Objectives To validate the presence and demonstrate the clinical value of the type I interferon (IFN)-signature during arthritis development.
Method In 115 seropositive arthralgia patients who were followed for the development of arthritis (Amsterdam Reade cohort), and 25 presymptomatic individuals who developed rheumatoid arthritis (RA) later, and 45 population-based controls (Northern Sweden cohort), the expression levels of 7 type I IFN response genes were determined with multiplex qPCR and an IFN-score was calculated. The diagnostic performance of the IFN-score was evaluated using Cox regression and Receiver Operating Characteristics (ROC)-curve analysis.
Results In 44 of the 115 at-risk individuals (38%) from the Amsterdam Reade cohort, arthritis developed after a median period of 8 months (IQR 5–13). Stratification of these individuals based on the IFN-score revealed that 15 out of 25 IFNhigh individuals converted to arthritis, compared with 29 out of 90 IFNlow individuals (p=0.011). In the Northern Sweden cohort, the level of the IFN-score was also significantly increased in presymptomatic individuals who developed RA compared with population-based controls (p=0.002).
Cox regression analysis of the Amsterdam Reade cohort showed that the hazard ratio (HR) for development of arthritis was 2.38 (p=0.008) for IFNhigh at-risk individuals after correction for anticitrullinated protein antibodies (ACPA) and rheumatoid factor (RF). The ROC-curve area under the curve (AUC) for the IFN-score combined with ACPA and RF in the prediction of arthritis was 78.5% (p=0.0001, 95% CI 0.70 to 0.87).
Conclusions The results demonstrated clinical utility for the IFN-signature as a biomarker in the prediction of arthritis development.
- Early Rheumatoid Arthritis
- Autoimmune Diseases
Statistics from Altmetric.com
Rheumatoid arthritis (RA) is a chronic inflammatory joint disease eventually leading to destruction of cartilage and bone.1 No curative treatment is currently available and prolonged treatment with Disease Modifying Anti-Rheumatic Drugs (DMARDS) or biologicals is required to suppress disease activity and joint damage. Early diagnosis in combination with timely initiation of treatment was demonstrated to increase the chance of remission and to prevent irreversible joint damage.2 Hence, recognition of individuals at high risk of developing RA may provide a major step forward towards strategies for the very early possibly preventive treatment of RA.
Accumulating evidence suggests a role for a dysregulated immune system prior to the appearance of clinical symptoms. Accordingly, the presence of anticitrullinated protein antibodies (ACPA) and rheumatoid factor (RF) up to 14 years before the onset of disease was demonstrated.3 ,4 Together with the demonstration of epitope spreading of the ACPA response,5 ,6 this suggests that immune tolerance has been lost many years before the disease manifests clinically. In a prospective study, it was demonstrated that 20% of ACPA-positive individuals at risk, and 40% of ACPA and RF-positive individuals at risk developed RA within 2 years.7 These findings show that the presence of these antibodies is important but not sufficient in itself to develop RA. Therefore, efforts to identify additional biomarkers, such as cytokines, chemokines and gene signatures have been made to improve the prediction of RA.8–11
Recently, we described gene signatures relevant to the development of RA. The results suggested a major role for type I interferon (IFN)-mediated immunity, as shown by the presence of an IFN-signature in blood cells in ACPA and/or RF-positive individuals, who developed arthritis, independent of ACPA positivity.9 The type I IFN signature consists of type I IFN response genes, and was previously shown to be also present in a subset of patients with established RA,12 suggesting that pathological processes in the preclinical phase of RA are reminiscent of those in established RA.
In the present study, we aim to study the association between type I IFN-signature and arthritis development in two independent cohorts, namely another cohort of seropositive persons at risk (Amsterdam Reade cohort), and a cohort of individuals before onset of symptoms of later diagnosed RA (Northern Sweden cohort), and demonstrate its clinical utility to predict RA development.
Materials and methods
The Amsterdam Reade cohort consisted of 115 newly included seropositive arthralgia patients at risk for RA without clinical arthritis, from the Jan van Breemen Research Institute Reade, that is, not the same patients as analysed in our previous publication.9 Inclusion criteria were the absence of arthritis despite joint complaints, as determined by two independent investigators (LvdS and DvS), a seropositive status (ACPA and/or IgM-RF positive3 ,5) and a minimal follow-up of 12 months. Exclusion criteria were: a history of arthritis ascertained by a physician, erosions on radiographs and previous use of DMARDs. Patients were followed biannually for the first year and then annually for the development of arthritis, defined as having one or more swollen joints by two independent investigators.
The Northern Sweden cohort consisted of 25 individuals at 2.9 years (IQR 2.2–5.5) before onset of symptoms of RA (referred to as presymptomatic individuals), and 45 population-based sex-matched and age-matched controls (PBC), identified from the Medical Biobank of Northern Sweden. Twenty-three of the individuals were also sampled when they were diagnosed with early RA (1987 American College of Rheumatology (ACR) criteria). They had a follow-up time of a median of 11 years (IQR 9–12).13 ACPA, RF, C-reactive protein (CRP), erythrocyte sedimentation rate (ESR) determinations were as previously described.4
All patients gave written informed consent, and the study was approved by the regional ethics committee in both countries.
RNA isolation and gene expression profiling
Total RNA was isolated from whole blood using the PAXgene RNA system (PreAnalytix, Hombrechtikon, Switzerland) for the Amsterdam Reade cohort and the Trizol (Invitrogen, Bleiswijk, The Netherlands) isolation method on buffy-coat cells for the Northern Sweden cohort, according to manufacturers’ instructions. RNA was purity tested and amplified as previously described.9 Multiplex q-PCR was performed using the 96.96 Biomark Dynamic Array systems (Fluidigm Corporation, San Francisco, USA) at ServiceXS (Leiden, The Netherlands), according to the manufacturers’ instructions. Expression levels of target genes were log2 transformed and calculated relative to GAPDH.
IFN-score calculation and statistical analyses
The IFN gene set that makes up the IFN signature, consisted of the seven strongest correlating type I IFN response genes, that is, IFI44L, IFI6, IFIT1, MXA, OAS3, RSAD2 and EPSTI (r=0.74), which discriminated between converting and non-converting arthralgia patients in a previous study using DNA microarray analysis.9 An IFN-score was calculated by averaging the relative expression of these genes (log2 based). Patients were stratified in a IFNhigh and IFNlow group based on a cut-off determined by receiver operating characteristics (ROC)-curve analysis using the IFN-score for arthritis development correlating with a specificity of 85%. Statistical analyses were done with Mann–Whitney U test, χ2 test, Cox regression and ROC-curve analysis using GraphPad PRISM 5.0 or SPSS V.15.0. p Values <0.05 were considered to be significant.
Demographic and clinical characteristics for both cohorts are shown in table 1.
First, we determined the IFN-score in seropositive arthralgia patients at risk for RA from the Amsterdam Reade cohort. During clinical follow-up (median follow-up time of 23 months (IQR 12–30)), 44 at-risk individuals developed arthritis, 40 (91%) RA according to the 2010 ACR/European League against Rheumatism (EULAR) criteria, and 4 (9%) undifferentiated arthritis, after a median of 8 months (IQR 5–13). Analysis of the linear IFN-scores revealed a trend for an elevated mean IFN-score in the group of converting at-risk individuals compared with the non-converting at-risk individuals (p=0.066) (figure 1A). When we stratified patients for a high and low IFN-score, a total of 25 patients were categorised as IFNhigh of whom 15 (60%) converted to arthritis compared with 29 out of 90 (32%) IFNlow patients, revealing a clear association of IFNhigh-status and arthritis conversion (χ2, p=0.011) (figure 1B).
Second, we studied the IFN-response gene expression in presymptomatic RA patients from the Northern Sweden cohort. The results revealed a significant elevated IFN-score in the RA patients compared with PBC (p=0.0012) (figure 1C). We also observed that the IFN-score was significantly increased in the presymptomatic individuals compared with PBC (p=0.0019). Straifying these individuals in an IFNhigh and IFNlow group, revealed that an IFNhigh status was present in 61% (14/23) of the RA patients, and 48% (12/25) of the presymptomatic individuals (median time before disease onset 2.9 years (IQR 2.2–5.5)) compared with 22% (10/45) in PBC (χ2, p=0.004) (figure 1D).
The results from the two independent validation cohorts confirm the association of an increased IFN-score with at-risk or preclinical individuals who develop RA.
In order to study the predictive value of the IFN-score for development of arthritis, we determined the HR for the IFNhigh versus IFNlow at-risk patients in relation to arthritis development in the Amsterdam Reade cohort. This analysis revealed that IFNhigh patients have a significantly higher risk of developing arthritis compared with IFNlow patients after correction for ACPA and RF status (HR of 2.38, p=0.008, 95% CI 1.26 to 4.49) (figure 2A). Age, shared epitope, CRP, ESR and non-steroidal anti-inflammatory drugs had no significant association with the conversion status.
Next we used the ROC-curve AUC analysis in the Amsterdam Reade cohort to determine the accuracy of ACPA/RF and/or IFN-score in separating arthritis converters from non-converters. First, we calculated the AUC for ACPA and RF as a predictor for arthritis development, which resulted in an AUC of 0.62 (p=0.032, 95% CI 0.514 to 0.724) (figure 2B). The IFN-score by itself gave an AUC of 0.602 (p=0.066, 95% CI 0.491 to 0.714), whereas the combination of ACPA/RF and IFN-score revealed an AUC of 0.785 (p=0.0001, 95% CI 0.699 to 0.872). This result demonstrates that the combination of ACPA/RF and IFN-score reached the highest AUC, and means that this combination correctly diagnoses 78.5% of randomly drawn pairs of arthralgia patients at risk for RA. Based on these data, a cut-off could be chosen to predict future arthritis with a specificity of 85%, and a sensitivity of 52.3% correlating with a Positive Predictive Value (PPV) of 65%.
Here, we support findings from a prior study suggesting that the IFN-signature genes are elevated in the blood cells of individuals at risk for RA.9 To this end, we used two independent validation cohorts of different nature. In the Amsterdam Reade cohort, we determined the IFN-score in seropositive individuals at risk for development of RA who were monitored for the development of arthritis. The other cohort consisted of presymptomatic individuals from the Medical Biobank of Northern Sweden, who subsequently developed RA. In both cohorts, we demonstrated a statistically significant association between high IFN-score and the risk of arthritis. Previously, we have shown that arthritis development is related to high-positive ACPA, or ACPA and RF double-positive status.14 Now we could demonstrate that the contribution of a high IFN-score to the risk of arthritis development is independent of ACPA and RF. These results reveal the utility of the IFN-signature to identify individuals at high risk for progression to arthritis.
In this study, the significance for the IFNhigh score in predicting the conversion to arthritis was already observed using a relatively short follow-up period (median follow-up of 23 months (IQR 12–30)). This relatively short follow-up period leaves open the possibility that future converters may still be present in the non-converter group, which may explain the finding that we did not reach significance (trend) for an association with the linear IFN-score values. Since we have the impression that the majority of conversions in the at-risk cohort takes place in the first 2 years after inclusion, it will be of interest to study the development of the IFN-activity in relation to the time to onset of arthritis.
The increased IFN-activity represented by the high IFN-score likely reflects various underlying processes that are associated with an activated immune status. This correlates with findings of elevated concentrations of proinflammatory cytokines and chemokines in the preclinical phase of RA.8 ,10 However, since cytokine and chemokine biomarkers are associated with autoantibodies, the IFN-score is likely to provide novel and additional clinical value.15 Underlying processes that may specifically be linked to the IFN-activity include a break in tolerance, dendritic cell (DC) differentiation, stimulation of the humoral and cellular arms of the immune system and chemokine activity.16–18 Of particular interest is the capacity of IFN-induced maturation of DCs. This may lead to the induction of costimulatory activity of immature DCs, leading in turn to a break in tolerance through the activation of autoreactive T cells.18 This process may be essential to facilitate epitope spreading of the ACPA response.
Our data demonstrated that an elevated IFN-signature represents an additional risk factor to predict RA. Since multiplex qPCR technology allows easy and accurate transcript quantification in peripheral blood cells, measurement of the IFN-signature represents an ideal methodology for biomarker assessment. Hence, the IFN-signature could be useful as biomarker for the prediction of RA in at-risk individuals, such as seropositive arthralgia patients and first-degree relatives of RA patients.
The Medical Biobank of Northern Sweden (head: professor Göran Hallmans) is gratefully acknowledged for the contribution of samples.
Contributors All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. CLV 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 conception and design: SV, DvS and CLV. Acquisition of patient material and of data: LvdS, MB, SR-D, JGW, JL. Data analysis: JL, SV, MB and LvdS. Analysis and interpretation of data: JL, SV, DvS and CLV.
Funding This research was performed with support from the ‘TRACER’ consortium in the framework of the Center for Translational Molecular Medicine (CTMM) (http://www.ctmm.nl), and the Dutch Arthritis Foundation (grant 04I-202 and grant 0801034).
Competing interests Dr D. van Schaardenburg (PhD, MD) and Prof. Dr C. Verweij (PhD) are inventors on a patent application, wherein the use of the gene signatures for the diagnosis of pre-clinical RA is described.
Patient consent Obtained.
Ethics approval Regional Ethics Committee.
Provenance and peer review Not commissioned; internally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.