Chromatin landscapes and genetic risk in systemic lupus

Systemic lupus erythematosus (SLE) is a complex trait believed to be caused by gene–environment interactions that lead to a perturbed immunologic state in which autoantibodies, immune complex deposition, and complement activation contribute to systemic inflammation and target tissue damage. The genetics of systemic lupus has been studied extensively, in particular its association with complement deficiencies. Although rare, C1q deficiency is the strongest genetic risk factor for SLE [1, 2]. C1r and C1s deficiencies are commonly inherited together, and over 50% of these patients develop SLE [3]. Moreover, homozygous C2 and C4 deficiencies have been shown to predispose toward SLE [4–6].

Other than complement deficiencies, however, associations between SLE and functions of specific genes have been harder to clarify. This situation became even more confusing as data began to emerge from genome-wide association studies (GWAS) and genetic fine mapping studies [7–9], where the majority of risk-associated single nucleotide polymorphisms (SNPs) occurred in noncoding regions of the genome, often considerable distances (in genomic terms) from protein-coding genes and their promoters. Thus, while it is still common in the literature to identify disease-associated SNPs by their nearest gene, most genetic risk for SLE does not appear to be within “genes,” as conventionally understood, at all. In this respect, SLE resembles almost every other complex trait studied by GWAS [10]. Maurano et al. [10] have shown that most SNPs for most complex traits lie within genomic regions identified by projects like ENCODE, Roadmap Epigenomics, and Blueprint Epigenomics as regulatory regions, often regions active during fetal life. This observation has been confirmed from studies of specific diseases. Recently, for example, Jiang et al. [11] demonstrated that regions of genetic risk for juvenile idiopathic arthritis (JIA) identified by genetic fine mapping using Illumina Immunochip arrays are enriched for H3K4me1 and or H3K27ac histone marks, epigenetic signatures associated with enhancer function. There is thus a broadly emerging consensus in the fields of genetics and functional genomics that genetic risk for complex traits likely involves specific aspects of transcriptional regulation and coordination rather than aberrant function of protein-coding genes.

In the current study, we examined the “epigenetic landscape” around known SLE-associated SNPs in an effort to better understand the potential significance of disease-associated SNPs. We focused on three cell types known to contribute to SLE pathogenesis: CD19⁺ B cells, CD4⁺ T cells, and neutrophils [12–16]. We used ENCODE and Roadmap Epigenomics data as well as data generated in our own laboratory (for neutrophils) to identify functional elements within these regions.

We queried the chromatin landscape around SNPs whose associations with SLE are well documented [17]. In addition, we queried recently reported SNPs found in a large Asian population [18]. CD19⁺ B-cell and CD4⁺ T-cell data were queried from ENCODE, while neutrophil RNA sequencing (RNAseq) and chromatin immunoprecipitation sequencing (ChIP-seq) data for H3K4me1/H3K27ac data were generated in our laboratory and have been reported recently [11]. Laboratory methods for ChIP-seq and RNAseq data are described briefly in the following.

Multiple GWAS in human disease have yielded surprising data demonstrating that a significant majority of disease-associated polymorphisms are located within noncoding regions of the genome [8, 9], i.e., those regions of the genome where transcription is coordinated on a genome-wide basis [32]. In fact, only 1–2% of the human genome is believed to span protein-coding genes [33, 34] and the remaining DNA is believed to incorporate an abundance of regulatory elements that contribute to maintenance of a cell’s identity and/or regulate specific cell functions.

Results of recent GWAS in SLE also illustrate this point. Of 46 SNPs identified by Bentham et al. and Sun et al. [17, 18] as conferring risk for SLE, only 16 lie within coding regions. In this study, we demonstrate that these SLE-associated SNPs lie within LD blocks containing histone marks commonly associated with enhancer function. We identified these marks in three cell types known to contribute to SLE pathogenesis and/or disease manifestations: CD4⁺ T cells, CD19⁺ B cells, and neutrophils.

Enhancers are cis-acting, active regions of DNA that promote gene transcription. Enhancers act by binding transcription factors and other transcriptional regulators that then alter the three-dimensional conformation of chromatin and facilitate the interaction between gene promoters and protein–DNA complexes. Enhancers may lie considerable distances from the promoters they regulate and may not regulate the genes closest to them [35]. Enhancers may also be cell-type specific and tissue specific, and may regulate more than one gene [19, 36]. For example, Martin et al. [35] recently used HiC chromosome capture approaches to identify long-range interactions between autoimmune disease risk loci and target genes. They demonstrated that SNPs lying large distances apart (in genomic terms) might either interact with the nearest gene or bypass multiple genes lying nearer to them to interact with those situated more distally.

The finding that the genetic risk for SLE lies largely within functional, noncoding regions of the human genome that contain regulatory elements in neutrophils, CD4⁺ T cells, and CD19⁺ B cells invites a new perspective in disease pathogenesis. All three cell types have been implicated in the pathogenesis of SLE [12–16]. Preliminary studies in adult SLE used RNAseq and found differentially expressed genes comprising different cellular functions from distinct leukocyte populations (in particular, from B cells and monocytes) [37]. Thus, cell-type-specific differences in gene expression may contribute to the pathogenesis of SLE. Similar findings of risk in the noncoding genome have been observed in the neutrophils and CD4⁺ T cells of JIA patients [11].

Our results demonstrate that, particularly in lymphocytes, there is copious transcription factor binding in H3K4me1/H3K4me3/H327ac-marked regulatory regions encompassed by the LD blocks containing SLE-associated SNPs, providing further evidence that these are important regulatory regions. Shi et al. [38] have shown that both promoters and enhancers exhibit significant changes in monocytes from SLE patients when compared with healthy controls. In particular, differentially methylated regions in SLE were significantly enriched in potential interferon-related TFBS. Furthermore, the importance of histone modifications (e.g., epigenetic marks) in regulating transcription is demonstrated in the recent work of Zhang et al. These authors identified distinct patterns of H3K4me3 methylation associated with aberrations in gene expression in monocytes from patients with SLE [39]. Their results demonstrated that genes overexpressed in SLE tended to respond to H3K4me3 changes downstream of transcription start sites.

Our findings, as well the expanded understanding of gene regulation that has emerged in the past 10 years, suggest a new paradigm of SLE pathogenesis that includes complex interactions between the innate and adaptive immune systems that may emerge because of disordered transcriptional regulation in both lymphoid and myeloid cells. The field of functional genomics is demonstrating that transcription is a complex process that must be regulated and coordinated on a genome-wide basis to maintain normal cellular function [32]. For example, transcription factors do not simply bind to DNA independently from other proteins, but rather interact with one another in layers of complexity. This newer finding suggests that transcriptional and regulatory networks are created for complex biological processes, and that even small perturbations of this system (e.g., from genetic variance or environmentally-induced epigenetic alterations) could accumulate over time. These cumulative small perturbations result in significant disorders in the regulation and coordination of transcription, ultimately leading to the development of disease. These newer data suggest that complex disorders may be due less to “bad genes” than to faulty gene regulation [40, 41]. Perhaps SLE, and many of the other conditions we refer to as “autoimmune diseases,” may be better understood as a disease of disordered transcription.

It is important to keep in mind the limitations of our study. First, both Bentham et al. and Sun et al. [17, 18] focused on regions of immunologic interest when performing GWAS, in that both groups used the Illumina Immunochip. The GWAS thus identified only selected genomic regions of specific immunologic interest that confer risk for SLE. It will be important to investigate whether additional risk loci would be revealed with a broadening of the query to genes that regulate chromatin access, or genes that regulate specific epigenetic processes (e.g., DNA methyl transferases, histone deacetylases, etc.).

In addition, as already mentioned briefly, enhancers appear to be cell-type specific and tissue specific. The enhancer marks used in this study for neutrophils and lymphocytes as mapped by the Roadmap Epigenomics project were detected in adult blood cells. There is the possibility that slightly different results could be obtained from cells in pediatric SLE patients. Thus, our results may not be generalizable or extrapolated, for example, to a pediatric population without further investigation. Indeed, we already know that the epigenomes of pathologically relevant cells may differ from epigenetic marks annotated in Roadmap data [42]. Coit et al. [40], for example, have shown that the methylome in CD4⁺ T cells of patients with SLE shows distinct differences from what is observed in healthy controls.

Moreover, the SLE patients included in the GWAS studies were generated from heterogeneous populations and most likely included patients with varying clinical manifestations. It is well known that SLE can affect different organ systems in the body with varying severity, and that treatment of these manifestations varies from milder immunosuppression with hydroxychloroquine alone to more aggressive immunosuppression with anti-neoplastic agents for disease control. Recently, Haddon et al. [43] demonstrated that pediatric SLE patients with kidney involvement possessed a different autoantibody profile than those without kidney involvement. Other groups then showed differences in expression levels of RNA and microRNAs in lupus nephritis biopsies [44, 45], suggesting that different clinical phenotypes may have individualized gene expression signatures.

Our results also have implications for the optimization of therapy in patients with SLE. Jiang et al. [20] demonstrated that response to treatment in JIA suggests that children on treatment for JIA can experience a long period of asymptomatic disease remission, but continue to have immune cell dysregulation. Our current treatments for SLE do not “normalize” immune cell function, as evidenced by continued periods of disease remission and flare. A better understanding of how epigenetic signatures drive gene expression signatures in SLE patients will allow us to better determine which patients with which SLE phenotypes will respond best to different treatments given their epigenetic profiles.

In this study, we have shown that disease-associated SNPs in SLE lie within LD blocks rich in functional elements regulating and coordinating gene transcription. These findings provide new insight into possible links between genetic and epigenetic risk factors for SLE.