Interferon-resistant Human Melanoma Cells Are Deficient in ISGF3 Components, STAT1, STAT2, and p48-ISGF3γ*

The mechanism of IFN resistance was examined in three long-term cell lines, SK-MEL-28, SK-MEL-3, and MM96, exhibiting significant variation in responsiveness to the antiproliferative and antiviral effects of type I IFNs. The JAK-STAT components involved in IFN signal transduction were analyzed in detail. After exposure to IFN, activation of the IFN type I receptor-linked tyrosine kinases, JAK-1 and TYK-2, was detected at similar levels in both IFN-sensitive and IFN-resistant cell types, indicating that IFN resistance did not result from a deficiency in signaling at the level of receptor-associated kinase activation. However, analysis of ISGF3 transcription factor components, STAT1, STAT2, and p48-ISGF3γ, revealed that their expression and activation correlated with cellular IFN responsiveness. The analysis was extended to also include IFN-sensitive primary melanocytes, three additional IFN-resistant melanoma cell lines, and seven cell cultures recently established from melanoma patient biopsies. It was consistently observed that the most marked difference in ISGF3 was a lack of STAT1 in the resistant versus the sensitive cells. Transfection of the IFN-resistant MM96 cell line to express increased levels of STAT1 protein partially restored IFN responsiveness in an antiviral assay. We conclude that a defect in the level of STAT1 and possibly all three ISGF3 components in IFN-resistant human melanoma cells may be a general phenomenon responsible for reduced cellular responsiveness of melanomas to IFNs.

The interferons (IFNs) comprise a family of multifunctional polypeptides, recognized for their antiviral, antiproliferative, and immunoregulatory functions (reviewed in Refs. 1 and 2). Type I IFNs, IFN-␣ and -␤, exert their biological actions by binding to high-affinity cell-surface receptors that stimulate phosphorylation of tyrosine residues on type I receptor components (3,4) and on the receptor-associated tyrosine kinases, TYK-2 and JAK-1 (5-7). Following tyrosine phosphorylation, the activated tyrosine kinases then induce the formation and activation by tyrosine phosphorylation of the latent cytoplasmic transcription factor IFN-stimulated gene factor 3␣ (ISGF3␣) (8). The ISGF3␣ transcription factor is a complex of STAT molecules, including STAT1 and STAT2. STAT1 exists in two forms of 91 kDa (STAT1␣) and 84 kDa (STAT1␤), and STAT2 is a 113-kDa protein (reviewed in Refs. 9 and 10). During the classical type I IFN response, the tyrosine-phosphorylated STAT1 and STAT2 proteins form heterodimers (11), which complex with p48-ISGF3␥, a DNA-binding protein. The ISGF3 complex then migrates to the nucleus and binds to the ISRE to activate transcription of IFN-regulated genes (reviewed in Ref. 12).
The effect of IFN in the treatment of advanced malignant melanoma has been demonstrated in several clinical trials with an overall response rate of approximately 22% (reviewed in Ref. 13). In addition, analysis of results from studies using tumor colony-forming assays in vitro suggests that approximately 50% of the tumors contain cells nonresponsive to the antiproliferative effects of IFNs (14,15). The reasons for the differences in responsiveness of human melanoma cells to IFN in vitro and in vivo remain unclear. However, many studies have reported various defects in the IFN system as being responsible for the different sensitivities to type I IFNs in cell lines established from other tumor types. Such defects include: i) IFN-␣/-␤ gene deletion in acute leukemia cell lines (16) and malignant T cells (17); ii) alteration or down-regulation of IFN-␣ receptor gene expression in hairy cell leukemia (18) and lymphoblastoid cells (19); iii) interference with the induction of the expression of IFN-stimulated genes in B lymphoid cell lines (20) and Burkitt's lymphoma cells (21); iv) defects in the activation of transcription factors in Daudi cells (22), primary leukemia cells (23), and in many other cancer cell types (24). In addition, melanoma cell lines with a wide variation in their responsiveness to the antiproliferative (25) and antiviral (26) activities of IFNs have been described. We have previously demonstrated that the difference in responsiveness among these melanoma cell lines is related neither to the receptor binding affinity nor to the numbers of IFN receptors on the cell surface (27). In addition, the cDNA sequence encoding the type I receptor subunit (28) isolated from IFN-resistant melanoma cell lines contained no significant difference from the prototypic Daudi cDNA sequence (29). Thus, the cellular differences in responsiveness to IFN of the melanoma cell lines could not be at the level of expression of the receptor, and the defect was likely to be the result of events downstream from the receptor (28). We have previously reported that the IFN responsiveness of human melanoma cells could be correlated with the levels of IFN-induced tyrosine-phosphorylated cellular proteins (30). In this study, the responsiveness of melanoma cell lines to IFN was further investigated by analyzing the level of IFN-␣ 2stimulated tyrosine phosphorylation, the expression and activation of components in the IFN signal transduction pathway, including the type I receptor-associated tyrosine kinases, JAK-1 and TYK-2, and the components of ISGF3. We present evidence that the loss of sensitivity to IFN in melanoma cells arises from a deficiency in the levels of the components of ISGF3, particularly STAT1 and a resulting deficiency in ISRE binding activity.
Immunoblotting-Cell lysates were prepared and subjected to Western blots. The blot was incubated with the relevant primary antibody at a dilution according to the manufacturer's instruction. After stringent washing, the blot was incubated with the relevant secondary antibody and developed using an enhanced chemiluminescence kit (Boehringer Mannheim).
Densitometric Analysis-An HP Scan Jet IIcx/T (Hewlett Packard) densitometer calibrated using Kodak optical density (0.0 -2.0 units) standard test strips was used to scan autoradiographs exposed to the immunoblot chemiluminescence signal. The MCID-M4 (version 3, Rev 1.3) microcomputer optical imaging system (Imaging Research Inc., St. Catherines, Ontario) was used for quantitation. The auto-contouring function was applied to quantitate the relative densities of bands. The optical density value for each band obtained was background subtracted before comparisons of the relative ratios of optical densities were determined.
Immunoprecipitation-Total cell lysates were subjected to immunoprecipitation with 1-5 g of either polyclonal or monoclonal antibodies in 500 l of 1 ϫ IP (immunoprecipitation) buffer (150 mM NaCl, 10 mM Tris-HCl, pH 7.4, 1 mM EDTA, 1 mM EGTA, 1% Triton X-100, 0.5% Nonidet P-40, 1 mM sodium orthovanadate, 1 mM phenylmethylsulfonyl fluoride, 10 g/ml aprotinin, and 10 g/ml leupeptin). If a monoclonal antibody was used, 50 l of 50% protein G-Sepharose suspension (Pharmacia Biotech Inc.) was added, and the mixture was incubated overnight at 4°C with constant agitation. When polyclonal antibody was used, 50 l of 50% protein A-Sepharose suspension (Pharmacia) was used instead of protein G-Sepharose. The immunoprecipitated complexes were washed three times with 1 ϫ IP buffer and the proteins were eluted by the addition of 60 l of 2 ϫ SDS-PAGE sample buffer followed by boiling for 5 min. Sepharose beads were pelleted by centrifugation in a microfuge for 5 min, and the supernatant containing proteins was subjected to SDS-PAGE.
RNA Preparation and Northern Blotting-Cytoplasmic RNA was isolated by the procedure of Gough (35) and subjected to Northern blot analysis (36). The blot was probed with an ␣-32 P-labeled full-length STAT1␣/␤ cDNA (a gift from Dr. Chris Schindler, Dept. of Medicine, College of Physicians and Surgeons, Columbia University, NY) prepared using a Gigaprime random priming kit (Bresatec, Adelaide, Australia). The blot was stripped of probe by incubating the membrane in 95°C distilled water for 30 min and subsequently reprobed with an ␣-32 P-labeled PstI fragment of a rat GAPDH cDNA as described previously (37).
Construction and Transfection of STAT1 Expression Vector-An expression plasmid for STAT1 was constructed by inserting the fulllength STAT1 cDNA into mammalian expression vector, pBPSTR1 (39), containing a puromycin resistance gene (a gift from Dr. Steve Reeves, Massachusetts General Hospital, Molecular Neuro-Oncology). Transfection of BING cells (40), an amphotropic counterpart to the BOSC23 viral packaging cell line, was performed by calcium phosphate precipitation as recommended (41). BOSC23 cells were grown in the presence of doxycycline (Sigma) during packaging of virus. MM96 cells that had been plated at a density of 5 ϫ 10 6 cells the day before transfection were infected with filtered medium containing virus in the presence of 4 g/ml polybrene (Sigma). Cells were allowed to grow for 48 h before addition of puromycin (2.5 mg/ml; Sigma) for selection of resistant clones. The expression levels of STAT1 in clones were determined by immunoblotting.
Induction of Tyrosine Phosphorylation of Tyrosine Kinases, TYK-2 and JAK-1, by IFN-␣ 2 in Melanoma Cell Lines-A detailed analysis of the activation of the signaling components of the IFN signal transduction pathway in melanoma cell lines with different responses to IFN was undertaken. Antibodies to JAK-1 and TYK-2 were used to immunoprecipitate proteins from total cell lysates prepared from the melanoma cell lines SK-MEL-28, SK-MEL-3, and MM96 following treatment with IFN-␣ 2 . All three cell lines exhibited an increase in the tyrosine phosphorylation of tyrosine kinases, JAK-1 (130 kDa) and TYK-2 (135 kDa) after a 30-min IFN-␣ 2 treatment (Fig. 1, A and C, respectively), whereas tyrosine-phosphorylated JAK-1 or TYK-2 were not observed in unstimulated cells. When the immunoblots were stripped and analyzed for the level of expression of the protein kinases, JAK-1 and TYK-2, both the untreated and 30-min IFN-␣ 2 -treated samples prepared from each cell line were found to contain similar levels of protein ( Fig. 1, B and D). These results indicate that after binding of IFN-␣ 2 to the type I IFN receptor, the activation by tyrosine phosphorylation of the tyrosine kinases, JAK-1 and TYK-2, was induced in all three melanoma cell lines.
Reduced Expression and Activation of the ISGF3 Component STAT1 in IFN-resistant Melanoma Cell Lines Relative to IFNsensitive Cell Lines-The level of STAT1 protein expressed in each of the three melanoma cell lines was examined (Fig. 2B). Total cell lysates were prepared from equal numbers of SK-MEL-28, SK-MEL-3, and MM96 cells and analyzed by immunoblotting with antisera to STAT1. The IFN-resistant cell lines SK-MEL-3 and MM96 contained markedly less STAT1 protein compared with the level detected in the IFN-sensitive cell line, SK-MEL-28. A comparatively high level of STAT1, similar to that found in SK-MEL-28, was detected in another IFN-sensitive melanoma cell line, Hs294T (data not shown). Thus, SK-MEL-28 is not an atypical cell line with respect to STAT1, despite it being the most IFN-responsive melanoma cell line that we have tested. Northern blotting analysis of STAT1 transcript levels (Fig. 3) also revealed decreased STAT1 mRNA expression in the IFN-resistant melanoma cell lines SK-MEL-3 and MM96. These results are consistent with a correlation between the IFN resistance of cells and a deficiency in the levels of STAT1.
Relative Levels of STAT1 Present in Other IFN-resistant Melanoma Cell Lines Compared with IFN-sensitive Cell Lines-Since the IFN-resistant cell lines SK-MEL-3 and MM96 were shown to be deficient in STAT1, the levels of STAT1 present in several other melanoma cell lines that had been previously characterized to be IFN-resistant (25,26) were also analyzed by immunoblotting with antiserum to STAT1. All three of the additional IFN-resistant cell lines examined, HT-144, Li-Br, and SK-MEL-1, had a reduced level of STAT1 protein compared with the level of STAT1 in the IFN-sensitive cells (Fig.  4A). Similar levels of the two different housekeeping control proteins, HSP90 and GAPDH (Fig. 4B), were detected by immunoblotting equivalent amounts of the cell samples indicating that similar amounts of total cellular proteins were loaded for each of these cell types. Thus, the reduced levels of STAT1 in IFN-resistant cells were not simply due to variation in sample loadings but reflected differences in cellular STAT1 expression. The relative levels of STAT1 were determined by comparing the ratio of densitometric values obtained for STAT1 to the densitometric values obtained for GAPDH (Table I).
Relative Levels of STAT2 and p48-ISGF3␥ in Melanoma Cell Lines-Cell lysates prepared from equal numbers of SK-MEL-28, SK-MEL-3, and MM96 cells were analyzed by immunoblotting to compare the levels of STAT2 (Fig. 2A). As with STAT1, the IFN-sensitive SK-MEL-28 cells contained the highest levels of STAT2; the less IFN-responsive SK-MEL-3 cells contained the intermediate levels and the IFN-resistant MM96 cells contained the lowest levels of STAT2. The difference in the level of STAT2 between SK-MEL-28 and SK-MEL-3 was not as significant as the difference found in the levels of STAT1. Whole cell RNA was isolated and 20 g was electrophoresed on a 1% agarose gel. RNA was then transferred onto zeta-probe membrane for hybridization with a ␣-32 P-labeled STAT1␣/␤ cDNA probe. The membrane was washed and autoradiographed. Subsequently, the blot was stripped and reprobed with a ␣-32 P-labeled GAPDH probe.
Densitometric analysis of immunoblot signals indicated that the STAT2 level of SK-MEL-28 was only 2 times higher than that of SK-MEL-3. MM96 had a much lower expression level of STAT2, some 4 times lower than that observed for SK-MEL-28. Examination of the levels of p48-ISGF3␥ expression (Fig. 2C) revealed that SK-MEL-28 and SK-MEL-3 have relatively similar levels of p48-ISGF3␥. The level of p48-ISGF3␥ in MM96 is approximately 2 times lower than SK-MEL-28 and SK-MEL-3 as determined by densitometric analysis. Thus, the IFN-resistant cell lines have mild deficiencies of STAT2 and p48-ISGF3␥ compared with the IFN-sensitive cell line, SK-MEL-28.

IFN-resistant Melanoma Cells Show a Greatly Reduced IFNactivated ISGF3-ISG15 Interaction
Detected by EMSA-To determine whether the deficiency of ISGF3 components particularly STAT1 in IFN-resistant cells affected the ability of IFN to activate ISGF3 binding to the ISRE, EMSA was performed using the standard ISGF3-ISG15 response element interaction (8). Cell lysates were prepared from cultures of SK-MEL-28, SK-MEL-3, and MM96 cells incubated with IFN-␣ 2 for 30 min, and then samples were subjected to EMSA. Lysates prepared from SK-MEL-3 and MM96 cells were found to have lower levels of activated ISGF3 complex in comparison to that prepared from SK-MEL-28 cells (Fig. 5). For some of the experiments, where indicated, cells were IFN-␥-primed for 16 h prior to the stimulation with IFN-␣ 2 because IFN-␥-priming has been shown to result in the up-regulation of cellular ISGF3 levels (42). A higher amount of ISGF3 was detected in IFN-␥primed SK-MEL-28 cells when compared with non-IFN-␥primed cells (Fig. 5). The low levels of activated ISGF3 in the IFN-resistant cell lines, SK-MEL-3 and MM96, could only be detected by EMSA when lysates were prepared from IFN-␥primed cells and when there were much greater numbers of cells (SK-MEL-3, 2ϫ; MM96, 4ϫ) relative to SK-MEL-28.
Deficiency in STAT1 Is Common in Melanoma Cells Recently Established from Patient Biopsies-The levels of STAT1 in cultured melanoma cells established from patient biopsies were also studied to determine whether the deficiency in ISGF3 components, in particular STAT1, was a common phenomenon. All of the melanoma cells established from patient biopsies were extremely unresponsive to the antiproliferative effect of IFN-␤ (IC 50 of Ͼ5000 IU/ml IFN-␤) (data not shown). Cultures of primary human melanocytes that were responsive to IFN were included as a positive control (IC 50 of approximately 20 IU/ml). These melanocytes contained levels of STAT1 protein similar to that of SK-MEL-28 (Fig. 6A). Total cell lysates prepared from recently cultured patient biopsies were examined on the basis of equivalent amounts of total cellular protein instead of equal numbers of cells because a great variation in the sizes of cells was found. Immunoblotting with GAPDH antiserum (Fig. 6B) showed almost identical levels of GAPDH present in each sample. Reduced levels of STAT1 were observed in the IFN-resistant melanoma cells compared with SK-MEL-28 (Fig. 6A). The relative levels of STAT1 were determined by comparing the ratio of densitometric values obtained for STAT1 to the densitometric values obtained for GAPDH (Table II). As shown in Table II, primary human melanocytes have a significantly higher level of STAT1 than was detected in the IFN-resistant melanoma cells from patients. As STAT1 protein in melanoma cells established from patients 3 and 5 could not be detected (Fig. 6A), samples from patients 3 and 5 containing greater amounts of protein were analyzed. For comparison, an equivalent sample was prepared from the moder-

FIG. 4. Levels of STAT1 in IFN-resistant melanoma cell lines.
Panel A, total cell lysates of 1 ϫ 10 5 cells prepared in 1 ϫ PAGE sample buffer were subjected to immunoblotting using anti-STAT1 as the primary antibody and sheep anti-mouse horseradish peroxidase-conjugated antibody. STAT1 proteins were detected using the chemiluminescence detection system. Panel B, samples of 1 ϫ 10 5 cells were subjected to SDS-PAGE analysis. Proteins were transferred to nylon membrane and immunoblotted with AC88 as the primary antibody and sheep anti-mouse horseradish peroxidase-conjugated antibody, and blots were developed using the chemiluminescence detection system. The blot was stripped and reprobed with GAPDH antiserum, and binding was detected using sheep anti-rabbit horseradish peroxidase-conjugated antibody.  Fig. 4, A and B. b Ratios of densitometric values of STAT1/densitometric values of GAPDH were standardized for comparison relative to SK-MEL-28 arbitrarily assigned a value of 1.00. ately IFN-sensitive (IC 50 of approximately 500 IU/ml) cells established from patient 7. The relative expression of STAT1 in these different melanoma cell samples (Fig. 6C) was analyzed by determining the comparative ratios of the densitometric values obtained for STAT1 to the densitometric values obtained for GAPDH ( Fig. 6E and Table III). A low level of STAT1 protein was detected in the sample from patient 3 but STAT1 could only be observed in the sample of patient 5 when the blot was developed and exposed for a longer time period (10 min instead of 1 min) (Fig. 6D). The sample from the moderately IFN-sensitive cells of patient 7 contained much more STAT1 protein than the IFN-resistant cells, consistent with the observation that cellular responsiveness to IFN correlates with cellular STAT1 levels.
Transfection of STAT1 Gene Improves Responsiveness to IFN-␤-Since STAT1 levels are low in IFN-resistant melanoma cells such as MM96, we examined whether overexpression of STAT1 would improve the responsiveness of MM96 to the antiviral effects of IFN-␤. The antiviral effects of IFN-␤ were examined in IFN-sensitive SK-MEL-28, IFN-resistant MM96 and STAT1 transfected MM96 cells (Fig. 7A). As expected, SK-MEL-28 is sensitive to treatment with IFN-␤ (CPE 50 of 5 IU/ml), which failed to protect the parental MM96 cells from the viral mediated cytopathic effects (CPE 50 of 1600 IU/ml). However, with the STAT1-transfected MM96 cells, the effect of IFN-␤ is significantly increased by approximately 10 times (CPE 50 of 150 IU/ml). The expression of STAT1 in the MM96 transfected cells was confirmed by immunoblotting (Fig. 7B) and found to be about 3 times higher than the level of STAT1 detected in the parental MM96 cells.

DISCUSSION
Several studies using protein tyrosine kinase inhibitors have revealed the importance of tyrosine phosphorylation in IFN signal transduction (43,44). In a recent study, we demonstrated that the responsiveness of melanoma cell lines to IFN correlated with the IFN-␣ 2 -induced changes in tyrosine phosphorylation (29). In this report, we have further investigated the nature of the reduction in type I IFN-induced tyrosine phosphorylation in IFN-resistant melanoma cell lines. We have also extended our investigation to include a comparison of the relative levels of STAT1 expression found in primary human melanocytes and melanoma cells recently established from patient biopsies.
Analysis of the receptor-associated tyrosine kinases JAK-1 and TYK-2 ( Fig. 1), both of which are involved in the type I IFN-signaling pathway, revealed that there were no significant differences in their activation between the three melanoma cell lines tested following IFN-␣ 2 stimulation. In addition, our results suggested that both TYK-2 and JAK-1 were constitutively expressed in all three cell lines. Thus, we conclude that the defect in IFN-resistant melanoma cells is neither at the level of the expression of the tyrosine kinases nor their activation. However, examination of the presence and activation of the components of the ISGF3 transcription factor, STAT1, STAT2, and p48-ISGF3␥ revealed significant differences between IFNsensitive and -resistant melanoma cell lines. In particular, a marked reduction in the levels of STAT1 was consistently observed in all of the IFN-resistant cell lines (Fig. 4A) and in low passage cultures of melanomas recently established from cancer patient biopsies resistant to the antiproliferative effects of type I IFNs. Additionally, we also observed mild deficiencies in the levels of the other ISGF3 components, STAT2 and p48-ISGF3␥ (Fig. 2) in the IFN-resistant cell lines. An analysis of the ISGF3 interaction with the ISG15 ISRE sequence by EMSA (Fig. 5) revealed significantly reduced levels of ISGF3 present in SK-MEL-3 and particularly in MM96 cells, the most IFNresistant cell type. The observation that IFN-resistant cell lines such as MM96 contain lower levels of STAT1, STAT2 and p48-ISGF3␥ (all three components that constitute ISGF3 activity) is consistent with the observed decrease in ISGF3 DNA binding activity and the lack of IFN responsiveness. Given the marked reduction in STAT1 protein levels in IFN-resistant cells, it follows that this form of transcription factor-DNA enhancer element interaction in the cells would be severely affected. Transfection of IFN-resistant MM96 cells, which have a marked deficiency of STAT1, with STAT1 cDNA resulted in cells having an increased responsiveness to IFN as judged by altered sensitivity to viral-mediated cytopathic effect (Fig. 7). Full restoration of IFN responsiveness equivalent to SK-MEL-28 may require further increasing levels of STAT1 or restoration of levels of STAT2 and p48-ISGF3␥. The molecular basis for the reduction in cellular levels of ISGF3 components in IFN-resistant cells remains to be determined. However, it is possible that it is caused either by the failure of the IFNresistant cells to constitutively express ISGF3 components or a loss of normal regulation of their expression.
IFN-stimulated gene (ISG) induction via activation of transcription at the ISRE is what ultimately dictates the cellular outcome in terms of IFN responses. Previous analysis of the IFN-induced expression of 2Ј,5Ј-oligoadenylate synthetase (OAS), ds-RNA-induced protein kinase (PKR) and MX antigen showed no significant differences between IFN-resistant and -sensitive melanoma cell lines (30). Thus, our observations with melanoma cell lines clearly demonstrate a determining role for ISGF3 in the establishment of cellular responses to type I IFNs, albeit that these responses must depend on ISG expres-sion other than that of OAS, PKR and MX. Our results provide an explanation for the previous observations obtained using antiproliferative and antiviral biological assays, which revealed widely differing responsiveness of melanoma cell lines to IFN (25,26). Also, the observed lack of ISGF3 in IFNresistant melanoma cell lines helps to explain the reduced level of tyrosine-phosphorylated proteins detected in the IFN-resistant cells reported previously (30).
In addition to our observations in melanoma cells, a similar type of ISGF3 deficiency has been described in other cancer cell types (22,24,45,46), and STAT1 deficiency has been reported (ibid.) to occur in IFN-resistant breast cancer cell lines (47). These observations lead to the suggestion that ISGF3 deficiency may be a more general mechanism underlying the development of IFN resistance in a wide variety of cancers. Thus, a diagnostic assay based on determining deficiency in ISGF3 components may serve as a useful clinical marker of IFN resistance in melanomas and other cancers. Finally, our results may have wider implications with respect to the role of the IFNs in carcinogenesis given that many cancers are resistant to IFN therapy (13) and that IFNs have a role as a tumor suppressor (48).