Background
Systemic lupus erythematosus (SLE) is a heterogeneous autoimmune disease. One hallmark of SLE is the presence of autoantibodies against nuclear constituents (antinuclear antibodies [ANA]), including anti-double-stranded DNA (anti-dsDNA). Organ involvement is diverse, but joints, skin, mucosa, kidneys, and the nervous system are commonly involved. Typical laboratory aberrations include cytopenias and complement activation. SLE diagnosis overlaps with other autoimmune conditions in subsets of patients, such as antiphospholipid syndrome (APS) or Sjögren’s syndrome. Predictors of specific organ involvement or damage are still insufficiently identified [
1‐
3].
Type I interferons (IFNs; particularly IFN-α) play a major role in SLE pathogenesis, and type II IFNs are also of importance [
4,
5]. A proportion of patients displays increased serum levels of IFN-α or an upregulation of the IFN-regulated genes (IFN signature) [
6‐
8]. There are 13 known subtypes of IFN-α, and this fact constitutes one of multiple technical challenges in detection of IFN-α subtypes. Thus, many researchers rely on indirect measurements of the IFN signature [
9]. Interestingly, the gene signatures of type III (IFN-λ) and type I IFNs overlap [
10]. Increased levels of IFN-λ have been reported in SLE [
11]. Therefore, both these IFN subsets are of interest in the context of SLE and the IFN signature.
Four molecules belong to the IFN type III (IFN-λ) group: IFN-λ1, IFN-λ2, IFN-λ3 (also referred to as IL-29, IL-28a, and IL-28b, respectively), and IFN-λ4 [
12,
13]. IFN-λ is typically produced by virus-infected epithelial cells or plasmacytoid dendritic cells, but other antigen-presenting cells, T-helper type 17 (Th17) cells, keratinocytes, and neutrophils are additional sources of IFN-λ [
14‐
16]. There is a single IFN-λ receptor, which is expressed mainly on cells of epithelial origin, such as skin, gut, kidney epithelium, and neutrophils [
12,
17].
Th17 cells are important in SLE, and we previously reported that high levels of Th17 cytokines (interleukin [IL]-17 and IL-23) are associated with poor renal prognosis [
18]. Also, Th17 cells have been reported to produce IFN-λ1 in psoriatic lesions [
15]. However, the relationship between Th17 and IFN-λ has not been explored in SLE.
Chemokine C-X-C motif chemokine 10 (CXCL10)/interferon-γ-induced protein 10 (IP-10), initially described as an IFN-γ-inducible protein, is often considered an indirect biomarker of type I IFNs. IP-10 is upregulated in SLE and is associated with disease activity and specific clinical manifestations [
9,
19,
20].
SLE is a heterogeneous disease. Approximately 40% of patients develop nephritis, 80% have arthritis and/or mucocutaneous manifestations, and 70% demonstrate hematological abnormalities [
1,
2]. Many investigators have hypothesized that different pathogenetic mechanisms and molecules may drive SLE subgroups and result in the observed clinical and serological diversity. Moreover, on the basis of early data derived from trials on IFN-α therapies, researchers have reported that only a subgroup of patients with SLE with the IFN signature respond to the therapy [
21].
The aim of this study was to investigate if type I and type III IFNs drive SLE independently or in parallel, as well as which clinical parameters are associated with upregulation of these cytokines. In addition, we explored if there are intercorrelations between the IFNs, including IP-10, and the IL-17/IL-23 system.
In this article, we present our findings on the levels and clinical associations of IFN-λ1 and IFN-α in a large cohort of patients with SLE and a matched population of control subjects. Additionally, we describe the relationship between the IFN subtypes and IL-17A, IL-23, and IP-10 in regard to clinical and laboratory parameters of SLE.
Methods
Study population
This cross-sectional study included 261 consecutive patients with SLE from the Karolinska SLE cohort. All patients fulfilled at least four of the 1982 revised American College of Rheumatology classification criteria for SLE [
22]. Patients were ≥18 years old, and no other exclusion criteria were applied. We identified 261 control subjects in the population registry who were matched for age, sex, and geographical region to the patients with SLE. Among control subjects, a diagnosis of SLE was the only exclusion criterion. All participants underwent a structured investigation by a rheumatologist. Clinical and routine laboratory data were documented at the time of inclusion. SLE disease activity was assessed by using both the Systemic Lupus Erythematosus Disease Activity Index (SLEDAI) and the Systemic Lupus Activity Measure (SLAM). The latter captures more subjective symptoms such as fatigue and musculoskeletal pain [
23]. Organ damage was assessed with the Systemic Lupus International Collaborating Clinics/American College of Rheumatology Damage Index (SDI) [
24‐
26]. Definitions of the specific organ manifestations were recorded according to the SLAM and SLEDAI instruments, with slight modifications of some items, as defined below [
26,
27]. Patients with SLAM or SLEDAI scores >6 were considered to have active disease. Mucocutaneous activity was defined as a positive score for any of SLAM items 4–7. Damage was assessed using SDI definitions, with the exception of renal damage, which was defined as a glomerular filtration rate (GFR) ≤60 ml/minute/1.73 m
2, according to the Modified Diet in Renal Disease formula, or terminal renal failure due to nephritis (on dialysis or with a transplant) [
28,
29]. Venous thromboembolism (VTE) included both pulmonary embolism and deep vein thrombosis. Our definition of any vascular event (VE) included any objectively verified arterial or venous event, including stroke, transitory ischemic attack, myocardial infarction, angina, peripheral vascular ischemia, or VTE, as previously specified [
30]. Serum samples were collected at inclusion after overnight fasting, aliquoted, and stored at −70 °C until analysis.
Laboratory methods
All blood and urine chemistry analysis was performed according to standard routine at the time of inclusion at the internationally certified Karolinska University Hospital laboratory. ANA were analyzed by indirect immunofluorescence on HEp-2 cells (Immuno Concepts, Sacramento, CA, USA). Antibodies to specific nuclear antigens (anti-dsDNA, antinucleosomes, anti-Ro52/SSA, anti-Ro60/SSA, anti-La/SSB, anti-Smith) and phospholipids (anticardiolipin antibodies immunoglobulin (aCL IgG), and anti-β
2-antiglycoprotein domain 1 antibodies IgG [aβ
2GP1IgG]) were analyzed by multiplex bead technology (Luminex, Austin, TX, USA) using the BioPlex 2200 system (Bio-Rad Laboratories, Hercules, CA, USA). The cutoff for aCL and aβ
2GP1 fulfilled the 99th percentile as described previously [
31]. Lupus anticoagulant (LA) was determined by the modified dilute Russell’s viper venom time method (Biopool, Umeå, Sweden) using Bioclot LA (Trinity Biotech, Co.Wicklow, Ireland). aCL, aβ
2GP1, and LA are together referred to as antiphospholipid antibodies (aPL).
Detection of cytokines
IFN-λ1 and IFN-α levels in sera were measured by enzyme-linked immunosorbent assay (ELISA) according to the manufacturer’s instructions. For the IFN-λ1 assay, a mouse anti-IFN-λ1 IgG2A capture monoclonal antibody (MAB15981; R&D Systems, Minneapolis, MN, USA) and affinity-purified goat polyclonal IgG (BAF1598; R&D Systems) were used for coating and detection, respectively. The reagents had up to 100% cross-reactivity with IFN-λ3. IFN-α subtypes were measured using a pan-IFN-α ELISA detection kit. The detected IFN-α subtypes included 1, 2, 4, 5, 6, 7, 8, 10, 13, 14, 16, and 17 (3425-1A-20; Mabtech AB, Nacka Strand, Sweden), representing all but one IFN-α (IFN-α21). Therefore, we refer to our findings hereinafter as IFN-α.
In short, high-binding 96-well Nunc plates (Thermo Fisher Scientific, Waltham, MA, USA) were coated with capture antibody at 8 μg/ml for IFN-λ1 and 4 μg/ml for IFN-α in a carbonate buffer, pH 9.6, and incubated at +4 °C overnight. Plates were blocked (5% milk powder, 0.05% Tween in PBS) for 1 h, then washed and incubated overnight at +4 °C with patient sera, diluted 4:1 for detection of IFN-λ1 and 2:1 for IFN-α in a dilution buffer (3652-D2; Mabtech AB). Recombinant human IFN-λ1 (1598-IL; R&D Systems) and IFN-α included in the kit were used for derivation of standard curves. Ten serial dilutions were performed. Wells containing only buffer were used to determine the level of background. Negative control samples from five healthy control subjects were used when setting up and titrating the ELISA. Samples were run in duplicates. Biotinylated detection antibodies were added at a concentration of 0.4 μg/ml for IFN-λ1 and 1 μg/ml for IFN-α and incubated for 1 h in room temperature (RT). After a washing step, streptavidin-alkaline phosphatase diluted at 1:1000 was added and incubated for 1 h at RT. Afterward, a substrate solution (N1891, SIGMAFAST Protease Inhibitor Cocktail; Sigma-Aldrich, St. Louis, MO, USA) was added, and optical density was measured after 6 h and overnight at 405 nm.
Human IL-17A, IL-23, and IP-10 were analyzed with commercial sandwich ELISAs (DY317, DY1290, and DY266; R&D Systems) according to the manufacturer’s instructions. Streptavidin-HRP followed by addition of substrate (P3804 o-phenylenediamine dihydrochloride; Sigma-Aldrich) was used as a detection reagent, and optical density was measured at 450 nm.
The titration allowed us to detect IFN-α levels down to 2 pg/ml and 16 pg/ml for IFN-λ1. However, accurate detection levels recommended by the manufacturer and also in relation to the estimated standard curves of the ELISA were set to the following: 36 pg/ml for IFN-α, 300 pg/ml for IFN-λ1, 10 pg/ml for IL-17, 100 pg/ml for IL-23, and 18 pg/ml for IP-10. Values below the cutoff were denoted 50% of the cutoff value and transformed logarithmically with base 10 (log10).
Statistics
Student’s t test was used to compare normally distributed continuous variables, and the Mann-Whitney U test or the Wilcoxon/Kruskal-Wallis tests were used for non-normally distributed and nonparametric variables. For comparison between proportions, we used Pearson’s chi-square test or a two-tailed Fisher’s exact test. Correlations were calculated by Spearman’s rank correlation analysis. p Values <0.05 were considered significant. JMP software (SAS Institute, Cary, NC, USA) was used for all statistical analyses.
Discussion
This is, to the best of our knowledge, the first comparative analysis of serum levels of types I and III IFNs in patients with SLE and matched population control subjects. The levels of both IFNs were higher in patients with SLE. Interestingly, increased and even high levels of IFN-λ1 and IFN-α could also be detected in a proportion of control subjects, both in patients with other diagnoses and in apparently healthy individuals. Our well-characterized SLE cohort enabled distinct phenotypic stratification of patients based on cytokine profiling. We report that both IFN types are increased in subgroups of patients with SLE, but only with a partial overlap and with no correlation. Furthermore, we demonstrate that high levels of IFN-λ1 and IFN-α are associated with different clinical and serological profiles. IFN-λ1 correlates with classical Th17 cytokines, and the subset of patients with high levels of IFN-λ1, IL-17A, and IL-23 is characterized by more organ damage, in particular renal impairment. Moreover, we identified two distinct subsets of patients with active SLE: one with high levels of both IFN-λ1 and IFN-α and another with high levels of IP-10.
Our data indicate that specific antibody patterns are associated with different cytokine profiles. We found two nonoverlapping subpopulations positive for antinucleosome antibodies: one IFN-λ1
high and another IP-10
high. The latter association has been reported before [
32]. Interestingly, antinucleosome antibodies have been suggested to be even more specific than anti-dsDNA and predict lupus nephritis [
33,
34]. We found that the IFN-α
high group, in line with previous reports, was more often positive for Ro/SSA and La/SSB antibodies [
8]. We report a novel observation that positivity for aPL and VEs are less common in the IFN-α
high group.
In SLE, type III IFNs have scarcely been studied, and this is the first large study where clinical associations could be thoroughly investigated [
11]. We found that musculoskeletal involvement is uncommon among IFN-λ1
high patients. In an earlier, smaller Asian study, researchers reported that levels of IFN-λ1 correlate to SLEDAI and that patients with renal and/or arthritic manifestations have higher serum levels of IFN-λ1 [
11]. We could not directly confirm these associations. This might be due to genetic differences but also to a different study design (ours being cross-sectional and the other one at disease exacerbation). Interestingly, other investigators have reported that IFN-λ2 had a therapeutic effect in collagen-induced arthritis model through reduction of Th17 cells [
12]. In addition, we found a correlation between IFN-λ1, IL-17A, and IL-23 and that the triple-high subgroup displays higher SDI scores and more renal impairment. Moreover, we found that positivity for all three aPL antibodies and a history of thrombocytopenia are common characteristics of this subgroup. We and others have previously reported that high IL-17, IL-23, and aPL are associated with poor nephritis outcomes [
18,
35,
36]. Thus, our data imply that type III IFNs, together with Th17 cytokines, rather than IFN-α, are associated with an unfavorable nephritis prognosis.
Data on types I and III IFNs and mucocutaneous lupus are somewhat conflicting [
8,
16]. Increased IFN-λ1 expression in serum and skin lesions in patients with cutaneous lupus erythematosus has been reported, but our present study and previous reports indicate that the mucocutaneous inflammation in SLE is most probably driven by type I IFNs [
8,
16].
We further report leukopenia in the IFN-αhigh group and lymphopenia in the IFN-λ1high group, whereas the double-IFNhigh groups were low in both leukocytes and lymphocytes. Triple-high IFN-λ1, IL-17A, and IL-23 patients displayed a higher frequency of thrombocytopenia. Hence, different hematological manifestations seem to be associated with different cytokine patterns.
So far, data regarding the association between IFN signature and disease activity have been of a dual nature [
6,
7,
37]. According to our findings, upregulation of both types I and III IFNs is associated with high disease activity and a phenotype comprising neuropsychiatric involvement, lymphadenopathy, leukopenia and lymphopenia, Ro/SSA positivity, and low C4 level. Further, in our cohort, upregulation of IP-10 was not, as often assumed, a proxy for IFN-α activity [
19]. Rather, levels of IP-10 were independently associated with high disease activity. In addition, active arthritis was more common in this group. IP-10 is also regarded as a disease activity marker in rheumatoid arthritis, and our findings indicate that it might be a marker of lupus arthritis [
38]. Altogether, our findings indicate that patients with active SLE, but with different phenotypes, can be identified by either high serum levels of both IFN-λ1 and IFN-α or high levels of IP-10.
The clinical profile of the IFN-α
high subgroup in our cohort is in line with previous studies [
8]. Our novel observations are that the IFN-α
high group had better preserved renal function, was less often positive for aPL, and had fewer VEs. Hence, only a few of them were receiving warfarin treatment. Our results support previous observations that the risk for cardiovascular morbidity and mortality (cardiovascular disease [CVD]) is associated with positivity for aPL, whereas patients with Ro/SSA and La/SSB positivity have reduced risk [
1,
39]. In the present study, we further define that the group with lower CVD risk had high IFN-α levels. Positive associations between activation of IFN-regulated genes and CVD were previously reported. However, specific IFN levels were not investigated in this context, which is of importance because type III IFNs also contribute to the IFN signature [
40].
Our study included measurement of IFN-α and IFN-λ1 levels in a large cohort of population control subjects. A proportion of control subjects had detectable—and some had high—levels of the investigated cytokines. The study design included population control subjects matched with patients with SLE for age, sex, and geographical region (i.e., individuals with other autoimmune diseases, malignancies, or chronic infection [hepatitis] were included). This might partially explain a fairly high proportion of control subjects with increased cytokine levels. IFNs are part of a physiological immune response against viruses, and no screening for possible subclinical viral infections was performed at inclusion, though no control individuals with acute infections were recruited. Our findings implicate that tracking and interpreting upregulation of IFNs in humans is complicated because these cytokines are part of physiological antiviral protection. Upregulation of IFN-α and IFN-λ1 is not specific for SLE. However, we observed certain associations among SLE phenotypes and cytokine patterns. Our findings might be of interest while tailoring treatment for some SLE subsets.
Until recently, commercial ELISAs for IFN-α detection were assumed to be not sensitive enough to detect circulating IFN-α. The majority of available data on type I IFNs in SLE is therefore based on assessments of IFN-regulated gene expression scores (signatures), either in patients’ own cells or in cell lines exposed to serum from patients with SLE. The readout of these assays is upregulation of combinations of IFN-regulated genes. These methods are indirect and unspecific, and studies are therefore difficult to interpret and compare [
41,
42]. Moreover, the signatures of types III and I IFNs overlap; for example,
IFIT1,
IFI44,
STAT1,
CXCL-10/IP-10,
IP-9, and
Mx1 are all induced by both IFN types. Therefore, an impact of IFN-λ subtypes should be considered in the context of a positive IFN signature in SLE [
7,
17,
43,
44]. Various impacts of IFN-λ and IFN-α subtypes could also explain contrasting clinical associations in previous reports [
7,
8,
23,
45]. The pan-IFN-α and pan-IFN-λ1 ELISAs used in this study represent a novel, direct approach to measuring circulating types I and III IFNs. We observed similar clinical associations as reported in an earlier study in which investigators used a direct method to measure IFN-α [
8]. This is reassuring and suggests that this method could serve as a direct, cost-effective way to measure circulating IFNs. Further evaluation is needed to confirm its reliability. To increase accuracy, we limited our comparisons to groups with substantially high IFN levels.
We analyzed the possible impact of steroids and/or DMARDs, but we could not identify any associations. A weakness of this study is the cross-sectional approach whereby the majority of patients were included in a stable phase of inactive or low-activity disease. Inclusion of patients at SLE diagnosis or during active flares is planned.
Importantly, therapeutic blockade of the IFN-α pathway by the IFN-α receptor (IFN-AR) antagonist anifrolumab is being evaluated in clinical trials as a novel therapy for SLE [
21]. Investigators reported that patients with an IFN signature and rash respond better to this treatment. Our findings suggest that probably not all patients with SLE will be candidates for targeting the type I IFN pathway. The results of this study imply that patients with SLE nephritis may derive limited benefit from IFN-AR blockade but could instead be candidates for targeting either type III IFNs or the Th17 axis.
Acknowledgements
We are grateful to Eva Jemseby for management of blood samples; to Jill Gustafsson, Susanne Pettersson, and Sonia Möller for coordination and blood sampling; and to Ola Börjesson and Marika Kvarnström for inclusion of patients and control subjects. We also thank Marie Wahren-Herlenius for providing access to the laboratory equipment and Leonid Padyukov for help with the statistical analysis.