Alternative activation of STAT1 and STAT3 in response to interferon-gamma.

Interferon-gamma (IFNgamma) is a pluripotent cytokine whose major biological effects are mediated through a pathway in which STAT1 is the predominant and essential transcription factor. STAT3 can also be activated weakly by IFNgamma, but the mechanism of activation and function of STAT3 as a part of the interferon response are not known. Here we show that STAT3 activation is much stronger and more prolonged in STAT1-null mouse embryo fibroblasts than in wild-type cells. In response to IFNgamma, SRC-family kinases are required to activate STAT3 (but not STAT1) through tyrosine phosphorylation, whereas the receptor-bound kinases JAK1 and JAK2 are required to activate both STATs. Tyrosine 419 of the IFNgamma receptor subunit 1 (IFNGR1) is required to activate both STATs, suggesting that STAT1 and STAT3 compete with each other for the same receptor phosphotyrosine motif. Activated STAT3 can replace STAT1 in STAT1-null cells to drive the transcription of certain genes, for example, socs-3 and c/ebpdelta, which have gamma-activated sequence motifs in their promoters. Work from Ian Kerr's laboratory reveals that the gp130-linked interleukin-6 receptor, which usually activates STAT3 predominantly, activates STAT1 efficiently when STAT3 is absent. Because STAT1 and STAT3 have opposing biological effects (STAT3 is an oncogene, and STAT1 is a tumor suppressor), the reciprocal activation of these two transcription factors in response to IFNgamma or interleukin-6 suggests that their relative abundance, which may vary substantially in different normal cell types, under different conditions or in tumors is likely to have a major impact on how cells behave in response to different cytokines.

Interferon-␥ (IFN␥) is a pluripotent cytokine whose major biological effects are mediated through a pathway in which STAT1 is the predominant and essential transcription factor. STAT3 can also be activated weakly by IFN␥, but the mechanism of activation and function of STAT3 as a part of the interferon response are not known. Here we show that STAT3 activation is much stronger and more prolonged in STAT1-null mouse embryo fibroblasts than in wild-type cells. In response to IFN␥, SRC-family kinases are required to activate STAT3 (but not STAT1) through tyrosine phosphorylation, whereas the receptor-bound kinases JAK1 and JAK2 are required to activate both STATs. Tyrosine 419 of the IFN␥ receptor subunit 1 (IFNGR1) is required to activate both STATs, suggesting that STAT1 and STAT3 compete with each other for the same receptor phosphotyrosine motif. Activated STAT3 can replace STAT1 in STAT1-null cells to drive the transcription of certain genes, for example, socs-3 and c/ebp␦, which have ␥-activated sequence motifs in their promoters. Work from Ian Kerr's laboratory reveals that the gp130-linked interleukin-6 receptor, which usually activates STAT3 predominantly, activates STAT1 efficiently when STAT3 is absent (Costa-Pereira, A. P., Tininini, S., Strobl, B., Alonzi, T., Schlaak, J. F., Is'harc, H., Gesualdo, I., Newman, S. J., Kerr . Because STAT1 and STAT3 have opposing biological effects (STAT3 is an oncogene, and STAT1 is a tumor suppressor), the reciprocal activation of these two transcription factors in response to IFN␥ or interleukin-6 suggests that their relative abundance, which may vary substantially in different normal cell types, under different conditions or in tumors is likely to have a major impact on how cells behave in response to different cytokines.
The major biological responses to interferon-␥ (IFN␥), 1 including inhibition of proliferation, antiviral responses, immune modulation, and tumor suppression, are mediated through the well known JAK-STAT pathway (1)(2)(3). Biochemical and genetic studies show that STAT1 plays a critical role in IFN␥-dependent signaling, and cells lacking STAT1 are defective in IFNmediated antiviral activity. Furthermore, overexpression of a constitutively active STAT1 mutant protein can enhance IFN␥induced antiproliferative activity (4). The phenotypes of STAT1-null mice also demonstrate that STAT1 is the major transcription factor in IFN␥-dependent signaling (5,6). However, recent work reveals that additional signals are required for the full range of responses to IFN␥ (7). First, expression of the c-myc gene was induced in STAT1-null cells but suppressed in wild-type cells, revealing an alternative pathway activated by IFN␥ (8). Second, microarray analysis showed that a number of genes are regulated by IFN␥ in STAT1-null primary bone marrow-derived macrophages or mouse embryo fibroblasts (MEFs) (9,10). Third, the physiological relevance of the STAT1-independent responses to IFN␥ has been demonstrated in vivo: IFN␥ induced the proliferation and survival of primary bone marrow cells from STAT1-null mice but not wild-type mice, and STAT1-null mice are more resistant to infection with murine cytomegalovirus or sindbis viruses than are mice lacking both type I and type II IFN receptors (10). In addition, Mycobacterium tuberculosis infection selectively inhibits the transcriptional responses to IFN␥ without inhibiting STAT1 function (11). More recently, IB kinases have been shown to be required for activation of a subset of IFN␥-stimulated genes, and this requirement is independent of NFB activation. Furthermore, STAT1 activation is not impaired in IB kinase ␣/␤-null MEFs in response to IFN␥ (12). STAT1-independent pathways are also important for responses to type I interferon. During viral infection IFN␣/␤ negatively regulates IFN␥ production through STAT1, and in the absence of STAT1, type I interferons promote IFN␥ production (13). In the central nervous system, STAT1 deficiency leads to earlier onset and more severe neurological disease with increased lethality in response to IFN␣. Thus, in the absence of STAT1, alternative signaling pathways mediate pathophysiological actions of IFN␣ in the living brain (14). Type I interferon can also inhibit interleukin-7 (IL-7)-mediated growth or survival of T and B lymphoid progenitors through a STAT1-independent pathway (15,16). It is likely that type I interferon activates at least two alternative pathways independently of STAT1; one is normally undetectable or inhibited when STAT1 is present (13,14), and the other is effective whether STAT1 is present or not (15,16).
Although the importance of the IFN␥-induced, STAT1-independent signaling is clear, the mechanism is not. Previous studies showed that phosphatidylinositol 3-kinase and its effector kinase AKT (17), mitogen-activated protein kinase (18,19), and protein kinase C␦ (20) are activated by IFN␥, but their activities are involved in the serine phosphorylation of STAT1, which in turn is important for its maximum transcriptional activation (21). STAT1 and STAT3 have similar structures, both are phosphorylated on tyrosine residues upon cytokine or growth factor stimulation, and both form dimers through the reciprocal SH2 domain/phosphotyrosine interactions, move to the nucleus, bind to ␥-activated sequence (GAS) elements, and activate transcription of many genes (2,22,23). STAT1 and STAT3 dimers bind selectively to very similar but not identical elements (24,25) and, thus, activate different but overlapping sets of genes, which is likely to account for their very different biological effects. In contrast to the tumor suppressor function of activated STAT1 (26 -28), STAT3 is an oncogene (29,30). In normal cells, STAT1 and STAT3 can be phosphorylated by the intrinsic kinase activity of receptors (31)(32)(33) or receptor-bound tyrosine kinases (2,34,35). Some oncogenic proteins with tyrosine kinase activity (v-SRC, v-ABL, and v-FES) can phosphorylate STAT1 and STAT3 (36 -38), and SRC and JAKs cooperate to mediate STAT3 activation (39 -41). Here, we report that, in the absence of STAT1, IFN␥ activates STAT3 strongly, SRCfamily kinases are required for the activation, and the activated STAT3 can drive the expression of some genes that normally respond to activated STAT1 in IFN␥-treated wildtype cells.
Retroviral Infection-Bosc packaging cells were transfected transiently with expression plasmids using LipofectAMINE (Invitrogen). Twenty-four hours after transfection, the supernatant medium was collected, filtered through sterile 0.2-m syringe filters, and used to infect the target cells in the presence of 5 g/ml Polybrene (Sigma). After 24 h, the cells were fed with fresh complete medium. After 48 h, the infected cells were trypsinized and seeded at a dilution of 1:10 to 1:100 in complete medium containing 2 g/ml puromycin.
Oligonucleotide Microarrays-STAT1-null MEFs were treated with IFN␥ for 2 or 4 h, and total RNA was prepared by using the TRIzol method (Invitrogen). cRNAs were prepared and hybridized to murine genome arrays (Affymetrix MG-U74Av2 chips) according to the manufacturer's instructions (Affymetrix, San Jose, CA). The stained arrays were read and analyzed by using an Affymetrix GeneChip scanner and the accompanying software.
Luciferase Assay-The murine socs-3 promoter constructs, clone 6 and clone 6M, were kindly provided by Dr. Shlomo Melmed (Cedars-Sinai Research Institute, UCLA School of Medicine, Los Angeles, CA). For transient transfections, 2 ϫ 10 5 cells were plated in 6-well plates, incubated for 24 h, and transfected by using the FuGENE 6 reagent (Roche Applied Science) with 0.5 g of clone 6 or clone 6M and 0.1 g of pSV-␤-galactosidase. After 24 h, transfected cells were treated with IFN␥ for 6 h, and luciferase activity was measured by using a luciferase assay kit (Promega). Transfection efficiencies were determined by measuring ␤-galactosidase activity. Results are shown for one of at least three independent experiments.

RESULTS
Phosphorylation of STAT3 in Response to IFN␥-STAT3 is phosphorylated transiently in response to IFN␥ in wild-type MEFs, but its phosphorylation is much stronger and more prolonged in STAT1-null cells (Fig. 1). When STAT1 was re-expressed in STAT1-null cells, the intensity of STAT3 phosphorylation was again attenuated (data not shown). Note the very strong induction of STAT1 expression in wild-type cells treated with IFN␥ for 6 or 16 h and the higher levels of STAT3 in STAT1-null cells.
Previous work has demonstrated that STAT3 is a substrate for SRC (38). To test for SRC activation in response to IFN␥, an in vitro kinase assay was performed with SRC immunoprecipitated from IFN␥-treated wild-type or STAT1-null MEFs ( Fig.  2A). IFN␥ treatment increased both the autophosphorylation of SRC and the SRC-dependent phosphorylation of the exogenous substrate enolase. The requirements for JAK1 and JAK2 were checked in the same in vitro kinase assay; SRC is not activated in response to IFN␥ in JAK1-or JAK2-null cells, although higher basal SRC activities were observed (Fig. 2B). To determine whether SRC mediates the activation of STAT3 in response to IFN␥, we performed Western analyses of cells pretreated with the SRC inhibitor SU6656 (44,45). IFN␥ induced the phosphorylation of STAT3 in both wild-type and STAT1null cells treated with the Me 2 SO solvent only, and this phosphorylation was strongly inhibited by SU6656 (Fig. 3A). We also tested SYF cells, which lack the three SRC-family members, SRC, FYN, and YES. Western analysis showed that, in response to IFN␥, STAT3 phosphorylation was abolished in SYF cells, whereas the activation of JAK2 and STAT1 was intact (Fig. 3B). When SRC was put back into SYF cells, the basal level of STAT3 phosphorylation increased, and treatment with IFN␥ caused a further increase (Fig. 3B). Cells with kinase-dead SRC behaved the same as cells infected with the vector only. Together, the data indicate that IFN␥ activates STAT3 through SRC in a manner that requires both JAK1 and JAK2.
Tyrosine 419 of Murine IFNGR1 Is Required for the Activation of Both STAT1 and STAT3 in Response to IFN␥-Previous work has shown that tyrosine 419 of murine IFNGR1 is required for STAT1 activation in response to IFN␥ (43). To determine whether the same site is also required for STAT3 activation, IFNGR1-null MEFs reconstituted with various receptor mutants were tested. Western analyses showed that STAT3 activation was observed only in cells expressing wildtype or mutant Tyr-419, in which residue 419 is the only tyrosine remaining in the cytoplasmic domain of IFNGR1 (Fig.  4). Mutation of Tyr-419 to phenylalanine abolished the activation of both STAT3 and STAT1. Tyr-419 is absolutely required for the activation of both STAT3 and STAT1 in response to IFN␥, and therefore, we conclude that these two STATs compete for this phosphorylated binding site on IFNGR1.
Identification of Genes Regulated by IFN␥ in STAT1-null MEFs-About 30 genes were found to be induced by IFN␥ in serum-starved STAT1-null MEFs (9). In our previous work, serum starvation was used to allow genes induced rapidly by growth factors to be observed. To analyze the response in the presence of serum, we performed another array experiment. Total RNA samples were prepared from STAT1-null MEFs

FIG. 2. Increase in SRC kinase activity upon IFN␥ treatment.
A, wildtype (WT) or STAT1-null MEFs were treated with IFN␥ for 10 or 20 min, and whole cell lysates were immunoprecipitated with anti-c-SRC followed by a kinase assay using [␥-32 P]ATP and enolase as substrates. Phosphorylation (P) of both enolase and SRC was analyzed by SDS-PAGE followed by autoradiography. B, wild-type, JAK1-null, or JAK2-null MEFs were treated with IFN␥ for 10 or 20 min followed by the kinase assay described above.

FIG. 1. STAT3 activation by IFN␥ is enhanced in STAT1-null cells.
Wildtype (WT) or STAT1-null MEFs were exposed to IFN␥ for the times indicated. Total cell extracts were prepared, and equal amounts of protein were analyzed by SDS-PAGE and the Western method for activated, phosphorylated STAT3 (pYSTAT3) and STAT1 (pYSTAT1). treated with IFN␥ for 2 or 4 h and analyzed by using murine U74Av2 arrays (Affymetrix), which contain sequences corresponding to about 6000 known genes. Eleven genes were in-duced within 2 h, and an additional six genes were induced after 4 h ( Table I). The induction of most of these genes was confirmed by Northern analysis (Table I). For the genes induced at 4 h, induction might be due to secondary rather than primary effects. Many of these genes had been identified previously as induced in serum-starved cells (9). Genes induced in both STAT1-null and wild-type MEFs and in which GAS elements have been identified (socs-3, c/ebp␦, mmp-1, metallothionein) (46 -49) are likely to be regulated alternatively by STAT1 or STAT3. For c-myc and c-jun, which are induced only in the FIG. 5. A, socs-3 mRNA expression is induced by IFN␥ in both primary STAT1-null and wild-type (WT) MEFs. The cells were treated with 1000 IU/ml murine IFN␥, and mRNA levels were analyzed by the Northern method. B, dominant negative STAT3 blocks socs-3 and c/ebp␦ gene expression in response to IFN␥. STAT3-null MEFs or stable pools of immortalized STAT1-null MEFs infected with retroviruses containing either empty vector (pLEGFP-N1) or dominant negative STAT3 in pLEGFP-N1 were treated with IFN␥. Total RNA was analyzed by the Northern method.

FIG. 3. SRC-family kinases are required for STAT3 activation in response to IFN␥.
A, wild-type (WT) or STAT1-null MEFs were pretreated with solvent (Me 2 SO (DMSO)) or SU6656 for 30 min before being stimulated with IFN␥ for 20 min. Total cell extracts were prepared, and equal amounts of protein were analyzed by SDS-PAGE and the Western method for activated, phosphorylated STAT3 (pYSTAT3) and STAT1 (pYSTAT1). B, SYF cells lacking SRC, FYN, and YES were infected with empty vector pBABEpuro3 or wild-type or a kinase-dead (KD) mutant of SRC in pBABEpuro3; stable pools were generated. Cells were exposed to IFN␥, total cell extracts were prepared, and equal amounts of protein were analyzed by SDS-PAGE and the Western method for activated, phosphorylated STAT3 (pYSTAT3), JAK2 (pY-JAK2), and STAT1 (pYSTAT1).
FIG. 4. Tyrosine 419 of murine IFNGR1 is required to activate both STAT1 and STAT3. IFNGR1-null MEFs were infected with retroviruses containing either empty vector (pBABEpuro3) or IFNGR1 in pBABEpuro3. Wild-type IFNGR1 and the mutants Y419F, 4XYF (Y for F at residues 285, 370, 419 and 441), and Tyr-419 (419Y) (Y for F at residues 285, 370, and 441) were used. Stable pools of infected cells were treated with IFN␥, total cell extracts were prepared, and aliquots containing equal amounts of protein were analyzed by SDS-PAGE and the Western method for phosphorylated STAT1 (pYSTAT1) and STAT3 (pYSTAT3).

STAT3 Activation by IFN␥
absence of serum in STAT1-null but not in wild-type cells (8), we conclude that STAT1 dimers bind to the GAS elements to regulate gene expression negatively. It is likely that, in the absence of STAT1, all of the induced genes in the Table I are regulated positively by STAT3, directly or indirectly.
Response of socs-3 to IFN␥ in STAT1-null Cells-Our experiments identified socs-3 as an IFN␥-induced gene both in wildtype cells and in primary macrophages and immortalized MEFs derived from STAT1-null mice (9, 10). The socs-3 gene is also induced in response to a variety of agents that activate STAT3, including interleukin 6 (IL-6), IL-10, leptin, leukemia inhibitory factor (LIF), IFNs, and bacterial lipopolysaccharide (50). The sequence of the murine socs-3 promoter has been characterized in response to LIF, and a STAT1/STAT3 binding site was identified to be essential. Mutation or deletion of this site completely abolished socs-3 promoter activity (46). This GAS element is conserved in mice, rats, and humans (51). Overexpression of wild-type STAT3 led to an increase of socs-3 mRNA levels after stimulation with LIF, whereas overexpression of a dominant negative mutant of STAT3 diminished the induction of socs-3 mRNA after stimulation with LIF (46).
In wild-type or STAT1-null primary MEFs treated with IFN␥, socs-3 mRNA increased rapidly (Fig. 5A). Consistent with previous results (52), ip-10 was induced only in wild-type and not in STAT1-null cells. The same experiment with immortalized MEFs gave the same result (data not shown). JAK1 and JAK2, required for STAT1 activation in response to IFN␥, are also both required for IFN␥-induced socs-3 gene expression (data not shown). In summary, the response of socs-3 to IFN␥ requires both JAK1 and JAK2 but not STAT1.
To test the role of STAT3 in the expression of the endogenous socs-3 gene in response to IFN␥, STAT1-null immortalized MEFs were infected with the dominant negative STAT3Y705F and treated with IFN␥ for 2 h. Because phosphorylation of tyrosine 705 is essential for STAT3 to function in response to external stimuli (2, 53), STAT3Y705F acts as a specific inhibitor for wild-type STAT3 (54,55). Northern analysis showed that, in response to IFN␥, socs-3 mRNA still increased in STAT3-null cells, and its induction was substantially attenuated in STAT1-null cells infected with dominant negative STAT3. In this experiment, we also analyzed the expression of c/ebp␦, which is regulated by STAT1 in response to IFN␥ and by STAT3 in response to IL-6-family cytokines (47). The induction of c/ebp␦ mRNA in response to IFN␥ in STAT1-null MEFs was almost completely lost when dominant negative STAT3 was present (Fig. 5B).
Our results showed that STAT3 activation in response to IFN␥ requires SRC (Fig. 3, A and B); the role of SRC in socs-3 gene expression was tested. Wild-type or STAT1-null MEFs were infected with a vector from which a dominant negative SRC protein was expressed, and the infected cells were treated with IFN␥. Dominant negative SRC suppressed socs-3 gene expression in STAT1-null cells but not in wild-type cells (Fig. 6).
A GAS element in the socs-3 promoter, essential for LIFinduced expression (46), was tested for its role in the IFN␥ response in STAT1-null cells. Clone 6 and clone 6M, which has mutations in the GAS site, were transfected transiently into wild-type or STAT1-null MEFs. The results, summarized in Fig. 7A, showed that the GAS element in the socs-3 promoter is FIG. 7. A, a GAS element in the socs-3 promoter mediates IFN␥ responses in both wild-type and STAT1-null MEFs. Clone 6 (wild-type (WT)) or clone 6M (mutant (MT)) constructs in which the murine socs-3 promoter drives luciferase expression were transfected transiently into cells. After 24 h the cells were stimulated with IFN␥, and luciferase (LUC) activities were measured with untreated cells as control. The mutant promoter contains the mutation 5Ј-TTCCAGGAA-3Ј to 5Ј-ATCGACGAT-3Ј. kb, kilobases. B, both STAT1 and STAT3 bind to a GAS element derived from the socs-3 promoter. Cells were untreated or treated with IFN␥ for 15 min, and nuclear extracts were prepared. Aliquots containing equal amounts (10 g) of each extract were analyzed by EMSA. A 32 Plabeled GAS derived from the socs-3 promoter (5Ј-75 CAGTTCCAG-GAATCGGGGGGC 55 -3Ј) was used as the probe. Ab, antibody.
FIG. 6. Dominant negative SRC suppresses IFN␥-induced socs-3 expression in STAT1-null but not wild-type (WT) MEFs. Cells were infected with retroviruses containing either empty vector (pBABEpuro3) or dominant negative (DN) SRC in pBABEpuro3, and stable pools of infected cells were treated with IFN␥. Total RNA was analyzed by the Northern procedure.

STAT3 Activation by IFN␥
required for all IFN␥-induced activation in the presence or absence of STAT1. To determine which proteins bind to the GAS element, cells were treated with IFN␥ for 30 min, and nuclear extracts were analyzed by EMSA (Fig. 7B). Specific binding was observed, but only in extracts from treated cells. Three complexes representing STAT1 homodimers, STAT1/ STAT3 heterodimers, and STAT3 homodimers were present in extracts of wild-type cells, with the complex involving STAT1 homodimers most intense. Extracts from IFN␥-treated STAT1null MEFs formed only the complex involving STAT3 homodimers. These results show that both STAT1 and STAT3 dimers bind to the same GAS element in the socs-3 promoter.

DISCUSSION
STAT3 Activation-Upon IFN␥ stimulation, STAT3 is phosphorylated in some wild-type cell types. It was reported that IFN␥ significantly increased sPLA (2)-IIA mRNA expression and secretion of the protein in human arterial smooth muscle cells in culture and that STAT3 is involved in this process (53). But the regulation of sPLA (2)-IIA by IFN␥ is cell type-specific since IFN␥ can not up-regulate sPLA (2)-IIA in HepG2 cells (56). Here we report that in cells lacking STAT1, STAT3 activation contributes to the transcription of several genes, i.e. socs-3 and c/ebp␦. These two genes are induced by STAT1 in wild-type cells. In wild-type MEFs, STAT1 is still the major driving force for transcription since EMSAs showed that in response to IFN␥, STAT1 homodimer is the predominant protein binding to the socs-3 GAS element, whereas STAT3 bound with much less affinity (Fig. 7B), consistent with previous findings that showed that STAT1 and -3 have different affinities for very similar but not identical GAS sites (24). The increase of socs-3 mRNA levels (Fig. 5A) is higher than that of the promoter activities (Fig. 7A) in STAT1-null cells after IFN␥ stimulation. The difference may result from different experimental approaches and usage of different cell lines; Northern analyses to examine socs-3 mRNA level increases can be done in both primary and immortalized MEFs, but we could only do transient transfection experiments to determine the promoter activities in immortalized but not primary MEFs. Both STAT1 and STAT3 activation by IFN␥ depends on the same residue on IFNGR1 (Fig. 4), revealing that STAT1 and STAT3 compete for the same binding site. Previous studies demonstrate that the SH2 domain of STAT1 has a much higher affinity for the phosphotyrosine 419 motif of IFNGR1 than the SH2 domain of STAT3 (57)(58)(59). We conclude that STAT1 is preferentially activated in wild-type cells even though STAT1 is less abundant than STAT3; however, changes in the relative abundance of STATs 1 and 3 will lead to changes in their relative degrees of activation.
SRC Activation-SRCs are non-receptor tyrosine kinases. Expression of v-SRC causes transformation of a variety of cell types (38). SRC is often observed to be either overexpressed or constitutively active in a large percentage of some tumor types, i.e. colon and breast (60,61). SRC is also activated by extracellular stimuli such as growth factors and cytokines. In NCI-H292 human alveoepithelial carcinoma cells, IFN␥ activates c-SRC, resulting in the activation of STAT1 and initiation of icam-1 gene expression and monocyte adhesion (62). Here we also show that SRC is activated by IFN␥ and, furthermore, provide evidence that it is involved in the activation of STAT3 but not STAT1. This apparent difference in the role of SRC with respect to STAT1 activation may be due to differences in the cell lines (Chang et al. (62) used human epithelial cells, and we studied MEFs) or in how the experiment was performed (Chang et al. (62) used the chemical inhibitor PP2, whereas we used SU6656). Furthermore, we confirmed the inhibitory effects of SU6656 by conducting experiments in SYF cells. The mechanism by which SRC is activated by IFN␥ is not known. However, it seems likely that it is recruited to IFN␥ receptor complex by binding to a phosphotyrosine residue, is activated by JAK1 and JAK2, and then phosphorylates bound STAT3.
The Balance between STAT1 and STAT3-Although STAT1 and STAT3 are very similar proteins often activated by the same stimuli, they have very different effects on cell growth and survival. For example, activated STAT3 antagonizes the proapoptotic effects of activated STAT1 in fibroblasts (63). Furthermore, IFN␥ can function as a growth factor for transformed/immortalized or primary bone marrow cells that lack STAT1 (10). It is possible that STAT3, activated in these cells, mediates the proliferative effect of IFN␥, especially when the growth-inhibitory protein STAT1 is missing. IFN␥ is also able to activate the mitogen-activated protein kinase and phosphatidylinositol 3-kinase pathways, which are linked to cell growth. When the inhibitory protein STAT1 is lacking, these additional pathways may come into play and activate components needed for cell growth. This mechanism is consistent with previous observations that, in STAT1-null MEFs, IFN␥ induces the expression of a group of genes that are proliferation-related (9). For type I interferons, STAT1 inhibits the IFN␣/␤-induced production of IFN␥ during viral infection (13) and also inhibits the pathological effects caused by IFN␣ in living brains (14). However, in the absence of STAT1, alternative pathways take over and promote the type I interferon effects, including induction of IFN␥ after viral infection and the pathological effects of IFN␣. A major alternative pathway is likely to be the activation of STAT3, which may, for example, be stronger and more prolonged in the brains of STAT1-deficient mice overexpressing IFN␣ (14).
Here we also show that, in the absence of STAT1, the activation of STAT3 by IFN␥ is much stronger and more prolonged and that STAT3 dimers contribute to the expression of at least several genes. In related work Costa-Pereira et al. (64) show that, in MEFs lacking STAT3, IL-6 mediates an IFN␥-like response including prolonged activation of STAT1, the expression of multiple IFN␥-stimulated genes, and an antiviral state. Additionally, Herrero et al. (65) report that pretreatment with IFN␥ switched the balance of IL-10 STAT activation from STAT3 to STAT1 (65). We see that STAT-activating cytokines can activate STAT1 and STAT3 alternatively and which STAT is activated depends on their relative affinities for the receptor (58, 59) as well as their relative abundance. It is well known that the expression of STAT1 or STAT3 is often deregulated in tumor cell lines or in samples from patients (66,67). In such patients with abnormal expression levels of STAT1 or STAT3, the responses to IFN␥ or IL-6 may even be the opposite of the normal responses.