Biochemical Journal

Research article

USP18 establishes the transcriptional and anti-proliferative interferon α/β differential

Véronique Francois-Newton , Mark Livingstone , Béatrice Payelle-Brogard , Gilles Uzé , Sandra Pellegrini


Type I IFNs (interferons) are pathogen-induced immunoregulatory cytokines that exert anti-viral and anti-proliferative activities through binding to a common cell-surface receptor. Among the 17 human IFN subtypes, IFNβ binds the IFNAR (IFNα receptor) 1/IFNAR2 receptor chains with particularly high affinity and is especially potent in select bioactivities (e.g. anti-proliferative and pro-apoptotic) when compared with IFNα2. However, no molecular basis has been ascribed to this differential action, since the two ligands are equipotent in immediate early signalling events. In the present study we report that IFNβ induces Stat (signal transducer and activator of transcription) phosphorylation and transcriptional activation of ISGs (interferon-stimulated genes), including two genes with pro-apoptotic functions, for a considerably longer time frame than does IFNα2. We show that the diversification of α2/β responses progressively builds up at the receptor level as a result of accumulating USP18 (ubiquitin specific protease 18), itself an ISG, which exerts its negative feedback action by taking advantage of the weakness of IFNα2 binding to the receptor. This represents a novel type of signalling regulation that diversifies the biological potential of IFNs α and β.

  • apoptosis
  • cytokine signalling
  • negative feedback
  • protease
  • ubiquitin-specific protease 18 (USP18)


Type I IFNs (interferons) form a family of secreted cytokines that regulate cellular functions as diverse as resistance to viral infection, innate and acquired immune responses, and normal and tumour cell survival and death [1]. One unique feature of this IFN family is its high level of complexity in all mammals. In humans 13 IFNαs and one each of IFNs β, κ, ω and ϵ bind the same receptor and operate through the same Jak (Janus kinase)/Stat (signal transducer and activator of transcription) pathway. The type I IFN receptor is made of IFNAR (IFNα receptor) 1 and IFNAR2, single membrane-spanning proteins belonging to the class 2 cytokine receptor superfamily [2]. Upon IFN binding, the receptor-associated tyrosine kinases Jak1 and Tyk2 (tyrosine kinase 2) are immediately activated and phosphorylate IFNAR2 on tyrosine residues, which serves as docking sites for Stat transcription factors. Once phosphorylated by the Jaks, activated Stat1/2 associate with IRF (IFN regulatory factor) 9 forming the prominent transcriptional ISGF3 (IFN-stimulated transcription factor 3) complex that induces expression of ISRE (IFN-stimulated response element)-containing ISGs (IFN-stimulated genes) [3]. Non-canonical Stat complexes and IRFs can bind to ISRE-related sequences and reinforce ISG induction by type I IFN [4].

It is an open question as to the reason of the existence of multiple type I IFN genes. Population geneticists have recently addressed this question by investigating how natural selection acted upon these genes [5]. Some IFN subtypes (α6, α8, α13 and α14) were found to have evolved under strong selective constraints, others (α2, α5, α21, β, κ and ω) were shown to have accumulated some diversity and a third group of IFNs (notably α10, α16, α17 and ϵ) display high frequencies of amino acid changes within the population. Thus different degrees of constraint and redundancy characterize the human type I IFN family members. In that respect, it is relevant that all living mammals possess a single or a small number of IFNβ genes and a larger number of IFNα-related genes and that the α and β genes are differently regulated [6,7]. These and additional observations point to unique physiological roles of IFNβ[8].

Several studies have reported on differential bioactivities of type I IFNs, but despite the identification of subtype-specific events at the receptor level [912], no molecular mechanism has been implicated. A differential is defined as a lack of correlation between two specific activities. Among the human subtypes, IFNβ is especially potent in bioactivities requiring long-term stimulation, such as proliferation inhibition, apoptosis and cell differentiation, where IFNβ can be over 50-fold more potent than IFNα2, but exhibits near equipotency with IFNα2 in antiviral activity [2]. Substantial differences exist with respect to the binding of these two IFNs to the receptor. Hence, in vitro, IFNβ binds more tightly to IFNAR1 and IFNAR2 than IFNα2, and forms a more stable ternary complex [2]. Recently, IFNs of differing affinities and potencies have been co-crystallized with IFNAR1/2 ectodomains [13]. The overall architecture of these solved ternary complexes is similar, confirming that the respective stabilities are relevant to differential potencies.

In cells the stability of ligand–receptor complexes appears to impact signals regulating receptor traffic. Within minutes of stimulation differential IFNAR2 routing can be appreciated: IFNβ induces the down-regulation and degradation of cell surface IFNAR2, whereas IFNα2 induces its recycling [14]. Related to this, mutants of IFNα2 that were designed to form a stable ternary complex are able to down-regulate surface IFNAR2 [15]. Signalling feedback controls operating at immediate–early times include serine/threonine kinase(s) and ubiquitin ligase(s) targeting the IFNAR1 subunit [16,17], as well as SOCS (suppressor of cytokine signalling)-mediated action on receptor–Jaks/Stats. Another negative feedback control involves USP18 (ubiquitin-specific protease 18), an IFN-induced isopeptidase able to cleave ubiquitin-like ISG15 from conjugates [18,19].

In our previous analyses of IFNα2 compared with IFNβ signalling in human transformed cells, we reported that the two subtypes activate the canonical Jak/Stat pathway, early ISG induction and cell-cycle arrest with similar magnitude, but that IFNβ induces more robust apoptosis than IFNα2 [14,20]. Our data suggested that the higher apoptotic potency of IFNβ requires the activation of signals additional to the early acting Jak/Stat signalling events. In continuation of these studies, we have assessed Stat phosphorylation and ISG expression at later phases (>8 h) of the response to IFNs. We show that the α2/β differential is progressively established at the level of Stat activation and gene induction through the expression of the negative feedback regulator USP18. Analyses of USP18-silenced cells demonstrate that USP18 is largely, if not entirely, responsible for establishing α/β differential bioactivities which require long-lasting stimulation.


Cells and reagents

The human amnion-derived WISH cells were cultured in DMEM (Dulbecco's modified Eagle's medium) and 10% heat-inactivated FBS (fetal bovine serum) and HLLR1-1.4 cells were cultured as described previously [21]. Recombinant IFNα2 was from Dr Dirk Gewert (Wellcome Foundation, Beckenham, U.K.); IFNβ was from Biogen Idec. IFNs were purified to specific activities >108 units/mg of protein. Jak inhibitor 1 (Calbiochem) was used at 800 nM.

Protein analysis

Cells were lysed in RIPA buffer and lysates (40 μg) were analysed as described previously [22]. The antibodies used were against phospho-Tyr701 Stat1, TRAIL (tumour-necrosis-factor-related apoptosis-inducing ligand), USP18, CASP (caspase) 8, cleaved CASP3, cleaved CASP9, Akt (all Cell Signaling Technology), Stat1 and phospho-Tyr689 Stat2 (Millipore), Stat2 (UBI), tubulin and actin (Sigma), ISG15 (a gift from Dr Ernest Borden, Cleveland Clinic and Case Comprehensive Cancer Center, Taussig Cancer Institute, Cleveland, OH, U.S.A.), IFIT (IFN-induced protein with tetratricopeptide repeats) 1, IFIT2, IFIT3 (gifts of Professor Ganes Sen, Cleveland Clinic and Case Comprehensive Cancer Center, Taussig Cancer Institute, Cleveland, OH, U.S.A.) [23], OAS2 p69 (a gift from Dr Ara Hovanessian, Institut Pasteur, Paris, France) [24], and MxA (a gift from Dr Otto Haller, University of Freiburg, Freiburg, Germany) [25]. Immunoblots were revealed using enhanced chemiluminescence detection reagents (Western Lightning, PerkinElmer) and bands were quantified with Fuji LAS-4000.

siRNA (small interfering RNA) silencing lysates

USP18 ON-TARGETplus SMARTpool and a control siRNA (ON-TARGETplus non-targeting pool) were from Dharmacon. Individual USP18 siRNAs (#7, 5′-CUGCAUAUCUUCUGGUUUAdTdT-3′; #8, 5′-GGAAGAAGACAGCAACAUGdTdT-3′; and #10, 5′-GCAAAUCUGUCAGUCCAUCdTdT-3′) were from Sigma. Cells were transfected with 25 nM of siRNA using Lipofectamine™ RNAi Max Reagent (Invitrogen), according to the manufacturer's instructions. At 24 h later, the cells were either left untreated or stimulated with IFNα2 or IFNβ.

qRT-PCR (quantitative real-time PCR)

Cells were harvested using RNeasy Mini Kit (catalogue number 74104, Qiagen) according to the manufacturer's instructions including RNase-Free DNase Set (catalogue number 79254, Qiagen) on column DNase digestion. Reverse transcription was performed using M-MLV (Moloney murine leukaemia virus) reverse transcriptase (catalogue number 28025-013, Invitrogen) according to the manufacturer's instructions with random primers (catalogue number 58875, Invitrogen) and rRNasin (catalogue number 29457913, Promega). cDNA was purified with QIAquick PCR Purification kit (catalogue number 28104, Qiagen). qRT-PCR was performed with Fast Start Universal SYBR Green Master mix (catalogue number 4913850001, Roche) and StepOne Plus Machine (Applied Biosystems) using standard curve-based quantification with a 60°C annealing temperature. PCR product standards were produced as above, purified with QIAquick PCR Purification kit and diluted in TE (Tris/EDTA) buffer with 10 μg/ml sheared salmon testes DNA (catalogue number D-9156, Sigma). PCR products for each primer pair were analysed by agarose gel to confirm the proper molecular mass and subjected to sequencing. Quantification data are presented as the 95% confidence limits of ratio to the GAPDH (glyceraldehyde 3-phosphate dehydrogenase) level (Student's t test, n=4).



Flow cytometry

For flow cytometry intracellular staining phospho-Tyr701 Stat1 antibodies (1:100 dilution, Cell Signaling Technology) and Alexa Fluor® 488-conjugated anti-rabbit secondary antibodies were used according to the manufacturer's instructions. Briefly, cells were trypsinized and washed with PBS prior to fixation in paraformadehyde (3.7%) for 10 min at 37°C and permeabilization in methanol (90%) for 30 min at −20°C. After blocking in 2% BSA in PBS, cells were subjected to staining with primary and secondary antibodies and then washed. Samples were analysed with a Becton Dickinson FACS Calibur flow cytometer.

Immunofluorescence/confocal microscopy

WISH cells plated on to glass cover slips and treated with IFNα2 or IFNβ (500 pM) for the indicated times were washed twice with cold PBS, prior to fixation (3.2% paraformaldehyde) at 37°C for 10 min followed by permeabilization in methanol (100%) at −20°C for 24 h. The cells were washed with PBS, blocked with 2% BSA in PBS and subjected to staining with phospho-Tyr701 Stat1 antibody (1:100 dilution) in blocking solution. After washing with PBS, cells were subjected to staining with Alexa Fluor® 488-conjugated anti-rabbit secondary antibody (1:500 dilution, Invitrogen) and DAPI (4′,6-diamidino-2-phenylindole, 200 ng/ml; Invitrogen) in blocking solution. The cells were washed and the coverslips were mounted using Fluoromount-G (Southern Biotech), prior to confocal image acquisition with an LSM510 Meta inverted confocal microscope (Zeiss).

Anti-proliferative/apoptosis assays

The anti-proliferative activity of IFNs was assessed as described previously [20]. Briefly, cells were seeded at 5×105/60 mm dish and left to attach. The cells were transfected with control or USP18 siRNA for 24 h and then seeded in 96-well plates. At 16 h later, the cells were treated with varying doses of IFNα2 or IFNβ for 72 h prior to Crystal Violet assessment of cell density. Apoptosis was assessed by an 7-AAD (7-aminoactinomycin D) assay as described previously [26].


IFNβ more potently induces ISGs at late time points than IFNα2

The extent of activation of the canonical Jak/Stat pathway after brief stimulation with IFNα2 and IFNβ is comparable, as seen in several human cell types [14]. Consistent with this, the accumulation of ISG transcripts in response to the two IFN subtypes is nearly equivalent within 2–8 h of stimulation ([20] and results not shown). Nonetheless, by virtue of its higher affinity for the receptor, IFNβ is a more potent inducer of apoptosis than IFNα2 [27,28], particularly in cells with a low density of receptors [20] and cells of non-haematopoietic origin [29]. To explain this conundrum, we monitored the steady-state mRNA levels of well-characterized ISGs in WISH cells after long periods of stimulation. As shown in Figure 1, at 8 h of IFNα2 and IFNβ stimulation, no consistent differences in IFIT1, MxA, USP18, CXCL11, OAS1 or ISG15 induction could be observed. A similar trend was seen for the induction profiles of pro-apoptotic TRAIL and FAS. However, after 16–36 h of continuous IFN stimulation, all of these ISGs were more highly expressed in IFNβ-stimulated cells. Some transcripts, such as IFIT1, USP18, FAS and TRAIL, had remarkably diminished between 8 and 16 h of IFNα2 stimulation. Other transcripts, like OAS1 and ISG15, had levelled off by 8–16 h of IFNα2 stimulation, but continued accumulating in response to IFNβ.

Figure 1 IFNβ induces prolonged expression (mRNA) of all of the assessed ISGs relative to IFNα2

mRNA levels relative to GAPDH are shown for IFIT1, MxA, USP18, CXCL11, OAS1, ISG15, TRAIL and FAS for WISH cells treated with 500 pM IFNα2 or IFNβ for the indicated times. Error bars indicate 95% confidence intervals (Student's t test).

In agreement with the above data, Western blot analyses of WISH cells stimulated from 8 to 36 h revealed a progressive α2/β differential accumulation of ISG-encoded proteins (IFIT1, IFIT3, MxA, USP18, and free and conjugated ISG15) starting after 16 h of treatment (Figure 2A). This was the case also in fibrosarcoma HLLR1-1.4 cells (Figure 2B, see also Figure 4B) that exhibit a rather poor anti-proliferative response to IFNs (results not shown).

Figure 2 IFNβ induces prolonged expression (protein) of all of the assessed ISGs relative to IFNα2

Western blot analysis of USP18, ISG15, MxA, IFIT2 (HLLR1-1.4 only), IFIT1, IFIT3, OAS2 (HLLR1-1.4 only) and β-actin levels for WISH (A) and HLLR1-1.4 (B) cells treated with 100 pM of IFNα2 or IFNβ for the indicated times.

IFNβ induces more persistent Stat1/2 phosphorylation than does IFNα2

In light of the above findings, we asked whether the activation levels of Stat1 and Stat2 could account for the delayed α2/β differential accumulation of ISG mRNA. For this, we monitored tyrosine-phosphorylated Stat1 and Stat2 in WISH cells stimulated with either IFNα2 or IFNβ (250 pM) for 1–12 h (Figure 3A). No difference in Stat activation levels induced by these two IFNs could be observed at early time points (0–4 h) as reported previously [14]. However, at later time points (8 and 12 h), phospho-Stat1 and phospho-Stat2 levels were indeed higher in IFNβ-stimulated cells. This was confirmed by monitoring the level of phospho-Stat1 by intracellular staining and flow cytometry (Figure 3B) as well as by confocal microscopy (Figure 3C). Both analyses showed that at the early time points there was an equivalent Stat1 activation level in response to IFNα2 and IFNβ. Conversely, after 12 h of stimulation phospho-Stat1 (total or nuclear) was detectable only in IFNβ-stimulated cells. In conclusion, activated Stats are more persistent and appear to correlate with higher levels of ISG transcripts in IFNβ-stimulated cells. To assess whether the persistent Stat1/2 activation requires continuous receptor activation, we tested the effect of the potent Jak inhibitor 1. The robust Stat phosphorylation detected at 15 min of IFNβ treatment was abrogated by a 15 min pre-incubation of the cells with the Jak inhibitor (Figure 3D). In the cells stimulated for 9 h with IFNβ, the level of phosphorylated Stats progressively decreased to baseline with increasing inhibitor incubation time (0–2 h). Similar half-lives (30–45 min) were measured for phospho-Stat1 and phospho-Stat2. These data suggest that the protracted Stat activation observed only in response to IFNβ is not due to slower deactivation mechanisms, but rather requires continuous activation of the receptor–Jak complex.

Figure 3 Sustained phosphorylation of Stat1/2 upon IFNβ treatment

(A) Western blot analysis of phosphorylated (P-) and total Stat1 and Stat2 and USP18 in WISH cells stimulated with 250 pM IFNα2 or IFNβ for the indicated times. (B) Flow cytometric analysis of intracellular phospho-Tyr701 Stat1 in WISH cells unstimulated or stimulated for 1 or 12 h with 500 pM IFNα2 or IFNβ. Mean fluorescence intensitites (MFI) are indicated for IFN-stimulated (open histograms) and unstimulated cells (shaded histograms). (C) Immunofluorescence staining with phospho-Tyr701 Stat1 (Alexa Fluor® 488) in the nuclei of WISH cells unstimulated (0 h) or stimulated with IFNα2 and IFNβ (500 pM) for 1, 4 and 12 h. Nuclei are stained with DAPI (DNA). Sec. Ab., cells were stained with secondary Ab only. (D) Western blot analysis of phosphorylated (P-) Stat1 and Stat2 in IFN-stimulated cells treated with Jak inhibitor 1. WISH cells were stimulated for 9 h with IFNβ (250 pM) and treated or not with Jak inhibitor 1 (800 nM) for the indicated times before the end of the IFN stimulation. The efficiency of the inhibitor was controlled by pre-treating cells for 15 min with the inhibitor and then adding IFNβ for 15 min. Nearly complete abrogation of the phosphorylated bands was obtained (98% and 80% for Stat1 and Stat2 respectively). (E) Western blot analysis of phosphorylated (P-) Stat1 and phosphorylated and total Stat2 for control and USP18 silenced HLLR1-1.4 cells treated with 100 pM of IFNα2 or IFNβ for the indicated times reveals that USP18 silencing results in persistent phosphorylation of Stats.

USP18 is responsible for the IFN α2/β differential signalling, transcriptional and anti-proliferative activities

As described above (Figure 3A), the α2/β differential in phospho-Stats evident at 8 and 12 h parallels the accumulation of ISG-encoded proteins, including USP18, a negative regulator of type I IFN responses [19]. Therefore we assessed the effect of silencing USP18 on Stat1/2 activation at various stages of stimulation (from 1 to 36 h). Efficient silencing of USP18 totally abrogated the α2/β differential and resulted in a long lasting Stat1 and Stat2 phosphorylation which, importantly, was equivalent for the two IFN subtypes (Figure 3E).

In order to assess the extent to which the control of Stat phosphorylation by USP18 regulates the α/β differential gene induction, we monitored ISG transcripts in cells silenced for USP18. Interestingly, cells lacking USP18 not only accumulated higher levels of ISGs (mRNA and protein) at late stimulation times (>8 h), but also responded similarly to the two IFN subtypes (Figure 4 and Supplementary Figure S1 at This demonstrates a major role of USP18 in the IFN α2/β differential induction of ISGs.

Figure 4 USP18 is responsible for differential late (8–72 h) ISG expression by IFNα2 and IFNβ

(A) mRNA levels relative to GAPDH are shown for IFIT2, MxA, TRIM22, ISG15, IFITM1 and Stat1 for HLLR1-1.4 cells treated with 100 pM of IFNα2 or IFNβ for the indicated times. (B) Western blot analysis of USP18, MxA, OAS2, IFIT1, IFIT3, ISG15, Stat1 and β-actin (control) and USP18-silenced HLLR1-1.4 cells treated with 100 pM of IFNα2 or IFNβ for the indicated times.

The biological consequence of the USP18-dependent establishment of differential ISG induction was first assessed by measuring the percentage of apoptotic cells after 72 h of stimulation. As shown by 7-AAD staining of the control cells, IFNβ was approximately 2-fold more potent than IFNα2 (Figure 5A, upper panels). USP18 silencing augmented considerably both IFNα2- and IFNβ-induced apoptosis and abolished the differential (Figure 5A, lower panels). Accordingly, the IFNα2- and IFNβ-induced levels of pro-apoptotic TRAIL and of cleaved CASPs 3, 8 and 9 were equalized in USP18-silenced cells (Figure 5B). Next, we measured the anti-proliferative activity of IFNα2 and IFNβ in the control and USP18-silenced cells. In cells transfected with control siRNA, the anti-proliferative potencies of IFNα2 and IFNβ were profoundly different (EC50 values of 780 pM and 26 pM respectively; EC50 α2/EC50 β=30). On the other hand, in USP18-silenced cells the differential was muted (EC50 values of 37 pM and 7 pM respectively; EC50 α2/EC50 β=5) (Figure 5C).

Figure 5 USP18 is responsible for differential apoptotic and anti-proliferative effects induced by IFNα2 and IFNβ

(A) Flow cytometric analysis of 7-AAD incorporation in control and USP18-silenced WISH cells untreated or treated with IFNα2 or IFNβ (500 pM) for 72 h. (B) Western blot analysis of full-length and cleaved (cl.) CASP8 and cleaved CASP9 and CASP3, TRAIL, ISG15, and MxA in WISH cells treated with IFNα2 or IFNβ (500 pM) for 72 h. (C) Cell density (Crystal Violet staining) was assessed after 72 h of IFNα2 (●) or IFN β (○) treatment at varying doses (between 0.1 pM and 3 nM) in control or USP18-silenced WISH cells.


We have investigated the molecular mechanism underlying the differential action of two human type I IFN subtypes, IFNα2 and IFNβ, towards apoptosis and proliferation control. We found that, in human transformed fibroblasts and epithelial cells, a low level of activated Stat1 and Stat2 is maintained upon stimulation with IFNβ, but not IFNα2, through continuous low level activation of the receptor–Jak complex. Moreover, the transcriptional potential of IFNβ persists for longer times. Importantly, we demonstrate that the α2/β differential Stat activation and ISG induction are dependent on the presence of USP18, since these differentials are abrogated upon silencing USP18. Furthermore, silenced cells exhibit a remarkably reduced IFNα2/β differential in long-term (72 h) apoptotic and anti-proliferative responses. These data illustrate that the IFN-regulated accumulation of USP18, a canonical ISG, progressively restrains IFNα2-induced signalling more than IFNβ signalling. On the basis of our previous study [22], USP18 is expected to restrain signalling by all of the IFNα/ω subtypes.

USP18 associates with IFNAR2 [19] and does not modify the level of IFNARs at the cell surface, but rather affects the assembly and/or stability of the receptor–ligand ternary complex [22]. This was shown in non-stimulated cells expressing exogenous USP18 as well as in cells primed for 8 h with type I IFN and then washed to eliminate residual IFN and secure full recovery of surface receptors. In cells under continuous IFN stimulation, as in the present study, the accumulation of USP18 may alter the properties of one or both receptors to the extent that pre-existing binding differences are magnified. In fact USP18 appears to specifically lower IFNα2 activity below threshold levels by hindering the already weak association of receptor and ligand. The binding affinities of IFNα2 for IFNAR1 and IFNAR2 have been demonstrated to be 100- and 50-fold lower respectively than those for IFNβ [30]. Thus the tighter ternary complex formed by IFNβ retains a moderate Jak/Stat signalling potential even in the presence of USP18. In support of this model, designer mutants of IFNα2 that form a tighter ternary complex with IFNAR1 and IFNAR2 have been shown to more potently induce ISG mRNA, to exhibit IFNβ-like anti-proliferative activities and to be less sensitive to USP18 action [15,22]. Thus we propose that the USP18-driven negative feedback loop is an integral part of the delayed IFN response, decoding ligand input specificity and setting the threshold of duration and amplitude of receptor activation induced by different ligands. It is likely that the dynamic range of the system may be very sensitive to the varying concentrations of USP18 in the cell.

The mechanisms by which IFN induces bioactivities requiring long-term stimulation are complex, as they involve the actions of multiple ISGs and can be very much cell-context specific [31]. Although subtle differences may exist and go undetected in the early phase of robust Stat phosphorylation and ISG transcription, these do not appear sufficient to explain the α2 compared with β anti-proliferative differential. Indeed, we consistently observed greater induction of IFIT2 (ISG54) after 8 h of IFNβ treatment than after IFNα2 treatment (Figure 2B), and IFIT2 (ISG54) has been shown to induce apoptosis [32]. However, the results of the present study and previous work by our laboratory show clearly that IFNβ, even at low doses (30 pM), exhibits more pronounced anti-proliferative activity than does IFNα2 at high doses (3 nM), even though at these respective doses IFNα2 induces much greater Jak/Stat phosphorylation than IFNβ [14]. This demonstrates that early Jak/Stat phosphorylation can not explain the anti-proliferative differential.

Previous work has shown that PI3K (phosphoinositide 3-kinase)/mTOR (mammalian target of rapamycin) signalling is critical for the induction of apoptosis by high doses (around 1.5 nM) of IFNα2 in two cancer cell lines [33], and yet ISG mRNA induction has been shown to be largely independent of PI3K signalling [34]. We cannot exclude that the difference in anti-proliferative potencies of IFNα2 and IFNβ results from differential modulation of the PI3K pathway; however, our observation that silencing USP18 reduces considerably the anti-proliferative differential of the two cytokines suggests that USP18-dependent control of ISG mRNA induction is the key determinant. Furthermore, we were not able to detect consistent PI3K/mTOR activation following the addition of IFN (results not shown). At the late stages of stimulation (>8 h) modification of this pathway, and of other non-Stat pathways, may probably result from autocrine acting factors (e.g. pro-apoptotic TRAIL) that are themselves ISG products.

In the present study, silencing of USP18 increased and also equalized (α2 compared with β) the induction of pro-apoptotic genes (e.g. TRAIL and FAS) and the percentage of apoptotic cells in the two stimulated populations. Interestingly, USP18 was recently identified in a screen as the most powerful isopeptidase capable to protect E1A-transformed embryonic fibroblasts from apoptosis induced by anti-cancer drugs and relying on basal IFN [35]. In these cells, IFNα, used as a 250 pM single dose, failed to induce apoptosis unless combined with USP18 silencing, allowing a very robust up-regulation of TRAIL transcripts. Induction of TRAIL by IFN has been recurrently associated with apoptosis in different cell types [27,28]. In human bladder cancer cells TRAIL knockdown was shown to reduce IFNα2-induced apoptosis, and similar effects were observed upon knockdown of FADD (Fas-associated death domain), CASP8, Stat1, IRF1 and CDKN1A (cyclin-dependent kinase inhibitor 1A) [36]. These results would be consistent with the TRAIL/TRAILR (TRAIL receptor)1/2-FADD/CASP8 pro-apoptotic pathway [37] being of key importance. In this model Stat1 and IRF1 function as transcription factors for TRAIL and other pro-apoptotic ISGs [36,38]. Accordingly, IRF1 was shown to be involved in IFNβ-specific apoptosis of Ewing's sarcoma-derived cell lines [39]. An attractive model to explain how even low dose IFNβ limits cell proliferation and induces robust apoptosis would invoke the continued formation of ISGF3 in order to secure critical levels of ISGs during the time period in which ISG-encoded transcription factors (e.g. IRF1 and Stat1) exert their co-operative actions.

The present study brings together past observations regarding the relative potencies of IFNα2 and IFNβ. For instance, in primary human umbilical vein endothelial cells, IFNβ, used at 2 to 5 pM doses, was found to be 2-fold more potent than IFNα in Stat1 activation, a small difference relative to the 2–3 log difference measured in long term antiviral and anti-proliferative activities [40]. A basal USP18 level in these cells could account for the differential induction of ISGs measured as early as after 4 h of IFN stimulation. A well-studied ISG encoding the chemokine CXCL11 (β-R1/ITAC) was previously shown to be specifically induced by IFNβ and to require NF-κB (nuclear factor κB) activation [41]. Accordingly, we did observe a greater induction of CXCL11 in WISH cells stimulated with IFNβ than with IFNα2, particularly at later time points when USP18-mediated attenuation of IFNα2 and autocrine acting factors may come into play. Interestingly, in a physiological differentiation process of human monocytes, the 100-fold higher inhibition of osteoclastogenesis by IFNβ with respect to IFNα2 was proposed to be mediated, at least in part, by autocrine-acting CXCL11, whose expression in monocytes undergoing osteoclastic differentiation was more efficiently up-regulated by IFNβ [42].

In conclusion, USP18 is able to shift to a different extent the dose dependence of late responses to IFNα2 and IFNβ. In that respect, it is conceivable that, in any given cell type, a physiological level of USP18, constitutively expressed or maintained by low level autocrine/paracrine IFNβ or IFNλ [22], may set the sensitivity threshold to a pathogen- or stress-induced high level of IFN. Although USP18-mediated attenuation of IFNα signalling may protect infected cells from apoptotic death, the exclusive property of IFNβ to signal more persistently may, in defined cellular contexts, allow the establishment of an adaptive immune response. A recent study in a murine infection model showed that USP18 can be critical to the establishment of antiviral immune responses [43]. By restraining IFN responses in macrophages resident in the splenic marginal zone, USP18 allows local permissive VSV (vesicular stomatitis virus) infection that is necessary to secure sufficient antigen production and activation of the adaptive immune response. On the other hand in clinical settings USP18 may counteract the efficacy of therapeutic IFNα as for example in hepatitis C virus chronically infected patients, whereas high USP18 levels in pre-treatment livers has been associated with a poor response to treatment [44,45].


Véronique Francois-Newton and Mark Livingstone designed and performed experiments, analysed the data and wrote the paper; Béatrice Payelle-Brogard designed and performed the experiments shown in Figure 3; Gilles Uzé provided critical insight into data obtention and interpretation; and Sandra Pellegrini conceived and supervised the study and wrote the paper.


This work was supported by the European Community's Seventh Framework Programme (FP7/2007-2010) [grant number 223608 (to S.P. and G.U.)], the Institut Pasteur and the Centre National pour la Recherche Scientifique. V.F.N. was supported by the Ministère de l'Education Nationale, de la Recherche et de la Technologie and by the Ligue contre le Cancer, M.L. was supported by the Pasteur Foundation.


We thank the following for providing materials: Dr Dirk Gewert (Wellcome Foundation, Beckenham, U.K.) Dr Ernest Borden (Cleveland Clinic and Case Comprehensive Cancer Center, Taussig Cancer Institute, Cleveland, OH, U.S.A.), Professor Ganes Sen (Cleveland Clinic and Case Comprehensive Cancer Center, Taussig Cancer Institute, Cleveland, OH, U.S.A.), Dr Ara Hovanessian (Institut Pasteur, Paris, France) and Dr Otto Haller (University of Freiburg, Freiburg, Germany). Also: Zhi Li and Béatrice Corre for expert technical assistance; Pascal Roux (Plateforme Imagerie Dynamique PFID, Imagopole, Paris, France) for assistance with confocal microscopy; all of the members of our laboratory for discussions; and Frédérique Michel for critical reading of the paper prior to submission.

Abbreviations: 7-AAD, 7-aminoactinomycin D; CASP, caspase; CXCL11, C-X-C motif chemokine 11; DAPI, 4′,6-diamidino-2-phenylindole; F, forward; FADD, Fas-associated death domain; GAPDH, glyceraldehyde 3-phosphate dehydrogenase; IFIT, interferon-induced protein with tetratricopeptide repeats; IFN, interferon; IFNAR, IFNα receptor; IRF, IFN regulatory factor; ISG, IFN-stimulated gene; ISGF3, IFN-stimulated transcription factor 3; ISRE, IFN-stimulated response element; Jak, Janus kinase; mTOR, mammalian target of rapamycin; OAS, 2′-5′-oligoadenylate synthetase, 40/46kDa; PI3K, phosphoinositide 3-kinase; qRT-PCR, quantitative real-time PCR; R, reverse; siRNA, small interfering RNA; Stat, signal transducer and activator of transcription; TRAIL, tumour-necrosis-factor-related apoptosis-inducing ligand; TRIM22, tripartite motif-containing protein 22; USP18, ubiquitin-specific protease 18


View Abstract