Constitutive Stat3 activity up-regulates VEGF expression and ... - Nature

13 downloads 62 Views 339KB Size Report
1Immunology Program, H. Lee Mo tt Cancer Center and Research Institute, Department of .... *Correspondence: H Yu or R Jove; E-mail: huayu@moffitt.usf.edu.
ã

Oncogene (2002) 21, 2000 ± 2008 2002 Nature Publishing Group All rights reserved 0950 ± 9232/02 $25.00 www.nature.com/onc

Constitutive Stat3 activity up-regulates VEGF expression and tumor angiogenesis Guilian Niu1, Kenneth L Wright1, Mei Huang2, Lanxi Song3, Eric Haura3, James Turkson2, Shumin Zhang1, Tianhong Wang4, Dominic Sinibaldi2, Domenico Coppola3, Richard Heller3, Lee M Ellis5, James Karras6, Jacqueline Bromberg7, Drew Pardoll4, Richard Jove*,2 and Hua Yu*,1 1

Immunology Program, H. Lee Mott Cancer Center and Research Institute, Department of Oncology, University of South Florida College of Medicine, Tampa, Florida, FL 33612, USA; 2Molecular Oncology Program, H. Lee Mott Cancer Center and Research Institute, Department of Oncology, University of South Florida College of Medicine, Tampa, Florida, FL 33612, USA; 3 Clinical Investigations Program, H. Lee Mott Cancer Center and Research Institute, Department of Oncology, University of South Florida College of Medicine, Tampa, Florida, FL 33612, USA; 4Department of Oncology, Johns Hopkins University School of Medicine, Baltimore, Maryland, MD 21205, USA; 5Department of Surgical Oncology, University of Texas M. D. Anderson Cancer Center, Houston, Texas, TX 77030, USA; 6Antisense Drug Discovery, Isis Pharmaceuticals, Inc., Carlsbad, California, CA 92008, USA; 7Department of Medicine, Memorial Sloan-Kettering Cancer Center, New York, NY 10021, USA

Non-receptor and receptor tyrosine kinases, such as Src and EGF receptor (EGFR), are major inducers of vascular endothelial growth factor (VEGF), one of the most potent mediators of angiogenesis. While tyrosine kinases signal through multiple pathways, signal transducer and activation of transcription 3 (Stat3) is a point of convergence for many of these and is constitutively activated with high frequency in a wide range of cancer cells. Here, we show that VEGF expression correlates with Stat3 activity in diverse human cancer cell lines. An activated Stat3 mutant (Stat3C) up-regulates VEGF expression and stimulates tumor angiogenesis. Stat3Cinduced VEGF up-regulation is abrogated when a Stat3binding site in the VEGF promoter is mutated. Furthermore, interrupting Stat3 signaling with dominant-negative Stat3 protein or Stat3 antisense oligonucleotide in tumor cells down-regulates VEGF expression. Consistent with an important role of Stat3 in VEGF upregulation induced by various oncogenic tyrosine kinases, v-Src-mediated VEGF expression is inhibited when Stat3 signaling is blocked. Moreover, chromatin immunoprecipitation assays indicate that Stat3 protein binds to the VEGF promoter in vivo and mutation of a Stat3-binding site in the VEGF promoter abrogates v-Src-induced VEGF promoter activity. These studies provide evidence that the VEGF gene is regulated directly by Stat3 protein, and indicate that Stat3 represents a common molecular target for blocking angiogenesis induced by multiple signaling pathways in human cancers. Oncogene (2002) 21, 2000 ± 2008. DOI: 10.1038/sj/ onc/1205260 Keywords: Stat3 activation; VEGF; tumor angiogenesis

*Correspondence: H Yu or R Jove; E-mail: huayu@mott.usf.edu or E-mail: richjove@mott.usf.edu Received 5 October 2001; revised 30 November 2001; accepted 12 December 2001

Introduction It has been well established that all successful tumors must undergo neovascularization, or angiogenesis, in order to acquire nutrients for continued growth and metastatic spread (Folkman, 1995). Vascular endothelial growth factor (VEGF) is one of the most important of all known inducers of angiogenesis (Grunstein et al., 1999; Millauer et al., 1994; Plate et al., 1992; Shweiki et al., 1992). Induction of VEGF and tumor angiogenesis can develop as a result of environmental conditions, notably hypoxia (Mukhopadhyay et al., 1995b; Shweiki et al., 1992), or genetic alterations (Jiang et al., 2000; Pages et al., 2000; Rak et al., 1995, 2000), including activation of oncogenic tyrosine kinases (Dankbar et al., 2000; Ellis et al., 1998; Mukhopadhyay et al., 1995a; Petit et al., 1997; Wiener et al., 1999). Constitutive activation of tyrosine kinases is highly prevalent in a wide range of cancers and their role as important VEGF inducers has been well established. For example, abnormal activation of EGFR and Src signaling pathways that induce VEGF expression (Ellis et al., 1998; Mukhopadhyay et al., 1995a; Petit et al., 1997; Wiener et al., 1999) has been demonstrated in breast cancer (Garcia et al., 1997, 2001; Khazaie et al., 1993; Luttrell et al., 1988, 1994; Muthuswamy and Muller, 1995), while activation of IL-6 receptor, which signals through another tyrosine kinase signaling pathway, is responsible for myeloma VEGF expression and vascularization (Dankbar et al., 2000). Although tyrosine kinases signal through multiple pathways, Stat3 is a point of convergence for many of these (Bowman et al., 2000; Bromberg and Darnell, 2000; Catlett-Falcone et al., 1999a). Accumulating evidence is de®ning a critical role for Stat3 in oncogenesis (Bowman et al., 2000; Bromberg and Darnell, 2000; Bromberg et al., 1999; Catlett-Falcone et al., 1999a,b; Grandis et al., 2000). It has been shown that constitutive activation of Stat3 signaling contri-

Stat3 induces VEGF expression and angiogenesis G Niu et al

butes to oncogenesis by preventing programmed cell death and enhancing cell proliferation (Bowman et al., 2001; Catlett-Falcone et al., 1999b; Grandis et al., 1998, 2000). Stat3 is constitutively activated at high frequency in a wide range of cancers, including various blood malignancies, breast cancer, head and neck cancer, prostate cancer and melanoma (Bowman et al., 2000; Bromberg and Darnell, 2000; Catlett-Falcone et al., 1999a,b; Garcia et al., 2001; Grandis et al., 1998, 2000; Ni et al., 2000; Niu et al., 1999). The list of cancers with constitutive Stat3 signaling continues to grow as more diverse tumor specimens are examined. Interestingly, recent studies show that VEGF induction in cardiac myocytes by glycoprotein (gp) 130, a subunit of the IL-6 receptor, requires Stat3 signaling (Funamoto et al., 2000). However, a requirement for Stat3 signaling in mediating gp130-induced VEGF expression does not necessarily indicate that constitutive Stat3 activity in tumor cells directly induces VEGF expression. Furthermore, because the oncogenic tyrosine

kinases known to induce VEGF expression also activate signaling pathways other than Stat3, the role of constitutive Stat3 activity in tumor VEGF upregulation and angiogenesis remains to be de®ned. We show here that VEGF expression in diverse cancer cells correlates with constitutive Stat3 activity. Enforced Stat3 activity in tumor cells up-regulates VEGF expression and enhances tumor angiogenesis. Conversely, interrupting Stat3 signaling in tumor cells inhibits VEGF expression, identifying Stat3 as a target for anti-angiogenic therapy. In addition, we demonstrate that Src-induced VEGF expression is reduced when Stat3 signaling is blocked. Moreover, VEGF promoter mutagenesis and chromatin immunoprecipitation assays indicate that the VEGF gene is directly regulated by the Stat3 protein. Our ®ndings thus provide a direct link between Stat3 activity and VEGF induction in tumors, suggesting that targeting Stat3 for therapeutic intervention in cancer may disrupt angiogenesis induced by multiple tyrosine kinases.

2001

Figure 1 Correlation of VEGF expression with Stat3 activity in diverse human cancer cell lines. (a) Upper panel: Stat3 DNAbinding activity determined by EMSA using nuclear extracts prepared from human breast cancer, head and neck carcinoma, and melanoma cell lines. EGFR-expressing NIH3T3 ®broblasts were stimulated with EGF and used here as markers for activated Stat3 and Stat1 dimers bound to labeled hSIE probe. Lower panel: Western blot analysis of lysates from the same cells probed with VEGF or b-actin antibodies. The relative levels of Stat3-DNA binding activity and VEGF protein expression were quanti®ed by ImageQuant software (Molecular Dynamics, Sunnyvale, CA, USA). The overall correlation coecient between VEGF expression and Stat3 activity in the three groups of tumor cell lines is 0.936 with a P-value of 50.001. (b) Supershift analysis using Stat1 and Stat3 speci®c antibodies incubated with nuclear extracts from the indicated cancer cell lines con®rmed that Stat3, but not Stat1, was activated. Previous studies demonstrated activation of Stat3, but not Stat1, in MDA-MB-453 and SK-BR-3 cell lines (Garcia et al., 1997) Oncogene

Stat3 induces VEGF expression and angiogenesis G Niu et al

2002

Results Stat3 activity is associated with VEGF up-regulation in diverse human cancer cell lines To evaluate if constitutively-activated Stat3 may contribute to VEGF up-regulation in cancer, we ®rst assessed VEGF expression and Stat3 activity in 10 human cancer cell lines, which span three di€erent types of cancer (Figure 1). EGFR and Src signaling pathways, both of which can induce VEGF expression (Ellis et al., 1998; Mukhopadhyay et al., 1995a; Petit et al., 1997; Wiener et al., 1999), are known to contribute to Stat3 activation in breast cancers (Garcia et al., 1997, 2001). In addition, EGFR signaling has been found to activate Stat3 in head and neck carcinoma (Grandis et al., 1998). As shown in Figure 1a, the

Figure 2 Constitutively-activated Stat3 mutant protein up-regulates VEGF expression in both normal ®broblasts and in B16 tumor cells. (a,b) Upper panels: Stat3 DNA-binding activity determined by EMSA in NIH3T3 clones (a) and in B16 cells (b) transfected with the Stat3C expression vector and selected by G418. Note that while clone NIH3T3/Stat3C-5 was resistant to G418 selection, it did not express activated Stat3. Lower panel: Western blot analyses with VEGF or b-actin antibodies. wt=parental NIH3T3 cells

expression of VEGF correlated with elevated Stat3 DNA-binding activity in both breast carcinoma and head and neck cancer cell lines. While the oncogenic events responsible for melanoma VEGF up-regulation and Stat3 activity remain to be de®ned, the expression of VEGF also paralleled elevated Stat3 DNA-binding activity in these cells. Supershift experiments with Stat3-speci®c antibodies demonstrate that Stat3 protein is the predominant STAT family member activated in every case (Figure 1b). Because VEGF expression correlates with Stat3 activity in all of the cancer cell lines examined, constitutive activation of Stat3, which occurs at high frequency in numerous diverse cancers, may be an important inducer of VEGF in cancer angiogenesis. A constitutively-activated Stat3 mutant induces VEGF expression and stimulates tumor angiogenesis To directly test the hypothesis that constitutivelyactivated Stat3 has a causal role in VEGF induction in tumor cells, we determined whether persistent Stat3 signaling by itself could up-regulate VEGF expression. Stable expression of Stat3C, an oncogenic mutant form of Stat3 that is constitutively activated without tyrosine phosphorylation (Bromberg et al., 1999), in NIH3T3 ®broblasts led to increased Stat3 DNA-binding activity and enhanced expression of VEGF (Figure 2a). Of the four Stat3C-transfected NIH3T3 clones that survived G418 selection, only one (Stat3C-5) did not contain elevated Stat3 DNA-binding activity. Lack of increased Stat3 activity in these cells correlated with the absence of enhanced VEGF expression when compared to that of wild-type NIH3T3 cells. B16 murine melanoma cells display relatively low levels of constitutive Stat3 activation (Niu et al., 1999), and enforced expression of Stat3C up-regulated the expression of VEGF (Figure 2b). When these Stat3C-transfected B16 tumor cells were mixed with Matrigel and implanted in vivo, a robust increase in vascularization was observed (Figure 3a). This increased angiogenesis was con®rmed by higher hemoglobin content (Figure 3b). These ®ndings,

Figure 3 Enforced constitutive Stat3 activity stimulates formation of tumor vascular capillaries. B16 tumor cells stably transfected with either pcDNA3 or pStat3C expression vector were placed in Matrigel and implanted in C57BL/6 mice (Coughlin et al., 1998). (a) Photomicrographs were taken 5 days after implanting Matrigel/tumor cells (n=2). (b) Hemoglobin assays were performed 7 days after injecting Matrigel/tumor cells. Values are means+s.d. Data shown are average of 10 samples from each group from three independent experiments (P50.05, Student's t-test, two tailed) Oncogene

Stat3 induces VEGF expression and angiogenesis G Niu et al

combined together, indicate that constitutively-activated Stat3 induces VEGF expression, which was accompanied by increased tumor angiogenesis in vivo.

2003

Stat3 up-regulates VEGF expression through the VEGF gene promoter

Figure 4 Stat3 up-regulates VEGF expression through the VEGF promoter. (a) Sequences of Stat3-binding sites. The mismatched bases in putative VEGF gene Stat3-binding sites compared to the known CRP gene Stat3-binding site are underlined, and lower case letters indicate mutations. Wt=wild type; Mut=mutant. (b) DNA-binding activity of Stat3-binding sequences (7848) as measured by competition EMSA. FIRE= the unrelated c-fos intragenic regulatory element oligonucleotide. (c) Chromatin immunoprecipitation (ChIP) assay was performed using either a Stat3-speci®c antibody, an irrelevant antibody (antiRho), or no antibody. PCR primers were designed to yield a 130bp product, which includes the Stat3-binding site (7848) of the VEGF promoter. As a negative control, PCR reactions using primers for the mouse albumin gene previously described (Wells et al., 2000) were included in these experiments. Input lane represents 0.02% of total chromatin used in ChIP assays. (d) Stat3C-induced VEGF promoter activities with and without mutations in the Stat3-binding site (7848). B16 cells were transiently transfected with the pGL3VEGF plasmid or its derivative containing site-speci®c mutations in the Stat3-binding site at position 7848 (Mut) within the VEGF promoter as shown

Our results thus far show that Stat3 activity stimulates VEGF expression and tumor angiogenesis. We next investigated if Stat3 could regulate the VEGF promoter directly. By searching for potential STATbinding sites with the consensus sequences, TT(N4)AA and TT(N5)AA (Seidel et al., 1995), six candidate Stat3-binding sites were identi®ed in the 2.4 kb VEGF promoter region. Two sites at positions 7848 and 7630 upstream of the transcription initiation site in the VEGF promoter most closely resembled a previously de®ned Stat3-binding site within the C-reactive protein (CRP) promoter (Turkson et al., 1998; Zhang et al., 1996). The site at 7848 has a one-base mismatch relative to the CRP Stat3-binding site, while the site at 7630 has two-base mismatches (Figure 4a). Because an individual natural STAT-binding site can have weak binding anity (Xu et al., 1996), we tested if candidate Stat3-binding sites in the VEGF promoter could compete for the binding of Stat3 to a high-anity variant of the Stat3-binding site in the c-fos promoter's sis-inducible element (hSIE) (Wagner et al., 1990). An oligonucleotide containing the putative VEGF Stat3binding site at 7848 competed e€ectively with labeled hSIE probe for Stat3 binding (Figure 4b). In contrast, the putative VEGF Stat3-binding site at 7630, which has two mismatch bases instead of one, could not compete for the binding of Stat3 to hSIE (data not shown). The competition against hSIE using an oligonucleotide containing the VEGF Stat3-binding site at 7848 was speci®c, since an irrelevant oligonucleotide (FIRE) (Turkson et al., 1998) and a mutated form of the 7848 Stat3-binding oligonucleotide (Figure 4a, line 3) failed to compete with hSIE for Stat3 binding (Figure 4b). To determine if Stat3 protein could directly bind to the Stat3-binding site in the VEGF promoter in vivo, chromatin immunoprecipitation (ChIP) assays were performed (Wells et al., 2000). This technique allows for the detection of speci®c genomic DNA sequences that are associated with a particular transcription factor in intact cells. As shown in Figure 4c, an association of Stat3 with the VEGF promoter in vivo in v-Src transformed NIH3T3 cells was detected. Immunoprecipitation with a Stat3 antibody followed by PCR using oligonucleotide primers that amplify a 130bp region spanning the Stat3-binding site at 7848

above. Co-transfection with either Stat3C expression vector or pcDNA3 control empty vector was as indicated. Values for luciferase activity represent means+s.d. of triplicate transfections representing at least three experiments. Luciferase activity was normalized to transfection eciency using b-galactosidase activity as an internal control Oncogene

Stat3 induces VEGF expression and angiogenesis G Niu et al

2004

Figure 5 Inhibition of Stat3 signaling in tumor cells reduces the expression of VEGF. (a,b) Western blot analyses of B16 and SCK tumor cells transfected with the indicated expression vectors or oligonucleotides. GFP=pIRES-EGFP control vector, Stat3b=pIRES-Stat3b expression vector. ASO=antisense oligonucleotide against Stat3, CO=control mismatch oligonucleotide. (c,d) The two murine tumor cell lines were transfected with pGL3VEGF luciferase reporter in the presence of additional expression vectors or oligonucleotides, as indicated. The levels of luciferase activity in tumor cells transfected with pGL3VEGF in the presence of pIRES-EGFP or control oligonucleotide were designated as 100%. These experiments were repeated at least twice with similar results

within the VEGF promoter yielded a 130-bp band. In contrast to this, immunoprecipitation with an irrelevant antibody (anti-RhoA) or no antibody resulted in the absence of this 130-bp band, demonstrating the speci®city of the interaction between Stat3 and the VEGF promoter. The speci®c interaction between Stat3 and the VEGF promoter was further indicated when PCR ampli®cation of the murine albumin promoter (which does not possess Stat3 sites) from all samples revealed no enrichment for nonspeci®c DNA sequences in the Stat3 immunoprecipitation reaction (Figure 4c). We next tested if mutating the Stat3-binding site in the VEGF promoter would result in loss of Stat3Cinduced VEGF promoter activity. The same mutations at the 7848 site shown in Figure 4a were introduced into the VEGF promoter linked to a luciferase reporter construct, pGL3VEGF (Akagi et al., 1998). Cotransfection of Stat3C expression vector and wild-type pGL3VEGF into B16 cells stimulated VEGF promoter activity (Figure 4d). In contrast, when the Stat3binding site at 7848 was mutated in pGL3VEGF, Stat3C-induced VEGF promoter activity was abrogated (Figure 4d). Taken together, these data provide Oncogene

evidence that Stat3 induces VEGF expression directly through the VEGF promoter. Blocking Stat3 in tumor cells down-regulates VEGF expression To determine if Stat3 is a potential target for antiangiogenic therapy, we assessed whether blocking Stat3 signaling in tumor cells down-regulates VEGF expression. B16 tumor cells were transiently transfected with either an expression vector encoding a dominantnegative variant of Stat3 protein, designated Stat3b (Caldenhoven et al., 1996; Turkson et al., 1998), or Stat3 antisense oligonucleotide. As shown in Figure 5a, inhibition of constitutive Stat3 signaling by either approach in B16 tumor cells down-regulates expression of the endogeneous VEGF gene. To con®rm that blocking Stat3 signaling represses transcriptional activity of the VEGF promoter under these conditions, B16 cells were transfected with Stat3b expression vector and a luciferase reporter construct based on the VEGF promoter (Akagi et al., 1998). In the absence of Stat3b expression, VEGF promoter activity was readily detectable in B16 tumor cells as indicated

Stat3 induces VEGF expression and angiogenesis G Niu et al

constitutively-activated Stat3 (data not shown). Transcriptional down-regulation of VEGF expression as a consequence of disruption of Stat3 signaling was veri®ed in SCK-1 cells (Figure 5b,d). These results demonstrate that interrupting constitutive Stat3 signaling in tumor cells down-regulates VEGF expression.

2005

Stat3 is obligatory for Src-induced VEGF up-regulation

Figure 6 Stat3 signaling is obligatory for VEGF expression induced by the Src oncoprotein. (a) Western blot analysis of vSrc-transformed NIH3T3 ®broblasts transiently transfected with indicated vectors and oligonucleotides. Antibodies to Stat3, VEGF and b-actin proteins were used as probes. (b) v-Srcinduced VEGF expression occurs at transcriptional level. NIH3T3 ®broblasts were transiently transfected with a luciferase reporter construct under the control of the VEGF promoter (pGL3VEGF) together with the indicated expression vectors (Turkson et al., 1998). In the presence of Stat3b, v-Src-stimulated VEGF promoter activity is abrogated. (c) v-Src-induced VEGF promoter activity requires the Stat3-binding site. NIH3T3 ®broblasts were transfected with pGL3VEGF and its derivative (Mut) containing site-speci®c mutations at the consensus Stat3-binding site (position 7848). Co-transfection with either a pcDNA3 empty control vector or v-Src expression vector was as indicated. Two additional experiments were performed with similar results

by expression levels of luciferase. Co-transfection of Stat3b vector, but not the control empty vector, decreased transcriptional activity of the VEGF promoter by 2 ± 3-fold (Figure 5c). This inhibitory e€ect on transcriptional activity of the VEGF promoter was also observed in B16 tumor cells transfected with Stat3 antisense oligonucleotide (Figure 5c). The murine mammary tumor cell line, SCK-1, also displays

Abnormal activation of the Src tyrosine kinase is prevalent in diverse cancers and its role in VEGF induction and tumor angiogenesis has been well documented (Ellis et al., 1998; Mukhopadhyay et al., 1995a; Wiener et al., 1999). Src has also been shown to be obligatory for hypoxia-induced VEGF expression (Mukhopadhyay et al., 1995b), further demonstrating a pivotal role of Src in tumor angiogenesis. While Src tyrosine kinase is known to activate multiple signaling pathways, like many oncogenic tyrosine kinases Src is upstream of Stat3 (Bromberg et al., 1998; Turkson et al., 1998; Yu et al., 1995). To assess whether Stat3 is necessary for VEGF up-regulation by v-Src, we determined if blocking Stat3 signaling could inhibit VEGF expression induced by Src tyrosine kinase activity. NIH3T3 ®broblasts stably transformed by the v-Src oncoprotein (Turkson et al., 1998) were transiently transfected with a Stat3b expression vector. Enforced Stat3b expression was accompanied by a reduction of VEGF expression in v-Src-NIH3T3 ®broblasts (Figure 6a). Stat3 antisense oligonucleotide or control mismatch oligonucleotide was also transfected into v-Src-NIH3T3 cells. A reduction of endogenous Stat3 protein levels as a result of Stat3 antisense oligonucleotide transfer caused inhibition of VEGF expression (Figure 6a). Furthermore, transcriptional activity of the VEGF promoter (Akagi et al., 1998) was diminished by the dominant-negative Stat3b protein, as shown by luciferase reporter gene expression assays (Figure 6b). We further determined if Stat3 binding to the VEGF promoter is required for Srcinduced VEGF expression. As shown in Figure 6c, sitespeci®c mutations of the Stat3-binding site at position 7848 within the VEGF promoter abrogated Srcinduced VEGF promoter activity. Taken together, these results demonstrate that Src-induced VEGF upregulation requires Stat3-mediated activation of VEGF promoter. Discussion Both VEGF over-expression and constitutive activity of Stat3 occur at high frequency in diverse human tumors. While previous work has shown that Stat3 signaling is required for gp130-induced VEGF expression in cardiac myocytes (Funamoto et al., 2000), whether constitutive Stat3 activation is sucient to induce VEGF up-regulation and thus contribute to tumor angiogenesis has not been addressed before. Furthermore, the signaling pathway downstream of Stat3 leading to VEGF induction has not been de®ned. Oncogene

Stat3 induces VEGF expression and angiogenesis G Niu et al

2006

Oncogene

Our results show that constitutively-activated Stat3 alone is capable of activating VEGF expression and stimulating tumor angiogenesis. Moreover, Stat3induced VEGF up-regulation requires the Stat3binding site at position 7848 in the VEGF promoter and Stat3 protein binds to the VEGF promoter region containing this site in vivo, providing evidence that VEGF is a direct target gene of Stat3. Our ®ndings also demonstrate that interrupting Stat3 signaling inhibits VEGF expression, identifying Stat3 as a promising molecular target for tumor anti-angiogenesis therapy. As Stat3 is downstream of several important angiogenic tyrosine kinases, such as Src and EGFR, our ®ndings suggest that blocking Stat3 may inhibit neovascularization mediated by multiple angiogenic signaling pathways. Consistent with this hypothesis, we show that Src-induced VEGF up-regulation requires Stat3 activation and the identi®ed Stat3-binding site in the VEGF promoter. Because Src activation is necessary for VEGF up-regulation induced by hypoxia, which has a pivotal role in solid tumor angiogenesis (Mukhopadhyay et al., 1995b; Shweiki et al., 1992), targeting Stat3 in tumors may potentially not only inhibit VEGF expression induced by many oncogenic tyrosine kinases prevalent in cancers, but also block hypoxia-mediated VEGF up-regulation. In addition to tyrosine kinases, other oncogenic signaling pathways that are independent of Stat3 are also known to induce VEGF expression in tumors (Jiang et al., 2000; Pages et al., 2000; Rak et al., 1995, 2000). Nevertheless, our ®nding that VEGF expression correlates with Stat3 activity in many tumor cell lines examined supports a critical role of Stat3 in VEGF up-regulation in a variety of cancers. The important role of VEGF in inducing tumor angiogenesis has been well established (Grunstein et al., 1999; Millauer et al., 1994; Plate et al., 1992; Shweiki et al., 1992). However, in addition to an enhanced expression of angiogenic mitogens, such as VEGF, a reduction of angiogenic inhibitors has also been observed during oncogenic progression from normal to cancerous cells (Volpert et al., 1997). Our recent studies indicate that disrupting Stat3 signaling in tumor cells activates the expression of IP-10 and IFN-b (L Burdelya et al., manuscript in preparation). Because both IP-10 and IFN-b are inhibitors of angiogenesis (Coughlin et al., 1998; Dong et al., 1999), constitutive activation of Stat3 may also stimulate tumor angiogenesis by down-regulating the expression of angiostatic mediators during malignant progression. Previous studies have demonstrated that activation of Stat3 by oncogenic tyrosine kinases regulates the expression of genes that are essential for cell growth and survival (Bowman et al., 2001; Bromberg et al., 1999; Catlett-Falcone et al., 1999b; Grandis et al., 2000; Sinibaldi et al., 2000). Our present ®ndings that constitutive Stat3 activation directly promotes VEGF expression and stimulates tumor angiogenesis indicate a far broader role of Stat3 in cancer pathogenesis

than previously anticipated. Thus, constitutive Stat3 activation confers multiple advantages on tumor cells that are essential for successful malignant progression. Because Stat3 represents a point of convergence for many oncogenic as well as angiogenic events, Stat3 is a promising molecular target for powerful intervention in cancer therapy (Turkson and Jove, 2000). Materials and methods Cell lines and mice NIH3T3 ®broblasts and v-Src-transformed NIH3T3 cells were grown in DMEM supplemented with 5% calf serum. B16 melanoma and SCK mammary carcinoma cell lines were grown in RPMI medium supplemented with 10% fetal bovine serum (FBS). Human cancer cell lines were obtained from ATCC and grown in DMEM supplemented with 10% FBS. Female, 6 ± 8 weeks old, C57BL/6 mice were purchased from the National Cancer Institute (Bethesda, MD, USA) and used for in vivo tumor angiogenesis assays. Mice were housed in American Association for Accreditation of Laboratory Animal Care-approved, speci®c pathogen- and viral antibody-free facility located at the H. Lee Mott Cancer Center and Research Institute. Transfection v-Src-transformed NIH3T3 ®broblasts (Turkson et al., 1998), B16 and SCK tumor cells (Coughlin et al., 1998; Niu et al., 1999) were seeded 18 ± 24 h before transfection with expression plasmids or oligonucleotides using Lipofectamine (Life Technologies). For plasmid transfection, 1 mg of either pIRES-EGFP or pIRES-Stat3b was used (Catlett-Falcone et al., 1999b). The sequence for Stat3 antisense oligonucleotide synthesized using phosphorothioate chemistry is 5'AAAAAGTGCCCAGATTGCCC-3'. The sequence for control oligonucleotide is identical to the antisense oligonucleotide except for three mismatched bases (italics), 5'AAAAAGAGGCCTGATTGCCC-3'. Transfection of Stat3C expression vector (Bromberg et al., 1999) into NIH3T3 ®broblasts and B16 tumor cells was carried out using calcium phosphate precipitation, followed by selection in medium supplemented with 1 mg/ml G418. Site-specific mutagenesis The human VEGF promoter reporter construct (pGL3VEGF) consists of a 2.4 kb genomic DNA fragment containing the 5' region of the VEGF gene upstream of the transcription initiation site cloned into the pGL3 vector (Akagi et al., 1998). Potential Stat3-binding sites in the VEGF promoter were determined by searching the 2.4 kb VEGF gene 5' region for consensus STAT-binding sites, TT(N4)AA and TT(N5)AA (Seidel et al., 1995) using sequence analysis software by DNAStar. Two sites at positions 7848 and 7630 upstream of the transcription initiation site were selected from this subset that closely resembled a previously-de®ned Stat3-binding site within the CRP promoter (Zhang et al., 1996). The putative Stat3binding site at 7848 in the pGL3VEGF plasmid was speci®cally mutated (from 5'-TTCCCAAA-3' to 5'gcgtCAAA-3') using the unique site elimination method (Deng and Nickolo€, 1992) and con®rmed by sequencing.

Stat3 induces VEGF expression and angiogenesis G Niu et al

Western blot analysis and electrophoretic mobility shift assays (EMSA) Equal amounts of total cellular proteins were separated by SDS-polyacrylamide gel electrophoresis followed by immunoblotting with anti-Stat3 (K-15, Santa Cruz Biotechnology), anti-VEGF (Neomarkers) or anti-b-actin (Sigma) antibodies, as described (Turkson et al., 1998). EMSA and competition assays were performed as previously described (Turkson et al., 1998). The oligonucleotides containing the putative Stat3binding sites in the VEGF promoter and their derivatives used in EMSA for competing with hSIE probe are as follows: (7848): wild-type: 5'-TGGACACTTCCCAAAGGACC-3' mutant: 5'-TGGACACgcgtCAAAGGACC-3' (7630): wild-type: 5'-CCCCTTTCCAAAGCCCATT-3' mutant: 5'-CCCCatggCAAAGCCCATT-3'. Chromatin immunoprecipitation (ChIP) assays ChIP assays were performed essentially as previously described (Wells et al., 2000). Solubilized chromatin was prepared from a total of 36107 asynchronously growing vSrc transformed NIH3T3 cells. The chromatin solution was diluted 10-fold with ChIP Dilution Bu€er (0.01% SDS; 1.1% Triton X-100; 1.2 mM EDTA; 16.7 mM Tris-HCl, pH 8.1; 167 mM NaCl), and precleared with Protein A beads blocked with 1% salmon sperm DNA and 1% BSA. The precleared chromatin solution was divided and utilized in immunoprecipitation assays with either an anti-Stat3 rabbit polyclonal antibody (sc-482, Santa Cruz Biotechnology) or an antiRhoA rabbit polyclonal antibody (sc-179, Santa Cruz Biotechnology). The immunoprecipitates were then pelleted, washed, and the antibody/protein/DNA complex was eluted o€ the beads by resuspending the pellets in 50 mM NaHCO3 and 1% SDS for 30 min. Crosslinking was reversed and protein and RNA were removed by adding 10 mg Proteinase K and 10 mg RNase A, followed by incubation at 428C for 3 h. Puri®ed DNA was subjected to PCR with primers speci®c for a 130-bp region (7913 to 7783) spanning the Stat3-binding site (7848) in the VEGF promoter. The sequences of the PCR primers used are as follows: VEGF forward(+): 5'-CTGGCCTGCAGACATCAAAGTGAG-3', and VEGF reverse(7): 5'-CTTCCCGTTCTCAGCTCCACAAAC-3'. PCR was run for 38 cycles (948C for 30 s, 588C for 30 s, 728C for 1 min), and ®nal products were resolved on a 2.5% agarose gel containing ethidium bromide. Luciferase assays Transfections of pGL3VEGF reporter construct in combination with various vectors for luciferase assays were carried

out using Lipofectamine reagent according to the manufacturer's protocol (Gibco). Brie¯y, transfections contained a total of 1 mg of DNA, including 0.1 mg of the indicated luciferase reporter construct, 10 ng b-galactosidase expression vector (internal control), and 0.9 mg of v-Src or Stat3C, or Stat3b expression vectors, or control empty vectors. In experiments where antisense oligonucleotides were used, 300 nM of either Stat3 antisense oligonucleotide or a mismatched oligonucleotide was added to cells. Cytosolic fractions were prepared at 48 h post-transfection (Turkson et al., 1998). Samples were analysed with a luminometer and normalized to b-galactosidase activity by colorimetric assay at A570 as an internal control for transfection eciency.

2007

Matrigel assays Matrigel assays were performed as described previously (Coughlin et al., 1998). Brie¯y, 26105 B16 tumor cells stably transfected with either pcDNA3 or pStat3C were resuspended in 100 ml PBS and mixed with 0.5 ml of Matrigel (Collaborative Biomedical Products) on ice, followed by injection subcutaneously into the abdominal midline of C57BL mice. Matrigel plugs were harvested on day 5 for photography or on day 7 for hemoglobin assays. Hemoglobin quanti®cation was carried out by the Drabkin method (Passaniti et al., 1992). Brie¯y, after dissecting away all the surrounding tissue, Matrigel pellets were melted at 48C and assayed for hemoglobin content (Drabkin's reagent kit, Sigma).

Abbreviations Stat3, signal transducer and activation of transcription 3; VEGF, vascular endothelial growth factor; IL-6, interleukin-6; EGF, epithelial growth factor; EMSA, electrophoretic mobility shift assay.

Acknowledgments This work was supported by National Institutes of Health (CA75243, CA89693, CA55652 and CA82533) and by the Dr Tsai-Fan Yu Cancer Research Endowment. We would like to thank Anita Larson for assisting with preparation of the manuscript, Dr Judith Abraham for providing the VEGF promoter and Wenbiao Liu for technical support. This work has been supported in part by the Molecular Biology Core Facility and Biostatistics Core Facility at the H Lee Mott Cancer Center and Research Institute.

References Akagi Y, Liu W, Zebrowski B, Xie K and Ellis LM. (1998). Cancer Res., 58, 4008 ± 4014. Bowman T, Broome MA, Sinibaldi D, Wharton W, Pledger WJ, Sedivy JM, Irby R, Yeatman T, Courtneidge SA and Jove R. (2001). Proc. Natl. Acad. Sci. USA, 98, 7319 ± 7324. Bowman T, Garcia R, Turkson J and Jove R. (2000). Oncogene, 19, 2474 ± 2488. Bromberg J and Darnell Jr JE. (2000). Oncogene, 19, 2468 ± 2473. Bromberg JF, Horvath CM, Besser D, Lathem WW and Darnell Jr JE. (1998). Mol. Cell. Biol., 18, 2553 ± 2558.

Bromberg JF, Wrzeszczynska MH, Devgan G, Zhao Y, Pestell RG, Albanese C and Darnell Jr JE. (1999). Cell, 98, 295 ± 303. Caldenhoven E, van Dijk TB, Solari R, Armstrong J, Raaijmakers JAM, Lammers JWJ, Koenderman L and de Groot RP. (1996). J. Biol. Chem., 271, 13221 ± 13227. Catlett-Falcone R, Dalton WS and Jove R. (1999a). Curr. Opin. Oncol., 11, 490 ± 496. Catlett-Falcone R, Landowski TH, Oshiro MM, Turkson J, Levitzki A, Savino R, Ciliberto G, Moscinski L, Fernandez-Luna JL, Nunez G, Dalton WS and Jove R. (1999b). Immunity, 10, 105 ± 115. Oncogene

Stat3 induces VEGF expression and angiogenesis G Niu et al

2008

Oncogene

Coughlin CM, Salhany KE, Gee MS, LaTemple DC, Kotenko S, Ma X, Gri G, Wysocka M, Kim JE, Liu L, Liao F, Farber JM, Pestka S, Trinchieri G and Lee WM. (1998). Immunity, 9, 25 ± 34. Dankbar B, Padro T, Leo R, Feldmann B, Krop€ M, Mesters RM, Serve H, Berdel WE and Kienast J. (2000). Blood, 95, 2630 ± 2636. Deng WP and Nickolo€ JA. (1992). Anal. Biochem., 200, 81 ± 88. Dong Z, Greene G, Pettaway C, Dinney CP, Eue I, Lu W, Bucana CD, Balbay MD, Bielenberg D and Fidler IJ. (1999). Cancer Res., 59, 872 ± 879. Ellis LM, Staley CA, Liu W, Fleming RY, Parikh NU, Bucana CD and Gallick GE. (1998). J. Biol. Chem., 273, 1052 ± 1057. Folkman J. (1995). N. Engl. J. Med., 333, 1757 ± 1763. Funamoto M, Fujio Y, Kunisada K, Negoro S, Tone E, Osugi T, Hirota H, Izumi M, Yoshizaki K, Walsh K, Kishimoto T and Yamauchi-Takihara K. (2000). J. Biol. Chem., 275, 10561 ± 10566. Garcia R, Bowman TL, Niu G, Yu H, Minton S, MuroCacho CA, Cox C, Falcone R, Fairclough R, Parsons S, Laudano A, Gazit A, Levitzki A, Kraker A and Jove R. (2001). Oncogene, 20, 2499 ± 2513. Garcia R, Yu CL, Hudnall A, Catlett R, Nelson KL, Smithgall T, Fujita DJ, Ethier SP and Jove R. (1997). Cell. Growth Di€er., 8, 1267 ± 1276. Grandis JR, Drenning SD, Chakraborty A, Zhou MY, Zeng Q, Pitt AS and Tweardy DJ. (1998). J. Clin. Invest., 102, 1385 ± 1392. Grandis JR, Drenning SD, Zeng Q, Watkins SC, Melhem MF, Endo S, Johnson DE, Huang L, He Y and Kim JD. (2000). Proc. Natl. Acad. Sci. USA, 97, 4227 ± 4232. Grunstein J, Roberts WG, Mathieu-Costello O, Hanahan D and Johnson RS. (1999). Cancer Res., 59, 1592 ± 1598. Jiang BH, Zheng JZ, Aoki M and Vogt PK. (2000). Proc. Natl. Acad. Sci. USA, 97, 1749 ± 1753. Khazaie K, Schirrmacher V and Lichtner RB. (1993). Cancer Metastasis Rev., 12, 255 ± 274. Luttrell DK, Lee A, Lansing TJ, Crosby RM, Jung KD, Willard D, Luther M, Rodriguez M, Berman J and Gilmer TM. (1994). Proc. Natl. Acad. Sci. USA, 91, 83 ± 87. Luttrell DK, Luttrell LM and Parsons SJ. (1988). Mol. Cell Biol., 9, 497 ± 501. Millauer B, Shawver LK, Plate KH, Risau W and Ullrich A. (1994). Nature, 367, 576 ± 579. Mukhopadhyay D, Tsiokas L and Sukhatme VP. (1995a). Cancer Res., 55, 6161 ± 6165. Mukhopadhyay D, Tsiokas L, Zhou X-M, Foster D, Brugge JS and Sukhatme VP. (1995b). Nature, 375, 577 ± 581.

Muthuswamy SK and Muller WJ. (1995). Oncogene, 11, 271 ± 279. Ni Z, Lou W, Leman ES and Gao AC. (2000). Cancer Res., 60, 1225 ± 1228. Niu G, Heller R, Catlett-Falcone R, Coppola D, Jaroszeski M, Dalton W, Jove R and Yu H. (1999). Cancer Res., 59, 5059 ± 5063. Pages G, Milanini J, Richard DE, Berra E, Gothie E, Vinals F and Pouyssegur J. (2000). Ann. N. Y. Acad. Sci., 902, 187 ± 200. Passaniti A, Taylor RM, Pili R, Guo R, Long PV, Haney JA, Grant DS and Martin GR. (1992). Lab. Invest., 67, 519 ± 528. Petit AM, Rak J, Hung MC, Rockwell P, Goldstein N, Fendly B and Kerbel RS. (1997). Am. J. Pathol., 151, 1523 ± 1530. Plate KH, Breier G, Weich HA and Risau W. (1992). Nature, 359, 845 ± 848. Rak J, Mitsuhashi Y, Bayko L, Filmus J, Shirasawa S, Sasazuki T and Kerbel RS. (1995). Cancer Res., 55, 4575 ± 4580. Rak J, Yu JL, Klement G and Kerbel RS. (2000). J. Investig. Dermatol. Symp. Proc., 5, 24 ± 33. Seidel HM, Milocco LH, Lamb P, Darnell Jr JE, Stein RB and Rosen J. (1995). Proc. Natl. Acad. Sci. USA, 92, 3041 ± 3045. Shweiki D, Itin A, So€er D and Keshet E. (1992). Nature, 359, 843 ± 845. Sinibaldi D, Wharton W, Turkson J, Bowman T, Pledger WJ and Jove R. (2000). Oncogene, 19, 5419 ± 5427. Turkson J, Bowman T, Garcia R, Caldenhoven E, de Groot RP and Jove R. (1998). Mol. Cell Biol., 18, 2545 ± 2552. Turkson J and Jove R. (2000). Oncogene, 19, 6613 ± 6626. Volpert OV, Dameron KM and Bouck N. (1997). Oncogene, 14, 1495 ± 1502. Wagner BJ, Hayes TE, Hoban CJ and Cochran BH. (1990). EMBO J., 9, 4477 ± 4484. Wells J, Boyd KE, Fry CJ, Bartley SM and Farnham PJ. (2000). Mol. Cell Biol., 20, 5797 ± 5807. Wiener JR, Nakano K, Kruzelock RP, Bucana CD, Bast Jr RC and Gallick GE. (1999). Clin. Cancer Res., 5, 2164 ± 2170. Xu X, Sun YL and Hoey T. (1996). Science, 273, 794 ± 797. Yu CL, Meyer DJ, Campbell GS, Larner AC, Carter-Su C, Schwartz J and Jove R. (1995). Science, 269, 81 ± 83. Zhang D, Sun M, Samols D and Kushner I. (1996). J. Biol. Chem., 271, 9503 ± 9509.