Tumor necrosis factor polymorphisms associated with tumor necrosis factor production influence the risk of idiopathic intermediate uveitis

• Introduction
• Methods
• Results
• Discussion

Abstract

Purpose: Idiopathic intermediate uveitis (IIU) is a potentially sight-threatening inflammatory disorder with well-defined anatomic diagnostic criteria. It is often associated with multiple sclerosis, and both conditions are linked to HLA-DRB1*15. Previously, we have shown that non-infectious uveitis (NIU) is associated with interleukin 10 (IL10) polymorphisms, IL10-2849A (rs6703630), IL10+434T (rs2222202), and IL10+504G (rs3024490), while a LTA+252AA/TNFA-238GG haplotype (rs909253/rs361525) is protective. In this study, we determined whether patients with IIU have a similar genetic profile as patients with NIU or multiple sclerosis.

Methods: Twelve polymorphisms were genotyped, spanning the tumor necrosis factor (TNF) and IL10 genomic regions, in 44 patients with IIU and 92 population controls from the UK and the Republic of Ireland.

Results: IIU was strongly associated with the TNFA-308A and TNFA-238A polymorphisms. We found the combination of TNFA-308 and -238 loci was more strongly associated with IIU than any other loci across the major histocompatibility complex, including HLA-DRB1.

Conclusions: TNF polymorphisms, associated with increased TNF production, are highly associated with IIU. These results offer the potential to ascribe therapeutic response and risk (i.e., the influence of HLA-DRB1*15 status and TNFR1 polymorphism) to anti-TNF therapy in IIU.

Intermediate uveitis is an anatomically defined diagnosis, reserved for patients who have intraocular inflammation primarily involving the vitreous, peripheral retina, and pars plana [1]. Intermediate uveitis is a common presentation to general ophthalmology practice, particularly among children [2], with a population prevalence of 1.4/100,000 [3]. Generally, the onset is insidious with symptoms of blurred vision and floaters, and remission is infrequent and transitory. Vision loss is most commonly caused by chronic cystoid macular edema or secondary glaucoma [4].

The diagnosis of idiopathic intermediate uveitis (IIU) is restricted to patients in whom there is no evidence of infection; pars planitis refers to a subset with concomitant snowball formation or pars plana exudation (snowbanking) [2]. The presenting clinical phenotype overlaps with infectious causes of uveitis, including syphilis, tuberculosis, Lyme disease, cat-scratch fever, toxoplasmosis, Whipple's disease, Epstein-Barr virus, human T lymphotropic virus type I, and human immunodeficiency virus. Furthermore, intermediate uveitis is commonly associated with several diseases that span the full spectrum of the autoimmune-autoinflammatory (AI) immunological disease continuum [5], such as sarcoidosis, multiple sclerosis (MS), thyroid disease, and inflammatory bowel disease (IBD) [6]. Moreover, the familial aggregation of cases with IIU and other AI diseases suggests the existence of common genetic variants that underlie susceptibility to AI disease and/or a common environmental agent [7-13]. While unconfirmed, a general hypothesis is that an infectious foreign agent (virus or bacterium) systemically activates self-reactive T-cells in genetically susceptible individuals. The mechanisms by which this occurs are likely to involve molecular mimicry or non-specific bystander activation of self-reactive T-cells that home to the eye, leading to chronic, relapsing, or recurrent intraocular inflammatory disease [14].

The connection between IIU and MS merits further interrogation because both have been associated with the HLA-DRB1*15 antigen, a subtype of DR2 [15,16]. MS develops in 14% to 16.2% of patients with IIU, and either disease can precede the other [16,17]. In one study, 31% of patients with IIU who were HLA-DRB1*15⁺ also had MS, and 25% had a positive family history of MS [15]. Patients with MS have autoreactive T-cells and antibodies directed against glial proteins, such as myelin basic protein, that are associated with actively demyelinating lesions [18]. Glial proteins are also detected in snowbanks [8]. The inference is that patients with MS and IIU have autoreactive T-cells directed toward a common glial epitope.

The experimental models of uveitis (experimental autoimmune uveoretinitis, EAU) and MS (experimental autoimmune encephalomyelitis, EAE) share many common features; in particular, the detrimental role of tumor necrosis factor-α (TNFα). A consistent feature of EAU is the increased TNFα expression found in inflammatory cell infiltrates [19,20], and similarly, high levels of TNFα are found in MS lesion sites in patients with MS [21]. Furthermore, cerebrospinal fluid levels of TNFα are increased in chronic progressive MS compared with controls and correlate with disease severity [22]. Consequently, anti-TNFα monoclonal antibodies and a TNF receptor 1:immunoglobulin G fusion protein to ameliorate disease were tested in EAU and EAE with promising results [23,24]. Although anti-TNFα therapies translate effectively to the clinic for uveitis [25], this is not the case for patients with MS in whom anti-TNFα agents worsen disease and precipitate demyelination in others [26]. As a result, the administration of anti-TNFα therapy to patients with IIU has been somewhat more tentative. However, treatment successes with anti-TNFα therapy for patients with IIU have been reported [27,28], while others have described episodes of central nervous system demyelination in patients with IIU for the first time following anti-TNF therapy [29-31].

Previously, we have shown that specific genotypes in three haplotype-tagging single nucleotide polymorphisms (htSNPs) in the IL10 gene (rs6703630, rs2222202, and rs3024490) are significantly associated with susceptibility to non-infectious uveitis (NIU), while a LTA+252AA/TNFA-238GG haplotype (rs909253 and rs361525) is protective [7]. Moreover, patients with two closely overlapping white dot syndromes, punctuate inner choroidopathy (PIC) and multifocal choroiditis with panuveitis (MFCPU), demonstrated identical associations with the IL10 haplotype, IL10-2849AX/+434TC, that were not observed in other subgroups [32]. MFCPU and PIC fall under the SUN (Standardized Uveitis Nomenclature) Working Group classification of posterior uveitis with clinical evidence of multifocal choroiditis [1], and both disorders are characterized by inflammatory microgranulomata at the chorioretinal interface [33]. Hence, our data on the genetic profile of these patients suggested that MFCPU and PIC may be manifestations of the same disease [32].

In this report, we sought to further interrogate our study population to determine whether patients with IIU (which is also defined anatomically by the SUN Working Group) demonstrate a characteristic genetic profile that differs from that of patients with non-specific NIU. In addition, we were interested to know whether this profile was similar to genetic associations identified previously in patients with MS.

The demographics of the IIU and control groups were similar for age and sex with no significant differences (Table 1). As we have found previously, there was a high prevalence of AI disease in patients’ self-reported personal medical histories (29.5% versus 0% in controls) and family history (20.5% versus 5.4% in controls) [7]. The conditions that patients reported in their personal histories were non-specified arthritis/arthralgia (four patients), asthma (three patients), primary hyperthyroidism/hypothyroidism (two patients), primary hypoparathyroidism (one patient), fibromyalgia (one patient), psoriasis and arthritis (one patient), and type 2 diabetes mellitus (T2D, one patient). In the family histories, only one patient had a relative with MS, and another had a family member who was also affected by IIU. Other conditions reported by patients in their family history were T2D (one patient), celiac disease (one patient), rheumatoid arthritis (two patients), non-specified arthritis (one patient), inflammatory bowel disease (one patient), and primary hyperthyroidism (one patient). Five controls had a family history of AI disease: four were unaffected relatives of patients in the study, and one had a strong family history of arthritis.

The results of this study have shown that IIU is strongly associated with polymorphisms of the TNFA-308 and -238 loci and that these associations appear to be independent of HLA-DRB1*15 and the ancestral haplotype, HLA-A1, B8, DR3. The relationship is best explained with an allelic model, in which the TNFA-308A and TNFA-238A minor alleles are associated with IIU. Moreover, the combined TNFA-308G/-238G haplotype confers resistance to IIU (p_c=0.000004). Collectively, these results are enticing since the TNFA-308 and TNFA-238 loci have been shown to influence TNF production levels and have been associated with several other diseases on the AI immunological disease continuum [43].

Our results are consistent with previous work that has shown that the LTA+252AA/TNFA-238GG haplotype is negatively associated with non-infectious uveitis (p=0.00031) in a cohort where patients with six different uveitic syndromes contributed equally to the analyses [7]. Moreover, two white dot syndromes, PIC and MFCPU, demonstrated an extended LT-TNF haplotypic association with disease [32]. Yet none of the other uveitic syndromes in our original study, including Behçet’s disease, sarcoidosis, and sympathetic ophthalmia, demonstrated significant associations with the TNFA-238 and TNFA-308 loci that were independent of their HLA associations, for example, sympathetic ophthalmia with HLA-DRB1*04 [7,32,44](data not shown). HLA-B27 is a well-recognized susceptibility allele for idiopathic anterior uveitis and linkage disequilibrium between HLA-B27 and other variants within the human MHC known to confound analyses for additional susceptibility loci in this region. Several previous studies investigating the prevalence of TNF polymorphisms in patients with anterior uveitis have suggested genetic associations, but the inconsistent outcomes may be a result of heterogenous patient cohorts with a mixed number of underlying systemic disease associations or incomplete stratification analyses for HLA-B27 in patient and control groups [45-47].

The investigation of genetic influences on TNF production is also complicated by the location of the TNF gene cluster within the MHC, the most polymorphic region of the human genome [48]. Only 1.2 kb separates the polyadenylation site of LTA and the transcription start site of TNFA within the TNF gene cluster of the MHC class III region, genes that encode the structurally homologous inflammatory cytokines, LTα and TNFα. Moreover, examples of long-range linkage disequilibrium (LD), extending more than 2 Mb, arise in a subset of MHC haplotypes because of LD between tightly linked segments of strong LD creating a unique microstructure [49,50]. Consequently, the TNFA-308A polymorphism is often linked to the LTA+252G polymorphism in several combined HLA-TNF-LT haplotypes, including the HLA 8.1 ancestral haplotype (A1, B8, DR3) [41] that has repeatedly been associated with higher TNF production levels and many immunopathological diseases [43,51-57].

Why some studies have failed to demonstrate a correlation between polymorphisms of TNFA and TNFα production levels might be partly attributed to the chance representation of different subsets of haplotypes in each study, and partly due to differences in experimental methods, including the cell-type investigated, culture conditions, type of stimulant used (if any), and means of measurement [56]. Analyzing chromatin structure is one method for determining regions of a gene involved in transcriptional activation, and DNase I hypersensitivity (HS) sites represent nucleosome-free regions of a gene that are accessible to transcription factors as well as the DNase I endonuclease [58-60]. DNase I HS sites within regions of a gene demonstrating a high degree of DNA sequence conservation with other species, known as conserved non-coding sequences, are most likely to represent important regulatory sequences. Consequently, it is relevant that the proximal TNFA and LTA promoters are highly conserved between species, and constitutively active DNase I HS sites have been identified in these regions in human monocyte and T-cell lines, whereas other inducible HS sites are cell-type and stimulus dependent [61-66]. Although further regulatory elements have been identified in conserved non-coding sequences elsewhere in the TNF gene cluster [62,67-71], the LTA and TNF proximal promoter regions attain the highest conservation scores. Moreover, the TNFA-308 polymorphism appears to affect transcription factor binding and TNF transcriptional activity in a cell-type and stimulus-specific fashion [72]. These data from functional chromatin studies combined with the evidence from disease association studies further implicate this region of the MHC in the immunopathogenesis of several AI diseases. However, the relevance of specific polymorphisms in disease pathogenesis likely depends on the importance of the cell-type and stimulatory conditions in which the polymorphisms have most influence. Although one limitation of this study is that we did not include an analysis of TNF production levels based on patient genotype, this would clearly require a systematic investigation of the response of different leucocyte subclasses to varying stimuli before and after systemic immunosuppression is administered, which is beyond the scope of this report.

Perhaps for this reason, the role of TNF in the pathogenesis of AI disease is not always straightforward. The timing and duration of TNF expression are important in determining the pathogenic versus protective roles of TNF, since prolonged TNF exposure can activate antigen-presenting cells (APCs), augment antigen-presentation capability, and upregulate the expression of costimulatory molecules, while in other situations, TNF can inhibit the function of mature dendritic cells (DCs), induce their apoptosis, and impair antigen presentation [73]. In addition, exposure of DCs to TNFα in vitro has the capacity to induce a tolerogenic “semimature” functional phenotype on these cells, which can themselves secrete further TNFα to act in an autocrine fashion. When administered in vivo, TNFα-activated semimature DCs critically depend on the microenvironment to determine whether they remain tolerogenic or become immunogenic [74-77]. Hence, APCs are influenced by environmental cytokine signals that can promote either immunity or tolerance depending on their timing. These data might explain the paradoxical effects of anti-TNF therapy in MS. In comparison to IIU, the HLA class II region contributes most to genetic susceptibility to MS by linkage, case-control, and genome-wide association studies [78-82], while the link between TNF polymorphisms and MS remains contentious: TNFA-308A was significantly associated with reduced risk of MS in one large meta-analysis [83], but not another [84]. Since anti-TNF therapy for Crohn’s disease and rheumatoid arthritis (conditions not known to be linked to MS) have been reported to precipitate subsequent demyelination [85-88], low levels of TNF per se, whether due to TNF blockade or genetic factors, may be the major risk for demyelination [89], and systemic anti-TNF therapy for IIU might present the same risk. MS patients with a TNF receptor 1 (TNFR1) polymorphism, rs1800693, may be at particular risk of an adverse response to anti-TNF treatment because this SNP results in a soluble TNFR1 isoform that naturally antagonizes the action of TNF [89]. Whether this SNP is also predictive of demyelination after anti-TNF therapy in patients with IIU remains to be determined. Nevertheless, the data suggest that TNF has different roles in initiating and maintaining the two disorders: high levels increase the risk of IIU, and low levels increase the risk of demyelination; in addition, these levels are determined by several genetic factors, such as TNF and TNFR1 polymorphisms, and environmental factors such as exposure to anti-TNF therapy.

In comparison, IL10 polymorphisms appeared to have less influence on IIU susceptibility in our cohort. IL10 has mainly anti-inflammatory properties: it downregulates the expression of MHC class II and costimulatory molecules, inhibits the maturation of DCs, and inhibits the release of proinflammatory cytokines, resulting in the suppression of Th1 cell responses [90-93]. Furthermore, IL10 production by antigen-specific CD4⁺ Tregs is enhanced by IL10 [94]. Nevertheless, we did not detect an association between IL10-1082AA and poor visual outcome, as reported previously [42]. Neither was there an association between IL10-1082 and change in visual acuity from disease onset to the census date. One caveat is that 33/102 patients in the previous report had a Snellen VA <6/12 in both eyes after five years of follow-up [42], while only 1/44 patients met these criteria in our cohort. Although we predicted a change in visual acuity would be a better predictor of disease severity, it is likely that both visual outcome measures are confounded by coexisting ocular disease, differences in treatment regimen, and response to treatment between patients.

A further limitation of this study is that patients with coexistent MS were excluded from the outset, and none of our cohort developed MS during the follow-up period. Furthermore, only 2/44 patients were on anti-TNF therapy during the period of the study, and neither patient developed central nervous system demyelination. Nevertheless, the strong association of TNF haplotypes with IIU demonstrates the future need to investigate further their correlation with TNF expression, disease severity, and response to treatment with anti-TNF agents. Additionally, the association is important in driving a future ability to tailor therapy to those who will benefit, and to investigate further the relationship in the context of HLA-DRB1*15 status and TNFR1 polymorphism, to direct treatment to those that will demonstrate maximum benefit, but also to ensure no adverse effects (as demonstrated in MS).