STAT3, p-STAT3 and HIF-1α are associated with vasculogenic mimicry and impact on survival in gastric adenocarcinoma

Vasculogenic mimicry (VM) formation is important for invasion and metastasis of tumor cells in gastric adenocarcinoma (GAC). The present study aimed to investigate the association between signal transducer and activator of transcription-3 (STAT3), phosphor-STAT3 (p-STAT3), hypoxia-inducible factor-1α (HIF-1α) and VM formation in GAC, and discuss their clinical significance and correlation with the prognosis of patients with GAC. The expression levels of STAT3, p-STAT3, HIF-1α and VM were assessed in 60 cases of patients with GAC and 20 cases of patients with gastritis on tissue microarrays by immunohistochemical methods. The expression levels of STAT3, p-STAT3, HIF-1α and VM were higher in patients with GAC (particularly in poorly differentiated GAC) than in those with gastritis (P<0.05). The expression levels of STAT3, p-STAT3 and HIF-1α were higher in VM tissues compared with non-VM tissues (P<0.05). Positive correlations existed between STAT3, p-STAT3, HIF-1α and VM expression (P<0.05). The expression levels of STAT3, p-STAT3 and HIF-1α, VM, status of lymph node metastasis and tumor differentiation degree were associated with the overall survival time of patients with GAC (P<0.05). However, only p-STAT3 and VM expression were identified as the independent risk factors of GAC OS when analyzed with multivariate analysis. p-STAT3 and VM play a significant role in indicating the prognosis of patients with GAC. STAT3 activation may play a positive role in VM formation of GAC by the STAT3-p-STAT3-HIF-1α-VM effect axis.


Introduction
A blood supply is essential for the growth and hematogenous metastasis of tumors. Maniotis et al (1) previously reported an angiogenesis-independent pathway known as vasculogenic mimicry (VM). This pathway is a novel phenomenon in which highly aggressive human melanoma cells imitate endothelial cells and form vascular channel-like structures to convey blood plasma and red blood cells without the involvement of endothelial cells. Periodic acid-Schiff (PAS)-positive patterns identify VM channels. Subsequently, VM was identified in lung cancer, hepatocellular carcinoma, gallbladder carcinoma, gastric adenocarcinoma (GAC) and other types of cancer (2)(3)(4)(5). A study by Li et al (5) described the expression of VM in GAC, particularly in poorly differentiated GAC. VM may play an extremely significant role in the biological behavior of multiple tumors (2)(3)(4)(5)(6). However, establishing the detailed mechanism of VM formation is required.
It has been reported that positive expression of hypoxia-inducible factor-1α (HIF-1α) is associated with the formation of VM in primary gallbladder, non-small cell lung cancer and hepatocellular carcinoma (2)(3)(4). STAT3 modulates the stability and activity of HIF-1α, and activated STAT3 increases the HIF-1α protein level by increasing HIF-1α stability through blocking HIF-1α degradation and accelerating its de novo synthesis (7). Pawlus et al (8) found that STAT3 exhibited specific binding to the promoters of HIF1 or HIF2 target genes respectively, even when overexpressed, and STAT3 interacted with HIF-1α to activate HIF1 target gene promoters. Taking into consideration the aforementioned details, STAT3 activation is possibly associated with VM formation. Therefore, investigating the association between STAT3 and VM formation in GAC is worthwhile to learn more about tumor development, invasion and metastasis.
In the present study, the expression levels of STAT3, p-STAT3, HIF-1α and VM were explored simultaneously for the first time. Firstly, the existence of VM in GAC was confirmed by a cluster of differentiation 31 (CD31)/PAS double-staining method. Subsequently, combining VM existence with the expression levels of STAT3, p-STAT3 and HIF-1α, the association was assessed between them and the possible formation mechanism of VM was investigated. Additionally, prognosis was assessed by Kaplan-Meier survival analysis for univariate analysis and by Cox proportional hazards model for multivariate analysis.

Subjects.
A total of 80 cases of paraffin-embedded specimens were collected in the Department of Pathology at the Qilu Hospital of Shandong University (Jinan, China). These cases included 60 GAC specimens (46 male and 14 female patients; median age, 60.0 years) and 20 gastritis specimens (11 male and 9 female patients; median age, 56.2 years). Primary gastric cancer in these patients was diagnosed and treated at the Qilu Hospital between July 2005 and December 2006. The patients with GAC had well-documented clinical histories and follow-up information. None of the patients underwent preoperative chemotherapy and/or radiation therapy. The follow-up time ranged between 6 and 72 months until July 2012, although the follow-up data of one case was lost. Overall survival (OS) time was defined as the interval between the dates of surgery and mortality. The gastritis cases were derived from gastritis biopsy specimens. All the cases were reviewed by two highly qualified pathologists. The study was approved by the ethics committee of Shandong University School of Medicine (Jinan, China) and written informed consent was obtained from the patients or their family.
Construction of the tissue microarray. A tissue microarray instrument (HT-1 type; Hengtai Technology Development Co., Ltd., Chaoyang, China) was used to construct a blank receptor wax block of six rows and seven columns. Marked and collected tissues from the paraffin-embedded specimens were inserted into the holes of the receptor wax block. From each case, two specimens were acquired to overcome the loss of tissue. The first two holes on the first line were filled with ash, which served as a 'blank' specimen-positioning reference. Each receptor wax block accommodated 40 specimens, which represented a total of 20 cases. The GAC specimens were built into the three tissue microarrays. Each was subjected to repeated wax melting at 56˚C to become a whole specimen. The tissue microarrays and gastritis tissue specimens were sectioned into 4-µm-thick slices that served as a continuous backup source.

Immunohistochemical analysis.
A positive result of immunohistochemical staining is characterized by the existence of yellow-to-brown granules. The positive staining of STAT3 was mainly located in the cytoplasm and partly in the nuclei, while the positive staining of p-STAT3 and HIF-1α was mainly located in the nuclei and partly in the cytoplasm. There were two factors that determined the final outcomes: The staining intensity observed under microscope (BX53, OLympus, Tokyo, Japan) and the proportion of positive cells estimated in an average of 100 cells counted in 10 high-magnification fields. The staining intensity was subjected to the following numerical scoring: Specimens were colorless, 0 points; pale yellow, 1 point; yellow, 2 points; or brown, 3 points. The proportion of positive cells was scored as follows: The number of positive cells was <5%, 0 points; 5-25%, 1 point; 26-50%, 2 points; 51-75%, 3 points; and >75%, 4 points. Immunostaining was considered positive when the product of the two types of scores was multiplied and was ≥4 (10). VM was identified in GAC tissues by CD31/PAS double-staining. VM, characterized by CD31-negative/PAS-positive vascular-like patterns and the presence of red blood cells, was formed by GAC cells, while typical blood vessels showed CD31-positive/PAS-negative in their vascular wall. All sections were scored blindly by two independent observers.
Statistical analysis. The statistical analysis was performed with the SPSS Graduate Park 19.0 software (SPSS, Inc., Chicago, IL, USA). The count column was analyzed by the χ 2 test. For the correlation analysis of STAT3, p-STAT3, HIF-1α and VM expression, Spearman's rank correlation test was applied; whereas for the survival analysis, the Kaplan-Meier method and Cox regression analysis were applied. P<0.05 was considered to indicate a statistically significant difference.

Results
STAT3, p-STAT3, HIF-1α and VM expression in GAC and gastritis tissues. VM ( Fig. 1A and B, arrow), characterized by CD31-negative/PAS-positive channels, and containing red blood cells, was only found in GAC specimens (31.7%; P<0.05). In the vascular wall of typical blood vessels from gastritis specimens, only CD31-positive/PAS-negative staining (Fig. 1C, arrow) was found instead of VM formation. STAT3-positive expression ( Fig. 2A and D) was detected mainly in the cytoplasm and partly in the nuclei of GAC tissue cells. p-STAT3- (Fig. 2B and E) and HIF-1α-positive expression (Fig. 2Cand F) was detected mainly in the nuclei and partly in the cytoplasm of GAC tissue cells.
Survival analysis of STAT3, p-STAT3, HIF-1α and VM. Using Kaplan-Meier univariate analysis, six factors were found to have statistically significant associations with the OS time of

A B C
patients with GAC following curative surgery, including STAT3, p-STAT3 (Fig. 3A), HIF-1α, VM (Fig. 3B), status of lymph node metastasis and degree of differentiation (P<0.05). In addition, VM combined with STAT3, p-STAT3 or HIF-1α, respectively, was also found to have statistically significant associations with the OS time of patients with GAC. Patients with p-STAT3-and VM-negative expression were more likely to have a longer median OS time compared with those with p-STAT3-and (or) VM-positive expression (P<0.05) (Fig. 3C and Table III). All the aforementioned six variables were analyzed by a multivariate Cox proportional hazards model (forward stepwise procedure). In this model, VM (HR, 3.021 and P=0.001), and p-STAT3 (HR, 2.520 and P= 0.006) showed significant correlations with the OS times of patients with GAC following curative surgery, which indicated that VM and p-STAT3 were the independent risk factors of the OS time of patients with GAC (Table IV).

Discussion
Recently, various factors have been studied to reveal the mechanism of VM formation. VM is considered to play a key role in tumor growth, progression and metastasis (5,6). Li et al (5) found the existence of VM in GAC, and that hypoxia may Table II. Correlation between STAT3, p-STAT3, HIF-1α and VM expression in GAC.   participate in VM formation of GAC, particularly in poorly differentiated GAC. In the present study, it was found that VM was detected only in GAC specimens, particularly in the poorly differentiated GAC tissues. Patients with VM formation had a significantly shorter median OS time than those without VM formation (P<0.001). By multivariate survival analysis, VM was found to be an independent risk factor of the OS time of patients with GAC. Therefore, VM was indicated to be a detective marker of GAC tissues. The most significant difference in the microenvironment between tumor and normal tissues is ischemia of the tumor due to structural imperfections of the tumor vessels, which induces to anoxia of tumor tissues (11). As a hypoxia-dependent protein, HIF-1α can be rapidly degraded when oxygen is normal, but when oxygen is not sufficient, it can upregulate cell proliferation at the transcription level, activate the expression of numerous hypoxia response genes, and be closely associated with energy metabolism, angiogenesis, infiltration and metastasis of the tumor by binding with the hypoxia response element of the hypoxia response (12). As a tumorigenesis factor, HIF-1α could induce angiogenesis of lung cancer when activated in hypoxia (13). In the present study, it was found that HIF-1α-positive expression was significantly increased in GAC specimens, particularly in VM GAC specimens, compared with the gastritis specimens (P<0.05). Similarly, the HIF-1α-positive expression was positively associated with VM formation (r=0.480 and P=0.001). These demonstrated that HIF-1α was a positive index of VM formation in GAC tissues. Patients with HIF-1α-and VM-positive expression were more likely to have a shorter median of OS compared with those with HIF-1α-and (or) VM-negative expression by survival analysis (P<0.05). Therefore, we propose that the phenomenon of HIF-1α-VM double-positive expression is a more promising index of prognosis than that of HIF-1α-or VM-positive expression.
Xu et al (15) demonstrated that HIF-1 expression induced by Src was inhibited when blocking STAT3 signaling in breast cancer and melanoma cell lines. STAT3 converted to p-STAT3, and p-STAT3 directly bound HIF-1α and upregulated HIF-1α stability through delaying protein degradation and accelerating protein synthesis (7). STAT3 can promote HIF-1α transcription and increase HIF-1α protein stability by inhibiting the expression of p53 (8,16,17). All these results reveal that STAT3 is a positive factor of HIF-1α. Furthermore, p-STAT3 can upregulate the expression of matrix metalloproteinase 2 (MMP2) to promote the formation of VM in tumor tissues (6,(18)(19)(20). Hypoxia is a possible mechanism of VM genesis by the induction of the expression of HIF-1α, MMP-2 and MMP-9 (6,19,20). Above all, the results of the present study concluded that STAT3 activation could upregulate and stabilize the expression of HIF-1α by various pathways intending to promote the VM formation under hypoxic conditions.
The results of the present study showed that STAT3-and p-STAT3-positive expression was increased in the VM group (P<0.05). Additionally, STAT3 expression was positively correlated with p-STAT3, HIF-1α and VM expression, respectively, in GAC tissues. By univariate and multivariate survival analysis, patients with both negative expression of p-STAT3 and VM were found to be more likely to have a longer median OS time compared with those with p-STAT3-and (or) VM-positive expression (P<0.05), and p-STAT3 was an independent risk factor of the OS time of patients with GAC. These indicated a specific type of association between STAT3, p-STAT3, HIF-1α and VM in GAC tissues.
Combining the aforementioned studies with the results of the present study, it was deemed that STAT3 may be a novel positive factor of VM formation in GAC tissues through the effect of p-STAT3. STAT3 and p-STAT3 were positive factors of HIF-1α expression and VM formation in GAC tissues. STAT3 was significantly associated with progression and prognosis of GAC. Combining the previous studies with the present study results, it can be concluded that STAT3 may promote the formation of VM to affect the invasion and metastasis in GAC tissue by a specific type of mechanism (STAT3-p-STAT3-HIF-1α-VM).
In conclusion, it was found that p-STAT3 and VM played a significant role in indicating the prognosis of patients with GAC. STAT3 activation may play a positive role in VM formation of GAC tissues by the STAT3-p-STAT3-HIF-1α-VM effect axis. These results provide opportunities to develop potential novel therapeutic targets for GAC.