Epistasis and immunity: the role of genetic interactions in autoimmune diseases

Abstract

Autoimmune disorders are a complex and varied group of diseases that are caused by breakdown of self-tolerance. The aetiology of autoimmunity is multi-factorial, with both environmental triggers and genetically determined risk factors. In recent years, it has been increasingly recognized that genetic risk factors do not act in isolation, but rather the combination of individual additive effects, gene–gene interactions and gene–environment interactions determine overall risk of autoimmunity. The importance of gene–gene interactions, or epistasis, has been recently brought into focus, with research demonstrating that many autoimmune diseases, including rheumatic arthritis, autoimmune glomerulonephritis, systemic lupus erythematosus and multiple sclerosis, are influenced by epistatic interactions. This review sets out to examine the basic mechanisms of epistasis, how epistasis influences the immune system and the role of epistasis in two major autoimmune conditions, systemic lupus erythematosus and multiple sclerosis.

In the era of systems biology, it is no longer possible to consider a gene in isolation: it is increasingly apparent that we must consider the interactions of a gene, including the interactions of its protein product, gene–environment interactions and gene–gene interactions. This is particularly relevant in complex human disease, where aetiology is multi-factorial. Autoimmune diseases are a varied group of complex disorders, with a strong genetic component, caused by breakdown of self-tolerance. This review sets out to examine the role of gene–gene interactions in autoimmunity: we review the known mechanisms of epistasis, and suggest where in the immune system they might act. We proceed to draw on disease-specific examples from two complex autoimmune conditions: multiple sclerosis (MS) and systemic lupus erythematosus (SLE).

Understanding epistasis: basic principles

The term epistasis has been used for over 100 years, being first used by Bateson to describe the masking of one disease-causing mutation by the co-inheritance of a mutation at a separate locus.¹ The idea has since been expanded to include any statistical deviation from the additive combination of two loci, the definition that is most commonly used today.^2,3 At first, this might seem a complex mathematical concept, but this is not the case. In simple terms, epistasis is the phenomenon whereby the effect of one genetic variant is altered by another. The term can be used interchangeably with gene–gene interaction, or genetic interactions.

It has been known for many years that disease-causing mutations show wide phenotypic variability, even within families, emphasizing that mutation outcome is dependent on the genetic background, as well as on environmental factors. There are many examples of simple Mendelian diseases showing phenotypic variability. An example in immunological disease comes from the study of X-linked lymphoproliferative disorder (XLP-1). Two families harbouring heterozygous deletion of SH2D1A demonstrated wide inter-familial and intra-familial variability of a clinical XLP-1 phenotype, despite identical disease mutations.⁴ The principle of epistasis is, however, just as relevant to complex diseases.

Complex genetic disorders are caused by polygenic inheritance of common variations, which are present within the normal population. Genome-wide association studies (GWAS) have begun to unravel the common traits underlying complex disease, with well over 1000 disease-linked loci identified.^5,6 There is, however, a continuing mystery: despite identification of many important loci, only 20–30% of heritability can be explained, with the cause of the majority of heritability of complex disease remaining obscure.⁷ It has been argued that the missing heritability relies on the identification of novel variants associated with disease, which might be a large number of common alleles with small effects, or a few very rare alleles with large effects.^8–14

It has been argued that this might not be the only answer: genetic interactions might explain a significant proportion of the missing heritability. For example, the inflammatory bowel disorder Crohn's disease has over 70 associated loci, which explain approximately 21·5% of the heritability of disease.¹⁵ The application of a relatively simple mathematical model that takes into account genetic interactions can, however, explain almost 80% of the current missing heritability.⁷ This is perhaps not surprising, as pervasive epistasis has been observed in several model organisms.^16–18 In yeast, the effects of more than half of primary quantitative trait loci were shown to be affected by a secondary locus and, in both Caenorhabditis elegans and Saccharomyces cerevisiae, it was shown that nearly all disease-causing mutations were affected by variation at secondary loci.^16,18–20 Furthermore, studies in yeast have revealed that it is not only binary epistatic interactions that need to be considered: the majority of phenotypic diversity seen within a yeast strain relies on complex multigenic epistatic interactions, involving at least five loci.^7,21

This said, it is undeniable that there are risk-alleles associated with complex diseases that remain to be identified. Studies of phenotypic variance between monozygotic and dizygotic twins have emphasized that a large proportion of continuous traits (such as body mass index, height and blood pressure) can be explained by additive effects of genetic loci, and that non-additive (i.e. epistatic) effects play a relatively minor role.²² This effect might well be because quantitative traits, such as those studied in the twin pairs, are determined by a large number of genetic factors, each making a very small contribution to the quantitative measurement. In these instances, epistatic interactions play almost no role, and the effect becomes additive.²³ We would argue, however, that in the cases of complex diseases there is a role for both additive effects and epistatic interactions. It is likely that for complex diseases there are a small number of major loci – many of which are likely to have been discovered already by GWAS – which interact with each other through epistasis, these interactions accounting for a substantial proportion of the missing heritability. In addition, some of the missing heritability will be caused by the additive effect of a large number of loci with individually very small contributions.

Gene–gene interactions within the immune system

The immune system shows marked complexity of phenotype and genotype and, as such, is a rich environment in which to observe epistasis. Networks of cell subsets – within and between which signalling molecules interact in cascades leading to many functional outcomes – give rise to a system in which there are many opportunities for epistasis. The understanding of these interactions at a functional level, and concomitantly at a genetic level, is key to understanding risk and susceptibility to autoimmune disease. The specific study of epistasis with regard to autoimmunity has so far been limited.²⁴ The wealth of basic research into autoimmunity can, however, be probed to identify candidate interactions, which might be subject to between-molecule epistasis.³ Comparison of these potential epistatic players with genes which have been implicated in autoimmune disease (by GWAS) paves the way for an integrated and functionally supported analysis of the genetic basis of autoimmunity, as has begun to occur in other fields such as cardiovascular disease.^25–27 Furthermore, genetic modelling that directly aims to detect epistasis might maximize the likelihood of identifying the susceptibility effects of individual genes.²⁸

At the most basic level, epistasis operates at direct interfaces between proteins, wherein alterations in the genetic code of either partner have the capacity to alter the character and affinity of their physical interaction.³ Direct molecular recognition is also the basis of immune function and so interactions between cytokines and their receptors, signalling molecules in cell activation cascades, and antigen receptors and their binding partners are all potentially subject to epistasis. Functional redundancy is another simple cause of epistasis and, again, is known to be rife in the immune system. A prime example is the MHC, in which functional redundancy is common, and epistasis is most likely to be crucial in the functioning of these multi-allelic, polymorphic genes.²⁹ The killer-cell immunoglobulin-like receptor system has similar functional redundancy between its members, and also directly binds MHC molecules.³⁰ It is clear, then, that there is ample scope for a multitude of within-molecule and between-molecule genetic interactions involving these two systems, with wide-reaching implications for immune disease.^31–33

As well as conceptually simple forms of epistasis between pairs of genes, or pairs of gene groups, more complex forms of epistasis operate within and between signalling pathways. Immune cell function is largely controlled by receptor-mediated activation of intracellular signalling pathways, which initiate transcriptional and non-transcriptional processes that affect the cell state and so the immune response. Epistasis might strongly govern the functional outcomes of pathway activation. In uni-directional signalling pathways, such as the mitogen-activated protein (MAP) kinase cascades (which are key to the function of both innate and adaptive immune cells), mutations in single genes can have adverse consequences for normal activity of the entire pathway.³⁴ Conversely, functionally opposing mutations in different pathway members can have masking or relieving consequences for potentially deleterious genetic variations.³ For example, it is plausible that increased enzymatic activity of a MAP-kinase-kinase could compensate for reduced activity of an upstream MAP-kinase. It is also theoretically possible – and perhaps inevitable – that epistatic interactions will occur that involve regulatory components of pathways, as true pathway linearity is rare. In immunity, this means consideration of intracellular regulatory components, such as phosphatase and ubiquitinase enzymes, as well as the immune-regulatory system formed by T-regulatory cells and immunosuppressive cytokines, such as IL-10 and TGF-β.

When it is considered that functional redundancy and co-operation are observed between entire systems of molecules – not simply individual molecules – it is clear that many more epistatic interactions are possible. In the immune system, this might arise when there are multiple pro-inflammatory and anti-inflammatory inputs to a cell; for example, the synergistic functioning of the NF-κB pathway, the MAP kinases and the phosphatidyl inositol 3 kinase pathway in activation of innate sentinel cells.^34–36 Another such possible source of epistasis is the stimulation of T cells through multiple cytokine receptors and their cogent Janus kinase/signal transducer and activator of transcription pathways, which occurs after recruitment of T cells by the cytokine milieu of an inflammatory focus.^37,38 These buffering interactions, in which epistasis can compensate for mutation in one pathway, are not confined to signalling pathways but might also occur between functional modules, such as the inflammasome and, as such, these functional units must be considered as candidates for epistasis.³⁹

The meta-interactions that are known to exist between pathways can be used to predict sets of epistatic partners for genes, and these ‘seed sets’ might prove invaluable in mapping epistatic networks onto the immunological genome. The mapping and modelling of epistasis has been proposed as a method which might prove invaluable in extracting useful hypotheses from GWAS datasets.⁴⁰ One important aspect of this approach, and another key tenet of epistasis that must be considered in research focused as such, is the presence of genetic hubs, which are known to act in most investigated cell systems.¹⁹ Genetic hubs are infrequent genes that have numerous epistatic interactions – vastly in excess of most genes in the system – and consequently, in which minor mutations have large effects on the overall phenotype. The molecular functions of these genes are often unexpectedly specialized, and this lack of clarity as to what predisposes a gene to being an epistatic hub means that candidates are hard to predict. The elucidation of which immune genes are hubs will, therefore, have far-reaching consequences for understanding the genetic basis of immunity, and must be a priority.

Epistasis in systemic lupus erythematosus

Systemic lupus erythematosus is a multi-system autoimmune disease, characterized by B-cell autoreactivity and deposition of immune complexes in tissues, including the muscles, joints, kidney and heart. The disease prevalence has been estimated at 25/100 000 in the UK, and the incidence is approximately eight times higher in women than men.⁴¹ Familial aggregation of cases has been noted since the early study of disease, long supporting a strong genetic aetiology.^42–45 A recent large survey confirmed that first-degree relatives of patients with SLE had an increased risk of SLE (odds ratio = 16·98), as well as of other autoimmune disorders.⁴⁶ Studies using twins showed that approximately 25% of monozygotic twins share an SLE phenotype, with a significantly higher level of concordance between monozygotic than dizygotic twins, supporting the importance of genetic factors.^47–49 The overall heritability of SLE was estimated at > 66%.⁵⁰

The recent advances in genomic technology have begun to unravel the complex genetic aetiology of SLE. There are currently more than 30 loci associated with increased SLE risk, identified (or confirmed) by seven large GWAS in different ethnic populations.^51–58 The two major loci associated with SLE are the HLA region on chromosome 6p and the FcγR gene cluster on chromosome 1q.^51–60

The HLA region is a well-characterized gene cluster at chromosome 6p21.4, containing over 200 genes with immunological roles. There has been consistent linkage of two HLA class II genes (HLA-DR2 and HLA-DR3) and two HLA class III genes (C4, C2) with SLE susceptibility, and there have also been reports of other associated genes, that have yet to be independently replicated (e.g. TNF, SKIV2L and MSH5).^50,52,59,60 A recent comprehensive study of epistatic interactions with the HLA region has confirmed that genetic interactions play an important role in SLE susceptibility.⁶¹ The two strongest epistatic effects were observed between CTLA4 and the HLA region. First, CTLA4 was shown to interact with rs3131379, a single nucleotide polymorphism in an intronic region of the MSH5 gene in the HLA Class 3 region, with a disease odds ratio of 1·19. Second, CTLA4 interacted with rs1270942, a single nucleotide polymorphism in an intronic region of the complement factor B (CFB2) gene, also in the HLA Class 3 region, this interaction increases the odds of disease by approximately 20%. Four other epistatic interactions were detected that associated less strongly with disease risk, although they were still statistically significant.

Cytotoxic T lymphocyte antigen 4 (CTL4A, also known as CD152) is a T-cell transmembrane protein responsible for negative regulation of T-cell activity following T-cell activation by antigen-presenting cells.⁶² The protein is critical for the maintenance of T-cell homeostasis and self-tolerance, acting through interactions with both conventional effector T-cells and regulatory T-cells.⁶³ The epistatic interaction between the HLA locus and CTLA4 underlines the importance of inappropriate antigen presentation and T-cell activation in the breakdown of self-tolerance and the pathogenesis of SLE.

The Fc receptors are a group of cell surface glycoproteins that bind the Fc portion of antibodies, thereby providing a link between humoral and cellular immune responses.⁶⁴ They regulate a variety of immune responses, including phagocytosis, mast cell degranulation, antibody-dependent cellular toxicity, B-cell activation and immune complex clearance.⁶⁵ The Fc receptors are encoded by a gene cluster at chromosome 1q21–24, which includes the Fc γ receptors (FcγR). Deficiency in FcγR has been implicated in several autoimmune diseases, including SLE, autoimmune glomerulonephritis and rheumatoid arthritis.^66,67 A recent study in SLE-prone mice defined two regions associated with defective self-tolerance, and epistatic relationships within one of these regions were further explored.⁶⁶ It was shown that a genetic interaction between FcγRIIB and slam genes was important in the development of SLE phenotype.⁶⁸

FcγRIIB is a negative regulator of B-cell receptor-mediated B-cell activation.⁶⁹ Deficiency of FcγRIIB in mice has been associated with the development of autoantibodies and autoimmune glomerulonephritis, whereas dysregulation of the human FcγrIIB during B-cell development has been implicated in SLE pathogenesis.^66,70,71 The slam gene cluster, located at chromosome 1q in both mice and humans encodes a family of transmembrane immune cell receptors. Slam gene family members participate in cell–cell interactions among many cell lineages in the adaptive and innate immune systems.⁷¹ The slam genes play an important role in the development of central tolerance in both B- and T-cells, predominantly through modulation of apoptosis, anergy and cell-cycle progression.⁷² Congenic mice containing autoimmune-type slam alleles were shown to produce autoantibodies – indicative of a failure to maintain self-tolerance – and the mice developed an SLE-like phenotype.^73,74

The discovery of Fcgr2b-slam interaction strengthens the argument that breakdown of induction of self-tolerance in lymphocytes is one of the key processes in autoimmune pathology. Despite the description of an epistatic interaction between Fcgr2b and slam, the study⁶⁸ did not explore other potential genetic interactions between the myriad of strong candidate genes that resided within the linked region (such as Fcgr3, Fcrla, Fcrlb) and, as such, it is likely that further experimental work will define more epistatic interactions between Fc receptors that are relevant to SLE pathogenesis.²⁴

Epistasis in multiple sclerosis

Multiple sclerosis is an inflammatory disease of the central nervous system in which activation of CD4⁺ T cells, predominantly of T helper type 1 polarization, leads to an influx of inflammatory cells, eventually causing demyelination, neuronal pathology and neurological dysfunction.⁷⁵ It has long been clear that development of the disease must involve both a genetic predisposition and an environmental contribution or trigger. The genetic susceptibility is indicated by high disease frequencies in particular ethnic groups such as northern Europeans,⁷⁶ and this is confirmed by twin studies, which show that monozygotic twins have a concordance rate of around 25%.⁷⁵ The environmental trigger is widely suggested to be Epstein–Barr virus infection, whereas the roles of exposure to sunlight and dietary factors (both previously implicated) have yet to be established.^77,78 The majority of research into the genetic basis of MS has been focused on identifying predisposing HLA alleles – notably variants at the HLA-DRB1 locus, particularly DRB1*15, as well as HLA-C.^79–81 The exponentially increased power of genomics in the past decade has, however, vastly expanded the list of disease-associated genes. The National Institutes of Health Catolog of Published GWAS⁸² lists 14 studies^83–96 that examined MS susceptibility; these studies implicated upwards of 100 distinct genes, although there was limited overlap of implicated loci between studies. Many of the identified genes – including HLA loci, which feature strongly – are immune-related. These include cytokines, chemokines and their receptors (such as IL-7, IL-12A, IL-12RA and CXCR4), signal transduction molecules and transcription factors (such as signal transducer and activator of transcription 3 and interferon-induced transcription factor 8) and co-stimulatory molecules (such as CD80, CD86 and CLECL1). Interestingly, molecules in the vitamin D metabolic pathway, such as CYP27B1, feature, whereas very few molecules involved in neurological pathways were associated.

Although GWAS have been invaluable in indicating potential contributors to the genetic risk for MS, and many loci have since been further established as risk factors in additional focused genetic studies, they have not been all-encompassing. First, identified loci account for < 50% of the known heritability of MS, meaning that there is substantial missing heritability. Second, direct testing and validation of implicated genes in vitro or in vivo has been limited, although models describing how they could act in scenarios such as CD4 T-cell maturation have been suggested, if not experimentally characterized.⁹⁶

Recent work has made progress in the understanding of the role that epistasis plays in MS. Combined with genome-based advances, this work has the potential to functionally characterize GWAS-identified risk loci, thereby accounting for a proportion of the missing heritability associated with MS. The majority of studies into the role of epistasis have been focused on understanding the interactions between risk-associated HLA alleles. A seminal study in humanized mice showed that linkage disequilibrium between DRB risk alleles might be the result of modifying effects of one allele on the T-cell response provoked by the other, highlighting the importance of balance in the immune response and providing insight into the classic relapsing and remitting phenotype of MS.⁹⁷

Further studies in mice have added to the understanding of how MHC alleles modify risk significance of co-expressed alleles.^98,99 Population studies have established the risk contribution of DQA and DQB HLA Class II alleles in concert with DRB1,¹⁰⁰ and have also defined the DRB1*15 risk contribution as being fundamentally subject to its overall HLA haplotype and hence potentially epistasis.¹⁰¹ Epistasis between HLA alleles has been identified in several ethnic populations, further cementing the importance HLA-region gene–gene interactions.^102–104

It is firmly established that HLA alleles play a major role in MS, and although studying the role of HLA-mediated epistasis in MS is valuable in understanding the complexities of inherited risk, it is unlikely to yield new therapeutic targets – the discovery of novel therapeutic targets in MS, and other complex diseases, will require the integration of genomic, transcriptional and functional research (Fig. 1). GWAS provides a vital list of genes that might play a role in the pathogenesis of complex disorders – the answer to the ‘what is involved?’ question. With the wealth of data available in the post-GWAS era, it is essential that we move from what? to how? and why?, or the resources used for the GWAS will not have been explored to their fullest potential.

Figure 1.

Approaches to research: one of the ‘pinnacles’ of medical research is the identification of novel therapeutic targets. In polygenic diseases, the most basic level (‘the foundation’) is genetic studies, which should include analysis of epistatic interactions. Genetic studies support transcriptional analysis, which provides a more comprehensive view of the gene within the cellular system. Finally, the top level of research is functional studies, which can lead to the discovery of novel therapeutic targets.

The first step post-GWAS is to analyse the epistatic interactions of the candidate genes identified, as this in itself might explain missing heritability in complex disorders. Furthermore, the interactions identified can be used to construct hypotheses and to guide future experimental work. For example, it might highlight important transcriptional or functional phenotypes, which could be explored further. This has been achieved in several studies that have used knowledge-driven analysis of genetic data to identify disease-associated non-HLA gene–gene interactions. Such strategies have highlighted previously unreported roles for T helper type 2 cytokines and complement factors in MS, a paradigmatically T helper type 1 disease, as well as highlighting the importance of dendritic cells in pathogenesis.^105,106 A further example of how GWAS data can be used in a constructive manner comes from a study into the genetics of osteoarthritis.¹⁰⁷ The study identified a gene, DOT1L, which was strongly linked to the osteoarthritis phenotype. Further studies were subsequently performed in mice, that demonstrated relevant tissue-specific expression of this gene (transcriptional/expression studies), and a new role for DOT1L in the process of chondrogenesis and connective tissue morphology (functional studies). The study concluded that DOT1L will be an interesting new target for osteoarthritis therapy. Another informative strategy is likely to be the integration of pathway and gene ontology databases with GWAS data, and this has recently provided the first genetic evidence implicating variants in neurological pathways in MS risk.¹⁰⁸

Genetic data might, furthermore, be used in this way to study autoimmunity as a paradigm. Genetic commonalities between conditions might be found by examining genes and groups of genes that are associated with several autoimmune diseases (outlined in Table 1). Predictably, genes associated with antigen presentation, the interferon pathways and T-cell signalling were found to be associated with several autoimmune conditions. Understanding the ways in which such genes interact epistatically, and the functional consequences of these interactions, might pave the way to a more global understanding of the development of autoimmune pathology and loss of self-tolerance that occurs in these debilitating conditions.

Table 1.

Common gene associations identified by genome-wide association studies in six major autoimmune disorders – rheumatoid arthritis (RA), inflammatory bowel disease (Crohn's disease and ulcerative colitis; IBD), systemic lupus erythematosus (SLE), multiple sclerosis (MS), type I diabetes mellitus (DM-I) and systemic sclerosis (SS). The Catalog of published genome-wide association studies (http://www.genome.gov/gwastudies/) was used to identify all genes that were associated with three or more of the autoimmune diseases of interest

Associated gene	Functions	RA	IBD	SLE	MS	DM-I	SS
Class II HLA alleles	Antigen presentation	x	x	x	x	x	x
IL2RA/B	Interleukin-2 receptor	x	x	x		x
IRF5	Interferon-induced transcription factors	x	x		x		x
IRF8		x	x	x			x
ETS1	Transcription factor, some roles in immune function shown	x	x		x
PTPN2	Apoptosis regulator	x	x			x
PTPN22	T-cell receptor signalling	x	x			x
TNFA1P3	Anti-inflammatory signalling molecule	x	x		x
TNFRSF14	T-cell signalling receptor. Herpes simplex virus entry receptor	x	x	x
AFF family members	Transcription factor expressed in lymphoid tissue	x			x	x
STAT4	Interleukin-2 induced transcription factor	x			x		x
BACH2	Transcription factor, some roles in immune function shown		x	x		x
IL7R	Interleukin-7 receptor		x	x		x

In summary, the phenomenon of epistasis is a crucial concept in the understanding of autoimmune disease. The abundant molecular interactions, functional redundancy within gene clusters, complex signalling pathways and genetic hubs mean that the immune system is a rich source of epistatic interactions. We are beginning to observe the extent to which gene–gene interactions alter the phenotypic outcome of risk-alleles for common autoimmune disorders, such as MS and SLE, although a great deal of further work remains to be done. It is highly likely that consideration of epistatic interactions will solve the paradigm of missing heritability in complex immune disorders, and further research in this field that uses genetic data as the basis for functional studies will, no doubt, clarify our understanding of the complex processes involved in autoimmune disorders.

Disclosures

The authors declare no conflict of interest.