|Home | About | Journals | Submit | Contact Us | Français|
High serum interferon α (IFNα) activity is a heritable risk factor for systemic lupus erythematosus (SLE). Auto-antibodies found in SLE form immune complexes which can stimulate IFNα production by activating endosomal Toll-like receptors and interferon regulatory factors (IRFs), including IRF5. Genetic variation in IRF5 is associated with SLE susceptibility; however, it is unclear how IRF5 functional genetic elements contribute to human disease.
1034 patients with SLE and 989 controls of European ancestry, 555 patients with SLE and 679 controls of African–American ancestry, and 73 patients with SLE of South African ancestry were genotyped at IRF5 polymorphisms, which define major haplotypes. Serum IFNα activity was measured using a functional assay.
In European ancestry subjects, anti-double-stranded DNA (dsDNA) and anti-Ro antibodies were each associated with different haplotypes characterised by a different combination of functional genetic elements (OR > 2.56, p >003C; 1.9×10−14 for both). These IRF5 haplotype-auto-antibody associations strongly predicted higher serum IFNα in patients with SLE and explained > 70% of the genetic risk of SLE due to IRF5. In African–American patients with SLE a similar relationship between serology and IFNα was observed, although the previously described European ancestry-risk haplotype was present at admixture proportions in African–American subjects and absent in African patients with SLE.
The authors define a novel risk haplotype of IRF5 that is associated with anti-dsDNA antibodies and show that risk of SLE due to IRF5 genotype is largely dependent upon particular auto-antibodies. This suggests that auto-antibodies are directly pathogenic in human SLE, resulting in increased IFNα in cooperation with particular combinations of IRF5 functional genetic elements.
SLE is a systemic autoimmune disorder affecting multiple organ systems including the skin, musculoskeletal, renal and haematopoietic systems. Humoral autoimmunity is a hallmark of SLE, and patients frequently have circulating auto-antibodies directed against dsDNA, as well as RNA binding proteins (RBP). Anti-RBP autoantibodies include antibodies which recognize Ro, La, Smith (anti-Sm), and ribonucleoprotein (anti-nRNP), collectively referred to as anti-retinol-binding protein). Anti-retinol-binding protein and anti-dsDNA auto-antibodies are rare in the healthy population.1 These auto-antibodies can be present in sera for years preceding the onset of clinical SLE illness2 and are likely pathogenic in SLE.3,4
Interferon regulatory factor (IRF)5 is a transcription factor that induces transcription of IFNα and IFNα-induced genes.5 Genetic association studies of IRF5 have defined haplotypes which confer either susceptibility to or protection from SLE in European ancestry individuals.6 These haplotypes are characterised by multiple functional genomic variants6 (figure 1A), which presumably alter IRF5-mediated transcription and subsequent risk of SLE.
IFNα is a pleiotropic type I interferon which can break self-tolerance by activating antigen presenting cells after uptake of self material.7 Serum IFNα activity is elevated in many patients with SLE,8–10 and high IFNα is associated with presence of anti-double-stranded DNA (dsDNA) and anti-RBP antibodies.10,11 Some patients treated with recombinant human IFNα for malignancy and viral hepatitis have developed de novo SLE, which typically resolves after the IFNα is discontinued.12,13 These data suggest a role for IFNα in SLE susceptibility.14
In our previous work, we have demonstrated that abnormally high serum IFNα is common in unaffected SLE family members, suggesting that IFNα is a heritable risk factor for SLE.8 The idea is supported by studies demonstrating that SLE-risk genetic variants in the IFNα pathway have been associated with high serum IFNα levels or increased sensitivity to IFNα.15–18 Further support for the genetic contribution to this trait in patients with SLE is provided by a recent genome-wide study that detected novel genetic variants associated with serum IFN levels in patients with SLE.19
We have previously shown that an SLE-risk haplotype of IRF5 was associated with high serum IFNα in patients with SLE.16 The differential effect of IRF5 genotype on serum IFNα was most prominent in patients with either anti-dsDNA or anti-RBP antibodies.16 We have recently demonstrated a similar model of association between genetic variants in the IRF7 locus and serum IFNα and particular auto-antibodies.15,20 Given that the genetic effect of IRF5 variants upon serum IFNα was dependent upon the presence of SLE-associated auto-antibodies, we hypothesised that these auto-antibodies may be required for a biological effect related to IRF5 variants and, if so, then SLE susceptibility related to IRF5 may depend upon these auto-antibodies. To test this hypothesis, we detected and modelled the associations between IRF5 haplotypes, auto-antibodies and serum IFNα in a large cohort of patients with SLE.
IRF5 genotype and auto-antibody data were obtained from 1034 patients with SLE and 989 controls of European ancestry, and 555 patients with SLE and 679 healthy controls of African American ancestry from the Oklahoma Medical Research Foundation (OMRF) Lupus Family Registry and Repository and OMRF Lupus Genetics Studies in collaboration with University of Uppsala, the PROFILE consortium of investigators21 and Imperial College London, as well as from the Hospital for Special Surgery (HSS) Lupus Family Registry and the University of Chicago Translational Research Initiative in the Department of Medicine (TRIDOM). 138 patients with SLE and 125 controls of African–American Gullah ancestry from the Medical University of South Carolina, and 73 patients with SLE of African ancestry from the University of the Witwatersrand, Johannesburg, South Africa were also studied. All patients with SLE met the American College of Rheumatology criteria for the diagnosis of SLE.22 In families multiplex for SLE, only one affected member was chosen randomly for inclusion to prevent bias. We studied four single nucleotide polymorphisms (SNPs) previously used to designate risk and protective haplotypes for SLE in European ancestry individuals (rs2004640, rs3807306, rs10488631 and rs2280714).6,23 Genotyping was performed at OMRF using the Illumina GoldenGate and Infinium systems (Illumina, Inc., San Diego, CA) and individuals in the TRIDOM, PROFILE and HSS registries were genotyped using TaqMan Assays-by-Design probes (Applied Biosystems, Carlsbad, CA). SNP genotyping was performed with >99% completeness at the rs3807306, rs10488631 and rs2280714 SNPs in both registries. The rs2004640 SNP was genotyped with 80.2% completeness at OMRF and >99% completeness in the TRIDOM and HSS registries. All tested SNPs had the expected Hardy–Weinberg proportions (p≥0.25 for all SNPs). The study was approved by institutional review boards in all institutions, and informed consent was obtained from all subjects.
Antibodies to Ro, La, Smith (anti-Sm) and ribonucleoproteins (anti-nRNP) in the OMRF samples at the OMRF Clinical Immunology Laboratory were measured by precipitin method, and those in the Hospital for Special Surgery and the University of Chicago samples, by ELISA in their respective clinical laboratories. Standard clinical laboratory cut-off points were used to categorise samples as either positive or negative. Anti-dsDNA antibodies were measured using Crithidia luciliae immunofluorescence at all sites, and detectable fluorescence was considered positive. Most subjects were derived from the OMRF registry, and there were no statistically significant differences in antibody prevalence between study sites (p≥0.10 for each, table 1).
Serum IFNα activity was measured in the same sample in which auto-antibodies were measured. We have developed a sensitive and reproducible bioassay to detect serum IFNα activity,8,24 as ELISA methods for detection of IFNα in human serum have been complicated by low sensitivity and low specificity.25 In this bioassay, reporter cells (WISH cells; American Type Culture Collection (ATCC), Manassas, VA, #CCL125) are used to measure the ability of sera to cause IFN-induced gene transcription. The reporter cells are cultured with patient sera for 6 h and then lysed, and three canonical IFNα-induced transcripts (interferon-induced protein with tetratricopeptide repeats 1 (IFIT-1), myxovirus (influenza virus) resistance 1 (MX-1) and eukaryotic translation initiation factor 2-alpha kinase 2 (PKR)) are measured using reverse transcriptase PCR. Relative expression data from the three transcripts are then normalised using the mean and SD of healthy donor sera (n=141) run in the same assay, and data are presented as an IFNα activity score. The IFN-induced transcriptional activity in the reporter cells can be blocked with anti-IFNα monoclonal antibodies, and no significant functional inhibitors have been detected to date.8 This assay has been highly informative when applied to SLE, as well as other human autoimmune disease populations.8,26,27
The χ2 statistic was used to analyze the categorical data, and ORs were calculated using standard parameters. The Haploview software V.4.2 was used for haplotype analysis (solid spine of linkage disequilibrium (LD) method) and calculation of r2 values. Logistic regression was performed using a backward logistic model including the five major haplotypes as predictors and each auto-antibody as an outcome variable serially, discarding non-significant predictors to achieve the best model fit. Attributable risk was calculated as the number of genotypes skewed toward SLE risk in a given patient category divided by the total number of risk-skewed genotypes in the SLE cohort as compared to controls. IFNα data were analyzed as a quantitative trait, with p-values calculated using the Mann–Whitney U test, as the data were not normally distributed.
European haplotype structures and frequencies were similar to those previously reported6,23 (figure 1B), and antibody prevalence data are summarised in table 1. In European ancestry patients with SLE, we compared haplotype frequencies between patients who either had or lacked particular auto-antibody specificities using logistic regression models. We found strong and strikingly distinct associations between anti-dsDNA and anti-Ro antibodies, and different IRF5 haplotypes (table 2). Anti-dsDNA antibodies were associated with two haplotypes, including the previously reported SLE-risk haplotype (TACA, #1), as well as a second haplotype which has been previously categorised as neutral with respect to SLE risk (TATA, #2, table 2).6 The two haplotypes associated with anti-dsDNA are characterised by a promoter insertion,23 splice site variation and a polyadenylation variant6 (table 2), while the exon 6 insertion is present on one of the associated haplotypes and absent on the other.
Anti-Ro antibodies were strongly associated with the previously reported SLE-risk6 TACA haplotype (table 2). This haplotype is characterised by presence of the promoter and exon 6 insertions, as well as the splice site variation and the polyadenylation variant. Notably, the main difference between the anti-Ro and anti-dsDNA associations is that anti-Ro is not associated with the TATA haplotype (#2), which lacks the exon 6 insertion. While this insertion may be required for the anti-Ro association, it is also found on other haplotypes that are not associated with anti-Ro such as haplotype #5, suggesting that this element alone is not sufficient. Anti-La antibodies were present in 7% of European ancestry patients with SLE and were almost exclusively found in combination with anti-Ro antibodies. Surprisingly, there was a third pattern of association between anti-La antibodies and IRF5 variants (table 2). Anti-La antibodies were associated with the rare haplotype #3. This haplotype lacks the promoter insertion, contains the splice site variation and exon 6 insertion and lacks the alternate polyadenylation site (table 2). Anti-nRNP and anti-Sm antibodies were not associated with IRF5 haplotypes, although the prevalence of these antibodies was low in European ancestry patients with SLE, and this could limit the power to detect associations with these antibody specificities (table 1).
Examining case–control data, we found out that IRF5 associations within subsets of patients defined by auto-antibodies were strong (table 3). The TACA and TATA haplotypes (#1 and #2) were strongly associated with anti-dsDNA positive patients versus controls (OR=2.79, p=2.9×10−20), and the TACA haplotype (#1) was strongly associated with anti-Ro positive patients versus controls (OR=2.57, p=1.8×10−14). In considering the overall risk of SLE due to IRF5 in our SLE cohort, 70.4% of SLE risk attributable to IRF5 was carried by subjects with either anti-Ro or anti-dsDNA auto-antibodies (37.4% and 33% due to anti-dsDNA and anti-Ro, respectively). Anti-La antibodies are found in a subset of patients with anti-Ro so stratifying subjects by anti-Ro captures the genetic effect of anti-La. Thus, a majority of the risk of SLE due to IRF5 was carried by a minority of the patients (393 of 1034 European ancestry patients with SLE in our cohort had anti-dsDNA, anti-Ro or both).
In African–American subjects, the four SNP haplotype structures were similar to those observed in European ancestry subjects (figure 1C), and many of the tested auto-antibodies were more prevalent than in European ancestry subjects (suppl. table 1). In case–control analysis, we found that the TACA haplotype (#1) was associated with anti-Ro and anti-dsDNA (OR=2.10, p=3.8×10−3), and no significant associations were observed with other haplotypes (suppl. table 2). This TACA haplotype reported as an SLE-risk haplotype in European ancestry6 is not found in the African ancestry HapMap Yoruba population. In our African–American controls, this haplotype is present at 2.3% frequency (suppl. table 2), which is consistent with 19.2% European admixture when compared to European controls. A similar admixture proportion is present in the African–American cases versus European ancestry cases (18.1%). Thus, the frequency of haplotype #1 is consistent with the expected 20% European admixture rate in African–American cases and controls, allowing for a similar disease association.
To explore this issue further, we genotyped 73 African patients with SLE from Johannesburg, South Africa, and found out that haplotype #1 (TACA) was not present, although the allele marking this haplotype in Europeans (rs10488361C) was found at 2.7% frequency on a different haplotype (suppl. table 3). Furthermore, in 138 patients with SLE and 125 controls of African–American Gullah ancestry with an average of 8% European admixture, the TACA haplotype is present in controls at 0.8% frequency (6.7% of European control frequency) and 3.3% frequency in cases (16.7% of European case frequency). We conclude that the association of these IRF5-tag SNP haplotypes with SLE in African–Americans28 is due to a European-derived haplotype and is the result of contributions from the sub-phenotypes of anti-dsDNA and anti-Ro. Interestingly, we did not observe any significant associations between IRF5 haplotypes and anti-Sm or anti-nRNP antibodies in this ancestral background, despite a much higher prevalence of these antibodies in African–American patients with SLE than in European ancestry patients with SLE (Suppl. table 1). Our African–American patient sample is smaller than the European ancestry cohort, and the tag SNPs used in this study are derived from studies of European ancestry populations. Thus, our more limited findings in African–Americans could relate to reduced statistical power and our more limited knowledge of IRF5 risk factors in this background. Further studies in this background are clearly warranted.
We next examined serum IFNα in the context of the auto-antibody-haplotype associations we observed (figure 2). Across both ancestral backgrounds, the haplotypes associated with particular auto-antibodies result in higher serum IFNα in the presence of that particular auto-antibody. This supports the contention that auto-antibodies are required for dysregulation of serum IFNα due to IRF5 risk variants. As noted above, the risk of SLE due to IRF5 was largely explained by these haplotype-autoantibody associations. Thus, the auto-antibody-positive patient group in which IRF5 exerts a differential effect on serum IFNα is the group that demonstrates significant risk of SLE due to IRF5 genotype. These data support a model of pathogenesis in which specific auto-antibodies cooperate with particular IRF5 variants to dysregulate IFNα production, resulting in an increased risk of SLE (figure 3).
The formation of SLE-associated auto-antibodies precedes clinical disease in many individuals.2 It is likely that SLE-associated auto-antibody immune complexes can stimulate IFNα production in humans in vivo via activation of the endosomal TLR system as has been demonstrated in some in vitro systems.3 Gain-of-function variations in IRF5 downstream of endosomal TLRs could then exacerbate the chronic endogenous stimulation provided by nucleic acid containing immune complexes. We have previously shown that subjects who have anti-Ro antibodies but do not have clinical autoimmunity also do not have high serum IFNα.26 These data suggest that background factors are important to the well-described association between serum IFNα and SLE-associated auto-antibodies, and we propose that IRF5 is one such important background factor.
While we can clearly show that SLE-associated auto-antibodies interact with IRF5 genetic variations to modulate serum IFNα in SLE, it is also possible that IRF5 genotype could directly predispose to the formation of these same auto-antibodies (figure 3, arrow with ‘?’). Endosomal TLR signalling is important in B cell maturation, and interestingly, IRF5 knockout mice show a deficit in the formation of SLE-associated auto-antibodies.29 This could suggest a feed-forward loop, in which IRF5 variants predispose to auto-antibodies that then cooperate with the same IRF5 elements to result in greater IFNα production. The fact that particular auto-antibodies are associated with specific IRF5 variants could indicate some preferential usage of IRF5 isoforms downstream of TLR7 versus TLR9, although this hypothesis is speculative at present. Plasmacytoid dendritic cells are one of the main IFNα producing cell types, and this may be this cell in which IRF5 variants exert their influence upon IFNα levels. In murine systems, plasmacytoid dendritic cells can activate B cells in multiple ways, including direct cell–cell contact,30 and IFNα produced by these cells can prime B cells to respond to TLR9 ligands.31 Thus, the feed-forward loop we suggest may manifest as an interaction between plasmacytoid dendritic cells and B cells in the presence of TLR ligands.
While anti-dsDNA antibodies and serum IFNα could potentially vary over time, we find strong associations with fixed genetic polymorphisms, supporting the concept of an inherent tendency toward these molecular traits. We and others have frequently observed genetic associations with anti-dsDNA antibodies and serum IFNα, and this precedent also suggests that there is some stability to both of these traits.15,19,32,33 It seems likely that there is an underlying propensity for some patients to have anti-dsDNA antibodies and high serum IFNα, upon which some variation over time can act as well. The prevalence of anti-dsDNA antibodies was somewhat low in our study, which may reflect the fact that most patients were recruited from the outpatient setting and which may have greater ability to assess the patients underlying intrinsic tendency toward anti-dsDNA antibodies with less variation related to disease activity. All sites in this study measured anti-dsDNA antibodies using the Crithidia luciliae immunofluorescence technique, and thus, our data are applicable to anti-dsDNA antibodies measured using this test.
Genetic associations with auto-antibody subsets within an autoimmune disease have been demonstrated previously, for example a PTPN22 variant is associated with anti-CCP positive rheumatoid arthritis.34 The strikingly detailed relationship we demonstrate between specific auto-antibodies and different IRF5 alleles is unprecedented. Our data support the concept that particular IRF5 functional genetic elements contribute to SLE risk through their relationship with auto-antibodies and IFNα production, and it is possible that IRF5 alleles help dictate the particular auto-antibodies produced. These results demonstrate the capacity of informative functional analyses and phenotypical variation to describe the genetic architecture of complex diseases. In a disease as complex as human SLE, it is likely that subset analyses incorporating detailed molecular phenotypes will be required to understand the various genetic contributions to pathogenesis.
Funding Funding was obtained from research grants from the National Institutes of Health (AR060861, AI083790, DK42086, AI071651 and RR024999 to TBN; AR62277, AR42460 and AI24717 to JAK and JBH; CA141700 and AR058621 to MEAR; AR002138, RR025741, AR30692 and Mo1-RR0048 to RRG; AR49084 to RRG, JCE and JBH; AR33062, AR42476 and MO1-RR00052 to JCE; RR-01070 to DLK; UL1RR029882 to DLK and GSG; AR049459 to GSG; AI065687 to TJV; RR15577, AR48940, AR045084 and AR053483 to JAJ; AR052125 and AI063274 to PMG; AI059893 to MKC; AI53747, AI31584, DE15223, RR20143, AI62629 and AR48940 to JBH) and from the Lupus Research Institute to TBN and MKC; the Alliance for Lupus Research to TBN, JBH, and MKC; the Arthritis National Research Foundation to TBN; the HHMI Gilliam Fellowship to SNK; the Connective Tissue Diseases Research Fund to JMF and MT; the Swedish Research Council and Gustaf Vth-80th-year Jubilee to MEAR; the US Department of Veterans Affairs to GSG and JBH; the Wellcome Trust and National Institute for Health and Research to TJV; the Mary Kirkland Center for Lupus Research to JAJ and MKC; and the Lou Kerr Chair in Biomedical Research to JAJ.
Competing interests None.
Patient consent All subjects signed informed consent to participate in the study, and no personal identifying information is presented for any of the patients.
Provenance and peer review Not commissioned; externally peer reviewed.