Expression of Activated Signal Transducer and Activator of Transcription 3 Predicts Expression of Vascular Endothelial Growth Factor in and Angiogenic Phenotype of Human Gastric Cancer Free

Gastric cancer remains the second most frequently diagnosed malignancy worldwide (1), and its aggressive nature is related to mutations of various oncogenes and tumor suppressor genes (2–5) and abnormalities in several growth factors and their receptors (3, 4). These abnormalities affect the downstream signal transduction pathways involved in the control of many aspects of cancer biology, including angiogenesis. Specifically, these perturbations may confer a phenotype of increased angiogenesis to gastric cancer cells (3–5). Previous studies indicated involvement of several angiogenic factors in gastric cancer angiogenesis, including interleukin-8, epidermal growth factor, fibroblast growth factor, insulin-like growth factor, matrix metalloproteinase-2, transforming growth factor, and vascular endothelial growth factor (VEGF; refs. 6–10). Previous studies also showed that microvessel density (MVD) and various angiogenic factors related to tumor development and progression are predictive of patient survival in gastric cancer (3, 11–15). However, the molecular mechanisms behind the abnormal expression of many of these angiogenic factors in gastric cancer, including VEGF, remain unclear.

Increasing evidence suggests that VEGF expression is regulated by various hormones and growth factors, including epidermal growth factor and fibroblast growth factor (5), and by oncogenic proteins such as Src and Ras (16, 17). Significantly, many tumors often exhibit overexpression of these growth factors and oncogenic proteins, and these molecules often transmit signals through the signal transducer and activator of transcription 3 (Stat3), a member of the Janus-activated kinase/STAT signaling pathway (16, 17). Therefore, it is highly probable that Stat3 activation is a common signaling intermediate leading to the overexpression of VEGF in malignant tumors. This notion is apparently supported by the fact that constitutively activated Stat3 protein has been found in various types of tumors, including leukemia; cancers of the breast, head and neck, pancreas, and prostate; and melanoma (16–19) . Our recent studies of human pancreatic cancer and melanoma have linked abnormal Stat3 activation to overexpression of multiple genes downstream of Stat3, including VEGF (17, 18) and MMP-2 (19). These findings led us to hypothesize that abnormal activation of Stat3 causes overexpression of multiple angiogenic molecules, which in turn render tumor cells highly angiogenic, as manifested by increased MVD. Recent studies also have revealed that altered Stat3 can contribute to oncogenesis, presumably through its critical role in the expression of many other genes key to the regulation of multiple aspects of tumor cell survival, growth, and angiogenesis (17–21). At present, it is not known whether abnormal Stat3 expression and activation critically contribute to gastric cancer development and progression. Because tumor MVD predicts patient survival (22), we asked whether Stat3 predicts tumor MVD, given the evidence that Stat3 may control the expression of many genes key to tumor angiogenesis, such as VEGF.

In the present study, we sought to determine whether Stat3 expression is related to VEGF expression and tumor MVD in gastric cancer specimens and whether these factors predict survival in gastric cancer patients. We found that elevated Stat3 activation and concomitant VEGF overexpression occurred in human gastric cancer and were directly correlated with tumor MVD and inversely correlated with patient survival. Furthermore, genetically enforced alterations of activated Stat3 expression led to altered VEGF expression and angiogenic potential in human gastric cancer cells. Therefore, abnormally activated Stat3 expression may be a potential molecular marker for poor prognosis and directly contribute to gastric cancer angiogenesis and aggressive gastric cancer biology.

Human Tissue Specimens and Patient Information. We used human gastric cancer tissue specimens preserved in the Gastric Cancer Tissue Bank at and obtained information about the patients from the University of Texas M.D. Anderson Cancer Center Upper Gastrointestinal Carcinoma Database. Primary gastric cancer in these patients was diagnosed and treated at M.D. Anderson Cancer Center from 1985 to 1998. The patients had well-documented clinical histories and follow-up information. None of them underwent preoperative chemotherapy and/or radiation therapy. Eighty-six patients with primary gastric cancer were randomly selected for this study. All of the patients had undergone gastrectomy with lymph node dissection. Gastric cancer centered at or above the gastroesophageal junction was not included in this study. Proximal gastric cancer centered below the gastroesophageal junction was included. Gastroesophageal junction cases associated with Barrett's were not included. All patients also were observed at M.D. Anderson Cancer Center through the end of 1999.The median follow-up period for the 86 patients was 25.7 months. At the end of 1999, 30 patients were still alive, whereas 56 patients had died. Fifty-three lymph node metastasis specimens from the 86 patients with gastric cancer and 57 normal gastric tissue specimens obtained from patients without gastric cancer were also included in this study.

Immunohistochemistry. Sections (5 μm thick) of formalin-fixed, paraffin-embedded tumor specimens were deparaffinized in xylene and rehydrated in graded alcohol. Antigen retrieval was done with 0.05% saponin for 30 minutes at room temperature. Endogenous peroxidase was blocked using 3% hydrogen peroxide in PBS for 12 minutes. The specimens were incubated for 20 minutes at room temperature with a protein-blocking solution consisting of PBS (pH 7.5) containing 5% normal horse serum and 1% normal goat serum and then incubated at 4°C in a 1:200 dilution of rabbit polyclonal antibody against human Stat3 (Phospho-Stat3 [tyr-705]; Cell Signaling Technology, Beverly, MA), or a 1:100 dilution of rabbit polyclonal antibody against human VEGF (clone A-20, SC-152, Santa Cruz Biotechnology, Santa Cruz, CA). The samples were then rinsed and incubated for 1 hour at room temperature with peroxidase-conjugated anti-rabbit IgG. Next, the slides were rinsed with PBS and incubated for 5 minutes with diaminobenzidine (Research Genetics, Huntsville, AL). The sections were washed thrice with distilled water, counterstained with Mayer's hematoxylin (Biogenex Laboratories, San Ramon, CA), and washed once each with distilled water and PBS. Afterward, the slides were mounted using Universal Mount (Research Genetics) and examined using a bright-field microscope. A positive reaction was indicated by a reddish-brown precipitate in the nuclei (Stat3) or cytoplasm (VEGF; ref. 18).

Stat3 and VEGF staining were classified as negative, weak positive, or strong positive according to the percentage of positive cells and staining intensity. Scores for percentage of positive cells were assigned as follows: ≤10% of cells positive, 0; 11% to 25% of cells positive, 1; 26% to 50% of cells positive, 2; 51% to 75% of cells positive, 3; and >75% of cells positive, 4. Scores for staining intensity were assigned as follows: no staining, 0; light brown, 1; brown, 2; and dark brown, 3. Overall scores were obtained by multiplying the percentage score by the intensity score. Overall scores ≤3 were defined as negative, overall scores >3 but ≤6 were defined as weak positive, and overall scores >6 were defined as strong positive. Two independent pathologists examined 5 random fields (1 field = 0.159 mm², at ×100 magnification) of each sample, and scored each sample without knowledge of patient outcome (double-blinded). An average value of the two scores was presented in the present study (23).

Quantification of Tumor Microvessel Density. For CD34 staining, tissue sections were processed and stained with a 1:100 dilution of monoclonal goat anti-CD34 (PECAM1-M20; Santa Cruz Biotechnology), stained with peroxidase-conjugated anti-goat IgG, and then counterstained with Mayer's hematoxylin (Biogenex Laboratories). The slides were mounted and examined using a bright-field microscope. A positive reaction was indicated by a reddish-brown precipitate in the cytoplasm. For quantification of tumor MVD, highly vascular areas were initially identified by scanning tumor sections using light microscopy at low power. Vessels count was assessed in areas of the tumor containing the highest numbers of capillaries and small venules, based on the criteria of Weidner et al. (22). Vessels in five high-power fields (×200 magnification [×20 objective and ×10 ocular]) were counted by two independent investigators without knowledge of the patient outcome (double-blinded). An average value of the two scores was presented in the present study MVD was divided into three groups: low (<50 vessels per five high-power fields), moderate (50-100 vessels), and high (>100 vessels) as described previously (23).

Western Blot Analysis. Whole cell lysates were prepared from human gastric cancer cell lines and human normal and gastric cancer tissue specimens. Standard Western blotting was done using a polyclonal rabbit antibody against activated human Stat3 (Phospho-Stat3 [tyr-705]; Cell Signaling Technology), a polyclonal rabbit antibody against human VEGF; and anti-rabbit IgG, a horseradish peroxidase-linked F(ab′)₂ fragment obtained from a donkey (Amersham Life Sciences, Arlington Heights, IL). Equal protein sample loading was monitored by incubating the same membrane filter with an anti-β-actin antibody. The probe proteins were detected using the Amersham enhanced chemiluminescence system according to the manufacturer's instructions (23).

Stable Dominant-Negative Stat3 and Constitutive Active Stat3 Transfection. Human gastric cancer AGS cells were transfected with dominant-negative Stat3 (Stat3-DN), constitutive active Stat3 (Stat3-CA), green fluorescent protein expression vector (GFP), and vector alone (control) using LipofectAMINE (Life Technologies, Inc., Rockville, MD). Stat3-CA was generated by substitution of the cysteine residues within the COOH-terminal loop of the SH2 domain of Stat3 (18, 19, 21). Stat3-DN construct was produced by a phenylalanine substitution of the tyrosine residue position at 705 resulting in a reduction of tyrosine phosphorylation of wild-type Stat3 and inhibition of endogenous Stat3 (18, 19, 24). Successful stable transfection was verified using anti-tag antibodies as described previously (18). Three independent clones of pStat3-CA (CA1, CA2, and CA3) and pStat3-DN transfection (DN1, DN2, and DN3) were used in this study. The dominant-negative Stat3 and control vector expression constructs were provided by Dr. Koichi Nakajima. (Osaka City University Graduate School of Medicine, Osaka, Japan).

Northern Blot Analysis. Total RNA was extracted using TRIzol reagent (Invitrogen, San Diego, CA). RNA (12 μg) was separated electrophoretically on a 1% denaturing formaldehyde agarose gel, transferred to a GeneScreen nylon membrane (DuPont, Boston, MA) in 20× SSC, and UV cross-linked using a UV-Stratalinker 1800 (Stratagene, La Jolla, CA). Additionally, the VEGF probe was labeled with [³²P]-dCTP using a random labeling kit (Boehringer Mannheim Biochemicals, Indianapolis, IN). Equal loading of RNA samples was monitored by hybridizing the same membrane filter with a human β-actin cDNA probe (18).

Vascular Endothelial Growth Factor Protein Measurement. The VEGF protein levels in the culture supernatants were determined using the Quantikine VEGF ELISA kit (R&D Systems, Minneapolis, MN), which is a quantitative immunometric sandwich enzyme immunoassay. A curve of the absorbance of VEGF versus its concentration in the standard wells was plotted. By comparing the absorbance of the samples with the standard curve, we determined the VEGF concentration in the unknown samples (18).

Endothelial Cell Tube Formation Assay. The tube formation assay was done as described previously (25). Briefly, 250 μL of growth factor-reduced Matrigel (Collaborative Biomedical Products, Bedford, MA) were pipetted into each well of a 24-well plate and polymerized for 30 minutes at 37°C. Human umbilical vein endothelial cells were harvested after trypsin treatment and suspended in conditioned medium from either 1 × 10⁶ AGS-Stat-DN cells or 1 × 10⁶ AGS-Neo cells cultured for 48 hours in modified Eagle's medium containing 1% fetal bovine serum. Then 2 × 10⁴ human umbilical vein endothelial cells in 300 μL of conditioned medium were added to each well and incubated at 37°C, 5% CO₂, for 20 hours. The cultures were photographed with bright-field microscopy using a Sony digital camera equipped with an Optimas 6.2 program. The degree of tube formation was assessed as the percentage of cell surface area versus total surface area. Control cell cultures were given arbitrary percentage values of 100 (25).

Statistical Analysis. The two-tailed χ² test was done to determine the significance of the difference between the covariates. Survival durations were calculated using the Kaplan-Meier method. The log-rank test was used to compare cumulative survival of patient groups. A Cox proportional hazards model was used to provide univariate and multivariate hazard ratios for the study variables. The expression levels of Stat3 and VEGF, MVD, age, sex, Lauren's histologic classification, stage (American Joint Committee on Cancer system), and completeness of surgical resection (R0 versus R1 or R2) were included in the model. In all of the tests, P < 0.05 was defined as statistically significant. SPSS version 11.05 (SPSS, Inc., Chicago, IL) was used for analyses.

Patient Characteristics. Eighty-six cases were selected to represent all stages and histologic types of malignant gastric cancer. The study included tissue samples from 56 men and 30 women. The mean patient age was 62 years. Proximal cancer localization was observed in 20 cases. Fifty-three patients had intestinal type and 33 had diffuse type cancer. Details of patient characteristics are provided in Table 1.

Signal Transducer and Activator of Transcription 3 Expression and its Association with Vascular Endothelial Growth Factor Expression and Microvessel Density in Human Gastric Tumor Tissue. Stat3 expression and MVD were evaluated by immunohistochemistry in the primary cancer tissue of 86 patients. Strong Stat3 expression was observed in 20 cases (23%). Stat3 expression was classified as weak or negative in 43 cases (50%) and 23 cases (27%), respectively. No significant differences were detected between the three Stat3 expression categories in the distribution of gender, race or ethnicity, tumor location, type of resection, residual disease status, extent of lymphadenectomy, or Lauren's histologic classification (Table 1). High MVD was observed in 31 cases (36%). Low or moderate MVD was observed in 45 cases (52%) and 10 cases (12%), respectively. Stat3 expression significantly correlated with VEGF expression and MVD. This finding was further confirmed by analysis of consecutive tissue sections, in which the Stat3 expression pattern was consistent with patterns of VEGF expression (Fig. 1), and by Western blot analysis of Stat3 and VEGF protein expression (Fig. 2). Therefore, these data provided clinical evidence to support our hypothesis that aberrant Stat3 activation causes increased angiogenesis through the overexpression of various angiogenic molecules such as VEGF.

Effects of Signal Transducer and Activator of Transcription 3 Expression, Vascular Endothelial Growth Factor Expression, and Microvessel Density on Patient Survival. Median survival durations according to Stat3 and VEGF expression are presented in Table 2. Elevated Stat3 expression, but not increased VEGF, was associated with inferior survival (Table 2). The median survival durations in patients with low, moderate, and high MVD were 64, 37, and 14 months, respectively. The effect of MVD expression on survival duration was statistically significant (Table 2). Other variables that affected survival in univariate analyses included stage and completeness of resection (Table 2). Age at diagnosis (as a continuous variable by univariate Cox proportional hazard analysis) and Lauren's histologic type did not have a statistically significant affect on survival.

Next, Stat3 expression, VEGF expression, MVD, stage, completeness of resection, age, and Lauren's histologic classification were entered into a Cox proportional hazards model for multivariate analysis (Table 2). After adjustment for the effect of covariates, strong Stat3 expression [P = 0.049; hazard ratio, 2.28; 95% confidence interval, 0.99-4.58] and advanced stage (P < 0.01) were independent predictors of poor survival. In multivariate analyses, trends toward poorer survival were observed among patients who had high MVD and strong VEGF expression. However, these affects on survival were not statistically significant.

Suppression of Gastric Cancer Angiogenesis by Blockade of Activated Signal Transducer and Activator of Transcription 3 Expression in Human Gastric Cancer Cells. To provide direct evidence of whether Stat3 regulates the angiogenic phenotype of gastric cancer, AGS human gastric cancer cells were stably transfected with pStat3-DN expression vector as we previously reported (18, 19). We found that AGS cells transfected with pStat3-DN (AGS-DN1, AGS-DN2, and AGS-DN3), but not AGS cells transfected with control vector (AGS-Neo1 and AGS-Neo2), exhibited a decreased expression of VEGF as determined using Northern blotting (Fig. 3A) and ELISA (Fig. 3B). Furthermore, consistent with decreased expression of VEGF, the supernatant from pStat3-DN-transfected (DN1 and DN2) cells seemed to be less angiogenic than that from control vector-transfected (Vector) cells as determined by an endothelial cell tube formation assay (Fig. 3C). Addition of exogenous VEGF to the conditioned medium at least partially overcame the Stat3-DN inhibition of tube formation (Fig. 3C). The representative pictures were taken in situ for tube formation in the supernatant of control transfected AGS cells (Fig. 3D,1), Stat3-DN-transfected AGS cells (Fig. 3D,2) and Stat3-DN-transfected AGS cells with the addition of 1 ng/mL VEGF (Fig. 3D 3). We found that Stat3-DN suppressed tube formation, which was partially overcome by addition of exogenous VEGF. These results suggest that blockade of Stat3 activity suppresses VEGF expression and impairs the angiogenic phenotype of gastric cancer cells.

In contrast, AGS cells transfected with pStat3-CA (AGS-CA1, AGS-CA2, and AGS-CA3), but not AGS cells transfected with control GFP expression vector (AGS-GFP), exhibited an increased expression of VEGF as determined using Northern blotting (Fig. 4A) and ELISA (Fig. 4B). Furthermore, consistent with increased expression of VEGF, the supernatant from pStat3-CA-transfected (AGS-CA) cells seemed to be more angiogenic than that from control vector-transfected (AGS-GFP) cells as determined by an endothelial cell tube formation assay (Fig. 4C).

Angiogenesis is essential for tumor growth and metastasis (26). Numerous lines of evidence have shown that angiogenesis, as quantitated according to MVD, plays significant clinicopathologic roles in cancer patients (22, 27). Notably, MVD can be an independent, highly significant, accurate prognostic indicator in cancer patients (22). In the present study, we found direct clinical evidence of a strong correlation between Stat3 expression and VEGF expression and MVD in human gastric cancer. Although VEGF, MVD, and Stat3 were all associated with poor patient survival, Stat3 was apparently more powerful than VEGF and MVD in predicting clinical outcome of gastric cancer patients as indicated by multivariate analysis showing that only Stat3 was an independent prognostic factor.

Tumor angiogenesis has been considered the most important predictor of overall survival in gastric cancer (6–10). Moreover, some key angiogenic factors have also been indicated to be independent prognostic factors in gastric cancer (3, 11–15). However, the clinicopathologic role of MVD and its major regulatory molecules in gastric cancer development and progression remain largely unclear. Angiogenesis is regulated by multiple factors, including VEGF (6–10), and many of these factors may individually predict MVD. For example, VEGF, which is considered critical to angiogenesis, has a close association with MVD (3, 11–15). The present study showed an overall close correlation between VEGF and MVD, which reflects the importance of VEGF in tumor angiogenesis and is clearly consistent with the earlier findings. However, expression of VEGF and its receptors has not correlated with MVD or malignancy in many other types of tumors, including digestive endocrine tumors (28). Presumably, different types of tumors may use angiogenic factors other than VEGF to induce vascular formation. Many lines of evidence from studies of gastric cancer and other tumor types indicate a relationship between MVD and other angiogenic molecules (4, 12, 13). Therefore, MVD should have better prognostic value than any particular angiogenic molecule, because increased MVD might result from overexpression of any major angiogenic factors. In the present study, Stat3 seemed to be powerful predictor of MVD. However, more studies are needed to determine whether Stat3 is a better MVD-predicting factor than individual angiogenic factors, considering that Stat3 is a critical transcription factor and may control the expression of many genes key to tumor angiogenesis, such as VEGF (17, 18).

Studies on the association between MVD and prognosis in gastric cancer have produced inconsistent results; some have found that MVD is a predictor of poor prognosis, whereas others have found that MVD predicts good prognosis (11, 13, 15, 29–35). In the present study, overall survival was significantly lower in patients with a high MVD than in patients with a low MVD, which is consistent with several previous reports (11, 14, 32, 33, 36, 37). However, multivariate analysis revealed that MVD was not an independent prognostic indicator, whereas Stat3 was. This superiority of Stat3 may underline the fact that prognosis is directly related to tumor biology manifested by its multiple aspects, including the angiogenic phenotype, invasive capacity, apoptosis resistance, and proliferation. Thus, an elevated MVD may not necessarily predict all other aspects of cancer biology (i.e., the overall biological behavior), besides the angiogenic phenotype, whereas Stat3 might most likely do so, because it is an important transcription factor for many genes that may regulate all aspects of cancer biology (20, 38). However, more studies are needed to substantiate whether that the Stat3 expression and activation status is a powerful and practical predictor of patient outcome.

Furthermore, Stat3 may be not only a useful molecular marker for selecting patients with a poor prognosis to receive more aggressive preoperative or adjuvant therapy in the setting of a clinical trial but also an effective therapeutic target for gastric cancer. In gastric cancer, multiple growth factor pathways are involved though multiple tyrosine kinases, such as epidermal growth factor receptor, HER2, VEGF receptor-1, VEGF receptor-2, platelet-derived growth factor receptor, and c-Met. Inhibition of one such pathway may not be sufficient for antitumor activity. With its central regulatory role in many of these pathways (20), Stat3 represents an attractive target for the development of effective therapies as evidenced by our recent studies in human pancreatic cancer and melanoma (18, 19).

Our current study provided the first evidence for a critical role of activated Stat3 in gastric cancer angiogenesis. Since activated Stat3 up-regulates matrix metalloproteinase-2 (19), cyclin D1 (21), c-myc, Bcl-xL, and Mcl-1 (17, 39), but down-regulates death receptor Fas (40), Stat3 may enhance tumor progression inclusively by affecting the expression of various genes related to cell survival, the cell cycle, invasion, and angiogenesis (39–43). Further exploring those molecular mechanisms that result in Stat3 overactivation may not only shed more light on abnormal Stat3 activation but also help improve understanding of Stat3's value as a prognostic factor and aid in the development of effective therapies targeting Stat3.

In summary, we discovered that the level of Stat3 expression in gastric cancer was directly related to VEGF expression level and MVD, which are closely related to the postoperative prognosis for gastric cancer. This study further showed that abnormally activated Stat3 expression represents a potential risk factor for poor prognosis and directly contributes to gastric cancer angiogenesis and progression. Therefore, preoperative determination of the level of Stat3 activity may be useful in deciding on the modality and extent of postoperative therapy. We are currently investigating the molecular mechanism by which the Stat3 signaling pathway regulates gastric cancer development and progression and whether this pathway is a therapeutic target for controlling gastric cancer growth and metastasis.

We thank Lydia Soto for help in the preparation of this article and Stephanie P. Deming for editorial comments.