Purpose Type I interferon (IFN-I) assays are an important emerging biomarker in SLE. A EULAR Task Force for IFN-I assays proposed a research agenda, which included key questions concerning the use of IFN-I stimulated gene (ISGs) and protein assays. These include: (i) influence of sample type; (ii) validated reference genes that are not influenced by IFN-I; (iii) for both gene expression and protein assays: confirmation that ISGs or proteins are specifically responsive to IFN-I, not IFN-II.

Methods To compare sample types, we extracted RNA from paired whole blood (TEMPUS) and PBMC samples obtained from 10 healthy controls, 7 at risk non-progressor, 6 at risk progressor, 10 inactive SLE and 12 active SLE. These were assessed for published IFN Scores A (module 1.2 and 3.4) and Score B (modules 3.4 and 5.12) using TaqMan. Bland-Altman agreement plots compared results.

To assess reference genes, we analysed the same set of samples for expression of a panel of 16 reference genes from the literature using TaqMan. Results were analysed using RefFinder software.

To compare the specificity of IFN response of candidate ISGs and proteins, we performed in vitro stimulation of healthy whole blood and PBMCs using IFN-α, IFN-β, IFN-κ, IFN-λ, IFN-γ, IL-1, IL-6, IL-10 and TNF. Samples were analysed at 0, 6 and 24h using TaqMan for a panel of 26 ISGs (summarised as IFN-Score-A and IFN-Score-B) as well as the transcripts for 17 IFN-stimulated proteins (mostly chemokines) described in the EULAR Systematic Literature Review.

Results IFN-Score-A correlated well between whole blood and PBMC samples (r2=0.8614). IFN-Score-B showed weaker correlation (r2=0.2024) and Bland-Altman plot showed greater deviation from line of agreement than for Score A.

There were marked difference in stability of published reference genes. Across several algorithms, the most consistently stable genes were: YWHAZ, PGKI and GUSB. The least stable were RPLPO, HMBS and ACTB (β-actin). Calculation of IFN Scores using the least stable reference genes demonstrated greater variability between samples and poor separation of SLE and HC samples compared to calculation using the most stable reference genes.

IFN-α strongly induced IFN-Score-A, IFN-Score-B and expression of CCL3, CCL4, CCL5, CCL7, CCL8 and CXCL12 as compared to IFN-γ. CXCL9 and CCL26 were more responsive to IFN-γ stimulation than IFN-α. CCL2, CCL19, CCL20, CCL21, CCL23 and CXCL11 demonstrated similar levels of response to IFNs. Expression of CXCL1, CXCL8 and CCL13 were suppressed by IFN-α. The chemokine transcripts CCL2, CCL7, CCL13, CCL20, CCL23, CXCL1 and CXCL8 were more responsive to IL-1 than IFN-α. CCL3, CCL19, and CCL21 responded to IL-1 similarly to IFNs.

Conclusions (i)The relative expression of different sets of ISGs varies between PBMCs and whole blood sample types. (ii)Some reference genes used in published IFN-I assays are not stable. (iii)Some gene expression and serum protein assays reported to measure IFN-I include components that are either not ISGs, or are more responsive to IFN-II or other cytokines than IFN-I. Our future work will develop a whole blood IFN-I assay optimised to avoid these artefacts and confounders.