USP18-Based Negative Feedback Control Is Induced by Type I and Type III Interferons and Specifically Inactivates Interferon α Response

Type I interferons (IFN) are cytokines that are rapidly secreted upon microbial infections and regulate all aspects of the immune response. In humans 15 type I IFN subtypes exist, of which IFN α2 and IFN β are used in the clinic for treatment of different pathologies. IFN α2 and IFN β are non redundant in their expression and in their potency to exert specific bioactivities. The more recently identified type III IFNs (3 IFN λ or IL-28/IL-29) bind an unrelated cell-type restricted receptor. Downstream of these two receptor complexes is a shared Jak/Stat pathway. Several mechanisms that contribute to the shut down of the IFN-induced signaling have been described at the molecular level. In particular, it has long been known that type I IFN induces the establishment of a desensitized state. In this work we asked how the IFN-induced desensitization integrates into the network built by the multiple type I IFN subtypes and type III IFNs. We show that priming of cells with either type I IFN or type III IFN interferes with the cell's ability to further respond to all IFN α subtypes. Importantly, primed cells are differentially desensitized in that they retain sensitivity to IFN β. We show that USP18 is necessary and sufficient to induce differential desensitization, by impairing the formation of functional binding sites for IFN α2. Our data highlight a new type of differential between IFNs α and IFN β and underline a cross-talk between type I and type III IFN. This cross-talk could shed light on the reported genetic variation in the IFN λ loci, which has been associated with persistence of hepatitis C virus and patient's response to IFN α2 therapy.


Introduction
Type I and type III (IL-28/29) IFNs form two multigenic families of pathogen-induced cytokines that exhibit common bioactivities through binding to unrelated cell surface receptors [1]. The numerous type I IFN subtypes (a/b/v) bind to a receptor made of the ubiquitously expressed IFNAR1 and IFNAR2 chains. Conversely, the type III IFNs (l1, 2, 3) bind to a receptor made of the broadly expressed IL-10R2 and of IFNLR1 (IL-28Ra) whose expression is cell type specific. Therefore, the response to type III IFNs is tissue specific and appears to be mainly restricted to epithelial cells [2].
Downstream of these two receptor complexes is a shared Jak/ Stat pathway, involving the Janus kinases Jak1 and Tyk2 that phosphorylate Stat1, Stat2 and Stat3. Activated Stat1/2 associate to IRF9 to yield the ISGF3 complex that induces transcription of IFN-stimulated genes (ISG) [1]. Thus, in humans, the 18 subtypes (13 a, 1 b, 1 v and 3 l) of the type I and type III IFN systems induce a same gene subset and exhibit antiviral and antiproliferative activities through two independent cell surface receptors. In addition, the type I IFNs are recognized as mediators linking innate and adaptive immunity via their effect on the differentiation and maturation of dendritic cells and T cells, activities not shared with type III IFNs [3].
Among the type I IFNs, the a/v subtypes on the one hand and the b subtype on the other are not equivalent, as they are differentially produced upon microbial infections and exhibit distinct bioactivities. The biological potency of any given subtype is determined by both receptor binding parameters and receptor density on target cells [4,5]. Hence, compared to IFN a2, IFN b binds the receptor with higher affinity, forms a longer-lived complex and is more potent at inducing translational control signals, inhibiting cell growth and osteoclastogenesis [6,7,8] (Moraga et al., submitted). Importantly, IFN a2 is routinely used in the clinic as in chronic HCV infection and several forms of cancer, whereas IFN b is approved for treatment of multiple sclerosis, considered an autoimmune disease [1].
The pleiotropic activities of IFNs must be tightly down regulated in time and space and several mechanisms have been shown to co-exist in order to attenuate IFN-initiated Jak/Stat signaling (reviewed in [9]). In an in vivo model, Sarasin et al showed that liver cells from mice repeatedly injected with murine IFN a become refractory to further IFN a stimulation [10]. The ISGencoded isopeptidase USP18/Ubp43 was found to be essential for the establishment of the desensitized state [10,11]. USP18 can remove the ubiquitin-like ISG15 from target proteins [12] and was found to inhibit IFN-induced Jak/Stat signaling when constitutively expressed in cultured cells [13]. Interestingly, USP18 expression was recently identified as a bad prognostic marker of the success of IFN a therapy in patients with chronic hepatitis C [14,15].
Here, we have studied how IFN induced desensitization integrates into the network built by the multiple type I and type III IFNs. We found that both type I and type III IFNs can induce a long lasting desensitization state in cells of different lineages, including human primary hepatocytes. Remarkably, the refractory state is targeted to the IFN a and v subtypes, leaving nearly intact the cells' responsiveness to b and l IFNs. We show that USP18 is necessary and sufficient to differentially desensitize cells by disturbing the assembly of a IFNs with the receptor complex.
Altogether, our findings emphasize the existence of differential activities within the type I IFN family and underline a novel type I/III IFN cross-talk acting at the receptor level that could have important consequences in the set up of clinical protocols, especially for the treatment of HCV-infected patients who are resistant to conventional pegylated IFN a2 therapy.

Type I and type III IFNs induce desensitization to IFN a
In a first set of experiments, we established to what extent cells that had responded to IFN a2, IFN b or IFN l1 could mount a response to a second stimulation. For this, we used HLLR1-1.4 cells, a clone derived from human fibrosarcoma HT-1080 cells stably expressing the IFNLR1 receptor chain and the luciferase reporter gene controlled by an ISGF3-dependent promoter [16]. Thus, HLLR1-1.4 cells are responsive to type I IFNs as well as to type III IFNs (Fig. 1B).
HLLR1-1.4 cells were left untreated (naïve) or stimulated (primed) with IFN a2, IFN b or IFN l1 for 24 hr, thoroughly washed and kept in fresh medium for another 24 hr (scheme in Fig. 1A). Following this resting period, the levels of Jak/Stat phosphorylation, luciferase activity and 29-59 oligo-adenylate-synthetase (OAS 69K) mRNA in primed cells had nearly returned to basal levels. Naïve and primed HLLR1-1.4 cells were challenged with IFN a2, IFN b or IFN l1 for 6 hr (Fig. 1A) and luciferase activity was quantified (Fig. 1B). The potency of IFN a2 in luciferase induction (expressed as EC 50 ) decreased 14.5-68.9 fold in primed cells as compared to naïve cells, whereas the potency of IFN b decreased only 2.1-3.2 fold (Fig. 1B). A similar trend was observed when naïve and primed cells were monitored by RT-qPCR for induction of OAS 69K mRNA in response to 10 pM of each IFN (Fig. 1C). As shown in Fig. 1D, upon desensitization the dose response relationship for IFN a2 had shifted down by a factor of 50-100. Interestingly, the activity of all the a/v subtypes assayed was decreased in type I and in type III IFN-primed.
Desensitization was also evident at the level of early signaling events. Fig. 2A shows that cells primed with IFN b or with IFN l1 were refractory to low doses (10 and 100 pM) of IFN a2 in terms of tyrosine phosphorylation of Stat1, Stat2 and Stat3 (compare lanes 2-3 with lanes 9-10 or [16][17]. Conversely, in primed cells Stat1, Stat2 and Stat3 were still phosphorylated upon treatment with IFN b and IFN l1. Moreover, in primed cells the activation of Jak1 and Tyk2, the earliest effectors of the pathway, was abrogated in the case of IFN a2, but still detectable in the case of IFN b and IFN l1 (Fig. 2B).
To better characterize the differential desensitization state, we asked if primed cells resumed their response to IFNa2 when stimulated for longer than 30 min. Naïve and primed cells were thus stimulated with IFN a2 and IFN b for up to 8 hr. As shown in Fig. 2C, the desensitization of primed cells to IFN a2 persisted, independently of treatment duration. To define for how long primed cells remained in a refractory state, we extended the interval between priming and restimulation (washing of 24, 48 and 72 hr). As shown in Fig. 2D, primed cells had regained IFN a2 sensitivity after 72 hr in the absence of cytokine. In conclusion, the differential desensitized state of the cell persists even at 8 hr of stimulation, but is reversible as seen when cells are kept in the absence of cytokine for 3 days.
Priming with type I IFN induced a similar a2/b differential desensitized state in cell lines, such as bronchial epithelial BEAS-2B and uroepithelial Hs 789.T cells, and in foreskin fibroblasts and T cell blasts ( Fig. 3A and 3B). Human primary hepatocytes respond to type I IFNs and more weakly to IFN l1 (Fig. 3C, compare level of P-Stat2, lanes 1-8). After 24 hr of priming with either IFN a2 or IFN l1, hepatocytes expressed higher Stat2 protein and detectable levels of USP18, both proteins being encoded by ISG (Fig. 3C, lanes 9-24). Importantly, primed hepatocytes were desensitized to IFN a2 and only marginally to IFN b (Fig. 3C, compare P-Stat2 in lanes 3, 11 and 19) and the extent of desensitization was related to the level of sensitivity to the priming cytokine. Of note, the basal phosphorylation level of Stat3 was reduced in IFN l1-primed cells with respect to naïve or IFN a2-primed cells (compare P-Stat3 in lanes 1, 9 and 17). Overall, these results demonstrate a previously unrecognized inhibitory cross-talk between the type I and type III IFN systems.
Desensitized cells are impaired in their ability to bind IFN a2 As shown above, cell desensitization to IFN a2 is manifest at the level of Janus kinase activation and thus may result from reduced surface level of the receptor chains or impaired binding of IFN a2. FACS analysis clearly showed that naïve and primed (i.e. desensitized) cells expressed equivalent levels of IFNAR1 and IFNAR2 (Fig. 4A). Therefore, we tested whether the ligand binding property of the receptor differed between naïve and primed cells. For this, we iodinated IFN a2 and, in place of IFN b which cannot be iodinated without loss of bioactivity, we made use of an engineered mutant of IFN a2 (IFN a2-HEQ) whose affinity for IFNAR1 is similar to that of IFN b and which recapitulates IFN b unique biological activities [6]. Accordingly, in primed HLLR1-1.4 cells, IFN a2-HEQ was as potent as IFN b in inducing Stat phosphorylation (Fig. 4B). On this basis, we compared the binding of 125 I-IFN a2 and 125 I-IFN a2-HEQ to naïve and to primed HLLR1-1.4 cells. Fig. 4C shows that the binding of 125 I-IFN a2 was reduced in both IFN b-primed cells and IFN l1-primed cells with respect to naïve cells. The reduction was most apparent for low 125 I-IFN a2 concentrations, matching the decrease in specific biological activity (Fig. 1D). In contrast, the binding of 125 I-IFN a2-HEQ was only marginally reduced in primed cells with respect to naïve cells (Fig. 4D). In conclusion, despite unaltered levels of IFNAR1 and IFNAR2, cells primed with either type I or type III IFNs (i.e. desensitized) are unable to assemble a functional IFN a2 receptor complex.
Expression of USP18 is necessary and sufficient to cause differential desensitization USP18 has been shown to downmodulate type I IFN activity through binding to IFNAR2 [13], and we therefore tested its role in differential desensitization. In HLLR1-1.4 cells USP18 mRNA was induced by a2, b and l1 IFNs, but not by other cytokines such as IFN c or IL-6 ( Fig. 5C). Accordingly, cell priming with IFN c or IL-6 did not induce USP18 nor lead to a desensitized state ( Fig. 5A and 5B). USP18 protein accumulated with similar kinetics in cells treated with IFN b or IFN l1, reaching maximum level between 8 and 16 hr of stimulation (Fig. 5D). As found in primary hepatocytes (Fig. 3C), two USP18specific bands of comparable intensity were consistently detected.
To study the involvement of USP18, its expression was silenced in HLLR1-1.4 cells (Fig. 5E). Remarkably, cells wherein USP18 was efficiently silenced were not desensitized to IFN a2. Of note, the level of Stat1/2 phosphorylation was higher in USP18silenced/primed cells (lanes 14, 15 and 17, 18) than in USP18silenced/naive cells (lanes 11,12) and most likely this is consequence of the higher content of these Stats in primed cells.
To determine if USP18 was sufficient to establish differential desensitization, HLLR1-1.4 cells were stably transfected with a USP18 expression vector or an empty vector. In the 10 positive clones analysed, the level of ectopic USP18 was higher (4 to 50 fold) than endogenous USP18 present in 8 hr-primed cells. Clone HU13 was chosen as it expressed the least USP18 (Fig. 6A). We checked the integrity of the deISGylase activity of USP18 expressed in clone HU13 by comparing the steady state level of ISGylated conjugates in IFN b-treated parental and HU13 cells (data not shown). The response to IFN a2, IFN b and IFN l1 was measured. Compared to naïve HLLR1-1.4 cells, clone HU13 was severely impaired in its phosphorylation response to IFN a2. Conversely, responses to IFN b and IFN l1 were preserved or slightly reduced (Fig. 6A-C). By directly comparing luciferase induction in a control clone (HP1) and in HU13 clone, it appeared that constitutive USP18 in HU13 shifted the dose response relationship so that higher concentration of IFN a2 was required to trigger a response of the same magnitude (Fig. 6D). All IFN a subtypes tested and IFN v were found to be between 25 to 85-fold less potent on HU13 cells than on control HP1 cells, whereas the activities of IFN b and IFN l1 were marginally reduced (Fig. 6E). The specific activity of IFN a2-HEQ, engineered for higher binding towards IFNAR1 (see above), was reduced only by a factor of 10 on HU13 cells (Fig. 6E), in accordance with the level of activity of this mutant on cells desensitized by IFN priming (Fig. 4B). On the other hand, the specific activity of IFN a2-a8tail, an IFN a2 mutant engineered for higher IFNAR2 binding [7], was decreased by a factor of 45 on HU13 cells as compared to HP1 cells (Fig. 6E). These results indicate that the affinity of a given IFN a subtype towards IFNAR1 determines the degree of USP18dependent desensitization to that subtype.
As shown for desensitization caused by IFN priming (Fig. 4A), desensitization caused by expression of USP18 was not dependent on a change of IFNAR1 or IFNAR2 cell surface expression (Fig. 7A). In studies analogous to those performed on IFN-primed cells, we measured binding of radiolabeled ligands on HU13 and   (Fig. 7B) and this to the same extent as for IFNprimed HLLR1-1.4 cells (see Fig. 4C). As expected, the binding of 125 I-IFN a2-HEQ was similar for the two clones (Fig. 7C). In conclusion, these data demonstrate that the sole expression of USP18 recapitulates the binding alteration observed in IFNprimed (i.e. desensitized) cells.

Discussion
The major findings of our study are hereafter summarized: i) type I IFN and type III IFN desensitize cells to several a IFNs but only marginally to IFN b; ii) cells of different lineages -including primary hepatocytes -undergo differential desensitization; iii) the extent of desensitization is controlled by the level of an ISG, USP18; vi) forced expression of USP18 in naive cells blunts IFN a response at the level of its assembly to the receptor complex.
USP18 is a cysteine protease specialized in the removal of ISG15 from ISGylated proteins. However, the phenotypic alterations caused by USP18 deletion in the mouse have been dissociated from ISG15-dependent mechanisms [17,18,19]. One group has proposed that USP18 attenuates IFN a signaling regardless of the isopeptidase activity of the protein by competitively displacing Jak1 from its interaction with IFNAR2 [13]. We indicated IFN doses. Cell lysates (50 mg) were analysed with the indicated Abs to evaluate tyrosine phosphorylation and content of Jak1 and Stats. The arrow points to the band corresponding to phosphorylated Jak1. The level of USP18 (bottom panel) was assessed in a 10% SDS PAGE. Of the two USP18 bands (apparent MW of 38 and 35 kDa), the faster migrating one results from proteolytic processing [46]. This latter comigrates with a non specific cross-reacting band detected in naïve cells and indicated by the asterisk (bottom panel). doi:10.1371/journal.pone.0022200.g003  have obtained preliminary evidence that, indeed, a catalytically inactive USP18 impairs IFN signaling when highly and stably expressed in naïve cells. In this context, however, desensitization is severe and not differential, as cells become refractory to IFN a and b subtypes. At present, we cannot exclude the possibility that the isopeptidase activity of USP18 could be required in certain physiological contexts, for instance when USP18 is below a given threshold or when the level of ISG15 conjugates is maximal. In fact, while it is remarkable that desensitization can be achieved by constitutive USP18 expression in naïve cells, it is also conceivable that the native protein acquires distinctive biochemical propertiesin terms of stability, partners and/or substrates -when it gradually accumulates in IFN-stimulated cells, along with the ISGylation manchinery. On that line, additional work is required to understand the functional link between USP18, ISG15 and the ISGylation machinery in desensitization. Interestingly, several features of the ISG15 system are closely related to the ubiquitin system [20]. Notably, ISG15 is the only ubiquitin-like molecule whose C-terminal residues (LRLRGG) are identical to those of ubiquitin. These similarities suggest functional or regulatory overlap between the two pathways and indeed, in vitro, murine USP18 can remove ubiquitin from substrates [21,22].
From in vitro studies, it is known that the assembly of the IFN a2receptor complex on artificial membranes is conditioned by the IFNAR1 concentration, whereas IFN b recruits IFNAR1 even if present at very low concentration [6]. Indeed, in cells the surface level of IFNAR1 is critical for the intensity of IFN a signaling [5,23,24]. Importantly and in accordance with other reports [13,25], we found that the presence of USP18 has no effect on the global cell surface level of IFNAR2 and IFNAR1. Nonetheless, our study shows that binding of IFN a is impaired in cells expressing USP18, whether IFN-primed or USP18-transfected. Overall, we favor a model whereby the interaction of USP18 with IFNAR2 ( [13] and our data) may lead to a re-organization of the architecture of the type I IFN receptor. A change in lateral mobility of the receptor chains, in their localisation in membrane microdomains or their physical preassociation could weaken assembling and signaling of IFN a2. Conversely, owing to its higher affinity for the receptor, IFN b would retain activity on USP18-expressing cells.
It is remarkable that, at least in humans, the 13 a IFNs exhibit non-optimal affinity to the receptor chains and it is precisely this weakness that allows a/b differential bioactivities and desensitization [6]. Thus, in a viral infection, abundant IFN a is likely to be induced from the multiple genes and limits the spread of the virus by exerting potent antiviral action in a timely regulated mode on cells that will then be desensitized. On the other hand, the single IFN bthat is induced alone or, in response to viral infection, co-induced with IFN a [26] -is optimized to bind the receptor chains with high affinity and retains activity on cells desensitized for IFN a. This exclusive property of IFN b may be critical for the stimulation of adaptive immune responses necessary to eradicate the virus.
Type I IFNs and type III IFNs are induced by similar stimuli, exhibit common bioactivities and synergyze in antiviral activity towards several viruses, including HCV [27,28,29]. Their functional overlap was somehow expected given the activation, through different receptors, of the same transcriptional factor ISGF3 [30]. This is the first report of an inhibitory effect exerted by IFN l upon IFN a activities. One particular context where cellular desensitization to IFN a could be relevant is the therapeutical control of chronic HCV infection. The current standard therapy is pegylated-IFN a2 and ribavirin, whose success is influenced by the virus genotype and multiple host factors. Among the strongest predictive factors of treatment outcome is the expression level of ISGs in liver tissue. Indeed, high baseline ISG expression in hepatocytes has been consistently associated with poor response to therapy [14,15,31,32,33]. Intrahepatic differences in ISG expression may reflect differences in host innate antiviral responses before and/or during the chronic phase. The ISG «signature» is likely to be driven and maintained by local innate cytokines, such as IFNs, and may ultimately result in failure to respond to therapeutic IFN.
Interestingly, USP18 is a component of the gene signature predictive of poor treatment response [15,32,33]. Moreover, the knockdown of USP18 in hepatoma cells was shown to potentiate the anti-HCV effect of IFN a [11]. Here, we provide evidence that primary human hepatocytes respond to IFN l and, when primed with it, they express USP18 and become desensitized to IFN a2. Thus, it is tempting to speculate that this USP18-mediated refractoriness to IFN a could contribute, at least in part, to lower the effectiveness of an IFN a-based therapy. In that event, IFN b or l would represent alternative therapeutic approaches.
Another strong predictive factor of successful treatment of chronically HCV infected patients (and spontaneous viral clearance) is the IFN l3 (IL28B) genotype. Paradoxically, the good response IFN l variant, ie predicting higher success rate of IFN a-based therapy, was found to be associated with higher viral load [34,35,36,37]. These consistent observations have spurred intensive studies to try to relate the IFN l3 (IL28B) genotype with the level of hepatic ISG [33,38,39,40]. To date contradictory conclusions have been reported that do not yet provide a clear picture. Likewise, we are still missing consistent analyses of which of the variants, if any, alters expression level and/or potency of IFN l.
In view of these and our present data, one can speculate that a patient with the hapless genotype may induce IFN l inappropriately (e.g. altered level, potency or timing) upon HCV infection. On the one hand, this will lower the viral load without however clearing the virus and, on the other hand, will maintain a high level of ISGs, including USP18. Sustained level of USP18 may contribute, at least in part, to desensitize liver cells to administered IFN a.

Materials and Methods
Cells HLLR1-1.4 cells are described in [16]. HLLR1-1.4 and derived clones were cultured in DMEM with 10% fetal calf serum (FCS), hypoxanthine, thymidine and aminopterin (HAT) and 400 mg/ml G418. To obtain HU and HUS clones, HLLR1-1.4 cells were cotransfected with pSVpuro and pMet7 empty vector or pMet7 encoding USP18 using FuGENE6 (Roche Applied Science). Colonies selected in 0.4 mg/ml puromycin were analysed by immunoblot for USP18 level as compared to primed HLLR1-1.4 cells. Primary hepatocytes were isolated as described [41] from a human liver sample obtained from a 51 y-o female with intrahepatic lithiasis. The French National Ethics Committee has authorized the use of these samples for research. The patient was free of any HCV, HBV and HIV markers at the time of surgery. Hepatocytes were plated at confluence in a 12-well plates at 10 6 cell/well precoated with collagen in culture medium consisting of Williams' E and Ham's F-12 (Sigma) (1/1 in volume). For the first 24 h, 5% FCS (Gibco) was added to favor cell attachment. The standard medium was then replaced with 1 ml of serum-free medium as described [41]. Cultures were incubated at 37uC and 5% CO2.

Plasmids and reagents
USP18 cDNA was cloned by PCR using as template the cDNA prepared from HLLR1-1.4 cells stimulated with IFN b-treated for 6 hr and as primers: forward 59TTTGATATCCTGGGGGTT-TTGGAGTGA39 and reverse 59TAGACCGGTCTGAAGG-TTTTGGGCATTTC 39. The PCR product was subcloned in pMET7 vector. The catalytic activity of USP18 was assessed by comparing the global protein ISGylation level in 293T cells transiently transfected with ISG15, E1, E2 and E3 enzymes of the ISGylation machinery in the presence or absence of USP18. Rec IFN a2b was a gift of D. Gewert (Wellcome, UK) and IFN b was from Biogen Idec (Boston, MA). Mutants IFN a2-HEQ and IFN a2-a8tail were described in [6,42]. IFN a1, a8 and a21 were produced as in [43]. IFN v was from G.R. Adolf (Bender, Wien). All IFNs were purified to homogeneity. Hyper-IL-6, chimeric fusion of human IL-6 and IL-6Ra, was a gift of Merck Serono S.A. Human IFN c was from PBL, Biochemical Laboratories and IFN l1 from Peprotech.

Luciferase reporter assay
To measure luciferase activity, cells were plated in 96-well plate and treated in triplicate for 6 hr with 9 dilutions of IFN in a concentration range covering 2.4 log. Cells were lysed and luciferase activity was quantified in a luminometer (LB960 Berthold). Non-linear regression fits and determination of EC 50 were done using Prism 5 (GraphPad software).

Quantitative real-time PCR
Cells were treated with IFN for 4 hr. Total RNA was purified with RNeasy columns (Qiagen). Reverse transcriptions were primed with random primers and performed using Moloney murine leukemia virus reverse transcriptase (Invitrogen). Quantitative real-time PCR (qRT-PCR) was performed using the TaqMan gene expression assay technology (Applied Biosystems) for USP18 (catalog no. Hs00276441). Each sample was run in triplicate, normalized to the 18S RNA amplification level in the same sample, and calculated relative to the expression of the target gene in unstimulated cells. For measuring OAS 69K mRNA, qRT-PCR assays were performed as in [26]. Quantification data are presented as the 95% confidence limits of ratio to the glyceraldehyde 3-phosphate dehydrogenase (GAPDH) level.
For binding assays ( Fig. 3C and D), naïve and 8 hr-primed HLLR1-1.4 cells were seeded on 6-well plates (8610 5 cells/well) and 16 hr later incubated for 1 hr at 37uC with different concentrations of either 125 I-IFN a2 or 125 I-IFN a2-HEQ only or in the presence of a 100 fold excess of unlabeled cold IFN a2-HEQ competitor. Cells were washed three times in DMEM and 5% serum to eliminate unbound IFN, trypsinized, and counted for 125 I using a c counter (Berthold). For binding assays on clones ( Fig. 6B and C), cells were seeded on 6-well plates (8610 5 cells/ well) and treated as above.

USP18 silencing
USP18 ON-TARGETplus SMARTpool and a control siRNA (ON-TARGETplus non-targeting pool) were from Dharmacon. Cells were transfected with 25 nM of siRNA using Lipofectamine RNAi max reagent (Invitrogen), according to manufacturer's instructions. Twenty-four hr later, cells were either left untreated or primed, washed and challenged with IFN a2 or IFN b for 30 min to measure activation of Stats (scheme in Fig. 1A).
The siRNAs constituting the USP18 ON-TARGETplus SMARTpool were also tested individually.