DNA hypomethylation leads to cGAS-induced autoinflammation in the epidermis

Abstract

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.

Keywords: DNA methylation; autoinflammation; cytosolic DNA; epigenetics; innate immune system.

Figure 1. Epidermal DNA hypomethylation results in a severe skin phenotype

1. A

   Immunofluorescence labeling of wild‐type mouse dorsal skin samples (Dnmt1^{f / f} , skin from P3) shows expression of DNMT1 throughout epidermal layers whereas keratinocyte‐specific knockout (Dnmt1^{Δ / Δep} ) 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.

2. B, C

   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).

3. D

   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.

4. E

   HPLC analysis of cytosine methylation in the epidermis of Dnmt1^{Δ / Δep} 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.

5. F, G

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

6. H

   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^{Δ / Δep} mice.

Data information: Scale bars, 50 µm.

Figure EV1. Keratinocyte‐specific deletion of Dnmt1 in mice results in early postnatal lethality associated with severe dermatopathologic alterations

1. Weight curve of control (Dnmt1^{f / f} ) and Dnmt1^{Δ / Δep} 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^{Δ / Δep} 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.

2. Kaplan–Meier plot comparing control (Dnmt1^{f / f} n = 13, Dnmt1^{f /+} n = 9, Dnmt1^{Δ /+ ep} n = 14) and Dnmt1^{Δ / Δep} (n = 11) mice using log‐rank test (Mantel–Cox). P < 0.0001.

3. H&E staining of dorsal skin sections from control and Dnmt1^{Δ / Δep} mice at P3, P5, and P7. Enlarged sections are indicated on the right. Dermatopathologic alterations found in Dnmt1^{Δ / Δep} 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).

4. 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).

Figure 2. Epidermal DNA hypomethylation results in induced immune response

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

2. 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.

3. 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).

4. Relative mRNA expression levels of immune‐related genes in control and Dnmt1^{Δ / Δep} 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.

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

6. 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^{Δ / Δep} 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

1. A

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

2. B

   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).

3. C–E

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

4. F

   The inside out barrier of the skin of 3‐ to 7‐day‐old Dnmt1^{Δ / Δep} and control pubs was tested by a dehydration assay, n = 5 (P3), n = 11 (P5, control), n = 6 (P5, Dnmt1^{Δ / Δep} ), n = 3 (P7, control), and n = 4 (P7, Dnmt1^{Δ / Δep} ). 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^{Δ / Δep} mice exhibit DNA hypomethylation and derepression of transposon/repeat families

1. A

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

2. B

   Expression of deregulated genes found by RNA sequencing of cultured primary keratinocytes that were isolated at P3 of control and Dnmt1^{Δ / Δe 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.

3. C

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

4. D

   Relative mRNA expression of transposable elements in the epidermis of control and Dnmt1^{Δ / Δep} 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.

5. E, F

   DNA methylation of CD45‐depleted epidermal keratinocytes isolated from newborn wild‐type (n = 3) and Dnmt1^{Δ / Δe 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^{Δ / Δe 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

1. A, B

   Immunofluorescence double‐labeling of G2/M cells (H3S10ph) and proliferating cells (Ki67) of dorsal skin sections from day 7 old control and Dnmt1^{Δ / Δep} 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^{Δ / Δep} mice (n = 4). Values were compared using two‐tailed Student's t‐test. Data are mean ± SEM. *P ≤ 0.05.

2. C

   Quantification of basal (left) and suprabasal (right) cells from day 7 old control (n = 4) and Dnmt1^{Δ / Δep} 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.

3. D–F

   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.

4. G

   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.

5. H

   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^{Δ / Δep} mice

1. 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.

2. Immunoblot analysis for expression of DNMT1 and cGAS in lysates of isolated mouse epidermis of control, Dnmt1^{Δ / Δep} , and Dnmt1^{Δ / Δep} Cgas ^−/− 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.

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

Figure 3. Epidermal deletion of DNMT1 results in derepression of transposable elements and activation of immune genes with minor impact of the MAVS pathway

1. 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.

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

3. Relative expression of immune‐related genes of the epidermis from P7 of control (n ≥ 7), Dnmt1^{Δ / Δep} , (n ≥ 8), MAVS knockout (Mavs ^−/−, n = 7), and Dnmt1^{Δ / Δep} Mavs ^−/− (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.

Figure 4. DNA hypomethylation results in release of cytoplasmic DNA and cGAS activation

1. A

   Paraffin skin sections (P5) of control and Dnmt1^{Δ / Δep} 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.

2. B

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

3. C–F

   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).

Figure 5. Additional deletion of cGAS attenuates the autoinflammatory Dnmt1^{Δ / Δep} skin phenotype

1. Representative images of P7 H&E‐labeled dorsal skin sections of control (Dnmt1^{f / f}), Dnmt1^{Δ / Δep} , and Dnmt1^{Δ / Δep} Cgas ^−/− mice.

2. 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.

3. Representative images of CD45 immunofluorescence labeling of P7 dorsal skin sections from Dnmt1^{f / f} , Dnmt1^{Δ / Δep} , and Dnmt1^{Δ / Δep} Cgas ^−/− mice. Dashed lines indicate the dermal‐epidermal border.

4. 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.

5. Epidermis obtained from P7 control, Dnmt1^{Δ / Δep} , Cgas ^−/−, and Dnmt1^{Δ / Δep} Cgas ^−/− 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).