Activated Signal Transducer and Activator of Transcription 3 (STAT3) Supports the Malignant Phenotype of Human Pancreatic Cancer

Transcription

1 GASTROENTEROLOGY 2003;125: Activated Signal Transducer and Activator of Transcription 3 (STAT3) Supports the Malignant Phenotype of Human Pancreatic Cancer ARNE SCHOLZ,* SANDRA HEINZE,* KATHARINA M. DETJEN,* MICHAEL PETERS,* MARTINA WELZEL,* PETER HAUFF, MICHAEL SCHIRNER, BERTRAM WIEDENMANN,* and STEFAN ROSEWICZ* *Department of Hepatology, Gastroenterology, Endocrinology and Metabolism, Campus Virchow-Klinikum Charité, Humboldt University, Berlin; and Research Laboratories, Schering AG, Berlin, Germany Background & Aims: Constitutive activation of signal transducer and activator of transcription 3 (STAT3) has been implicated in regulation of growth and malignant transformation.we therefore analyzed the expression and biologic significance of STAT3 in human pancreatic cancer cells. Methods: Expression and activation of STAT3 were investigated by immunohistochemistry and immunoblotting.functional inactivation of STAT3 was achieved by stable transfection of dominant-negative STAT3 constructs in 2 pancreatic cancer cell lines and confirmed by electrophoretic mobility shift assay and immunoblotting.cell proliferation and tumorigenicity were evaluated by cell counting, colony formation in soft agar, and xenotransplantation in nude mice.stat3- dependent cell cycle distribution was monitored by flow cytometry, immunoprecipitation, immunoblotting, and histone H1 and GST-Rb kinase assays. Results: Compared with nontransformed human pancreas, activated STAT3 is overexpressed in ductal carcinoma cells but not in ducts from chronic pancreatitis.constitutive activation was also observed in all human pancreatic cancer cell lines examined.functional inactivation of STAT3 resulted in significant inhibition of anchorage-dependent and -independent proliferation in vitro and reduced tumor growth in vivo.cell cycle analysis showed a delay of G 1 /S-phase progression due to inhibition of cyclin-dependent kinase 2 activity based on increased expression of p21 WAF1 in vitro and in vivo.blocking of the STAT3 upstream activator Janus kinase 2 by tyrphostin also resulted in growth arrest because of delayed G 1 /S-phase progression and increased expression of p21 WAF1. Conclusions: On malignant transformation, activated STAT3 promotes cellular proliferation by acceleration of G 1 /S-phase progression and thereby contributes to the malignant phenotype of human pancreatic cancer. Adenocarcinoma of the pancreas represents the fifth most common cause of cancer-related death in industrialized western countries. 1 Approximately 95% of pancreatic adenocarcinomas are classified as a ductal phenotype. Pancreatic cancer is associated with a poor prognosis; median survival is 4 8 months, and the overall 5-year survival rate does not exceed 2% in most studies. 2,3 Because of advanced stage of disease at the time of diagnosis, 20% of tumors are amenable for resection, which is still considered the only curative therapy. 4 Alternative therapeutic approaches are of limited efficacy because pancreatic cancer is characterized by a primary resistance to a broad spectrum of proapoptotic stimuli such as chemotherapy, radiation, and combinations thereof. 3 Thus, the molecular mechanisms involved in pancreatic carcinogenesis have to be uncovered to further identify novel targets for therapeutic intervention. In this context, genetic alterations play a predominant role in tumorigenesis and apoptosis susceptibility. In recent years, the prevalent genetic alterations associated with malignant transformation of the pancreas have been identified. Among others, the 4 most common alterations comprise activation of the K-ras oncogene as well as functional inactivation of the tumor suppressor genes p16, p53, and DPC4. 5 Most pancreatic adenocarcinomas have accumulated 3 or 4 genetic alterations simultaneously. 6 In addition to these well-defined genetic lesions, there is increasing evidence for a potential involvement of signal transducer and activator of transcription 3 (STAT3) in malignant transformation. STAT3, which belongs to the 7 family members of latent cytosolic transcription factors, was first characterized in cytokine- Abbreviations used in this paper: CDK, cyclin-dependent kinase; EGF-R, epidermal growth factor receptor; FCS, fetal calf serum; IL, interleukin; JAK, Janus kinase; SDS-PAGE, sodium dodecyl sulfate polyacrylamide gel electrophoresis; SOCS, suppressor of cytokine signaling; STAT3, signal transducer and activator of transcription by the American Gastroenterological Association /03/$30.00 doi: /s (03)

2 892 SCHOLZ ET AL. GASTROENTEROLOGY Vol.125, No.3 and growth factor dependent signal transduction. 7 Molecular insights into STAT signaling were mainly derived from investigations of the Janus kinase ( JAK)- STAT pathway. 8,9 Receptor-associated specific JAKs, activated by cognate ligand binding, subsequently activate STAT proteins by tyrosine phosphorylation, which then dimerize and translocate to the nucleus where they bind to specific DNA sequences and regulate gene transcription. 9,10 Additional tyrosine kinases that mediate STAT activation include growth factor receptors and cytoplasmic tyrosine kinases, particularly the src kinase family. 11 In addition to tyrosine phosphorylation, the transcriptional activity of STAT3 has been shown to be enhanced by phosphorylation of a serine residue in the C-terminal transcriptional activation domain. 12 The transcriptional targets of STAT3 include genes that control inflammatory responses, survival, and proliferation. In addition, they encompass genes of the suppressor of cytokine signaling (SOCS) family, which encode inhibitors of the JAK-STAT signaling pathway (for review, see Greenhalgh et al. 13 ). Thus, a negative feedback loop is initiated that tightly regulates STAT3 activation under physiologic conditions. Several studies have proposed that STAT3 plays a crucial role in the regulation of cell proliferation, apoptosis, and differentiation that is further supported by early embryonic lethality at day 7 of STAT3 knockout mice. 14 Apart from the transient activation of STAT3 signaling under physiologic conditions, constitutive activation of STAT3 has recently been observed in various human carcinomas and tumor cell lines Several lines of evidence indicate that constitutive activation of STAT3 contributes to malignant transformation because abrogation of STAT3 activity resulted in inhibition of proliferation and induction of apoptosis in vitro and decreased tumorigenicity in vivo. 11,19 Furthermore, genetic evidence of the intrinsic oncogenic potential of STAT3 was provided by an artificially mutated constitutively active STAT3 variant that was stably expressed in rodent fibroblasts and induced colonies in soft agar and tumors in nude mice. 20 There is currently no information available about the role of activated STAT3 in human pancreatic cancer. However, prominent candidates involved in aberrant STAT3 activation are represented by the tyrosine kinase src and epidermal growth factor receptor (EGF-R) signaling Both overexpression and activation of src and EGF-R in turn have been documented in human pancreatic cancer. 24,25 Therefore, in the current study, we examined the expression and biological function of activated STAT3 in human pancreatic cancer cells. Materials and Methods Materials The following were purchased: human pancreatic carcinoma cell lines AsPC-1, BxPC-3, CAPAN-1, CAPAN-2, MIA PaCA-2, and PANC-1 from American Type Tissue Culture Collection (Manassas, VA) and DAN-G cells from Deutsches Krebsforschungszentrum (Heidelberg, Germany); Dulbecco s modified Eagle medium, RPMI 1640 medium, phosphate-buffered saline, glutamine and Geneticin (G418), Iscove s modified Dulbecco s modified Eagle medium, and Hyclone fetal calf serum (FCS) from Life Technologies (Karlsruhe, Germany); FCS and penicillin/streptomycin from Biochrom (Berlin, Germany); UltraCulture medium from Bio- Whittaker (Verviers, Belgium); Sepharose A beads from Sigma (Deisenhofen, Germany); nitrocellulose, enhanced chemiluminescence reagent, adenosine triphosphate, and poly[d(i-c)]/ poly[d(i-c)] from Amersham (Braunschweig, Germany); Effectene Transfection Reagent from Qiagen (Hilden, Germany); T4 polynucleotide kinase from Promega (Madison, WI); interleukin (IL)-6 from Calbiochem (Bad Soden, Germany); ribonuclease A and calf histone H1 from Roche (Mannheim, Germany); Rb substrate from Santa Cruz Biotechnology (Santa Cruz, CA); and tyrphostins AG490 and AG1517 from Alexis Biochemicals (Grünberg, Germany). The following antibodies were purchased: p(tyr)stat3 (9E12) and cyclin A (BF683) from Upstate Technologies (Hamburg, Germany); HA-tag antibody from Eurogentec (Seraing, Belgium); STAT3 and p27 Kip1 from BD Transduction Laboratories (Heidelberg, Germany); BCL-X L/S, cyclin D1 (G ), c-myc (9E10), p21 WAF1 (Ab-1), cyclin E (HE12), and Rb (G3-245) from BD Pharmingen (Heidelberg, Germany); STAT3 and STAT1, cyclin-dependent kinase (CDK)-2 (C-22), CDK-4 (M-2), p(tyr) (PY20), JAK2, and proliferating cell nuclear antigen from Santa Cruz Biotechnology; Socs-1 and Socs-3 from Zytomed (Berlin, Germany); p(tyr)stat3 (used for immunohistochemistry) and p(ser)stat3 from New England BioLabs (Frankfurt, Germany); and secondary antibodies from Dianova (Hamburg, Germany). Cell Culture Cell lines were grown as subconfluent monolayers in Dulbecco s modified Eagle medium containing 10% FCS (CAPAN-2, MIA PaCA-2, PANC-1) or in RPMI 1640 containing 20% FCS (AsPC-1), 15% FCS (CAPAN-1), and 10% FCS (BxPC-3, DAN-G). Culture media were supplemented with 2 mmol/l glutamine and 100 U/mL penicillin/streptomycin. Cells were maintained in 95% air and 5% CO 2. LCC-18 cells were generously provided by Kjell Öberg and were cultured as described. 26 Immunohistochemistry Immunohistochemical staining was performed on paraffin-embedded specimens by using the alkaline phosphatase/ anti alkaline phosphatase method described previously 27 except for a modified antigen retrieval procedure. Sections were

3 September 2003 ACTIVATED STAT3 AND HUMAN PANCREATIC CANCER 893 boiled for 1 minute in a 0.01 mol/l citrate buffer (ph 6.0) by using a pressure cooker before they were incubated at 4 C overnight with the antibody to p(tyr)stat3 diluted 1:50 in Tris-buffered saline. Sections treated with Tris-buffered saline instead of the primary antibody were used as a negative control. Digital images were captured with an AxioCam digital camera (Zeiss, Jena, Germany). Immunoblotting Whole cell lysates were prepared by washing cells with ice-cold phosphate-buffered saline and directly lysed in a buffer (20 mmol/l Tris, ph 7.8, 150 mmol/l NaCl 2, 2 mmol/l ethylenediaminetetraacetic acid, 50 mmol/l -glycerolphosphate, 0.5% TNP-40, 1% glycerin, 1 mmol/l Na 3 VO 4, 1 mmol/l dithiothreitol, 10 KIU/mL aprotinin, 10 mmol/l NaF, 2 mol/l leupeptin, and 2 mmol/l phenylmethylsulfonyl fluoride). Extracts were boiled in Laemmli s sample buffer for 5 minutes at 95 C. Aliquots were separated by sodium dodecyl sulfate/polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to nitrocellulose membranes by electroblotting. Membranes were blocked in phosphate-buffered saline/0.1% Tween 20 containing 5% nonfat dry milk for 1 hour at room temperature and incubated with primary antibodies overnight at 4 C. After washes in phosphate-buffered saline/0.1% Tween 20, membranes were exposed to appropriate secondary horseradish peroxidase conjugated antibodies at a dilution of 1:20,000 for 1.5 hours at room temperature. Bands were visualized by enhanced chemiluminescence. Primary antibodies were diluted 1:1000 except for p(tyr)stat3 (1:5000), STAT3 (BD Transduction Laboratories; 1:2500), and c-myc (1:250). For protein extraction of a tumor specimen from nude mice, 200 mg of frozen tissue was homogenized with an UltraTurrax (IKA-Werke, Staufen, Germany) and lysed by incubation with a detergent mix (50 mmol/l Tris-HCl, ph 7, 150 mmol/l NaCl, 0.1% NaN 3,3% Nonidet P-40, 1.5% sodium desoxycholate, and 0.3% SDS) for 20 minutes on ice. Stable Transfection The dominant-negative constructs HA-STAT3D and HA-STAT3F in pcaggs-neo were kindly provided by Professor Hirano (Osaka University, Osaka, Japan). 28 In HA- STAT3F, tyrosine 705, which is responsible for STAT3 activation, was replaced with phenylalanine. In HA-STAT3D, E434 and E435 of the DNA-binding domain were replaced with alanines. The HA tag located at the amino terminus of the dominant-negative STAT3 complementary DNA permits discrimination from endogenously expressed STAT3. Stable transfection of CAPAN-1 and BxPC-3 cells was performed using Effectene Transfection Reagent as recommended by the manufacturer. After 13 days of selection with the neomycin analogue G418 (1 g/ L CAPAN-1; 0.5 g/ L BxPC-3), clones were isolated and expanded. To generate vector controls, both cell lines were transfected with the empty vector. Electrophoretic Mobility Shift Assays Electrophoretic mobility shift assays were performed using the STAT3 consensus oligonucleotide m67sie with the sequence 5 -CGGGAGGGATTTACGGGAAATGCTG Double-stranded oligonucleotide was end labeled with [ - 32 P]adenosine triphosphate in a T4 polynucleotide kinase reaction. Extraction of nuclear proteins was performed as described previously. 30 For DNA-binding reactions, 10 g of nuclear extracts was incubated with 3 ng of the labeled oligonucleotide in 20 L binding buffer (0.02 g/l poly[d(i-c)]/ poly[d(i-c)], 5 mmol/l dithiothreitol, 0.5 g/l bovine serum albumin, 10 mmol/l HEPES, ph 7.9, 100 mmol/l NaCl, 10% glycerol, and 1 mmol/l ethylenediaminetetraacetic acid) for 30 minutes at room temperature. Competition experiments were performed by preincubation with a 100-fold molar excess of either unlabeled consensus or mutated oligonucleotide deficient in STAT3 DNA binding (Santa Cruz Biotechnology) for 30 minutes at room temperature before the radiolabeled probe was added. For supershift experiments, 2 L of STAT1 or STAT3 antibodies were included. Protein/DNA complexes were separated on a 5% Tris/borate/ethylenediaminetetraacetic acid polyacrylamide gel, and bands were detected on autoradiographs of the gel. Electrophoretic mobility shift assay analyses for SP1 DNA binding were conducted as previously described. 30 Cell Growth Assays For evaluation of anchorage-dependent growth, cells were counted by using a hemocytometer. Viability of the cells was confirmed by trypan blue exclusion. Anchorage-independent growth was estimated by colony formation in soft agar as previously described. 31 A total of 10 3 cells per clone were plated in triplicate as a single cell suspension in methylcellulose. Colonies of more than 20 cells were counted after 10 days. Tumor Formation in Nude Mice Ten female BALB/c mice were injected subcutaneously with cells of each cell clone. Tumor size was measured twice weekly by calculating the product of the largest diameter and its perpendicular. At day 21, the mice were killed and the tumors removed for further analysis. The local animal research authorities approved the animal experiments. Cell Cycle Analysis For cell cycle analysis, cells were synchronized by serum starvation for 24 hours, resulting in G 0 /G 1 -phase accumulation. Cells were then released by incubation with the appropriate serum-supplemented medium. For flow cytometry, 10 6 cells were fixed in 70% ethanol for 30 minutes at 20 C and stained with phosphate-buffered saline containing 50 g/ml propidium iodide and 300 g/ml ribonuclease A at 37 C. DNA content of 10 4 cells was analyzed on a FACSCalibur, and cell cycle distribution was quantified by using CellQUEST software (BD Biosciences, Heidelberg, Germany).

4 894 SCHOLZ ET AL. GASTROENTEROLOGY Vol.125, No.3 Immunoprecipitation and Histone H1 and GST-Rb Kinase Assays JAK2, CDK-2, and CDK-4 immunoprecipitation as well as kinase assays were performed as previously described. 31,32 For GST-Rb kinase assay, CDK-4 immunoprecipitates from whole cell lysates were exposed to 20 L kinase buffer and 50 mol/l unlabeled adenosine triphosphate, 7 g GST-Rb substrate, and 10 Ci of [ - 32 P]adenosine triphosphate for 30 minutes at 33 C. Kinase reaction was terminated by boiling the samples in 5 Laemmli buffer for 5 minutes. The immunoprecipitates were then subjected to SDS-PAGE, and CDK-4 activity was determined by autoradiography of the dried gel. Statistical Analysis Data were analyzed by t test using GraphPad statistical software (San Diego, CA). All tests were 2 sided, and differences were considered to be statistically significant at P Data represent means SEM from 3 independent experiments unless otherwise specified. Results Activated STAT3 Is Overexpressed in Human Pancreatic Carcinoma To evaluate the expression of activated STAT3 in human pancreatic cancer, we initially analyzed the prevalence of activated STAT3 in nontransformed pancreatic tissue, chronic pancreatitis, and ductal adenocarcinoma by using a phosphotyrosine-specific STAT3 antibody (p[tyr]-stat3) that exclusively detects active STAT3. All carcinoma samples analyzed (n 11) showed distinct nuclear p(tyr)-stat3 staining in at least 35% of the ductal tumor cells (Figure 1C and D). In contrast, ductal epithelia of nontransformed tissue derived from the resection margins (n 6) (Figure 1A) or from patients with chronic pancreatitis (n 7) (Figure 1B) showed negligible or no p(tyr)-stat3 signal. Consistent with the results obtained in the primary tumors, immunostaining for activated STAT3 was also observed in perineural tumor infiltration (Figure 1E) and malignant transformed ductal cells of lymph node metastasis (Figure 1F). In contrast to the results obtained for activated STAT3, signals observed with an antibody that detects STAT3 expression irrespective of its phosphorylation status were similar in nontransformed and carcinoma samples (data not shown). Therefore, these data suggest that malignant transformation in the human pancreas is associated with activation of STAT3. Expression of Activated STAT3 in Human Pancreatic Carcinoma Cell Lines To characterize STAT3 expression and activation status in vitro, a representative panel of 7 human exocrine pancreatic cancer cell lines was investigated by immunoblotting using specific antibodies for STAT3 (Figure 2). All cell lines examined expressed STAT3 and STAT3 isoforms, although to various extents. In addition, tyrosine-phosphorylated activated STAT3 was expressed in all exocrine pancreatic cancer cell lines but not in neuroendocrine LCC-18 cells, which served as a negative control. In general, the expression pattern of STAT3 and activated STAT3 correlated well in between cell lines, suggesting that most of the cellular STAT3 is expressed in its activated, tyrosine-phosphorylated form. In contrast, serine phosphorylation of STAT3 was mainly detected in DAN-G, MIA-PaCA-2, and PANC-1 cells and did not necessarily correlate with the extent of total STAT3 expression. In analogy to the in vivo situation, these data suggest that constitutive activation of STAT3 represents a common phenomenon in human pancreatic cancer cell lines. STAT3 phosphorylation occurs in response to a broad spectrum of physiologic stimuli, including activation of cytokine receptors and receptor-tyrosine kinases. In an attempt to identify potential upstream signals responsible for STAT3 activation in pancreatic cancer cells, we explored the contribution of EGF-R signaling. A panel of 5 human pancreatic carcinoma cell lines was treated with the highly selective EGF-R antagonist tyrphostin AG for 48 hours, and STAT-3 phosphorylation was evaluated (Figure 3A). Compared with their respective controls, 4 of 5 cell lines showed an almost complete loss of STAT3 tyrosine phosphorylation, indicative of severely reduced STAT3 activity. A very moderate reduction of STAT3 protein was also noted; however, this could not account for the loss of activated STAT3 expression. Thus, EGF-R inhibition and/or upstream regulatory components prevented STAT3 activation. Because SOCS proteins act as endogenous inhibitors of cytokine signaling, we compared the expression of SOCS-1 and SOCS-3 in AG1517-treated cells and controls (Figure 3, lower panels). Neither SOCS-1 nor SOCS-3 was up-regulated in AG1517-treated cells, suggesting that EGF-R inhibition did not affect STAT3 phosphorylation via regulation of SOCS expression. SOCS proteins were instead down-regulated to a variable extent in cells with reduced STAT3 phosphorylation, consistent with SOCS genes representing a transcriptional target of STAT3. MIA-PaCa cells were refractory to treatment with AG1517, suggesting that (1) STAT3

5 September 2003 ACTIVATED STAT3 AND HUMAN PANCREATIC CANCER 895 Figure 1. Expression of active, tyrosine-phosphorylated STAT3 in human pancreatic cancer. Sections of paraffin-embedded pancreatic tissue were immunostained with a p(tyr)stat3 antibody by using the alkaline phosphatase/anti alkaline phosphatase detection method. Representative examples of immunohistochemical staining are shown for (A) normal pancreas, (B) chronic pancreatitis, and (C F ) pancreatic carcinoma. (A, B, and C [left side]) Ductal epithelia of nontransformed pancreas are predominantly negative. In malignant transformed ductal structures of pancreatic carcinoma, (C [right side] and D) strong nuclear immunostaining was observed as well as (E) in perineural tumor infiltration and (F ) lymph node metastasis. activity in this cell line was independent of EGF-R signaling and (2) AG1517 did not suppress STAT3 phosphorylation via unspecific, non EGF-R related mechanisms. To characterize the impact of EGF-R signaling on STAT3 activity in more detail, time dependence and dose dependence of STAT3 inhibition by AG1517 was determined in CAPAN-1 cells. Loss of STAT3 tyrosine

6 896 SCHOLZ ET AL. GASTROENTEROLOGY Vol.125, No.3 corroborate the contribution of EGF-R signaling to constitutive STAT3 binding activity, we also used a neutralizing EGF-R antibody. 35 In support of the results obtained with the AG1517 compound, the EGF-R antibody but not the immunoglobulin G matched control was capable of suppressing STAT3 DNA binding (Figure 3C, lanes 19 and 20), indicating that EGF-R signaling Figure 2. Expression and activation of STAT3 in human pancreatic carcinoma cell lines. Whole cell lysates of 7 exocrine human pancreatic cancer cell lines and a human neuroendocrine cell line (LCC-18) were analyzed by immunoblotting. Equal amounts of protein (26 g) were subjected to SDS-PAGE and were immunoblotted by using an antibody against STAT3. The blot was stripped and sequentially incubated with antibodies against the tyrosine- and serine-phosphorylated form of STAT3, respectively. phosphorylation was time dependent, starting at 24 hours of treatment with AG1517 and maintained throughout the observation period of 3 days (Figure 3B). The reduction of STAT3 phosphorylation was also dose dependent, with effects noted at 1 and 5 mol/l at 72 hours (data not shown). To provide more functional evidence of impaired STAT3 function in AG1517-treated CAPAN-1 cells, we assessed STAT3 DNA-binding activity by electrophoretic mobility shift assay (Figure 3C). Nuclear extracts prepared from vehicle- or tyrphostin-treated cells were incubated with the STAT3 consensus oligonucleotide m67sie 29 and analyzed for protein/dna interactions. Because m67sie contains binding sites for STAT1 and STAT3, we confirmed STAT3 binding by supershift experiments using the appropriate antibodies (Figure 3C, lanes 1 6). A single distinct DNA/protein complex could be detected. This complex was completely competed by an excess of unlabeled consensus oligonucleotide but not affected by a mutated STAT3 oligonucleotide. Supershift analysis using STAT1 and STAT3 antibodies showed no mobility shift by addition of a STAT1 antibody but a complete supershift by 2 independent STAT3 antibodies, thereby confirming the specificity of the STAT3/DNA interacting complex (Figure 3C, lanes 5 and 6). Furthermore, STAT3 DNA binding increased in response to IL-6, a prototype STAT3 activator 34 (Figure 3C, lanes 7 and 8). In contrast to IL-6, AG1517 profoundly reduced STAT3 DNA binding. This inhibition could be noted as early as 12 hours following AG1517 stimulation (Figure 3C, lanes 9 13). At 48 hours, a distinct inhibition was obtained with 100 nmol/l AG1517 and DNA binding was virtually abrogated at 10 mol/l (Figure 3C, lanes 14 18). To further Figure 3. Autocrine EGF-R activation stimulates STAT3 activity of human pancreatic cancer cells. (A) Effect of tyrphostin AG1517 (10 mol/l, 48 hours) on STAT3 expression and phosphorylation as well as on the expression of regulatory SOCS-1 and SOCS-3 molecules. (B) Time course of AG1517-induced effects on STAT3 and SOCS. CA- PAN-1 cells were cultured overnight before stimulation with 10 mol/l AG1517 for the indicated periods. Representative immunoblots from 2 experiments that yielded similar results are shown. (C) Effects of EGF-R inhibition on STAT3 DNA binding. Nuclear extracts from CAPAN-1 cells were incubated with a [ - 32 P]adenosine triphosphate labeled STAT3 consensus oligonucleotide (wt) and electrophoretic mobility shift assays were conducted (lanes 1 6). For competition and supershift experiments, extracts were preincubated with a 100-fold molar excess of either unlabeled consensus (wt) or mutated (mut, deficient in STAT3 DNA binding) oligonucleotide or 2 g of antibodies to STAT1 or STAT3. For stimulation with IL-6 (lanes 7 and 8), cells were serum starved for 4 hours and incubated with 100 ng/ml of IL-6 in serum-free medium for 15 minutes before nuclear extracts were prepared. Treatment with AG1517 (lanes 9 18) and treatment with the anti EGF-R antibody (10 g/ml) (lanes 19 and 20) were performed in the presence of medium. Concentrations are in mol/l. Co, control; V, vehicle (dimethyl sulfoxide).

7 September 2003 ACTIVATED STAT3 AND HUMAN PANCREATIC CANCER 897 represents at least 1 mechanism for the elevated STAT3 activity in human pancreatic cancer cells. Functional Inhibition of Activated STAT3 in Human Pancreatic Carcinoma Cells To explore the biologic relevance of increased STAT3 activity in human pancreatic cancer, we established an in vitro model to suppress STAT3 activation. CAPAN-1 and BxPC-3 cells were stably transfected with 2 different dominant-negative STAT3 complementary DNA constructs. 28 The dominant-negative effect created by point mutations was based on either inactivation of the DNA-binding domain (HA-STAT3D) or by replacement of tyrosine 705, which inhibits activation by tyrosine phosphorylation (HA-STAT3F). Three independent clones for each cell line were generated, and expression of the dominant-negative STAT3 protein was confirmed by the use of an HA-specific antibody (Figure 4A). Analysis of the dominant-negative STAT3 effect on STAT3 expression and phosphorylation status in CAPAN-1 cells showed a slightly elevated expression level of STAT3 in the dominant-negative STAT3-transfected clones compared with vector-transfected or wildtype controls (Figure 4A). Based on the loss of tyrosine 705, overexpression of dominant-negative HA-STAT3F (clone F2 and F5) resulted in decreased expression levels of tyrosine phosphorylated, activated STAT3, indicating the presumed dominant-negative function. In contrast, STAT3 tyrosine phosphorylation of clones expressing HA-STAT3D (clone D4) was comparable to wild-type and vector-transfected controls because the construct contains an intact tyrosine phosphorylation site (Figure 4A). Increased p(ser)-stat3 signals were noted in clones that expressed the dominant-negative STAT3 molecules, which despite their modifications still allow for serine phosphorylation (Figure 4A). Very similar results were obtained in the BxPC-3 cell line on transfection with both dominant-negative STAT3 constructs (data not shown). To then confirm that the dominant-negative approach impaired STAT3 function in the transfected clones, we again assessed STAT3 DNA-binding activity in clones that expressed the dominant-negative constructs and the appropriate vector-transfected and wild-type controls (Figure 4B). Basal STAT3 DNA binding was readily detectable in CAPAN-1 and BxPC-3 cells, although DNA binding was consistently weaker in BxPC-3 cells (Figure 4B, lanes 1 and 2). Vector-transfected variants did not differ from their respective wild-type controls (Figure 4B, lanes 3, 4, 8, and 9). In contrast, the dominant-negative STAT3 clones of either CAPAN-1 or BxPC-3 cells failed to show basal DNA binding (Figure Figure 4. Functional inhibition of STAT3 in CAPAN-1 cells by stable transfection with dominant-negative STAT3 constructs. Two HA-tagged dominant-negative constructs and empty vector controls were stably transfected in CAPAN-1 cells. (A) Expression of dominant-negative STAT3 in CAPAN-1 cells. Whole cell lysates of dominant-negative STAT3 transfectants (D4, F2, and F5) in CAPAN-1 (D16, D19, and F8 in BxPc-3 cells), wild-type (WT), and vector-transfected (V) controls with equal amounts of protein (20 g) were subjected to SDS-PAGE and immunoblotted with an antibody against the HA tag. Blots were stripped and reprobed with STAT3-, p(tyr)stat3-, and p(ser)stat3- specific antibodies. For loading control, expression of proliferating nuclear antigen (PCNA) was evaluated. (B) DNA-binding activity of dominant-negative STAT3 clones from CAPAN-1 cells (lanes 5 7; clones D4, F2, and F5) and BxPC-3 cells (lanes 10 12; clones D16, D19, and F8) was compared with wild-type (WT) and vector-transfected controls (V) (lanes 1 4, 8, and 9) by electrophoretic mobility shift assay. Ten micrograms of nuclear extracts from serum starved (4 hours) cells was incubated with a [ - 32 P]adenosine triphosphate labeled STAT3 consensus oligonucleotide (cons). Nuclear extracts from CAPAN-1 were incubated in parallel with a [ - 32 P]adenosine triphosphate labeled SP1 consensus oligonucleotide and SP1 DNA binding was examined (lanes 13 17). 4B, lanes 5 7 and 10 12). To further confirm that the inhibition of DNA binding was specific for STAT3- DNA interaction, SP1 DNA binding was evaluated in CAPAN-1 clones. No differences in SP1 binding to its consensus oligonucleotide sequence were noted between the STAT3-competent and STAT3-deficient CAPAN-1 cell clones (Figure 4B, lanes 13 17). Taken together, these data indicated successful selective functional inhibition of activated STAT3 in both cell lines. Inhibition of Activated STAT3 Blocks Pancreatic Cancer Cell Growth We next investigated the functional consequences of STAT3 inactivation for growth of CAPAN-1 and

8 898 SCHOLZ ET AL. GASTROENTEROLOGY Vol.125, No.3 BxPC-3 cells. Dominant-negative inhibition of STAT3 function resulted in a profound decrease of anchoragedependent growth when compared with wild-type or vector-transfected controls in both cell lines (Figure 5A and B). Detailed time-course analyses in CAPAN-1 clones documented a time-dependent reduction in cell numbers by more than 60% of control in all 3 dominantnegative STAT3-transfected clones (Figure 5A). Because the tumorigenic potential of adherent cancer cells is better reflected by anchorage-independent growth, we next performed soft agar assays. In both cell lines, colony formation of dominant-negative STAT3-expressing clones was significantly inhibited (Figure 5C and D). These data indicate that activated STAT3 contributes to enhanced anchorage-dependent and -independent proliferation of pancreatic cancer cells in vitro. We next evaluated the antiproliferative effects of dominant-negative STAT3 in vivo. For these experiments, CAPAN-1 cells were injected subcutaneously into nude mice and tumor growth was monitored (Figure 6A). Compared with vector-transfected controls, expression of dominant-negative STAT3 resulted in a moderate timedependent growth inhibition of all STAT3-inactivated cell clones (Figure 6A). Continuous expression of dominant-negative STAT3 during the in vivo experiments was confirmed by performing immunoblot analysis of tumor lysates using an HA-specific antibody (Figure 6B). Because of inconsistent take rates and large variability in Figure 5. Growth inhibition of CAPAN-1 and BxPC-3 cells by dominantnegative STAT3. (A and B) For anchorage-dependent growth, 10 5 (CAPAN-1) or 10 4 (BxPC-3) cells per clone were seeded. Viable cells were counted manually with a hemocytometer at the indicated times (CAPAN-1) or after 96 hours (BxPC-3), respectively. (C and D) Anchorage-independent growth was estimated by colony formation in soft agar. Cell agar suspension of 10 3 cells per clone were seeded. Colonies ( 20 cells) were scored 10 days after plating. (A D) Each experiment was performed in triplicate. Data represent mean SEM of 3 independent experiments. *Significantly different from wild-type (WT) and vector-transfected controls (V) (P 0.05). Figure 6. Activated STAT3 promotes tumor growth in vivo. (A) Ten mice were injected with CAPAN-1 cells of vector-transfected control (V) and dominant-negative STAT3 clones (D4, F2, F5). Tumor volume was measured twice weekly by calculating the product of the largest diameter and its perpendicular. Data represent mean SEM. *Significantly different from vector controls (P 0.05) at indicated times. (B) Immunodetection of dominant-negative STAT3 expression in tumor tissue of mice. A total of 200 mg frozen tissue of the xenotransplanted tumor at day 21 was lysed and homogenized. A total of 50 g protein was subjected to SDS-PAGE and incubated with the antibody against HA tag. tumor growth, a reliable analysis of BxPC-3 clones in vivo was not feasible. Functional Inhibition of Activated STAT3 Delays G 1 /S-Phase Progression To further elucidate the molecular mechanisms responsible for STAT3-regulated growth in pancreatic cancer cells, we examined induction of apoptosis and regulation of cell cycle progression as 2 key components responsible for growth inhibition. Transfection of dominant-negative STAT3 did not result in an increased induction of apoptosis in either cell line as judged by a variety of biochemical and morphologic criteria (data not shown). We therefore used CAPAN-1 cells to investigate the role of activated STAT3 for cell cycle progression in more detail. Cell populations of vector-transfected controls and 3 dominant-negative STAT3-expressing clones were synchronized in G 0 /G 1 -phase by serum starvation for 24 hours. Cells were then released from synchronization and analyzed by flow cytometry after 0, 6, 12, 18, and 24 hours. We consistently observed a distinct delay of G 1 /S-phase transition as a consequence of functional inhibition of STAT3 (Figure 7A and B). Vector-transfected controls entered S phase after 12 hours to a sig-

9 September 2003 ACTIVATED STAT3 AND HUMAN PANCREATIC CANCER 899 Figure 7. Effect of dominant-negative STAT3 on cell cycle progression. Vector (control) and dominant-negative STAT3-transfected CAPAN-1 cells (clone F2) were synchronized in G 0 /G 1 phase by serum starvation for 24 hours. Following stimulation with medium/15% FCS, cells were prepared for flow cytometry analysis after 0, 6, 12, 18, and 24 hours. Fluorescence-activated cell sorter analysis was performed based on DNA content of 10 4 cells. (A) A representative analysis of a time-course experiment is shown. (B) Statistical analysis of cell cycle distribution. Data represent the mean SEM of 5 independent experiments. *Significantly different compared with vector-transfected controls (P 0.05). nificant extent and reached G 2 /M phase after 24 hours. In contrast, dominant-negative STAT3-expressing clones (representative results for F2 are shown) barely entered S phase after 18 hours with a statistically significant amount of cells remaining in the G 0 /G 1 phase at this point (Figure 7B), suggesting that functional inhibition of activated STAT3 results in a transient delay of G 1 /Sphase progression of the cell cycle. Inhibition of Activated STAT3 Selectively Inhibits CDK-2 Activity G 1 /S-phase transition is mainly regulated by CDK-4 and CDK-2 as well as their regulatory components. 36 To determine which of these proteins serve as targets for activated STAT3 and account for the observed delay of cell cycle progression, we performed kinase assays and immunoblot experiments in synchronized vector controls and dominant-negative STAT3 clones. In general, CAPAN-1 cells had very little CDK-4 kinase activity and we were unable to detect differences in CDK-4 content, cyclin D1 expression, or CDK-4 activity between STAT3-deficient cells and controls (Figure 8A). In sharp contrast, release of CAPAN-1 cells from synchronization resulted in a dramatic increase in CDK-2 kinase activity (Figure 8B). This induction became clearly visible after 12 hours, exactly when vector-transfected cells entered S phase (see Figure 7). In contrast, dominant-negative STAT3-expressing cells failed to induce a profound increase in CDK-2 activity (Figure 8B). Analyzing the various components of the CDK-2 complex, we were unable to detect consistent differences between vector and dominant-negative STAT3 clones for CDK-2, cyclin E, and cyclin A as well as p27 KIP1 or proliferating cell nuclear antigen. However, we consistently observed an increase in p21 WAF1 expression in clones with impaired STAT3 activity compared with vector-transfected controls (Figure 8B). This difference in p21 WAF1 expression was not yet apparent at 6 hours, when a robust induction of p21 occurred in both STAT3- competent and STAT3-deficient cells. Accordingly, the retinoblastoma-associated tumor suppressor protein (prb) was found hypophosphorylated at this time in both cell populations. Subsequently, however, the p21 content of control cells decreased, whereas prb phosphorylation increased and CDK-2 activation started (Figure 8B, upper rows). In STAT3-deficient cells, elevated p21 levels were maintained during this period, preventing CDK-2 activation (Figure 8B). To confirm that changes in p21 WAF1 expression translated into an altered p21 WAF1 complement of CDK-2 complexes, we performed CDK-2 immunoprecipitation experiments (Figure 9A). Indeed, we observed that p21 WAF1 was sequestered into the CDK-2 complexes and that the relative amount of p21 WAF1 was increased in the STAT3-deficient clones from 12 hours onward. This observation

10 900 SCHOLZ ET AL. GASTROENTEROLOGY Vol.125, No.3 putative target genes of STAT3 such as BCL-X L, cyclin D1, and c-myc, as has been described in other cell systems. Performing immunoblot analysis, we observed no relevant alterations of these proteins in the 3 dominantnegative STAT3 clones compared with wild-type or vector-transfected CAPAN-1 cells (Figure 9C). Inhibition of JAK2 Reproduces Functional Inhibition of STAT3 To further corroborate the results obtained by expression of dominant-negative STAT3, we explored an Figure 8. Inactivation of STAT3 inhibits CDK-2 activity in CAPAN-1 cells. (A and B) Vector-transfected controls (V) and dominant-negative STAT3-transfected F2 cells were synchronized in G 0 /G 1 phase by serum starvation for 24 hours. Cells were then released and whole cell lysates prepared at the indicated times. Cell cycle progression was controlled by simultaneous flow cytometry analysis (see Figure 7). Expression levels of cell cycle components were analyzed by immunoblotting using 50 g of protein lysate per lane and the respective specific antibodies. Representative results of 3 4 independent experiments are shown. For histone H1 and GST-Rb kinase assays, lysates were immunoprecipitated with CDK-2 or CDK-4 specific antibodies, respectively. After the kinase reaction was terminated, the whole reaction volume was subjected to SDS-PAGE and kinase activity determined by autoradiography of the dried gels. Each kinase experiment was independently repeated at least 3 times. suggested that a selective increase of p21 WAF1 expression was most likely responsible for the CDK-2 inhibition in response to functional inhibition of activated STAT3. The central role of p21 WAF1 as a determinant of STAT3-dependent growth regulation was further substantiated by the in vivo experiments described in Figure 6. All xenotransplanted CAPAN-1 tumors expressing a dominant-negative STAT3 showed elevated p21 WAF1 expression when compared with vector-transfected controls (Figure 9B). In contrast to the pronounced induction of p21 WAF1 expression, we were unable to show regulation of other Figure 9. Inhibition of activated STAT3 increases the p21 WAF1 content of CDK-2 complexes. (A) In analogy to the experimental protocol previously outlined, CDK-2 immunoprecipitation of vector control (V) and F2 cells was performed and sequentially immunoblotted with antibodies against p21 WAF1 and CDK-2. (B) Up-regulation of p21 WAF1 expression in vivo. Tumor tissue lysates derived from nude mice xenotransplanted with dominant-negative STAT3 (D4, F2, F5) and vector-transfected CAPAN-1 cells were subjected to immunoblotting using p21 WAF1 antibody. (C) Effect of dominant-negative STAT3 on expression levels of putative STAT3 target genes investigated in CAPAN-1 cells by immunoblot analysis of whole cell lysates using specific antibodies as indicated. Proliferating cell nuclear antigen content of CDK-2 complexes was not altered, suggesting that p21 WAF1 regulation was specific. Arrows indicate the molecular sizes of the detected bands.

11 September 2003 ACTIVATED STAT3 AND HUMAN PANCREATIC CANCER 901 independent experimental approach by blocking JAK2, which acts as an endogenous upstream activator of STAT3. 9 For these experiments, we used tyrphostin AG490, a specific inhibitor that has been reported to block JAK2 phosphorylation. 37 Treatment of CAPAN-1 cells with tyrphostin resulted in a dose-dependent inhibition of JAK2 activity, as shown by an anti phospho-tyrosine immunoblot from JAK2 immunoprecipitates (Figure 10A). Concentrations of 100 and 200 mol/l were sufficient for a substantial reduction of JAK2 tyrosine phosphorylation (Figure 10A). These concentrations also resulted in a significant inhibition of anchorage-dependent growth (Figure 10B). In analogy to the experiments performed with dominantnegative STAT3-expressing cells, we conducted cell cycle analysis in response to AG490. CAPAN-1 cells were synchronized in G 0 /G 1 phase, exposed to 100 mol/l AG490 supplemented in medium/15% FCS, and prepared for flow cytometry analysis after release from synchronization. Whereas controls exited G 0 /G 1 phase at 18 and 24 hours, cells treated with AG490 did not enter S phase and were retained in G 0 /G 1 phase of the cell cycle (Figure 10C). Immunoblot analyses of cells that had been treated with increasing concentrations of AG490 confirmed a dose-dependent inhibition of STAT3 activation that was paralleled by an increase in p21 WAF1 expression (Figure 10D). Thus, inhibition of STAT3 activity by blocking the upstream activator JAK2 also resulted in growth arrest due to a delay of G 1 /S-phase progression that is most likely mediated by an increase in p21 WAF1 expression. Discussion In the current study, we show constitutive activation of STAT3 in human pancreatic cancer cells compared with their nontransformed counterparts. The mechanisms responsible for STAT3 activation on malignant transformation in the human pancreas are currently unknown. Activation of STAT3 can be observed in response to a plethora of stimuli, including growth factors, cytokines, oncoproteins, activated cytoplasmic kinases, or mitogenic receptors. 11,38 Our observation that considerable STAT3 activation is absent in chronic pancreatitis suggests that events intimately linked to the transformation process are involved. Given that human pancreatic cancer cells overexpress EGF 39 and are characterized by the existence of a superagonistic autocrine mitogenic loop mediated by transforming growth factor, EGF, and the EGF-R, 40 we chose to further explore the role of EGF-R signaling in STAT3 activation. Using either an EGF-R specific antagonist or a neutralizing antibody, Figure 10. Effect of tyrphostin AG490 on CAPAN-1 cells. (A) Inhibition of JAK2 tyrosine phosphorylation. Cells were incubated with AG490 or vehicle (dimethyl sulfoxide) for 24 hours. A total of 2200 g of whole cell lysate was immunoprecipitated with a JAK2 antibody and immunoblotted using an antibody against p-tyrosine. To control for equal amounts of immunoprecipitates, blots were stripped and reprobed with JAK2 antibody. (B) Dose-dependent inhibition of anchorage-dependent growth. Triplicates of 10 4 cells were incubated with the indicated concentrations of AG490. After 96 hours, viable cells were counted manually with a hemocytometer. Data represent mean SEM of 3 independent experiments. *Significantly different from vehicle-treated control (P 0.05). (C) Cell cycle distribution. After synchronization as previously outlined, cells were exposed to 100 mol/l AG490 and submitted to flow cytometry analysis at the indicated times. A representative fluorescence-activated cell sorter analysis of 4 independent experiments, all yielding similar results, is shown. (D) Down-regulation of STAT3 activity and up-regulation of p21 WAF1 expression. CAPAN-1 cells were treated with the indicated concentrations of tyrphostin for 24 hours. Whole cell lysates (40 g) were analyzed by immunoblotting using specific antibodies as indicated. Proliferating cell nuclear antigen remained unchanged, suggesting that induction of p21 WAF1 was not due to unspecific changes in protein synthesis. Representative blots of 4 independent experiments are shown. we were able to suppress constitutive STAT3 DNA binding in human pancreatic cancer cells. Because these experiments were performed without exogenous addition

12 902 SCHOLZ ET AL. GASTROENTEROLOGY Vol.125, No.3 of growth factors, they clearly identify autocrine or paracrine EGF-R stimulation as a major upstream stimulus of STAT3 activity in pancreatic cancer cells. They are also consistent with previous observations that EGF-R mediated growth requires activation of STAT3. 21 Furthermore, although STAT3 DNA-binding activity was clearly EGF-R dependent under resting conditions, it could still be stimulated by addition of IL-6. Because elevated IL-6 serum levels have been reported in patients with pancreatic cancer 41 and IL-6 may be produced by human pancreatic cancer cells, 42,43 endocrine, autocrine, and/or paracrine IL-6 stimulation may also support STAT3 activation in the context of human pancreatic cancer disease. Furthermore, the cytosolic src family of tyrosine kinases has been shown to induce constitutive activation of STAT3 in multiple src transformed cell lines. 22,23 Overexpression and activation of src has been shown in 100% of pancreatic cancers investigated, 24 raising the possibility that the src family may also contribute to constitutive activation of STAT3 in pancreatic cancer cells. Apart from upstream activators, STAT3 activity is also regulated by SOCS family members, including SOCS-1 and SOCS Of these, SOCS-1 has been characterized as a tumor suppressor and was found frequently inactivated in hepatocellular carcinomas. 45 Thus, loss of SOCS-1 in pancreatic cancer could disrupt the physiologic negative feedback loop that limits STAT3 activation and thereby accounts for elevated STAT3 activity. However, SOCS-1 and SOCS-3 could both be readily detected in CAPAN-1 and BxPC-3 cells (Figure 3 and data not shown). Furthermore, their expression was down-regulated on STAT3 inhibition, suggesting that SOCS-mediated regulatory loops remained functional in pancreatic cancer cells. Although inhibition of STAT3 activation results in growth inhibition in a variety of tumor cell types, 15,46 growth inhibitory effects of activated STAT3 have also been described in other cell systems, 28,47 indicating cell and tissue context-specific biologic functions for activated STAT3. Therefore, 2 independent approaches were used to examine the biologic significance of persistently activated STAT3 in pancreatic cancer cells: expression of 2 different dominant-negative STAT3 constructs and pharmacologic inhibition of JAK2, which acts as a central endogenous upstream activator of STAT3. 7 In all experimental settings, we observed that functional inactivation of STAT3 resulted in significant inhibition of anchorage-dependent and -independent growth in vitro as well as tumor growth in vivo. In pancreatic cancer, growth regulatory phenomena need to be interpreted in light of the genetic background of the model under investigation. In this context, both tumor cell lines analyzed in this study express functionally inactivated tumor suppressors p16, p53, and DPC4 as well as an oncogenic (CAPAN-1) or wild-type K-ras (BxPC-3), thereby reflecting exactly the genetic alterations observed in most pancreatic cancers. 6 It is therefore remarkable that inhibition of STAT3 activation is still capable of overriding the combined mitogenic and antiapoptotic effects provided by these coexisting genetic alterations. These considerations therefore further support the concept that activated STAT3 provides a relevant contribution to the malignant phenotype of human pancreatic cancer. Depending on the cell type investigated, 2 mechanisms of action have been described for growth regulation by activated STAT3: cell cycle regulation and modulation of apoptosis. In some experimental models, activated STAT3 functions in an antiapoptotic manner; accordingly, inhibition of activated STAT3 results in induction of apoptosis. 51,52 This induction of apoptosis was observed to occur in a BCL-2/BCL-X L dependent or independent fashion. 51,53 In pancreatic cancer cells, inhibition of activated STAT3 did not alter the rate of spontaneous programmed cell death as examined by a variety of biochemical and morphologic criteria (data not shown). Furthermore, we were unable to detect an increased sensitivity for a variety of proapoptotic stimuli such as chemotherapeutic agents on functional inhibition of STAT3 (data not shown). Accordingly, expression of BCL-X L, a known target in tumor cells sensitive for STAT3-mediated apoptosis, 54 was not altered on STAT3 inactivation. In line with what has previously been described in other tissues, the capability of activated STAT3 to act antiapoptotic is therefore cell type dependent. 55 Several explanations for the lack of antiapoptotic action of STAT3 in pancreatic cancer cells seem feasible. (1) Oncogenic K-ras as well as inactivated p53, p16, and DPC4, which predominantly occur in most pancreatic cancers, have all been shown to act antiapoptotic in a variety of models (2) Pancreatic cancer cells overexpress nuclear factor B and prb, both of which protect the tumor cell against proapoptotic stimuli. 27,59,60 Therefore, although activated STAT3 is capable of stimulating proliferation, its functional inhibition cannot overcome the antiapoptotic protection provided by several coexisting genetic alterations in pancreatic cancer cells. Although STAT3 was unable to modulate apoptosis, it profoundly regulates cell cycle distribution. On inhi-