Background
Signal transducer and activator of transcription (Stat) proteins are a family of transcription factors that are activated by phosphorylation of a conserved tyrosine residue in response to a host of growth factors and cytokines. Phosphorylated Stat dimers translocate into the nucleus to activate of several target genes that are involved in diverse cellular processes such as cytokine signaling [
1], cell proliferation and development [
2], abnormal tumorigenesis [
3‐
5], and suppression of the immune response in the tumor microenvironment [
6]. Stat3, one of seven members of the Stat family, has been most strongly implicated in tumorigenesis [
3‐
5]. Stat3 regulated genes include cyclin D1 [
7] and c-myc [
8], which are involved in cell proliferation; Bcl-X
L [
9], survivin [
10], and Bcl-2 [
11], which mediate apoptosis; several matrix metalloproteases involved in invasion [
12‐
14]; and growth factors and cytokines such as VEGF [
15] and HGF [
16]. Up- or down-regulation of many of these genes has obvious implications in the development of cancer. Constitutive activation of Stat3 has been observed in a number of human cancers and cancer cell lines [
3]. Given that no naturally occurring Stat3 mutations that result in constitutive activity have been identified, the persistent Stat3 activation in tumors is likely due to a differences in expression or activity of proteins that regulate Stat3 or signaling molecules involved in the Stat3 pathway. Potential candidates include suppressor of cytokine signaling (SOCS), a negative regulator of cytokine signaling that is silenced by methylation in some tumors [
17,
18] and various receptor tyrosine kinases such as EGFR and c-met that are activated in cancers [
19,
20]. More recent evidence indicates that Stat3 may be a required for the maintenance of stem cell-like characteristics of glioblastoma stem cells [
21]. Given the central role of stat3 in positively effecting multiple biological processes involved in malignant cell behavior, extensive effort has been made to target Stat3 and suppress its activity in cancer cells [
22]. One recent preclinical study using myeloid and B cell-specific targeting of Stat3 by siRNA showed that silencing Stat3 expression lead to a strong antitumor immune response [
23].
Stat3 has a critical function in the development of skin cancer [
24]. Using both skin-specific Stat3 knockout models [
25,
26], and a skin specific Stat3 gain of function transgenic model (K5.Stat3C mice) [
27], we have shown in collaboration with the DiGiovanni laboratory that Stat3 is indispensable for the initiation, promotion and malignant progression stages of skin carcinogenesis. More recently it has been shown, using skin stem cell-specific knockout of Stat3, that Stat3 is required for survival of skin stem cells during tumor initiation in the mouse skin 2-stage chemical carcinogenesis protocol, and that it is indeed these stem cells that form the initiated cell population that eventually gives rise to tumors [
28]. These results are reviewed in [
29,
30].
Malignant progression and tumor cell invasion are often the result of uncontrolled cell motility and adhesion. There is considerable evidence for a role for Stat3 signaling in cell migration and invasion under normal and pathological situations. Overexpression and activity of Stat3 has been linked to the invasion and metastasis of several cancers in humans, including cutaneous squamous cell carcinoma (SCC) [
31,
32], colorectal adenocarcinoma [
33], and melanoma [
14]. Stat3 activation is required for induction of genes encoding matrix metalloproteases-1, 2, and 9 (MMP-1, MMP-2, and MMP-9) [
12‐
14], as well as several other genes important in the metastatic cascade. In addition, Stat3-deficient keratinocytes are unable to migrate, partly due to deregulated p130
CAS phosphorylation [
34], a protein which is involved in the formation of focal adhesion complexes and reorganization of the cytoskeleton. These studies, as well as our studies with mouse skin-specific Stat3 gain of function and knockout models [
29,
30], suggest that Stat3 activation plays a critical role in the invasion and metastasis of carcinomas.
Scatter factor, or hepatocyte growth factor (HGF), can regulate cell survival, growth, migration, and angiogenesis upon binding its cell surface receptor, c-met [
35]. c-met is often constitutively active in human tumors [
36‐
38], and HGF/c-met signaling is at least partially mediated by Stat3 [
20,
39]. In particular, Stat3 is transiently phosphorylated after HGF treatment, likely mediated by Stat3/c-met direct or indirect interaction [
39]. Stat3 was coimmunoprecipitated with phosphorylated c-met from intact cells, and the specificity of this interaction was demonstrated in
in vitro competition experiments with peptides mimicking the c-met docking sequence and Stat3 phosphorylation site [
39]. Activated Stat3 mediates signals downstream of the HGF/c-met pathway [
20], and Stat3 has been shown to be essential for HGF-induced morphogenesis and for invasive behavior driven by c-met in both fibroblasts and breast carcinoma cells [
39‐
41]. A Stat3 binding site has been identified in the HGF promoter and this site is required for Stat3-mediated activation of HGF transcription in breast epithelial cells [
16,
42]. Overall, these observations suggest that positive feedback signals in the HGF/c-met/Stat3 signaling pathway could contribute to tumorigenesis.
In order to investigate the role of Stat3 in the induction of cell motility and invasion in greater detail, we have employed a cell culture model in which a dominant-negative form of Stat3 was expressed in a tumorigenic human skin SCC cell line, SRB12-p9 [
43]. Stat3α, the full-length form of Stat3, is abundant and is constitutively phosphorylated (active) in these cells. It has been reported that Stat3β, a naturally occurring Stat3 splice variant that lacks the C-terminal transcriptional activation domain, can act in a dominant-negative fashion to inhibit the transcriptional activity of Stat3α [
44,
45]. It was subsequently shown that mutating the critical tyrosine 705 phosphorylation site to phenylalanine generated a form of Stat3β (Stat3β-Y705F) that could block DNA binding by all endogenous forms of Stat3 [
46]. In this study we compare the malignant characteristics of SRB12-p9 cells with that of stably transfected clones that express FLAG-tagged Stat3β-Y705F (S3DN cells), which have suppressed Stat3 activity [
43]. We confirm for the first time that interference with the HGF/c-met/Stat3 signaling pathway can block tumor cell invasion in an
in vivo model, and we present evidence for a positive feedback loop between Stat3 and c-met. Finally we correlate membrane localization of phospho-Stat3 with invasion
in vivo.
Methods
Cell culture
The human skin SCC cell line SRB12-p9 was derived by single cell cloning from SRB-12 cells (a gift from Dr. Janet Price, Department of Cancer Biology, University of Texas M.D. Anderson Cancer Center). Cells were cultured in a humidified atmosphere at 5% CO
2, in Dulbecco's Modified Eagle's Media-F12 supplemented with 10% fetal calf serum. Generation of the stable S3DN expressing SRB12-p9 cell lines and vector control cell lines was previously described [
43].
Invasion and scattering assays
8 μm pore-sized membrane inserts for 12 well plates were coated with 50 μL of 1:25 diluted Matrigel basement membrane matrix (Becton Dickinson) following the manufacturer's protocol. 25,000 WT, NEO4 and S3DN cells were seeded in triplicate onto inserts and incubated for 10 hours at 37°C in 5% CO2. No chemo-attractant or serum gradient was used. After incubation, membranes were fixed with 4% paraformaldehyde for 20 minutes and stained with hematoxylin. Cells that migrated through the Matrigel to the bottom of the insert were photographed on a Nikon TE300 microscope and counted. For the scattering assay, cells growing in 10% FCS-containing media were treated with 100 ng/ml HGF for 24 or 48 hours and photographed with phase contrast at 100× magnification on a Nikon TE300 microscope. Percent cell scattering was determined by counting the total number of cells in a field (T), and the number of cells which were visually determined to have no cell-cell contacts (M) in the field, and applying the following formula: percent scattering = (M/T) × 100%. At least 6 fields were counted for each treatment group and the fields were marked on the bottom of the dish so that they could be located at the 24 and 48 hour time points.
Western blotting
Cells were lysed and tumors homogenized on ice in RIPA lysis buffer (150 mM NaCl, 50 mM Tris-HCl pH 7.5, 1 mM EDTA, 1% NP-40, 1 mM PMSF, 1 mM Na-orthovanadate) supplemented with 40 μl complete protease inhibitor cocktail (Roche) according to manufacturer-provided instructions. Extracted protein was quantified using the BioRad Protein Assay kit. Proteins were separated by SDS acrylamide (4-20%) gel electrophoresis and transferred to nitrocellulose membranes (BioRad). Blots were blocked with 5% BSA for 1 hour at room temperature, followed by incubation overnight at 4°C with antibodies against phosphorylated forms of c-met (Tyr1234-1235) and Stat3 (Tyr705), total Stat3 (Cell Signaling Technology), total c-met and b-actin (Santa Cruz), and MMP-2 and MMP-9 (Chemicon). Blots were washed with TBS/0.1% Tween 20 and incubated with a horseradish peroxidase-conjugated secondary antibody for 1 hour at room temperature, followed by an additional 3 washes with TBS/0.1% Tween 20. Chemiluminescence detection was performed according to the manufacturer's instructions (Millipore) followed by exposure to X-ray film or using the Chemidoc gel documentation system (BioRad).
Immunoprecipitation
After overnight incubation in serum free media, cells were treated with 100 ng/ml HGF for 30 minutes prior to harvest, and lysed with the following immunoprecipitation buffer: 1% NP-40 in PBS with 40 μl complete protease inhibitor cocktail (Roche), 2 mM Na-orthovanadate, 10 μM lactacystin, 10 mM NaF. Cell lysates (500 μg for each sample) were rocked at 4°C for 1 hour, followed by centrifugation for 5 minutes to pellet cellular debris. Supernatants were collected and samples were precleared with Protein A/G Plus-Agrarose (Santa Cruz) and normal IgG (Santa Cruz) for 1 hour at 4°C, followed by centrifugation for 5 minutes at 20,000 g. Total Stat3 (Cell Signaling Technology) and c-met (Santa Cruz) antibodies were added at a dilution of 1:1000 to a total volume of 1 ml, and immunoprecipitations were performed at 4°C overnight with constant rocking. 20 μl of Protein A/G Plus-Agarose (Santa Cruz) beads were added, and the solution was allowed to rock for 4 hours at 4°C. Reactions were washed 3 times using immunoprecipitation buffer and then brought up in 6× Laemli buffer for protein electrophoresis and Western blot analysis.
Mouse xenograft model
Groups of 6-7 week old female athymic NCR Nu/Nu (nude) and SCID/bg mice were housed in a temperature and humidity controlled Association for Assessment and Accreditation of Laboratory Animal Care facility with a 12 hour light/dark cycle. All procedures were approved by the LSUHSC Institutional Animal Care and Use Committee in accordance with NIH guidelines. Mice were maintained on LM-485 diet (Harlan Teklad) and allowed access to food and water ad libitum. Mice were injected subcutaneously in the interscapular region with 1 × 106 cells in 0.1 ml PBS. The growth of the subcutaneous tumors was evaluated twice weekly by caliper. Animals were sacrificed and tumors were excised at the end of the experiment (21 days after cell injection) and fixed in 4% paraformaldehyde for histological observation, western blotting, and immunohistochemistry.
Immunohistochemical examination of tumors
Tumors were isolated and fixed in formalin and embedded in paraffin prior to sectioning. Paraffin sections of 4 μm were deparaffinized in xylene, 3 × 7 minutes, and rehydrated by stepwise washes in decreasing ethanol/H2O ratio (100% to 50%, followed by soaking in water). Sections were either stained with hematoxylin and eosin (H&E) or boiled for 1 minute for antigen retrieval. For immunohistochemical staining, sections were incubated in Superblock (Pierce) blocking reagent for 1 hour at room temperature. After washing 3× in PBS, slides were incubated for overnight at 4°C with antibodies against p-c-met (Tyr1234-1235), p-Stat3 (Tyr705), and total Stat3 (Cell Signaling Technology), total c-met (Santa Cruz), and MMP-2 and MMP-9 (Chemicon). Slides were photographed on a Nikon TE300 fluorescence microscope under oil immersion, at 600× magnification, with a CCD camera (Roper Scientific). Images were processed with IPLabs v3.55 software (Scanalytics).
Statistical analysis
The statistical significance between experimental values was assessed by Student's T test. A P value of < 0.05 was considered as statistically significant.
Discussion
Given the cumulative evidence supporting Stat3 as a potential therapeutic target in cancer and that there are ongoing clinical trials testing Stat3 inhibitors [
49], our group has attempted to better understand the molecular mechanisms of Stat3 activity and the role of HGF/c-met in transducing its signal. To this aim, we have used the S3DN system, which provides a workable cell culture model to study multiple aspects of Stat3 function in detail. Here we successfully demonstrate that even partial reduction in Stat3 activity is sufficient to restrict motility and invasion. S3DN (DN2 and DN5) cells were significantly less invasive in both
in vitro and
in vivo assays, and this phenotype was accompanied by marked suppression of MMP-2 and MMP-9 expression, c-met activity, as well as inhibition of Stat3/c-met interaction.
The expression of S3DN reduced HGF-induced scattering and invasion through Matrigel-coated membranes (Figure
1). This assay indicates that the motile phenotype of the P9WT cells in this context is cytokine dependent and mediated by Stat3, which is relevant for tumors growing
in vivo where they respond to a variety of cytokines and growth factors from host cells or from autocrine stimulation. The lack of complete suppression of invasiveness for the S3DN tumors indicates less than full penetrance for the effect of S3DN expression. This can be attributed to the fact that there is still active Stat3 in the S3DN cells (Figure
4, data not shown). It is also possible that the percentage of attached and invasive tumors would have been even higher for the WT and NEO4 groups had the experiment been carried out for a longer time, since it is likely that those tumors designated as attached would have eventually become invasive if allowed to grow longer.
We originally hypothesized that the mechanism for the reduced invasive behavior of S3DN cells is due to a suppression of c-met-mediated events. c-met is frequently constitutively active in human tumors [
36‐
38], and several reports have shown that Stat3 is one of the downstream effectors of HGF/c-met signaling [
20,
39]. HGF treatment enhances Stat3 activity, likely due to an interaction between c-met and Stat3 [
39]. We show that expression of S3DN leads to reduced baseline as well as inducible c-met activity in S3DN cells and tumors (Figure
2,
4). Inhibition of c-met activity in S3DN cells is also associated with reduced localization of p-c-met to focal adhesions (Figure
2). These findings provide a strong indication for a positive feedback loop involving HGF, c-met, and Stat3. The existence of such a feedback loop is further supported by our finding that S3DN cells are less responsive to HGF-induced motility. An analysis of the levels of endogenous HGF secretion in S3DN cells by ELISA revealed no detectable difference between the WT, NEO4 and S3DN cell lines (data not shown). However, the levels of secreted HGF in these cells were difficult to compare since they were below the detection threshold of the ELISA (< 40 pg/ml). We note that based on numerous reports concerning the biological effects of HGF, which typically fall in the ED50 range of 20-100 ng/ml, we do not expect that these cells produce sufficient amounts of HGF to ellicit an autocrine response. Our initial hypothesis was supported by our finding that there was an interaction between Stat3 and c-met proteins, and that this interaction is reduced in cells that express S3DN (Figure
6). It is possible that Stat3/c-met interaction occurs through FAK, since there are known associations between c-met and FAK [
50] and Stat3 and FAK [
51]. We did find an association between c-met and FAK in IP experiments, but there was no detectable difference in the degree of this interaction between WT and S3DN cells (data not shown). We speculate that S3DN is preventing the activation of c-met at the cell membrane by interfering with normal Stat3/c-met interaction, possibly by competing with Stat3 for binding to c-met.
Interestingly, we observed membrane localization of p-Stat3 (Tyr705) in WT and NEO4 tumors, but not in the S3DN tumors, which had only nuclear staining (Figure
4A). This is in line with our previously published results with human SCC samples where we observed p-Stat3 (Tyr705) staining at the cell membrane and nuclei in tumors, but only in nuclei in the immediately adjacent non-malignant skin [
48]. It is possible that the membrane localization of p-Stat3 (Tyr705) is specific to cells that are more malignant. Recent published reports indicate that there are potential non-nuclear functions for Stat3. The functional flexibility of Stat3 was outlined in a review published by Gao and Bromberg [
52], specifically describing a non-transcriptional, cytoplasmic role that may potentiate cell motility. A Stat3-stathmin interaction has been documented [
53], as well as Stat3 localization at focal adhesions with FAK and paxillin [
51]. A recent publication describes a metabolic function in the mitochondria that supports Ras-dependent transformation [
54] and cellular respiration [
55]. These reports indicate non-nuclear roles for Stat3 that contribute to the oncogenic properties of cancer cells and that these functions are separate from tyrosine-phosphorylation dependent transcriptional activity. We provide further support for a novel function for Stat3 by describing invasive tumors with a distinct Stat3 membrane localization. Although we do see suppression of MMP-2 and MMP-9 expression, which is likely due to transcriptional downregulation caused by S3DN expression, we propose that Stat3 may also be acting at the membrane to enhance motility and invasion (Figure
6). We further speculate that the Stat3 protein engaged in the non-transcriptional function at the membrane may be exerting a fast-acting response to external activators, whereas the transcriptional activation of genes like the MMPs by nuclear Stat3 could be a slower mechanism to potentiate cell motility. Thus, it is possible that the transcriptional and non-transcriptional functions of Stat3 are working in concert to influence cell migration, especially in malignant, invasive cells.
The lack of diminished phospho-Stat3 staining in the S3DN cells appears at first contradictory to previous data showing a positive correlation between Stat3 phosphorylation and activity. However we suggest that the S3DN protein blocks endogenous Stat3 activity at a point downstream of its phosphorylation. We have previously shown that Stat3 DNA binding, as measured by EMSA, is reduced in the S3DN cells compared to WT cells, confirming that S3DN expression has a suppressive effect on Stat3 activity [
43].
Our observation that S3DN tumors grew to similar size as WT and NEO4 tumors was unexpected. A previous report, using HT-29 colon carcinoma cells in mouse xenograft experiments, showed a reduction in tumor size for cells stably expressing a similar Stat3 dominant negative form compared to parental HT-29 tumors [
56]. One possible explanation for this discrepency could be the fact that the parental HT-29 cells, unlike the SRB12-p9 cells, were negative for constitutive activation of Stat3 in the absence of exogenously added growth factors. From our initial studies of the S3DN cells we observed that the proliferation rate was unaffected by S3DN expression
in vitro, while sensitivity to exogenous growth factor deprivation was increased [
43]. We hypothesized that the partial suppression of Stat3 activity in the S3DN cells was sufficient to affect survival under growth factor deprived conditions, but not sufficient to affect proliferation rate. In the case of HT-29 cells, where dominant negative Stat3 expression appears to completely block Stat3 activity[
56], the tumor size is reduced even though host supplied growth factors are present. To observe a similar effect in our system would likely require stronger suppression of Stat3 activity.
Conclusions
We have shown for the first time in an in vivo tumor model that the HGF/c-met/Stat3 signaling cascade is critically involved in the motility and invasion of cancer cells, and disruption of this pathway by expression of S3DN suppresses invasion, reduces c-met activity and inhibits c-met/Stat3 interaction. Tumor cells producing high levels of HGF can activate c-met by either autocrine or paracrine signaling mechanisms, which leads to Stat3 activation and gene transcription. Based on the indications for a positive feedback loop between c-met and Stat3, we speculate that abnormally high levels of c-met activation could trigger this loop and drive cells toward invasive behavior. We also show that there is a possible role for Stat3 at the cell membrane that contributes to the malignant phenotype of invading tumor cells. We must express a high degree of caution in generalizing these results to all tumor cell types or even to other skin SCC cells, since these findings are confined to a single cell line. It will be necessary to perform a similar series of experiments on other skin SCC cell lines, as well as other tumor cell types. Attempts to generate additional stably expressing S3DN cells using other skin-derived cell lines, and cell lines of other tumor types, have been unsuccessful so far. This is perhaps due to a S3DN-mediated suppression of proliferation and/or survival that is too strong in those cells. The SRB12-p9 cells are likely to be unique in having a very strong constitutive Stat3 activity when grown in the presence of serum, thereby allowing survival of stable S3DN clones. Never-the-less, based on these results we suggest that targeting the HGF/c-met/Stat3 pathway could be an especially effective strategy for cancer therapy. Future work will focus on exploring the interaction between Stat3 and c-met, and the relationship of the HGF/c-met/Stat3 signaling loop to the invasive potential of skin SCC cells.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
ZAS carried out in vitro invasion, motility, adhesion, and soft agar assays, immunohistochemistry, Western blotting, and contributed to the draft of the manuscript. WY generated and characterized the S3DN cell lines. KH assisted with the immunohistochemistry, Western blotting, and in vitro cell assays. JNG maintained the mice, assisted with tumor harvest, and immunohistochemistry. RS performed statistical analysis for the design and interpretation of the mouse tumor Xenograft experiments. JLC contributed to the conception and design of the study, coordinated the study, contributed to the microscopy, tissue harvesting, and final editing of the draft of the manuscript. All authors have read and approved the final manuscript.