Reciprocal Regulation of c-Src and STAT3 in Non-Small Cell Lung Cancer

Lung cancer accounts for 29 of all cancer deaths in the United States, with a 5-year overall survival rate of 15 for all stages (1). Although chemotherapy remains the standard treatment for advanced or metastatic nonsmall cell lung cancer (NSCLC), response rates do not exceed 35 with frontline therapies and are even lower in the second-line setting (2). Improving our understanding of the signaling pathways that drive tumor behavior is essential for improving clinical outcomes. One potential therapeutic target in NSCLC for which clinical inhibitors have been developed is cellular Src (c-Src; ref. 3). The Src family consists of nonreceptor tyrosine kinases involved in signal transduction in both normal and cancer cells (4). c-Src is the best characterized and most often involved in cancer progression. c-Src overexpression has been shown in multiple tumor types, in which its activation correlates with shorter survival (reviewed in ref. 3). In NSCLC, c-Src is expressed and activated in both adenocarcinomas and squamous cell carcinomas (5, 6).

c-Src participates in several normal cellular functions during development and adulthood, including cell cycle progression, immune recognition, adhesion, spreading, migration, apoptosis regulation, and differentiation (reviewed in refs. 3, 7). In cancer cells, constitutive activation of c-Src disregulates many of these processes. Inhibition of c-Src activity using both molecular approaches and pharmacologic inhibitors in multiple cancer cell types has been found to lead to reduced anchorage-independent growth (8), decreased proliferation (9), cell cycle arrest (10), decreased tumor growth in vivo (11, 12), apoptosis (9), increased susceptibility to anoikis (13), diminished in vitro invasion and migration (14, 15), decreased in vivo metastasis (12, 16), and decreased in vivo vascularity (17). In NSCLC specifically, c-Src inhibition leads to decreased hypoxia-induced vascular endothelial growth factor expression (18). Inhibition of c-Src with a pharmacologic inhibitor (dasatinib) leads to profound and universal in vitro inhibition of migration and invasion of NSCLC cells. However, its effect on viability and proliferation is more variable and occurs at concentrations of dasatinib that are near or above the peak plasma concentrations possible in humans (14).

c-Src has multiple downstream substrates that mediate its biological functions in cancer cells. The interaction between c-Src and its substrate focal adhesion kinase (FAK) is essential for normal cell migration and invasion (19). c-Src also regulates downstream proliferation induced by growth factor receptors. Following activation by growth factor receptors, c-Src promotes survival via phosphorylation of the p85 subunit of phosphatidylinositide 3 kinase (PI3K) and thus the AKT pathway, signal transducer and activator of transcription-3 (STAT3), STAT5, and Shc, and thus the Ras/mitogen-activated protein kinase pathway (13, 20, 21). The STAT family of transcription factors, especially STAT3, regulates oncogenic signaling in many different tumor types (22). Indeed STAT3 is required for viral Src-mediated transformation (23). STAT3 can be activated by growth factor receptors or cytokine receptors, usually via nonreceptor tyrosine kinases such as c-Src or janus-activated kinase (JAK) proteins. STAT3 activation leads to the increased expression of downstream targets (e.g., Bcl-XL, cyclin D1, survivin) and increased cell survival, proliferation, and tumor growth in vivo (24). Inhibition of STAT3 results in increased apoptosis, decreased proliferation, and decreased tumor size (25, 26). STAT3 activation can also contribute to angiogenesis (27). Hypoxia-induced vascular endothelial growth factor expression is dependent on c-Src activation; this activation of c-Src leads to the downstream activation of STAT3, which binds to the vascular endothelial growth factor promoter with hypoxia-inducible factor-1 (HIF-1).

Although targeting growth factors and signal transduction pathways is a successful strategy in several tumor types, feedback and parallel signaling pathways can limit the efficacy of this approach. Despite c-Src expression in epithelial tumors, including NSCLC, and robust inhibition of c-Src with clinically relevant agents (e.g., dasatinib), the effect of c-Src inhibition on cell survival and proliferation has been modest (14). Defining mechanisms that limit the cytotoxic effects of c-Src inhibitors may result in the development of combinations of therapeutic agents for NSCLC that inhibit metastasis and enhance cytotoxicity. Because c-Src mediates its effects on cancer cell survival and proliferation via diverse substrates, including STATs, in this study we tested our hypothesis that STAT3 may not be inhibited sufficiently by c-Src inhibition in NSCLC to result in clinical effects. We determined that STAT3 was not durably inhibited in NSCLC cell lines and xenografts following c-Src inhibition, making it a candidate pathway for resistance to chemotherapy. Consistent with this result, we found that depletion of STAT3 enhanced the cytotoxicity of c-Src inhibition. In addition, we observed an inverse correlation between c-Src and STAT3 activation levels in untreated primary lung tumors by reverse-phase protein array. The results of these studies support a model in which c-Src and STAT3 are reciprocally regulated in NSCLC tumors, allowing for cancer cell survival following c-Src inhibition.

Dasatinib for in vitro studies was provided by Bristol-Myers Squibb and was prepared as a 10 mmol/L stock solution in DMSO. Dasatinib for animal studies was purchased from The University of Texas M. D. Anderson Cancer Center pharmacy. Pyridone 6 was purchased from Calbiochem.

Human NSCLC cell lines A549 and H226 were obtained from the American Type Culture Collection. H1299, H2009, and H1792 were gifts from Dr. John Minna (Hamon Center for Therapeutic Oncology Research, University of Texas Southwestern Medical Center, Dallas, Texas). Cell lines with wild-type epidermal growth factor receptors (EGFR) were selected for these studies because EGFR mutations can profoundly affect the response of NSCLC to c-Src inhibition. Three KRAS mutant cell lines were included (A549, H2009, and H1792) because this mutation is common in patients with NSCLC. Cells were grown in monolayer cultures in RPMI 1640 medium containing 10 fetal bovine serum (A549, H226, H1299, and H1792) or RPMI 1640 medium supplemented with hydrocortisone, insulin, transferrin, estadiol, and selenium (HITES) containing 5 fetal bovine serum (H2009) at 37C in a humidified atmosphere of 95 air and 5 CO₂.

Western blot analysis was done to measure protein expression and phosphorylation at 30 min and 7 h after inhibition or specific knockdown of c-Src, STAT3, and/or JAK. Protein levels in treated cells were compared with those in untreated cells at these same times to control for the effect of confluence or vehicle effects on STAT3 activation. Antibodies used in the Western blot analysis included c-Src and pSTAT3 S727 (Santa Cruz Biotechnology); pSrc Y419, pSTAT3 Y705, STAT3, pFAK Y861, Lyn, Yes, Bcl-XL, survivin, and STAT5 (Cell Signaling Technology); and -actin (Sigma Chemical Company).

For the Western blot analysis, cells were rinsed and lysed as previously described (28). Equal protein aliquots from cleared lysates were resolved by SDS-PAGE, transferred to nitrocellulose membranes, immunoblotted with primary antibody, and detected with horseradish peroxidaseconjugated secondary antibody (Bio-Rad Laboratories) and enhanced chemiluminescence reagent (Amersham Biosciences).

To knock down STAT3, the NSCLC cells were harvested, washed, and suspended at a density of 1 million cells/100 L of Nucleofector-V solution (Amaxa Corp.). Small interfering RNA (siRNA; 200 pmol/100 L) was added to the cell suspension and electroporated using the U-31 Nucleofector program (Amaxa). Immediately after electroporation, 500 L of prewarmed RPMI medium were added to the cuvette, and the cells were transferred to 6-well plates. The medium was changed after 16 h. STAT3 and c-Src siRNA were predesigned by siGenome Smartpool (Dharmacon, Inc., a part of Thermo Fisher Scientific) and obtained from Ambion. Controls included cells that were mock-transfected (i.e., without siRNA) and those transfected with a nontargeting (scrambled) siRNA.

The MTT assay was used to assess cytotoxicity as previously described (28). For each cell line, eight wells were treated with 0, 1, 2, 4, or 8 mol/L dasatinib and the IC₅₀. Cell cycle analysis was done as described previously (29). Briefly, cells were fixed and stained with propidium iodide. DNA content was analyzed by fluorescence-activated cell sorting analysis (Becton Dickinson) using ModFit software (Verity Software House). Apoptosis was measured using terminal deoxynucleotidyl transferase biotin-dUTP nick-end labeling staining using the manufacturer's protocol (APO-BRDU kit, Phoenix Flow Systems) as described previously (30). Briefly, fixed cells were incubated with terminal deoxynucleotidyl transferase and Br-dUTP and subsequently incubated with a fluorescein-labeled antiBr-dUTP antibody and analyzed by fluorescence-activated cell sorting analysis.

All animal procedures were done in accordance with the policies of M.D. Anderson's Institutional Animal Care and Use Committee. Ten female Swiss nu/nu strain, 6-week-old mice were used for each xenograft model. Each athymic nude mouse was injected s.c. with 4 million A549 or H226. When visible tumors had developed, dasatinib was administered by oral gavage at a dose of 20 mg/kg/d for 5 d. The mice were euthanized 2 h after the last dose of dasatinib, tumors were dissected, and the mice were examined for distant metastases. The tumors were homogenized and subjected to Western blotting as previously described (30).

Forty-six paired normal lung and NSCLC tumor samples were obtained from surgical specimens in the M. D. Anderson Cancer Center Thoracic Tissue Bank (Table 1). Of the tumors, 22 were squamous cell carcinomas and 24 were adenocarcinomas. The median age of the patients from whom the tumors had been excised was 67 y (range, 48-81 y); 22 (48) were women, and 44 (96) were former or current smokers. Six (13) patients had stage IA cancer, 14 (30) stage IB, 2 (4) stage IIA, 13 (28) stage IIB, 3 (7) stage IIIA, 6 (13) stage IIIB, and 2 (4) stage IV.

Protein lysate was prepared from pellets from tumor tissues as previously described (31). Briefly, lysis buffer [1 Triton X-100, 50 mmol/L HEPES (pH 7.4), 150 mmol/L NaCl, 1.5 mmol/L MgCl₂, 1 mmol/L EGTA, 100 mmol/L NaF, 10 mmol/L NaPPi, 10 glycerol, 1 mmol/L phenylmethylsulfonyl fluoride, 1 mmol/L Na₃VO₄, and 10 g/mL aprotinin] was added to the samples, followed by microcentrifugation at 14,000 rpm for 10 min. Clear supernatants were collected, followed by protein quantitation using the BCA reaction kit (Pierce Biotechnology, Inc.). The cell lysate was mixed with SDS sample buffer without bromophenol blue [three parts cell lysate plus one part 4 SDS sample buffer, which contained 35 glycerol, 8 SDS, 0.25 mol/L Tris-HCl (pH 6.8)]. Before using the cell lysate, 10 -mercaptoethanol was added. The samples were boiled for 5 min. Then, the samples (each in duplicate) were serially diluted (1:2-1:16) with SDS sample buffer. To each of the diluted samples an equal amount of 80 glycerol/2 PBS solution (8 mL of glycerol mixed with 2 mL of 10 PBS without Ca²⁺ and Mg²⁺) was added, after which the diluted samples were transferred to 384-well plates. Reverse-phase protein arrays (RPPA) were produced and analyzed as described (52), with slight modifications. Protein arrays were printed on nitrocellulose-coated glass FAST Slides (Schleicher & Schuell BioScience Inc.) by a GeneTAC G3 arrayer (Genomic Solutions) with 48 200-m-diameter pins arranged in a 4 12 format. Forty-eight grids were printed at each slide with each grid containing 24 dots. Protein dots were printed in duplicate with five concentrations. Arrays were produced in batches of 15, and occasional low-quality arrays (e.g., with many spot dropouts) were discarded.

Antibody staining of each array was done using an automated BioGenex autostainer. Briefly, each array was incubated with a specific primary antibody: pSrc Y527, pSTAT3 Y705, pSrc Y416 (which binds the pSrc Y419 activation site in humans), pFAK Y576, p-p130Cas Y249, and pPaxillin Y118; Cell Signaling); c-Src and STAT3 (Upstate). The signal was detected using the catalyzed signal amplification (CSA) system according to the manufacturer's recommended procedure (DakoCytomation California, Inc.). In brief, the RPPA slides were blocked for endogenous peroxidase, avidin, and biotin protein activity with I-block (Applied Biosystems) at room temperature for 15 min. After the blocking procedure, the slides were incubated with primary antibody and secondary antibody and diluted in DAKO antibody diluent with a background-reducing compound at room temperature for 20 min each. The slides were then incubated with streptavidin-biotin complex and biotinyl-tyramide (for amplification) for 15 min each, streptavidin-peroxidase for 15 min, and 3,3-diaminobenzidine tetrahydrochloride chromogen for 5 min. Between steps, each slide was washed with TBS containing 0.1 Tween-20 (TBST). Spot images were quantified using imaging analysis with an HP Scanjet 8200 scanner (Hewlett Packard) with a 256-shade gray scale at 600 dots per inch.

RPPA data were quantified using a SuperCurve method which detects changes in protein level by Microvigene software (VigeneTech) and an R package developed in-house (32). Briefly, the SuperCurve method generates a common logistic curve by pooling data from all samples on the slide. Individual dilution series numbers for each sample are then mapped onto the SuperCurve for quantification. After quantification, data were logarithm-transformed (base 2) for further processing and analyses. Then median-control normalization was applied on the dataset. The statistical analyses were done using R (version 2.7.0). All samples were done in duplicate, and average values were used for analysis. Two sample t-tests were used to compare protein levels between normal and tumor tissue; Pearson correlations were used to analyze the association between protein levels in the samples. In all analyses, P 0.05 was considered significant.

Western blot analysis of five NSCLC cell lines showed that c-Src phosphorylation was rapidly (30 minutes) and durably (7 hours) inhibited at a site essential for c-Src activation (pSrc Y419 in human c-Src; Fig. 1A). Total c-Src levels were not changed by dasatinib treatment. In contrast to pSrc Y419, in which inhibition was prolonged, STAT3 activation (as shown by pSTAT3 Y705 levels) was transiently inhibited at 30 minutes in A549 and H226 (0.60 and 0.57 times the control value, respectively), followed by reactivation by 7 hours (1.25 and 1.18 times control). In H2009, H1299, and H1792 activation was seen beginning at 30 minutes (range, 1.20-1.78 times control) and persisted at 7 hours (1.19-1.63 times control; Fig. 1B).

After 1, 5, or 7 hours and after 5 days of dasatinib treatment, tumors from mouse xenografts of NSCLC cells were grossly dissected and examined by immunohistochemistry. Tumors were confirmed to consist primarily of NSCLC cells (>90) with no distant metastases (data not shown). Protein expression and phosphorylation were then measured by Western blot analysis and compared with control (vehicle-treated mice sacrificed at 7 hours or 5 days; Fig. 2). As expected, dasatinib treatment resulted in pSrc Y419 inhibition at all times. In A549 xenografts, pFAK inhibition was seen beginning at 5 hours. In contrast, but consistent with the in vitro data, pSTAT3 Y705 was inhibited at early times (0.58-fold at 5 hours after treatment) but was strongly induced by 7 hours (3.1-fold). pSTAT3 Y705 levels returned to baseline after 5 days of continuous daily treatment. Unlike the in vitro studies, the in vivo studies showed that total STAT3 levels were also elevated at 7 hours after treatment but were not significantly different from control levels. In H226 xenografts, pSTAT3 Y705 was inhibited between 5 and 7 hours (0.32- and 0.44-fold at 5 and 7 hours, respectively), but rose above baseline by 5 days (1.7-fold).

To determine whether STAT3 reactivation was downstream of c-Src and not caused by an off-target effect of dasatinib, we examined the effect of c-Src depletion by siRNA at 72 and 96 hours in the NSCLC cell lines A549 and H226. Cell lines were transfected with either scrambled siRNA or c-Srcspecific siRNA. In A549 cells, c-Src knockdown decreased total and activated c-Src levels but increased pSTAT3 Y705 by 6.7- and 7.2-fold at 72 and 96 hours, respectively (Fig. 3). In H226 cells, total and activated c-Src levels were suppressed at both time points; pSTAT Y705 was strongly induced by 72 hours (3.5-fold) and returned near baseline by 96 hours (1.5-fold; Fig. 3). Other Src family members, such as Yes and Lyn, were not affected by c-Src knockdown in either cell line.

To determine whether STAT3 reactivation was mediated by JAK proteins, we tested the effect of JAK inhibition on STAT3 activation. NSCLC cell lines were treated with the JAK inhibitor pyridone 6 alone and in combination with dasatinib (Fig. 1A). JAK inhibition alone had no effect on pSrc Y419 but led to complete inhibition of pSTAT3 Y705 and its downstream targets survivin and Bcl-XL at 7 hours. When combined with dasatinib, JAK inhibition by pyridone 6 prevented STAT3 phosphorylation.

To evaluate the relationship between c-Src and STAT3 in clinical samples, we used reverse-phase protein array, due to its sensitivity and the small amount of protein required, to quantify the levels of total and phosphorylated Src and STAT3 as well as downstream targets of Src in paired samples of NSCLC tumors and normal lung tissue. Paired t-test of tumor and normal tissue from the same patients showed significantly higher c-Src activity in tumors as illustrated by decreased levels of autophosphorylated and inactive c-Src (pSrc Y527; P = 1.09 10⁹) in tumor tissue as compared with normal lung (Fig. 4A). In contrast, activated STAT3 (pSTAT3 Y705) levels were significantly lower in tumor tissue (P = 0.006; Fig. 4A).

Among the tumor specimens, there was a statistically significant correlation between levels of pSrc Y527 and pSTAT3 Y705, with a Pearson correlation coefficient of 0.32 (P = 0.03; Fig. 4B). This correlation was also seen when the ratio of pSrc Y527 to total c-Src in tumors was compared with the ratio of pSTAT3 Y705 to total STAT3 (r = 0.56; P = 5.25 10⁵; Fig. 4B). In addition, the decrease in pSrc Y527 from normal to tumor correlated directly with the decease in pSTAT3 Y705 from normal to tumor tissue (r = 0.33; P = 0.03).

Downstream targets of Src (FAK, p130Cas, and paxillin) were also evaluated by RPPA. As expected, activation of these targets (as measured by phosphorylation) was inversely correlated with pSTAT3 Y705 and pSrc Y527. Specifically, correlation coefficients between pSTAT3 Y705 and phosphorylated FAK, p130Cas, and paxillin were r = 0.50 (P = 0.0005), 0.30 (P = 0.04), and 0.42 (P = 0.004), respectively (Fig. 4C).

Despite testing several antibodies for RPPA, there is not yet a validated antibody for this assay for the activated form of c-Src (human pSrc Y419). However, using the best performing antibody available, we observed a 1.6-fold higher level of pSrc Y419 in tumors compared with normal tissue (P = 0.06). The ratio of pSrc Y419 to total Src was also correlated with the ratio of pSTAT3 Y705 to total STAT3, although this was not statistically significant (P = 0.67).

Because of its role in mediating survival and proliferation, STAT3 induction following c-Src inhibition may represent a mechanism of drug resistance. To examine the biological effects of STAT3 reactivation in NSCLC cells, we evaluated the effect of STAT3 knockdown on cytotoxicity when combined with dasatinib. A549 and H226 cells were transfected with STAT3 siRNA, scrambled (nontargeting) siRNA, or mock-transfected. A549 and H226 cells transfected with STAT3 siRNA showed an 87 and 79 decrease in STAT3 protein levels (respectively) at 48 hours after transfection with STAT3 siRNA as compared with scrambled siRNA (Fig. 5A and B). To determine the biological effect of c-Src inhibition combined with specific depletion of STAT3, cells were treated with dasatinib 48 hours after transfection. A MTT assay was then used to estimate the number of living cells remaining after 72 hours of treatment. Cells with depleted STAT3 were significantly more sensitive to dasatinib than those transfected with scrambled siRNA. In A549 cells, the IC₅₀ values were 0.7 and 4 mol/L in control and STAT3 siRNAtransfected cells, respectively, for A549 (Fig. 5C), and 5 and 38 mol/L for H226 (Fig. 5D). Unlike dasatinib alone, the combination of STAT3 knockdown with dasatinib strongly induced apoptosis in both cell lines (Fig. 5E). Dasatinib alone induced cell cycle arrest in H226 cells, but this was not significantly affected by the addition of STAT3 depletion.

In this study, we found that c-Src and STAT3 activation, as shown by phosphorylation status, were reciprocally regulated in NSCLC cell lines, xenografts, and human tumors. Despite an initial inhibition of STAT3 phosphorylation in the A549 and H226 models, prolonged c-Src inhibition resulted in an increase in STAT3 phosphorylation in all NSCLC cell lines tested both in vitro and in vivo. STAT3 reactivation was JAK-dependent, as illustrated by the observation that reactivation was inhibited by the addition of a JAK inhibitor, pyridone 6. Finally, we showed that inhibition of STAT3 reactivation (by STAT3 siRNA) enhanced the cytotoxicity of dasatinib, showing that this pathway has biological significance. Taken together, these results suggest that STAT3 reactivation may be an important mechanism of resistance to c-Src inhibitors in NSCLC and may be a clinically relevant target for combination therapy.

Previous studies have shown that c-Src is overexpressed in NSCLC and that increased c-Src activity is associated with worse clinical outcome. Clinical investigators are enthusiastic about c-Src inhibitors because specific and potent kinase inhibitors are well tolerated in humans (33). Two such approved anticancer drugs, imatinib and erlotinib, use ATP-competitive kinase inhibition to inhibit Bcr-Abl (34) and EGFR (35), respectively, proving that kinase inhibition of signal transduction molecules can lead to profound tumor responses. However, thus far c-Src inhibitors have shown limited activity in NSCLC patients. Understanding the mechanisms of resistance to c-Src inhibition in NSCLC will be extremely important for understanding how these drugs can be used more effectively in this disease.

In NSCLC, resistance to tyrosine kinase inhibitors, such as those targeting EGFR, is often due to either activation of the signaling pathway downstream to the drug target (e.g., k-Ras mutations) or signaling through alternate pathways (e.g., c-Met; refs. 36, 37). Therefore, because STAT3 activity is an important downstream target of c-Src and necessary for c-Src signal transduction, characterization of its relationship to c-Src activity and response to c-Src inhibition was of particular interest. Interesting, these studies did show reactivation of STAT3 in the setting of c-Src inhibition. In cell lines from head and neck squamous cell cancers (30), squamous cell carcinoma of the skin,⁸

Three negative feedback loops regulate STAT function after cytokine signaling: SH-2containing phosphatases, which inactivate JAK by dephosphorylation; protein inhibitors of activated STAT, which are negative regulators of STAT-induced transcription (i.e., downstream of STATs); and suppressors of cytokine signaling, which inhibit JAK activity, facilitate proteosomal degradation of JAK, and compete with STATs for binding to cytokine receptors (39). Although there are no known positive feedback loops leading to STAT3 activation after its inhibition, loss of a negative feedback loop could play the same role. For example, v-Abl leads to JAK/STAT activation via its disruption of suppressor of cytokine signaling 1 function (SOCS-1) (40).

The concentration (100 nmol/L) of dasatinib was chosen for these studies because it has been shown to completely inhibit c-Src and is relatively specific (41). For example, in intact cells, we observed 90 reduction in phosphorylated c-Src with >40 nmol/L dasatinib.⁹

Although we showed an inverse correlation between c-Src and STAT3 activity in human tumors, human NSCLC samples were not available from the post-dasatinib setting to confirm a reactivation of STAT3 in patients following c-Src inhibition. Nevertheless, we believe that our in vitro and murine models sufficiently support a rationale for combining c-Src and STAT inhibition in the clinical setting. Three ATP-competitive c-Src inhibitors are being studied in clinical trials: dasatinib, AZD0530 (AstraZeneca), and SKI-606 (Wyeth). A nonATP-competitive c-Src inhibitor is also in clinical trial (Kinex pharmaceuticals). Preclinical studies of c-Src inhibitor AZD0530 also showed a reactivation of STAT3 in A549 cells 24 hours after treatment with this drug, further supporting that this may be an important mechanism of resistance across this class of drugs (46). Many JAK inhibitors are being studied in the laboratory and several are in early clinical trial (47,50). Our long-term goal is to use the results of these studies to design clinical trials of these or other more specific c-Src and JAK inhibitors, as available, to improve the survival of patients with NSCLC.

No potential conflicts of interest were disclosed.

We thank Doris R. Siwak for her contribution to the RPPA experiments.