Dr. Harley has received consulting fees, speaking fees, and/or honoraria from Bio-Rad Industries, Inc., and IVAX Diagnostics, Inc. (more than $10,000 each), and owns stock or stock options in IVAX Diagnostics, Inc.
Dr. Crow has received consulting fees, speaking fees, and/or honoraria from Genentech, Novo Nordisk, and ZymoGenetics (less than $10,000 each), owns stock or stock options in XDx, Inc., and has a patent pending for an interferon assay.
A haplotype of the interferon regulatory factor 5 (IRF5) gene has been associated with the risk of developing systemic lupus erythematosus (SLE), and our previous studies have demonstrated that high levels of serum interferon-α (IFNα) activity are a heritable risk factor for SLE. The aim of this study was to determine whether the IRF5 SLE risk haplotype mediates the risk of SLE by predisposing patients to the development of high levels of serum IFNα activity.
IFNα levels in 199 SLE patients of European and Hispanic ancestry were measured with a sensitive functional reporter cell assay. The rs2004640, rs3807306, rs10488631, and rs2280714 single-nucleotide polymorphisms (SNPs) in IRF5 were genotyped in these patients. Haplotypes were categorized as SLE risk, neutral, or protective based on published data.
SLE patients with risk/risk and risk/neutral IRF5 genotypes had higher serum IFNα activity than did those with protective/protective and neutral/protective genotypes (P = 0.025). This differential effect of IRF5 genotype on serum IFNα levels was driven largely by SLE patients who were positive for either anti–RNA binding protein (anti-RBP) or anti–double-stranded DNA (anti-dsDNA) autoantibodies (P = 0.012 for risk/risk or risk/neutral versus protective/protective or neutral/protective). The rs3807306 genotype was independently associated with high serum IFNα in this autoantibody group. We found no difference in IFNα activity according to IRF5 genotype in patients lacking either type of autoantibody or in patients positive for both classes of autoantibody.
The IRF5 SLE risk haplotype is associated with higher serum IFNα activity in SLE patients, and this effect is most prominent in patients positive for either anti-RBP or anti-dsDNA autoantibodies. This study demonstrates the biologic relevance of the SLE risk haplotype of IRF5 at the protein level.
The pathogenesis of systemic lupus erythematosus (SLE) is likely driven by a combination of genetic risk factors and environmental events that lead to an irreversible break in immunologic self tolerance. Interferon-α (IFNα) is a pleiotropic type I interferon with the potential to break self tolerance by activating antigen-presenting cells after uptake of self material (1). Serum levels of IFNα are elevated in SLE patients (2, 3) and are correlated with disease activity (4). Additionally, a number of patients treated with recombinant human IFNα for malignancy and chronic viral hepatitis have developed de novo SLE, which typically resolves after the IFNα is discontinued (5, 6). These data suggest a potential role of IFNα in susceptibility to SLE. In previous studies, we demonstrated that abnormally high serum IFNα activity is common in SLE families, both in healthy and SLE-affected members, as compared with unrelated individuals (7). These data implicate high levels of serum IFNα as a heritable SLE risk factor; however, the causative genes underlying this risk factor are not known.
Interferon regulatory factor 5 (IRF-5) is a transcription factor that can induce transcription of the IFNα transcript itself as well as many IFNα-induced genes (8). Genetic association studies examining single-nucleotide polymorphisms (SNPs) in IRF5 have defined distinct haplotypes that confer either susceptibility to, or protection from, SLE in persons of European and Hispanic ancestry (9–11). These haplotypes are characterized by functional changes in IRF5 messenger RNA (mRNA) (10, 11), which presumably will alter IRF5-mediated transcription, resulting in the risk of SLE. The SLE risk haplotypes of IRF5 could potentially alter serum IFNα activity in vivo. We hypothesized that the IRF5 SLE risk haplotype predisposes to high serum IFNα activity, which subsequently results in the risk of SLE. In this study, we examined serum IFNα activity and IRF5 genotypes in SLE patients to determine whether the IRF5 genotype influences IFNα activity in vivo in patients with SLE.
PATIENTS AND METHODS
Patients and samples.
Serum, plasma, and genomic DNA samples were obtained from the Hospital for Special Surgery (HSS) Lupus Family Registry, the HSS Lupus Registry, and the Lupus Family Registry and Repository (LFRR) at the Oklahoma Medical Research Foundation (OMRF). A total of 199 SLE patients of European American and Hispanic American ancestry were studied, including 108 patients from the HSS registries (58 of European ancestry and 50 of Hispanic ancestry) and 91 patients from the LFRR (65 of European ancestry and 26 of Hispanic ancestry). The patients of Hispanic ancestry in both registries are self-identified as Hispanic American, with ancestral origins in Mexico, South America, and the Caribbean. The allele frequencies and haplotype structures of IRF5 did not differ significantly between patients of European ancestry and patients of Hispanic ancestry, and both populations are presented in the study in aggregate. Control samples from 141 healthy subjects were obtained commercially from healthy blood donors (n = 59), as well as from healthy unrelated individuals who had donated to the 2 registries as control subjects (50 from the HSS registries and 32 from the LFRR). The study was approved by the institutional review boards at all institutions, and informed consent was obtained from all subjects in the study.
Measurement of autoantibodies.
Antibodies to double-stranded DNA (dsDNA) and RNA binding proteins (anti-Ro, anti-La, anti-Sm, and anti-RNP, collectively referred to as anti-RBP antibodies) were measured in samples from the HSS registries by the Clinical Laboratory at the HSS, using enzyme-linked immunosorbent assay (ELISA) methods. Antibodies were measured in samples from the LFRR by the Clinical Immunology Laboratory at the OMRF, using precipitin techniques. Standard clinical laboratory cutoff points from each laboratory were used to categorize samples as positive or negative, and a sample with a positive result for 1 or more of the anti-Ro, anti-La, and anti-Sm, or anti-RNP antibodies was considered anti-RBP positive in this study.
Stimulation of reporter cells.
ELISAs for the detection of IFNα in human serum have been complicated by poor correlation of the results with those of functional assays (12), possibly due to detection of a similar epitope on a non-IFNα protein or a stable, but biologically inactive, IFNα breakdown product (12). We have developed and validated a sensitive and reproducible bioassay for the detection of serum IFNα activity (7, 13). In this assay, reporter cells are used to measure the ability of patient sera to cause IFN-induced gene expression. The reporter cells (WISH cells) (no. CCL-25; American Type Culture Collection, Rockville, MD) are an epithelial-derived cell line that is highly responsive to IFNα.
WISH cells were cultured in minimum essential medium (Invitrogen, Carlsbad, CA) with Earle's salts, 10% fetal bovine serum, 10 mM HEPES buffer, 2 mML-glutamine, 100 units/ml of penicillin, and 100 μg/ml of streptomycin. Cells were then plated in 96-well culture plates at a density of 5 × 105/ml and incubated with 50% patient plasma or serum for 6 hours. Recombinant IFNα was used as a positive control, and healthy sera and culture media were used as negative controls. After 6 hours, the cells were lysed.
The ability of sera to cause IFN-induced gene expression was largely abrogated by the addition of monoclonal anti-IFNα antibody (Chemicon, Temecula, CA), confirming that IFNα was the major active type I IFN causing the IFN-induced gene expression (7). In a set of samples with low levels of IFNα activity, the addition of recombinant IFNα (BioSource International, Camarillo, CA) resulted in IFNα-induced gene expression proportional to the amount of IFNα added, excluding any frequent significant endogenous inhibitors of IFNα in the samples (7). Further details on the validation of the assay are described elsewhere (7, 13).
Purification of total cellular mRNA and synthesis of complementary DNA (cDNA).
Total cellular mRNA was purified from the WISH cell lysates using the Qiagen Turbocapture RNA purification kit (Qiagen, Chatsworth, CA) according to the manufacturer's protocol. The cDNA was prepared from the mRNA using the Invitrogen oligo(dT) primer and Superscript III reverse transcriptase system (Invitrogen).
Real-time polymerase chain reaction (PCR) quantification of mRNA-derived cDNA.
Ten microliters of a 1:40 dilution of the cDNA was then quantified using real-time PCR with the Bio-Rad SYBR Green fluorophore system (Bio-Rad, Hercules, CA). Forward and reverse primers for the genes encoding myxovirus resistance 1 (MX-1), RNA-dependent protein kinase (PKR), and IFN-induced protein with tetratricopeptide repeats 1 (IFIT-1), which are known to be highly and specifically induced by IFNα (14), were used in the reaction: for MX1, 5′-TACCAGGACTACGAGATTG-3′ (forward) and 5′-TGCCAGGAAGGTCTATTAG-3′ (reverse); for PKR, 5′-CTTCCATCTGACTCAGGTTT-3′ (forward) and 5′-TGCTTCTGACGGTATGTATTA-3′ (reverse), and for IFIT1, 5′-CTCCTTGGGTTCGTCTATAAATTG-3′ (forward) and 5′-AGTCAGCAGCCAGTCTCAG-3′ (reverse). The housekeeping gene GAPDH was also amplified to control for background gene expression: 5′-CAACGGATTTGGTCGTATT-3′ (forward) and 5′-GATGGCAACAATATCCACTT-3′ (reverse). Each sample and control was run in duplicate. Melt curves were analyzed to ensure the specificity of the amplified product, and standard curves were generated for each PCR experiment.
Real-time PCR data analysis.
The amount of PCR product of the IFNα-induced gene was normalized to the amount of product for the housekeeping gene GAPDH in the same sample. The relative expression of each of the 3 IFN-induced genes tested was calculated as the fold increase compared with its expression in WISH cells cultured in medium alone. Healthy unrelated donor sera were tested in the WISH assay to establish a normal value for IFNα activity, and the mean and SD of the relative expression of the IFNα-induced gene induced by healthy donor sera in the WISH assay were calculated. The ability of serum samples from the patients to cause IFN-induced gene expression in the reporter cells was then compared with the mean and SD induced by healthy unrelated donor sera. To accomplish this, the number of SDs of the relative expression above the mean in healthy donors was calculated for each gene. The mean relative expression of the gene in cells exposed to healthy donor sera was subtracted from the mean relative expression of the gene in cells exposed to experimental sera, and then the remainder was divided by the SD value for the same gene in healthy donors, which gave the relative number of SDs above that in healthy donors. This number was generated for each of the 3 genes and then summed to yield the final score. This normalization by the SD value in healthy donors was needed because each gene is up-regulated to a different degree following stimulation, and without normalization, data from the most highly up-regulated gene would dominate in the data set. Despite this difficulty, we think that measurement of 3 genes is a more robust indicator of downstream coordinate IFNα pathway activation than is measurement of just 1 highly induced transcript.
Individuals in the HSS registries were genotyped at the rs2004640, rs3807306, rs2070197, rs10488631, and rs2280714 SNPs using ABI TaqMan Assays-by-Design primers and probes on an ABI 7900HT PCR sequencer (Applied Biosystems, Foster City, CA). Individuals in the LFRR registry were genotyped at the rs2004640, rs3807306, rs10488631, rs10954213, and rs2280714 SNPs using the Illumina GoldenGate system (Illumina, San Diego, CA). SNP genotyping was performed with >98% completeness in both registries. All tested SNPs conformed to Hardy-Weinberg equilibrium; P values for departure from Hardy-Weinberg equilibrium were as follows: for rs2004640, P = 0.94; for rs3807306, P = 0.83; for rs2070197, P = 1.0; for rs10488631, P = 0.92; for rs2280714, P = 0.95; and for rs10954213, P = 1.0. There were no significant differences in allele frequencies between the 2 study sites; P values for the difference in allele frequency by site were as follows: for rs2004640, P = 0.67; for rs3807306, P = 0.27; for rs10488631, P = 0.74; and for rs2280714, P = 0.49 (by chi-square analysis).
Two-sided Fisher's exact test (sum of small P's method for observed ≥ expected) was used to analyze the categorical data, and the Mann-Whitney nonparametric t-test was used to compare quantitative data. Haploview 4.0 software was used for haplotype analysis and for calculation of D′ and r2 values. Haplotypes observed in the patients were categorized as risk, neutral, or protective as they related to SLE susceptibility, based on previous genetic association studies in SLE patients of European and Hispanic ancestry (10, 15).
IRF5 genotyping results.
All individuals were genotyped at the rs2004640, rs3807306, rs10488631, and rs2280714 SNPs in IRF5. Haplotype structures and frequencies based on these 4 SNPs in our cohort were very similar to published data (10), as illustrated in Figure 1. The rs2070197 SNP was genotyped only in the HSS cohort. SNP rs10488631 was a perfect proxy for rs2070197 (D′ = 1, r2 = 1), and the latter SNP was not genotyped in the LFRR patients. SNP rs10954213 is a functional polyadenylation site variant associated with SLE risk (10), and this SNP was genotyped in the LFRR cohort. The rs2280714 and rs10954213 SNPs were in high linkage disequilibrium (D′ = 1, r2 = 0.79), which is consistent with previous studies (10). The rs2280714 and rs10488631 SNPs were used as proxies for rs10954213 as it relates to SLE risk, and combinations of the 4 markers tested in all patients captured >98% of the variation at rs10954213.
Association of IRF5 SLE risk genotypes with higher levels of serum IFNα activity as compared with protective genotypes.
Among SLE patients stratified by genotype at single SNPs in IRF5, there were no significant differences in serum IFNα activity (P > 0.35 for all single-marker analyses). When IRF5 genotypes were analyzed as combinations of risk, neutral, and protective haplotypes, SLE patients with the risk/risk or risk/neutral IRF5 genotypes had higher serum IFNα activity than did those with the protective/protective or protective/neutral genotypes (P = 0.025). Patients with neutral/neutral or risk/protective genotypes demonstrated an intermediate level of serum IFNα activity, as shown in Figure 2. There was a trend toward higher IFNα activity in the risk/risk and risk/neutral group as compared with the neutral/neutral and risk/protective genotype group (P = 0.075). The 3 genotype groups presented in Figure 2B were formed by combining genotypes that had the most similar median values for serum IFNα activity; median serum IFNα activity values for each genotype group were as follows protective/protective = 1.94, neutral/protective = 2.35, neutral/neutral = 3.14, risk/protective = 3.13, risk/neutral = 5.81, and risk/risk = 9.59.
Most pronounced differential effect of IRF5 genotype on serum IFNα activity in SLE patients with either anti-RBP or anti-dsDNA antibodies.
In previous studies, we showed that anti-RBP autoantibodies, such as Ro, La, Sm, and RNP, as well as anti-dsDNA autoantibodies are independently associated with high levels of serum IFNα activity in SLE patients (7). Anti-RBP and anti-dsDNA antibodies influence serum IFNα in an additive manner, and patients who are positive for both classes of autoantibodies (double-positive) have higher mean levels of serum IFNα than those who are positive for either anti-RBP or anti-dsDNA alone (single-positive) (7). The 2 single-positive groups (RBP+/dsDNA– and RBP–/dsDNA+) have similar mean serum IFNα activity (7).
When SLE patients in the present study were stratified by double-negative, single-positive, or double-positive autoantibody status and IRF5 genotype, the greatest difference in serum IFNα activity by IRF5 genotype was seen in the single-positive autoantibody group, as shown in Figure 3. Double-positive patients had significantly higher serum IFNα activity as a group than single-positive patients (P = 0.015); however, there were no significant differences in serum IFNα activity by IRF5 genotype in the double-positive group. The strong influence of double-positive autoantibody status appeared to obscure the influence of IRF5 genotype, although the apparent lack of influence of IRF5 genotype in this group could also be due to the fact that there were fewer patients in the double-positive group to drive the statistical significance (42 double-positive patients versus 82 single-positive patients). There were no significant differences in serum IFNα activity by IRF5 genotype in the double-negative patient group (P = 0.46 for protective/protective plus neutral/protective versus risk/neutral plus risk/risk), despite having similar numbers of patients as the single-positive group. There were no differences in haplotype or genotype frequencies between the different autoantibody strata, suggesting that these 2 variables are independent (P for dependence = 0.29).
The median level of serum IFNα activity by antibody status and genotype in the double-negative and single-positive groups were as follows: 0.54, 0.88, and 1.01 in the double-negative group with the protective/protective or neutral/protective genotype, the neutral/neutral or risk/protective genotype, and the risk/neutral or risk/risk genotype, respectively, and 2.55, 5.42, and 19.52 in the single-positive group with the protective/protective or neutral/protective genotype, the neutral/neutral or risk/protective genotype, and the risk/neutral or risk/risk genotype, respectively. Thus, the double-negative risk/risk plus risk/neutral group had 1.8 times higher median serum IFNα activity than the double-negative protective/protective plus protective/neutral group. The single-positive risk/risk plus risk/neutral group had 7.6 times higher median serum IFNα activity than the single-positive protective/protective plus neutral/protective group. This would suggest a roughly 4-fold multiplicative interaction between single autoantibody positivity and IRF5 genotype on the median serum IFNα activity when comparing the double-negative and single-positive groups. Variances are wide in the data set, however, as shown in Figure 3, so this interaction estimate should be interpreted with caution.
Two-marker analysis based on rs3807306 and rs10488631.
After completion of our study, new data were published suggesting that a novel insertion/deletion in the promoter region could be a major contributor to the SLE risk signal arising from the 5′ region of IRF5 (15). In their study, Sigurdsson et al (15) found the most parsimonious designation of SLE risk using the promoter insertion/deletion as a marker for SLE risk in the 5′ region of the gene, and the rs10488631 SNP as a marker for SLE risk in the 3′ region of the gene. The risk variant of the promoter insertion/deletion was highly correlated with the A allele of the rs3807306 SNP (r2 = 0.61), which was also genotyped in the present study.
Given these data, we undertook a reanalysis of our serum IFNα activity data using the rs3807306 risk allele to designate the contribution of the 5′ region of IRF5 to SLE risk and the rs10488631 allele as a 3′ marker of SLE risk. As shown in Figure 4, serum IFNα activity varied significantly by IRF5 genotype at these 2 SNPs in the single-positive autoantibody group, with an increasing number of risk alleles corresponding to higher serum IFNα activity. Interestingly, in the single-positive autoantibody group, the rs3807306 SNP shows an independent association with high levels of IFNα in the absence of the rs10488631 risk variant (P = 0.03 for AA/TT versus CC/TT genotypes at rs3807306/rs10488631). In our cohort, the SLE risk variant of rs10488631 was always found with the SLE risk variant of rs3807306, so an independent influence of rs10488631 and rs3807306 could not be tested, although the highest levels of serum IFNα activity were seen in the presence of both SLE risk variants.
We have previously demonstrated familial aggregation of high serum IFNα activity in SLE families, suggesting that high serum IFNα is a heritable SLE risk factor (7). In this study, we demonstrated that the SLE risk haplotype of IRF5 is associated with higher serum IFNα activity in SLE patients than the protective haplotypes. These data suggest that IRF5 haplotypes contribute to the heritability of serum IFNα activity and play a role in the variance in IFNα activity observed between different SLE patients. Thus, it seems likely that IRF5 mediates SLE risk, at least in part, by modulating serum IFNα activity, which provides biologic relevance at the protein level for the SLE risk haplotype of IRF5. The protective/protective and protective/neutral genotype groups showed similar levels of serum IFNα activity, as did the risk/risk and risk/neutral genotype groups, and the risk/protective genotype group had similar serum IFNα activity as the neutral/neutral genotype group. These data suggest that risk and protective IRF5 haplotypes may have a dominant influence on serum IFNα activity when combined with a neutral haplotype and that the risk/protective haplotype combination may have a balancing effect on serum IFNα activity.
We demonstrated that separation of patients by autoantibody status is important, since the differential effect of IRF5 genotype on serum IFNα activity was detectable only in patients who were positive for either anti-RBP or anti-dsDNA and may have been obscured by higher background IFNα activity in patients who were positive for both categories of autoantibodies. It is interesting that the risk/risk and risk/neutral genotypes did not show significantly higher serum IFNα activity than the protective genotypes in the double-negative autoantibody group, despite having similar numbers of patients as the single-positive group. An effect of IRF5 genotype on IFNα activity in the double-negative patients is not ruled out by this study, however, since IFNα levels were much lower in this group in general than in the single-positive or double-positive patients, and greater numbers of patients may be required in order to detect a significant difference. The number of patients in the double-positive autoantibody group was small, and we were not able to exclude an association of IRF5 genotype with serum IFNα in this group.
The increased ratio of median serum IFNα levels between the single-positive protective/protective and neutral/protective genotypes and the risk/risk and risk/neutral genotypes as compared with the double-negative patients suggests the possibility of a gene–autoantibody interaction. In vitro models have shown that the addition of sera containing anti-RBP or anti-dsDNA antibodies to dendritic cells in culture results in brisk production of IFNα (16). This may result from the nucleic acid contained within these autoantibody–immune complexes triggering endosomal Toll-like receptors (TLRs) after uptake into cells via Fc receptors. If autoantibodies in SLE patient sera are required to see a differential effect of IRF5 genotype on serum IFNα activity, then IRF5 may be operative in SLE downstream of the activation of endosomal TLRs by nucleic acid–containing autoantibody–immune complexes. If IFNα is the protein mediator of SLE risk due to IRF5 genotype, then we would expect that if data from existing case–control genetic association studies are reanalyzed according to the autoantibody strata used in the present study, the IRF5 SLE risk haplotype would show a higher odds ratio for disease in single-positive patients than in double-negative patients. This may or may not be the case in the double-positive autoantibody group, but either result in this group would also be of high interest.
Emerging data suggest that an insertion/deletion polymorphism in the promoter region of IRF5 may play an important role in SLE susceptibility (15). When we reanalyzed our serum IFNα activity data using the model proposed in the present study and using rs3807306 as a proxy for the promoter insertion/deletion, we again saw significant differences in serum IFNα activity by IRF5 genotype in the single-positive autoantibody group. Linkage is strong between the promoter insertion/deletion and previously described SLE risk variants in the 5′ region of the gene, including both rs3807306 and rs2004640, and therefore, the findings of this reanalysis share many similarities with the findings of our initial analysis. However, the 2-marker analysis by rs3807306/rs10488631 genotype categorized the 6% frequency TCTA haplotype as a relatively lower SLE risk haplotype than did our previous analysis, and this new analysis allowed for an association of genotype in the 5′ region of IRF5 with high levels of IFNα, which is independent of 3′ risk alleles. This result could suggest that the promoter insertion/deletion plays a larger role in serum IFNα activity than does the 5′ splice-site variation; however, future in vitro and mechanistic studies are required to more definitively address this question.
High levels of serum IFNα activity show complex inheritance as a trait, and modeling the number of factors involved using relative recurrence-to-risk ratios suggests that 3–4 independent factors will be operative in a given SLE patient (7). Therefore, while the effect of IRF5 genotype on serum IFNα activity appears modest and limited to certain autoantibody groups in this study, we expect that a number of other genetic factors will also be important. It is likely that combinations of high IFNα–predisposing genetic variants will be required to fully manifest the high serum IFNα trait, and the other genetic factors that underlie high levels of serum IFNα as an SLE risk factor are as yet unknown.
Dr. Niewold had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study design. Niewold, Harley, Crow.
Acquisition of data. Niewold, Kelly, Flesch, Espinoza, Harley.
Analysis and interpretation of data. Niewold, Harley, Crow.
Manuscript preparation. Niewold, Harley, Crow.
Statistical analysis. Niewold.
We would like to thank Karen Onel, MD, and Kenan Onel MD, PhD, for their contribution in establishing the Hospital for Special Surgery Lupus Family Registry, and Gail Bruner, BSN, for her assistance in obtaining data and materials from the Lupus Family Registry and Repository at Oklahoma Medical Research Foundation.