Overexpression of ZEB2 at the Invasion Front of Colorectal Cancer Is an Independent Prognostic Marker and Regulates Tumor Invasion In Vitro

The ability of tumor cells to detach from the main tumor bulk and to invade into the surrounding tissue is a crucial step for tumor dissemination and metastasis. At the tumor invasion front, these processes are often associated with loss of an epithelial cell phenotype and acquisition of a mesenchymal cell phenotype (1, 2). By this epithelial-to-mesenchymal transition (EMT), tumor cells are endowed with an increased migratory capacity and an augmented invasiveness (3).

Although epithelial cells undergo EMT, loss of E-cadherin and concomitant expression of distinct mesenchymal markers like vimentin (VIM), cadherin-11 (CDH11), or fibroblast-specific protein (FSP1) play a central role in this reversible transdifferentiation (4–7). Besides to cytokines like TGF-β, extracellular proteins of the tumor microenvironment or hypoxia can induce the downregulation of E-cadherin and the upregulation of mesenchymal markers (8). This process is further mediated intracellularly by several transcription factors such as Snail1, Snail2, zinc finger E-box–binding homeobox 1 (ZEB1) and 2 (ZEB2) and the basic helix-loop-helix transcription factor E47. Theses transcription factors can directly bind the E-cadherin promoter and inhibit E-cadherin transcription (9–13).

Because the process of EMT most probably occurs at the edge of a tumor, we hypothesized an overexpression of a pattern of genes at the invasion front, which are involved in the control of EMT in colorectal cancer. Therefore, we started our study with an expression analysis of EMT-related genes in the tumor invasion front of colorectal liver metastases. These genes included the extracellular matrix-associated genes COL1A1, COL3A1, the mesenchymal markers Vimentin, FSP1 and OB-Cadherin, the transcription factors HIF1α, ZEB1, and ZEB2, EMT-repressing microRNAs (miRNA) miR-141, miR-200a, miR-200b, miR-200c, and MMP3. Among other genes, we could show ZEB2 to be overexpressed in the tumor invasion front on mRNA and protein expression level. On the basis of these findings, we have further investigated the impact of ZEB2 on migration and invasion in vitro in DLD-1 tumor cells. In this study, we show that repression of ZEB2 by siRNA in DLD-1 cells results in a reduced capacity of migration and invasion in vitro. Furthermore, we show the prognostic significance of ZEB2 in a large group of patients with primary colorectal cancer.

Paraffin-embedded samples of primary colorectal adenocarcinomas were included from a consecutive series of 175 patients, who underwent tumor resection between 1996 and 2004 at the Department of General, Visceral, and Transplantation Surgery, University of Heidelberg. Sixty-seven patients died during follow-up by a cancer-related death. Prior to our immunohistochemical study, we selected only patients with sporadic colorectal cancer. Patients with a positive medical history for hereditary nonpolyposis colorectal cancer or familial adenomatous polyposis were excluded for our analysis. Paraffin-embedded and snap-frozen samples of colorectal liver metastases were obtained from 30 patients who underwent curative liver resection between 2005 and 2009 at the Department of General, Visceral, and Transplantation Surgery, University of Heidelberg. The median follow-up time for patients with primary colorectal cancer was 124 months, the median follow-up time for patients with colorectal liver metastases was 32 months. A written informed consent from all patients regarding tissue sampling had been obtained. Our study protocol was approved by the local ethics committee. Table 1 A and B displays the clinical and histopathologic characteristics of the patients.

Paraffin-embedded tumor samples were provided from the tissue bank of the National Center for Tumor Disease (NCT) Heidelberg. Frozen tissue of colorectal liver metastases was taken immediately after resection. Normal liver tissue without evidence of liver metastases was acquired at least 10 cm away from the tumor. One piece of pure liver tissue and 1 piece consisting of metastases and adjacent liver tissue were fixed in Tissue-Tek (Satura Finektek), frozen in liquid nitrogen and stored at −80°C for further preparations.

Tissue preparation and laser microdissection (LMD) were carried out as recently described (14). Briefly, 16 μm sections of frozen tissue were cut using a cryostat (Leica). After staining with cresyl violet according to the Ambion laser capture microdissection (LCM) staining kit protocol, LMD was conducted within 2 hours using LCM equipment (Molecular Machines & Industries or PALM). Normal liver tissue was microdissected at least 10 cm away from the host-tumor interface. The liver invasion front included the 10 cell layers of liver tissue adjacent to the tumor. Likewise, tissue from the tumor invasion front was obtained from the 10 tumor cell layers next to the adjacent liver tissue. Tissue from the tumor center was microdissected at least 100 or more cell layers away from the invasion front.

Total RNA from cell lines and microdissected tissue was extracted employing miRNeasy Mini Kit (Qiagen) following the manual's instructions. RNA quality was evaluated by using an Agilent 2100 Bioanalyzer. Total RNA was reversely transcribed using the miScript Reverse Transcription Kit (Qiagen). Five nanogram of the resulting cDNA was further used for quantification by qPCR (SYBR Green PCR Kit, Qiagen) in a Roche Light Cycler (Roche Diagnostics GmbH). Ready-specific primer pairs were purchased from Qiagen (Supplementary Table S1). Samples were normalized to 18s RNA and fold change of expression was calculated according to the ΔΔC_t method as previously described (15).

Immunohistochemical staining for ZEB2 was carried out as previously described (16). In summary, 1 μm sections from paraffin-embedded tissue blocks were placed on capillary gap microscope slides (SUPERFROST PLUS microscope slides, Menzel) and baked overnight at 37°C. Sections were deparaffinized by xylene, rehydrated in graded concentrations of ethanol and boiled in a microwave oven for 5 minutes [pH 6.0, 0.94 mL Antigen Unmasking Solution (Vector Laboratories, Inc.)/100 mL distilled water] for 3 times for heat-induced antigen retrieval. Endogenous peroxidase activity of the tissue was inhibited by treatment with 3.0% hydrogen peroxidase solution in methanol for 20 minutes. Nonspecific-binding sites were blocked in 1 mol/L PBS with 10% normal goat serum and an Avidin/Biotin Blocking Kit (Vector Laboratories). Primary antibodies against ZEB2 (Rabbit polyclonal, Novus Biologicals, 1:200) and IgG-negative control (Mouse IgG, BD Pharmingen, 1:200) were incubated at 4°C overnight. After washing with PBS, sections were loaded with secondary antibody coupled with peroxidase-conjugated polymers (EnVision+ System, DakoCytomation A/S) for 30 minutes. Subsequently, the primary antibodies were detected by using AEC Substrate Chromogen (DakoCytomation A/S) according to the instructions of the manufacturer. The sections were counterstained with hematoxylin, dehydrated in graded concentrations of ethanol and mounted. The staining intensity of each slide was separately scored for tumor invasion front and tumor center by scanning the whole section at medium (50×) and high magnification (200×) on a blind basis by 2 independent researchers (C.K. and S.L.) and 1 pathologist (C.M.) as absent: 0, weak: 1, medium: 2 and strong: 3. A multihead microscope was used and consensus was reached for each slide.

Microsatellite instability analysis was carried out using the markers BAT25, BAT26, and CAT25 as described previously (17). BRAF and KRAS mutation analysis was carried out using the following primers: sense 5′-TCATAATGCTTGCTCTGATAGGA-3′ and antisense 5′-GGCCAAAAATTTAATCAGTGGA-3′ for BRAF, sense 5′-AAGGCCTGCTGAAAATGACTG-3′ and antisense 5′-AGAATGGTCCTGCACCAGTAA-3′ for KRAS. Briefly, PCR products were purified using the QIAquick PCR purification kit (Qiagen). Subsequently, sequencing reaction was carried out using BigDye terminator kit (v1.1, Applied Biosystems) according to the manufacturer's instructions. Sequencing products were separated on an ABI 3100-automated sequencer (Applied Biosystems).

The colon cancer cell line DLD-1 was purchased from the Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbH (ACC274, DSMZ). In August 2010, the cell line was retested and reauthenticated by analyzing cell samples for 8 polymorphic short random repeats loci (DSMZ). DLD-1 cells were maintained in RPMI-1640 (Sigma), supplemented with 10% (v/v) fetal calve serum, 100 U/mL penicillin and 100 μg/mL streptomycin in a humidified atmosphere of 5% CO₂ at 37°C. ZEB2 siRNA [siZEB2 NM_014795_813 (sense): 3′-GCACAACAACGAGATTCTA-5′ and siZEB2NM_014795_1503 (sense): 3′-GCACATCAGCAGCAAGAAA-5′] and scrambled siRNA [NM_053208_siRNA_control_1224 (sense): 3′-GCUUAACCCGUAUUGCCUATT-5′; both Invitrogen GmbH] were transfected using Lipofectamine 2000 (Invitrogen GmbH) according to the manufacturer's instruction.

Eight hours after transfection, the medium containing Lipofectamine 2000 was discarded and DLD-1 cells were maintained in serum-free RPMI–1640 medium overnight. After overnight starvation, ability of cell migration and cell invasion was evaluated by the CytoSelect 24-Well Cell Migration and Invasion Assay (8 μm pores, Colorimetric Format, Cell Biolabs, Inc.) according to the manufacturer's instructions. For the Cell Migration Assay, DLD-1 cells were incubated for 24 hours; for the Invasion Assay, DLD-1 cells were incubated for 72 hours. Spectrophotometry for the Cell Migration and Cell Invasion Assay was conducted on a microtiter plate reader at 540 nm. All assays were carried out 3 times in quadruplicates.

The software package SPSS, version 18.0 (SPSS), and the statistical software environment R, version 2.13.0, were used for all calculations. Wilcoxon-signed rank test was employed for determining a different regulation between 2 different tissue compartments, respectively. Student t test was conducted for analysis of the migration and invasion assays. Fisher's exact test was applied to examine independence of ZEB2 expression and clinical and pathologic parameters. The Kaplan–Meier method was used to estimate cancer-specific survival curves, and differences between survival curves were assessed with the log-rank test. Multivariate survival analysis was carried out using Cox proportional hazards regression including the following covariates: gender, age (dichotomized at median age), tumor stage according to the UICC classification, dichotomized tumor grade (grade 1, 2 vs. 3, 4), type of resection (curative vs. noncurative), microsatellite stability, KRAS mutation status, expression of ZEB2 in the tumor invasion front and in the tumor center (dichotomized at median expression). The survival endpoint for univariate and multivariate analysis was time from surgery until cancer-related death, where death due to other reasons was counted as censored event. The proportional hazards assumption was tested as proposed by Grambsch and Therneau (18). Results were considered significant at a P value less than 0.05.

We first evaluated the expression profile of 13 EMT-associated genes and miRNAs in 30 colorectal liver metastases. All liver metastases had been subjected to laser microdissection to obtain compartment-specific tissue from normal liver, liver invasion front, tumor invasion front, and tumor center. The following genes were included in the evaluation: the extracellular matrix-associated genes COL1A1, COL3A1, the mesenchymal markers Vimentin, FSP1 and OB-Cadherin, the transcription factors HIF1α, ZEB1 and ZEB2, EMT-repressing miRNAs miR-141, miR-200a, miR-200b, miR-200c, and MMP3. Expression analysis revealed COL3A1, Vimentin, HIF1α, ZEB1, and ZEB2 to be significantly upregulated in the liver invasion front compared with normal liver and tumor center (Supplementary Fig. S1A–M, Supplementary Table S2). Besides, there was a statistically significant increased expression of ZEB2 in the tumor invasion front compared with the tumor center. HIF1α and ZEB1 were found to be significantly upregulated in the tumor invasion front compared with liver and tumor center. MMP3 displayed also an overexpression at the tumor edge compared with tumor center and liver in 20 clinical specimens, but was not detectable in 10 samples (Supplementary Fig. S1A–M, Supplementary Table S2).

HIF1α displays an increased expression in tumor cells and proliferating fibroblasts at the invasion front (2, 19, 20). ZEB1 has been found to be upregulated in dedifferentiated tumor cells of colorectal cancer at the invasion front and in tumor-associated fibroblasts.

These results of the upregulation of HIF1α and ZEB1 at the invasion front in colorectal cancer are in good accordance with our findings of the expression analysis of EMT-associated genes. However, the increased expression of ZEB2 at the tumor edge of colorectal liver metastases has not yet been described. Therefore, we focused on this EMT-inducing transcription factor which displayed an upregulation at the liver invasion front and at the tumor invasion front compared with pure liver and tumor center, respectively, based on the results of our qPCR expression analysis. To further localize the cellular origin of ZEB2, immunohistochemical analysis on matched paraffin-embedded colorectal liver metastases of the same clinical specimens, which had been subjected to laser microdissection, was done. Immunhistochemical analysis confirmed an overexpression of ZEB2 at the invasion front in 18 of 30 samples. In 10 clinical specimens there was a homogenous expression of ZEB2 in the tumor center and the invasion front, in 1 sample the tumor center displayed a stronger expression than the invasion front and 1 sample was not evaluable (Fig. 1A–D). Hepatocytes from the liver invasion front and from areas not adjacent to the tumor displayed a strong cytoplasmic expression of ZEB2. Immunohistochemical analysis of 175 primary colorectal cancer samples revealed an invasion front-specific overexpression of ZEB2 in 128 specimens when compared with ZEB2 expression in the tumor center. In 95 of these 128 cases, the difference in grade of intensity between tumor center and tumor invasion front was 1, in 32 cases the difference amounted to 2 and in 1 case the invasion front was strongly positive whereas the tumor center was negative. Thirty-nine specimens displayed a homogenous expression of ZEB2 in the tumor invasion front and tumor center, 4 samples had a stronger expression of ZEB2 in the tumor center than tumor invasion front and 6 samples were not evaluable (Fig. 1 E–H). In all samples, ZEB2 was predominantly expressed in the cytoplasm, whereas nuclear staining was observed only in a few cells. Immunohistochemical analysis of ZEB2 in healthy control colon tissue adjacent to the colorectal tumor displayed mainly no expression of ZEB2 or only a very weak expression.

Due to the overexpression of ZEB2 at the invasion front in vivo, we hypothesized that ZEB2 regulates migration or invasion in colon cancer cells in vitro. ZEB2 was antagonized by siRNA in vitro in DLD-1 cells. Transfection efficiency was evaluated by quantitative real-time PCR (qRT-PCR) and reached more than 90% (Fig. 2A). Twenty-four hours posttransfection, DLD-1 cells, which had been transfected with ZEB2 siRNA, displayed an impaired migration capacity of 25% compared with the cells having been transfected with scrambled siRNA (P = 0.01; Fig. 2B). Likewise, 72-hour posttransfection, DLD-1 cells transfected with ZEB2 siRNA had a reduced invasion capacity of 26% compared with DLD-1 cells transfected with scrambled siRNA (P = 0.001; Fig. 2C).

The association of clinical and histopathologic parameters with ZEB2 expression was evaluated separately for the tumor invasion front and tumor center. To compare the expression of ZEB2 with clinical and pathologic parameters, samples were grouped as ZEB2^Low (staining intensity ≤ median immunohistochemical score = 2) and as ZEB2^High (staining intensity > median immunohistochemical score = 2). Fisher's exact test revealed statistically significant associations of ZEB2 expression in the tumor invasion front with grading (P < 0.001) and tumor stage according to the UICC classification (P = 0.014; Table 1).

Log-rank tests showed a highly significant impact of tumor stage (P < 0.001) and tumor grade (P < 0.001) on cancer-specific survival (Table 2). Furthermore, high expression of ZEB2 at the tumor invasion front correlated significantly with a worsening prognosis (P < 0.001; ZEB2^High: estimated cancer-specific 5-year survival rate: 57.3%, ZEB2^Low: estimated cancer-specific 5-year survival rate: 76.2%; Fig. 3A, Table 2). Likewise high expression of ZEB2 in the tumor center was associated with grading (P = 0.036) and tumor stage according to the UICC classification (P = 0.048; Table 1). Although there was a tendency for poorer prognosis in patients with high ZEB2 expression in the tumor center, this correlation was not statistically significant (Fig. 3B, Table 2); estimated cancer-specific 5-year survival rate of 60.3% for ZEB2^High and of 71.8% for ZEB2^Low).

Multivariate analysis by Cox proportional hazards regression was carried out to identify prognostic markers for cancer-specific survival. Expression of ZEB2 at the invasion front was found to be an independent prognostic marker for cancer-specific survival (P = 0.02) besides tumor stage (P < 0.001) whereas ZEB2 expression at the tumor center failed to be significant (P = 0.33; Table 3).

The invasion front of colorectal cancer is characterized by a dynamic process referred to as EMT (1, 2, 21). During this process, tumor cells loose their epithelial cell–cell junctions and develop a mesenchymal-like phenotype. This process can be activated by cytokines, proteins of the extracellular matrix or hypoxia and is further mediated by several transcription factors. For a first overview, we started our study in 30 colorectal liver metastases by analyzing the expression of 13 genes and miRNAs at the invasion front, which are involved in the regulation of EMT. Here, we were able to show an overexpression of MMP3, HIF1α, ZEB1, and ZEB2 at the invasion front. These genes are involved in the repression of E-cadherin, which is a hallmark of EMT (11, 22–25). Brabletz and colleagues have shown that membranous expression of E-cadherin is lost at the invasion front of colorectal primary cancer and liver metastases (1). In this context, our data support the assumption that EMT occurs predominantly at the invading tumor edge. Interestingly, none of the EMT-repressing miRNAs of the miR-200 family (26–30) displayed a significantly different expression between tumor center and tumor invasion front, though they were highly upregulated compared with normal liver. Our results do not provide a more profound mechanistic explanation for these observations. However, it could be conjectured that the upregulation of pro-EMT factors tends to be more essential for tumor invasion than the inhibition of EMT-repressing molecules.

By immunohistochemistry, we confirmed the upregulation of ZEB2 at the tumor invasion front. ZEB2 was predominantly expressed in the cytoplasm, whereas nuclear staining was observed only in a few cells. These findings are consistent with a recent report, comparing the expression of ZEB2 by 2 monoclonal antibodies in a small cohort of patients with colorectal cancer (31). Due to the overexpression of ZEB2 at the invasion front, we postulated an invasion-promoting role of this transcription factor. In fact, downregulation of ZEB2 by siRNA resulted in a decreased migration and invasion capacity of DLD-1 cells in vitro. These data are in good accordance with a previous report, in which Vanderwalle and colleagues showed that doxycycline-induced overexpression of ZEB2 is associated with a more invasive type in DLD-1 cells (32).

One of our main findings is the significant correlation between overexpression of ZEB2 at tumor center and invasion front with the tumor stage according to the UICC classification. Furthermore, multivariate analysis revealed a prognostic impact of invasion front-specific overexpression of ZEB2 in primary colorectal cancer besides UICC stage, whereas the expression of ZEB2 in the tumor center was not significantly correlated to cancer-specific survival when analyzed together with other risk factors. Although previous reports have already described increased expression of ZEB2 as a prognostic marker in bladder cancer (33), ovarian cancer (34), squamous cell carcinoma (35), and lung cancer (36), these studies have not distinguished between the expression at the tumor edge and tumor center. However, our in vitro data and other studies indicate strongly that ZEB2 plays an important role in tumor migration and invasion (27, 32). These abilities are required at the invasion front to break down the basement membrane and to extravasate into lymph and blood vessels in the adjacent tissue. In this context, our data suggest that overexpression of ZEB2 at the invasion front is relevant for tumor cell dissemination and progression in primary colorectal cancer.

In summary, we report about an expression profile of a panel of EMT-associated genes in the tumor invasion front of colorectal liver metastases. These data support the significant role of EMT in vivo at the tumor edge of colorectal cancer. Moreover, our study is the first to show that overexpression of ZEB2 correlates significantly with the tumor stage according to the UICC classification and tumor grading. In our study, overexpression of ZEB2 at the invasion front is associated with poor prognosis in primary colorectal cancer besides known risk factors including age, tumor stage, and grading. These results suggest that ZEB2 might be a useful prognostic marker and should furthermore be investigated as a potential therapeutic target in patients with colorectal cancer.

No potential conflicts of interest were disclosed.

The authors thank Bettina Walter from the tissue bank of the National Center for Tumor Disease (NCT) Heidelberg for technical help in tissue sections.

The Clinical Research Unit KFO 227 “Colorectal cancer: From primary tumor progression towards metastases” was funded by the German Research foundation (DFG).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.