Activation of Stat3 by Cell Confluence Reveals Negative Regulation of Stat3

Activation of Stat3 by cell conﬂuence reveals negative regulation of Stat3 by cdk2

Richard A Steinmann,1, Abbey Wentzel2, Yalin Lu1, Christine Stehle1 and Jennifer Rubin Grandis2

1Department of Medicine, University of Pittsburgh School of Medicine and University of Pittsburgh Cancer Institute, USA; 2Department of Otolaryngology, University of Pittsburgh School of Medicine and University of Pittsburgh Cancer Institute, USA

The signal transducing protein Stat3 activates gene Stat3knockouts have underscored a critical role for transcription in cells in response to multiple cytokines. Stat3in the function of T cells, macrophages, keratino- Constitutive activation of Stat3 has been observed in solid cytes, and the mammary gland (Takeda and Akira, tumors including head and neck squamous cell carcinoma. 2000). Stat3 activation in cancer has been associated with The control of Stat3activity has been shown to be autocrine stimulatory loops and is believed to convey a important in several biological settings. Stat3activation growth advantage to cells. We now demonstrate ligand- has been linked to both growth arrest and to aberrant independent activation of Stat3 by high cell density in cell growth, to apoptosis and antiapoptosis, and to cell multiple cancer cell lines. Activation of Stat3 is associated migration (Akira, 2000). During hematopoiesis, Stat3is with antiproliferative rather than proliferative conditions. required for myeloid differentiation (Steinman and Iro, Interference with cdk2 activity upregulates Stat3 phos- 1999; McLemore et al., 2001) and facilitates differentia- phorylation and Stat3-directed DNA-binding activity. tion-coupled growth arrest by inducing the transcription Our data supports a model in which Stat3 activity is of the cell cycle inhibitor p27 (de Koning et al., 2000). In partially suppressed by cdk2 in growing cells and contrast, several leukemogenic oncogenes have been derepressed upon cell confluence. found to promote constitutive Stat3activation in cell Oncogene (2003) 22, 3608–3615. doi:10.1038/sj.onc.1206523 line models and in leukemic blasts. Overexpression and/ or heightened activation of Stat3is a prominent feature Keywords: Stat3(signal transducer and activator of of cancers in addition to leukemias. Fibroblasts transcription-2) cdk2; confluence; cyclin-dependent engineered to express a dominant-active Stat3demon- kinase; protein transduction strated anchorage independence and formed tumors in nude mice, thus establishing Stat3as an oncogene (Bromberg et al., 1999). We have similarly found that xenografts of head and neck cancer cells constitutively expressing Stat3C form more aggressive and larger Introduction tumors (Kijima et al., 2002). This may result from an antiapoptotic function of Stat3, which has been STAT proteins are a family of transcriptional activators, observed in many (but not all) Chapman et al., 2000) which have been shown to modulate gene expression in cell types. We have previously shown that constitutive response to multiple cytokines (for review see Bromberg Stat3signaling inhibits apoptosis in a model of and Darnell, 2000; Jove, 2000 #803). STAT proteins can squamous cell carcinoma (Grandis et al., 2000a), and be activated by phosphorylation either by JAK kinases that growth of head and neck cancer cells is inhibited by that are associated with active cytokine receptors, or by dominant-negative Stat3(Grandis et al., 1998a). In nonreceptor tyrosine kinases. The essential role of squamous cell carcinoma, breast cancer, and prostate STAT proteins in mediating cytokine signaling is cancer, high levels of constitutive Stat3activation are manifested by the phenotype of STAT knockout mice, thought to arise from autocrine stimulatory routes which parallel defects that arise from cytokine receptor (Song and Grandis, 2000; Berclaz et al., 2001; Giri or cytokine knockouts (for review, see Takeda and et al., 2001; Li and Shaw, 2002). Akira, 2000). Unlike other STAT knockouts, the loss of While Stat3activation by soluble cytokines has been Stat3results in embryonic lethality, indicative of a well described, there is little known about modulation of nonredundant requirement for this factor. Conditional the Stat3signaling pathway by cell–cell interactions, and particularly by cellular signals associated with cell confluence. In this work, we report ligand-independent *Correspondence: RA Steinman, Suite 2. 18 UPCI Research Pavilion, activation of Stat3by cell confluence. Activation results Hillman Cancer Center, 5117 Centre Avenue, Pittsburgh, PA 15213, in part from inhibition of cyclin-dependent kinase-2, USA; E-mail: [email protected] Received 9 September 2002; revised 24 February 2003; accepted 24 suggesting the existence of an ‘inside-out’ signaling February 2003 pathway involving cell cycle inhibitors and Stat3. Stat3 activation by cell density and cdk2 inhibition RA Steinman et al 3609 Results contact inhibition of growth. The density-dependent growth of several SCCHN cell lines was therefore Upregulation of Stat3 DNA-binding activity with cell characterized, as well as their ability to induce the confluence cyclin-dependent kinase inhibitors (cdki’s) p21 and p27 upon cell confluence. We studied the relation of Stat3signaling and cell Figure 1a demonstrates that SCCHN cell lines UM- confluence in squamous cell carcinoma of the head and 22B (22b) and PCI-37A (37a) continued to grow for neck (SCCHN) cell lines, in which we had shown that several days after attaining visual cell confluence, Stat3was required for cell growth (Grandis et al., 1998a, whereas the SCCHN cell line PCI-15B (15b) exhibited 2000a). In these cell lines, Stat3activation was believed confluence-dependent growth arrest. Analysis of cell to result from an autocrine loop involving EGFR-TGF- cycle modulator expression showed that the cdki’s p21 alpha (Grandis and Tweardy, 1993; Grandis et al., and p27 were upregulated in confluent UM-22B and 1998b), which was necessary for the growth of SCCHN PCI-37A cells and further upregulated in postconfluent cells in vitro and in vivo (He et al., 1998; Endo et al., cells harvested 2 days after visual confluence (Figure 1b). 2000). We hypothesized that autocrine induction of Interestingly, PCI-15B cells, which grew most slowly, Stat3would facilitate the ability of these cells to escape did not significantly alter cdki expression at confluence.

Figure 1 Stat3is activated following cell confluence and growth arrest. ( a) Growth curves of SCCHN cell lines. Cell line PCI-15B arrests at visual confluence (day 4, arrow), whereas lines UM-22B and PCI-37A continue to grow after cell-cell contact and arrest at higher density. (b) Expression of Stat3and of cell cycle modulators p21 and p27 in cell lines at 50%density (subconfluence), at confluence, and 3days after attaining confluence (postconfluence). ( c) Gel shift demonstrating Stat binding to the SIE DNA sequence. Stat3: 3 homodimer (A), Stat3: 1 heterodimer (B), and Stat 1 : 1 homodimer (C) bands are indicated. An increase in Stat3 DNA binding activity between low (L, 50%confluence) and high (H, postconfluent) density cultures is shown for three SCCHN cell lines. (d) Association of Stat3activation with Rb hypophosphoryation. 22b cells were harvested at indicated times after seeding. ( e) Cell cycle profile of UM-22B cells at subconfluence (50%) and postconfluent (4100%) density. Cells were doubly stained with BrDU (vertical axis) and 7AAD (horizontal axis) and cell cycle profile as shown was calculated. (f) Supernatant swapping experiment. Representative gel shift results following harvest of low density UM-22B cells exposed to low density (L) or high density (H) supernatants for 8 h. High-density (postconfluent) cultures were similarly exposed either to supernatant from low- or high-density cultures

Oncogene Stat3 activation by cell density and cdk2 inhibition RA Steinman et al 3610 The expression of Stat3protein was constant at Confluence-mediated Stat3 activation in SCCHN cells is all cell densities. Surprisingly, Stat3activation increased EGFR-independent dramatically for each cell line in proportion to cell density (Figure 1c). This was not a function of cell The inability of supernatant swapping to recapitulate culture duration, because the same effect was seen Stat3activation associated with confluence suggested whether cells were allowed to grow to high density, or if that this phenomenon was ligand-independent. In order cells were split and seeded in equivalent numbers to further evaluate this possibility, we determined into a small (confluent) and large (subconfluent) whether the EGFR-specific inhibitor PD153035 was dish and harvested concurrently (data not shown). capable of blocking density-dependent Stat3activation Stat3was activated at high cell density both in cells in UM-22B SCCHN cells. EGFR activation is the that grew past confluence (UM-22B, PCI-37A) and primary mechanism for ligand-mediated Stat3activa- et al cells that did not (PCI-15B), indicating that acti- tion in these cells (Grandis ., 2000b). vated Stat3was not sufficient to overcome contact Results are shown in Figure 3a. TGF-alpha was inhibition of growth. As shown in Figure 1e, even added in the absence of serum in order to minimize basal though UM-22B cells continued to accumulate levels of Stat3activation and optimize responsiveness to postconfluence, their rate of cell cycling slowed. exogenous TGF-alpha. Figure 3a demonstrates that In fact, Stat3was activated concurrently with cell under serum-starved conditions, Stat3activation was cycle inhibition, as manifested by Rb hypophosphoryla- lower than in the presence of serum. There was a tion (Figure 1d). Stat3activation could not be marked increase in Stat3activation when cells were accounted for by changes in the level of Stat3protein confluent. Confluence-activated Stat3was not inhibited (Figure 1b, d). by the EGFR inhibitor PD153035 at a concentration at We considered that Stat3activation at high cell which it effectively blocked Stat3activation by high confluence could reflect disproportionate expression of a levels of exogenous TGF-alpha. This finding further stimulatory cytokine in confluent cell cultures. There- supported the conclusion that confluence increased fore, supernatant swap experiments were conducted in Stat3through a ligand-independent mechanism. It is which supernatants from subconfluent cells were applied notable that whereas confluence upregulated Stat3, to confluent cells, and confluent cell supernatants were other antiproliferative conditions (e.g. serum starvation) applied to subconfluent cells for 8 h. As shown in did not do so. Figure 1f, the level of Stat3activation was not affected The lack of a requirement for EGFR signaling for by switching of cell supernatants, indicating that it did Stat3upregulation was confirmed by measuring Stat3 not result from differential expression of a soluble activity in EGFR-knockout mouse fibroblasts grown to factor. different densities. As shown in Figure 3c, Stat3 activity Stat3activation by cell confluence was not confined to also increased with density in these knockout cells. SCCHN cells. Figure 2 demonstrates that substantial Receptor-mediated activation of Stat3has been upregulation of Stat3DNA binding occurred with associated with the formation of Stat3–Rac1 complexes; confluence in transformed cell lines of bladder, kidney, for instance, IL-6 activation of Stat3appears to require et al et al breast, and fibroblast origin. Rac1 (Simon ., 2000; Faruqi ., 2001). Figure 3b shows immunoprecipitation experiments in UM-22B cells that demonstrated that the amount of tyrosine 705- phosphoStat3in complex with Rac1 did not change with cell density. Confluence-mediated Stat3activation therefore did not depend on altered interactions with Rac1. It is notable that increased cell density was associated with an increase in phosphotyrosine-Stat3which immunoprecipitated with antibody directed against Stat3. This indicates that there is an increase in Stat3phosphoryla- tion with increased cell density (see also Figures 4c, 5b).

Stat3 phosphorylation is induced by cdk2 inhibition As shown in Figure 1, the level of cdk inhibitors rose as SCCHN cells became confluent and continued to rise as they were postconfluent. Analysis of cdk2-kinase activity in UM-22b and in PCI-15B cells demonstrated that cdk2-kinase activity decreased concurrently with an increase in Stat3activity (Figure 4a). Since Stat3 activation did not precede the decrease in cdk2 kinase Figure 2 Activation of Stat3by increased cell density in multiple activity, we considered that changes in Stat3activity cell lines. Gel shift assay for Stat3activity is shown. MDCK type I might occur downstream of cdk2 inhibition. In order to or II cells were seeded at low (L), intermediate or high (H) cell densities. Other cell lines were harvested at subconfluence (L) or test whether cdk2 inhibition caused a feedback increase postconfluent (H). The gel shift band in the ZRP cell assay was in Stat3activity, UM-22B cells were incubated with the supershifted by anti-Stat3antibody, as shown cdk2 inhibitor roscovitine or were transduced with

Oncogene Stat3 activation by cell density and cdk2 inhibition RA Steinman et al 3611

Figure 3 (a) EGFR independence of Stat3activation at confluence. SCCHN cells (37a)were maintained in low serum at low (L) or high (H) cell density. A marked increase in Stat activity upon confluence was evident. Activity was further increased by the addition of exogenous EGFR ligand TGF-alpha. However, while an EGFR inhibitor could negate TGF-alpha upregulation of Stat3activity, it could not decrease the confluence-mediated activation. The first two lanes show levels of Stat3activity in 37acells maintained in 10%serum-containing medium at low or high confluence. (b) Equivalent association of phospho-Stat3with Rac1 at low and high cell densities. SCCHN cells at low (L) or high (H) density were harvested, immunoprecipitated with antibody as indicated and blotted for phosphoTyr705-Stat3. A strong heavy chain band is also evident. (c) Upregulation of Stat3in EGFR-knockout fibroblasts with cell density. Supershifted Stat3bands from gelshifts performed on extracts harvested at days 1–6 of growth are shown. Cells attained visual confluence on day 4 dominant-negative cdk2 protein (TAT-cdk2DN) or a trypsin lowered phospho-Stat3levels and activated cdk2 transducible GFP control protein (TAT-GFP). kinase. Figure 4b demonstrates that inhibition of cdk2 specifi- cally upregulated Stat3activity. Similar results were observed with NIH-3T3 cells (data not shown). Stat3 activation by cdk2 inhibitors and by cell confluence Discussion occurred at the level of Stat3phosphorylation on tyrosine 705 (Figure 4c). The present investigation has identified ligand-independent upregulation of Stat3phosphorylation and Stat3 Stat3 activation and cell–cell interactions DNA-binding activity by cell confluence in numerous cell lines. Stat3activation is density-dependent and We considered it likely that cell–cell interactions continues to increase following cell confluence. Stat3 occurring at high cell density contributed to Stat3 activation coincided with a decrease in Rb phosphoryla- activation. In order to examine this in more detail, tion and with downmodulation of cdk2-kinase activity. experiments were conducted using MDCK cells, which Dominant-negative cdk2 increased Stat3tyrosine phos- grow in a tight monolayer which can be disrupted with phorylation, indicating that cdk2 inhibition could be brief exposure to trypsin. Cells that were 48 h postcon- contributing to confluence- dependent increases in Stat3 fluence were exposed to PBS for 10 min or to trypsin to activity. disrupt cell–cell contacts. Trypsin-treated cells contin- The finding of increased Stat3activity in confluent ued to adhere to the flask but assumed a rounded head and neck cancer cells compared with faster- morphology (Figure 5a). Cells were then re-exposed to growing subconfluent cells was somewhat surprising, growth medium for 30 min, following which adherent given the association of Stat3signaling with oncogenesis cells were harvested for gel shift analysis. As shown in in these cells (Grandis et al., 1998a). Since activated Figure 5a, exposure to trypsin decreased Stat3binding Stat3is insufficient to sustain SCCHN cell proliferation to DNA. This was associated with decreased Stat3 at high density (Figure 1), the oncogenicity of Stat3 phosphorylation (Figure 5b). Based on our data more likely results from its antiapoptotic function suggesting an inverse relation between cdk2 activity (Grandis et al., 2000a) than by augmenting prolifera- and Stat3activation, we measured cdk2 kinase activity tion. in low- and high-density MDCK cells and in trypsin- Our evidence indicated that confluence-mediated disrupted cells. As shown in Figure 5b, high-density cells Stat3activation was ligand-independent. Stat3was not contain higher levels of phospho-Stat3and low cdk2 activated by incubation of subconfluent cells with kinase activity. Disruption of cell–cell contacts with confluent cell supernatent, nor was Stat3suppressed in

Oncogene Stat3 activation by cell density and cdk2 inhibition RA Steinman et al 3612

Figure 5 (a) Requirement for mature cell–cell contacts for Stat3 upregulation. Confluent MDCK type II cells (lower left) were incubated with PBS (H-PBS) or with trypsin/EDTA (H-trypsin) for 10 min which caused loss of cell–cell contacts. They were then reincubated for 30 min in media and harvested for gel shift assay. The Stat3: DNA band was confirmed by supershift of PBS-treated extract as indicated. Photographs of adherent cells immediately after PBS or trypsin treatment are shown. Typical result of three experiments is shown. (b) Stat3activation and cdk2-kinase activity. Cells were harvested following growth at low (L) or high (H) densities or after treatment with PBS or trypsin as in panel a. Equivalent micrograms of cell extract from each sample were immunoprecipitated with IgG or Stat3antibody as shown and immunoblotted with phosphotyrosine-Stat3specific antibody. An equivalent aliquot of each extract was also immunoprecipitated with anti-cdk2-antibody for determination of cdk2 kinase activity as shown Figure 4 (a) Inverse relation of cdk2-kinase and Stat3activities. UM-22B and PCI-15B cells were grown for up to 10 days as indicated. Cells were confluent on day 4. Cdk2 activity was measured for each timepoint; a parallel sample was harvested for et al., 2001), we have found focal adhesion kinase Stat3. activity. Stat-3 antibody supershifted bands on gelshift are (FAK) to exist in complex with phospho-Stat1 in cells shown. (b) Direct transduction of dominant-negative cdk2 protein used in this study (data not shown), but have no activates Stat3. UM-22B cells were exposed overnight to 150 nm cell-permeable DN-cdk2 (TAT-cdk2DN) or to GFP control (TAT- evidence of FAK–Stat3complex formation at high or GFP) and Stat3activity determined on gel shift assay. Activation low cell densities. of Stat3by 12 mg/ml roscovitine is also evident. Two independent The association of Stat3activation with cell density experiments are shown. (c) Increased phosphorylation of Stat3by and decreased growth led us to speculate that there dominant-negative cdk2 and by confluence. Western blot of 50 mg might be a feedback mechanism between cell cycle protein extract is shown. Cells were exposed to media alone or to TAT fusion proteins as indicated overnight at indicated cell regulators and Stat3. We noted decreased activity of densities and extracts were blotted with Try705-specific anti-Stat3 cyclin-dependent kinase-2 at high cell confluence, when antibody. Ponceau staining confirmed equivalent protein loading in Stat3is activated. Direct dominant-negative protein each lane transduction and cdk2-inhibitor experiments indicated that cyclin-dependent kinase-2 suppresses Stat3phos- phorylation. Since cdk2 activity affects Stat3phosphor- confluent SCCHN cells by EGFR inhibitors. Although ylation, it most likely acts at a proximal point in the these manipulations would not prevent Stat3activation transduction pathway. In rat hepatocytes, Cdk2 has by an internal autocrine loop, it is unlikely that such an been reported to modulate receptor activation at the autocrine mechanism would be coupled to cell density. plasma membrane by regulating endocytosis (Gaulin Autocrine activation of Stat3and confluence-mediated et al., 2000). We have shown that the cdk2-inhibitor p27 activation may represent independent pathways whose also may interact with signaling complexes by localizing relative importance varies with the cellular environment. to lipid rafts (Yaroslavskiy et al., 2001) Such localiza- Cell–cell adherence facilitates Stat3activation in con- tion could facilitate ‘inside-out’ signaling of cell cycle fluent cells, as manifested by the ability of brief trypsin regulators to alter signal transduction at the membrane. exposure to downmodulate Stat3activation in MDCK Indeed, recent reports that Stat3localizes to rafts cells. The upregulation of Stat3with cell density may (Sehgal et al., 2002) and that stat3undergoes endocy- not be universal – we have become aware that in Caco-2 tosis during signaling (Bild et al., 2002) indicate that cells Stat3activation increases with confluence and Stat3may share topographic proximity with cdk2 and decreases postconfluence (Wang and Evers, 1999). cdk2 inhibitors at the cell membrane. Our finding of Additional investigation will determine which cell–cell Stat3upregulation by cdk2 inhibitors is consistent with signaling pathways are involved in Stat3activation at the report that antisense p27Kip1 decreased Stat3 high cell density. In keeping with other reports (Xie activity in keratinocytes (Hauser et al., 1998).

Oncogene Stat3 activation by cell density and cdk2 inhibition RA Steinman et al 3613 Inhibition of cdk2 augmented Stat3activation in transfection of constitutively active Stat3constructs subconfluent SCCHN cells. This indicates that Stat3was (Shen et al., 2001). Protein interactions involved in the far from maximally stimulated in growing cells despite cell–cell activation of Stat3may therefore constitute a the presence of a TGF-alpha/EGFR autocrine loop. reasonable therapeutic target. Our data support a model in which Stat3activity is Other biologic processes may be affected by con- partially suppressed by cdk2 in growing cells and is fluence-associated changes in Stat3activity. For in- derepressed upon cell confluence. Even in cells that have stance, increased Stat3signaling may modulate attained visual confluence and concommitant Stat3 expression and/or activity of integrins involved in cell activation, inhibition of residual cdk2 activity further adherence (Wooten et al., 2000). To the extent that a augments Stat3activation. On the other hand, growth cell-density-associated pathway of Stat3activation is arrest through serum starvation did not suffice to recruited at distinct stages in development, this pathway activate Stat3, indicating that a serum component as could also contribute to the critical role for Stat3in well as low cdk2 activity is required for optimal Stat3 embryogenesis. activation. The mechanism through which cdk2 suppresses Stat3 phosphorylation is unknown. We have not been able to Materials and methods demonstrate a direct interaction between cdk2 and Stat3. It is conceivable that cdk2 acts in tandem with Cell lines known negative regulators of Stat3. These include Head and neck squamous cell carcinoma cell lines PCI-37A, suppressor of cytokine signaling-3) (SOCS3) (Suzuki UM-22B and PCI- 15B, NIH-3T3 cells, MDCK type II kidney et al., 2001; for review Krebs and Hilton, 2001; cells, ZRP breast cancer cells, and TSU bladder cancer cells Yasukawa et al., 2000), protein inhibitor of activated were grown in DMEM medium (Mediatech, Herndon, VA, Stat3(PIAS) (Junicho et al., 2000), cyclin D1 (Bienvenu USA) containing 10% fetal bovine serum (FBS, Life et al., 2001), and the cell cycle modulator p21 (Coqueret Technologies, Frederick, MD, USA), 100 mg/ml penicillin, and Gascan, 2000). Of these, only SOCS3has been and streptomycin. Medium was supplemented with 10 nm B- noted to inhibit Stat3phosphorylation; the other estradiol for ZRP cells. MDCK type I cells were cultured in inhibitors either inhibit nuclear translocation of phos- DMEM/F12 medium (ATCC, Manassus, VA, USA) with phorylated Stat3or inhibit transcriptional activation by 10% FBS. All cells were split to sustain log-phase growth and DNA-bound Stat3. We therefore considered whether new vials were thawed every 2 months. cdk2 might suppress Stat3by increasing the activity of SOCS3. We noted a potential cdk2 phosphorylation site Immunoprecipitations (SPPR; Kitagawa et al., 1996) and cyclin-interacting Cells were washed with PBS and lysed in cold immunopreci- sites (RPL; Kitagawa et al., 1996) in the SOCS3 pitation buffer (50 mm Tris, pH 7.5, 150 mm NaCl, 50 mm sequence. However, we could not demonstrate changes NaF, 1 mm NaVO3, 1 mm DTT, 0.5%NP-40, 0.1% protease in SOCS3protein levels with confluence, nor association inhibitor, 0.1% phosphatase inhibitor, 5 mg/ml PMSF, 1 mg/ml of SOCS3with cdk2 nor confluence-dependent changes leupeptin, from Sigma, St. Louis, MO, USA). Precleared lysate in SOCS3phosphorylation (data not shown). (100 mg) in 1 ml IP buffer was incubated overnight with 1 mgof Stat3upregulation following cell confluence may rabbit IgG (Santa Cruz, Santa Cruz, CA, USA), 1 mgof polyclonal anti-Stat3(Cell Signaling Technology, Beverly, contribute to the decreased growth rate of these cells. MA, USA) or 1 mg of anti- Rac1 (sc271AC, Santa Cruz) and Among Stat3target genes, some are antiproliferative nutated at 41 overnight. Protein A/G Plus (40 ml) beads (Santa including p27 (de Koning et al., 2000) and p21 (Sinibaldi Cruz) were added, and after an additional 2 h of mixing at 41, et al., 2000). Suppression of Stat3activation in growing beads were pelleted followed by five washes in IP buffer, beads cells may help to suppress such antiproliferative gene were boiled in Laemmli buffer, protein fractionated on SDS– expression. On the other hand, it is possible that the PAGE, transferred to nitrocellulose, and immunoblotted with increase in Stat3activation represents a feedback phospho-tyrosine 705-specific anti-Stat antibody (Cell Signal- mechanism to oppose the pathways limiting the growth ing) as per the manufacturer ’s recommendations. of confluent cells. Such a premise would raise the question of why Stat3would be insufficient to promote Western blotting proliferation. It is notable that levels of antiproliferative Western blotting was conducted using standard techniques and Stat3targets (e.g. p27) are upregulated following cell the following antibodies, listed by target: p21 (sc397-HRP, confluence, whereas Stat3-responsive proliferative genes Santa Cruz); p27 (p27-HRPO, BD Pharmingen, San Diego, (e.g. myc, cyclin D1) are not (data not shown). Whether CA, USA); Rb (14001A, BD Pharmingen), phosphoTyr705- these target gene expression patterns reflect selective Stat3(Cell Signaling). Stat3binding to target promoter sequences as a function of cell confluence is under study. Kinase assays The enhanced oncogenicity of Stat3-expressing cells MDCK lysates immunoprecipitated with anti-cdk2 (Labvi- may arise from density-dependent increase in Stat3 sion, Freemont, CA, USA, Ab-3) were washed with IP buffer activation. Cancer cell lines have been shown to exhibit and then with kinase buffer (50 mm HEPES, pH 7. 5, 10 mm increased resistance to apoptosis upon cell confluence MgCl2,5mm MnCl2,1mm DTT). Samples were resuspended I (Garrido et al., 1997; Dimanche-Boitrel et al., 1998) in 10 ml kinase buffer with 10 ml of ATP mix (20 mm ATP, 10 mg Such apoptosis resistance has been mimicked by histone, 10 mCi 34P-gamma-ATP) and incubated for 30 min at

Oncogene Stat3 activation by cell density and cdk2 inhibition RA Steinman et al 3614 301C. Proteins were fractionated on SDS–PAGE and exposed TAT protein production to film. TAT-cdk2DN and TAT-gfp plasmids were a generous gift from Steven Dowdy. Fusion proteins were essentially prepared Growth curves and cell cycle determination as described in Nagahara et al. (1998), with the exception that PCI-37A, UM-22B, and PCI-15B cells were seeded at fusion proteins purified on nickel columns were directly 3 Â 104 cells/ml in 24-well plates. For each cell line, cells from desalted on PD10 columns. Intracellular transport was duplicate wells were counted daily for 11 days using trypan verified. Fusion proteins were applied at 150 nm final blue vital dye staining. Log of cell number (normalized to day concentration, reapplied 8 h later, and cells were harvested 1 cell number) was calculated. Following a 3-h BrDU pulse, 16 h after initial addition of fusion proteins. the cell cycle profile was determined using dual BrdU/7AAD staining (BrDU Flow Kit, BD Plarmingen, San Diego, CA, Gel shift assays USA) as per the manufacturer ’s instructions. Analysis was performed on Coulter XL flow cytometer (Beckman-Coulter, Gel shift assays were conducted using 15 mg of whole-cell Miami, FL, USA). extract in each reaction and radiolabeled hSIE duplex oligonucleotide (GATCCATTTCCCGTAAATC) as pre- TGF-alpha stimulation and EGFR inhibition viously described (Steinman and Iro, 1999). As indicated, Stat3: DNA complexes were supershifted by 30min preincu- PCI-37A cells were plated at low and high densities in DMEM bation with 4 mg anti-Stat3antibody (Santa Cruz, sc482X). containing 12%FBS. Cells were simultaneously changed to serum-free media when ‘low density’ cells were 50% confluent and ‘high density’ cells were 48 h post-100% confluence. Acknowledgements Following 24 h of serum starvation, cells were treated with We thank John Marshall for ZRP cells, Ora Weiss, and 100 nm of EGFR-specific tyrosine kinase inhibitor PD153035 Thomas Kleyman for MDCK cells, James James Johnston for (Parke-Davis Pharmaceuticals, Ann Arbor, MI, USA) for 2 h; 3T3-SOCS cells, Schlomo Melmed for a SOCS-luciferase or with 100 nm PD153035 for 2 h in addition to 30 ng/ml TGF- construct, Peter Nissley for a GST-SOCS plasmid, Sarah alpha (Oncogene Research Products, Boston) during the final Dunn and Martin Myers for a SOCS3-expression plasmid, 30 min; or with 30 ng/ml TGF-alpha for 30 min. All treatments Jacqueline Bromberg for a Stat3C expression plasmid, and were in serum-free DMEM. Control cells received fresh serum- Steven Dowdy for GFPTat and dominant-negative cdk2Tat free DMEM alone. Cells were harvested by scraping and plasmids. This work was supported in part by NIH Grants whole-cell extracts were prepared. EMSA was performed with RolHL65172 (RAS), RO1CA77308 (JRG), and an UPCI pilot radiolabeled hSIE duplex oligonucleotide. grant (RAS).

References

Akira S. (2000). Oncogene, 19, 2607–2611. Gaulin JF, Fiset A, Fortier S and Faure RL. (2000). J. Biol. Berclaz G, Altermatt HJ, Rohrbach V, Siragusa A, Dreher E Chem., 275, 16658–16665. and Smith PD. (2001). Int. J. Oncol., 19, 1155–1160. Giri D, Ozen M and Ittmann M. (2001). Am. J. Pathol., 159, Bienvenu F, Gascan H and Coqueret O. (2001). J. Biol. Chem., 2159–2165. 276, 16840–16847. Grandis JR, Drenning SD, Chakraborty A, Zhou MY, Zeng Bild AH, Turkson J and Jove R. 2002. EMBO J., 21, Q, Pitt AS and Tweardy DJ. (1998a). J. Clin. Invest., 102, 3255–3263. 1385–1392. Bromberg J and Darnell Jr JE. (2000). Oncogene, 19, Grandis JR, Drenning SD, Zeng Q, Watkins SC, Melhem MF, 2468–2473. Endo S, Johnson DE, Huang L, He Y and Kim JD. (2000a). Bromberg JF, Wrzeszczynska MH, Devgan G, Zhao Y, Proc. Natl. Acad. Sci. USA, 97, 4227–4232. Pestell RG, Albanese C and Darnell Jr JE. (1999). Cell, 98, Grandis JR, Melhem MF, Gooding WE, Day R, Holst VA, 295–303. Wagener MM, Drenning SD and Tweardy DJ. (1998b). J. Chapman RS, Lourenco P, Tonner E, Flint D, Selbert S, Natl. Cancer Inst., 90, 824–832. Takeda K, Akira S, Clarke AR and Watson CJ. (2000). Adv. Grandis JR and Tweardy DJ. (1993). Cancer Res., 53, 3579– Exp. Med. Biol., 480, 129–138. 3584. Coqueret O and Gascan H. (2000). J. Biol. Chem., 275, Hauser PJ, Agrawal D, Hackney J and Pledger WJ. (1998). 18794–18800. Cell Growth Differ., 9, 847–855. de Koning JP, Soede-Bobok AA, Ward AC, Schelen AM, He Y, Zeng Q, Drenning SD, Melhem MF, Tweardy DJ, Antonissen C, van Leeuwen D, Lowenberg B and Touw IP. Huang L and Grandis JR. (1998). J. Natl. Cancer Inst., 90, (2000). Oncogene, 19, 3290–3298. 1080–1087. Dimanche-Boitrel MT, Micheau O, Hammann A, Haugg M, Jove R. (2000). Oncogene, 19, 2466–2467. Eymin B, Chauffert B and Solary E. (1998). Int. J. Cancer, Junicho A, Matsuda T, Yamamoto T, Kishi H, Korkmaz K, 77, 796–802. Saatcioglu F, Fuse H and Muraguchi A. (2000). Biochem. Endo S, Zeng Q, Burke NA, He Y, Melhem MF, Watkins SF, Biophys. Res. Commun., 278, 9–13. Lango MN, Drenning SD, Huang L and Rubin Grandis J. Kijima T, Niwa H, Steinman RA, Drenning SD, Gooding (2000). Gene Therapy, 7, 1906–1914. WE, Wentzel AL, Xi S and Grandis JR. (2002). Cell Growth Faruqi TR, Gomez D, Bustelo XR, Bar-Sagi D and Reich NC. Differ., 13, 355–362. (2001). Proc. Natl. Acad. Sci. USA, 98, 9014–9019. Kitagawa M, Higashi H, Jung HK, Suzuki-Takahashi I, Ikeda Garrido C, Ottavi P, Fromentin A, Hammann A, Arrigo AP, M, Tamai K, Kato J, Segawa K, Yoshida E, Nishimura S Chauffert B and Mehlen P. (1997). Cancer Res., 57, and Taya Y. (1996). EMBO J., 15, 7060–7069. 2661–2667. Krebs DL and Hilton DJ. (2001). Stem Cells, 19, 378–387.

Oncogene Stat3 activation by cell density and cdk2 inhibition RA Steinman et al 3615 Li L and Shaw PE. (2002). J. Biol. Chem., 21, 21. Suzuki A, Hanada T, Mitsuyama K, Yoshida T, Kamizono S, McLemore ML, Grewal S, Liu F, Archambault A, Hoshino T, Kubo M, Yamashita A, Okabe M, Takeda K, Poursine-Laurent J, Haug J and Link DC. (2001). Immunity, Akira S, Matsumoto S, Toyonaga A, Sata M and 14, 193–204. Yoshimura A. (2001). J. Exp. Med., 193, 471–481. Nagahara H, Vocero-Akbani AM, Snyder EL, Ho A, Latham Takeda K and Akira S. (2000). Cytokine Growth Factor Rev., DG, Lissy NA, Becker-Hapak M, Ezhevsky SA and Dowdy 11, 199–207. SF. (1998). Nat. Med., 4, 1449–1452. Wang S and Evers BM. (1999). J. Gastrointest. Surg., 3, Sehgal PB, Guo GG, Shah M, Kumar V and Patel K. (2002). 200–207. J. Biol. Chem., 277, 12067–12074. Wooten DK, Xie X, Bartos D, Busche RA, Longmore GD and Shen Y, Devgan G, Darnell Jr JE and Bromberg JF. (2001). Watowich SS. (2000). J. Biol. Chem., 275, 26566–26575. Proc. Natl. Acad. Sci. USA, 98, 1543–1548. Xie B, Zhao J, Kitagawa M, Durbin J, Madri JA, Simon AR, Vikis HG, Stewart S, Fanburg BL, Cochran BH Guan JL and Fu XY. (2001). J. Biol. Chem., 276, and Guan KL. (2000). Science, 290, 144–147. 19512–19523. Sinibaldi D, Wharton W, Turkson J, Bowman T, Pledger WJ Yaroslavskiy BB, Stolz DB, Watkins SC, Alber SM, Bradbury and Jove R. (2000). Oncogene, 19, 5419–5427. NA and Steinman RA. (2001). Mol. Med., 7, 49–58. Song JI and Grandis JR. (2000). Oncogene, 19, 2489–2495. Yasukawa H, Sasaki A and Yoshimura A. (2000). Annu. Rev. Steinman RA and Iro A. (1999). Leukemia, 13, 54–61. Immunol., 18, 143–164.

Oncogene