/SVTERT3‐5 (Wagner et al, 2018) (EVERCYTE, Vienna, Austria, CLHT‐011‐0026‐5) were grown in KGM2 (C‐20011, PromoCell) adjusted to 0.06 mM CaCl2 and propagated according to the manufacturer’s instruction.

Treatment and immunofluorescence (IF) analysis of keratinocytes

Cells were seeded either onto glass coverslips (for IF) placed in 6‐well plates or into 12‐well plates (for RNA expression analyses) one day before treatment. On the next day, cells were treated with either the vehicle control PBS or the DNMT1 inhibitor 5‐Aza‐2′‐deoxycytidine (DAC; Sigma‐Aldrich). To inhibit the cGas/Sting pathway, the cells were treated with the inhibitor C‐178 (generous gift from Andrea Ablasser (Haag et al, 2018)) 24 h after the first DAC treatment. Keratinocytes were harvested either 24 h (IF) or 48 h (IF, RNA) after DAC treatment. For IF analysis, cells were washed in ice‐cold PBS, fixed in 4% paraformaldehyde for 20 min at 4°C, washed with PBS, and stored at 4°C until further analysis. For IF, cells were permeabilized using 0.1% Triton X in PBS for 20 min at room temperature than blocked in 2% BSA in PBS with 10% goat serum for 30 min at room temperature. Antibodies specific for human cGAS (#D1D3G, Cell signaling Technology, 1:200) and ph‐Histone Histone H3, Ser 10 (SC‐8658‐R, Santa Cruz Biotechnology, 1:2,000) were overnight incubated at 4°C. After washing three times with PBS containing 0.1% Tween‐20, the coverslip sections were incubated with Alexa Fluor 546‐conjugated secondary antibody (Fisher Scientific) for 1 h at room temperature. After washing three times with PBS containing 0.1% Tween‐20, the samples were counterstained with 1 µg/ml 4′,6‐diamidino‐2‐phenylindol (DAPI, Sigma‐Aldrich) and PicoGreen dsDNA Reagent (Fisher Scientific, 1:2,000) for 20 min in TE buffer (10 mM Tris–HCl, 1 mM EDTA, pH 8.0). Coverslips were subsequently washed 3 times in PBS and mounted on SuperFrost Plus Adhesion slides (Thermo Scientific) using ProLong Gold (Thermo Fisher Scientific).

Determination of chromosomal aberrations during mitosis

Human keratinocytes treated with DAC for 24 or 48 h were immunolabeled for histone H3S10ph and counterstained with PicoGreen. About 100 mitotic H3S10ph‐positive nuclei were analyzed in a blinded and randomized manner, classified into mitotic phases, and assessed for chromosomal aberrations.

Hematoxylin and eosin (H&E) labeling, immunohistochemistry (IHC), immunofluorescence (IF), and TUNEL analyses

Tissue samples were fixed overnight in 4% PFA and further embedded in paraffin. Stainings were performed on 4‐µm sections using SuperFrost Plus Adhesion slides (Thermo Scientific). H&E stainings were carried out according to the standard procedure with a COT 20 Linear Slide Stainer (Medite). IF and IHC analyses were carried out according to protocols published previously (Fischer et al, 2005) with modifications. Briefly, paraffin sections were prepared for staining by heat antigen retrieval using a pressure cooker (2100 Antigen Retriever, Aptum Biologics) for 20 min at 121°C. Depending on the used antibody, the slides were cooked in 10 mM Citrate buffer (pH 6) or in Dako Target Retrieval Solution (pH 9; Agilent). Samples were cooled down and incubated with phosphate‐buffered saline, pH 7.2, plus 2% BSA and plus 10% goat serum (DAKO) for 20 min to block nonspecific binding. The slides were incubated overnight at 4°C using a humid chamber. For IHC, samples were incubated with the respective secondary antibody, stainings were visualized with the VECTASTAIN ABC kit and DAB substrate (Vector labs) and sections were counterstained with hematoxylin. For IF, sections were incubated with the appropriate Alexa Fluor 488/546‐conjugated secondary and counterstained with 1 µg/ml 4′,6‐diamidin‐2‐phenylindol (DAPI, Sigma‐Aldrich) and PicoGreen dsDNA Reagent (Fisher Scientific, 1:2,000) for 20 min in TE buffer (10 mM Tris–HCl, 1 mM EDTA, pH 8.0). The slides were mounted with Prolong Gold (Thermo Fisher Scientific). For in situ labeling of 3′‐OH ends of DNA, the TUNEL assay (Roche) was performed according to the manufacturer’s instructions (Fischer et al, 2005).

Primary antibodies used for IHC

Ki67 (1:500, Novocastra), Keratin 10, Keratin 6, Loricrin (1:1,000, Covance), p53 (1:50, NCL‐p53‐CM5p, Novocastra), cleaved‐caspase 3 (1:200, CC3, cs‐96615, Cell Signaling), PCNA (1:200, DAKO), DNMT1 (1:200, Abcam, ab188453), CGAS (1:100 Cell Signaling, 31659; 1:200 Cell Signaling 15102) 5 mC (1:500, Diagenode 006‐500), CD45 (1:1,000, Abcam ab10558), and histone H3S10ph (1:2,000, Santa Cruz Biotechnology SC‐8656‐R).

Microscopy, image processing, and image analysis

H&E, IHC, and immunolabeling were analyzed using a slide scanner (VS120 S6, Olympus) equipped with a XM10 monochrome camera (Olympus) and a Pike F‐505C color camera (Allied Vision). Image scans were obtained and further processed using the following Programs: VS‐ASW 2.9, OlyVia (Olympus) and Adobe Photoshop CS6 (Adobe). For some IF analyzes, images were taken using a LSM Meta 510 confocal microscope (Zeiss). It was ensured that identical exposure times and gray value settings were used for the samples to be compared. For automated image quantification, a script was developed using Fiji software (Schindelin et al, 2012). In brief, a tissue mask was interactively created corresponding to the tissue region of interest. Particle analysis was done after applying a fix threshold to allow a comparative analysis. For enhanced particle separation, a watershed algorithm was used. Particle analysis revealed the parameters particle count, area, and area fraction of the tissue region of interest. For the quantification of Ki67, positive cells in the interfollicular epidermis (IFE), positive cells of the basal, and suprabasal layers of the IFE were counted and set in relation to the total cell number.

Histopathology

Paraffin sections of back skin which were stained by H&E and by azan were scanned using a slide scanner (VS120 S6, Olympus) equipped with 40x objective and a Pike F‐505C color camera (Allied Vision). Scanned images were analyzed by a certified pathologist in a blinded manner using the OlyVia software (Olympus). Samples were scored for nine pathological alterations and graded from 0 to 5. For details of analyzed alterations and grading system, see Appendix Fig S4 and Appendix Table S1. All scores were summed up, and total H&E score is given in the figure.

Quantification of 5‐methylcytosine (5‐meC)

For 5‐meC quantification in the mouse epidermis, genomic DNA was extracted using the phenol/chloroform method. Briefly, mouse epidermis (as isolated described above) was incubated overnight in 4.5 ml proteinase K buffer with 300 μl proteinase K (20 mg/ml, 50 mM Tris–HCl pH8.0, 100 mM NaCl) and 250 μl SDS at 55°C. Then, 200 μl of 5 M NaCl and 5 ml phenol/chloroform were added. This was vortexed for 30 sec and then centrifuged for 5 min at 1,150 g at 4°C. The aqueous phase was subsequently transferred to a new tube, and 5 ml of chloroform was added. This was vortexed and centrifuged again, and the aqueous phase was put into a fresh tube. Three volumes of 96% ethanol were added, and DNA was then fished out with a glass hook, was centrifuged again, and washed in 70% ethanol. After drying the pellet, it was resuspended and dissolved in 400 μl TE buffer (1 M Tris–HCl pH 8.0, 0.5 M EDTA) for 5 min at 65°C. HPLC analysis of isolated DNA was performed as previously described (Ramsahoye, 2002). Genomic methylation was assessed by measuring methyl group carrying and unmodified cytidine nucleotides. HPLC shows the percentage of deoxy‐methyl‐cytidine‐mono‐phosphate (dmCMP) from total cytidines (dmCMP and dCMP). For 5‐meC quantification in human keratinocytes, DNA methylation was analyzed by quantitative dot blot analysis as previously described (Zupkovitz et al, 2021).

DNA methylation analysis of selected repetitive elements

Epidermis of P7 mice was prepared by treatment with Dispase II solution (5 mg/ml in 1x PBS; Roche) for 45 min, 37°C. Epidermal cells were isolated by shaking in DNase I solution containing 250 µg/ml DNase I (DN25, Sigma‐Aldrich), 8% chelated fetal bovine serum (F2524, Sigma‐Aldrich) in MEM (Sigma‐Aldrich) for 30 min at 37°C. After filtration using a 70‐µm cell strainer, CD45‐positive cells were depleted using Mojosort Mouse CD45 Nanobeads (480027, BioLegend) according to manufacturer’s instructions. Purity of epidermal keratinocytes was checked by flow cytometry using anti‐CD45. Genomic DNA of keratinocytes was prepared using Genelute Mammalian Genomic DNA Kit (G1N70 Sigma‐Aldrich). DNA methylation status of selected repetitive elements was analyzed using deep amplicon bisulfite sequencing (IAP‐LTR1a) or deep hairpin‐bisulfite sequencing (major Satellites) according to Arand et al, 2012 (Arand et al, 2012), with slight modifications. Briefly, after bisulfite conversion, DNA was bound to MagBinding Beads (Zymo Research) in 7 M guanidine‐HCl buffer. After washing with 80% EtOH in 10 mM Tris–HCl pH 8.0, DNA was desulfonated using 0.1 M NaOH in 70% isopropanol followed by two washes with 80% EtOH in 10 mM Tris–HCl pH 8.0 and elution in water. Amplicons were amplified using fusion primers of the repeat‐specific primer and part of the TruSeq adaptor sequence. A second amplification step with six cycles was performed using full TruSeq adaptor sequences involving custom dual‐index barcodes. Analysis for IAP‐LTR1a (specific primer sequences: GAT AGT TGT GTT TTA AGT GGT AAA TAA A and TCC CTA TTA AAA AAA ATT ATC CTT C) omitted restriction and ligation reaction before bisulfite treatment. After equimolar pooling of purified amplicons, the library was sequenced on a MiSeq nano 2 × 150 bp run. Reads were demultiplexed and trimmed for adaptor sequences using cutadapt v2.10 (Martin, 2011). Paired reads were merged using flash (Magoc & Salzberg, 2011), and different amplicons were separated using cutadapt v2.10 by primer sequence. CpG methylation was determined using BiQAnalyzerHT (Lutsik et al, 2011) according to Arand et al (2012). More than 1,300 reads were obtained for every sample, with an average conversion rate of 99.5% as determined from the Hairpin adaptor sequence (see Appendix Table S2).

DNA methylation profile at selected genes upregulated in Dnmt1 / mice

DNA methylation status of the genomic loci of immune genes derived from wild‐type keratinocytes was obtained from Chatterjee et al, 2014 (Chatterjee et al, 2014) and visualized in IGV browser (http://software.broadinstitute.org/software/igv/home), Broad Institute and UC San Diego, CA.

RNA isolation and real‐time PCR analysis

Total RNA was isolated following the manufacturer’s instructions (TRIzol, Invitrogen). RNA was reversely transcribed with the iScript cDNA Synthesis Kit (Bio‐Rad). Real‐time PCR analysis was performed with the KAPA SYBR FAST qPCR MasterMix (Peqlab) on an iCycler IQ system (Bio‐Rad). Relative expression levels were normalized to HPRT (mouse) or B2 M (human).

qPCR primer for mRNA expression of mouse genes

Hprt forward (5′) and reverse (3`) (GCTGGTGAAAAGGACCTCT and CACAGGACTAGAACACCTGC Cxcl1 forward and reverse (CTG CAC CCA AAC CGA AGT C and AGC TTC AGG GTC AAG GCA AG); Cxcl10 forward and reverse (TTC TGC CTT CAT CCT GCT G and AGA CAT CTC TGC TCA TCA TTC); Cldn1 forward and reverse (CTT CTC TGG GAT GGA TCG and GGG TTG CCT GCA AAG TAC TGT); Cldn3 forward and reverse (GCG GCT CTG CTC ACC TTA GT and GAC GTA GTC CTT GCG GTC GTA); Cldn8 forward and reverse (TCA GAA TGC AGT GCA AGG TC and AGC CGG TGA TGA AGA AGA TG), Cdkn2a forward and reverse (GAA CTC TTT CGG TCG TAC CCC and CAG TTC GAA TCT GCA CCG TAG); Cdkn2b forward and reverse (CCC TGC CAC CCT TAC CAG A and CCC TGC CAC CCT TAC CAG A); Isg15 forward and reverse (AAG CAG ACT CCT TAA TTC C and CTG TAC CAC TAG CAT CAC); Tnf forward and reverse (CAG GAG GGA GAA CAG AAA C and ACA AGC AGG AAT GAG AAG AG); Oasl1 forward and reverse (ACC AGC AGT ATG TGA GAG and GGT CCA GTA GAT ACA GAG G); Ccr2 forward and reverse (ACG ATG ATG GTG AGC CTT GTC and CAG GAA GAG CAG GTC AGA GAT G); IAP forwards and reverse (ACT AAC TCC TGC TGA CTG G and TGT GGC TTG CTC ATA GAT TAG); Dnmt1ex3‐4 forward and reverse (GAG GAA GGC TAC CTG GCT AAA G and TCA CTG TCC GAC TTG CTT CTC); MLV forward and reverse (CCA GAC TTG ATC CTG CTA CA and CTC AGT CAG CCA TCT CTG AC); LV30 forward and reverse (CTC GCA CCT TTT CGC GCT CG and GCG AGG GGA TCA TCA TAA CA); Ccl5 forward and reverse (ATA TGG CTC GGA CAC CAC TC and CCC ACT TCT TCT CTG GGT TG); Mavs forward and reverse (GGA CAC ACT CTG GGG ACT CT and GGT CAG GGA TGT TGT GAC CT); Ifit1 forward and reverse (GCT CTG CTG AAA ACC CAG AG and CCC AAT GGG TTC TTG ATG TC); Ccl5 forward and reverse (CCC TCA CCA TCA TCC TCA CT and GAG CAC TTG CTG CTG GTG TA); Aim2 forward and reverse (CCT GAT TCA AAG TGC AGG TG and GAG GCA GCA GAG CAG TTT TC); Il1b forward and reverse (ACT CAT TGT GGC TGT GGA GA and TTG TTC ATC TCG GAG CCT GT); Il6 forward and reverse (GAG GAT ACC ACT CCC AAC AGA CC and AAG TGC ATC ATC GTT GTT CAT ACA); Ifna forward and reverse (CGC AGG AGA AGG TGG ATG and TGC TGG TGG AGG TCA TTG); Ifng forward and reverse (GCT GAC CTA GAG AAG ACA CAT C and TTC CAC ATC TAT GCC ACT TGA G).

qPCR primer for mRNA expression of human genes

B2M forward and reverse (GAG TAT GCC TGC CGT GTG AAC and TCT AAG TTG CCA GCC CTC CTA G); CCL5 forward and reverse (CTG CTT TGC CTA CAT TGC CC and TCG GGT GAC AAA GAC GAC TG); IFT2 forward and reverse (CAA AGG GCA AAA CGA GGC AG and CCC AGG CAT AGT TTC CCC AG).

RNA sequencing and data analysis

For RNA sequencing, epidermis of three control and three Dnmt1 / mice was isolated at P3 and keratinocytes were cultured as described above for 5 days. RNA was extracted as described above and then used for RNA sequencing. For transcriptome sequencing (RNA‐seq) experiments, RNA was subjected to poly(A) selection with a Dynabeads mRNA purification kit (Invitrogen), followed by reverse transcription using a NEB RNA Ultra kit and library generation using a TruSeq library generation kit (Illumina). We performed full‐length mRNA‐seq experiments in three biological replicates for Dnmt1 / and the corresponding control keratinocytes and applied the “union” model of the htseq‐ count script (Anders et al, 2015) to calculate the number of reads associated with each of the 21,608 mouse RefSeq genes for each sample. We used these counts to compute reads per kilobase per million (RPKM) values for each gene and determined Spearman’s correlation coefficient (p) for each set of biological replicates. Based on the high correlation of the replicates (P = 0.99 between each 2 of the 3 control samples and between each 2 of the 3 knockout samples), we used the log‐transformed means of RPKM values under each condition to plot the distribution of gene expression levels by using kernel density estimation. Based on this distribution, we set the threshold for gene expression to 1 RPKM (log‐RPKM value equal to zero). This is consistent with data from previous studies, which estimated that the value of 1 RPKM corresponds to 1 transcript per cell (Mortazavi et al, 2008). The analysis of differentially expressed genes across the two conditions was performed by using htseq‐count and the Bioconductor edgeR package (Gentleman et al, 2004; Robinson et al, 2010). Genes that showed a minimum of a 2‐fold change in expression levels (adjusted P‐value of ≤ 0.05) were classified as upregulated, whereas genes displaying a fold change of ≤ 0.5 (adjusted P‐value of ≤ 0.05) were categorized as downregulated. The cutoffs for differential expression were set as absolute fold change > 2, and a corrected P‐value of > 0.05. GO analysis was performed with the DAVID (https://david.ncifcrf.gov/). The EntrezGeneIDs of the genes of interest were used as input for the web‐based analysis tools. The genes were functionally classified according to the Gene Ontology Consortium annotation category GO biological process.

Repeat analysis

To analyze repeat enrichment, we mapped the reads with bowtie (Langmead et al, 2009) (v 1.2.0, ‐m1 ‐‐max multi.fastq) against the human reference genome (GRCh38) or the mouse reference genome (MGSCv37). The uniquely aligning reads and the multi‐mappers were assigned to repeat classes, families, and types with RepEnrich (Criscione et al, 2014). The significantly differentially detected repeats were called using DESeq2 (Love et al, 2014) (v 1.10.1). The repeat count table from repseq was annotated with the class/family name of the repeat and plotted in a scatter plot of log fold change versus log10(adjusted P‐values).

RNA‐seq analysis

The 3' adaptors were removed using cutadapt (v1.4.2, Read1: cutadapt ‐‐match‐read‐wildcards ‐O 1 ‐a AGA TCG GAA GAG CAC ACG TCT GAA CTC CAG TCA C), and trimmed reads with a length of less than 18 bp were discarded. The trimmed reads were filtered with a contaminants database consisting of rDNA sequences (KY962518.1, NR_023363.1 and the ERCC.fa) with bowtie2 (v2.1.0 ‐‐very‐sensitive‐local). The recovered nonmatching reads were aligned to the genome/transcriptome using STAR‐align (v2.4.2a, STAR ‐‐outSAMstrandField None ‐‐outFilterIntronMotifs RemoveNoncanonical ‐‐outFilterMismatchNoverLmax 0.1 ‐‐outFilterMismatchNmax 10 ‐‐outFilterScoreMinOverLread 0.30 ‐‐outFilterMatchNminOverLread 0.30 ‐‐outFilterMatchNmin 30 ‐‐chimSegmentMin 15 ‐‐quantMode TranscriptomeSAM ‐‐chimJunctionOverhangMin 15 ‐‐twopassMode Basic ‐‐outSAMtype SAM ‐‐outSAMattributes All ‐‐outReadsUnmapped Fastx intronMotif ‐‐alignIntronMax 200000 ‐‐outSJfilterIntronMaxVsReadN 10000 20000 30000 50000 ‐‐outSJfilterOverhangMin 20 12 12 12 ‐‐outFilterType BySJout ‐‐alignMatesGapMax 0 ‐‐outFilterMultimapNmax 20). The STAR‐index used for alignment was generated from the human genome build NCBI GRCh38 (GCA_000001405.15) and a transcriptome GTF file from ENSEMBL build 78. Unique mappers from the resulting BAM file were used to create unstranded, stranded and reverse stranded coverage tracks for visualization using bedtools (v2.27, bedtools genomecov ‐split ‐bg, bedtools genomecov ‐split ‐bg ‐strand “+”, bedtools genomecov ‐split ‐bg ‐strand “‐”) and bedGraphToBigWig from the kent‐tools (bedGraphToBigWig v4, bedGraphToBigWig ‐blockSize=256 ‐itemsPerSlot=1024). Normalized coverage tracks were created by dividing the coverage at each position with the total number of aligned reads per million. The differentially expressed genes were called using DESeq2 (Love et al, 2014) (v 1.10.1).

Protein and histone isolation and immunoblot analysis

Protein and isolation from skin and cultured keratinocytes for immunoblot analysis and subsequent immunoblot signal quantification were performed as previously described (Winter et al, 2013).

Primary antibodies for immunoblotting

DNMT1 (1:500, sc‐20701, Santa Cruz Biotechnology), DNMT3a (1:1,000, sc‐20703, Santa Cruz Biotechnology), DNMT3b (1:1,000, sc‐376043, Santa Cruz Biotechnology), Actin (1:20,000; A5316, Sigma), Lamin B (1:1,000, sc‐6216, Santa Cruz Biotechnology), H3 C‐terminal (1:500,000, ab1791, Abcam), IAP (1:500, kindly provided by the Bryan R. Cullen lab (Bogerd et al, 2006), Center for Virology and Department of Molecular Genetics and Microbiology, Duke University Medical Center, Durham), PCNA (1:1,000, C1 PC10, Dako), γH2A.X S139ph (1:1,000, ab2893, Abcam), and cGas (1:1,000, 31659, Cell Signaling).

FACS analysis

Dermis and epidermis from pups were separated after overnight incubation in dispase solution (see above) at 4°C. Epidermis was cut into small pieces and incubated for 30 min at 37°C in 0.2 mg/ml DNAse I (Sigma) in RPMI. Dermis was cut and treated with RPMI containing DNAse and 1.6 mg/ml collagenase IV (Worthington) for 30 min at 37°C. A single cell suspension was prepared by passing the dissociated tissue through a 70‐μm cell strainer. Cells were washed with FACS buffer (1% BSA, 0.1% NaN3, PBS), and unspecific antibody binding was blocked with anti‐mouse CD16/CD32 (eBioscience). Approximately 1 × 10E6 cells per tube were stained with respective antibodies for 30 min at 4°C. 7AAD (Sigma) was added shortly before measurement for dead cell discrimination. Data were acquired on an FACS‐Aria and analyzed with FlowJo software according to the gating strategy shown in Appendix Fig S10.

Primary antibodies used for FACS

CD45 APC‐Cy7 (1:100, BioLegend, clone 30‐F11), MHC‐II PE (1:1,000, BioLegend, clone M5), CD11b PE‐Cy7 (1:100, BioLegend, clone M1/70), F4/80 APC (1:50, eBioscience, clone BM8), Ly6G PE (1:50, BioLegend), and 7AAD PerCP (2.5 μl/sample, Sigma).

Quantifications and Statistical analysis

Quantification of Ki67 and Histone H3S10ph (H3S10ph) epidermal cells

Paraffin sections of back skin from P7 wild‐type and Dnmt1 / mice were double immunolabeled for Ki67 and H3S10ph. For quantification, three random images per mouse were chosen and the number of basal and suprabasal epidermal H3S10ph‐positive and Ki67/H3S10ph double‐positive nuclei were counted in a blinded manner. The obtained triplicate values were averaged and compared with an unpaired two‐tailed t‐test.

Statistical analysis

In general, measurements were taken from distinct samples. qRT–PCR and quantification experiments were evaluated with Microsoft Excel. Immunoblot signal intensities were quantified using the ImageQuants software, and relative protein levels were normalized to beta‐Actin or Lamin B. The significance between groups was determined by the unpaired Student’s t‐test or by ANOVA. P‐values were calculated with the GraphPad Prism software, and standard deviation (s.d.) or standard error of the mean (SEM) is shown. *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ns = not significant (P > 0.05).

Quantification of micronuclei

PicoGreen‐stained cells were scanned using a 100x oil immersion objective. The scans were randomized, and 6 images each were blindly selected for analysis. Micronucleus‐associated nuclei were counted along with the total number of nuclei.

Author contributions

CSe, MAB, HF, and MW conceptualized the study. CSe, MAB, HF, MW, MS, and AA designed the study. MAB, HF, MW, LG, SS, PW, TG, TM, BZ, CFi, CFo, JA, BR, SL, GM, AP, MK, GE, MM, and CSch carried out experiments. ST, UR, MG, PP, and LK analyzed data. IT and RST carried out bioinformatic analysis. CSe, MAB and HF wrote the manuscript.

Conflict of interest

The authors declare that they have no conflict of interest.

Appendix

Click here for additional data file.

Expanded View Figures PDF

Click here for additional data file.

Dataset EV1

Click here for additional data file.

Dataset EV2

Click here for additional data file.