|Home | About | Journals | Submit | Contact Us | Français|
Osteopontin (OPN) is a multifunctional cytokine involved in long bone remodeling and immune system signaling. Additionally, OPN is critical for interferon-α (IFN-α) production in murine plasmacytoid dendritic cells. We have previously shown that IFN-α is a heritable risk factor for systemic lupus erythematosus (SLE). Genetic variants of OPN have been associated with SLE susceptibility, and one study suggests that this association is particular to men. In this study, the 3′ UTR SLE-risk variant of OPN (rs9138C) was associated with higher serum OPN and IFN-α in men (P = 0.0062 and P = 0.0087, respectively). In women, the association between rs9138 C and higher serum OPN and IFN-α was restricted to younger subjects, and risk allele carriers showed a strong age-related genetic effect of rs9138 genotype on both serum OPN and IFN-α (P<0.0001). In African-American subjects, the 5′ region single nucleotide polymorphisms, rs11730582 and rs28357094, were associated with anti-RNP antibodies (odds ratio (OR) = 2.9, P = 0.0038 and OR = 3.9, P = 0.021, respectively). Thus, we demonstrate two distinct genetic influences of OPN on serum protein traits in SLE patients, which correspond to previously reported SLE-risk variants. This study provides a biologic relevance for OPN variants at the protein level, and suggests an influence of this gene on the IFN-α pathway in SLE.
Systemic lupus erythematosus (SLE) is a prototype systemic autoimmune disorder, typically involving the skin, musculoskeletal, renal and hematologic organ systems. Susceptibility to SLE likely arises from a complex network of gene–gene and gene–environment interactions.1 Although many genetic loci have been linked to disease susceptibility, much work remains to integrate these loci into the pathophysiology of the disease. Recent studies suggest a role for interferon-α (IFN-α) in SLE pathogenesis. IFN-α is a pleiotropic type I IFN with the potential to break self tolerance by activating antigen-presenting cells after uptake of self material.2 Serum IFN-α levels are elevated in many SLE patients, and elevations may correlate with disease activity.3,4 In addition, a number of patients treated with recombinant human IFN-α for malignancy and chronic viral hepatitis have developed de novo SLE, which typically resolves after the IFN-α is discontinued.5,6 We have previously shown that abnormally high serum IFN-α is clustered in SLE families in both healthy and SLE-affected members, suggesting that high serum IFN-α is a heritable risk factor for SLE.7 In follow-up studies, we have demonstrated two SLE-associated genetic polymorphisms, IRF5 and PTPN22, predispose to high serum IFN-α in SLE patients,8,9 however, much of the genetic variation resulting in heritability of serum IFN-α remains to be discovered.
Osteopontin/secreted phosphoprotein 1 (OPN) is a phosphorylated extracellular matrix protein with a variety of functions, including wound healing, bone formation and remodeling,10 as well as immunological functions such as T-cell activation, Th1 differentiation, B-cell activation,11 and macrophage activation and chemotaxis.12 Studies have demonstrated high levels of OPN in biopsies of inflamed tissues in SLE and other autoimmune diseases,13,14 and variants of the OPN gene have been associated with SLE susceptibility.15–17 In a study of 394 Italian SLE patients, two single nucleotide polymorphisms (SNPs) in OPN were associated with SLE, including rs11439060 in the 5′ flanking region (odds ratio (OR) = 2.35, P = 0.006) and rs9138 in the 3′ UTR (OR = 1.57, P = 0.00094).16 Linkage disequilibrium between these two SNPs was low (r2 = 0.23), and logistic regression models supported an independent genetic association between each of these SNPs and SLE. A follow-up study including 707 European-American and 434 African-American SLE patients did not find significant associations between 5′ SNPs and SLE, although the rs11439060 variant was not genotyped directly.17 This study replicated the association between rs9138 and SLE, although surprisingly the association was particular to the male patients in their cohort (13% of the SLE patients studied). The previous Italian study did not separate patients by gender for analysis, however, the male/female ratio in that study was similar (14% male).
The potential influence of OPN genetic variants on serum OPN levels is not well described to date. One previous study found that the SLE-associated rs9138 C allele in the 3′ UTR was associated with increased serum OPN levels in healthy controls, however, the same association was not seen in the SLE patients.16 In murine models, OPN is essential for IFN-α production downstream of the endosomal Toll-like receptor 9 in plasma-cytoid dendritic cells, likely via interaction with the MyD88 adaptor protein.18 Plasmacytoid dendritic cells are thought to be the major IFN-α producing cells in the body, and this striking impact of OPN on the murine Toll-like receptor/IFN-α system suggests a potential relevance for OPN in IFN-α production in human disease.
In this study, we investigate the association of nine OPN SNPs with serum IFN-α activity and serum OPN levels in SLE patients, to determine whether genetic variation at the OPN locus is associated with altered serum levels of IFN-α, OPN or autoantibodies in SLE patients in vivo. SNPs previously associated with SLE15–17 and tag SNPs collectively representing >95% of the known variation in the gene region were included in the study.
Given that a gender-specific genetic association between the rs9138 SNP in OPN and SLE was detected in one study,17 we separated male and female patients for these analyses. Additionally, we have previously shown that younger SLE patients have higher serum IFN-α than older SLE patients.19 Given the role of OPN in long bone remodeling and the potential for age-related influences on OPN expression, we also looked for age-related trends in serum IFN-α and serum OPN levels related to OPN genotype. Cytokines were studied as quantitative traits in analyses stratified by OPN SNPs, gender and previously described high IFN-α age groups.19 SLE-associated autoantibodies were studied in the patients as categorical traits, analyzed as outcome variables in logistic regression models using OPN SNPs as predictor variables.
We analyzed serum IFN-α data in the 40 male SLE patients stratified by OPN genotype at each of the 9 genotyped SNPs. Diagrams showing the haplotype structure between the nine SNPs in each ancestral background are shown in Figure 1, and allele frequencies are shown in Supplementary Table 1. The rs9138 C allele was associated with increased serum IFN-α activity in male patients in a dose–response fashion (P = 0.0087; Figure 2a). rs9138 is the most consistently replicated SLE-risk SNP in previous case–control genetic studies,15–17 and this is the SNP which demonstrated a preferential association with SLE in men in one previous study.17 Other OPN SNPs did not show significant associations with serum IFN-α activity in male patients (P>0.4 for other SNPs, data not shown). Male patients of all ancestral backgrounds were analyzed in aggregate, as the number of male subjects was too small to support separate analyses in each ancestral background. Published work supports the idea that rs9138 C is a risk factor for SLE in male subjects of both European and African-American ancestry, with a similar OR in each ancestral background.17
We next examined serum IFN-α activity in the 283 female SLE patients in relation to each of the 9 OPN SNPs. No significant relationships were seen between any of the OPN SNPs and serum IFN-α (P>0.15 for all SNPs, data not shown). The rs9138 C allele has been associated with SLE in cohorts composed of 86% female subjects previously,16 however, there was no significant relationship between serum IFN-α and rs9138 in the female patients, in contrast to the above result in the male patients (Figure 2b). We undertook additional analysis to explore the possibility of an age-related genetic effect in female patients at rs9138. Given the role of OPN in bone remodeling and long bone growth,10 we hypothesized that the OPN gene may exhibit age-specific regulation, and that OPN may be differentially expressed in younger patients. We stratified the female patients by age, using cutoffs established in our previous work defining the highest IFN-α age group (age range 12–23).19 In this analysis, we saw that the rs9138 C allele was associated with increased serum IFN-α in a dose–response fashion in the younger patients (Figure 3; P = 0.018). There was a strikingly strong age-related difference in serum IFN-α comparing younger vs older patients with the AC genotype at rs9138, which was not present in subjects with the AA genotype (Figure 3; P = 0.0001 for higher serum IFN-α in subjects aged ≤23 years with AC genotype vs subjects aged >23 years with AC genotype, P = 0.87 for an age-related difference in serum IFN-α in subjects with AA genotype). When patients are analyzed separately by ancestry, similar age-related trends in serum IFN-α are seen in each ancestral background in patients with rs9138 AC and AA genotypes (Supplementary Figure 1).
We next measured serum OPN levels in the same sample in which IFN-α was measured for 180 SLE patients of representative age, sex and OPN genotype. There were no differences in OPN levels between male and female patients, although both male and female patients had higher serum OPN levels than the healthy unrelated controls (Figure 4a). Comparing OPN levels to IFN-α levels in the same sample, the two cytokines were weakly correlated in the overall cohort (Spearman's rho = 0.18, P = 0.016). When patients were separated by age and gender, there was an OPN/IFN-α correlation in 53 women aged ≤23 years with OPN data (Spearman's ρ = 0.30, P = 0.032) and there was a nonsignificant similar tendency in the 29 male patients with OPN data (Spearman's rho = 0.28, P = 0.15; combining female patients aged ≤23 years and male patients, Spearman's rho = 0.33, P = 0.0037). This relationship was not present in the 98 female patients aged >23 years (Spearman's rho = 0.09, P = 0.37).
We then examined serum OPN levels in patients stratified by gender, age in women and genotype at the rs9138 SNP to determine if serum OPN demonstrated similar age- and gender- related influences in relation to this SNP. Male SLE patients showed a dose–response increase in serum OPN in relation to the rs9138 C allele, similar to the pattern seen with IFN-α in male patients (P = 0.0062; Figure 4b). Female patients did not show any overall difference in serum OPN when stratified by rs9138 genotype (Figure 5a). However, after stratification of the cohort by age, the subjects ≤23 years old showed a dose–response relationship between rs9138 C and higher serum OPN (P = 0.021; Figure 5b). Older patients showed no statistical difference in serum OPN levels related to rs9138 genotype. Similar to the IFN-α analyses, younger patients with the AC genotype at rs9138 showed a large difference in serum OPN when compared to older patients with the AC rs9138 genotype (P< 0.0001), and patients with the AA genotype did not show any age-related difference in serum OPN levels (Figure 5b). Similar age-related patterns of serum OPN are seen in relation to the AC and AA genotypes of rs9138 in each ancestral background (Supplementary Figure 2).
Comparing allele frequencies at rs9138 in female SLE patients, there was a nonsignificant tendency toward increased frequency of the C allele in those ≤23 years old at the time of sample vs those >23 years old at the time of sample (European-American OR = 1.48, P = 0.25; African-American OR = 1.35, P = 0.53; Hispanic-American OR = 1.50, P = 0.41). The modified Breslow–Day test for OR homogeneity20 across ancestral backgrounds was not significant (χ2 = 1.99 with 2df, P = 0.37), indicating no significant heterogeneity of genetic effect in this sub-analysis. Thus, we also performed a combined analysis across all ancestries, which demonstrated a marginally significant enrichment of the rs9138 C allele in the younger subjects (OR = 1.49, χ2 = 4.22 with 1df, P = 0.040). Power is limited with our sample size, and this finding requires replication in larger independent cohorts.
We have previously shown associations between SLE-specific autoantibodies and serum IFN-α activity,7,8 and our group and others have demonstrated associations between SLE-specific autoantibodies and SLE risk alleles.21,22 We tested for autoantibody-OPN genotype relationships in our cohort using logistic regression modeling, and we found two SNPs in the 5′ region flanking OPN, rs11730582T and rs28357094 G, were associated with anti-ribonucleoprotein (anti-RNP) autoantibodies in African-American ancestry patients (P = 0.0038, OR = 2.9, 95% confidence interval (CI) = 1.4–6.1 for rs11730582 and P = 0.021, OR = 3.9, 95% CI = 1.1–11.8 for rs28357094). The rs11730582T allele was common in African-American subjects, and thus had a lower P value despite also having a lower OR than the rare rs28357094 G allele. These two SNPs are in LD with an insertion/deletion (−156 −/G, rs11439060) in European ancestry which has previously been associated with risk of SLE, and the association of the −156 indel was independent of the rs9138 C allele in this previous study16 We saw similar nonsignificant trends toward association of these two 5′ SNPs with anti-RNP antibodies in the other ancestral backgrounds. Anti-RNP antibodies were found at lower frequency in other ancestral backgrounds (62% in African-American vs 18% in European-American and 41% in Hispanic-American), and it is possible that we were underpowered to detect this association in other ancestral backgrounds.
The other SNPs, including rs9138, did not show any significant relationships with autoantibody traits in our SLE cohort. There was no significant linkage disequilibrium between the 5′ flanking SNPs, rs11730582 and rs28357094, and the 3′ UTR SNP, rs9138 (r2<0.2 for all pairwise comparisons in all ancestral backgrounds; Figure 1). The prevalence of anti-RNP antibodies did not differ by age, and there was no evidence for an age- or sex-related effect at these two 5′ flanking region SNPs (data not shown). Genotype at the two 5′ SNPs was not associated with statistical differences in serum IFN-α or serum OPN in the overall cohort, in African-American patients, or in patients stratified by anti-RNP antibodies (P>0.25 for each SNP in each of the above analyses). Although anti-RNP antibodies have been associated with higher IFN-α in SLE patients,7 we were likely underpowered to detect an overall difference in serum IFN-α in our cohort due to these alleles. There were no significant differences in serum OPN comparing subjects with anti-RNP antibodies to those lacking anti-RNP antibodies (P = 0.90 in overall cohort, P = 0.73 in African-American subjects).
In this study, we demonstrate an association of the OPN rs9138 variant with two cytokine traits, serum IFN-α activity and serum OPN. The SLE-risk rs9138 C allele is associated with higher levels of both cytokines, suggesting that these cytokines may mediate SLE susceptibility related to this locus. Strikingly, this effect at rs9138 was observed in a small cohort of male SLE patients with an age range of 8–68 (mean age = 29, s.d. = 18), whereas the same effect appears to be restricted to female SLE patients ≤23 years old. The male SLE cohort is not large enough to perform age stratification, and it is possible that a similar age-related effect is driving the relationship we demonstrate in male patients. However, a difference between men and women in OPN regulation is also possible, and interestingly a recent large-scale genetic association study found that risk of SLE due to OPN genotype at rs9138 was specific to male patients.17 We would hypothesize that reanalysis of the case-control genetic data in female patients by age may demonstrate an association between the rs9138 C allele and SLE susceptibility in women aged ≤23 years. Our current data suggest that the previously reported SLE-risk allele rs9138 C is enriched in younger female SLE patients, in whom the genetic effect on cytokine profiles is observed. This preliminary observation requires validation in larger case-control datasets, and if confirmed would support the idea that OPN variants predispose to SLE via their impact on cytokine traits.
We have previously demonstrated age-related trends in serum IFN-α activity in SLE patients and their first degree relatives,19 and this study would suggest that OPN genotype contributes to this phenomenon at least in women. Physiologic reasons for age-specific expression of OPN are unclear, however, OPN plays a role in long bone remodeling,10 and the period of long bone development during adolescence coincides with the age range in which we see differences in serum OPN due to OPN genotype. Additionally, work in both murine and human in vitro systems suggests that estrogen can stimulate OPN expression via the nonclassical estrogen receptor-related receptor-α (ERR-α) pathway.23,24 It is possible that age-related differences in estrogen signaling in the female SLE patients could underlie some of the age-related differences in OPN we have observed.
The mechanism by which OPN genotype modulates serum IFN-α is also unclear, although murine data suggest an essential role for OPN in IFN-α production by plasmacytoid dendritic cells.18 The rs9138 C SLE-risk variant in the 3′ UTR is associated with increased serum OPN levels in our study. We favor the hypothesis that both intracellular and extracellular OPN levels are regulated by the 3′ UTR region, possibly by variation in mRNA polyadenylation or stability. Other influences, such as promotion of gene expression due to increased estrogen/ERR-α signaling, could cooperate with increased mRNA stability to result in higher protein levels. Increased OPN levels in the setting of the risk variant could subsequently augment downstream signaling from endosomal Toll-like receptors, resulting in greater IFN-α production, and the correlation between serum IFN-α and serum OPN, which we observed in certain patient subgroups.
The association between 5′ flanking SNPs and anti-RNP antibodies in the African-American subjects is interesting, and this will be important to replicate and extend to other ancestral backgrounds. Anti-RNP antibodies are strongly associated with higher serum IFN-α in SLE patients,7 and the nucleic acid component of anti-RNP antibody immune complexes is thought to activate the endosomal Toll-like receptor pathway of IFN-α production.25 Our data suggest two distinct influences of OPN genotype on the IFN-α system in SLE, one variant in the 3′ UTR which likely influences OPN production and subsequently IFN-α production in plasmacytoid dendritic cells, and a second variant which is associated with anti-RNP antibodies capable of activating the endosomal Toll-like receptor pathway. Both of these variants have been previously associated with SLE susceptibility, suggesting that the protein phenotypes and patterns of association we present relate to underlying disease pathogenesis and susceptibility. The complex patterns of association we demonstrate may be helpful in reconciling previous case–control genetic studies, as differences in age, gender or autoantibody composition of the cases could have influenced the outcome of previous studies.
Serum and genomic DNA samples were obtained from the Translational Research Initiative in the Department of Medicine (TRIDOM) at the University of Chicago and the Hospital for Special Surgery (HSS) Lupus Registry. A total of 323 SLE patients (283 female, 40 male) were studied. Patients in the study were of African-American (n = 146), European-American (n = 108) and Hispanic ancestries (n = 69). All patients met four or more of the American College of Rheumatology criteria for the diagnosis of SLE.26 To standardize the IFN-α assay, 141 healthy donor serum samples were used, as previously described.7 The study was approved by the institutional review boards at all institutions, and informed consent was obtained from all subjects in the study.
The reporter cell assay for IFN-α has been described in detail elsewhere.7,27 Reporter cells were used to measure the ability of patient sera to cause IFN-induced gene expression. The reporter cells (Wistar Institute, Susan Hayflick (WISH) cells, ATCC no. CCL-25 (American Type Culture Collection, Manassas, VA, USA http://www.atcc.org) were cultured with 50% patient sera for 6 h, and then lysed. mRNA was purified from cell lysates, and cDNA was made from total cellular mRNA. cDNA was then quantified using real-time PCR using an Applied Biosystems 7900HT PCR machine with the SYBR Green fluorophore system. Forward and reverse primers for the genes MX1, PKR and IFIT1, which are known to be highly and specifically induced by IFN-α, were used in the reaction.7 GAPDH was amplified in the same samples to control for background gene expression.
The amount of PCR product of the IFN-α-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 three tested IFN-induced genes was calculated as a fold increase compared to its expression in WISH cells cultured with media alone. Healthy unrelated donor sera (n = 141) were tested in the WISH assay to establish a normal value for IFN-α activity, and the mean and standard deviation (s.d.) of IFN-α-induced gene relative expression induced by healthy donor sera were calculated. The ability of patient serum samples to cause IFN-induced gene expression in the reporter cells was then compared to the mean and s.d. induced by healthy unrelated donor serum. The number of s.d. above healthy donors for each gene was calculated to generate a serum IFN-α activity score as described in reference.7
OPN was measured in serum samples of 180 of the SLE patients of representative age, sex and genotype within the cohort using the Assay Designs, Ann Arbor, MI, USA human OPN enzyme-linked immunosorbent assay (ELISA), per manufacturer instructions. This ELISA kit has been shown to have good performance characteristics when used to measure OPN in human serum.28 Healthy unrelated donor samples were tested (n = 15) and performed as expected (mean = 1.71 ng ml−1, s.d. = 1.02 ngml−1).
Antibodies to anti-Ro, anti-La, anti-Sm and anti-RNP were measured by ELISA methods in the University of Chicago and the Hospital for Special Surgery samples in the respective clinical laboratories. Standard clinical lab cutoff points were used to categorize samples as positive or negative. Anti-dsDNA antibodies were measured using Crithidia lucidae immunofluorescence at both sites, and detectable fluorescence was considered positive.
Individuals in the HSS registries were genotyped at the previous SLE-associated SNPs rs9138, rs11730582 and rs28357094,15–17 as well as the tag SNPs rs10516800, rs2853749, rs11728697, rs6532040, rs1126772 and rs7655182 to cover >95% of the variation in the gene region. Genotyping was performed using ABI Taqman (Applied Biosystems, Inc., Foster City, CA, USA) Assays-by-Design primers and probes on an ABI 7900HT PCR machine. Diagrams showing the haplotype structure between the nine SNPs in each ancestral background are shown in Figure 1, generated using Haploview 4.0 software. None of the observed genotype frequencies at the studied SNPs deviated significantly from the expected Hardy–Weinberg proportions (P>0.01 for all SNPs in each ancestral background; Supplementary Table 1).
To analyze categorical data, χ2 distribution was used, and nonparametric Mann–Whitney U was used to compare quantitative IFN-α and serum OPN ELISA data between two genotype subgroups. Analysis of quantitative data across multiple genotype subgroups, which demonstrated a consistent dose–response trend, was done using a semi-log linear model, with equal weighting of genotype categories on the x axis and log-distributed PCR data on the y axis. Genotype categories were arrayed on the x axis by increasing SLE-risk designation, and P values were calculated as the probability that a horizontal line (null hypothesis) was a better fit of the data than a semi-log line with a non-zero slope between genotype categories using the extra sum-of-squares F test. The modified Breslow–Day test of OR homogeneity as corrected by Tarone is performed as described in.20 Data for the cross-ancestral analysis of rs9138 allele frequencies in younger vs older female patients are analyzed using an allelic test with a fixed-effects model. Logistic regression models were tested using the alleles of OPN SNPs as predictor variables in an additive model with the presence or absence of each SLE-associated autoantibody as the outcome variable. All P values shown in the paper are uncorrected for multiple comparisons.
Grant support and disclosures: SN Kariuki – none; JG Moore – none; KA Kirou – patent pending for interferon assay; MK Crow – NIH AI059893, Alliance for Lupus Research, Mary Kirkland Center for Lupus Research, Lupus Research Institute, and patent pending for interferon assay; TO Utset – Lupus Clinical Trials Consortium, Genentech; TB Niewold – NIH CTSA K12 Scholar Award RR025000-02, NIAID Clinical Research Loan Repayment AI071651, Arthritis Foundation Post-Doctoral Fellowship Award, Arthritis National Research Foundation Scholar Award.