SOCS1 is an inducible negative regulator of interferon λ (IFN-λ)–induced gene expression in vivo

Type I (α and β) and type III (λ) IFNs are induced upon viral infection through host sensory pathways that activate IFN regulatory factors (IRFs) and nuclear factor κB. Secreted IFNs induce autocrine and paracrine signaling through the JAK-STAT pathway, leading to the transcriptional induction of hundreds of IFN-stimulated genes, among them sensory pathway components such as cGAS, STING, RIG-I, MDA5, and the transcription factor IRF7, which enhance the induction of IFN-αs and IFN-λs. This positive feedback loop enables a very rapid and strong host response that, at some point, has to be controlled by negative regulators to maintain tissue homeostasis. Type I IFN signaling is controlled by the inducible negative regulators suppressor of cytokine signaling 1 (SOCS1), SOCS3, and ubiquitin-specific peptidase 18 (USP18). The physiological role of these proteins in IFN-γ signaling has not been clarified. Here we used knockout cell lines and mice to show that IFN-λ signaling is regulated by SOCS1 but not by SOCS3 or USP18. These differences were the basis for the distinct kinetic properties of type I and III IFNs. We found that IFN-α signaling is transient and becomes refractory after hours, whereas IFN-λ provides a long-lasting IFN-stimulated gene induction.

Type I and type III IFNs are induced in virus-infected cells and provide an important first line of defense through the rapid induction of hundreds of IFN-stimulated genes (ISGs) 3 that collectively establish an antiviral state. Type I IFNs (IFN-␣s and IFN-␤) bind to the ubiquitously expressed IFN-␣ receptor (IFNAR), which consists of two major subunits: IFNAR1 and IFNAR2c. Each receptor subunit constitutively binds to a single specific member of the JAK family: IFNAR1 to tyrosine kinase 2 (TYK2) and IFNAR2c to JAK1. Upon binding of the two chains by type I IFNs, TYK2 and JAK1 transactivate each other by mutual tyrosine phosphorylation and then initiate a cascade of tyrosine phosphorylation events on the intracellular domains of the receptors and on STAT1 and STAT2. STAT1 and STAT2 combine with a third transcription factor, IFN regulatory factor 9 (IRF9), to form interferon-stimulated gene factor 3 (ISGF3). ISGF3 binds to interferon-stimulated response elements (ISREs) in the promoters of ISGs. Alternatively, IFN-activated STAT1 can form homodimers. These STAT dimers bind a different class of response elements, the ␥-activated sequence elements (1). Activation of the JAK-STAT pathway is tightly controlled by several negative regulatory mechanisms. SOCS1 and 3 are rapidly induced and inhibit STAT activation (2). Mechanistically, SOCS proteins simultaneously bind the receptors and the JAKs and prevent STATs from access to the receptor kinase complex (3). Ubiquitin-specific peptidase 18 (USP18) is induced later but remains highly expressed for days (4). USP18 is responsible for the long-lasting refractoriness of IFN-␣ signaling in the liver (5). The physiological role of these inducible negative feedback loops is to prevent tissue damage caused by the potent pro-inflammatory effects of type I IFNs (6).
In humans, the type III IFN family consists of four members, IFN-1-4. They all bind to the IFN-receptor (IFNLR), consisting of the ubiquitously expressed IL-10R2 chain (shared with the IL-10 receptor) and a unique IFN-receptor chain (IFNR1) whose expression is mainly restricted to epithelial cells (7)(8)(9). Activation of the IFNLR by ligand binding results in the activation of ISGF3 and STAT1 homodimers, the same transcription factor complexes that are induced by type I IFNs. As a consequence, despite using different receptors, IFN-s induce highly similar sets of ISGs as IFN-␣/␤s (10 -14). However, at least in cell culture, ISG induction followed distinct kinetic profiles for IFN-␣ and IFN-. Although gene induction was rapid and transient with IFN-␣, IFN-induced a slower but more sustained increase in ISG expression (15,16). There is indirect evidence that this difference could result from the lack of USP18 binding to the IFNLR. In cell culture and in mice, USP18 was induced by both IFN-␣ and IFN-, but only IFN-␣ signaling was inhibited by USP18 (17,18). Contradictory to these findings, a more recent report identified USP18 as a novel inhibitor of IFN-signaling (19). The role of SOCS1 and SOCS3 in regulating IFN-signaling remains to be clarified as well. Overexpression of SOCS1 in Huh7 human hepatoma cells inhibited the antiviral activity of both IFN-␣ and IFN- (20). However, whether physiological expression levels of SOCS1 and/or SOCS3 can inhibit IFN-signaling remains to be shown. The IFN-system has been intensely studied in the context of hepatitis C virus (HCV) infections because genetic variants of the IFN-gene locus are strongly associated with the ability of the host response to clear the viral infection in the acute phase and with response to treatment with pegylated IFN-␣ and ribavirin in chronic hepatitis C (21). Genetic evidence points to IFN-4 as the key regulator of the host response to HCV. A variant allele (TT at position rs368234815) with an inactivating frameshift mutation in exon 1 of the IFN-4 gene is found with a frequency that increases from Africa (0.29 -0.44) to the New World (0.51-0.65) and Europe (0.58 -0.77) and reaches 0.94 -0.97 in East Asia (22). Homozygosity for the functionally inactive IFN-4 gene (TT/TT) is associated with spontaneous clearance of HCV in the acute phase and a high cure rate of therapies with pegylated IFN-␣ and ribavirin (23)(24)(25). It is not entirely clear why the host benefits from lack of a functional IFN-4. A pertinent observation relates to the activation status of the endogenous IFN system in the liver. Patients with wildtype IFN-4 (encoded by the rs368234815 ⌬G allele) mount a strong innate immune response and have a permanent and strong expression of hundreds of ISGs in the liver (26). Among these ISGs are also inducible negative regulators of IFN signaling such as USP18. The continuous high expression of USP18 makes it highly unlikely that type I IFNs are drivers of the innate immune response in chronic hepatitis C. Indeed, there is genetic evidence that IFN-4 is the driver of ISG expression in chronic hepatitis C (25). To better understand the mechanisms of ISG induction in chronic hepatitis C, differences in negative regulation of type I and type III interferons have to be elucidated.
In this work, we investigated the role of SOCS1, SOCS3, and USP18 in IFN-signaling in cells with physiological IFNLR expression and deletion or overexpression of all three inhibitors and in mice deficient for SOCS1 or USP18. We identify SOCS1 as a physiologically relevant inducible inhibitor of IFN-induced JAK-STAT signaling. SOCS3 and USP18 are important for regulating IFN-␣ but not IFN-signaling.

IFN-induces sustained gene expression in cells despite strong induction of USP18
The selective and restricted expression of IFNLR1 limits the biological activity range of IFN-primarily to mucosal epithelial tissues (27). We have previously shown very low expression of IFNLR1 in the human liver that can, however, be induced in patients with chronic hepatitis C, particularly with the IFN-4 wild-type genotype (TT/⌬G and ⌬G/⌬G), to levels that make hepatocytes responsive to IFN- (28). To analyze IFN-signaling in a cell line with similar IFNLR1 expression, we used a clone of Huh7 cells (a widely used human hepatoma cell line) that has been stably transfected with IFNLR1, Huh7 LR clone 3 (designated Huh7 LR in this manuscript) (28). Huh7 LR cells express around 10,000 copies of IFNLR1 per 40 ng of total RNA (Fig. 1A). Huh7 LR cells were then stimulated with saturating concentrations of human IFN-␣ and human IFN-1 for up to 48 h. The kinetics of the activation of JAK-STAT signaling were assessed with antibodies specific for tyrosine-phosphorylated STAT1 (p-STAT1), STAT2 (p-STAT2), and STAT3 (p-STAT3). IFN-␣ induced strong but transient STAT1 and STAT2 activation (Fig. 1B). Induction of the two early ISGs, RSAD2 and GBP5 (4), was also transient, with an expression peak 8 h after IFN-␣ stimulation (Fig. 1C). IFN-1 showed prolonged activation of STAT1 and STAT2 despite an even stronger induction of USP18 at the 24 h time point (Fig. 1B). Consistently, ISG induction was sustained and increased compared with IFN-␣ (Fig. 1C). The induction of pSTAT3 was not significantly different between IFN-␣ and IFN-1. These results are in accordance with published data obtained with other cell lines and primary human hepatocytes (11,15,16).

IFNsignaling is inhibited in cells overexpressing SOCS1, SOCS3, or USP18
Huh7 LR cells were then transiently transfected with expression plasmids for SOCS1, SOCS3, and USP18 and stimulated with saturating concentrations of IFN-␣ and IFN-1 for 30 min. All three proteins were strongly expressed and inhibited both IFN-␣-and IFN-1-induced STAT1 phosphorylation ( Fig.  2A). The inhibitory effect of USP18 on IFN-signaling was unexpected, given the sustained p-STAT1 and ISG induction shown in Fig. 1. We therefore systematically compared the expression levels of the negative regulators obtained by physiological stimulation with IFNs versus those obtained in cells transfected with expression plasmid. As shown in Fig. 2B, transfection resulted in expression levels that were 2-3 orders of magnitude higher than those induced by maximal physiological stimulation with IFNs. The mRNA results were also confirmed at the protein level (Fig. 2B). We conclude that all three inhibitors have the potential to inhibit both IFN-␣ and IFN-signaling at supraphysiological expression levels.

SOCS1-deficient cells are hyperresponsive to IFN-
To assess the physiological relevance of SOCS1, SOCS3, and USP18 for IFNsignaling, we generated knockout cell lines from Huh7 LR cells using the CRISPR-Cas9 technology (supplemental Fig. 1). The rationale for this loss-of-function experiment was the expectation that loss of a physiologically relevant inhibitor of IFN signaling must result in hyperactivation of the signal transduction pathway with increased ISG induction. By definition, any inducible inhibitor of IFN signaling has to restrict ISG induction at some point and to a relevant degree. We first tested this hypothesis using a luciferase reporter gene under the control of an ISRE promoter (i.e. the Mx1 promoter, ISRE-Mx1-Luc). To do so, we transfected control and knockout cells with the pGL3-ISRE-Mx1-Luc plasmid together with a constitutively expressed Renilla luciferase construct for normalization. 20 h after transfection, they were stimulated with IFN-1, IFN-3, IFN-4, IFN-␣, and IFN-␤ or left untreated, and luciferase activity was quantified 4, 8, and 24 h later. The luciferase activity in untreated cells did not differ significantly between cell lines but increased at least 15-fold within 4 h after IFN addition. At 4 h and 8 h, but not at 24 h, SOCS1 knockout cells showed significantly higher expression of the reporter constructs when stimulated with IFN-1, IFN-3, or IFN-4 ( Fig. 3 and supplemental Fig. 2). IFN-␣ signaling was enhanced in USP18 knockout cells at all time points (Fig. 3), whereas IFN-␤ signaling was enhanced at the 8 h time point only (supplemental Fig. 2). Type I IFN-induced expression of the reporter plas-Negative regulation of type III interferon signaling mid was also significantly enhanced at some time points in SOCS1 and SOCS3 knockout cells ( Fig. 3 and supplemental Fig.  2). We confirmed these findings by quantifying the expression levels of ISGs, including RSAD2, GBP5, and IFI27, as well as the three inducible regulators, SOCS1, SOCS3, and USP18, in the knockout cell lines after stimulation with IFN-1 and IFN-␣ ( Fig. 4 and supplemental Fig. 3). Expression of RSAD2, GBP5, and IFI27 in untreated control and knockout cells was detectable and did not differ significantly between the cell lines. The same ISGs were up-regulated 10-to 100-fold 4 h after IFN-␣ or IFN-1 stimulation (Fig. 4). Importantly however, IFN-1induced ISG expression was significantly enhanced in SOCS1 knockout cells at 8 and 24 h, but this increase was not observed in SOCS3 or USP18 knockout cells. For GBP5, this enhancement could already be observed 4 h after IFN-1 stimulation, although it did not reach statistical significance. Furthermore, we confirmed the known negative regulatory effect of USP18 on IFN-␣ signaling. We conclude that, in Huh7 LR cells, SOCS1 is an inducible and physiologically relevant negative regulator of IFN-1, IFN-3, and IFN-4.

Figure 1. IFN-␣-but not IFN--induced STAT1 phosphorylation becomes refractory to continuous stimulation.
A, liver biopsies from chronic hepatitis C patients (n ϭ 16) were divided into three groups based on their IFNL4 genotype (rs368234815; TT/TT, TT/dG, and dG/dG). Total RNA from biopsies and Huh7 and Huh7 LR cells were prepared. Expression of the IFNLR1 transcript was analyzed by quantitative PCR. Results (mean Ϯ S.D.) are shown as copy numbers per 40 ng of total RNA. B and C, Huh7 LR cells were stimulated with 1000 IU/ml IFN-␣ or 100 ng/ml IFN-1 for the indicated times. B, p-STAT1, STAT1, p-STAT2, STAT2, p-STAT3, STAT3, USP18, SOCS1, SOCS3, and actin were visualized using specific antibodies. Shown are representative blots from two independent experiments. C, transcripts of interferon-stimulated genes (RSAD2, IFI27, and GBP5) were quantified by PCR. Results (mean Ϯ S.D., n ϭ 3) are shown as relative expression to GAPDH. ut, untreated.

Negative regulation of type III interferon signaling SOCS1 is a physiological inhibitor of IFN-signaling in vivo
To test whether the results obtained in Huh7 LR cells are valid in vivo, we analyzed IFNand IFN-␣ signaling in mice deficient for Socs1 or Usp18 and their corresponding control mice. Because Socs1 knockout mice suffer from IFN-␥-mediated lethal toxicity, they were bread as Ifng/Socs1 double knockout mice, with Ifng knockout mice serving as controls, as described previously (5,29). The mice were sacrificed 4 and 8 h after IFN injections, and the liver, lung, and gut were analyzed for induction of seven different ISGs (Rsad2, Oas1, Stat1, Usp18, Gbp5, and Ifi27). Socs1 knockout resulted in significantly increased expression of all IFN-2-induced ISGs at 8 h in the lung and at 4 h in the gut, the two IFN-responsive organs, but not in the liver, which is devoid of IFNLR1 in mice (Fig. 5). Although Socs1 knockout also resulted in enhanced expression of three of six ISGs in the lung and gut upon IFN-␣ injection (Fig. 5), it did not affect IFN-␣-mediated ISG induction levels in the liver (Fig. 5 and supplemental Fig. 4A). In contrast, IFN-␣ induced all tested ISGs to significantly higher levels in the livers of Usp18 knockout mice compared with control mice at all time points (Fig. 6 and supplemental Fig. 4B). The effect of Usp18 knockout on IFN-␣ signaling in the lung occurred mainly at 8 h, and it was much less pronounced in the gut. The effect of Usp18 knockout on IFN-signaling was minimal, with only Rsad2 and Ifi27 being significantly increased at one time point in the gut and lung, respectively ( Fig. 6 and supplemental Fig. 4B). From these results, we conclude that IFN-␣ signaling in these mice is predominantly inhibited by Usp18, whereas IFN-signaling is regulated by Socs1.

Discussion
Type I IFNs (IFN-␣/␤) are potent and critically important cytokines that control innate and adaptive immune responses to infection, cancer, and other inflammatory stimuli. Positive feedback amplification through autocrine and paracrine induction of IFN-␣ gene transcription allows a very rapid and strong host response to infections (30). The extent and duration of this response is tightly controlled by several mechanisms to avoid adverse effects on tissue homeostasis. The fact that several clinically important autoimmune diseases, such as systemic lupus erythematosus and Aicardi-Goutières syndrome, are associated with uncontrolled type I IFN activities demonstrates the important role of negative regulators in keeping the system in balance (6,31). SOCS1 is an inducible negative regulator of IFN-␣ signaling that transiently restricts phosphorylation of STATs in the first hours. However, its role as a negative regulator of IFN-␣ signaling is limited. The fact that the lethal phenotype of Socs1 knockout mice is rescued by additional knockout of Ifng demonstrates that the main role of SOCS1 in the IFN system is to control IFN-␥ (29). Furthermore, deletion of Socs1 does not prevent the induction of refractoriness to IFN-␣ in the mouse liver (5). The main negative regulator of IFN-␣ seems to be USP18. This is demonstrated by this work, by previous work that showed that refractoriness to IFN-␣ stimulation is abrogated in Usp18 knockout mice (5), and by a rare genetic interferonopathy, pseudo-TORCH syndrome, which is caused by human USP18 deficiency (32).
Negative regulation of type III IFNs is fundamentally different from type I IFNs. The inducible up-regulation of USP18 does not induce refractoriness of IFNsignaling (this work and Refs. 11, 15-18).). Of note, as shown in Fig. 2, USP18 indeed has the potential to inhibit IFNsignaling, but only at expression levels that are not achieved by physiological induction

Negative regulation of type III interferon signaling
stimuli. We conclude that USP18 is not a physiological inhibitor of IFNsignaling. IFN-signaling is rather controlled by SOCS1, and contrary to IFN-␣, not only in the first hours of stimulation but also at later time points (Fig. 4). Of note, a long-lasting negative regulatory effect of SOCS1 on IFN-1 signaling has been found previously in shRNA knockdown experiments in A549 human alveolar epithelial cells (33). SOCS1 not only controls the closely related IFN-1 and IFN-3 but also IFN-4. We postulate that the prolonged inhibitory effects of SOCS1 on IFNsignaling result from its sustained up-regulation for at least 48 h, which contrasts the transient induction of SOCS1 by IFN-␣ (Fig. 2B).
In conclusion, we show that the IFN-␣ and the IFN-systems not only differ in terms of tissue distribution of their cognate receptors (27) but are also controlled by distinct negative regu-latory mechanisms of their signal transduction through the JAK-STAT pathway. IFN-␣ signaling is transient and shut down after 6 -8 h by USP18, a very strong inhibitor of STAT tyrosine phosphorylation at the IFNAR-kinase complex. IFNsignaling is not affected by USP18. It is controlled by SOCS1, but SOCS1 does not shut down STAT phosphorylation completely, allowing long-lasting stimulation of ISG transcription by IFN-. We postulate that these differences between IFN-␣ and IFN-in the negative regulation of JAK-STAT signaling are responsible for the previously described differences in the kinetics of ISG induction. Long-lasting activation of JAK-STAT signaling allows the IFN-system to mount a sustained antiviral state in mucosal epithelial cells that are constantly exposed to pathogens. IFN-␣ is a more powerful defense system that is activated when pathogenic viruses have breached

Negative regulation of type III interferon signaling
the mucosal surfaces and invaded the systemic circulation and other organs. In most instances, it is only transiently activated because prolonged activation of IFN-␣ signaling is detrimental for tissue homeostasis and can also negatively impact the cellular immune response.

Figure 5. Depletion of Socs-1 increased IFN--induced ISGs expression in vivo.
Control and Socs1 Ϫ/Ϫ mice were subcutaneously injected with PBS, 1000 units/g mouse IFN-␣, or 50 ng/g mouse IFN-2. The liver, the lung, and the gut were collected 4 h and 8 h after injection, and total RNA was prepared. The expression of Rsad2, Oas1, Stat1, Usp18, and Socs1 was measured by quantitative PCR. The results (mean Ϯ S.D.) are shown as relative expression to Rpl19. Three to four animals were used per time point and condition. Unpaired t test with Welch's correction; *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001. ut, untreated.

Negative regulation of type III interferon signaling
mal facility of the Department of Biomedicine of the University Hospital Basel under specific-pathogen-free conditions on a 12-h day and 12-h night schedule with ad libitum access to food and drinking water. Experiments were conducted with the approval of the Animal Care Committee of the Canton Basel-Stadt, Switzerland. Six-to eight-week-old males were used, and the animals were euthanized by CO 2 narcosis. Resected organs were immediately frozen in liquid nitrogen and kept at Ϫ70°C until further processing. Subcutaneous injections with PBS, mouse IFN-␣, or mouse IFN-2 were performed between 8 a.m. and 5 p.m.

Human liver biopsies
Liver biopsies from chronic HCV-infected patients (n ϭ 16, patient characteristics are shown in supplemental Table 1) were obtained in the outpatient clinic of the Division of Gastroenterology and Hepatology, University Hospital Basel, Switzerland. Biopsy material that was not needed for routine histopathology was used for research purposes after obtaining written informed consent. The use of biopsy material for this project was approved by the Ethikkommission Nordwest-und Zentralschweiz, Basel, Switzerland, protocol number M989/99. Total DNA was isolated from liver biopsies using the DNeasy Blood & Tissue Kit (Qiagen, Hombrechtikon, Switzerland) according to the instructions of the manufacturer. IFN-4 genotype was determined as described previously (25).

Whole-cell lysates and Western blotting analysis
Whole-cell lysates and immunoblots were prepared and used as described previously (38).

Total RNA extraction and quantitative PCR
Total RNA from cell lines was isolated using NucleoSpin RNA (Macherey-Nagel AG, Oensingen, Switzerland) according to the instructions of the manufacturer. cDNA was gener-ated from 1 g of total RNA using Moloney murine leukemia virus reverse transcriptase (Promega AG, Dübendorf, Switzerland). Total RNA from human biopsies and mouse samples was isolated using TRIzol according to the instructions of the manufacturer. 1 g of total RNA was incubated with rDNaseI using the DNA-free kit (Ambion). cDNA was generated using the TaqMan reverse transcription reagent kit (Applied Biosystems) according to the recommendations of the manufacturer. Realtime quantitative PCR was performed using FastStart Universal SYBR Green Master Mix (Roche Diagnostics AG, Rotkreuz, Switzerland) and the ABI 7500 detection system (Applied Biosystems, Thermo Fisher Scientific, Zug, Switzerland). Primers are listed in supplemental Table 2. The specificity of the PCR primers was assessed by sequencing the PCR product. Gene transcript expression levels were calculated using the ⌬Ct method. To quantitate IFNLR1 transcript levels, dilutions of plasmids containing the IFNLR1 ORF (28) were used as standard curves (dilutions ranged from 10 8 to 10 0 copies of plasmid).

SOCS1, SOCS3, and USP18 overexpression
The pCMV6 plasmid containing the SOCS1 gene (RC220847) was purchased from Origene Technologies Inc. (Rockville, MD) and used to overexpress SOCS1. Human USP18 and human SOCS3 coding sequences were cloned into the pCMV6-entry and pCMV-MIR vectors, respectively, using XhoI and BamHI restriction sites for USP18 and XhoI and KpnI restriction sites for SOCS3. Huh7 LR cells (2 ϫ 10 5 /well) were seeded onto a 6-well plate, and 2 g of expression plasmid was transfected using JetPrime TM (Polyplus Transfection, VWR International GmbH, Dietikon, Switzerland) according to the instructions of the manufacturer.

Generation of SOCS1, SOCS3, and USP18 knockout cell lines
Briefly, the CRISPR design tool from the Zhang laboratory (http://crispr.mit.edu) 4 was used to design single-guide RNA constructs (supplemental Table 3). Phosphorylated and annealed single-guide RNAs were cloned into pSpCas9(BB)-2A-GFP (PX458) (Addgene, 48138, deposited by Feng Zhang) using BbsI restriction sites. Plasmids were verified by sequencing. 48 h post-transfection, GFP-positive cells were single cellsorted by FACS. To confirm successful gene targeting in sorted clones, genomic DNA was extracted using the DNeasy Blood & Tissue Kit (Qiagen AG) and subjected to PCR amplification using specific primers for the region of interest (supplemental Table 2). PCR fragments were then cloned into a pGEMT-easy expression vector (Promega AG) and transformed into Escherichia coli TOP-10 chemically competent cells (Invitrogen). Colonies were analyzed by sequencing using the T7 primer. Finally, the absence of SOCS1, SOCS3, or USP18 protein was confirmed by immunoblotting.

Statistical analysis
Prism4 (GraphPad Software Inc., La Jolla, CA) was used for statistical analysis.