Different requirements for the cytostatic and apoptotic effects of type I interferons. Induction of apoptosis requires ARF but not p53 in osteosarcoma cell lines.

The regulation of cell growth is one of the most important effects of type I interferons (IFNs). This response may involve a cytostatic effect or the induction of apoptosis depending on the cell context. Often the growth-inhibitory response of type I IFNs is studied in tumor cell lines carrying mutations of tumor suppressor genes, and therefore, the growth-inhibitory effect can be influenced by inactivation of these important regulators of cell proliferation. In this report, we explored the role of the ARF-p53 pathway in the growth-inhibitory effect of type I IFNs. We found that p53 is only induced in cells that express p14(ARF) (p19(ARF) in mouse cells). Surprisingly, mouse embryonal fibroblasts that are null for p19(ARF) or P53, even after transformation with oncogenic RAS, respond as well as wild type to the growth-inhibitory effect of type I IFNs. Similarly, human ARF(-/-) U2OS and P53(-/-) SAOS-2 cells show a significant decrease in cell proliferation. However, only SAOS-2 or U2OS reconstituted with inducible p14(ARF) undergo apoptosis in response to IFN beta treatment, and this effect was not inhibited by expression of dominant negative p53. These data suggest that (i) at least in specific cell types, the induction of apoptosis by type I IFNs requires an ARF pathway that is p53-independent and (ii) the cytostatic and pro-apoptotic effects of type I IFNs employ different pathways.

The product of the retinoblastoma gene, RB (for recent reviews see Refs. [1][2][3][4], regulates the passage from G 1 to S phase by tightly controlling the restriction point (reviewed in Refs. [1][2][3][4][5][6]. Malignant transformation always involves deregulation of the restriction point by either the absence of functional RB, the amplification of cyclin D or CDK4, or the loss of p16 INK4a (1,4). However, additional events are required for the emergence of a malignant phenotype, because hyperproliferative signals trigger built-in safety mechanisms that result in cell cycle arrest/senescence or apoptosis. The main effector of this safety mechanism is the ARF-MDM2-p53 pathway. For instance, the absence of RB triggers not only the transcription of S phase genes but also p19 ARF , which in turn inhibits MDM2, allowing p53 to stop cell cycle progression and induce apoptosis or senescence (7)(8)(9). The ARF-MDM2-p53 pathway is also a safety net for the activation of oncogenes such as RAS, MYC, v-Abl, and E1A. These mechanisms explain why most tumors not only have alterations of the RB pathway (i.e. either inactivation of tumor suppressors RB or INK4A, or amplification of cyclin D or CDK4/6) but also mutations in P53 or ARF that allow cells to escape cell cycle arrest and/or apoptosis. Thus, it is not surprising to find in many cancers the INK4A locus, which encodes both p16 and p19 ARF deleted and inactivated by methylation and/or mutated (10 -20).
Type I interferons (IFN␣, -␤, and -) 1 possess a wide variety of actions among which the antiviral and antiproliferative activities have captured most of the attention because of their potential therapeutic uses (21)(22)(23). Binding of type I IFNs to their receptors results in the activation of kinases of the Jak family (Jak1 and Tyk2), which are responsible for tyrosine phosphorylation of latent cytoplasmic transcription factors designated as signal transducers and activators of transcription 1 and 2 (STAT1 and STAT2). Phosphorylated STAT1, STAT2, and a protein of the IRF family, p48/IRF9, form the IFNstimulated gene (ISG) F3 complex, which translocates to the nucleus where it binds specific promoter regions (interferonstimulated response elements (ISRE)) to regulate gene transcription. Type I IFNs also induce the formation of STAT1-STAT1 and STAT1-STAT3 dimers that bind to specific DNA elements (IFN␥-activated sequence and c-sis-inducible element, GAS and SIE, respectively) (24,25) and together with ISGF3 regulate transcription of the specific genes responsible for the antiviral and antiproliferative effects of IFNs. Although the integrity of the Jak-STAT pathway is required for the antiviral effect (24,26,27), STAT activation does not appear to be required for the growth-inhibitory effect in some cell types (28).
The regulation of cell proliferation by type I IFNs is more complex than expected because it can involve a cytostatic effect or induction of apoptosis (29 -37). Interestingly, most studies dealing with the characterization of the signaling events that lead to the antiproliferative effect were performed in tumor cell lines with alterations of tumor suppressor genes or activation of oncogenes, which could affect cell cycle regulation and apoptosis. Therefore, it should not be surprising to find that type I IFNs have different effects on cell proliferation depending on the cell context. For example, type I IFNs can induce G 1 arrest in some responsive cell lines by inducing p21, p27, and INK4D or decreasing expression of cyclin D3 and cdc25A (33,34,(37)(38)(39). In other cases, IFN␣2 and IFN␤ induce apoptosis (27,31,32). In this scenario, the genotype of a particular tumor could determine whether IFN induces cell cycle arrest, apoptosis, or no response.
Although a recent report indicates that p53 is involved in the antiproliferative and antiviral effects of type I IFNs, the role of one of the most important activators of p53, ARF, is not addressed (40). Because most oncogenic agents such as RAS and MYC stabilize p53 via ARF, we sought to determine whether the ARF-p53 pathway was involved in the antiproliferative response triggered by type I IFNs. We found that type I IFNs up-regulate p53 via ARF, as demonstrated by the absence of p53 induction in ARF Ϫ/Ϫ mouse embryonal fibroblasts (MEFs) and in ARF null human U2OS cells. Yet IFN␣ or IFN␤ significantly reduced the proliferation rate in ARF-and p53-deficient MEFs and human ARF Ϫ/Ϫ U2OS cells. In these cells, the growth-inhibitory effect was characterized by a cytostatic response without signs of apoptosis. Surprisingly, reconstitution of ARF Ϫ/Ϫ U2OS with an inducible form of ARF allowed type I IFNs to induce apoptosis in a p53-independent manner. These results demonstrate that (i) at least in some cell types such as U2OS, ARF is responsible for the induction of apoptosis in a p53-independent manner and (ii) the apoptotic and cytostatic effects of type I IFNs do not share the same linear pathway.
Cell Proliferation and Apoptosis Assays-Cell proliferation was assessed by performing MTT assays (42) and cell counts using an automated cell counter. Briefly, the cells were seeded at 10,000 and 2,000 cells/well in 24-and 96-well plates, respectively, and treated with the indicated concentrations of IFNs. The numbers of cells/well were determined by trypsinization and counting of duplicate wells in a Coulter Counter. MTT assay was performed by adding 50 l of a 5 mg/ml MTT solution for the last 4 h of the experiment. Experiments were performed at least three times with similar results. Apoptosis was studied using Hoechst, trypan blue, and annexin V (Alexis Corp., Lausanne, Switzerland) staining, scoring at least 150 cells in random fields.
Induction of ISGs-Total RNA was isolated from SAOS-2 and U2OS cells treated with or without 10,000 units/ml huIFN␤ with TRIzol Reagent (Invitrogen). A total of 5 g from each sample was reverse transcribed using the cloned avian myeloblastosis virus first-strand cDNA synthesis kit from Invitrogen following the protocol suggested by the manufacturer. Following reverse transcription, the samples were processed for PCR using the MultiGene-12 reverse transcriptase-PCR profiling kit for human interferon response genes (SuperArray, Frederick, MD) following the protocol suggested by the manufacturer. The PCR program consisted of an initial incubation at 94°C for 5 min to denature the samples followed by 30 cycles of 94°C for 30 s, 50°C for 30 s, and 72°C for 45 s. After completion of PCR, 10 l of each sample was separated by agarose gel electrophoresis and stained and scanned as a digital image using a CCD camera. The levels of the reverse transcriptase-PCR gene expression profile were quantified by NIH Image (version 1.62). Normalized values are represented as -fold increases.

Type I IFNs Induce p53 via ARF in MEFs-
The ample variability among cancer cell lines to the antiproliferative effect of type I IFNs raises the question as to whether the integrity of specific tumor suppressor genes, which are commonly altered during the oncogenic process, influences the outcome of IFN treatment. Because activation of the ARF-p53 tumor suppressor pathway by oncogenic insults leads to cell cycle arrest and/or apoptosis, which are also attributes of the growth-inhibitory response of IFNs, we sought to determine whether type I IFNs activated this tumor surveillance pathway. To avoid the presence of mutations commonly observed in tumor-derived cell lines, these experiments were performed using wild type primary MEFs maintained in culture following a typical 3T3 protocol (Fig. 1A). Fig. 1B shows that p19 ARF is induced 24 h (lane 3) after muIFN␣4 treatment (1,000 units/ml) and is still expressed at high levels at 48 h (lane 4) in early passage MEFs (P2, passage 2). The induction of p19 ARF is still detected above the elevated base-line levels present in late passage MEFs (Fig.  1C, lane 2). As expected, the increase in p19 ARF was accompanied by a concomitant increase in p53 (Fig. 1C, lanes 3 and 5). The requirement of p19 ARF for the induction of p53 in MEFs was confirmed by finding that ARF Ϫ/Ϫ MEFs stimulated with muIFN␣4 for 48 h did not up-regulate the expression of p53 (Fig. 1D). However, muIFN␣4 induced p19 ARF in P53 Ϫ/Ϫ MEFs indicating that type I IFNs require p19 ARF for the induction of p53. It was not surprising that the induction of ARF in p53 Ϫ/Ϫ MEFs was not as pronounced as in wild type MEFs, because it has been reported previously that higher levels of ARF are normally detected in cells that lack p53 (43,44).
To determine whether ARF was also required for p53 induction in human cells, we treated NARF2 cells with huIFN␣.2. NARF2 cells correspond to ARF Ϫ/Ϫ P53 ϩ/ϩ U2OS cells in which expression of human p14 ARF is induced by isopropyl-␤-D-thiogalactopyranoside (IPTG) (43,45). Fig. 2A shows that IFN␣2 fails to increase the amount of p53 above base-line levels in NARF (lanes 4 and 5), whereas IPTG induces significant amounts of p14 ARF , p53, and its downstream target, p21 (lane 2). Identical results were observed when the experiments were performed with huIFN␤ (data not shown). The combination of IFN␣2 and IPTG did not further increase the levels of either ARF or p53 ( Fig. 2A, lane 3). However, IFN␤ was able to induce p14 ARF (Fig. 2B) but not p53 in SAOS-2 cells (ARF ϩ/ϩ P53 Ϫ/Ϫ ). The data obtained with MEFs and NARF2 cells indicate that p53 induction by type I IFNs requires functional ARF.
The Growth-inhibitory Effect Promoted by Type I IFNs Is Not Altered by the Lack of ARF and p53 or by Transformation with Oncogenic RAS V12 -We next determined the role of the ARF-p53 pathway in the cell growth-inhibitory effect of type I IFNs. Wild type MEFs derived from independent embryos show significant growth inhibition after treatment with muIFN␣4 (Fig.  3A). This growth-inhibitory response corresponded to a cytostatic effect, because no significant increase in apoptotic cells was observed (Fig. 3E). Interestingly, Fig. 3B shows that cells that lack p53 or ARF respond to muIFN␣4 with the same efficiency as wild type cells. This result suggests that either there are redundant pathways that compensate for the absence of ARF and p53 or that these tumor suppressors are not required for the cytostatic component of the growth-inhibitory effect of IFN␣ in MEFs.
Because oncogenic RAS V12 can transform P53 Ϫ/Ϫ and ARF Ϫ/Ϫ MEFs in the absence of MYC, we next tested whether RAS transformation affected the growth-inhibitory effect of type I IFN of ARF Ϫ/Ϫ or P53 Ϫ/Ϫ MEFs. To address this ques-tion, ARF or p53 null MEFs were infected with either empty pBabe-hygromycin vector or vector encoding RAS V12 selected with hygromycin B, and the surviving cells were kept as a pool to avoid the variability that could be observed when individual clones are isolated. The expression of oncogenic RAS in ARF Ϫ/Ϫ and P53 Ϫ/Ϫ MEFs was confirmed by Western blot analysis (Fig.  3C). Fig. 3D shows that ARF Ϫ/Ϫ and P53 Ϫ/Ϫ MEFs transformed with oncogenic RAS respond to muIFN␣4 (1,000 units/ml) just as well as or even better than control cells. Although muIFN␣4 induced slightly more apoptosis in retrovirus-transduced cells than in wild type controls, treatment with muIFN␣4 did not significantly increase the number of apoptotic cells in either ARF Ϫ/Ϫ and P53 Ϫ/Ϫ MEFs expressing oncogenic RAS V12 compared with vector controls (Fig. 3E). These data suggest that the cytostatic component of the growth-inhibitory effect of type I IFNs is not affected by oncogenic RAS or the null mutation of either ARF or p53.

ARF (but Not p53) Contributes to Type I IFN-induced Apoptosis in Human Cells-
The lack of requirement of either ARF or p53 for the growth-inhibitory effect of type I IFNs in MEFs raised the possibility that these tumor suppressors could be important in the growth-inhibitory effect in cells other than MEFs. Therefore, we next determined the role of ARF and p53 in the IFN response in human cells. We first studied the effect of IFN␤ on U2OS (ARF Ϫ/Ϫ P53 ϩ/ϩ ) and SAOS-2 (ARF ϩ/ϩ P53 Ϫ/Ϫ ) cells because they lack ARF and p53, respectively. Although both human cell lines showed a good response to huIFN␤ treatment (Fig. 4), SAOS-2 cells were more sensitive, and after a few days in culture, the presence of floating dead cells was evident. The study of apoptosis by annexin V staining confirms that almost 50% of SAOS-2 (but not U2OS) cells undergo cell death after 3 days of treatment with 1,000 units/ml IFN␤ (Fig. 4C). These data suggest that although a growth-inhibitory response to type I IFNs can be observed in the absence of ARF or p53 in human osteosarcoma cell lines, apoptosis was only observed in ARF ϩ/ϩ SAOS-2 cells.
To determine whether ARF plays a role in type I IFN-induced apoptosis through a pathway that is independent of p53, we used NARF2 cells stably expressing either dominant negative p53 (NARF-DNp53) or empty vector (NARF-Puro). To confirm the inactivation of wild type p53 by the DNp53, we studied the induction of p53 and its downstream target, p21, by ARF in control and NARF-DNp53 cells. Fig. 5 shows that IPTG induces ARF, p53, and p21 in control NARF-Puro cells (lane 2). Treatment with 1,000 units/ml IFN␤ alone or in association with 1 mM IPTG did not affect the induction of ARF, p53, or p21 observed with IPTG treatment alone (Fig. 5, lanes 3 and 4,  respectively). Although in NARF-DNp53 the levels of p53 are significantly elevated even in the absence of ARF, the induction of p21 is almost completely obliterated, demonstrating that DNp53 effectively inhibits the function of the endogenous wild type p53 protein.
We next determined whether IFN␤ would induce cell death in NARF cells in the presence of ARF. Trypan blue or annexin V staining (Fig. 6, A and B, respectively) showed that treatment with IPTG and IFN␤ for 3 days induced higher levels of apoptosis than either drug alone in NARF-Puro cells. Interestingly, the presence of DNp53 did not inhibit the apoptotic effect of IFN␤ plus IPTG, suggesting that the induction of ARF, rather than p53, was responsible for this effect. Surprisingly, NARF-DNp53 was consistently more sensitive to the induction of apoptosis by IFN␤ (Fig. 6B, compare treatment with 100 units/ml IFN␤ alone and plus IPTG). The reason for this higher sensitivity could be the low levels of ARF present in the absence of IPTG (leaky promoter), which cannot be detected by Western blotting. Alternatively, it could be because of other Type I Interferons Activate the ARF-p53 Pathway factors, as recently reported for NARF cells expressing the human papilloma virus E6 protein, which showed increased sensitivity to ARF-mediated apoptosis in response to tumor necrosis factor (46).
Finally, we determined whether the induction of apoptosis correlated with higher levels of the induction of ISGs. To achieve this goal, we performed semiquantitative reverse transcriptase-PCR after stimulation of SAOS-2 and U2OS with 10,000 units/ml IFN␤ for 24 h. Fig. 7 shows the fold increase over control in the level of expression of guanylate-binding protein 1 (GBP1), interferon-induced protein 35 (IFI35), interferon-inducible 56-kDa protein (IFIT1), interferon regulatory factor 7 (IRF7), and oligoadenylate synthetase 1 (OAS1). As observed, there are marginal differences in the fold induction of GPB1 and IFI35 between SAOS-2 and U2OS. However, the fold induction of IFIT1, IRF7, and OAS1 was substantially greater in SAOS-2 than in U2OS. These data suggest that ARF may modulate the expression of select ISGs such as OAS1, which is required for the induction of apoptosis by type I IFNs in some cell lines (47,48). DISCUSSION One of the most important biological effects of type I IFNs is the regulation of cell proliferation either through cytostatic or pro-apoptosis mechanisms (29 -37). This property was the basis for the use of type I IFNs for the treatment of different malignancies (49). However, there is ample variability in the response to type I IFN treatment in the clinical setting and even among tumor cell lines in vitro. One possible explanation for this variability is that genes that are commonly mutated in cancers, such as tumor-suppressor genes, may be required for the antiproliferative effect of IFNs.
ARF and p53 are commonly activated by oncogenic viruses and genotoxic agents, respectively, and therefore constitute an important safety mechanism against tumor development (19, 50 -53). Therefore, it is not surprising to find that a large proportion of cancers contain mutations of ARF or p53. To test the hypothesis that ARF or p53 could be important in the growth-inhibitory effect of type I IFNs, we first studied the induction of these proteins in wild type MEFs to avoid the presence of mutations of these genes. We found that type I IFNs induce ARF and p53 in wild type but not in ARF Ϫ/Ϫ MEFs, supporting the concept that the induction of p53 is via ARF. Similarly, IFN was unable to induce p53 in human NARF cells in which endogenous ARF had been inactivated, confirming that a similar mechanism for the induction of p53 is employed in human cells. Surprisingly, the absence of ARF or P53 did not impair the antiproliferative effect of type I IFNs in MEFs. Additionally, transformation of ARF Ϫ/Ϫ or P53 Ϫ/Ϫ MEFs by oncogenic RAS did not reduce the response to type I IFNs. These findings raised different possibilities. First, the activation of the ARF-p53 pathway was not required for the antiproliferative effect, or second, there are redundant mechanisms that can compensate for the absence of ARF and p53. Alternatively, the role of ARF and/or p53 could be cell typespecific. To address the latter possibility, we characterized the role of ARF and p53 in the type I IFN response in human cells.

FIG. 3. Effect of IFN on the proliferation of MEFs.
Wild type (A), ARF Ϫ/Ϫ , and p53 Ϫ/Ϫ (B) MEFs were grown in the absence or presence of varying concentrations of muIFN␣4, and the rate of proliferation was assessed as a percent of control using an MTT assay as described under ''Materials and Methods.'' C, ARF Ϫ/Ϫ and p53 Ϫ/Ϫ MEFs were transduced with either empty vector (Vec) or oncogenic RAS V12 (RAS) viruses, and the protein expression of RAS and tubulin was determined by immunoblotting as described under ''Materials and Methods.'' D, ARF Ϫ/Ϫ and p53 Ϫ/Ϫ MEFs, transduced with either empty vector or oncogenic RAS V12 viruses, were grown in the absence or presence of varying concentrations of muIFN␣4, and the rate of proliferation was assessed as in A. The coefficient of variability in all proliferation assays was less than 10%, and similar results were obtained in at least three independent experiments. E, wild type, ARF Ϫ/Ϫ , and p53 Ϫ/Ϫ MEFs transduced with either empty vector or oncogenic RAS V12 were grown in the absence or presence of muIFN␣4 (1,000 units/ml), and the level of apoptosis was determined by annexin V staining.
These studies also revealed that a significant antiproliferative response could be obtained in P53 Ϫ/Ϫ SAOS-2 cells or in ARF Ϫ/Ϫ U2OS cells. Surprisingly, the type I IFN response in ARF ϩ/ϩ SAOS-2 cells (but not in ARF Ϫ/Ϫ U2OS cells) involved the induction of apoptosis, raising the possibility that the presence of ARF was responsible for this effect. Interestingly, select ISGs were substantially induced in SAOS-2 compared with U2OS after IFN treatment. Among the ISGs profiled, OAS1 may play a critical role in apoptosis. Previous work has shown that splice variants of OAS have pro-apoptotic functions (48). Although the role of ARF in the induction of ISGs is unclear at this time, our data demonstrate an increase in the expression of select ISGs in SAOS-2 cells, which express p14 ARF . Further studies performed in NARF cells, which contain IPTG-inducible ARF, demonstrate that the presence of ARF allowed IFN␤ to induce apoptosis. Moreover, the apoptotic effect was not blocked by dominant negative p53, supporting the concept that IFN induces apoptosis via a p53-independent pathway.

FIG. 4. Effect of IFN on the proliferation of SAOS-2 and U2OS.
A, SAOS-2, deficient in p53, was grown in the absence or presence of varying concentrations of human IFN␤ for up to a period of 7 days, and the rate of proliferation was determined as described under ''Materials and Methods.'' B, U2OS, deficient in ARF, was similarly grown in the absence or presence of the indicated concentrations of human IFN␤, and the rate of proliferation was determined using MTT assays. C, the percentage of apoptotic cells was determined in both U2OS and SAOS-2 after treatment with IFN␤ (100 units/ml) for 3 days by annexin V staining. A, U2OS, NARF-Puro, and NARF-DNp53 were seeded in duplicate 60-mm dishes, treated with IFN␤ (10 units/ml) in the presence or absence of IPTG (1 mM) for 3 days, and scored for apoptosis using Hoechst staining as described under ''Materials and Methods.'' Similar results were obtained in two independent experiments. The standard deviation was less than 5% in both experiments. B, NARF-Puro and NARF-DNp53 were treated with IFN␤ (100 units/ml) or a combination of IFN␤ and IPTG (1 mM) for 3 days, and apoptosis was measured by annexin V staining.

FIG. 7. Induction of ISGs is increased in cells expressing ARF.
SAOS-2 and U2OS were treated with IFN␤ (10,000 units/ml) for 24 h. Following treatment, total RNA was isolated and reverse transcribed. After reverse transcription, the samples were amplified by PCR using gene-specific primers for known ISGs and analyzed as described under ''Materials and Methods.'' The data obtained with MEFs do not contradict a previous report indicating that the induction of p53 by type I IFNs plays a role in tumor suppression (40). For example, the authors reported that the effect of IFN␤ via p53 was only observed when MEFs were previously challenged with human papilloma virus E6 or a DNA-damaging agent, whereas we assessed the effect of p53 under normal culture conditions. Moreover, different methods were used to assess the growth-inhibitory effect of IFN␤ (soft agar assays versus growth curves in tissue culture plates). More importantly, Takaoka et al. (40) did not address the long term effect of IFN␤ treatment (beyond 6 -9 h) or the role of ARF in the induction of p53. It is possible that an initial direct effect on the p53 promoter is responsible for an early induction of the protein, whereas the sustained induction of p53 requires ARF.
Altogether, our data demonstrate that (i) the cytostatic effect of type I IFNs does not require ARF or p53 in MEFs or human osteosarcoma cell lines; (ii) the antiproliferative effect of type I IFNs is not inhibited by oncogenic RAS; (iii) the induction of apoptosis requires ARF (but not p53) in U2OS or SAOS-2 cells, and (iv) there are clear differences in the roles of the IFN-ARF pathway among cell types, because primary MEFs did not undergo cell death after type I IFN treatment even though ARF and p53 were induced, whereas apoptosis was induced in human osteosarcoma lines. It remains to be established what targets play a critical role in the cytostatic effect of type I IFNs. It has been reported that type I IFNs can affect other cell cycle regulators such as p27, INK4D, cyclin D3, and CDC25A (33,34,(37)(38)(39). However, an important question not yet addressed is whether the changes induced by type I IFNs on some cell cycle regulators are a consequence or are directly responsible for the antiproliferative effect. For example, the increase in the level of expression of p27 may only reflect an exit of the cell cycle rather than being responsible for the growth inhibition. The use of knock-out models for those genes believed to be involved in the growth-inhibitory effect of IFNs will help to settle these issues.