Clinical value of DNA methylation markers in autoimmune rheumatic diseases

Abstract

Methylation of cytosine residues in DNA, the best studied epigenetic modification, is associated with gene transcription and nuclear organization, and ultimately the function of a cell. DNA methylation can be influenced by various factors, including changes in neighbouring genomic sites such as those induced by transcription factor binding. The DNA methylation profiles in relevant cell types are altered in most human diseases compared with the healthy state. Given the physical stability of DNA and methylated DNA compared with other epigenetic modifications, DNA methylation is an ideal marker for clinical purposes. However, few DNA methylation-based markers have made it into clinical practice, with the notable exception of some markers used in the field of oncology. Autoimmune rheumatic diseases are genetically complex entities that can vary widely in terms of prognosis, subtypes, progression and treatment responses. Increasing reports showing strong links between DNA methylation profiles and different clinical outcomes and other clinical aspects in autoimmune rheumatic diseases reinforce the usefulness of DNA methylation profiles as novel clinical markers. In this Review, we provide an updated discussion on DNA methylation alterations in autoimmune rheumatic diseases and the advantages and disadvantages of using these markers in clinical practice.

The past decade has seen an increasing awareness of the importance of epigenetic mechanisms in the pathogenesis of various autoimmune rheumatic diseases, and epigenetic markers are emerging as potential clinical biomarkers. Among the various epigenetic modifications, DNA methylation is the best studied (Box 1). Alterations in DNA methylation profiles are linked to the occurrence of autoimmune rheumatic diseases as well as other complex diseases¹. Over the past decade, many different reports have highlighted the widespread occurrence of altered DNA methylation patterns in autoimmune rheumatic diseases, and a wealth of data are now available on DNA methylation alterations that occur in diseases such as systemic lupus erythematosus (SLE)^2,3 and rheumatoid arthritis (RA)^4,5. An increasing amount of data are also available for other conditions such as systemic sclerosis (SSc)⁶ and primary Sjögren syndrome (pSS)⁷.

Box 1 |. Relationship between DNA methylation and transcription.

Epigenetic processes are molecular mechanisms that influence the expression of a DNA sequence and can include DNA methylation, histone post-translational modifications, regulation by non-coding RNAs and nuclear organization. DNA methylation involves the covalent transfer of a methyl group to DNA residues. methylation of cytosines (5mC) is the most common form of DNA methylation and mainly takes place on cytosines followed by guanines (CpG sites). DNA methyltransferases (DNMTs) are the enzymes responsible for de novo or maintenance of DNA methylation. DNA demethylation can occur passively, owing to inefficient maintenance of DNA methylation, or through an active mechanism that involves the activity of ten-eleven translocation (TET) enzymes, thymine DNA glycosylases and base excision repair. Transcription factors are essential regulators of gene expression and are recruited to particular DNA sequences (such as promoters or enhancers) to promote or inhibit transcription of a gene. both the process of cytosine methylation and transcription can influence each other. For example, the most well-known link between these two processes is the association between methylation of CpG islands in gene promoters and transcriptional silencing of the gene. In this scenario, DNA methylation in the gene promoter can block the binding of the transcription factor and influence gene expression. Conversely, several transcription factors can recruit DNMTs and TET enzymes and have a direct effect on the methylation status of CpG sites in the vicinity of the transcription factor binding site. Furthermore, a transcription factor bound to its target sites can indirectly prevent recruitment of DNA methylation-related enzymes. methylation of cytosines in CpG sites in other genomic locations has variable effects on gene transcription depending on whether the modification occurs in enhancers or in other regulatory regions.

Global and sequence-specific changes in other epigenetic modifications, such as post-translational modifications of histones, can also occur in autoimmune rheumatic diseases^8–10. Different nuclear factors and upstream signalling pathways account for sequence-specific targeting of both DNA methylation and histone modifications. Such epigenetic alterations might function well as clinical markers as these changes provide an instant output of environmental effects, including effects that influence the disease activity, the level of systemic inflammation or the effect of a drug in a patient. However, as clinical biomarkers, chemical modifications of DNA have more practical value than histone modifications owing to the physical association between the methylated cytosine and a specific DNA sequence and the chemical stability of this modification. By contrast, histone modifications, despite being associated with specific DNA sequences, occur in a different macromolecule (a DNA-bound protein), which has practical consequences when it comes to sample preparation in the clinical setting. Preserving such a physical association at its precise genomic location requires methods that stabilize the physical interaction between histones and DNA by introducing covalent bonds, such as by using formaldehyde crosslinking. For this reason, DNA methylation analysis is much simpler and more convenient than the study of histone modifications in clinical practice.

In the past few years, different research groups have shown that various DNA methylation alterations in autoimmune rheumatic disease are associated with disease activity, disease subtypes, different cell types and physical locations in lesions and response to different drugs. Research efforts using large patient cohorts are leading to the identification of DNA methylation-based markers that can ultimately benefit patients and improve personalized treatments. In fact, in the field of oncology, several DNA methylation markers are already used for clinical purposes, for instance as therapeutic predictors¹¹. The fact that DNA methylation changes are reversible also make them a relevant target for therapeutic intervention, particularly in immune-oncology¹².

In this Review, we discuss the latest evidence concerning DNA methylation-related alterations in autoimmune rheumatic diseases and their relationship with different clinical aspects that support their relevance as clinical markers. We hope that this discussion will encourage further investigations to identify and validate novel and more robust clinical markers.

DNA methylation variability

Patterns of DNA methylation can vary between individuals, and these varying patterns can be caused by genetic or environmental factors. Approximately 20% of inter-individual variation in DNA methylation patterns have been attributed to differences in DNA segments or genes^13,14. Polymorphisms (or mutations) in promoters and enhancers can directly or indirectly affect the deposition or removal of methyl groups on DNA (Fig. 1a). Those genetic variants that are linked to an increased risk of an autoimmune disease might influence the phenotype of an individual through altering the levels of DNA methylation at regulatory regions of target genes^15,16. The existence of strong associations between genetic variants and DNA methylation patterns, known as methylation quantitative trait loci, across a variety of tissues strongly supports such a causal relationship^17–20. Epigenetic patterns can vary between populations^21–24, with genetic ancestry explaining most of the variation in methylation between ancestral groups^24,25. Environmental cues, such as increased levels of inflammatory cytokines and activation of certain pathways following viral or bacterial infection, can also modify DNA methylation patterns^5,26 (Fig. 1b). Dietary factors, such as the uptake of folates²⁷, can influence DNA methylation. Tissue-specific DNA methylation levels strongly correlate with chronological age^23,28 and sex²⁹, which is particularly relevant in autoimmune diseases that have a strong sex bias. In fact, the higher frequency of autoimmune diseases in females than in males might be explained in part by differences in DNA methylation of the X chromosome that are determined in a parent-of-origin manner³⁰. Some changes in DNA methylation can influence the expression levels of a given gene, and therefore have a functional effect, whereas other changes do not have functional effects but merely provide traces of a genetically or environmentally driven alteration in an upstream pathway or factor. In both cases, the analysis of DNA methylation patterns can be used to infer mechanisms and pathways and to gain insightful information on the state of a cell.

Fig. 1 |. Genetic and environment influences on DNA methylation.

a | A single nucleotide polymorphism (SNP) can influence DNA methylation through various mechanisms. This figure provides some examples of proximal regulation (when the SNP is located near the transcription start site) or distal regulation (when the SNP is located in a distant enhancer). For these examples, two alternative alleles are shown: one allele prevents the binding of the transcription factor (SNP A), and the other allele enables the binding of the transcription factor (SNP B). The transcription factor can recruit enzymes involved in DNA demethylation (such as ten-eleven translocation 2 (TET2)) or DNA methylation (such as DNA methyltransferase 3A (DNMT3A)), which can affect the methylation status of the promoter and ultimately the transcriptional activity of the gene. b | The environment, herein represented by increased levels of inflammatory cytokines, can activate signalling cascades and downstream transcription factors, which can recruit enzymes involves in DNA methylation (such as TET2, shown here) or DNA demethylation (such as DNMT3A), affecting the methylation status of the gene promoter and transcriptional activity. CTCF, CCCTC-binding factor; TF, transcription factor.

Genetic predisposition does not necessarily lead to autoimmune disease development, as exemplified by the discordant occurrence of autoimmune disease in pairs of genetically identical monozygotic twins³¹, reflecting the important influence of environmental cues on disease development. Studies in twins are powerful models to elucidate epigenetic modifications resulting from gene–environment interactions. In fact, the first high-throughput study on DNA methylation in autoimmune disease was performed by studying SLE in monozygotic twins², and other studies of autoimmune diseases have used such an approach^32–35, some of which will be discussed in later sections of this Review.

Hence, DNA methylation variability is an important consideration when developing DNA methylation-based clinical biomarkers. Understanding the mechanisms underlying such variability is not only relevant for the identification of potential mechanisms responsible for pathogenic DNA methylation alterations but also for the optimal design of studies to minimize confounding factors, such as age, sex and genetic ancestry.

Insights from DNA methylation studies

Cell-type-specific DNA methylation

A critical aspect to consider for DNA methylation, as well as other epigenetic modifications, is the cell type specificity of this modification. Different cell types have unique DNA methylomes, which are in part the result of the activity of cell-specific signalling pathways and downstream transcription factors that shape the deposition of methyl groups along the DNA sequence (see Box 1). Cell-specific methylomes are associated with specific transcriptomes, which are ultimately linked to cell identity and function. In fact, alterations in DNA methylation can lead to, or are associated with, a pathological cellular behaviour³⁶.

In the past decade, studies have not only confirmed the relevance of DNA methylation patterns to disease development but also have shed light on their contribution to the pathogenic behaviour of different immune cells, including B cells, T cells and monocytes, as well as specific cell types relevant to some diseases such as synovial fibroblasts in RA⁹ and salivary gland epithelial cells in pSS^35,37, as discussed in the next section.

T cells and B cells.

Most research efforts on the effect of DNA methylation in autoimmune rheumatic disease have focused on lymphocytes and in particular on T cells. DNA methylation alterations in these cell types are associated with exacerbated cellular responses, as well as the generation of inflammation. The majority of studies in T cells have focused on CD4⁺ T cells. For example, initial investigations of CD4⁺ T cells from patients with SLE showed that DNA hypomethylation in the promoter and enhancer regions of ITGAL³⁸, TNFSF7 (REF.³⁹) and CD40L⁴⁰ were associated with T cell hyperactivation. In the past 5 years, regulatory T cells (T_reg cells) have received much attention in the context of DNA methylation studies in autoimmune diseases. One study found that the reduced expression of the transcription factor FOXP3 in collagen-induced arthritis, a model of experimental arthritis generally associated with chronic inflammation, is dependent on reduced activity of TNF receptor 2 (REF.⁴¹). This receptor signals to promote FOXP3 expression by restricting DNA methylation on the FOXP3 promoter. During inflammation, loss of TNF receptor 2 resulted in decreased FOXP3 expression and pathogenic conversion of the T_reg cells into a proinflammatory T helper 17 (T_H17) cell-like phenotype. A relationship between decreased DNA methylation of the FOXP3 promoter and autoimmunity has also been shown in other autoimmune rheumatic diseases. For instance, in CD4⁺ T cells from patients with SSc, treatment with all-trans retinoic acid, a natural derivative of vitamin A, increases the expression of FOXP3, and subsequently the proportion of T_reg cells, by promoting demethylation of the FOXP3 promoter⁴².

Pharmacological induction of DNA demethylation as a way to experimentally induce autoreactivity was initially reported by Richardson⁴³. In this initial study, treatment of T cells with a DNA methyltransferase (DNMT) inhibitor, 5-azacytidine, removed the requirement of an antigen for the T cells to become activated, with T cell activation only requiring the presence of autologous macrophages. Subsequent studies by this research team and others led to the notion that drug-mediated DNA demethylation can induce T cell abnormalities similar to those occurring in SLE⁴⁴. In 2018, one group of researchers proposed that cell-specific inhibition of DNA methylation might actually ameliorate disease in lupus-prone MRL/lpr mice. In MRL/lpr mice with established disease, targeted delivery of 5-azacytidine to either CD4⁺ T cells or CD8⁺ T cells using nanolipogel microspheres decreased various manifestations of disease, including proteinuria and kidney pathology⁴⁵. Specifically, targeted delivery of 5-azacytidine to CD4⁺ T cells promoted the expansion and function of FOXP3⁺ T_reg cells, whereas delivery to CD8⁺ T cells maintained or increased the cytotoxic activity of the CD8⁺ T cells (thought to be important for controlling autoimmunity) while restraining their conversion to pathogenic TCRαβ⁺CD4⁻CD8⁻ double-negative T cells⁴⁵. The apparent contradiction of these results compared with prior findings could be explained by the coexistence of aberrant hypomethylation and hypermethylation patterns in SLE.

The increased autoreactivity of CD4⁺ T cells in SLE has been linked to upregulated expression of the transcription factor RFX1 (REF.⁴⁶). This transcription factor binds to the promoters of CD11A and CD70 and recruits DNMT1, histone deacetylase 1 (HDAC1) and the histone methyltransferase SUV39H1, resulting in downregulated expression of both genes. These two genes are associated with autoreactive responses and overstimulation of IgG synthesis in B cells. Reduced expression of RFX1 could explain the aberrant overexpression of CD11A and CD70 in T cells from patients with SLE⁴⁶. In mice, conditional deletion of Rfx1 increases the severity of pristane-induced lupus and promotes the induction of T_H17 cells. RFX1-deficient naive CD4⁺ T cells from the mice have an increased propensity to differentiate into T_H17 cells in vitro compared with RFX1-sufficient naive CD4⁺ T cells, an effect that can be reversed by restoring the expression of RFX1 using a gene vector. Notably, the data highlighted a non-canonical IL-6–STAT3 signalling pathway that regulates T_H17 cell differentiation by inhibiting RFX1 expression⁸. An additional epigenetic mechanism involved in the regulation of IL-2 and IL-17A production by CD4⁺ T cells involves the regulation of the cAMP response element modulator (CREM). In CD4⁺ T cells from patients with SLE, CREM is overexpressed owing to a SET domain containing 1 (SET1)-dependent increase in the amount of histone H3 trimethylated at lysine 4 (H3K4me3; an activating histone mark) within the gene promoter of CREM. This histone mark prevents the binding of DNMT3A and subsequent DNA methylation within this region, promoting CREM expression. Overexpression of CREMα results in decreased IL-2 production and increased IL-17A production by CD4⁺ T cells, leading to a pro-inflammatory cell phenotype⁴⁷.

DNA methylation in CD8⁺ T cells has also been widely studied in the context of SLE. Compared with CD8⁺ T cells from healthy individuals, CD8⁺ T cells from patients with SLE contain hypomethylated CpG sites in HLA-DRB1, as well as in many genes involved in type I interferon responses, such as STAT1 (REF.⁴⁸), that are upregulated in SLE. Hence, CD8⁺ T cells could be epigenetically primed at the DNA methylation level to respond to type I interferons.

In addition to its function in CD4⁺ T cells, CREM has also been implicated in the expansion of a rare subpopulation of TCRαβ⁺CD3⁺CD4⁻CD8⁻ T cells (known as double-negative T cells) in SLE⁴⁹. The underlying mechanism is thought to involve CREM-mediated recruitment of DNMT3A and histone methyltransferase G9a to the CD8 cluster, chromatin remodelling and subsequent epigenetic silencing of CD8A and CD8B.

B cells are important contributors to the majority of autoimmune diseases, as evidenced by the characteristic presence of autoantibodies in autoimmune rheumatic diseases and the success of B cell-depleting therapies in some diseases⁵⁰. Various studies of B cells in different diseases, including in RA, SLE and pSS, have implicated DNA methylation alterations in the dysregulated expression of many critical genes in these cells. For example, in CD19⁺ B cells, many relevant genes are differentially methylated in patients with RA compared to healthy individuals⁵¹. These genes include CD1C, TNFSF10, PARVG, NID1, DHRS12, ITPK1, ACSF3 and TNFRSF13C, all of which were identified in a discovery cohort of patients with RA and validated in an independent cohort⁵¹. Some of these genes were also differentially methylated in patients with SLE, suggesting that some epigenetic alterations, and potentially some underlying epigenetic mechanisms, are shared across these two autoimmune diseases that could perhaps be targeted with similar therapeutic approaches⁵¹. Dysregulated DNA methylation in B cells has also been implicated in pSS. In CD19⁺ B cells, interferon-regulated genes (such as MX1, IFI44L and PARP9) are hypomethylated in patients with pSS compared with healthy individuals, a feature associated with increased expression of these genes⁷.

Monocytes.

Compared with lymphocytes, the DNA methylation profiles of monocytes during autoimmunity have been studied less frequently; however, the important contribution of monocytes to local and systemic inflammation highlights the important relevance of this line of research. In one genome-wide DNA methylation analysis of CD4⁺ T cells, CD19⁺ B cells and CD14⁺ monocytes in SLE, the researchers found that all three cell types had an aberrant DNA methylation signature, including hypomethylation of genes involved in type I interferon signalling⁵². Various other studies have shown that the DNA methylome of peripheral blood monocytes is altered in patients with RA compared with healthy individuals^4,5, which is partly caused by a disease-associated increase in pro-inflammatory cytokines⁵. The DNA methylation profile of monocytes is also altered in SLE. An analysis of monocytes, as well as of B cells and T cells, from pairs of monozygotic or dizygotic twins discordant for SLE found that various interferon-regulated genes contained differentially methylated CpGs in the affected twin compared with the unaffected twin³². Although this study found that all the major immune cell compartments shared this interferon signature, the analysis also identified distinct cell population-specific gene sets, highlighting the importance of studying different immune cell types in epigenetic studies.

Synovial fibroblasts.

Synovial fibroblasts have an altered DNA methylation pattern in RA (reviewed elsewhere⁵³) that is associated with various pathogenic features of the cells. Phenotypic changes in these cells promote joint damage and destruction in RA. In an effort to understand these changes, researchers have comprehensively characterized the epigenome of synovial fibroblasts in RA, including analysing the genome-wide DNA methylation patterns, as well as genome-wide histone modifications, chromatin accessibility and RNA expression⁹. This analysis identified a number of expected pathways that were epigenetically dysregulated in synovial fibroblasts in RA, including pathways related to immune responses or matrix regulation. Some novel and unexpected pathways were also uncovered, such as the ‘Huntington disease signalling’ pathway, which included genes encoding phosphatidylinositide 3-kinases, activator protein 2, histone deacetylases and heat shock protein 70. These differentially modified genes could serve as potential therapeutic targets.

Location-specific DNA methylation

The DNA methylome is highly plastic and can provide information on the surrounding environment, which is of particular relevance for joint-residing cells such as tissue-resident macrophages or synovial fibroblasts, especially in the context of an autoimmune rheumatic disease. For example, levels of cytokine are increased in the inflamed joint and can directly influence the DNA methylome of different immune cell types⁵, as well as modulate the phenotypes acquired during their terminal differentiation to macrophages or other effector cells^54,55. RA and osteoarthritis are characterized by distinctive topographical patterns of joint involvement⁵⁶, affecting the small joints, such as the joints of the hands and feet, and the proximal joints, such as the knees and shoulders, to differing extents. Emerging data suggest that the DNA methylome of joint cells similarly has joint-specific patterns, which might reflect the surrounding environment and/or developmental differences. For example, the DNA methylome and transcriptomes of synovial fibroblasts can differ depending on the anatomical location⁵⁶. The existence of distinct DNA methylation signatures of synovial fibroblasts associated with different locations provides mechanistic insights into their transcriptional and functional diversity and, most importantly, their differential progression and therapeutic response.

Macrophages and monocytes, as well as other immune cell types, probably also have divergent DNA methylomes depending on their spatial location. For instance, the environment in the inflamed joints in RA probably alters the DNA methylome, transcriptome and phenotype of macrophages, so that their profiles differ from cells residing in the peripheral blood. This idea is supported by data showing that the transcriptional profiles of synovial monocytes are heterogeneous and that certain transcriptionally distinct monocyte subsets, activated by particular inflammatory cytokines and interferons, are enriched in RA⁵⁷. The presence of different combinations of cytokines in different joints might result in a wide range of monocyte phenotypes. Other tissues that are damaged in autoimmune rheumatic diseases, such as the kidneys or even the skin, also likely influence the methylome of residing cells. Understanding the specific methylation profiles in these locations should improve specific or localized treatments. Hence, anatomical location is a critical factor in disease development, and the differences in local microenvironments determine the DNA methylation and transcriptional signature of cells at these locations, the understanding of which is necessary for optimizing diagnosis, treatment and prognosis of diseases such as RA and osteoarthritis.

Overlapping disease mechanisms

For years, data from genetic studies have supported the notion that various disease pathways and clinical features are common across different autoimmune diseases. In fact, many susceptibility genes are shared among different diseases, including genes identified by genome-wide association studies and meta-analysis of genome-wide association studies^58–60. Similarly, DNA methylation analyses have revealed common pathways among different autoimmune diseases. Given the links between DNA methylation signatures and external influences that promote disease-associated pathways, DNA methylation analyses might be well positioned to help delineate the different contributions of such pathways.

As an example, SLE and Sjögren’s syndrome were initially considered as two forms of the same disease owing to their clinical similarities and coexistence, rather than the two different clinical entities they are viewed as today. Notably, a cross-comparative analysis of DNA methylation profiles in patients with SLE or pSS and healthy individuals identified the presence of widespread shared signatures between the two diseases, as well as a limited amount of disease-specific DNA methylation patterns. These data support the idea that SLE and Sjögren syndrome have largely similar epigenetic landscapes and share similar underlying disease mechanisms⁶¹. The presence of rare alterations that are specific to either disease could reflect disease-specific pathways and clinical manifestations. For example, further analysis of the SLE-specific DNA methylation patterns revealed an enrichment in certain pathways, including pathways involved in haemostasis and innate immune responses, as well as genes involved in apoptotic responses and NF-κB activation. Given that SLE is associated with an increased rate of apoptosis, unlike Sjögren syndrome, such epigenetic changes might be involved in pathogenesis pathways unique to SLE⁶¹.

Mixed connective tissue disease is a rare, complex, systemic autoimmune disease that has clinical features that resemble SSc, RA and SLE. Indeed, DNA methylation profiling analysis has revealed the existence of a common type I interferon-regulated epigenetic signature in these clinical entities. Notably, such analysis has also identified genes and pathways that are uniquely involved in mixed connective tissue disease, which could help improve the diagnosis and treatment of this disease⁶².

Overall, DNA methylation signatures are helping to uncover a multiplicity of connections between different cell types and diseases and shared pathways that are aberrantly activated during inflammatory responses. The contribution of some of these changes probably varies according to the stage of disease.

DNA methylation as a clinical marker

Various DNA methylation alterations are associated with clinical outcomes and other clinical aspects in autoimmune diseases and hence have potential as clinical biomarkers (Fig. 2; TABLE 1). The participation of different signalling and metabolic pathways and cell–cell interactions during the course of a disease can ultimately affect DNA methylation-mediated mechanisms; therefore, DNA methylation can be a direct readout of such events, which can serve a number of purposes.

Fig. 2 |. Clinical applications for DNA methylation markers in autoimmune rheumatic disease.

Measuring the DNA methylation profile of patients can provide useful information in multiple clinical scenarios. DNA methylation analysis can uncover insights into the pathogenesis of autoimmune rheumatic diseases and inform treatment strategies. For example, DNA methylation of different joints can be relevant for understanding the differences between healthy and inflamed joints. DNA methylation analysis can also be useful in the diagnosis of patients and can help classify patients into clinically useful subgroups, providing important information prior to treatment. DNA methylation markers can be used to monitor disease activity and predict good or poor responses (shown as green and red arrows, respectively) to first-line conventional synthetic DMARDs or biologic DMARDs.

Table 1 |.

Potential DNA methylation markers in autoimmune rheumatic diseases

Disease	Cell type	Main findings	Ref
Markers for disease activity and progression
SLE	Whole blood	Hypomethylation of two CpG sites in the IF/44L promoter were associated with high disease activity	66
	CD4+ T cells, monocytes and granulocytes	SLE was associated with hypomethylation in interferon-regulated genes in all cell types; this study was performed on twins discordant for SLE, and the effect was more pronounced when the affected twin had experienced a disease flare in the past 2 years	32
	CD4+T cells	Disease activity (as assessed by the SLEDAI score) was associated with hypomethylation of genes involved in T cell activation and differentiation, and hypermethylation in genes involved in inhibitory signalling pathways such as TCFβ signalling	64
	PBMCs	Disease activity (as assessed by the SLEDAI score) was associated with hypomethylation of interferon-regulated genes	65
RA	CD14+ monocytes, CD19+ B cells, memory CD4+ T cells and naive CD4+ Tcells	Hypomethylation of CpG sites in the promoters of CYP2E1 and DUSP22 were associated with active and erosive disease, respectively	4
	Monocytes	A cluster of several thousand CpG sites was associated with disease activity (as assessed by DAS28 score); researchers developed a DNA methylation-based formula for predicting disease activity in RA, which is potentially applicable in other chronic inflammatory diseases	5
Markers for disease subtype
SLE	Whole blood	Hypomethylation at two CpG sites within the IFI44L promoter was associated with renal disease in SLE	66
	CD4+T cells	Differential methylation of particular genes was associated with particular cutaneous manifestations	71
		Particular DNA methylation alterations correlated with particular clinical phenotypes (that is, cutaneous manifestations without systemic pathology, cutaneous manifestations with renal disease or cutaneous manifestations with renal and polyarticular disease)	3
RA	Whole blood	Differentially methylated region in the promoter of PCDHB14 was associated with the onset of ACPA positivity in RA	33
SSc	Whole blood	Methylation of 153 and 266 distinct CpG sites was associated with limited cutaneous SSc and diffuse cutaneous SSc, respectively	35
Markers for spatial localization
RA	Synovialfibroblasts	Identification of distinctive pathogenic epigenetic signatures of synovial fibroblasts in joints at different locations	56
Markers for drug response
RA	Whole blood	The methylation status of four CpG sites following 4 weeks of methotrexate therapy correlated with improvement in disease activity after 6 months of therapy	76
		A higher baseline level of global DNA methylation was associated with a reduced response to methotrexate	77
		Five CpG sites were differentially methylated between patients with a good response and patients with no response to 3 months of anti-TNF therapy	79
	T cells	The pre-treatment methylation levels of various CpG sites differed between patients with a good response and patients with a poor response to 6 months of DMARD therapy	78

ACPA, anti-citrullinated peptide antibody; DAS28, 28-joint disease activity score; PBMCs, peripheral blood mononuclear cells; RA, rheumatoid arthritis; SLE, systemic lupus erythematosus; SLEDAI, Systemic Lupus Erythematosus Disease Activity Index; SSc, systemic sclerosis.

Disease activity and progression

Autoimmune diseases are characterized by periods of high activity or flares and periods of remission, the latter of which is an important therapeutic goal. For some diseases such as RA and SLE, indexes for measuring disease activity are available that enable the monitoring and scoring of treatment responses. However, for other diseases such as SSc, disease activity scores are insufficient or unavailable and/or the main manifestations of the disease are difficult to precisely score. Hence, robust biomarkers of disease activity or progression are still needed. Initially, the apparently stable nature of DNA methylation profiles over long periods suggested that such profiles would not necessarily reflect the current status of a cell or be used as a measurement of the current disease activity status of a patient. However, emerging data suggest that DNA methylation has a fast turnover and is responsive to extracellular and/or environmental changes^54,63 and so could provide a useful measurement of the status of cells and of disease.

Data from twins discordant for SLE suggest that hypomethylation of interferon‐regulated genes is a robust feature of SLE that is present across various cell types, including CD4⁺ T cells, CD19⁺ B cells, monocytes and granulocytes³². Notably, in this study, hypomethylation was more pronounced when the affected twin had experienced a disease flare within the past 2 years, which suggests that this epigenetic signature is somewhat stable but tends to fade gradually. Other studies have also reported a link between SLE activity and DNA methylation in CD4⁺ T cells⁶⁴ or in peripheral blood mononuclear cells⁶⁵. Notably, one analysis of peripheral whole blood cells from patients with active SLE, patients with SLE who were in remission and healthy individuals found that the levels of methylation at two CpG sites within the IFI44L promoter could help distinguish patients with SLE from healthy individuals with a high sensitivity and specificity, and also had promise as a marker for monitoring changes in disease activity⁶⁶.

Various studies have investigated the relationship between disease activity and various DNA methylation patterns in RA in an effort to identify potential markers of disease activity. In peripheral blood samples from patients with RA, hypomethylated regions in the promoters of CYP2E1 and DUSP22 were associated with active and erosive disease⁴, respectively, suggesting that these alterations could function as biomarkers of disease progression and severity. In another study of peripheral blood monocytes from patients with RA, the methylation levels of a cluster of several thousand CpG sites correlated with the 28-joint disease activity scores of the patients⁵. Such disease-associated changes seem to be induced by several pro-inflammatory cytokines known to be increased during the active stages of disease (such as TNF, IFNα and IFNγ)⁵. On the basis of these findings, the researchers developed a formula for predicting disease activity, which could have clinical utility for RA as well as for other chronic autoimmune diseases that lack a sufficiently accurate disease activity score and that are also characterized by flares with high levels of pro-inflammatory cytokines.

The highly variable DNA methylation profiles of patients with RA or other conditions³⁴ is thought to reflect changes in the local environment of these cells, such as increased levels of cytokines and induction of oxidative stress⁵, and, as such, DNA methylation profiles provide a direct marker of the disease status of the patient as well as the response pathways activated during the disease course.

Disease subtyping

Autoimmune rheumatic diseases are complex multi-system diseases with various heterogeneous manifestations. For example, many patients with SLE have renal disease (up to 60% of patients with SLE), but not every patient with SLE develops renal complications⁶⁷. For patients with undifferentiated arthritis, the disease can remain undifferentiated for years or can eventually evolve into RA or psoriatic arthritis⁶⁸. One of the aims of modern medical practice is to provide personalized approaches to treatment. This strategy requires the availability of tools that can classify patients into clinically useful subgroups. Biomarkers can help identify disease subtypes for optimized therapy and can also help predict disease progression and treatment responses of individual patients. DNA methylation patterns are sensitive enough to provide accurate markers that reflect biological differences between individuals and so could help differentiate patients into clinically useful subgroups.

Rheumatoid arthritis.

In RA, the presence of disease-specific autoantibodies, including anti-citrullinated peptide antibodies (ACPAs), can be detected years before the onset of clinical disease⁶⁹. Various data support the notion that ACPA-positive and ACPA-negative RA are in fact two distinct disease subsets that differ in severity and response to treatments⁷⁰. To better understand the development of ACPA-positive RA, and the temporal contributions of epigenetic factors, one study investigated the DNA methylation profiles of two sets of twin pairs: five healthy monozygotic twin pairs discordant for ACPA positivity and seven monozygotic twin pairs discordant for ACPA-positive RA³³. Given the inherent limitations associated with twin studies of small size, they developed a method to improve the specificity and minimize the number of false positives, which enabled the prioritization and identification of candidate biomarkers of relevance. Using this approach, these researchers identified a differentially methylated region associated with the onset of ACPA-positive RA, which was located in the promoter of PCDHB14. Future studies with large cohorts including both ACPA-positive and ACPA-negative patients with RA should help delineate the DNA methylation differences between these two clinical entities in the pathway towards personalized treatments.

Systemic lupus erythematosus.

In SLE, alterations in the methylation status of particular genes are associated with certain cutaneous manifestations, suggesting the existence of unique epigenetic susceptibility loci for these types of manifestations. Manifestation-specific methylation patterns could be linked to different environmental cues and the activation of distinctive pathways, which might flag the need for different treatment strategies. For example, in one analysis of CD4⁺ T cells from patients with history of malar rash, discoid rash or no cutaneous manifestations, hypomethylation of MIR886 and TRIM69 and hypermethylation of RNF39 were associated with malar rash whereas hypomethylation of RHOJ was associated with discoid rash⁷¹. Discoid rash was also associated with hypomethylation in genes involved in antigen processing and presentation such as TAP1 and PSMB8. Notably, network analyses of hypomethylated regions identified an enrichment in genes involved in TAP-dependent processing and MHC class I antigen presentation for both types of cutaneous manifestations. These findings aid in our understanding of the pathogenesis of cutaneous manifestations in SLE and might pave the way for potential therapeutic strategies. One genome-wide comparison of the DNA methylation patterns of patients with various SLE phenotypes (that is, cutaneous manifestations alone in the absence of systemic pathology, cutaneous manifestations together with renal disease, and cutaneous manifestations together with renal and polyarticular disease) not only revealed the presence of common methylation changes across all the SLE phenotypes compared with the DNA methylation pattern of healthy individuals, but also specific DNA methylation changes that correlated with each clinical phenotype³. Gene ontology analysis of the differentially methylated genes identified enrichment in pathways relating to leukocyte extravasation signalling and mitochondrial dysfunction in patients with SLE and renal disease, implicating these two pathways in SLE-associated renal pathology. In another study, the levels of DNA methylation in the IFI44L promoter were much lower in patients with SLE with renal involvement than in patients with SLE without renal involvement⁶⁶, consistent with results from another study showing more robust hypomethylation at interferon-regulated genes in patients with SLE with renal involvement than in patients with SLE without renal involvement⁷².

SLE-specific methylation patterns can also vary across different ethnicities. For example, one analysis identified differences in the methylation patterns of African–American patients and European–American patients in interferon-regulated genes and EBF1 binding sites⁷³, confirming an earlier report also suggesting the presence of ethnicity-specific DNA methylation changes in SLE⁷⁴. Hence, ethnicity is an important consideration for personalized treatment in SLE.

Systemic sclerosis.

SSc can be divided into two subtypes on the basis of the extent of skin involvement: limited cutaneous SSc (lcSSc) and diffuse cutaneous SSc (dcSSc). Alterations in DNA methylation profiles can differ between these two subtypes. For example, in a comparison of the DNA methylation profiles of 27 pairs of twins discordant for SSc, each disease subtype had a distinct DNA methylation pattern³⁵. Specifically, the researchers identified 153 unique differentially methylated CpG sites in lcSSc and 266 distinct sites in dcSSc. The negligible overlap and occurrence of distinct epigenetic landscapes in each disease subset was consistent with a previous epigenome-wide association study of dermal fibroblasts in SSc⁷⁵. In this report, in which the researchers compared the methylation profile of patients with dcSSc, patients with lcSSc and age-matched, sex-matched and ethnicity-matched healthy individuals, only 203 CpG sites were commonly differentially methylated in both patients with dcSSc and in patients with lcSSc (representing 118 hypomethylated and six hypermethylated genes), whereas a total of 3,528 sites were differentially methylated in patients with either SSc subtype compared with healthy individuals.

Epigenetic studies in SSc are still limited, in part because of the rarity of this condition. The examples presented herein support the existence of SSc subtype-specific differences and highlight the need for studies with larger cohorts and collaborative efforts to increase the strength of these findings and validate these and other markers.

Predicting drug response

The identification of markers that can predict drug responses should not only benefits patients but should also help optimize the use of resources by public health systems. Emerging evidence suggests that DNA methylation could be used as a predictive marker for patients with rheumatic diseases, in particular for patients with RA. For instance, methotrexate is the first-line conventional synthetic DMARD for controlling active inflammation in patients with RA. However, ~40% of patients do not respond to treatment with methotrexate or cannot tolerate the drug. Given the relative ease of collecting whole blood samples, many studies have assessed the value of DNA methylation levels in the blood for predicting drug responses. In one such study, which characterized the methylation profiles of blood samples from patients with a good therapeutic response to methotrexate and patients with a poor therapeutic response to methotrexate by 6 months⁷⁶, the methylation profiles of the responders and non-responders were indistinguishable prior to treatment. However, following 4 weeks of treatment, the methylation status of four CpG sites correlated with improvement in disease activity at 6 months⁷⁶. Although these DNA methylation markers alone are not sufficient to be of clinical use in predicting drug responses, these data suggest that these markers could be incorporated into algorithms that also include clinical and genetic data to provide a more robust predictor of the treatment response. In another study of patients with early RA, higher baseline levels of global DNA methylation in the blood were associated with a decreased clinical response to methotrexate, as measured by the reduction in 28-joint disease activity score following treatment initiation⁷⁷, suggesting that the baseline DNA methylation level could have some utility as a predictor of response to methotrexate.

In addition to investigating the utility of DNA methylation levels in whole blood, some investigators have also studied the utility of measuring the DNA methylation profiles of circulating T cells for predicting treatment responses. In one such study of patients with early RA, the methylation status of 21 CpG sites in T cells prior to treatment were predictive of the subsequent therapeutic response to conventional synthetic DMARD therapy; these sites were differentially methylated between patients with a good therapeutic response and patients with a poor therapeutic response after 6 months of therapy⁷⁸.

As well as predicting responses to conventional synthetic DMARDs, the ability of DNA methylation profiles to predict responses to biologic drugs, such as TNF inhibitors (including etanercept), has also been assessed in RA. For example, in a comparison of blood samples from patients with a good response or a poor response to etanercept therapy at 3 months, five CpG sites were differentially methylated between the two groups prior to treatment, including two CpG sites that mapped to LRPAP1 (REF.⁷⁹). In an analysis of methylation quantitative trait loci, three single nucleotide polymorphisms correlated with the methylation levels at either of the two CpG sites on LRPAP1. In particular, an allele of one of these single nucleotide polymorphisms (the A allele of rs3468) correlated with higher levels of methylation at both sites and non-response to treatment in both the discovery cohort and in an independent replication cohort. Hence, DNA methylation marks on this gene have potential as biomarkers for predicting responses to anti-TNF therapy.

With the increasing number and variety of available biologic drugs that have different specificities and high costs, promptly identifying which patients will respond well to which drugs is important. The use of isolated cell types and bigger cohort sizes should provide further insights on the utility of DNA methylation signatures for predicting drug responses.

Future directions

A wide range of high-throughput methods for analysing DNA methylation in bulk are now available (Box 2). The availability of single-cell technologies is providing new challenges and opportunities for studying DNA methylation and transcriptional patterns in autoimmune rheumatic diseases, which could help in the understanding of pathogenesis and identify new clinical markers. For example, the incorporation of single-cell multi-omic methods⁸⁰, including single-cell methylomics, should help identify markers associated with rare pathogenic subpopulations that are difficult to distinguish using standard cell sorting-based purification protocols. For instance, single-cell transcriptomic analysis has already revealed the presence of a type I interferon and fibrotic signature in tubular cells and keratinocytes in lupus nephritis, which served as potential prognostic markers of poor response to treatment⁸¹. In this sense, single-cell methylation analysis would be useful to identify better and more robust markers for such rare cell populations. Future approaches should also consider assessing cell–cell interactions, which can be inferred from single-cell transcriptomics using tools such as CellPhoneDB⁸², given the importance of immune cell–cell communication in determining immune responses.

Box 2 |. Common methods for analysing bulk DNA methylation in human samples.

Different methods for characterizing the DNA methylation profiles of human cells are available and compatible with routine clinical use. In a community-wide benchmarking study that compared the performance of various DNA methylation assays⁸⁶, most of the assays were accurate and reproducible across laboratories, suggesting that these assays are well suited for clinical applications. For discovery studies, methods that have a high coverage of the genome, such as whole-genome bisulfite sequencing, should be considered and can be especially useful for analysing well-defined small cohorts or studies involving a few pairs of monozygotic twins. Whole-genome bisulfite sequencing protocols are available that require only small amounts of input DNA⁸⁷, which enables the analysis of rare cell populations. For larger discovery studies, with vast cohort sizes, methods such as reduced representation bisulfite sequencing⁸⁸ or DNA methylation bead arrays can be convenient and cost-effective. bisulfite amplicon sequencing and bisulfite pyrosequencing are useful methods for validating DNA methylation patterns, initially identified in discovery cohorts, using larger patient cohorts.

A limitation of several of the aforementioned studies is the small size of the cohorts used. Hence, additional studies that incorporate large cohorts of patients with various autoimmune rheumatic diseases are needed to better characterize disease subtypes and to predict responses to drugs. In addition to increasing the size of the cohorts, it would be relevant to focus on specific cell types instead of mixed populations. However, the high costs and technical limitations of cell sorting and single-cell technologies can limit the cohort sizes of studies. As a way to avoid this issue, researchers have developed a methodology for identifying novel cell type-specific associations between methylation patterns and disease status (in this case, RA status) using bulk methylation data⁸³. Such a methodology could be useful for analysing the large number of bulk methylation data already available and for extracting useful information from these data; however, the application of this method requires knowledge of the proportions of each cell type, which is often unknown. Paradoxically, the experimental costs of single-cell approaches can actually sometimes be lower than the costs of bulk experiments on sorted cells. For example, bulk RNA sequencing requires the generation of a library for each sorted cell population, whereas one single-cell RNA sequencing library contains all the information needed⁸⁴. Sample multiplexing and multimodal approaches can reduce the costs of single-cell approaches. Furthermore, an increasing number of protocols enable the simultaneous profiling of DNA methylation and other measurements from the same cell, further reducing costs. For example, different combinatorial methodologies can simultaneously generate single-cell data relating to DNA methylation, transcriptomics, chromatin accessibility and genotype⁸⁵.

Conclusions

Emerging data suggest that a wide range of alterations in DNA methylation are linked to a diversity of factors including genetic predisposition (genetic–epigenetic links) and dysregulation in epigenetic regulators, such as those related to the downstream effects of extracellular pro-inflammatory signalling. Notably, various studies have highlighted a relationship between DNA methylation alterations and various clinical outcomes or features (such as disease subtypes, response to drugs and disease activity). All these findings help elucidate the contribution of DNA methylation-mediated mechanisms in the acquisition of pathological cell behaviours, including that of B cells, T cells, monocytes and specific cell types relevant to affected tissues. The increasing number of studies linking DNA methylation and clinical features of the patients reinforces the usefulness of DNA methylation in the development of clinical markers. The incorporation of DNA methylation markers might help stratify patients on the basis of the disease subtypes. For some diseases, DNA methylation could provide useful markers to monitor disease activity and response to therapy. In addition to the association of DNA methylation profiles with a number of disease features, the stability and robustness of DNA methylation and the convenience of measuring DNA methylation profiles support the use of DNA methylation-based markers in clinical settings. Further efforts, such as the study of cell-specific alterations using large cohorts, are needed to translate these findings into clinical practice.

Key points.

• DNA methylation patterns are cell type specific, and particular patterns are associated with gene transcription and cellular function; aberrant DNA methylation profiles are associated with pathogenic cell phenotypes.

• Genetic susceptibility variants and environmental cues, in conjunction with nuclear factors, can influence the DNA methylation profiles of immune cells.

• Particular DNA methylation alterations are associated with the subtype, activity, progression and/or response to therapy of various autoimmune rheumatic diseases.

• The stability of methylated cytosines, and their physical association with the DNA, make these markers ideally suited for clinical purposes.

• For the future development of practical and useful DNA methylation-based markers, further studies that include large cohorts of patients and relevant cell types are needed.

Glossary

DNA methylation

The covalent attachment of a methyl group to a DNA residue (most commonly to a cytosine nucleotide followed by a guanine nucleotide, known as CpG sites).

Methylation quantitative trait loci

Genetic variants that are associated with DNA methylation levels at particular CpG sites.

Synovial fibroblasts

Also known as fibroblast-like synoviocytes, synovial fibroblasts are the main stromal cells of the joint synovium. These cells produce the extracellular matrix components of the synovial fluid and are critical for cartilage integrity and lubrication of the joint.

Regulatory T cells

(T_reg cells). A subpopulation of T cells that are immunosuppressive and generally suppress proliferation of effector T cells; these cells are essential for preventing autoimmune disease.

Genome-wide association studies

Observational studies of genome-wide sets of genetic variants in different individuals to identify variants associated with a trait; these studies typically focus on associations between single nucleotide polymorphisms and the occurrence of a disease.

Differentially methylated region

Genomic regions with a differential DNA methylation status across different biological samples. Differentially methylated regions usually involve adjacent sites or a group of sites close together.

Whole-genome bisulfite sequencing

A next-generation sequencing-based method for assessing the DNA methylation level of all cytosines in a genome, using sodium bisulfite treatment and DNA sequencing.

Reduced representation bisulfite sequencing

A method for analysing the DNA methylation profiles of genome regions that have a high CpG content using a combination of restriction enzyme digestion and bisulfite next-generation sequencing.

DNA methylation bead arrays

A high-throughput microarray-based method for measuring DNA methylation levels at single CpG site resolution.

Bisulfite pyrosequencing

A method designed to quantitatively determine the methylation status of individual cytosines in CpG sites from short PCR amplicons of DNA pre-treated with sodium bisulfite.

Epigenome-wide association study

An observational study examining a genome-wide set of epigenetic marks, including DNA methylation, and their association with a particular phenotype or trait (such as the occurrence of a disease).