Purpose: Colorectal cancers are composed of phenotypically different tumor cell subpopulations within the same core genetic background. Here, we identify high expression of the TALE transcription factor PBX3 in tumor cells undergoing epithelial–mesenchymal transition (EMT), analyze PBX3 regulation, and determine clinical associations in colorectal cancer.

Experimental design: We used transcriptomic and in situ analyses to identify PBX3 expression in colorectal cancer and cell biology approaches to determine its regulation and function. Clinical associations were analyzed in independent tissue collections and gene expression datasets of colorectal cancers with recorded follow-up data.

Results: PBX3 was expressed in tumor cells with high WNT activity undergoing EMT at the leading tumor edge of colorectal cancers, whereas stromal cells were PBX3 negative. PBX3 expression was induced by WNT activation and by the EMT transcription factors SNAIL and ZEB1, whereas these effects were mediated indirectly through microRNA miR-200. PBX3 was required for a full EMT phenotype in colon cancer cells. On the protein level, PBX3 expression indicated poor cancer-specific and disease-free survival in a cohort of 244 UICC stage II colorectal cancers, and was associated with metastasis in a case–control collection consisting of 90 cases with or without distant metastasis. On the mRNA level, high PBX3 expression was strongly linked to poor disease-free survival.

Conclusions: PBX3 is a novel indicator of EMT in colorectal cancer, part of an EMT regulatory network, and a promising prognostic predictor that may aid in therapeutic decision making for patients with colorectal cancer. Clin Cancer Res; 24(8); 1974–86. ©2018 AACR.

Translational Relevance

High WNT activity and epithelial–mesenchymal transition (EMT) are drivers and hallmarks of colon cancer progression. Here, we identify PBX3 as a new indicator of EMT and part of an EMT regulatory network that is associated with poor prognosis in colon cancer. Assessment of PBX3 in colon cancer tissues may improve risk stratification for patients with colon cancer through gauging EMT, which is not yet adequately reflected by other routinely assessed clinical parameters.

Colorectal cancer ranks third in cancer incidence among men and women and is a major cause of cancer morbidity and mortality (1). More than 80% of colorectal cancers have driver mutations in APC or β-catenin that over activate WNT signaling in these tumors (2). Although these mutations are present in all clonally derived colon cancer cells, WNT signaling still remains regulated in these tumors, which is reflected by coexistence of tumor cell subpopulations with relatively low or high WNT activity (3).

Colon cancer cells with high WNT activity have especially been associated with a specific phenotype. These tumor cells are typically located at the leading tumor edge, show a less differentiated morphology, and infiltrate surrounding stromal tissue (4). Moreover, they are characterized by loss of epithelial differentiation with reduced expression of the cell adhesion molecule E-cadherin, and increased expression of mesenchymal markers, including vimentin, fibronectin, and LAMC2 (5, 6). This phenotype is termed epithelial–mesenchymal transition (EMT) and is regulated by transcription factors of the ZEB and SNAIL families, including ZEB1, which is induced by WNT signaling in colon cancer (7). Downstream effects of ZEB1 then are either transduced through direct transcriptional activation, or indirectly through repression of microRNAs of the miR-200 family that target EMT-related genes (8). Because of the promotion of tumor infiltration by EMT, high WNT activity therefore is a crucial driver of colon cancer invasion and progression. Identifying factors that are involved in EMT regulation through WNT, thus, may hold keys for a better understanding of the malignant biology of colorectal cancer.

Pre–B-cell leukemia homeobox transcription factor 3 (PBX3) belongs to a family of evolutionary conserved three-amino-acid-loop-extension (TALE) homeodomain transcription factors. These are known to serve as cofactors for homeobox (HOX) proteins and are physiologically involved in regulation of gene expression during embryonic development (9, 10). In cancer, PBX3 has been functionally linked to the development of certain forms of leukemia (11). Furthermore, PBX3 has been associated with tumor progression and metastasis in gastric and colon cancer (12–14). In gastric cancer, PBX3 has been shown to induce an infiltrative tumor cell phenotype with upregulation of vimentin and repression of E-cadherin (15). In addition, microRNAs that were shown to regulate PBX3 included members of the miR-200 family (16). These data suggested oncogenic features of PBX3 in colorectal and other cancers, and provided a first link between PBX3 and EMT. However, the regulation of PBX3 in colon cancer and its contribution to tumor progression still are incompletely understood.

Here, we aimed to identify transcription factors that are linked to tumor cell heterogeneity and differential WNT signaling activity in colorectal cancer. In this context, we identified PBX3 and subsequently determined its regulation in colon cancer in detail. We found that PBX3 expression is regulated by WNT signaling and EMT, and that it is required for a full EMT phenotype in colon cancer cells. Furthermore, we evaluated its potential as a prognostic biomarker for patients with colorectal cancer.

Clinical cases

Study design was based on the REMARK criteria (17). Formalin-fixed, paraffin-embedded (FFPE) specimens were obtained from the archives of the institute of pathology of the Ludwig-Maximilians-University Munich (LMU). A frozen sample of normal mucosa and corresponding colon cancer tissue was provided by the biobank under administration of the foundation Human Tissue and Cell Research (HTCR; ref. 18). For patients with cancer, follow-up data were recorded by the Munich Cancer Registry. Specimens were anonymized, and the need for consent was waived by the institutional ethics committee of the Medical Faculty of the LMU. Tissue microarrays of FFPE cancer specimens (TMAs) were generated with six representative 1-mm cores, including tumor edges and tumor centers of each case, using hematoxylin and eosin–stained tissue sections as templates. Individual samples of normal colonic mucosa also were included.

For the survival collection, we included 260 randomly selected patients of a cohort of 949 patients diagnosed with UICC stage II colorectal cancer at the LMU between 1994 and 2007 that underwent surgical resection. Sixteen cases dropped out due to lack of sufficient tumor material, resulting in a final collection of 244 patients with a median follow-up time of 4.9 years. Forty-three patients (17.6%) had received adjuvant or neoadjuvant radiation and/or chemotherapy. For tumor-specific survival, colorectal cancer attributed death was defined as clinical endpoint, and was documented in 52 patients (21.3%). For analysis of disease-free survival, tumor progression was the clinical endpoint, documented in 76 patients (31.1%) either as tumor recurrence or metastasis. Survival data were censored when case follow-up was discontinued, or when patients had died of reasons other than colorectal cancer. Variables considered for inclusion models were age, sex, T-stage, tumor grade, and neo-/adjuvant therapy. At a power of 0.8 with α=0.05, relative risks of 2.2 for cancer-specific survival and 1.9 for disease-free survival could be detected in this collection (19), indicating sufficient sample size for our study design.

For the metastasis collection, we chose a case–control design that included tumor specimens of 90 patients who were surgically resected or biopsied at the LMU between 1994 and 2005. None of them were part of the survival collection. Half of the patients had colon cancers with synchronous liver metastasis (UICC stage IV), diagnosed by clinical imaging or liver biopsy. Controls consisted of patients with colon cancer without distant metastasis at the time of diagnosis (UICC stages I–III) and with disease-free survival of at least 5 years after primary surgical resection. Cases and controls were matched by tumor grade (according to WHO 2010), T-category, and tumor location (all tumors were right-sided colon cancers), resulting in 45 matched pairs. Sample size was based on previous experience and limited by the availability of tumor material.

Immunohistochemistry and immunofluorescence

Immunohistochemistry was done on 5-μm tissue sections on a Ventana Benchmark XT autostainer with ultraView Universal DAB detection kits (Ventana Medical Systems) using the antibodies and concentrations listed in Supplementary Table S1. Expression of PBX3 was evaluated blinded from clinical outcome by averaging the percentage of PBX3-stained tumor cells in all TMA cores of a tumor. To determine intratumoral co-localization of PBX3 and β-catenin or LAMC2, we measured staining intensities in tumor cells of different areas, including the leading tumor edge and the tumor center, on serial tumor sections using ImageJ (20). For immunofluorescence, cultured cells were fixed in 4% paraformaldehyde for 10 minutes, permeabilized with 0.2% TritonX100 for 15 minutes and blocked with 3% BSA in PBS for 30 minutes. Cells then were incubated sequentially at room temperature with primary antibodies listed in Supplementary Table S1 and then with Alexa Fluor 488- or 568-labeled secondary antibodies (Abcam). F-actin was visualized with Alexa Fluor 647-labeled Phalloidin (Invitrogen). Nuclei were counterstained with DAPI (Vector Laboratories, 1:500). Confocal fluorescence images were taken on an LSM 700 laser scanning microscope using the ZEN software (Carl Zeiss).

Gene-expression datasets and GSEA

Gene-expression datasets were retrieved from the Gene Expression Omnibus (21). For comparative analyses of colon cancer cell subpopulations with low and high WNT activity, datasets GSE32408 and GSE17375 were used (3, 22). Microarray data were normalized simultaneously with Robust Multi-array Average (RMA; ref. 23) using custom brainarray CDF (v19, ENTREZG; ref. 24) in R (www.r-project.org), which yielded one optimized probeset per gene, as previously described (25, 26). Data then were filtered for transcription factors based on annotations by the human protein atlas (27). For the analysis of colon cancer samples, we retrieved datasets GSE14333 and GSE39582, based on the availability of matched transcriptome and clinical data, and normalized them simultaneously in the same manner as described previously. Pearson correlations of PBX3 expression and expression of all other genes represented within these datasets then were calculated, and genes were ranked accordingly. GSEA analyses (28) then were done using this ranked gene list against curated sets of EMT core signatures (29, 30). Heatmaps for selected genes were generated with GENE-E (software.broadinstitute.org/GENE-E/).

Genetic vectors

The DOX inducible episomal pRTR vector system and pRTR-SNAIL-VSV were described previously (31). For pRTR-ZEB1-VSV, we excised the ZEB1 coding sequence from pcDNA4hismaxCZEB1, a gift from Janet Mertz (32), and inserted it into SfiI restriction sites of pRTR. For microRNA-binding luciferase reporter constructs, we amplified the 3′-UTR of the human PBX3 gene containing the miR-200–binding site by PCR, using primers 5′-GATCAGAGACTGGTAGCATCG-3′ and 5′-AATCATGAAAGCAAAAAGTTTATTC-3′, and cDNA from SW480 cells as template, and inserted it into pGL3 (Promega). To insert mutations into the miR-200 seed-matching sequence, we then applied site directed mutagenesis using the QuikChange Mutagenesis Kit according to the manufacturer's protocol (Agilent). For analysis of WNT/β-catenin activity within the PBX3 promoter, 2.5 kbp upstream of the PBX3 transcription start site containing TCF4-binding sites were amplified from human BAC clone CTD-2309J17 (CalTech BAC library) using Pfu Polymerase (Thermo Scientific) and primers 5′-CTCTAAGCGCTTTGCGATTG-3′ and 5′-AGCATCCTGGATTGATCGTC-3′. PCR products then were inserted into pBV-Luc vector (a gift from Bert Vogelstein, Addgene plasmid #16539). Synthetic DNA sequences (IDT) then were used to replace TCF4-binding sites by mutated sites. Modified vector elements were verified by restriction analysis and sequencing.

Cell culture, transfections, and luciferase assays

SW1222 were a gift from the Ludwig Institute for Cancer Research (New York, NY), and LS174T dnTCF4 and DLD-1 dnTCF4 were a gift from Marc van de Wetering (33). Other cell lines were from the ATCC. All cell lines were obtained between 2009 and 2014, authenticated using short-tandem repeat profiling, and tested negative for mycoplasma contamination. Cells were maintained in DMEM supplemented with 10% FBS, 100 U/mL penicillin, and 0.1 mg/mL streptomycin (Biochrom). For doxycycline (DOX, Sigma-Aldrich) induced expression, cells were stimulated with final concentrations of 100 ng/mL. Polyclonal cell pools for conditional expression were generated by transfection of pRTR-SNAIL-VSV or pRTR-ZEB1-VSV using FuGENE 6 (Promega) and selected in 2 μg/mL Puromycin (Sigma) for 14 days. WNT stimulation was achieved using WNT3a (R&D Systems) at a final concentration of 20 ng/mL. Transient knockdown was done using predesigned siRNAs targeting β-catenin (Qiagen), PBX3 (Ambion) or ZEB1 (Ambion and Dharmacon), or, as control, scrambled siRNA (siCtrl, Qiagen). hsa-miR-200c (Ambion) was used for transient miRNA expression. siRNAs and miRNAs were transfected at 10 nmol/L and 30 nmol/L final concentrations, respectively, using HiPerFect (Qiagen). For luciferase reporter assays, cells were transfected or co-transfected in 24-well plates with 10 ng Renilla luciferase control vector and 100 ng TOPflash or FOPflash luciferase reporter constructs carrying either wild-type or mutant TCF-binding sites (34) in the presence of 0.5 μL FuGENE 6 (Promega). Firefly luciferase activity was measured with dual-luciferase Reporter Assays (Promega) after 24 hours and normalized to Renilla luciferase activity. Luminescence was measured with an Orion II luminometer (Berthold).

Immunoblotting

Immunoblotting was done using whole-cell lysates supplemented with protease and phosphatase inhibitors (Roche), as previously described (35). Proteins were separated by SDS-PAGE, transferred onto polyvinylidene difluoride membranes (Merck Millipore) and incubated with primary antibodies listed in Supplementary Table S1. Bands were visualized using horseradish peroxidase (HRP)-‐conjugated secondary mouse (Promega) or rabbit (Sigma) antibodies and chemiluminescent HRP Substrate (Millipore). Bands then were quantified using ImageJ (20) and normalized to α-tubulin or β-actin.

Real-time quantitative PCR

Total RNA was isolated using QIAzol Lysis Reagent (Qiagen) and cDNA was generated from 500 ng total RNA per sample using QuantiTect Reverse Transcription (Qiagen). Real-time quantitative PCR (qRT-PCR) then was done using Fast SYBR Green Mix (Applied Biosystems) on a LightCycler 480 (Roche), applying 40 cycles of amplification at 95°C (1 sec), 60°C (20 sec), and 72°C (1 sec). Primers are listed in Supplementary Table S2. qRT-PCR results were first normalized to GAPDH mRNA levels in the same sample and then to levels of the corresponding transcript in control-treated tumor cells.

Statistical analyses

Survival was analyzed by the Kaplan–Meier method and groups were compared with the log-rank test. Optimal cutoffs for continuous variables were selected by receiver operating characteristic (ROC) curve analyses and Youden's index. Cox proportional hazards model was used for uni- and multivariate analyses. Appropriate statistical tests were used to compare data with similar variances and are referenced in figure legends. Biological replicates are given as n values. All graphs show mean and error bars represent standard deviation (s.d.). Differences were considered statistically significant when P < 0.05. P values are given within figures or figure legends. There were no missing data. Statistics were calculated with GraphPad Prism or SPSS (IBM).

PBX3 is overexpressed in colon cancer cells with high WNT activity

To find transcription factors linked to WNT signaling activity in colon cancer, we screened previously published gene-expression datasets that were derived from human colon cancer cell subpopulations with low and high WNT activity (3, 22). Of 950 represented genes that encoded for known or putative transcription factors, 67 (7.1 %) were significantly (P < 0.05 by t test) differentially expressed by 1.25-fold or more. Among those with most significant overexpression in tumor cells with high WNT activity that expectedly included known WNT pathway components or target genes such as LEF1, TCF7, and PROX1, we identified PBX3 (Fig. 1A; Supplementary Table S3). Indeed, in this dataset increased PBX3 expression coincided with high expression of WNT pathway components and target genes and, conversely, with repression of genes associated with a differentiated tumor cell phenotype (Fig. 1B). We then examined tissue sections of colon cancers and found that PBX3 was heterogeneously expressed in the cytoplasm of tumor cells, with strongest expression at the leading tumor edge, where it overlapped with strong expression of nuclear β-catenin, indicating high WNT activity (Fig. 1C and D). Interestingly, despite being known as a transcription factor, we observed no nuclear staining for PBX3; also, not in normal colonic mucosa that showed weak cytoplasmic positivity only (Supplementary Fig. S1A). To confirm specificity of PBX3 detection, we therefore examined its expression in normal colonic mucosa, primary colon cancer tissue, and colon cancer cell lines by immunoblotting and found that on the protein level PBX3 isoform A was mainly detected in tumor tissues and also responded to PBX3 knockdown by siRNA (Supplementary Fig. S1B–S1D). Collectively, these findings indicated upregulation of PBX3 in colorectal cancer cells with high WNT activity on mRNA and protein levels, and suggested a possible regulation of PBX3 by WNT.

Figure 1.

Expression of PBX3 in colon cancer cells with high WNT activity. A, Volcano plot for differentially expressed genes encoding for transcription factors in colon cancer cells with high and low WNT activity (GSE32408 and GSE17375). Colored dots denote genes that are significantly (P < 0.05) upregulated (red) or downregulated (blue) by 1.25-fold or more. Genes are listed in Supplementary Table S3. B, Heat maps of PBX3, selected WNT targets, and differentiation factors in these data sets (D1=GSE32408, D2=GSE17375). C and D, Representative (C) immunohistochemistry and (D) quantification of PBX3 and nuclear β-catenin on serial sections of colon cancer. Arrowheads indicate co-staining at the leading tumor edge; scale bar, 100 μm. For quantification n = 300 cells from six colorectal cancers were scored. P and r values are results of linear-regression analysis.

Figure 1.

Expression of PBX3 in colon cancer cells with high WNT activity. A, Volcano plot for differentially expressed genes encoding for transcription factors in colon cancer cells with high and low WNT activity (GSE32408 and GSE17375). Colored dots denote genes that are significantly (P < 0.05) upregulated (red) or downregulated (blue) by 1.25-fold or more. Genes are listed in Supplementary Table S3. B, Heat maps of PBX3, selected WNT targets, and differentiation factors in these data sets (D1=GSE32408, D2=GSE17375). C and D, Representative (C) immunohistochemistry and (D) quantification of PBX3 and nuclear β-catenin on serial sections of colon cancer. Arrowheads indicate co-staining at the leading tumor edge; scale bar, 100 μm. For quantification n = 300 cells from six colorectal cancers were scored. P and r values are results of linear-regression analysis.

Close modal

PBX3 expression is regulated by WNT signaling in colorectal cancer

Next, we determined whether PBX3 expression depended on high WNT activity in colon cancer. Reducing WNT activity by depletion of β-catenin with two different siRNAs reduced PBX3 on the protein level in five different colon cancer cell lines with known APC or β-catenin mutations (Fig. 2A; Supplementary Fig. S2). We confirmed these effects on the mRNA level by qRT-PCR in two cell lines, SW1222 and DLD-1, in which β-catenin knockdown significantly downregulated PBX3 expression and that of the WNT target genes AXIN2, NKD1 and LGR5 (Fig. 2B). Furthermore, we tested the effects of a DOX-inducible dominant negative TCF4 (dnTCF4), a potent inhibitor of the β-catenin/TCF4 transcription factor complex (33). In LS174T and DLD-1 colon cancer cell lines, dnTCF4 induction strongly reduced transcription from β-catenin/TCF4-binding sites, as seen in TOPflash luciferase reporter assays (Fig. 2C) but also decreased PBX3 protein expression and downregulated PBX3 mRNA levels among the panel of WNT target genes (Fig. 2D and E). On the contrary, WNT3a stimulation of HEK293T, a human embryonic kidney derived cell line with low intrinsic WNT activity, led to strong overexpression of PBX3 and active β-catenin on the protein level, as well as upregulation of PBX3 and WNT target gene mRNA (Fig. 2F and G). Taken together, these findings suggested that PBX3 expression is regulated by WNT signaling in colon cancer, whereas this effect is not cell type specific.

Figure 2.

Effects of WNT signaling on PBX3 expression in colon cancer cells. A and B, Effects of β-catenin or control (siCtrl) knockdown by siRNA in SW1222 and DLD-1 colon cancer cells, harvested 72 hours after transfection. A, Immunoblotting for indicated proteins on whole-cell lysates. B, qRT-PCR expression analyses for indicated genes. C–E, Induction of a conditional dominant negative TCF4 allele (dnTCF4) by DOX in LS174T and DLD-1 cells. Cells were harvested 24 hours after DOX treatment. C, TOPflash dual luciferase assays. D, Immunoblotting for indicated proteins. E, qRT-PCR expression analyses for indicated genes. F and G, Stimulation of HEK293T cells with WNT3a or without stimulation (Ctrl) for 24 hours. F, Immunoblotting for indicated proteins. G, qRT-PCR expression analyses for indicated genes. Numbers below immunoblots indicate normalized fold change by densitometry. Data are mean and error bars indicate s.d. *, P < 0.05; **, P < 0.01; ***, P < 0.001; n.s. P > 0.05 by t test; n ≥ 3.

Figure 2.

Effects of WNT signaling on PBX3 expression in colon cancer cells. A and B, Effects of β-catenin or control (siCtrl) knockdown by siRNA in SW1222 and DLD-1 colon cancer cells, harvested 72 hours after transfection. A, Immunoblotting for indicated proteins on whole-cell lysates. B, qRT-PCR expression analyses for indicated genes. C–E, Induction of a conditional dominant negative TCF4 allele (dnTCF4) by DOX in LS174T and DLD-1 cells. Cells were harvested 24 hours after DOX treatment. C, TOPflash dual luciferase assays. D, Immunoblotting for indicated proteins. E, qRT-PCR expression analyses for indicated genes. F and G, Stimulation of HEK293T cells with WNT3a or without stimulation (Ctrl) for 24 hours. F, Immunoblotting for indicated proteins. G, qRT-PCR expression analyses for indicated genes. Numbers below immunoblots indicate normalized fold change by densitometry. Data are mean and error bars indicate s.d. *, P < 0.05; **, P < 0.01; ***, P < 0.001; n.s. P > 0.05 by t test; n ≥ 3.

Close modal

To test for direct transcriptional regulation by WNT, we then screened the PBX3 promoter sequence and identified three putative β-catenin/TCF4-binding motifs (WWCAAAG; ref. 36) within 2.5 kb, upstream of the first exon of the PBX3 gene (Supplementary Fig. S3A). To determine whether PBX3 is induced by WNT activation via these β-catenin/TCF4 motifs, we subjected 2.5 kb of the PBX3 promoter region, including these motifs or mutated motifs as control, to dual luciferase reporter assays. Unexpectedly, WNT3a stimulation did not increase luciferase expression from the wild-type reporter (Supplementary Fig. S3B), whereas TOPflash assays confirmed strong transcriptional activation of WNT signaling by WNT3a (Supplementary Fig. S3C). These data suggested that PBX3 is no direct β-catenin/TCF4 target gene but instead modulated by other WNT-dependent downstream factors.

PBX3 is strongly associated with EMT in colon cancer

Because PBX3 expression was strongest in tumor cells with high WNT activity at the infiltrative tumor edge (Fig. 1C), and WNT signaling is known to induce EMT in colon cancer (7), we hypothesized that PBX3 might be linked to EMT. To test for a general association of PBX3 with an EMT phenotype, we assembled and normalized publicly available mRNA expression data of 856 colon cancers. Gene Set Enrichment Analyses (GSEA) revealed highly significant (P < 0.001) correlations of PBX3 expression and the expression of two published core EMT gene signatures (Fig. 3A; refs. 29, 30). Moreover, markers that reportedly indicate EMT in colon cancer were significantly overexpressed in tumors with high PBX3 levels, among them most prominently ZEB1 (r = 0.68, P < 0.0001; Fig. 3B). In contrast, CDH1, which encodes for E-cadherin, negatively correlated with PBX3 in this dataset, further supporting the association of PBX3 and EMT. In addition, to shed more light on the intratumoral distribution of PBX3 and EMT, we assessed colon cancer tissues for PBX3 and LAMC2, a factor regulated by ZEB1 (7), by immunohistochemistry. The expression of both markers overlapped at the leading tumor edge (Fig. 3C) and strongly correlated when scored in individual colon cancer cells (Fig. 3D). Taken together, these data demonstrated that PBX3 is associated with EMT in colon cancer.

Figure 3.

Association of PBX3 and EMT in colon cancer. A, GSEA for genes ranked by Pearson correlation (Pearson r) to PBX3 expression for two core EMT gene signatures by Anastassiou et al. and Taube et al. (29, 30) in gene-expression datasets of n = 856 colon cancers; P < 0.001. B, Heat map indicates clustering and positive correlation of PBX3 expression with colon cancer relevant EMT markers and negative correlation with CDH1. Colors indicate Pearson r from −1 (blue) to 1 (red). C and D, Representative (C) immunohistochemistry and (D) quantification of co-expression for PBX3 and LAMC2 on serial sections of colon cancer. Arrowheads indicate co-staining at the leading tumor edge; scale bar, 100 μm. For quantification n = 300 cells from six colorectal cancers were scored. P and r values are results of linear-regression analysis.

Figure 3.

Association of PBX3 and EMT in colon cancer. A, GSEA for genes ranked by Pearson correlation (Pearson r) to PBX3 expression for two core EMT gene signatures by Anastassiou et al. and Taube et al. (29, 30) in gene-expression datasets of n = 856 colon cancers; P < 0.001. B, Heat map indicates clustering and positive correlation of PBX3 expression with colon cancer relevant EMT markers and negative correlation with CDH1. Colors indicate Pearson r from −1 (blue) to 1 (red). C and D, Representative (C) immunohistochemistry and (D) quantification of co-expression for PBX3 and LAMC2 on serial sections of colon cancer. Arrowheads indicate co-staining at the leading tumor edge; scale bar, 100 μm. For quantification n = 300 cells from six colorectal cancers were scored. P and r values are results of linear-regression analysis.

Close modal

PBX3 is required for EMT in colon cancer cells

EMT can be induced in colon cancer cells by ectopic expression of SNAIL or ZEB1 (37, 38), and we applied this approach to test whether PBX3 expression is EMT dependent. We used a DOX-inducible episomal vector system to overexpress either SNAIL or ZEB1 in DLD-1 and LS174T cells, two colon cancer cell lines with low EMT marker expression and pronounced epithelial phenotypes (39). In both cell lines, induction of SNAIL caused upregulation of PBX3 protein levels within 12 hours (Fig. 4A) and also increased PBX3 mRNA levels together with VIM and ZEB1, while repressing CDH1 and miR-200c (Fig. 4B), indicating an EMT phenotype. Immunofluorescence further confirmed upregulation of PBX3 together with increased presence of F-actin stress fibers, another previously reported feature of EMT (Fig. 4C; ref. 39). Similarly, ZEB1 induction also caused upregulation of PBX3 protein and mRNA levels with downregulation of miR-200c in both cell lines, and CDH1 repression in DLD-1 cells, whereas it had less effect on the other EMT markers (Supplementary Fig. S4). Because this suggested a regulation of PBX3 through a SNAIL–ZEB1 signaling axis, we next examined the effects of ZEB1 depletion on PBX3 in SW480 and LoVo colon cancer cells, both of which have high endogenous levels of ZEB1. ZEB1 depletion by siRNA decreased PBX3 protein and mRNA whereas miR-200c significantly increased in both cell lines (Fig. 4D and E). We confirmed these findings using a second siRNA-depleting ZEB1 in LoVo cells (Supplementary Fig. S5). PBX3 expression therefore not only correlates with a mesenchymal phenotype but also depends on SNAIL/ZEB1-mediated induction of EMT in colon cancer.

Figure 4.

Effects of EMT induction on PBX3 expression in colon cancer cells. A–C, Induction of a conditional SNAIL allele in LS174T and DLD-1 cells by DOX. A, Immunoblotting for indicated proteins at indicated time points after DOX treatment. B, Gene-expression analyses by qRT-PCR for indicated genes after 72 hours with or without DOX treatment. C, Representative confocal immunofluorescence for indicated proteins after 72 hours with or without DOX treatment; scale bars, 50 μm. D and E, Effects of ZEB1 or control (siCtrl) knockdown by siRNA in SW480 and LoVo colon cancer cells, harvested 72 hours after transfection. D, Immunoblotting of indicated proteins on whole-cell lysates. E, Gene-expression analyses by qRT-PCR for indicated genes. Numbers below immunoblots indicate normalized fold change by densitometry. Data are mean and error bars indicate s.d. *, P < 0.05; **, P < 0.01; ***, P < 0.001 by t test; n ≥ 3.

Figure 4.

Effects of EMT induction on PBX3 expression in colon cancer cells. A–C, Induction of a conditional SNAIL allele in LS174T and DLD-1 cells by DOX. A, Immunoblotting for indicated proteins at indicated time points after DOX treatment. B, Gene-expression analyses by qRT-PCR for indicated genes after 72 hours with or without DOX treatment. C, Representative confocal immunofluorescence for indicated proteins after 72 hours with or without DOX treatment; scale bars, 50 μm. D and E, Effects of ZEB1 or control (siCtrl) knockdown by siRNA in SW480 and LoVo colon cancer cells, harvested 72 hours after transfection. D, Immunoblotting of indicated proteins on whole-cell lysates. E, Gene-expression analyses by qRT-PCR for indicated genes. Numbers below immunoblots indicate normalized fold change by densitometry. Data are mean and error bars indicate s.d. *, P < 0.05; **, P < 0.01; ***, P < 0.001 by t test; n ≥ 3.

Close modal

Because EMT induction by ZEB1 causes repression of miR-200, and PBX3 is a recently identified miR-200 target, we asked whether the effects of ZEB1 on PBX3 may be indirectly mediated through this miRNA. Using TargetScanHuman (40) we identified a highly cross-species conserved 7-mer seed-matching sequence of miR-200b/c within the PBX3 3′-UTR (Fig. 5A). Transfection of SW480 colon cancer cells with miR-200c repressed both ZEB1 and PBX3, with stronger effects on protein than on mRNA levels, as expected for direct miRNA effects (Fig. 5B and C). We then cloned the 3′-UTR of PBX3, including the miR-200 seed-matching sequence, downstream of a luciferase open reading frame and found significant downregulation of the reporter activity upon transfection with miR-200c (Fig. 5D and E), or upon siRNA mediated knockdown of ZEB1 (Fig. 5F). Both effects were abolished when using a luciferase reporter containing the 3′-UTR of PBX3 with a mutated seed-matching sequence (Fig. 5D–F). These findings demonstrated that PBX3 is targeted by miR-200c and suggested that ZEB1-mediated induction of PBX3 occurs indirectly through de-repression by miR-200c.

Figure 5.

Modulation of PBX3 expression by miR-200c and requirement of PBX3 for EMT in colon cancer cells. A, Illustration of the PBX3 3′-UTR indicating phylogenetic conservation of a miR-200 seed-matching sequence. B and C, Transfection of SW480 cells with miR-200c or non-targeting (Ctrl) oligonucleotides for 48 hours. B, Immunoblotting for indicated proteins on whole cell lysates. C, qRT-PCR expression analyses for indicated genes. D, 3′-UTRs of PBX3 with wild-type (WT) and mutated (MUT) miR-200 seed matching sequences. E and F, Effects on dual luciferase reporter assays in SW480 cells transfected with pGL3 carrying WT or MUT 3′-UTRs of PBX3, or with empty vector (pGL3). E, Co-transfection of miR-200c or non-targeting (Ctrl) oligonucleotides for 72 hours. F, Co-transfection of ZEB1 or control (siCtrl) siRNA for 72 hours. G, Induction of a conditional SNAIL allele in LS174T and DLD-1 cells by DOX and immunoblotting for indicated proteins on whole-cell lysates after PBX3 or control (siCtrl) knockdown by siRNA for 72 hours. H, Immunoblotting for indicated proteins on whole cell lysates of SW480 and LoVo colon cancer cells after PBX3 or control (siCtrl) knockdown by siRNA for 72 hours. Numbers below immunoblots indicate normalized fold change by densitometry. Data are mean and error bars indicate s.d.; *, P < 0.05; **, P < 0.01; ***, P < 0.001; n.s. P > 0.05 by t test; n ≥ 3.

Figure 5.

Modulation of PBX3 expression by miR-200c and requirement of PBX3 for EMT in colon cancer cells. A, Illustration of the PBX3 3′-UTR indicating phylogenetic conservation of a miR-200 seed-matching sequence. B and C, Transfection of SW480 cells with miR-200c or non-targeting (Ctrl) oligonucleotides for 48 hours. B, Immunoblotting for indicated proteins on whole cell lysates. C, qRT-PCR expression analyses for indicated genes. D, 3′-UTRs of PBX3 with wild-type (WT) and mutated (MUT) miR-200 seed matching sequences. E and F, Effects on dual luciferase reporter assays in SW480 cells transfected with pGL3 carrying WT or MUT 3′-UTRs of PBX3, or with empty vector (pGL3). E, Co-transfection of miR-200c or non-targeting (Ctrl) oligonucleotides for 72 hours. F, Co-transfection of ZEB1 or control (siCtrl) siRNA for 72 hours. G, Induction of a conditional SNAIL allele in LS174T and DLD-1 cells by DOX and immunoblotting for indicated proteins on whole-cell lysates after PBX3 or control (siCtrl) knockdown by siRNA for 72 hours. H, Immunoblotting for indicated proteins on whole cell lysates of SW480 and LoVo colon cancer cells after PBX3 or control (siCtrl) knockdown by siRNA for 72 hours. Numbers below immunoblots indicate normalized fold change by densitometry. Data are mean and error bars indicate s.d.; *, P < 0.05; **, P < 0.01; ***, P < 0.001; n.s. P > 0.05 by t test; n ≥ 3.

Close modal

To further determine whether EMT in colon cancer also depended on PBX3 expression, we induced EMT by SNAIL in DLD-1 and LS174T cells and concomitantly depleted PBX3 by siRNA. SNAIL expression caused downregulation of E-cadherin and upregulation of VIM, indicating loss of epithelial and gain of mesenchymal features. Depletion of PBX3 partially reversed this effect for E-cadherin but interestingly had no significant effects on VIM expression (Fig. 5G; Supplementary Fig. S6). Moreover, depleting PBX3 in SW480 and LoVo cell lines caused upregulation of E-cadherin but again did not reduce vimentin expression (Fig. 5H). These data implied that PBX3 expression is at least in part required for the induction and maintenance of EMT in colon cancer cells.

PBX3 expression is associated with colon cancer progression

Because of its dependence on WNT signaling and EMT, both drivers of colon cancer progression (5), we examined clinical associations of PBX3 expression in a collection of 244 colorectal cancer cases, all of which were UICC stage II with clinical follow-up records (Supplementary Table S4). We semiquantitatively assessed the percentage of PBX3-positive tumor cells, and defined scores with complete absence of staining (score 0), staining in less than 10% of tumor cells (score 1), 10% to 15% of tumor cells (score 2), and 50% or more tumor cells (score 3; Fig. 6A). Of note, although PBX3 staining was strongest in tumor cells at the leading tumor edge, it also extended to gland forming colon cancer cells in the tumor center, especially in cases with high levels of PBX3 expression (Fig. 6A). Kaplan–Meier statistics revealed similar outcome for cancer-specific and disease-free patient survival for cases with PBX3 scores 1–3, when compared with cases with PBX3 score 0 (Fig. 6B and C). We therefore re-classified into two categories only and found that PBX3-positive cases (scores 1–3) showed significantly worse cancer-specific and disease-free survival than PBX3 negative (score 0) cases (Fig. 6A–C). We then evaluated co-occurrences with other clinical/pathological variables, and observed that PBX3 positivity was more frequent in colorectal cancers of younger patients, whereas there was no significant correlation with sex, T-stage, tumor grade, or neo-/adjuvant therapy (Supplementary Table S4). Including these variables into proportional hazards regression analyses demonstrated independent prognostic power of PBX3 positivity for both cancer-specific and disease-free survival (Supplementary Tables S5 and S6). Because tumor outcome of colon cancer mainly depends on distant metastasis, we further investigated PBX3 expression in a second, independent matched case–control collection of 45 pairs of colon cancers with and without synchronous liver metastasis. In this collection, PBX3 positivity was significantly associated with liver metastasis (odds-ratio = 3.0), further strengthening the link of PBX3 and poor prognosis (Fig. 6D; Supplementary Table S7).

Figure 6.

PBX3 expression in colorectal cancer indicates poor prognosis. A, Assessment of PBX3 immunostaining in primary human colorectal cancers. Tumors were assigned semiquantitative expression scores from 0 (no staining) to 3 (strong staining) and accordingly categorized as PBX3 negative (score 0) and positive (scores 1–3). Arrows indicate stained tumor cells. Case numbers per score are indicated; scale bars, 100 μm. B and C, Kaplan–Meier plots for different PBX3 expression scores and categories in a collection of n = 244 stage II colorectal cancers for (B) cancer-specific survival and (C) disease-free survival. HR, hazard ratio for PBX3-positive cases. P values are log-rank test results. Ratios on curves indicate the number of events over the number of patients per group. D, Association of PBX3 expression and liver metastasis in a matched case control collection of colon cancers. The P value is χ2 test result. E,PBX3 mRNA expression and survival association in two individual and combined datasets of a total of n = 786 colon cancers. GEO accession numbers of individual datasets are indicated. Kaplan–Meier plots for cases with low and high PBX3 expression. HR for PBX3 high cases. P values are log-rank test results. Ratios on curves indicate the number of events over the number of patients per group.

Figure 6.

PBX3 expression in colorectal cancer indicates poor prognosis. A, Assessment of PBX3 immunostaining in primary human colorectal cancers. Tumors were assigned semiquantitative expression scores from 0 (no staining) to 3 (strong staining) and accordingly categorized as PBX3 negative (score 0) and positive (scores 1–3). Arrows indicate stained tumor cells. Case numbers per score are indicated; scale bars, 100 μm. B and C, Kaplan–Meier plots for different PBX3 expression scores and categories in a collection of n = 244 stage II colorectal cancers for (B) cancer-specific survival and (C) disease-free survival. HR, hazard ratio for PBX3-positive cases. P values are log-rank test results. Ratios on curves indicate the number of events over the number of patients per group. D, Association of PBX3 expression and liver metastasis in a matched case control collection of colon cancers. The P value is χ2 test result. E,PBX3 mRNA expression and survival association in two individual and combined datasets of a total of n = 786 colon cancers. GEO accession numbers of individual datasets are indicated. Kaplan–Meier plots for cases with low and high PBX3 expression. HR for PBX3 high cases. P values are log-rank test results. Ratios on curves indicate the number of events over the number of patients per group.

Close modal

Finally, for independent confirmation of these results, we tested for clinical correlations of PBX3 mRNA levels in the assembled gene-expression dataset of 856 colon cancer cases, 786 of which had follow-up data on tumor progression. Using ROC curve analyses and Youden's index, we identified an ideal cutoff value at a normalized expression intensity of 155 (natural scale) of PBX3 mRNA (Supplementary Fig. S7A). Dichotomal classification of cases by this score revealed a highly significant positive correlation of high PBX3 expression and poor disease-free survival that also was independent of other core clinical variables (Fig. 6E; Supplementary Table S8). Of note, the prognostic power of PBX3 was comparable with ZEB1 and outperformed SNAIL in this dataset (Supplementary Fig. S7B; Supplementary Tables S9 and S10). Collectively, these findings suggest that PBX3 is associated with tumor progression and poor survival in patients with colorectal cancer.

The ability of epithelial cancer cells to loose cellular junctions and polarity with subsequent infiltration of tumor surrounding stromal tissue is a main aspect of EMT and hallmark of cancer invasion (41). Here, we identify strong overexpression of PBX3 in tumor cells with high WNT activity undergoing EMT at the leading tumor edge of colorectal cancer. We demonstrate that PBX3 expression is induced in this tumor cell subset by WNT and the EMT regulating transcription factors SNAIL and ZEB1, whereas this induction—at least partially—occurs indirectly through a decreased repression of PBX3 mRNA by miR-200. These findings are in agreement with recent data that demonstrated targetability of PBX3 by different microRNAs (16), and therefore place its expression in colon cancer within a WNT and EMT regulatory network (42). Furthermore, we demonstrate that PBX3 expression is required for a full EMT phenotype in colon cancer cells, since its depletion partially blocked EMT induction by ZEB1 and SNAIL, and increased the expression of E-cadherin, indicating a shift toward more epithelial differentiation. However, because PBX3 depletion had no significant effects on vimentin expression, we suggest that only certain aspects of EMT depend on PBX3. Nevertheless, considering recent findings that PBX3 induces EMT in gastric cancer cells (15, 43), this indicates that it may generally be involved in EMT regulation in gastrointestinal cancers. Because PBX3 also has been shown to increase migration and invasion of colon cancer cells (14) both of which are features of EMT, this further supports the notion that PBX3 directly contributes to the infiltrative phenotype of colon cancer cells at the leading tumor edge. However, the exact mechanism by which PBX3 influences EMT in colon cancer still remains to be determined, keeping in mind that different isoforms of PBX3 and in some organs variable intracellular distributions are known and that PBX3 may function as cofactor for other homeobox proteins (11, 44).

In primary colon cancer tissues, we found that PBX3 can easily be visualized in situ by immunostaining. Importantly, labelling in these tumors was restricted to cancer cells while tumor surrounding stromal cells were PBX3 negative. Given that markers which robustly indicate EMT in colon cancer are scarce, and detection of ZEB1, SNAIL and vimentin can be difficult and confounded by labelling of stromal cells (45–47), we propose that PBX3 may be a useful marker to highlight and further study colon cancer cells undergoing EMT in situ. Furthermore, we demonstrate that PBX3 mRNA levels strongly correlated with EMT in a large gene-expression dataset derived from 856 colon cancer samples. Considering the restriction of PBX3 expression to cancer cells, we therefore propose that on the gene-expression level PBX3 may indicate the overall degree of EMT in colon cancer specimens with little confounding by the amount of stromal tissue within each sample. Of note, PBX3 expression was not completely restricted to infiltrative tumor cells at the leading tumor edge but also extended to glandular differentiated colon cancer cells, especially in cases with high levels of PBX3 expression. Because similar observations also were made for ZEB1 and SNAIL (45, 47), it remains to be determined to what extent infiltrative tumor cell morphology and such gradually expressed EMT-related factors indicate identical or only partially overlapping colon cancer cell subpopulations.

Our findings in patient series with clinical follow-up data suggest that PBX3 expression is linked to poor outcome in patients with colorectal cancer. PBX3 expression was significantly associated with poor cancer-specific survival and strongly correlated with an increased risk for cancer progression in a collection of 244 stage II colorectal cancers, whereas this was independent of other core clinical variables. Stage II colorectal cancer is characterized by local disease with full-thickness involvement of the bowel wall but absence of lymphatic or distant metastasis (48). Accordingly, most of these patients can be cured by surgical resection alone. However, disease progression after surgery still is observed in 25% to 30% of these cases, and patients may eventually die from their disease (49). We therefore suggest that assessing PBX3 expression may be promising to identify potentially aggressive cases of stage II colorectal cancer that may benefit from adjuvant therapy and increased clinical attention despite low tumor stage (50). Moreover, PBX3 expression also significantly correlated with metastasis in our case–control collection, and thus also may indicate progression in late-stage disease. In addition, we found that on the mRNA level PBX3 expression was highly significantly associated with poor outcome in a combined dataset with clinical information on 786 colon cancers, including all stages. This not only supported the findings in our tissue collections but also confirmed results from a previous study that suggested an association of PBX3 mRNA expression and poor patient survival in a smaller patient series (14). However, before PBX3 expression analysis may be introduced into routine pathology workup of colorectal cancer specimens, robust multicenter validation studies in prospectively recruited patient series will be required to determine its true prognostic biomarker potential. If the independent prognostic power of PBX3 can be validated, we suggest that this may be due to gauging EMT, which currently is not sufficiently reflected by other routinely assessed clinical and pathological variables.

No potential conflicts of interest were disclosed.

Conception and design: S. Lamprecht, M. Kaller, D. Horst

Development of methodology: S. Lamprecht, M. Kaller, H. Hermeking, D. Horst

Acquisition of data, analysis and interpretation of data, writing, review, and/or revision of the manuscript, and administrative, technical, or material support: S. Lamprecht, M. Kaller, E.M. Schmidt, C. Blaj, T.S. Schiergens, J. Engel, A. Jung, H. Hermeking, T.G.P. Grünewald, T. Kirchner, D. Horst

Study supervision: D. Horst

This study was supported by grants from the Deutsche Krebshilfe (11169; to D. Horst), the Wilhelm Sander-Stiftung (2012.031; to D. Horst), and the Rudolf Bartling Stiftung (to D. Horst and H. Hermeking). T.G.P. Grünewald was supported by grants from the Deutsche Krebshilfe (111886 and 70112257). We are grateful to A. Heier, J. Kövi, M. Melz, and A. Sendelhofert for experimental assistance.

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.

1.
Ferlay
J
,
Soerjomataram
I
,
Dikshit
R
,
Eser
S
,
Mathers
C
,
Rebelo
M
, et al
Cancer incidence and mortality worldwide: sources, methods and major patterns in GLOBOCAN 2012
.
Int J Cancer
2015
;
136
:
E359
86
.
2.
Cancer Genome Atlas N
. 
Comprehensive molecular characterization of human colon and rectal cancer
.
Nature
2012
;
487
:
330
7
.
3.
Horst
D
,
Chen
J
,
Morikawa
T
,
Ogino
S
,
Kirchner
T
,
Shivdasani
RA
. 
Differential WNT activity in colorectal cancer confers limited tumorigenic potential and is regulated by MAPK signaling
.
Cancer Res
2012
;
72
:
1547
56
.
4.
Cernat
L
,
Blaj
C
,
Jackstadt
R
,
Brandl
L
,
Engel
J
,
Hermeking
H
, et al
Colorectal cancers mimic structural organization of normal colonic crypts
.
PLoS One
2014
;
9
:
e104284
.
5.
Brabletz
T
,
Jung
A
,
Spaderna
S
,
Hlubek
F
,
Kirchner
T
. 
Opinion: migrating cancer stem cells - an integrated concept of malignant tumour progression
.
Nat Rev Cancer
2005
;
5
:
744
9
.
6.
Kalluri
R
,
Weinberg
RA
. 
The basics of epithelial-mesenchymal transition
.
J Clin Invest
2009
;
119
:
1420
8
.
7.
Sanchez-Tillo
E
,
de Barrios
O
,
Siles
L
,
Cuatrecasas
M
,
Castells
A
,
Postigo
A
. 
beta-catenin/TCF4 complex induces the epithelial-to-mesenchymal transition (EMT)-activator ZEB1 to regulate tumor invasiveness
.
Proc Natl Acad Sci U S A
2011
;
108
:
19204
9
.
8.
Wellner
U
,
Schubert
J
,
Burk
UC
,
Schmalhofer
O
,
Zhu
F
,
Sonntag
A
, et al
The EMT-activator ZEB1 promotes tumorigenicity by repressing stemness-inhibiting microRNAs
.
Nat Cell Biol
2009
;
11
:
1487
95
.
9.
Laurent
A
,
Bihan
R
,
Omilli
F
,
Deschamps
S
,
Pellerin
I
. 
PBX proteins: much more than Hox cofactors
.
Int J Dev Biol
2008
;
52
:
9
20
.
10.
Longobardi
E
,
Penkov
D
,
Mateos
D
,
De Florian
G
,
Torres
M
,
Blasi
F
. 
Biochemistry of the tale transcription factors PREP, MEIS, and PBX in vertebrates
.
Dev Dyn
2014
;
243
:
59
75
.
11.
Li
Z
,
Zhang
Z
,
Li
Y
,
Arnovitz
S
,
Chen
P
,
Huang
H
, et al
PBX3 is an important cofactor of HOXA9 in leukemogenesis
.
Blood
2013
;
121
:
1422
31
.
12.
Han
HB
,
Gu
J
,
Zuo
HJ
,
Chen
ZG
,
Zhao
W
,
Li
M
, et al
Let-7c functions as a metastasis suppressor by targeting MMP11 and PBX3 in colorectal cancer
.
J Pathol
2012
;
226
:
544
55
.
13.
Li
Y
,
Sun
Z
,
Zhu
Z
,
Zhang
J
,
Sun
X
,
Xu
H
. 
PBX3 is overexpressed in gastric cancer and regulates cell proliferation
.
Tumour Biol
2014
;
35
:
4363
8
.
14.
Han
HB
,
Gu
J
,
Ji
DB
,
Li
ZW
,
Zhang
Y
,
Zhao
W
, et al
PBX3 promotes migration and invasion of colorectal cancer cells via activation of MAPK/ERK signaling pathway
.
World J Gastroenterol
2014
;
20
:
18260
70
.
15.
Wang
S
,
Li
C
,
Wang
W
,
Xing
C
. 
PBX3 promotes gastric cancer invasion and metastasis by inducing epithelial-mesenchymal transition
.
Oncology letters
2016
;
12
:
3485
91
.
16.
Han
H
,
Du
Y
,
Zhao
W
,
Li
S
,
Chen
D
,
Zhang
J
, et al
PBX3 is targeted by multiple miRNAs and is essential for liver tumour-initiating cells
.
Nat Commun
2015
;
6
:
8271
.
17.
McShane
LM
,
Altman
DG
,
Sauerbrei
W
,
Taube
SE
,
Gion
M
,
Clark
GM
, et al
Reporting recommendations for tumor marker prognostic studies (REMARK)
.
J Natl Cancer Inst
2005
;
97
:
1180
4
.
18.
Thasler
WE
,
Weiss
TS
,
Schillhorn
K
,
Stoll
PT
,
Irrgang
B
,
Jauch
KW
. 
Charitable state-controlled foundation human tissue and cell research: ethic and legal aspects in the supply of surgically removed human tissue for research in the academic and commercial sector in germany
.
Cell Tissue Bank
2003
;
4
:
49
56
.
19.
Schoenfeld
DA
,
Richter
JR
. 
Nomograms for calculating the number of patients needed for a clinical trial with survival as an endpoint
.
Biometrics
1982
;
38
:
163
70
.
20.
Schneider
CA
,
Rasband
WS
,
Eliceiri
KW
. 
NIH Image to ImageJ: 25 years of image analysis
.
Nat Methods
2012
;
9
:
671
5
.
21.
Edgar
R
,
Domrachev
M
,
Lash
AE
. 
Gene Expression Omnibus: NCBI gene expression and hybridization array data repository
.
Nucleic Acids Res
2002
;
30
:
207
10
.
22.
Vermeulen
L
,
De Sousa
EMF
,
van der Heijden
M
,
Cameron
K
,
de Jong
JH
,
Borovski
T
, et al
Wnt activity defines colon cancer stem cells and is regulated by the microenvironment
.
Nat Cell Biol
2010
;
12
:
468
76
.
23.
Irizarry
RA
,
Hobbs
B
,
Collin
F
,
Beazer-Barclay
YD
,
Antonellis
KJ
,
Scherf
U
, et al
Exploration, normalization, and summaries of high density oligonucleotide array probe level data
.
Biostatistics
2003
;
4
:
249
64
.
24.
Dai
M
,
Wang
P
,
Boyd
AD
,
Kostov
G
,
Athey
B
,
Jones
EG
, et al
Evolving gene/transcript definitions significantly alter the interpretation of GeneChip data
.
Nucleic Acids Res
2005
;
33
:
e175
.
25.
Grunewald
TG
,
Bernard
V
,
Gilardi-Hebenstreit
P
,
Raynal
V
,
Surdez
D
,
Aynaud
MM
, et al
Chimeric EWSR1-FLI1 regulates the Ewing sarcoma susceptibility gene EGR2 via a GGAA microsatellite
.
Nat Genet
2015
;
47
:
1073
8
.
26.
Sahay
D
,
Leblanc
R
,
Grunewald
TG
,
Ambatipudi
S
,
Ribeiro
J
,
Clezardin
P
, et al
The LPA1/ZEB1/miR-21-activation pathway regulates metastasis in basal breast cancer
.
Oncotarget
2015
;
6
:
20604
20
.
27.
Uhlen
M
,
Fagerberg
L
,
Hallstrom
BM
,
Lindskog
C
,
Oksvold
P
,
Mardinoglu
A
, et al
Proteomics. Tissue-based map of the human proteome
.
Science
2015
;
347
:
1260419
.
28.
Subramanian
A
,
Tamayo
P
,
Mootha
VK
,
Mukherjee
S
,
Ebert
BL
,
Gillette
MA
, et al
Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles
.
Proc Natl Acad Sci U S A
2005
;
102
:
15545
50
.
29.
Anastassiou
D
,
Rumjantseva
V
,
Cheng
W
,
Huang
J
,
Canoll
PD
,
Yamashiro
DJ
, et al
Human cancer cells express Slug-based epithelial-mesenchymal transition gene expression signature obtained in vivo
.
BMC Cancer
2011
;
11
:
529
.
30.
Taube
JH
,
Herschkowitz
JI
,
Komurov
K
,
Zhou
AY
,
Gupta
S
,
Yang
J
, et al
Core epithelial-to-mesenchymal transition interactome gene-expression signature is associated with claudin-low and metaplastic breast cancer subtypes
.
Proc Natl Acad Sci U S A
2010
;
107
:
15449
54
.
31.
Siemens
H
,
Jackstadt
R
,
Hunten
S
,
Kaller
M
,
Menssen
A
,
Gotz
U
, et al
miR-34 and SNAIL form a double-negative feedback loop to regulate epithelial-mesenchymal transitions
.
Cell Cycle
2011
;
10
:
4256
71
.
32.
Ellis
AL
,
Wang
Z
,
Yu
X
,
Mertz
JE
. 
Either ZEB1 or ZEB2/SIP1 can play a central role in regulating the Epstein-Barr virus latent-lytic switch in a cell-type-specific manner
.
J Virol
2010
;
84
:
6139
52
.
33.
van de Wetering
M
,
Sancho
E
,
Verweij
C
,
de Lau
W
,
Oving
I
,
Hurlstone
A
, et al
The beta-catenin/TCF-4 complex imposes a crypt progenitor phenotype on colorectal cancer cells
.
Cell
2002
;
111
:
241
50
.
34.
Veeman
MT
,
Slusarski
DC
,
Kaykas
A
,
Louie
SH
,
Moon
RT
. 
Zebrafish prickle, a modulator of noncanonical Wnt/Fz signaling, regulates gastrulation movements
.
Curr Biol
2003
;
13
:
680
5
.
35.
Woischke
C
,
Blaj
C
,
Schmidt
EM
,
Lamprecht
S
,
Engel
J
,
Hermeking
H
, et al
CYB5R1 links epithelial-mesenchymal transition and poor prognosis in colorectal cancer
.
Oncotarget
2016
;
7
:
31350
60
.
36.
Korinek
V
,
Barker
N
,
Morin
PJ
,
van Wichen
D
,
de Weger
R
,
Kinzler
KW
, et al
Constitutive transcriptional activation by a beta-catenin-Tcf complex in APC-/- colon carcinoma
.
Science
1997
;
275
:
1784
7
.
37.
Kaller
M
,
Hermeking
H
. 
Interplay between transcription factors and MicroRNAs regulating epithelial-mesenchymal transitions in colorectal cancer
.
Adv Exp Med Biol
2016
;
937
:
71
92
.
38.
Peinado
H
,
Olmeda
D
,
Cano
A
. 
Snail, Zeb and bHLH factors in tumour progression: an alliance against the epithelial phenotype?
Nat Rev Cancer
2007
;
7
:
415
28
.
39.
Jackstadt
R
,
Roh
S
,
Neumann
J
,
Jung
P
,
Hoffmann
R
,
Horst
D
, et al
AP4 is a mediator of epithelial-mesenchymal transition and metastasis in colorectal cancer
.
J Exp Med
2013
;
210
:
1331
50
.
40.
Agarwal
V
,
Bell
GW
,
Nam
JW
,
Bartel
DP
. 
Predicting effective microRNA target sites in mammalian mRNAs
.
Elife
2015
;
4
.
41.
Hanahan
D
,
Weinberg
RA
. 
Hallmarks of cancer: the next generation
.
Cell
2011
;
144
:
646
74
.
42.
Thiery
JP
,
Acloque
H
,
Huang
RY
,
Nieto
MA
. 
Epithelial-mesenchymal transitions in development and disease
.
Cell
2009
;
139
:
871
90
.
43.
Li
B
,
Zhang
S
,
Shen
H
,
Li
C
. 
MicroRNA-144–3p suppresses gastric cancer progression by inhibiting epithelial-to-mesenchymal transition through targeting PBX3
.
Biochem Biophys Res Commun
2017
;
484
:
241
7
.
44.
Ramberg
H
,
Grytli
HH
,
Nygard
S
,
Wang
W
,
Ogren
O
,
Zhao
S
, et al
PBX3 is a putative biomarker of aggressive prostate cancer
.
Int J Cancer
2016
;
139
:
1810
20
.
45.
Spaderna
S
,
Schmalhofer
O
,
Hlubek
F
,
Berx
G
,
Eger
A
,
Merkel
S
, et al
A transient, EMT-linked loss of basement membranes indicates metastasis and poor survival in colorectal cancer
.
Gastroenterology
2006
;
131
:
830
40
.
46.
Toiyama
Y
,
Yasuda
H
,
Saigusa
S
,
Tanaka
K
,
Inoue
Y
,
Goel
A
, et al
Increased expression of Slug and Vimentin as novel predictive biomarkers for lymph node metastasis and poor prognosis in colorectal cancer
.
Carcinogenesis
2013
;
34
:
2548
57
.
47.
Franci
C
,
Gallen
M
,
Alameda
F
,
Baro
T
,
Iglesias
M
,
Virtanen
I
, et al
Snail1 protein in the stroma as a new putative prognosis marker for colon tumours
.
PLoS ONE
2009
;
4
:
e5595
.
48.
Edge
SB
,
Compton
CC
. 
The American joint committee on cancer: the 7th edition of the AJCC cancer staging manual and the future of TNM
.
Ann Surg Oncol
2010
;
17
:
1471
4
.
49.
Dotan
E
,
Cohen
SJ
. 
Challenges in the management of stage II colon cancer
.
Semin Oncol
2011
;
38
:
511
20
.
50.
Gunderson
LL
,
Jessup
JM
,
Sargent
DJ
,
Greene
FL
,
Stewart
AK
. 
Revised TN categorization for colon cancer based on national survival outcomes data
.
J Clin Oncol
2010
;
28
:
264
71
.