Objectives The aetiology of primary Sjögren's syndrome (pSS), also referred to as autoimmune epithelitis, is incompletely understood but includes an epigenetic contribution. Accordingly, the aim of this study was to investigate DNA methylation in salivary gland epithelial cells (SGEC), and to compare results with those publicly available from pSS B and T cells.
Methods Long-term cultured SGEC were selected to conduct an epigenome-wide association study (EWAS) in patients with pSS with comparison to controls using the HumanMethylation 450 K array from Illumina.
Results The analysis of differentially methylated CpG (DMC) uncovered 4662 positions corresponding to 2560 genes, and 575 genes with two or more DMC sites (DMCs), in SGEC as compared with controls. Further analysis highlighted an important proportion of interferon-regulated genes (61%), the calcium pathway (hypomethylated) and the Wnt pathway (hypermethylated). When comparing SGEC with pSS T and/or B cell results, an important overlap was observed with respect to differentially methylated genes (38.8%) and pSS risk factors (71.4%), although such assertion was not true when comparing DMCs.
Conclusions This study conducted in SGEC emphasises the role of DNA methylation in pSS pathogenesis and supports the necessity to conduct pure cell analysis for future EWAS studies when analysing salivary glands from patients with pSS.
- Sjøgren's Syndrome
- Gene Polymorphism
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
In primary Sjögren's syndrome (pSS), several lines of evidence suggest a role for DNA methylation in the disease's pathophysiology, and this is reinforced by recent epigenome-wide association studies (EWAS).1 With this in mind, quantitative analysis of 485 577 CpG sites in pSS peripheral blood naïve CD4+ T cells and CD19+ B cells was conducted using the Illumina HumanMethylation (HM) 450 K array, leading to the characterisation of new pathways.2–4 In contrast, the EWAS study performed in minor salivary glands (MSG) showed modest variations in DNA methylation,2 ,5 while we have recently observed global DNA methylation alterations in salivary gland epithelial cells (SGEC) from patients with pSS leading to, among other perturbations, sicca syndrome (SS) B/La autoantigen deregulation.6 ,7 As the cellular heterogeneity in the MSG biopsies may explain this paradox, long-term cultured SGEC were selected to conduct an EWAS study.
Samples were obtained from 16 individuals who met the revised American-European SS classification criteria,8 and for which clinical characteristics are summarised in table 1. The non-pSS control group patients (n=4) did not meet the aforementioned SS criteria. The institutional review board at the University of Athens approved this study, and participants signed an informed consent form. Non-neoplastic long-term cultured SGEC were established by standard technique as previously described.9
DNA methylation analysis
DNA was extracted from cells (Qiagen, Valencia, California, USA); 500 ng of DNA from each sample was bisulfite-converted (Zymo Research, Irvine, California, USA), and DNA methylation was evaluated by hybridising bisulfite-converted DNA to the HM 450 K BeadChip (Illumina, San Diego, USA). These steps were performed by NXT-DX Company (Gent, Belgium) according to manufacturer's directions. Validation of the array data was performed using bisulfite DNA sequencing in a replicative pSS cohort (n=8) and in the human salivary gland (HSG) cell line cultured for 48 hours with increasing amounts of the DNA demethylating drug 5-azacytidine (5-aza).7
Methylation data processing and analysis
DNA methylation data analysis was performed in the R environment (v3.2.3) using the Minfi (v1.16.1) statistical analysis package. This included correction for background signal and normalisation of each channel to control probes included in each array. Signal intensity was parsed into the Minfi R package quality control and Subset-quantile Within Array Normalization. The methylated and unmethylated probe values were used to calculate both M and β value ratios, and the following formulae were used:
The phenotype (pSS vs control) was defined as categorical, and the dmpFinder function was used (F-test). Based on the recommendations of Du et al,10 M values for each position with a p<0.01 were considered statistically significant. Next, another cut-off was applied, and only probes with a median difference using β value up to 7% were kept for analysis.3 The β value was selected due to its more intuitive biological interpretation. To visualise results, a heatmap was drawn.
For pathway analysis, only genes with at least two differentially methylated CpG (DMC) were considered. Pathway identification was performed considering, respectively, hypomethylated and hypermethylated probes and using the Database for Annotation, Visualization and Integrated Discovery (DAVID V.6.7, https://david.ncifcrf.gov).11 Parameters used in DAVID were a minimum gene group membership of two, a modified Fisher's exact p value (EASE score) maximum of 0.1 and the human genome as background. Moreover, the Interferome database V.2.01 (http://www.interferome.org) was used for the analysis of the possible involvement of significantly differentially methylated interferon-regulated genes.12
To search for epigenetic modifications in SGEC from eight patients with pSS as compared with four controls, we performed an epigenome-wide DNA methylation analysis. CpG sites were selected, as previously done by Miceli-Richard et al.3 Thereby, 4662 DMC sites (DMCs) were characterised (21% hypomethylated and 79% hypermethylated), corresponding to 2560 unique and annotated genes in SGEC from patients with pSS (figure 1A and online supplementary table S1). Among them, the majority of DMCs were present in genomic areas far from promoters (29.5% vs 41% of total probes; p<10−4, χ2 test) and CpG islands (12.4% vs 30.9% of total probes; p<10−4), and within open sea regions (48.8% vs 36% of total probes; p<10−4) and/or intergenic regions (29.1% vs 25% of total probes; p<10−4). Almost identical results were reported in peripheral blood CD19+ B cells from two studies (S1/S2),2 ,3 while differences were observed with CD4+ T cells from patients with pSS (figure 1B).4
supplementary table 1
Next, genes and DMCs obtained in SGEC from patients with pSS were compared with publicly available data from peripheral blood CD19+ B cells (S1/S2 lists pooled) and from CD4+ T cells isolated from patients with pSS.2–4 As represented in figure 2A and online supplementary table S2, 64 genes including 16 annotated to calcium ion binding (GO:0005509; p=8.9×10−9) and only one DMC (cg09152866, androgen-dependent tissue factor pathway inhibitor (TFPI)-regulating protein, ADTRP) were common to the three cell types. With regard to genes found in two cell types, they were 993/2560 (38.8%), and only 209/12 244 (1.7%) when regarding DMCs.
supplementary table 2
supplementary table 5
Thereafter, to identify the common functional characteristics of genes with aberrant DNA methylation, genes with a minimum of two DMCs per gene were considered in SGEC and lymphocytes to perform a pathway analysis (see online supplementary table S3). In SGEC, the calcium (hsa04020, p=0.02) and the Wnt (hsa04310, p=0.004) pathways were identified as enriched gene pathways for hypomethylated and hypermethylated DMCs, respectively. In contrast and when considering lymphocytes, methylated genes were enriched for the T cell receptor (TCR), B cell receptor (BCR) and T1 diabetes pathways in peripheral blood mononuclear cells (PBMCs) and B cells, while for hypomethylated genes, the allograft rejection, tight junction and T1 diabetes pathways were associated within PBMC, B cells and T cells, respectively.
supplementary table 3
Subsequently, as an interferon signature was found to be an important upstream regulator of genes in pSS, we submitted gene lists, with a minimum of two DMCs per gene, to the Interferome database. Compared with the results obtained with lymphocytes from patients with pSS, SGEC from patients with pSS showed a similar and important proportion of differentially methylated interferon-regulated genes (61%) that were related to the Wnt pathway (5/11) but not the calcium pathway.
Since single nucleotide polymorphisms in patients suffering from autoimmune diseases are strongly associated with epigenetic markers in long-range regulatory sequences,13 the overlap between the 43 gene risk factors associated with pSS14 and genes differentially methylated was tested in patients with pSS. In SGEC, 7/43 (16.3%) gene risk factors showed at least one DMC: CXCR5, GTF2I, ICA1, NRLP3, SLC25A40, tumour necrosis factor and MBL2. Among them, five gene risk factors and one DMC (cg04537602, CXCR5) were shared between SGEC and B cells (figure 2B). To validate HM450 K results, we quantitatively analysed DNA methylation patterns of the CpG of interest. Relative to the HSG cell line used as reference, a dose-dependent effect of 5-aza in HSG and variations in SGEC from patients with pSS were statistically confirmed for GTF2I, ICA1, SLC25A40 and NRLP3 (figure 2C).
Finally, patients with pSS were subdivided according to disease activity (ESSDAI ≥5 vs <5), lymphocyte infiltration (Chisholm ≥3 vs <3) and hydroxychloroquine intake. Variations were not found when comparing patient subgroups with the controls (figure 1A), while variations were observed when comparing pSS subgroups (see online supplementary table S4).
supplementary table 4
Our study revealed major epigenetic changes when analysing pure cell populations and long-term cultured SGEC from patients with pSS. We also provide arguments to consider the importance of DNA methylation on interferon-regulated genes in SGEC, which is in accordance with other EWAS studies conducted in lymphocytes. Moreover, the calcium pathway, not previously reported in pSS, was highlighted, and it is known that this pathway induces changes in DNA methylation.15 In addition, an important proportion of genes previously shown to be differentially methylated in T and/or B cells (41.3%) was also found when analysing SGEC, thus suggesting similar pathways between cellular subsets. However, such associations fall below 2% when comparing DMCs with only one DMC detected in the three cell subsets, and present in the 3′ untranslated part of the ADTRP.
When considering genes sensitive to the strong DNA hypomethylating drug 5-aza-2′-deoxycytidine (5-Aza) in the salivary gland cell lines such as HRM3, AQP5, E-cadherin, SSB/La and KRT19,7 ,16–19 we found no statistical difference in the differential methylation in SGEC between patients with pSS and controls. In fact, important individual variations exist between DMCs as addressed for the KRT19 locus in SGEC between patients with pSS,19 and during lifetime development in monozygotic twins20 leading to a negative result when performing EWAS analysis. As a consequence, both individual and cell-type variations may explain the paradoxical overlap observed between the three cell types when testing genes, while such an overlap was not confirmed with regard to DMCs. As a consequence, for MSG methylome analysis in pSS, we recommend using pure cell populations.
Last but not least, the simultaneous genomic–epigenomic analysis also revealed significant associations between pSS-associated genetic risk factors and DMCs in SGEC, suggesting that pSS risk factors have the potential to affect DNA methylation-sensitive pathways in lymphocytes and in SGEC. Now, future studies are necessary in order to characterise precisely the pathways, the cellular specificity, and to determine the functional consequences and the interactions between DMCs and genetic risk factors.
The authors expressed gratitude for comments and editorial assistance to Dr. E. Ballestar (IDIBELL, Spain) and to Dr. W.H. Brooks (University of South Florida, FL). They are also grateful to Simone Forest and Geneviève Michel for their help in typing the paper. Part of the work has been supported by the Research Accounts of the University of Athens (Grant #9988). This work is supported by the Institut Français pour la Recherche Odontologique, the Region Bretagne and the Association Française du Gougerot Sjögren et des syndromes secs.
Handling editor Tore K Kvien
AB and YR are the last two senior authors contributed equally to this work.
Contributors AC, ODK, CLD, CB, AB and YR analysed the data, drafted and revised the paper. EKK, AGT, J-OP and YR initiated the collaborative project and revised the project. ODK, EKK and AGT collected patient material and clinical data.
Funding Research Accounts of the University of Athens (9988).
Competing interests None declared.
Patient consent Obtained.
Ethics approval The institutional review board at the University of Athens.
Provenance and peer review Not commissioned; externally peer reviewed.