Epithelial-Mesenchymal Transition Phenotype, Metformin, and Survival for Colorectal Cancer Patients with Diabetes Mellitus II

Objectives We aimed to explore the association between metformin treatment and epithelial-mesenchymal transition (EMT) phenotype and further appraise the prognostic values of metformin and EMT markers E-cadherin and vimentin for colorectal cancer (CRC) in clinical practice. Methods We collected specimens and evaluated clinicopathological parameters of 102 stage I to III CRC patients with prediagnosed type 2 diabetes mellitus (DM II). Expression of E-cadherin and vimentin in tumors was detected by immunohistochemistry (IHC), and statistical analysis was performed using SPSS 19.0. Results In correlation tests, we found a lower tumor cell EMT degree (more E-cadherin (P = 0.014) and less vimentin (P = 0.011) expression in patients who used metformin, and the expression of E-cadherin and vimentin was associated with serum CA19-9 (P = 0.048, P = 0.009), tumor invasive depth (T) (P < 0.001, P = 0.045), and lymph invasion (N) (P = 0.013, P = 0.001). In Cox multivariate regression analysis, E-cadherin was identified as a prognostic factor for disease-free survival (DFS) (P = 0.038) and metformin use (P = 0.015P = 0.044) and lymph invasion (P = 0.016P = 0.023) were considered as the prognostic factors for both DFS and overall survival (OS). Conclusion Our study suggested that metformin may impede the EMT process and improve survival for stage I–III CRC patients with DM II.


Introduction
Colorectal cancer (CRC) is the third most common cancer and the fourth leading cause of cancer-related death worldwide [1]. With proper treatments, CRC patients in the early stages have been able to achieve long diseasefree survival (DFS). Several studies have indicated that CRC and many other cancer patients with diabetes mellitus (DM) tend to have a worse prognosis than patients without DM [2][3][4][5]. Over the last decade, metformin, as a classical oral hypoglycemic agent, has been discovered to possess antineoplastic activities [6,7] and has been proven effective in survival improvement for CRC patients with type II DM (DM II) [8][9][10][11][12].
Epithelial-mesenchymal transition (EMT) is a process by which epithelial cells lose epithelial characteristics, such as polarity and adhesion, gain migratory properties, and transform into mesenchymal cells [13]. EMT is widely occurring in embryonic development, tissue regeneration, fibrosis, and cancer [14]. EMT has been found to be associated with invasion, metastasis, recurrence, and drug resistance in cancer [15,16]. Based on a series of antineoplastic evidence from previous reports, the anti-EMT activity of metformin was also tested in a few studies. With metformin exposure, EMT inhibition was observed in breast cancer, melanoma, prostate cancer, lung cancer, and thyroid cancer cells [17][18][19][20][21][22][23][24]. However, all the abovementioned anti-EMT studies were performed using cell lines in vitro, and there is still little analogous research for CRC EMT.
Here, we aim to explore the association between metformin use and EMT degree for CRC patients with DM II in a clinical setting. In our study, clinicopathologic parameters were evaluated and the expression of EMT markers E-cadherin and vimentin in tumor tissue was detected by immunohistochemistry (IHC). Conclusions were based on the outcomes from statistical analysis including correlation tests and survival modeling for variables.

Materials and Methods
2.1. Patients. Patients with DM II, diagnosed pathologically with stage I to III colorectal cancer who underwent radical surgery in Qilu Hospital of Shandong University between 2008 and 2012, were eligible for our study. The clinical parameters of patients were obtained from the electronic medical record (EMR) system and case follow-ups. The exclusion criteria were set as follows: (1) secondary primary malignant tumor; (2) inflammatory bowel disease (IBD) or Peutz-Jeghers syndrome; (3) diagnosed with DM II less than 1 year before tumor diagnosis [25]; (4) administered metformin less than 6 months before tumor diagnosis [2]; (5) pathologic type other than adenocarcinoma; (6) accepted anticancer treatment before surgery; (7) did not reach R0 resection; (8) and incomplete medical records. The primary tumor was evaluated by the tumor-nodemetastasis (TNM) staging system (American Joint Committee on Cancer (AJCC), version 7.0), and therapy outcome was assessed by the Response Evaluation Criteria in Solid Tumor (RECIST, version 1.1).

Specimen and IHC.
The paraffin-embedded tumor tissues from radical surgery were cut into slices. After dewaxing and dehydrating, slides were subjected to antigen retrieval by citric acid treatment (pH 6.0) and microwave. Slides were then treated with H 2 O 2 to block endogenous peroxidase activity and subsequently incubated with 5% bovine serum albumin phosphate-buffered saline (BSA-PBS) solution, diluted monoclonal rabbit anti-E-cadherin (Cell Signaling Technology, USA), and antivimentin antibody (Cell Signaling Technology, USA) at 4°C overnight. The poly-horseradish peroxidase (Poly-HRP) immunoglobulin G (IgG) detection kit (ZSGB-BIO, China P. R.) and DAB kit (ZSGB-BIO, China P. R.) were used for detection and staining. The DAB-stained slides were counterstained with hematoxylin, dehydrated with alcohol, and cleared with xylene. Slides of normal colon and rhabdomyosarcoma tissue were also tested for the positive controls of E-cadherin and vimentin, respectively. BSA-PBS solution (5%) without antibody was used for negative controls.

Statistical Analysis.
The primary endpoint of the study was overall survival (OS). Student's t-test was used for continuous variable analysis, and Pearson's chi-squared (χ 2 ) test was used for comparison of categorical variables. Survival data were analyzed using the Kaplan-Meier method and compared by log-rank statistics. Cox regression was used for multivariate survival analysis. Outcomes with a P < 0 05 were considered significant. All statistical analyses were performed with SPSS 19.0 software.

Ethics
Statement. This study was approved by the Ethics Committee of Qilu Hospital, and all patients provided informed consent.

Patient Characteristics.
A final total of 102 of 115 patients were included in the study (7 patients were excluded because the DM II diagnosis date was less than 6 months before surgery, 2 patients were excluded for receiving presurgery chemotherapy, and 4 patients were lost to follow-up). The median duration of follow-up was 1678 days. Clinicopathologic parameters of the cohort, such as age, gender, smoking and drinking history, family history, tumor location, staging, invasive depth, lymph invasion, and E-cadherin and vimentin expression are shown in Table 1. Of the total patients, 28 (27.45%) were metformin users, 49 (48.04%) were detected as positive for E-cadherin expression, and 55 (53.92%) were detected positive for vimentin expression (representative pictures in Figure 1).

Association between Metformin Use, E-cadherin and
Vimentin Expression, and Clinicopathologic Parameters. As presented in Table 1, insulin injection (P = 0 018) and Ecadherin (P = 0 014) and vimentin (P = 0 011) expression were significantly associated with metformin use. Using the Mann-Whitney U test for IHC scores (Figure 2), we further observed trends that patients using metformin expressed higher levels of E-cadherin (P < 0 001) and lower levels of vimentin (P = 0 001) than those without metformin use. Patients with metformin treatment showed less lymph invasion (P = 0 041). Significant relationships were not observed with gender, age, body mass index (BMI), tumor markers, and other variables except family malignant history (P = 0 016).
The prognostic values of variables were further tested by multivariate Cox regression (Table 4). Metformin use (P = 0 015, P = 0 044) and lymph invasion (P = 0 016, P = 0 023) were confirmed as significant independent prognostic factors for both DFS and OS. E-cadherin expression (P = 0 038) was considered to be a prognostic factor only for DFS. Alcohol drinking (P = 0 035) was determined to be a risk factor for DFS, whereas tumor invasive depth (P = 0 066, P = 0 108) and vimentin expression (P = 0 369, P = 0 900), which were significant in Kaplan-Meier analysis, were not identified as prognostic factors in the multivariate model.

Discussion
Our study explored the association between metformin use and EMT marker (E-cadherin and vimentin) expression in 102 CRC patients with DM II and further estimated the prognostic values of metformin, EMT markers, and other clinicopathologic parameters.
As common molecular markers for EMT studies, Ecadherin performs cell to cell adhesive functions for the epithelium and is conventionally used to measure cell epithelization [26]; vimentin is a major cytoskeletal component for mesenchymal cells and is routinely applied in identifying mesenchyme [27]. Correlation analysis in our study revealed the expression discrepancy of E-cadherin and vimentin between the metformin group and the nonmetformin group: patients in the metformin group showed higher E-cadherin and lower vimentin expression in tumors, indicating lower proportions of EMT cells and lower EMT degree of malignant tissues. The results suggested that metformin use may impede the EMT process of CRC. Among numerous studies on metformin's antiplastic mechanisms, EMT inhibition was first reported in MDA-MB-231 and transforming growth factor beta-(TGF-β-) induced MCF-7 breast cancer cells [17,18]. This effect was discovered with the regulation of the cancer stem cell phenotype and has been further verified in other cell lines by previous studies [19][20][21][22][23][24]. To the best of our knowledge, this is the first study to reveal the anti-EMT potential of metformin in CRC. Although analogous results had not been directly reported in previous laboratory studies, the following clues were adequate to support our conclusion.
According to existing studies, the main mechanism of metformin action in DM II treatment is AMP-activated protein kinase (AMPK) activation [28]. By activating AMPK and downstream messengers, metformin suppresses gluconeogenesis, downregulates blood sugar and insulin levels, and moderates the metabolism of glucose and fat [29]. Similarly, in the anticancer studies, researchers have found that metformin can impede tumor growth by activating the AMPK pathway [30,31]. For human colorectal cell lines HT-29 and drug-resistant cell line LoVo, metformin was observed to activate AMPK and inhibit cell growth [32,33]. Additionally, AMPK pathway activation was considered as an alternative approach to inhibit EMT [34], and metformin has also been proven to perform such inhibition through AMPK activation for melanoma and breast cancer cells [19,20].
Researchers have also confirmed that several other signaling pathways such as TGF-β, JAK-STAT, PI3K/Akt/   In addition to metformin use, our study revealed that normal presurgery serum CEA and CA19-9 levels, tumor invasion beneath the visceral peritoneum (T1-3), and negative lymph invasion (N0) were associated with positive E-cadherin expression, and high presurgery serum CEA levels, insulin injection, invasion through visceral peritoneum (T4), and positive lymph invasion (N1-2) were associated with positive vimentin expression. A few studies also investigated the relationship between CA19-9 and EMT. For CEA and EMT, in one study, we found that soluble CEA molecules can not only enhance colorectal cell growth but also bind to the TGFβ receptor, which is a common trigger for EMT and inhibits TGF-β signaling [42]. This case was in contrary to our results, but we propose that the inhibition effect was not sufficient to block the mainstream of EMT because there was no further supporting evidence. However, in clinical practice for CRC, abnormally high CEA and CA19-9 levels in serum are more common in patients with advanced disease [43,44]. According to previous studies, E-cadherin and vimentin expression was associated with tumor stage, which was in accordance with our results. Yagasaki et al. discovered that E-cadherin and vimentin are associated with axillary metastases in breast cancer [45], and Liu et al. discovered that low E-cadherin expression is correlated with TNM stage for basal-like breast cancer  [46]. Moreover, E-cadherin and vimentin were also determined to be associated with lymph invasion and TNM stage for lung squamous carcinoma and oral squamous carcinoma [47,48]. It should be noted that 3 of the studies mentioned above [45][46][47] also found the correlation between E-cadherin and vimentin expression and tumor histologic grade; however, such correlation was not significant in our study.
Although patients with normal serum CA19-9 showed significantly longer DFS (P = 0 006) and OS (P = 0 005) in the Kaplan-Meier analysis, in Cox regression, neither CEA nor CA19-9 was considered as prognostic factors (Table 4), which did not accord with the common views of biomarkers' clinical values. We attribute the results to the following three aspects: (1) limited sample size; (2) low mortality (36.37%) of local stage (I-III) colorectal cancer in the follow-up period (median: 1678 days); (3) Kaplan-Meier curve and Cox regression were used only for survival data classified by categorical factors or cut-off values. As continuous variables, CEA and CA19-9 were redefined to categorical factors by cut-off values; thus, the original characteristics were inevitably ignored.
The prognostic values of E-cadherin and vimentin have been widely reported in previous studies. E-cadherin and vimentin were found to be prognostic factors for both DFS and OS for lung cancer [47,49] and oral squamous cancer [50]. E-cadherin has also been suggested as a positive indicator for OS for cervical squamous carcinoma [51] and CRC [52]. In our Kaplan-Meier survival analysis, E-cadherin and vimentin expression was demonstrated to be significant for both DFS and OS. However, in further multivariate Cox regression, only E-cadherin was identified as a prognostic factor for DFS. The contradictory results from Cox regression might be attributed to our limited sample size and inclusion criteria, as CRC patients without DM II were excluded from our study. However, the positive result of Kaplan-Meier analysis could still indicate the prognostic potential of E-cadherin and vimentin.
Patients with metformin not only revealed less EMT in tumors but also showed longer DFS and OS than patients who did not choose metformin for DM II treatment. However, it remains controversial whether metformin could bring survival benefits for cancer patients with DM II. For CRC, our results were supported by several studies [8][9][10][11][12] but do not agree with the results of 2 recent studies [53,54]. On the other hand, 6 of 7 studies were retrospective and the remaining study is a meta-analysis [12]. Therefore, the conflicting conclusions need to be verified by a multicenter prospective study with a large cohort.
In summary, combining the result of correlation analysis and survival analysis described above, we suggest that EMT is correlated with CRC progression and can be impeded by metformin. Metformin use and positive E-cadherin expression indicate better prognosis for CRC patients with DM II. Additionally, it is reasonable to believe that the antineoplastic activity of metformin as partially implemented by EMT interference. To further understand the anti-EMT effect of metformin, more inner molecular mechanisms need to be explored in future studies.

Conflicts of Interest
The authors declare that they have no conflict of interests.