The immortalized N/TERT keratinocyte cell line was used with the kind permission of Dr James G Rheinwald (15). N/TERTs were grown in Keratinocyte-SFM medium (ThermoFisher #17005–042) supplemented with 30 µg/ml bovine pituitary extract, 0.2 ng/ml epidermal growth factor and 0.3 mM calcium chloride (16) and passaged at ∼60% confluency to avoid differentiation. For treatment with Poly(I:C) and UVB, keratinocyte cultures were grown to 80% confluence and then treated with Poly(I:C) at 10 ug/ml, or irradiated in PBS with 50 mJ/cm2 UVB (310 nm) via a UV-2 irradiator (Tyler Industries, Alberta, Canada) followed by media replacement and harvest at indicated time points. Primary keratinocytes were isolated and cultured from 6 mm punch biopsies as we have previously reported (4, 5).

The generation of knock-out (KO) cell lines using non-homologous end joining (NHEJ) via CRISPR/Cas9 and baricitinib treatment to permit efficacious transfection was described previously (4, 17) and was completed with the assistance of the Functional Analytics CRISPR Core of the University of Michigan Skin Biology and Diseases Resource Center. Briefly, for the IFNB1 KO, the following oligonucleotides were used for annealing: IFNB1PCRF1: TGCTCTGGCACAACAGGTAG, IFNB1PCRR1: AGTCTCATTCCAGCCAGTGC. For the Mock−/− line, a sgRNA was designed for the tubulin alpha pseudogene using the following oligonucleotides: GTATTCCGTGGGTGAACGGG. The annealed oligonucleotides were inserted into the cloning vector pSpCas9 (BB)−2A-GFP (PX458) (Addgene # 48138) following the Ran et al. (18) protocol. Ligated plasmids were transformed into competent Escherichia coli (ThermoFisher Catalog #C737303) and then plated on LB-agar. Twelve colonies from each of the groups were selected and cultured in LB medium, and plasmids were purified using Qiagen mini-prep kit (Cat #27106), and then the proper insertion of sgRNA target sequences were verified by Sanger sequencing. Purified plasmid was transfected into an immortalized keratinocyte line (N/TERTs) using the TransfeX transfection kit (ATCC, Cat# ACS4005). Single cells positive for green fluorescent protein (GFP) were sorted into 96-well plates using a MoFlo Astrios #1 cell sorter and grown up to ∼50% confluence. Cells from 96-well plates were transferred into 12-well plates and grown to 50% confluence. DNA was extracted and PCR-amplified using specific primers. Homozygous or heterozygous IFNB1 mutations were verified by Sanger sequencing of the PCR product. For validation of findings, a total of four independent CRISPR/Cas9 KO mutants were generated for IFNB1. Similar experimental design and method was used to generate the STAT1 KO by CRISPR/Cas9 as described previously (19).

Total RNA was isolated with an RNeasy Mini kit (Catalog no. 74,104; Qiagen, Germantown, MD) according to the manufacturer's instructions. One microgram of total RNA was reverse transcribed using the iScript™ cDNA Synthesis Kit (Bio-Rad, Catalog# 1708891), and RT-qPCR was performed on a 7900HT Fast Real-time PCR system (Applied Biosystems) as described previously (4). All primer sequences are listed in Supplementary Table S1.

Total cellular extracts were prepared in HEPES buffer containing protease inhibitor mixture (Catalog no. 11836170001; Sigma-Aldrich) as described previously (20). Cell lysates were then gently resuspended and incubated at 4°C with gentle rocking for 40 min to 1 h, followed by microcentrifugation at 13,000 rpm for 10 min at 4°C. The supernatants were transferred to new tubes, and protein concentrations were determined by Bradford method using Bio-Rad Protein Assay Dye (Cat# 5000006) and serial dilution of BSA as standard. Proteins were separated by SDS-PAGE and transferred onto polyvinylidene difluoride membranes, and immunoblotting was performed using antibodies to IFNB1 (Cat#73671), P-STAT1 (Tyr701, Cat#9167), STAT1 (Cat#14995) and β-actin (Cat#4967) from Cell Signaling Technology. The primary and secondary antibodies were diluted with signal enhancer HIKAR solution 1 and solution 2, respectively, as described previously (20). Detection by enhanced chemiluminescence was carried out with a SuperSignal West Dura kit (Thermo Fisher Scientific, Catalog #34075), followed by imaging on Omega Lum C (Aplegen). The ratio of phosphorylated to total protein was quantified using Image J.

Poly(I:C) was purchased from Tocris Bioscience™ (Cat# 42-871-0). IFN-beta Mouse anti-Human, Clone: 76,703, was from R&D Systems™ (Cat#MAB814100). Cell supernatants were collected after 1 to 24 h post Poly(I:C). IFN-κ levels in keratinocyte supernatants were measured via ELISA, according to the manufacturer's protocol (#MBS936153, MyBiosource, San Diego, CA). IFN-β level was determined using Human IFN-β ELISA Kit (from PBL Assay Science, Catalog# 41410).

In silico analysis was used to identify IFNK-correlated genes and motifs enriched in the IFNK and IFNB1 upstream sequences. IFNK-correlated genes were identified by evaluating gene expression across 118 healthy control KC cell line or primary cell microarray samples using methods previously described (21) The 118 samples had been generated using the same commercial microarray platform (Affymetrix Human Genome Plus 2.0 array) and were compiled from 13 Gene Expression Omnibus series submissions (GSE7216, GSE18590, GSE21364, GSE21567, GSE27186, GSE30355, GSE32685, GSE33495, GSE33536, GSE34528, GSE36222, GSE36287 and GSE37637). Microarray samples were generated using RNA from cultured KCs following various forms of treatment (e.g., genetic mutations, siRNA knockdown, cytokines) see Supplementary Table S2.

13,007 genes were identified with detectable expression (p < 0.05, Signed rank test) in at least 5% (≥6/118) of the 118 microarray samples. Of these 13,007 genes, expression of 7,856 was positively correlated with IFNK (rs > 0). Spearman rank correlation estimates varied continuously among these 7,856 genes. To determine an appropriate threshold, a graphical approach was used to identify a critical point in the decay of correlations among the 7,856 genes (Supplementary Figure S1A). This defined a set of 570 genes with IFNK-correlated expression (rs ≥ 0.60). Enrichment of biological processes among the set of 570 IFN-correlated genes (Supplementary Figure S1E) was evaluated using a conditional hypergeometric test (22).

The upstream sequences of IFNK and IFNB1 were further inspected to identify matches to 2,935 binding sites known to interact with transcription factors or unconventional DNA binding proteins (23). The analysis was repeated with respect to regions 1 kb, 2 kb, 5 kb, and 10 kb upstream of both genes. The total number of motif occurrences in the IFNK1 and IFNB1 upstream regions was also evaluated for each of the 2,935 binding sites and Fisher's exact test was used to assess whether the number of occurrences differed significantly between IFNK1 and IFNB1 (Supplementary Tables S3–S6). The consensus motif sequence is shown for each motif along with the International Union of Pure and Applied Chemistry (IUPAC) nucleotide code.

For statistical calculations, we used GraphPad Prism version 7. For most studies, we compared mean values for experimental variables between groups using the Student's unpaired 2-sided t-test for normally distributed variables, and 1-way ANOVA for multiple comparisons. In all experiments, p less than 0.05 was considered statistically significant.

IFN-κ is a critical IFN in cutaneous lupus pathology via promotion of enhanced IFN responses and photosensitivity (4). To study the regulation of IFN-κ, we first examined a time course of the upregulation of type I IFNs in the N/TERT human keratinocyte line. When cells were stimulated with poly(I:C), a mimic RNA activator of RIG-I, MDA5, and TLR3 pathways, IFNB1 transcription upregulation was robust and rapid, peaking at 3 h after treatment and dropping close to baseline by 12 h (Figure 1A). This is consistent with other cell types in which poly(I:C) induces a rapid upregulation of IFNB1 in an IRF-3 dependent manner (24). Surprisingly, IFNK upregulation was much slower, with increased transcription starting around 12 h after stimulation and peak expression noted at 24 h after poly(I:C) stimulation (Figure 1B). Activation of IFNA transcription (as measured by a poly-specific primer for all alpha subtypes) was less robust and also delayed until 24 h after stimulation (Figure 1C). A time course of protein production of IFN-β and IFN-κ also followed a similar timeline (Figures 1D,E). These data suggest that regulation of IFNK occurs through an alternate, slower mechanism from IFNB, which is rapidly upregulated following treatment with poly(I:C).

UV radiation is an important stimulus for type I IFN production in keratinocytes and is also an important trigger for SLE-related skin disease. We thus next examined whether the kinetics of IFNB and IFNK regulation were similar in keratinocytes after UVB exposure. N/TERT keratinocytes were treated with increasing doses of UVB followed by RNA harvest at the indicated time points. Similar to poly(I:C), IFNB1 transcription started early after 20 or 50 mJ/cm² stimulation but upregulation of IFNK did not peak until 24 h after treatment (Figures 1F,G). Similar results for timing of IFNB and IFNK expression in response to poly(I:C) and UV light were obtained in human primary keratinocytes (Supplementary Figure S1).

The delay in IFNK upregulation suggested that a second signal from the initial stimulus was needed for its upregulation. Given that IFN genes can be activated by other IFNs, we first tested whether signaling through the IFNAR receptor was required. We pretreated the N/TERTs with a blocking IFNAR antibody or control IgG one hour before and during poly(I:C) stimulation. RNA was harvested for analysis at 3 and 24 h. As shown in Figure 2A, anti-IFNAR antibody treatment strongly blocked IFNK upregulation but had much smaller (but significant) effects on overall IFNB1 expression. We then tested to see whether IFN-β treatment of N/TERTs could indeed result in IFNK upregulation. As shown in Figure 2C, IFNK expression was increased after 2 h of IFN-β stimulation, similar to the upregulation of known interferon stimulated genes (ISGs) IRF7 and MX1 (Figure 2C). We then tested whether neutralization of IFN-β in N/TERT cultures blocked the upregulation of IFNK. Indeed, use of IFN-β antibody decreased the upregulation of both IFNK and IRF7 expression after Poly(I:C) (Figure 2D). Together, these data suggest that the induction of IFNK may be through IFNB signaling.

We next took a genetic approach to examine the relationship between IFNB and IFNK. To this end, we generated a knockout line in N/TERTs for IFNB1 and compared it to a line in which a non-coding region was targeted by CRISPR (“mock”). Both lines were generated with the use of baricitinib to avoid selection of low IFNK expression as previously demonstrated (17). As shown in Figure 3A, the basal expression of IFNB1 is very low, but was detectable at 3 and 6 h when cells were induced by Poly(I:C). No expression of IFNB was detected even with stimulation in four knockout lines of IFNB (B1KO #13, #14, #17, and #18), which validated the IFNB1 gene knockout (Figure 3A and Supplementary Figure S2). We then compared Mock−/− and two IFNB1 KO lines to investigate the effect of loss of IFNB1 on IFNK and ISG expression. Strikingly, deletion of IFNB1 resulted in a complete inhibition of IFNK upregulation after poly(I:C) and after UVB treatment (Figures 3B,E, respectively). This also resulted in a significant reduction in the expression of downstream IFN genes such as MX1, IRF7 and ISG15 (Figures 3C,D,F–H). These data suggest that in response to stimuli such as UVB and poly(I:C), IFN-β production is rapid and required for further downstream IFN production. Indeed in this context, IFNK behaves like an ISG.

The type III interferon, IFNλ, is important for epidermal defense against viruses. It is produced by, and acts on, keratinocytes. IFNλ is also produced in the skin in established SLE, and it has structural resemblance to IL-10, but similar signaling and downstream cellular effects to type I interferons (3, 25). We thus also assessed IFNL1 and IFNL3 (we do not detect IFNL2 or IFNL4 in N/TERTs). Our data identified a similar pattern of IFNL3 and IFNB1 regulation where both are early stimulated genes after poly(I:C) treatment. In contrast, IFNL1 behaved like IFNK where expression was seen 24-h post treatment (Figures 4A,B). Similarly, expression of IFNL1 was entirely dependent on IFNB1 whereas only the later expression of IFNL3 had any dependence on IFNB1. Taken together, these data suggest that IFN-β signaling may be also required for IFN-λ1 but not early IFN-λ3 production in keratinocytes.

Given the difference in time courses of IFNB and IFNK upregulation as well as the essential role of IFN-β in the production of IFN-κ, we next analyzed publicly available microarray data (n = 118 samples, see Supplementary Table S2) generated from cultured keratinocytes and identified 7,856 genes positively correlated with IFNK expression and with detectable expression in at least 5% of microarray samples (Supplementary Figure S3). We defined a set of 570 IFNK-co-expressed genes and identified several overrepresented IFN-related biological processing GO terms (Supplementary Figures S1D,E). Indeed, the most closely correlated genes were known ISGs such as MX1 and 2, IFI44, and ISG15. These data support IFNK as an ISG. We then examined the promoters of IFNB1 and IFNK to determine whether regulatory mechanisms for each IFN's activation would be apparent. For this, we used a screen of 2,935 binding sites known to interact with transcription factors or unconventional DNA binding proteins (23) and focused in on interferon regulatory factor (IRF) binding sites, which are important for both interferon production (especially IRF3) and ISG induction (especially IRF9). As shown in Supplementary Figure S1F, both IFNB and IFNK promoters exhibited binding sites for IRF9 (isgf3f motif), consistent with both genes being activated by signaling through the type I IFN receptor via activated STAT1/2/IRF9 complex formation. However, when we analyzed for binding sites for IRF3, which drives IFN production downstream of poly(I:C) and UV, there were 3 predicted sites for IFNB1 and no predicted sites for IFNK at −5 kb. Further away, at −10 kb total, 3 and 2 were predicted, respectively. While we cannot rule out overlap between these predicted sites, these data suggest that IFNB1 upregulation is more expedient downstream of activation by STING-driven IRF3 activation as there are IRF3 binding sites near the transcription start site whereas IFNK does not have similar availability for IRF3 binding.

Type I interferons bind to a shared cell surface receptor, the type I interferon receptor (IFNAR). IFNs binding to the IFNAR initiate activation of Janus kinase 1 (JAK1) and tyrosine kinase 2 (TYK2), leading to phosphorylation, dimerization and nuclear translocation of STAT proteins. We thus investigated the activation of type I IFN signaling after poly(I:C) and UV stimulation in the presence or absence of IFNB. Strikingly, the deletion of IFNB resulted in decreased activation of STAT1 after poly(I:C) (Figures 5A–D). Less robust activation of STAT1 was noted after UVB at the 6 h time point shown (and at 3 h not shown) (Figures 5E–H). Loss of IFNB resulted in a reduction in total STAT1 expression (Figures 5A,C,E,G and Supplementary Figure S4). We then examined IFN and ISG expression in STAT1 KO KCs. Consistent with our hypothesis, deletion of STAT1 decreased production of IFNK and the ISGs IRF7, ISG15, and MX1 for both poly(I:C) and UVB treatment (Figures 6A,B). However, expression of IFNB1 was not significantly affected by deletion of STAT1 and may have been enhanced, especially with poly(I:C) treatment. These data suggest that IFNB is produced upstream of STAT1 activation, likely through IRF3-dependent means, while IFNK transcription is downstream of IFN-β signaling and relies on IFNAR and STAT1 (Figure 7).