Epigenetic alterations in skin homing CD4+CLA+ T cells of atopic dermatitis patients

T cells expressing the cutaneous lymphocyte antigen (CLA) mediate pathogenic inflammation in atopic dermatitis (AD). The molecular alterations contributing to their dysregulation remain unclear. With the aim to elucidate putative altered pathways in AD we profiled DNA methylation levels and miRNA expression in sorted T cell populations (CD4+, CD4+CD45RA+ naïve, CD4+CLA+, and CD8+) from adult AD patients and healthy controls (HC). Skin homing CD4+CLA+ T cells from AD patients showed significant differences in DNA methylation in 40 genes compared to HC (p < 0.05). Reduced DNA methylation levels in the upstream region of the interleukin-13 gene (IL13) in CD4+CLA+ T cells from AD patients correlated with increased IL13 mRNA expression in these cells. Sixteen miRNAs showed differential expression in CD4+CLA+ T cells from AD patients targeting genes in 202 biological processes (p < 0.05). An integrated network analysis of miRNAs and CpG sites identified two communities of strongly interconnected regulatory elements with strong antagonistic behaviours that recapitulated the differences between AD patients and HC. Functional analysis of the genes linked to these communities revealed their association with key cytokine signaling pathways, MAP kinase signaling and protein ubiquitination. Our findings support that epigenetic mechanisms play a role in the pathogenesis of AD by affecting inflammatory signaling molecules in skin homing CD4+CLA+ T cells and uncover putative molecules participating in AD pathways.

. Descriptive characteristics of the participants at 450 ml blood donation. a Given the sample size and to avoid confounding effects by gender, only males were analyzed. b SCORAD: SCORing Atopic Dermatitis. Moderate eczema (score 15 to 40); severe eczema (> 40); n.a = not applicable. c Determined by ImmunoCAP (Thermo Fisher, Uppsala, Sweden). Reference value in this population is 122 kUA/L; IQR = interquartile range. d Phadiatop = analyses of IgE antibodies in plasma to a mix of aeroallergens including birch, timothy, mugwort, mite, cat, dog, horse and mold (Thermo Fisher). e The presence of physician-diagnosed asthma and/or rhinitis was obtained by questionnaire at the time of blood sampling.

Healthy controls AD patients
Sample size (n) a 10 10 Age, years (average ± SD) 37 ± 13.9 34.9 ± 14.7 SCORAD (objective) b n.a 44 ± 5.5 Total plasma IgE kU A /L, (median, IQR) c 16.5 (8.2 -22.5) 3 Vol.:(0123456789) Scientific Reports | (2020) 10:18020 | https://doi.org/10.1038/s41598-020-74798-z www.nature.com/scientificreports/ Comparative analysis of DNA methylation between AD patients and HC. The analysis of the genome-wide DNA methylation levels in the four sorted T cell populations revealed that skin-homing CD4 + CLA + T cells were the subset with the largest numbers of differentially methylated CpG probes (DMPs, n = 49) between AD patients and HC (Fig. 1a). These cells expressed the CD3 + marker, as well as the CLA + and the CD45RO + markers, and in about 60% of the population the CCR4 receptor (Fig. 1b). The 49 DMPs in these circulating CD4 + CLA + T cells (BH corrected p value < 0.05) mapped to 35 genes (Table 2). Further analysis of DMRs revealed 5 genes with more than 3 DMPs in the CD4 + CLA + T cells of the AD patients compared to HC (Table 3). These results refined the signals detected in GPR55, MAN1A1 and CDHR3 by using the DMP analysis ( Table 2) and uncovered methylation differences in the genes encoding the transcription factor estrogen receptor alpha (ESR1) and the nuclear receptor co-repressor 2 (NCOR2) ( Table 3), suggesting that in AD patients the epigenome of circulating CD4 + CLA + T cells is affected in regions encoding transcription factors. The annotated list of 40 differentially methylated genes between AD patients and HC with their cell location and known function are presented in Table 4. Some of the differentially methylated genes in CD4 + CLA + T cells of AD patients are implicated as key regulators of cytokine signaling pathways and immune inflammation (ARHGEF3, ASB2, DAPP1, IL10RA, PDE4A, SH2B3, STIM1 and TOX2, Fig. 1c). We also found that CD4 + CLA + T cells from AD patients have significantly decreased methylation in the IL13 promoter (Fig. 2). DNA methylation levels at the CpG site cg14523284 in the upstream region of IL13 were significantly lower compared to the levels in HC (Fig. 2a), by contrast, mRNA levels for IL13 were increased in AD patients (Fig. 2b). Spearman correlation tests showed a significant inverse correlation between DNA methylation and IL13 mRNA levels (Spearman rho −0.63, p = 0.006) (Fig. 2c), mainly driven by AD cases. This CpG site with reduced methylation levels in AD patients was at the proximal upstream of the Th2-specific DNA hypersensitive site in the IL13 promoter but within the Th2 locus-control long noncoding RNA 37 (Fig. 2d), indicating that this epigenetic modification might functionally explain the augmented capability of CD4 + CLA + T cells of AD patients to produce IL-13. Correlations computed within each group further strengthen the distinct AD vs HC responses, showing a clear trend within the former group (Spearman rho < −0.92, p < 0.0002, FDR < 0.003, n = 10) but no statistically significant relationship in the latter (Spearman rho = 0.39, p > 0.38, FDR > 0.38, n = 7). A linear regression to each of the two datasets shows a statistically significant relationship between delta-Ct and M-value in AD (p < 0.0005, R 2 > 0.82) but not in HC (p > 0.65, R 2 < 0.05) (Fig. 2c). miRNA deregulation in CD4 + CLA + T cells of AD patients. The analysis of global miRNA expression levels in the four T cell populations revealed that only the CD4 + CLA + T cells contain differentially expressed miRNAs (n = 16) between AD patients and HC (BH corrected p value < 0.05). In AD patients, 10 miRNAs were up-regulated, and 6 miRNAs were down-regulated, allowing a clear distinction between AD patients and HC (Fig. 3a). We selected 8 differentially expressed miRNAs from the microarray analysis (miR-7-5p, miR-21-3p, miR-93-5p, miR-130b-3p, miR-145-5p, miR-150-5p, miR-181b-5p and miR-1275) for technical validation by qPCR. Significant differences between AD patients and HC could be confirmed by qPCR for four of them, miR-21-3p, miR-130b-3p, miR-150-5p and miR-1275 (Fig. 3b,c). Next, we performed gene set enrichment analysis on the predicted miRNA targets of upregulated and downregulated miRNAs in AD (Fig. 4) and found 202 biological processes associated with the targets of the miRNAs dysregulated in AD (Supplementary Table S2 online). The top pathways (FDR < 1.1 × 10 −5 ) included cell differentiation and migration, apoptosis ubiquitin-dependent protein catabolic process, transforming growth factor beta receptor signaling pathway and positive regulation of MAP kinase activity. We found that ESR1, NDFIP2, ASB2 and TNRC6A genes which were differentially methylated in AD patients (Table 4) were also targeted by upregulated miRNAs in AD patients (Supplementary Table S2 online), suggesting complex interactions between these epigenetic layers.
Network analysis of coordinated epigenetic responses reveals potentially perturbed pathways in circulating CD4 + CLA + T cells from AD patients. To further explore the coordinated epigenetic relationships, we performed network analysis using the full set of miRNAs, as well as the top ~ 1% CpG sites showing the largest differences between AD and HC. An association analysis (absolute Spearman ρ > 0.75, FDR < 0.001) was used to identify strongly interconnected communities of miRNA/CpG sites. Communities are sets of miRNA/CpG sites that tend to display globally coordinated expression patterns, thus highlighting potentially harmonized regulatory effectors 38 . Importantly, the set of miRNA/CpG sites within a community are more strongly associated with their respective community than between communities. Our analyses highlighted six communities based on the global coordinated associations between miRNA and CpG sites. Interestingly, we identified one miRNA/CpG community (C5) that recapitulated the genes showing significant differences in the AD group (Fig. 5a). This community has 640 elements: 122 miRNAs and 518 CpG sites that included 8 out of 12 CpG sites with increased DNA methylation in AD patients (Table 2) but none with decreased methylation. This community also contained all 10 miRNAs that had been identified as upregulated in the AD group by the differential expression analysis (Fig. 3a), suggesting that our approach was powered enough to capture AD associated features and its neighbors including three miRNAs targeting IL13 (hsa-miR-98-5p, hsa-let-7d-5p, hsa-let-7f.-5p) (Fig. 5a). On the other hand, community C3 has 765 elements: 15 miRNAs and 765 CpG sites including CpG probes with increased DNA methylation in HC (DAPP1, STIM1, PDEA4 and TOX2) and 4 miRNAs with increased expression in HC ( Fig. 1b & Fig. 3). Overall, communities C3 and C5 reflect the antagonistic behavior that we had observed in HC and AD patients (Fig. 1b). Detailed information on the CpG sites and miRNAs detected in each community is presented in Supplementary Table S3 online.
The list of miRNAs targets together with the genes annotated to contain CpG sites associated with each community were further functionally characterized (Supplementary Table S4  www.nature.com/scientificreports/ many processes related to cell signaling. Indeed, top pathways (FDR < 10 −8 ) associated with community C5 revealed 17 involved in protein ubiquitination, positive regulation of intracellular signal transduction, protein phosphorylation, positive regulation of IKK/NFkB signaling, Ras protein signal transduction, stress-activated MAPK cascade and cytokine-mediated signaling (Fig. 5b). Overall, this suggests that epigenetic alterations in CD4 + CLA + T cells of AD patients may be affecting receptors and adaptors that are crucial for the regulation of cytokine signaling. Genes that showed increased DNA methylation in AD were represented in these pathways, and included ASB2 in protein ubiquitination, IL10RA in cytokine mediated signaling and ESR1 as a common gene in several C5 pathways (Fig. 5b).

Discussion
Several studies have shown disturbed biology in skin homing CLA + T cells in AD patients but the underlying mechanisms explaining the alterations in this cell population remain unclear. We here analyzed the combined profiles of DNA methylation and miRNA expression in sorted peripheral blood T cell populations from AD patients compared to HC. This study revealed for the first-time significant differences in the DNA methylation levels of several key immune genes in skin homing CD4 + CLA + T cells from AD patients. The most significant differences among the 40 differentially methylated genes were found in ASB2, DAPP1, FBXO34 and NDFIP2 ( Table 2). We also found significant DNA methylation differences in genes known to be genetically associated with AD predisposition including IL13 39,40 , IL10RA 41 , ZNF365 42 and STIM1 43 . Our results revealed for the first time a significant inverse correlation between reduced DNA methylation in the IL13 promoter and increased IL13 mRNA expression in CD4 + CLA + T cells of AD patients (Fig. 2c), providing insights into the molecular events that might lead to the remarkable ability of CLA + T cells from AD patients to secrete IL-13 22,24,44 . The relationship between reduced DNA methylation in the IL13 promoter of AD patients and increased IL-13 expression support that this cytokine is a central pathogenic mediator in AD 45 and therapies targeting this molecule or its receptors (e.g. Dupilumab) might be beneficial in patients with severe atopic phenotypes 46 . The increased DNA methylation levels in the IL13 promoter observed in HC could explain why the IL13 mRNA expression was significantly reduced in this group (Fig. 2). DNA methylation differences in the IL13 gene (cg04303330) has been also described by Boorgula et al., in whole blood samples from patients with the phenotype of AD with eczema herpeticum 47 .
We also found reduced methylation in the gene PDE4A in AD patients, a molecule that promotes downstream inflammatory pathways and is amenable to be targeted with phosphodiesterase inhibitors (Fig. 1b) 48 . Since CLA + T cells recirculate between skin and blood and mediate pathogenic inflammation, antibodies targeting adhesion molecules used by CLA + T cells to mediate their transendothelial migration 49 or anti-CLA antibodies are being attempted and proposed as promising therapeutic options in AD 50 . However, more studies are needed because CLA + T cells are involved in immunosurveillance and preventing their migration to skin produces CLA + leukocytosis and possibly alterations in other tissues.
We discovered 16 miRNAs to be differently expressed in CD4 + CLA + T cells from AD patients (Fig. 3), all except for miR-21 previously undescribed in AD 12 . The magnitude of the expression differences between patients and controls was small, but the effect sizes for miRNA variation are unknown. The up-regulation of miR-21 and miR-145 may reflect the pro-inflammatory status of CD4 + CLA + T cells in AD patients. Indeed, miR-21 is involved in the polarization of adaptive immune responses and has been found upregulated in lesional skin biopsies of AD patients and in lesional skin of patients with contact dermatitis after challenge with diphenylcyclopropenone 51 . The altered miRNA signatures in AD may differ depending on the cell type or tissue studied. We did not find Table 3. Differentially methylated regions (DMRs) with decreased DNA methylation in CD4 + CLA + T cells from AD patients compared to HC. a DMR analysis is based on minfi to collapse connected DNA methylation probes by distance rules; width in base pairs. b Sites in ESR1 cg21157690, cg17264271, cg15543523, cg26089753, cg08884395, cg01715172, cg21608605, cg20627916, cg07671949, cg23164938, cg23165623, cg21614759, cg19411146, cg21950534, cg11813455, cg24900983, cg05171584, cg23467008, cg22839866, cg23009221, cg27316393, cg00655307, cg01777019. CpG sites indicated in bold were also found as differentially methylated CpG sites in the DMP analysis (see Table 2). Chr: chromosome.  www.nature.com/scientificreports/ any differences in the expression of miR-146a or miR-155 between AD and HC, previously reported to be up-regulated in lesional skin biopsies of AD patients 52,53 consistent with that in those studies the significant differences were attributed to the keratinocytes. Moreover, we discovered 202 biological processes significantly enriched in targets of AD-associated miRNAs of which many were implicated in cell signaling, transforming growth factor beta production and interferon responses (Fig. 4). These analyses highlighted ESR1 as a target of several AD upregulated miRNAs, involved in several processes such as cell differentiation and migration, transforming growth factor beta receptor signaling pathway and T cell differentiation (Supplementary Table S2 online); suggesting for the first time that this transcription factor is a dysregulated mediator of several putative disturbed pathways in CLA + T cells from AD patients. These observations are in line with recent studies showing that estrogen-responsive genes may influence IL-13 production in patients with eosinophilic esophagitis 54 . We also evaluated putative targets of the dysregulated miRNAs in CD4 + CLA + T cells, not only for genes expressed in CD4 + T cells but using the entire miRTarBase targets, since these altered miRNAs may be released from the T cell and exert their effects on skin cells, endothelium or other immune cells.
The integrative network analysis of coordinated changes of DNA methylation and miRNA in CD4 + CLA + T cells revealed several pathways and new candidate dysregulated genes for AD (Fig. 5b). These included ASB2 a gene encoding the ankyrin-repeat suppressor of cytokine signaling (SOCS) box-containing protein 2, an E3 ubiquitin ligase that promotes protein degradation by coupling SOCS proteins with the elongin BC complex and mediate the ubiquitination of Notch targets such as E2A and Janus kinase (Jak) 2 55 and Jak3 56 . ASB2 is known to be expressed in T helper 2 cells (Th2), is regulated by GATA3 and promotes Th2 phenotype in vivo 57 . Further studies are needed to elucidate the role of ASB2 in CLA + T cells from AD patients. In contrast, HC showed increased methylation in NDFIP2 a gene encoding for the Nedd4 Family Interacting Protein 2 which has been described to limit the cytokine signaling and expansion of effector Th2 cells 58 by promoting degradation of JAK1, probably by ITCH-and NEDD4L-mediated ubiquitination 59 . Two other genes from the F-box family (FBXO18 and FBXO24) implicated in protein ubiquitination were found differentially methylated in AD patients (Table 2), altogether suggesting that genes implicated in protein ubiquitination are epigenetically altered in CD4 + CLA + www.nature.com/scientificreports/ T cells from AD patients, and may promote inflammation by altering signaling. Ubiquitin ligases also mediate inhibition of activation of induced cell death (AICD) and contribute to lymphocyte accumulation 60 , therefore, altered epigenetic signatures in genes involved in protein ubiquitination may contribute to the resistance to apoptosis that acquires the CLA + T cells upon skin infiltration 25 . The coordinated epigenetic changes highlight several genes and pathways involved in intracellular signaling and were consistently associated with AD features in the CD4 + CLA + T cells from patients. These included cytokine-mediated signaling associated with IL10RA and its intracellular adaptor SH2B3 which constrains cytokine signals and might influence inflammatory immune responses in peripheral lymphoid tissues 61 . Also, the Ras signaling transduction pathway implicating ARHGEF3; as well as positive regulation of IKK/NFKB signaling implicating NDFIP2 (Fig. 5b). Among genes in community C3, TOX2 showed increased methylation in HC (Fig. 1b & Fig. 5a); this transcription factor is being recognized as a regulator of T cell differentiation but further studies are needed to evaluate its role in CLA + T cells 62,63 .
We analyzed CD4 + CD45RA + naïve T cells since we hypothesized that epigenetic signatures predisposing to T cell dysregulation in AD patients may already be present at this cell stage before the effector or memory T helper cell differentiation. In agreement with Han et al. 64 , we found almost no differences in DNA methylation levels between CD4 + naïve T cells from AD patients compared to HC (Fig. 1a); neither in the unfractionated CD4 + T cells as described by Rodriguez et al. 11 , nor in CD8 + T cells supporting that epigenetic alterations are principally detected in CD4 + CLA + T cells. The CD8 + CLA + T cells also have an important role in AD 65,66 , although recent studies highlighted their pathogenic role in psoriasis 67 . Further studies are needed to evaluate epigenetic signatures in the CD8 + CLA + T cell populations in AD patients.
Provided that T cell turnover between skin and blood involves active de-homing from skin and migration through the lymph nodes and peripheral circulation 26 , it is conceivable that CD4 + CLA + T cells isolated from blood could be informative on pathogenic processes occurring in skin. It remains to be determined at which developmental state the CD4 + CLA + T cells acquired these altered epigenetic signatures in AD patients. Most probably by signals received after antigen priming or during their recirculation within skin, lymph nodes and the peripheral blood. Exposure to environmental signals from the skin microbiome or allergens could also be involved. We do not know the TCR specificity of the CD4 + CLA + T cells analyzed here but previous studies have shown that CLA + T cells in AD preferentially respond to allergens and Staphylococcal enterotoxin B 68 .
We did not compare frequencies of CLA + T cells between AD patients and HC in peripheral circulation, however, in agreement with previous reports 25 , we found abundant infiltrates of CLA + cells in the lesional skin biopsy specimens of the AD patients compared to HC (Fig. 6a-c), which were also dominated by CD3 + and CD4 + cells in consecutive sections (data not shown). Given the larger numbers of these epigenetically altered CD4 + CLA + T cells in the skin of patients (Fig. 6b) and the recirculation of these cells, we hypothesize that the epigenetic alterations detected in CLA + T cells from peripheral blood promote pro-inflammatory functions in skin of patients and contribute to AD immunopathology (Fig. 6d). It should be noted that AD is a heterogeneous www.nature.com/scientificreports/ disease with different immunophenotypes in the cellular infiltrates, including not only Th2 inflammation but also other cell populations [69][70][71] . We here included adult patients with a phenotype of severe AD (objective SCORAD above 40) and allergen specific IgE sensitization. Therefore, the epigenetic signatures detected in this study and their related genes may be associated with this phenotype of patients, and more research is needed to evaluate their implications in other AD cohorts. Several methodological aspects add strength to this study. To avoid the confounding effects that cell heterogeneity has on the interpretation of epigenetic and expression studies, we analyzed sorted T cell populations in peripheral blood. In addition, we used a robust methodology that interrogated DNA methylation levels in ~ 450 000 CpG sites through the genome and global miRNA expression in ~ 2000 miRNAs. At several loci, the magnitude of the epigenetic differences between AD and HC were relatively small, yet statistically significant after   Table S2 online. This circular layout was created with the free R package circlize (https ://cran.r-proje ct.org/web/packa ges/circl ize/index .html). Vol.:(0123456789)

Scientific Reports
| (2020) 10:18020 | https://doi.org/10.1038/s41598-020-74798-z www.nature.com/scientificreports/ correction by multiple testing. Several of the loci detected in the comparison between AD patients and HC (Fig. 1b & Table 2) showed coordinated changes when analyzed with a different algorithm for network analysis (Fig. 5a). Our study has also some limitations, to avoid the confounding effects of gender, we included only male AD patients that even though they were selected to be as much homogeneous as possible had some differences in their total serum IgE levels, AD severity and the presence of comorbidities such as asthma (Table 1). It is worth noting that a larger sample size would assist in identifying important biological signals in the etiology of AD. Due to the limitations in the amount of cells that were sorted per individual, we could only measure mRNA expression in few genes by TaqMan. We detected significant differences in IL13 expression, but we could not interrogate how the coordinated changes in miRNA and CpG methylation affects global gene expression. Nevertheless, this study revealed new genes and biological processes that deserve to be further validated at mRNA Figure 5. Integrated network analysis of miRNAs and CpG sites (CpGs) highlights coordinated epigenetic changes in CD4 + CLA + T cells. (a) Based on the set of associated miRNAs and CpG sites (absolute Spearman ρ > 0.75, FDR < 0.001) we identified 6 communities (C1-C6) of highly interconnected miRNAs/CpGs. These shows coordinated expression patterns, with strong interconnections within but not between communities. Note that communities 3 (yellow) and 5 (blue) show many of the differential miRNAs/CpGs detected in HC and AD patients: CpG sites (black), miRNAs (shadow), and miRNAs targeting IL13 (with asterisk). The node size for each community is proportional to the number of elements in each community. Line thickness connecting the communities is proportional to the number of connections between them. Detailed information on these communities is presented in Supplementary Table S3 online. (b) Functional enrichment analysis on putative regulated targets revealed 17 key pathways associated with AD (community C5) and are indicated as blue circles. Genes highlighted in pink and yellow were also detected as differentially methylated in the comparison between AD patients and HC ( Table 2). The networks were created in Cytoscape 3.7.2 (https ://cytos cape.org/).

Scientific Reports
| (2020) 10:18020 | https://doi.org/10.1038/s41598-020-74798-z www.nature.com/scientificreports/ and protein levels in other cohorts of AD patients and functionally evaluated at the cellular level to dissect their role in AD pathogenesis. In summary, we here discovered putative altered molecular pathways in circulating CD4 + CLA + T cells from AD patients, involving disease-associated signatures in DNA methylation and miRNA levels. The identified loci highlight new candidates in AD, including genes mediating intracellular cell signaling and adaptor molecules of the IL-10/IL-13 interleukin signaling pathway as well as genes involved in protein ubiquitination. Our findings support epigenetic profiling as a valuable tool to uncover putative molecules participating in disease pathways. Further studies are needed to define the downstream effects of these epigenetic alterations in AD immunopathology and evaluate if environmental signals at the target organ (e.g., skin microbiota) induce the detected epigenetic differences in circulating CD4 + CLA + T cells from AD patients.

Methods
Study population and samples. Twenty age-matched adult male individuals (10 AD patients and 10 HC) were selected from a Swedish eczema study 72 . All participants were asked to visit the Dermatology Unit at the Karolinska University Hospital in Solna and were examined by a dermatologist for their general physical conditions, in case of AD patients, also for the severity of the eczema. Afterwards, all provided a 450 ml blood donation (see below). The cases fulfilled the following inclusion criteria: (1) a physician-diagnosis of moderate or severe AD as determined by the objective Scoring Atopic Dermatitis index (SCORAD) (2) elevated levels of total plasma IgE > 122 kU A /L (ImmunoCAP, Thermo Fisher Scientific, Uppsala, Sweden), (3) positive allergenspecific IgE > 0.35 kU A /L as determined by Phadiatop (Thermo Fisher, Uppsala, Sweden), and (4) besides AD, a clinical history of asthma and/or rhinitis. All AD patients had used topical corticosteroids at least 30 days before blood sampling. Controls were healthy individuals without clinical history of eczema, asthma or other allergic diseases and plasma IgE levels below 122 kU A /L. On the same day as the 450 ml blood donation, skin biopsies were obtained from AD patients (n = 5) and from HC (n = 5). Punch biopsies (4 mm) were taken under local anesthesia (Lidocaine 5 mg/ml with epinephrine 5 µg/mL, Astra, Södertälje, Sweden) from eczema lesions in the AD patients located in the popliteal fossa, upper/mid back or shoulders and healthy skin of HC from the lowback (lumbar) region. The biopsies were snap-frozen on dry ice and stored at −80 °C until immunohistochemistry analysis (see below). This study was conducted in accordance with the Helsinki Declaration ethical principles for medical research and was approved by the Regional Ethical Review Board in Stockholm (Dnrs 04-593/1, 2008/1122-32, 2010/754-32, and 2011/1051-31). All participants gave their written informed consent.

Isolation of T cell populations. A 450 ml blood donation was obtained at the Blood Transfusion Center
Karolinska University Hospital and processed immediately for cell isolation. Peripheral blood mononuclear cells (PBMCs) were separated by density centrifugation on Ficoll-Paque Plus (GE Healthcare, Uppsala, Sweden) and then labeled for magnetic associated cell sorting (MACS) to obtain CD4 + T cells and CD8 + T cells, CD4 + CD45RA + naïve T cells and skin-homing CD4 + CLA + T cells (Miltenyi Biotec, Gladbach, Germany). Cell viability after sorting was > 90% in all populations (as determined by trypan blue exclusion) and the purity was examined by flow cytometry (see below). Two aliquots were stored at −80 °C: one containing 5 × 10 6 cells for DNA extraction as described previously 36 and another containing a maximum of 10 × 10 6 cells homogenized in QIAzol lysis reagent (Qiagen, Hilden, Germany) for total RNA extraction. . The data were normalized using the subset-quantile within array normalization (Swan) method 74 . The probes overlapping with known SNPs were removed, leaving 398 494 CpG sites for further analysis. Differentially methylated CpG sites between AD patients and controls were obtained using linear models (y ~ group + subject + age) and pairwise comparisons with empirical Bayes as implemented in limma 75 . DMRs were analyzed with the minfi package 73 .
All the CpG sites presented as having differential methylation between AD patients and HC had a p value < 0.05 after Benjamini-Hochberg (BH) correction for multiple testing.
RNA isolation and miRNA measurements. mRNA and miRNA were obtained by phenol/chloroform extraction. RNA integrity was evaluated using the Nanochip kits in Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA) resulting in RIN values (mean ± SD) of 9.31 ± 0.38. miRNA levels were measured for 4774 probes (representing 2006 human miRNAs, miRBase database release 19) using the SurePrint G3 (8 × 60 K) miRNA array (Agilent Technologies). The raw .txt files obtained from the Agilent feature extraction software were imported in R and analyzed with limma 75 . After quality check 71 samples remained for the analysis of differential miRNA expression: 17 CD4 + (HC = 8, AD = 9), 15 CD4 + CD45RA + naïve (HC = 6, AD = 9), 19 CD4 + CLA + (HC = 9, AD = 10) and 20 CD8 + (HC = 10, AD = 10). To ensure homogeneity of analytical strategy with the DNA methylation, the same statistical approach described above was also used to find differentially expressed miR-  (d) Circulating CD4 + CLA + T cells from AD patients (light blue cell) show significant differences in DNA methylation and miRNA levels compared to CD4 + CLA + T cells from HC (purple cell). The main differences were detected in the reduced DNA methylation of the IL13 gene, the increased expression of proinflammatory miRNAs and coordinated epigenetic changes in genes involved in protein ubiquitination and cytokine signaling in AD patients. Since these CD4 + CLA + T cells can recirculate between skin and blood 25 , these altered epigenetic marks might contribute to AD immunopathology.
Scientific Reports | (2020) 10:18020 | https://doi.org/10.1038/s41598-020-74798-z www.nature.com/scientificreports/ ther filtered for redundancy based on their similarity through Revigo 79 . Associations with miRNA targets were plotted using circus through the R circlize package 80 and Cytoscape 3.7.2 (https ://cytos cape.org/). qPCR validation of miRNA and mRNA expression in CD4 + CLA + T cells. To validate differentially expressed miRNAs, cDNA was prepared from the total RNA of CD4 + CLA + T cells (15 ng) from 10 AD patients and 9 HC (the same samples as used in the Agilent array) using the miRCURY LNA Universal RT micro-RNA PCR kit (Exiqon A/S, Vedbaek, Denmark) according to the manufacturer's protocol. miRNAs were amplified using the miRCURY LNA SYBR Green master mix and specific PCR primer sets for 8 miRNAs (Exiqon) according to the manufacturer's instructions. SNORD44 (snRNA) was used as endogenous control as described previously 81 . To validate mRNA targets, cDNA was prepared using the high capacity RNA to cDNA kit (Life Technologies) starting from 80 ng of total RNA from 8 HC and 10 AD patients. Gene expression levels (mRNA) were measured using validated Taqman probes (Applied Biosystems, Foster City, CA, USA) according to manufacturer's instructions using beta-2 microglobulin (B2M) as the reference gene. Quantitative real-time PCR was performed in the Biorad CFX96 system (BioRad Inc., Hercules, CA, USA) with each sample in duplicate, and the gene expression data were exported from the DataAssist Software v3.01. Differential expression between AD patients and HC was calculated using the comparative Ct method. In brief, the average Ct value for each miRNA was subtracted from the Ct value of SNORD44 to obtain the delta-Ct. The comparisons of delta-Ct between AD patients and HC were calculated by unpaired t-tests and a p value < 0.05 was considered significant.
Immunohistochemistry. The frozen skin biopsy specimens were embedded in optimal cutting temperature compound and six µm thin sections were prepared in a microtome-cryostat, two sections per glass slide. The tissue architecture and degree of cell infiltrates were evaluated by hematoxylin and eosin staining. For the evaluation of T cell markers, the sections were acetone fixed and stained using the avidin-biotin complex method Statistical analysis. Statistical approaches were chosen according to the diverse data types in this study and explained in their respective methods sections. Hypothesis testing was performed by considering the null hypothesis of the absence of an association between the compared variables. The associations were tested according to the nature of the data: continuous vs continuous (Spearman rank correlation test); continuous vs categorical (t-test or Mann Whitney test according to data distribution). p values were corrected by the Benjamini-Hochberg procedure for multiple testing and a value < 0.05 was considered statistically significant.