Dominant Negative Signal Transducer and Activator of Transcription 2 (STAT2) Protein: Stable Expression Blocks Interferon α Action in Skin Squamous Cell Carcinoma Cells¹

Introduction

Non-melanoma skin cancer is the most common cancer in the United States, with over one million new cases of the two most common forms, SCC³ and basal cell carcinoma, anticipated in 2002 (1). The more clinically aggressive form is SCC of the skin (2), which has been increasing in incidence since the 1960s at annual rates from 4% to as much as 10% in recent years. Unlike early-stage SCC, advanced SCC is aggressive, often resistant to local therapy, requires repeated surgical resections and courses of radiotherapy, and accounts for approximately 2000 deaths each year in the United States. Better control of advanced skin SCC is clearly necessary, presenting a great challenge to clinical oncologists.

IFNs regulate proliferation, differentiation, and immune function (3). IFN-α has been shown to be an active chemopreventive agent in the treatment of IEN of the skin (4, 5). IFN-α as well as the other type 1 IFN, IFN-β, binds to cell surface receptors composed of two distinct subunits, IFNAR1 and IFNAR2, causing their oligomerization (6). This results in the activation (autophosphorylation and/or transphosphorylation) of receptor-associated kinases JAK1 and tyk2, members of the JAK family of receptor-associated tyrosine kinases. JAK1 and tyk2 phosphorylate the receptor subunits along with several different STATs, members of a family of latent cytoplasmic transcription factors, resulting in STAT dimerization and translocation to the nucleus, where they modulate gene transcription. STAT1 and STAT2 are likely to be the most important STATs mediating IFN-α effects and are phosphorylated on tyrosine 701 and tyrosine 690, respectively (3). Upon phosphorylation, STAT1 and STAT2 proteins complex with a third protein, IRF9 (also called p48), to form the ISGF-3 transcription factor (7). After translocation to the nucleus, the IRF9 component of ISGF-3 binds to specific DNA elements found in the promoters of most type 1 IFN-responsive genes (7).

The activation of both STAT1 and STAT2 by IFN-α can be abolished by mutating tyrosine 701 or tyrosine 690, respectively (8, 9). The mutant form of STAT1 (STAT1-Y701F) can act in a dominant negative fashion to block ISGF-3 formation (8). The corresponding dnSTAT2 (STAT2-Y690F) prevented the phosphorylation of STAT1 upon IFN-α treatment, thereby presumably blocking IFN-α responses mediated by ISGF-3 (9). Current evidence suggests that STAT2 is specific for the IFN-α pathway, unlike other STAT family members, which can mediate signaling by multiple inducers (6). It is therefore expected that expression of dnSTAT2 protein will only block the IFN-α pathway and not others, clearly linking any phenotypic change to the IFN-α pathway. It should also be noted that IFN-α can activate other STATs besides STAT1 and STAT2 (3). For example, IFN-α can induce the phosphorylation and DNA binding of STAT3 and STAT4 proteins, thereby regulating additional sets of genes not common to the ISGF-3 pathway (10, 11).

In an effort to better understand the role of IFN-α in skin carcinogenesis, we have previously determined the expression pattern of the ISGF-3 proteins (STAT1α/β, STAT2, and IRF9) in normal skin, skin SCC, and actinic keratoses from patient biopsies (12, 13). Our results indicated a suppressed expression of one or more of these proteins in the majority of patients tested, as determined by manual scoring and quantitative densitometry. We have observed a similar decrease in expression of the same set of IFN signaling proteins in actinic keratoses (skin IEN; Ref. 14), indicating that the suppression may be an early event in skin carcinogenesis (13). These data have led to the hypothesis that the suppressed expression of these proteins may result in reduced responsiveness to type 1 IFNs, conferring a growth or survival advantage to those cells. To test this hypothesis, we have attempted to permanently block IFN-α signaling in a skin cell-based system through the forced expression of the dnSTAT2 protein. We demonstrate that expression of the dnSTAT2 protein can specifically suppress IFN-α responses in the IFN-α-sensitive SRB12-p9 human skin SCC cell line.

Materials and Methods

Cell Culture and Generation of Stable dnSTAT2-expressing Cell Lines.

The human skin SCC cell line SRB12-p9 was derived by single cell cloning from SRB-12 cells (a gift from Dr. Janet Price; Department of Cancer Biology, The University of Texas M. D. Anderson Cancer Center). Cells were cultured in a humidified atmosphere at 5% CO2, in a 1:1 mixture of DMEM and Ham’s F-12 medium, plus 10% fetal bovine serum (15). The pSG5-dnSTAT2 expression vector was constructed by ligating the EcoRI insert fragment of pBSK-STAT2 (kindly provided by J. E. Darnell; Rockefeller University) into the EcoRI site of the pSG5 expression vector (16), followed by sequential PCR-based site-specific mutagenesis using the oligonucleotides 5′-ATGCCTTCTGACTTCAGATCTAGGAACCACATTTC-3′ and 5′-GAAATGTGGTTCCTAGATCTGAAGTCAGAAGGCAT-3′ to introduce a BglII site immediately downstream of codon 851 and the oligonucleotides 5′-GGAACGGAGGAAATTCCTGAAACACAGGC- TC-3′ and 5′-GAGCCTGTGTTTCAGGAATTTCCTCCGTTCC-3′ to convert tyrosine 690 into phenylalanine. The FLAG epitope extension was created by inserting a DNA fragment consisting of a complementary oligonucleotide pair (5′-GATCGGACAAAGACGATGACGATAAATAGTAGATC-3′and 5′-GATCGATCTACTATTTATCGTCATCGTCTTTGTCC-3′) into the newly generated XhoI site. The resulting vector, pSG5-dnSTAT2-FLAG, was linearized with AatII and electroporated along with the bacterial neomycin phosphotransferase gene expression vector pKJ1, and stably expressing cell clones were isolated essentially as described previously (17). Cells (5 × 10⁶) suspended in 800 μl of PBS were electroporated with 5 μg of linearized, purified pSG5-dnSTAT2 and 0.5 μg of the bacterial neomycin phosphotransferase gene (neo^R) expression vector pHR56 (kindly provided by P. Chambon) with a Bio-Rad Gene Pulser set at 200 V and 960 μF. Cells were then plated at a density of approximately 1 × 10⁶ cells/10-cm culture plate and, after 24 h, subjected to neomycin selection (300 μg/ml G418 sulfate, Life Technologies Inc., Rockville, MD) for up to 14 days. Individual colonies were isolated, propagated, and divided into two aliquots, one for freezing and the other for expansion and Western blotting.

Western Blotting.

Whole cell extracts of stably transfected cells were purified as described previously (17). Proteins were quantitated by the Bradford assay (Pierce, Rockford, IL), and 50–100 μg protein/lane were electrophoresed by SDS-PAGE and electrophoretically transferred to nitrocellulose membranes. After transfer, blots were blocked with 3% milk powder for 1 h at room temperature, followed by incubation for 1 h at room temperature with rabbit polyclonal antibodies to the FLAG epitope (Sigma-Aldrich, St. Louis, MO), STAT2 (Santa Cruz Biotechnology, Santa Cruz, CA), STAT1 (Santa Cruz Biotechnology), phospho-STAT2 (Upstate Biotechnology, Lake Placid, NY), phospho-STAT1 (Cell Signaling Technology, Beverly, MA), and β-actin (Sigma-Aldrich). Blots were then washed three times (15 min each time) with PBS and 0.05% Tween and incubated with horseradish peroxidase-conjugated donkey antirabbit secondary antibody for 1 h at room temperature, followed by an additional three washes with PBS and 0.05% Tween. The blot was then incubated for 10 s to 1 min in chemiluminescence detection solution (Amersham Life Science Inc., Piscataway, NJ) and autoradiographed.

Cell Growth Inhibition Assays and IFN Treatments.

Cell growth inhibition assays were carried out essentially as described previously (18). Cells were plated in 24-well plates at identical densities in normal culture media 1 day before treatment with 100 IU/ml human IFN-α (PBL Biomedical Laboratories, New Brunswick, NY). Media were changed after 2 days, and the cells received fresh IFN-α for an additional 2 days. The percentage of cell growth inhibition was determined by cell counting using a Coulter cell counter (Coulter Electronics Inc., Hialeah, FL). The percentage of growth inhibition was calculated using the equation: (1 − R/C) × 100, where R and C represent the number of cells in IFN-α-treated and control culture, respectively. For all other treatments with IFN-α, cells were grown in normal culture media and switched to 0.5% FBS for 48 h followed by a 2-h incubation in serum free media, immediately before treatment with human IFN-α. For phospho-STAT Western blot experiments, cells were treated with 100 IU/ml IFN-α for 30 min. For RNA isolation for cDNA microarray screening, cells were treated with 100 IU/ml IFN-α for either 1 or 12 h, and for RT-PCR assays, cells were treated with 100 or 500 IU/ml IFN-α for 12 and 24 h.

cDNA Microarray Screening.

Total RNA was isolated by homogenization of fresh or snap-frozen epidermal cells using TriReagent (Molecular Research Center Inc., Cincinnati, OH), followed by standard organic extraction and precipitation. Purity and yield were determined by UV absorbance spectra over the range of 220–320 nm. Samples of total RNA (10–30 μg) were fractionated on 1% agarose gels containing 0.66 m formaldehyde to determine integrity, and 100 mg total RNA/sample was used for fluorescent probe synthesis. Probe synthesis and array hybridization were performed within The University of Texas M. D. Anderson Cancer Center Cancer Genomics Core Laboratory (Dr. W. Zhang, director), using established methods. Briefly, RNA was reverse transcribed and labeled by an indirect labeling method according to the manufacturer’s instructions (Clontech, Palo Alto, CA). The probe from untreated cells was labeled with Cy5, a red fluorescent dye, and the probe from IFN-α-treated cells was labeled with Cy3, a green dye. Slide hybridization was also conducted in the core facility according to the manufacturer’s instructions (Clontech). Fluorometric detection of probed slides was carried out for Cy5 (A_{650 nm}) and Cy3 (A_{550 nm}) with gray scale images. A colorized, merged image showed a red signal for spots hybridizing primarily with the control skin probe, a green signal for spots hybridizing primarily with the treated skin probe, and a yellow signal for spots hybridizing equally with both probes. Signal:noise ratio adjustment and normalization of signals were conducted within The University of Texas M. D. Anderson Cancer Center Cancer Genomics Core Laboratory using Arrayvision quantification software (Imaging Research, St. Catherines, Ontario, Canada).

Semiquantitative RT-PCR.

Semiquantitative RT-PCR was performed essentially as described previously (19). RNA was purified as described above and quantitated by absorbance. RNA (1 μg/reaction) and the appropriate 20-mer oligonucleotides (50 pmol/reaction) were combined with a 10× PCR mix [final concentrations, 50 mm KCl, 10 mm Tris (pH 8.3), 1.5 μm Mg₂Cl, and 200 μm each of dATP, dCTP, dGTP, and dTTP] to a final volume of 100 ml and subjected to the following PCR parameters: (94°, 3 min; 94° to 50° slope, 10 min; 50°C, 22 min) × 1 cycle; followed by (94°, 1 min; 55°, 30 s; 72°, 1 min) × 15–40 cycles. A mix of 5 ml of a Taq polymerase (2.5 units/tube) and avian myeloblastosis virus reverse transcriptase (4 units/tube) was added to each tube immediately after the 94° to 50° slope. Aliquots of each reaction were collected over a broad range of cycle numbers and electrophoresed in a 2% agarose gel containing ethidium bromide. RT-PCR products that were just below the visual limit of detection were blotted onto nylon membranes by capillary transfer in high-salt buffer. Blots were probed with [γ-³²P]ATP]-end-labeled oligonucleotide probes complementary to sequences contained between the oligonucleotides used for the RT-PCR. The expression of the glyceraldehyde phosphate dehydrogenase gene, which is ubiquitously expressed, was determined for each RNA sample to control for variations in RNA quantity.

Results

Generation of dnSTAT2-expressing SCC Cells.

Initially, SRB12-p9 cells, which were derived from the SRB-12 cell line (15), were stably transfected with an expression vector encoding a dnSTAT2 protein linked to the FLAG octapeptide at the COOH terminus, controlled by the SV40 promoter (Fig. 1A). The expression of dnSTAT2-transfected cell clones was determined by Western blotting with an anti-FLAG antibody (Fig. 1B, top panel). Note that the faint bands observed in the neo^R lane are nonspecific. The same blots were stripped and reprobed with an anti-STAT2 antibody that recognizes both endogenous and dnSTAT2, revealing a >10-fold increase in total STAT2 (endogenous STAT2 plus dnSTAT2) immunoreactivity in the dnSTAT2-expressing cell clones (Fig. 1B, middle panel; data not shown). The increase in total STAT2 is likely due to high expression levels of dnSTAT2, not to an increase in endogenous STAT2 expression (see explanation below).

dnSTAT2 Blocks Cell Growth Inhibition by IFN-α through the ISGF-3 Pathway.

The parental SRB12-p9 cells (Fig. 2, Par) are typically growth inhibited 45–75% after 4 days of treatment with 100 IU/ml IFN-α, a pharmacologically achievable concentration. For the independent dnSTAT2 cell clones shown in Fig. 1B (c1 and c3), there was approximately 15% growth inhibition compared with 47.5% and 43.9% for the parental SRB12-p9 and neo^R cells, respectively, averaged from 3 independent experiments (Fig. 2). We have previously demonstrated that the cell growth inhibitory action of IFN-α in the SRB12-p9 parental cells is primarily due to apoptosis induction and not to inhibition of proliferation (5).

We next sought to determine whether the observed differences in IFN-α-induced growth inhibition are due to a specific suppression in IFN-α signaling through the IFNAR/ISGF-3 pathway. We compared the IFN-α-induced phosphorylation of STAT2 and STAT1, which occurs rapidly in response to IFNAR activation, between parental and dnSTAT2-expressing cells using antibodies specific for the tyrosine-phosphorylated forms of STAT2 and STAT1. Tyrosine phosphorylation of both STAT2 and STAT1, induced by a 30-min treatment with 100 IU/ml IFN-α, was almost completely blocked in the dnSTAT2-expressing cells [Fig. 3A, phospho STAT2 and phospho STAT1, compare Lanes α (parental) and neo^R (controls) with corresponding lanes for clones 1 and 3 (Lanes c1 and c3)]. The level of total STAT1 and STAT2 proteins was unchanged by the IFN-α treatment, and STAT1 levels were unaffected by dnSTAT2 expression (Fig. 3A, STAT2 and STAT1). Total STAT2 immunoreactivity was markedly higher in the dnSTAT2-expressing cells and was also unchanged by IFN-α treatment (Fig. 3A, second panel). We believe that these higher total STAT2 levels were not due to an increase in endogenous STAT2 expression because the phospho-STAT2 antibody (Fig. 3A, top panel) will only recognize the endogenous STAT2 (because dnSTAT2 cannot be phosphorylated at tyrosine 690). The markedly lower signal for phospho-STAT2 in Lanes α for c1 and c3 is consistent with a reduction in the proportion of endogenous STAT2 protein in those cells, compared with the parental (Par) and neo^R lines. These findings parallel the growth inhibition results and strongly suggest that the ISGF-3 complex mediates this effect. The ability of IFN-γ to induce phosphorylation of STAT1 was unaffected in the c3 cells (Fig. 3B), indicating that the presence of dnSTAT2 did not interfere with the activation of STAT1 by the IFN-γ receptor.

dnSTAT2 Suppresses Induction Several IFN-α-responsive Genes.

The recent development of cDNA microarray technology has made it possible to compare gene expression profiles between two different cell or tissue samples. In collaboration with The University of Texas M. D. Anderson Cancer Center Cancer Genomics Core Laboratory, we have identified several early IFN-α-inducible genes in SRB12-p9 cells. Three separate cDNA array screens were conducted. The first compared the expression of genes between control cells and cells treated with 100 IU/ml IFN-α for 1 h, using an array containing 2304 duplicate spotted unique cDNA fragments. The second and third screens compared control cells with cells treated with 100 IU/ml IFN-α for 12 h using two arrays, one containing 1127 duplicate spotted cDNA fragments corresponding to genes previously shown to be active in several signal transduction pathways, and the other array containing 5700 spotted expressed sequence tag fragments. All screens identified a small number of gene expression changes (where 0.5 ≥ Cy5/Cy3 ≥ 2.0), as was expected with the short treatment time and relatively low dose of IFN-α used. A total of 16 genes were shown to be up-regulated by IFN-α in the screens, and 5 genes were suppressed by IFN-α in the screens. To date, individual gene expression changes were confirmed for five of the up-regulated genes by RT-PCR analysis of SRB12-p9 cells treated with 100 or 500 IU/ml IFN-α for 1, 12, and 24 h (Fig. 4, A and B; data not shown). All five of the confirmed genes, myxovirus resistance 1 (MxA1), IFN-α-inducible protein 6-16 (IFI-6-16), IFN- induced protein 56 (IFI-56), IFN-regulatory factor 7 (IRF-7), and the IFN-inducible protein 9-27 (9-27), have previously been shown to be IFN-α inducible in other systems (3).

We next determined whether the IFN-α induction of any of these five genes was suppressed in the c1 and c3 cells. All five genes were still induced by a 12-h treatment with 100 or 500 IU/ml IFN-α in the c1 cells and by a 12-h treatment with 500 IU/ml IFN-α in the c3 cells (Fig. 4B). Three of five genes were less induced at both IFN-α concentrations for c1 cells, and all five genes were less induced for c3 cells. After a 24-h treatment with either 100 or 500 IU/ml IFN-α, all five genes are less induced in both dnSTAT2 cell lines, with the exception of IRF-7, for c1 cells treated with 500 IU/ml IFN-α (Fig. 4B). There is an overall greater suppression of IFN-α-induced gene expression in the c3 cells compared with c1 cells (Fig. 1B).

Discussion

In this study we demonstrate by several criteria that stable expression of dnSTAT2 can reduce IFN-α responsiveness of otherwise highly IFN-α-sensitive SRB12-p9 cells. Of the two dnSTAT2-expressing cell lines analyzed, c3 cells, which express the higher amount of dnSTAT2 protein, show a stronger overall suppression of IFN-α-induced gene expression. The c3 cells also show a slightly greater suppression of IFN-α-induced phosphorylation of STAT1, which was observed in several independent experiments (Fig. 3A; data not shown). These results provide further evidence for the specificity of the dnSTAT2 effect.

The specific mechanism by which the dnSTAT2 (Y690F) phosphorylation site mutation blocks IFN-α signaling remains to be determined. Wild-type STAT2 normally interacts constitutively with the IFNAR2 chain, in both its tyrosine-phosphorylated and unphosphorylated forms (20). Upon binding of IFN-α to the receptor complex, STAT2 is recruited by way of its Src homology domain 2 to the phospho-tyrosine 466 site of the IFNAR1 chain (21). This event is followed by the phosphorylation of STAT2 on tyrosine 690 and recruitment of STAT1 to the complex, whereupon STAT1 is tyrosine phosphorylated (8). Our observation that IFN-α- induced phosphorylation of STAT1 is suppressed in the dnSTAT2-expressing cells suggests that the dnSTAT2 protein acts proximal to the STAT1 tyrosine phosphorylation step. We observe a partial block of IFN-α responses, consistent with the findings of other investigators who have observed a partial suppression of IFN-α-induced transcription by dnSTAT2 in transient transcription assays (22). Those investigators speculated that the dnSTAT2 protein exerts its effect by competing with endogenous STAT2 for binding to the IFNAR1 chain, thereby only partially blocking recruitment of STAT1 (22). Nevertheless, we cannot rule out the possibility that alternate pathways besides ISGF-3 activation are available for induction of the growth inhibitory response or the induction of ISGs by IFN-α in the c1 and c3 cells.

The inability of the dnSTAT2 protein to block IFN-γ- induced phosphorylation of STAT1 rules out other possible effects of dnSTAT2 expression on STAT1 function and allows us to conclude that STAT1 function remains intact. Further study will be necessary to determine the exact point of action of the dnSTAT2 protein. It should be noted that STAT3 phosphorylation, which can also be induced by IFN-α (10), is constitutive in SRB12-p9 cells and is not affected by expression of dnSTAT2 (data not shown).

Importantly, the cell growth inhibitory effect of 100 IU/ml IFN-α, a pharmacologically relevant concentration, on the dnSTAT2-expressing cells is markedly reduced compared with the parental cells. We have recently shown that IFN-α-induced growth inhibition of the parental SRB12-p9 cells does not correlate with changes in cell cycle distribution but rather with an induction of apoptosis (5). A recent study by Guo et al. (23) provides a possible explanation for this effect. Those investigators showed that IFI-56, one of the IFN-inducible genes that we find suppressed in the dnSTAT2 cells (Fig. 4, A and B), plays a role in inhibiting protein translation by binding to the P48 subunit [not to be confused with the p48 (IRF9) subunit of ISGF-3] of the translation initiation factor eIF3 (23). It is possible that IFI-56 could act in a manner similar to another IFN-inducible gene, double-stranded RNA-dependent protein kinase (PKR), which has also been shown to suppress protein translation by binding to eIF2α (24). In this latter study, PKR action was blocked through expression of a dominant negative variant, resulting in resistance to apoptosis induction by treatment with double- stranded RNA for 3T3 L1 cells (24). The exact mechanism of action of PKR in mediating apoptosis induction is not completely understood, but it appears to involve the up-regulation of several proapoptotic proteins. We speculate that IFI-56 may act in a similar manner. Additional studies are planned to determine the mechanisms of IFN-α-induced apoptosis and the possible involvement of IFI-56 in this system.

A role for IFN signaling in skin SCC development and progression is not clear at present. Previous studies have shown that IFN-α can suppress the proliferation of primary cultured keratinocytes (25) and induce apoptosis in skin SCC cells (5, 12). A requirement for STAT1 activation for the antiproliferative effects of IFN-α, as well as the apoptotic effect of tumor necrosis factor α, has been demonstrated (26, 27). Further support for the possibility that reduced IFN-α responsiveness could be related to malignancy comes from studies aimed at determining the mechanism of IFN resistance in various cancer cell types. Other investigators have shown that cultured cancer cells have deficiencies in ISGF-3 activity, and in one case, a series of melanoma cell lines exhibited reduced IFN response relative to primary melanocytes (28, 29). It was also shown that IRF9 and STAT1 levels and DNA binding activity increased during monocytic differentiation of U937 cells (30). Recently, it was found that arginine methylation of STAT1 is required for IFN-α-induced transcription and that this modification was inhibited in certain transformed cell lines (31). This indicates an inverse correlation between STAT1 and STAT2 expression and malignancy. These results, along with the finding that some DNA tumor viruses inhibit IFN action, are in keeping with the assertion that IFNs have a role as tumor suppressors (32).

IFN-α also could potentially play a role in tumor surveillance in normal or premalignant skin. In a mechanism similar to type 1 IFN signaling, type 2 IFN (IFN-γ) activates a distinct set of genes through the formation of STAT1 homodimers (3, 6). Kaplan et al. (33) have previously demonstrated that mice lacking either IFN-γ receptors or STAT1 protein develop spontaneous and chemically induced tumors more frequently than wild-type mice. They also demonstrated defects in IFN-γ signaling in several lung adenocarcinoma cell lines, but they did not find similar defects in IFN-α signaling, leaving a question as to the potential role of IFN-α in tumor surveillance. Because STAT1 is shared between the type 1 and type 2 IFN pathways, our previous finding that its expression is reduced in human tumors (18, 19) could reflect a defect in both pathways. Whether impaired IFN-α signaling results in diminished tumor surveillance capacity in normal skin remains to be investigated.

We have demonstrated the utility of dnSTAT2 expression for permanently suppressing IFN-α signaling in a skin cell-based model system. Current and future studies using the dnSTAT2 protein expressed in HaCaT cells, an immortalized nontumorigenic skin cell line that serves as a model for skin premalignancy (34), and transgenic mice expressing dnSTAT2 under the control of the skin-selective cytokeratin 5 promoter, along with STAT2 knockout mice, are aimed at directly testing our hypothesis that reduced IFN responsiveness can lead to skin malignancy. In addition, other investigators have generated a fusion protein between IRF9 and the COOH-terminal transcription activation domain of STAT2, which can constitutively activate ISGF-3 target genes and produce a type I IFN response in human fibrosarcoma cells (35). We plan to further test our hypothesis by expressing this fusion protein in tumorigenic skin SCC cells and in mouse skin to determine whether constitutive activation of type I IFN signaling has a tumor-suppressive or -preventive effect. This mechanistic study, together with our earlier report in skin IEN (13), supports STAT2 as an important molecular target for skin cancer chemoprevention drug development.