• lupus erythematosus
• systemic
• autoimmune diseases
• autoimmunity

Dysregulation of type I interferon (IFN-I) signalling plays a major role in systemic lupus erythematosus (SLE) pathogenesis.1 Selected IFN-stimulated genes (ISGs) are used to generate scores and were shown to be associated with specific clinical phenotypes, SLE activity, risk of flares and response to treatment targeting IFN-I.2 3 IFN-I gene scores are highly heterogeneous in the number of included ISGs and are not standardised for the use in routine clinical practice. Serum IFN-α levels detected by digital ELISA by single molecule array were shown to be a promising biomarker of SLE activity4 and predictor of flares among patients with SLE in remission.5 IFN-γ may also play a role in SLE pathogenesis and it has been shown that several genes that are upregulated by IFN-α are upregulated also by IFN-γ.6 In the present study, we aimed at assessing whether IFN-I gene score in blood and IFN-α or IFN-γ levels quantified by digital ELISA in serum performed similarly as biomarkers, mirroring the clinical activity of SLE. Moreover, we investigated by correlative evidence the contribution of IFN-α and IFN-γ to the expression levels of different ISGs and of an IFN-I gene score.

Gene expression was assessed by mRNA profiling using the NanoString nCounter gene expression system (NanoString Technologies, Seattle, Washington). Serum IFN-α and IFN-γ levels were quantified by digital ELISA technology (Quanterix Simoa, Lexington, Massachusetts, USA). Detailed methodology is available in online supplemental document S1. The clinical characteristics of the 133 patients with SLE included in the present study are reported in online supplemental table S1. Median age was 45.6 (range 19–78.8) years, 111 (83%) were women, 98 (74%) were Caucasians and 75 patients (56%) had an active disease using clinical Systemic Lupus Erythematosus Disease Activity Index (cSLEDAI), the contribution of low serum complement and elevated anti-dsDNA autoantibodies with a cut-off>0 to define active disease was excluded.4

Using the predefined cut-offs,3 4 the prevalence of high IFN-I gene scores, elevated IFN-α and IFN-γ serum levels were 44% (58/133), 45% (60/133) and 14% (18/133), respectively (figure 1A). Serum IFN-α levels showed a highly positive correlation with the IFN-I gene scores (Spearman’s correlation coefficient: rho=0.82), as well as with the expression level of individual ISGs except for CXCL10 (figure 1B,C). In contrast, IFN-γ levels showed a weak positive correlation with IFN-I gene scores (rho=0.32) (figure 1D) and IFN-α levels (rho=0.35), as well as with the expression level of individual ISG, except for CXCL10 which showed a stronger positive correlation (rho=0.60) in accordance with a preferential induction of CXCL10 by IFN-γ (figure 1C). Using Cohen’s kappa coefficient, serum IFN-α levels showed substantial agreement to classify SLE with high or low IFN-I gene scores κ=0.72 (95% CI: 0.60 to 0.84), whereas the agreement was low for IFN-γ (figure 1B,D). The sensitivity, specificity, negative and positive predictive values of serum IFN-α levels to classify SLE with high or low IFN-I gene score were 86%, 87%, 89% and 83%, respectively.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Prevalence, correlation and agreement between serum IFN-α and IFN-I gene score. (A) Prevalence of high serum IFN-α, serum IFN-γ and IFN-I gene score. A custom panel of 22 genes was developed and we calculated one IFN-I gene score (IFI27, IFI44, IFI44L, RSAD2)2 with a cut-off≥17.5 to define high score.3 Cut-off for high serum IFN-α was 136 fg/mL (blue line) as previously defined,4 cut-off for high IFN-gene score was 17.5 (red line) as previously defined3 and cut-off for high IFN-γ was 2558 fg/mL based on 3 SD above the mean from 74 HC to define elevated IFN-γ levels. (B) Spearman correlation coefficient (rho) between IFN-I gene scores and serum IFN-α values and Cohen’s kappa to assess agreement between IFN-I gene score and serum IFN-α to classify patients with SLE. (C) Correlation matrix diagram of individual IFN genes with serum IFN-α and serum IFN-γ levels. Spearman correlation analysis (rho) was applied and values were condensed in a colour scale. (D) Spearman correlation (rs) between IFN-I gene score and serum IFN-γ and Cohen’s kappa to assess agreement between IFN-I gene score and serum IFN-γ to classify patients with systemic lupus erythematosus (SLE). IFN, interferon.

Moreover, elevated serum IFN-α levels and IFN-I gene scores were associated with active SLE, as defined by cSLEDAI>0 or SLEDAI≥4 (online supplemental figure s1–s3) and were both associated with active skin lesions, arthritis and positive anti-dsDNA Abs in multivariable analysis (online supplemental table s2). In contrast, IFN-γ was neither associated with active SLE (online supplemental figure s1) nor with active SLE characteristics (online supplemental figure s2).

Finally, IFN-I gene score AUC=0.63 (95% CI: 0.53 to 0.72) and serum IFN-α AUC=0.63 (95% CI: 0.53 to 0.72) performed similarly and significantly better than C3 levels AUC=0.42 (95% CI: 0.32 to 0.52) to discriminate inactive versus active SLE adjusted p value=0.03 and 0.03, respectively (online supplemental figure s3 and table s3).

In this study, for the first time, we show that IFN-α assessed by digital ELISA and IFN-I gene score perform equally for identifying the association of IFN-I with SLE disease activity and clinical manifestations. Remarkably, this was specific to IFN-α, since no such association was observed with serum IFN-γ levels. Of importance, we observed no association of IFN-γ serum levels with active SLE clinical features and SLEDAI. This may suggest that IFN-γ serum levels may not perform optimally as SLE biomarkers and may not support the choice of IFN-γ as therapeutic target. However, further studies are needed to explore this issue. The limitations of our study are the cross-sectional design and the relatively low number of highly active patients with SLE, which reflects real-life practice in Switzerland.

IFN-α levels measured by digital ELISA could be easier to standardise than IFN-I gene scores to characterise IFN-I overexpression in clinical practice.