Expression of the urokinase plasminogen activator (uPA) increases during the progression of colorectal tumors from adenomas to carcinomas. The highest amounts of uPA are found at the invasion front of carcinomas, which also displays a strong expression of nuclear β-catenin and is therefore a region expressing β-catenin target genes at high levels. Here we show that β-catenin contributes to the transactivation of uPA. Therefore, β-catenin might have an impact on the capacity of colorectal tumors for invasion and metastasis, as well as dormancy, which are hallmarks of cancer.
In most if not all of the colorectal carcinomas, the Wnt pathway is deregulated, which is caused in up to 85% of all of the cases by mutations in the tumor suppressor adenomatous polyposis coli, or more seldomly in the oncogene β-catenin or the tumor suppressor gene axin-2/conductin (1) . As a consequence of these mutations, β-catenin is no longer degraded. Thus, there is a high selection pressure to stabilize β-catenin during colorectal carcinogenesis. β-Catenin participates in the generation of the epithelial phenotype by maintaining adherens junctions as it interacts with E-cadherin and the actin skeleton (1 , 2) . However, in conjunction with DNA-binding factors of the T-cell factor (TCF)/lymphoid enhancer factor family, β-catenin works as a transcription factor (1 , 3) . In the human gut, TCF-4 seems to be the relevant member of the TCF/lymphoid enhancer factor-1 family (4) . Moreover, TCF-4-deficient mice die shortly after birth due to depletion of the intestinal stem cell compartment (5) . Here, β-catenin imposes a dedifferentiated stem cell-like phenotype on the cells keeping them in a state of proliferation (6) . A prerequisite for the function of transcription factors is their nuclear localization. For β-catenin this situation is observed frequently at the invasion front in the majority of well-differentiated human colorectal adenocarcinomas (7) . Therefore, this zone highly expresses β-catenin target genes, which are important contributors to the hallmarks of cancer, proliferation (c-myc and cyclin D1), invasion and metastasis [CD44, laminin-5γ2, matrix metalloproteinase (MMP)-7, and MT1-MMP], angiogenesis (vascular endothelial growth factor), and others (8) .
The serine protease urokinase plasminogen activator (uPA) converts inactive plasminogen into plasmin and is therefore a kingpin in the initiation of a cascade of proteolytic steps accumulating in the degradation of the extracellular matrix (9) . Secondly, uPA is found in cellular structures at the leading edge of migrating cells that are involved in adhesion, migration, invasion, and intravasation (9) . Thirdly, at high-density, membranous complexes of uPA/uPA receptor and fibronectin/α5β1-integrin activate the extracellular receptor kinase 1/2 pathway leading to growth stimulation, whereas at low densities the p38 mitogen-activated protein kinase pathway takes over, and dormancy is induced (10) . Taken together, uPA is a coordinator of cellular growth properties. A strong expression of uPA and the uPA receptor, another target gene of β-catenin, at the invasion front of colorectal carcinomas indicates poor survival of patients and aggressive tumor growth (11) . In the present work we demonstrate that uPA is a β-catenin target gene in colorectal carcinomas.
Materials and Methods
Immunohistochemistry and Statistical Analysis.
Twenty seven formaldehyde-fixed, paraffin-embedded, well-differentiated sporadic colorectal adenocarcinomas displaying nuclear localization of β-catenin at their invasion front were drawn from the archive of the Institute of Pathology in Erlangen. Immunohistochemistry was done as has been described (12) using antibodies specific for uPA and β-catenin (Supplementary Data). The Mann-Whitney test was applied for correlating β-catenin or uPA expression with localization in the tumor, invasion front or central area, respectively. The χ2 test was used for correlating uPA and nuclear β-catenin expression. Results are presented graphically as box plot. Calculations and graphics were done using SPSS version 10.0.5 software (SPSS Inc., Chicago, IL).
Electrophoretic Mobility Shift Assay.
Unlabeled competitor, or water in the case of controls, was incubated with 1 μl of crude purified recombinant glutathione S-transferase (GST)-TCF-4 (amino acid 265–496) or GST in binding buffer [10 mm HEPES (pH 7.9), 60 mm KCl, 1 mm EDTA, 1 mm DTT, and 4% (w/v) Ficoll] in the presence of 12.5 ng/μl poly-(deoxyinosinic-deoxycytidylic) acid (Sigma, Munich, Germany) and 62.5 μg/ml BSA (Sigma) in a total volume of 16 μl. After an incubation period of 5 min, 0.5 ng of [γ32P]ATP end-labeled double-stranded oligonucleotides (Table 1) ⇓ with a specific activity of 3 × 108 dpm/μg (Hartmann Analytik, Braunschweig, Germany) were added as indicated and the mixture incubated for another 20 min. Fifteen ng (30-fold molar amount) of unlabeled oligonucleotides (Supplementary Data; Table 1 ⇓ ) were used for competition when indicated. Products were separated using 5% (w/v) 0.25× TCF-4 binding element (TBE) polyacrylamide gels (13) .
A SmaI-EcoRI blunted fragment of pEMBL8-uPA containing ∼2100 bp of the human uPA promoter/enhancer was subcloned into the SmaI-cleaved dephosphorylated luciferase reporter pBV-luc, yielding the plasmid puPA. Next, the TBEs in puPA were mutated alone (puPA1M and puPA2M) or in combination (puPA1/2M) using double-stranded oligonucleotides (Table 2) ⇓ together with QuikChange mutagenesis kits (Stratagene, Heidelberg, Germany) following the user’s manual. Luciferase reporter constructs containing four tandem repeats of each of the wild-type (WT) TBEs (p4×1, p4×2) or mutant TBEs (p4×1M, p4×2M) were subcloned by ligating the corresponding polymerized oligonucleotides (T1, T2 or M1, M2; Table 1 ⇓ ) into the dephosphorylated SmaI site of pBV-luc.
Cell Lines, Cell Culture, and Transfections.
293T, DLD-1, SW480 (American Type Culture Collection, Manassas, VA), 293 β-cat, and 293 β-gal (gift of Paul Polakis, Departments of Molecular Oncology and Molecular Biology, Genentech Inc., San Francisco, CA) were maintained in DMEM with 10% (v/v) fetal bovine serum, 2 mm Glutamax I, and 50 μm 2-mercaptoethanol (all from Invitrogen, Karlsruhe, Germany) without antibiotics. For transient transfections, 293T and SW480 cells were seeded in 12-cluster well plates and transfected 24 h later with 600 ng luciferase reporter plasmids, 50 ng standard plasmid pSV40-β-galactosidase (Promega, Heidelberg, Germany), and 600 ng of pcDNA or pcI-neo-derived vectors expressing β-catenin, TCF-4, c-jun, c-fos, or ets-1 as indicated, using 7.5 μl Superfect (Qiagen, Hilden, Germany) following the user’s instructions. pcDNA3-Cat (Invitrogen) was added to make up DNA to constant amounts of 1.25 μg in total. After an additional 16–28 h, cells were harvested, lysed, and luciferase assays done using Dual Light kits (Applied Biosystems, Darmstadt, Germany) following the supplier’s manual. Results were normalized on β-galactosidase activity. For knockdown experiments DLD-1 or SW480 cells were transfected with 3.9 μg double-stranded small interfering (si)RNA specific for β-catenin or (14) cdc2 and lamin A/C (Table 3) ⇓ as controls, respectively, using 9.6 μl TransMessenger (Qiagen) following Qiagen’s instructions. Cells were harvested ∼48 h after transfection.
RNA Isolation, Real Time PCR, and cDNA Microarrays.
Total RNA from transfected cell lines was prepared using RNeasy kits (Qiagen). One μg RNA was reverse transcribed using oligo(dT)15 mixed with random hexamer primers (1:1) and Superscript II (Invitrogen) following the user’s manuals. For the analysis of semiquantitative reverse transcription-PCRs 1 μl of cDNA or RNA was amplified in 25 cycles using primer pairs specific for GAPDH or uPA (Table 4) ⇓ . After electrophoresis bands were quantified by densitometry with the help of an Eagle Eye device (Stratagene). cDNA and RNA from siRNA transfection experiments were analyzed using real-time PCR using an ABI Prism 7700 Sequence Detection System (Applied Biosystems) together with intron spanning primer pairs specific for β-actin, β-catenin, GAPDH, c-myc, RelA, or uPA (Table 5) ⇓ and SYBRGreen PCR Master Mix kits (Applied Biosystems) following the user’s instructions. All of the reactions were repeated at least twice and in duplicates. Because the RNA microarray data will be presented elsewhere 4 a very brief description of the method is given only (Supplementary Data).
Chromatin Immunoprecipitation (ChIP).
Chromatin was isolated from formaldehyde-treated colorectal tumor cells DLD-1 and SW480, fragmented to a mean size <600 bp and subjected to ChIP using Chromatin Immunoprecipitation Assay kits (Upstate Biotechnology, Hamburg, Germany) together with 1 μg of β-catenin-specific antibody (clone 14; BD Transduction Laboratories, Heidelberg, Germany) following Upstate’s protocol. Several primer pairs specific for β-actin, melanoma growth stimulatory activity α, c-myc, and uPA (Table 6) ⇓ were used for investigating binding of β-catenin to DNA. For analyzing chromatin input, one-fortieth and for all other reactions one-tenth of the precipitated chromatin was taken as a template.
Results and Discussion
Gene expression profiles of human colorectal adenocarcinoma tissue taken from central areas or invasion fronts using Affymetrix-RNA microarrays showed that RNA levels of uPA were 2.5-fold up-regulated at invasion fronts. These results were confirmed by immunohistochemistry of 27 well-differentiated colorectal adenocarcinomas from another collection using antibodies specific for uPA or β-catenin. Prominent uPA expression was seen in tumor cells at the invasion front (Fig. 1A) ⇓ strongly expressing nuclear β-catenin (Fig. 1B) ⇓ , compared with low uPA expression in central parts of the same tumors (Fig. 1C) ⇓ displaying membranous and cytoplasmic but missing nuclear β-catenin staining (Fig. 1D) ⇓ . Moreover, both nuclear β-catenin and uPA expression correlated strongly with the localization of tumor cells at the invasion front (Mann-Whitney test, pMW < 0.001; Fig. 1E ⇓ ) and with each other (χ2 test, pχ < 0.001; Fig. 1E ⇓ ). Besides confirming data of high uPA (11) and nuclear β-catenin expression (7) , it was suggested that uPA may be a bona fide β-catenin target gene.
The human uPA promoter/enhancer (accession no. X12641) contains two consensus TBE motifs (WWCAAAG; Ref. 15 ) at positions, −737 (TBE1) and −562 (TBE2). They are embedded in a variety of other responsive elements (Fig. 2A) ⇓ important for the activity of this promoter/enhancer (16) . Firstly, we investigated whether both TBEs of the uPA promoter/enhancer bind GST-TCF-4 fusion proteins using electrophoretic mobility shift assays (Fig. 2B) ⇓ . GST-TCF-4 bound to radioactively labeled WT TBEs (T1 and T2; Fig. 2B ⇓ , lanes 2 and 11) but not to labeled mutant TBEs (M1 and M2; Fig. 2B ⇓ , lanes 9 and 18). Binding was competed on addition of a 30-fold molar amount of either unlabeled WT oligonucleotides from the uPA (T1 and T2; Fig. 2B ⇓ , lanes 3, 5, 12, and 14) or the c-myc promoter/enhancer (MY; Fig. 2B ⇓ , lanes 7 and 16) but not with uPA promoter/enhancer-derived mutant oligonucleotides (M1 and M2; Fig. 2B ⇓ , lanes 4, 6, 13, and 15). No binding to GST alone was seen (G; Fig. 2B ⇓ , lanes 8 and 17). Thus, using defined biochemical conditions, both TBEs of the human uPA promoter/enhancer bound specifically to TCF-4.
Secondly, transient transfection experiments using luciferase reporter constructs were done to investigate the role of both TBEs for the transactivation of the human uPA promoter/enhancer. Firstly, puPA (Fig. 3A) ⇓ , containing a 2100-bp fragment upstream of the uPA transcription start site, was transiently transfected with increasing amounts of expression plasmids encoding a degradation-resistant form of β-catenin in 293T cells. A dose-dependent increase in the activity of the uPA promoter/enhancer was observed, which was maximum after the addition of 0.3 μg of β-catenin expression plasmids (Fig. 3B ⇓ ; 7.4-fold). The importance of the transcription factors AP1- and Ets-1 for transactivating the human uPA gene was described (16) . Thus, we investigated the effects of combinations of c-jun and c-fos (both gene products combine to AP-1), ets-1, and β-catenin on the transactivation of the puPA promoter/enhancer. A combination of AP-1 and ets-1 induced a stronger activation than β-catenin alone (Fig. 3B ⇓ ; 15.8-fold compared with 7.4-fold), but together with β-catenin, a synergistic stimulation was seen (Fig. 3B ⇓ ; 42.9-fold). Next, either TBE alone (puPA1M and puPA2M; Fig. 3A ⇓ ) or in combination (puPA1/2M; Fig. 3A ⇓ ) were mutated and transiently cotransfected in 293T cells together with β-catenin expression plasmids. A loss of ∼65% of transactivation activity for puPA1M or puPA2M and ∼90% for puPA1/2M was observed (Fig. 3C) ⇓ . Moreover, dominant-negative TCF-4 (dnTCF-4) expression clones, encoding a TCF-4 molecule, that binds to DNA but not β-catenin, suppressed the transactivation of the uPA promoter/enhancer luciferase reporter construct in SW480 cells to a residual activity of 20%. In contrast, the already low activity of puPA1/2M was almost unaffected (Fig. 3D) ⇓ . Finally, luciferase reporter constructs containing tetrameric tandem repeats of the WT (p4×1, p4×2; Fig. 3A ⇓ ) or mutated TBEs (p4×1M, p4×2M; Fig. 3A ⇓ ) were transiently cotransfected with dnTCF-4 expression plasmids in SW480 cells. As expected, the luciferase activity of the WT constructs p4×1 and p4×2 was strongly suppressed (Fig. 3E) ⇓ , whereas the low basal activity of the mutant constructs p4×1M and p4×2M was only slightly affected (Fig. 3E) ⇓ . Taken together β-catenin transactivates the human uPA promoter/enhancer directly via both TBEs independently and works in synergy with other transcription factors, like AP-1 and Ets-1. For cloning the uPA promoter/enhancer luciferase reporter constructs, the luciferase reporter vector pBV-luc was used as it shows low background activity in β-catenin-regulated transfections (13) . Thus, our uPA luciferase reporter constructs contain two promoters, one from the minimum thymidine kinase promoter of herpes simplex virus in pBV-luc and the other from the uPA gene. Because all of the data are based on the same (Fig. 3B) ⇓ or TBE-mutated derivatives (Fig. 3, C and D) ⇓ as the control, possible side effects of the thymidine kinase promoter should be self-eliminating. Moreover, K-ras, a well-known inducer of uPA gene expression (17) , may partly drive the uPA promoter/enhancer activity in cultured SW480 cells, as this cell line harbors a point mutation in codon 12 of its K-ras gene. Nevertheless, mutations of both TBEs lead to a strong reduction in the transactivation activity (Fig. 3D) ⇓ , which clearly demonstrates a direct role for β-catenin in transcriptional control of the human uPA promoter/enhancer. The suppressive effect of dnTCF-4 is not due to toxic effects of the DNA, as the same amount of control expression plasmids was used in the controls. Moreover, we have shown previously that the dnTCF-4 amounts used in our experiments exert specific suppressive effects when compared with TCF-4 expression plasmids (18) . Notably, dnTCF-4 cannot shut down the uPA promoter/enhancer activity completely, which again indicates the involvement of other transcription factors.
Thirdly, the effect of β-catenin on the regulation of uPA transcription in the context of the native chromatin context was investigated. Firstly, 293 cells stably expressing a stabilized form of β-catenin (293 β-cat) showed a 3.2-fold increased expression of uPA mRNA compared with β-galactosidase expressing cells (293 β-gal; Fig. 4A ⇓ ). RNA template controls done without reverse transcription did not give rise to PCR products (data not shown). Thus, highly overexpressed β-catenin transactivates the uPA gene. Secondly, β-catenin mRNA was knocked down in DLD-1 and SW480 cells to residual amounts <25% depending on the siRNA and cell lines used (Fig. 4C ⇓ ; β-catenin, cat1, and cat2). mRNA copy numbers of the β-catenin target gene c-myc (13) but also uPA were down-regulated to ∼10% and 20%, respectively (Fig. 4B ⇓ ; c-myc, uPA, cat1, and cat2). Conversely, the expression of the housekeeping gene GAPDH and the transcription factor RelA (nuclear factor κB p65), which is no target gene of β-catenin, (14) are only unspecifically affected by β-catenin specific siRNAs (Fig. 4B ⇓ , relA, GAPDH, cat1, and cat2). Moreover, Lamin A/C-specific siRNA has no specific knockdown effect on any of the genes under investigation (Fig. 4B ⇓ , lam1 and lam2), although this siRNA induces loss of lamin expression (Supplementary Data; Fig. 1 ⇓ ). For RNA control templates the relative copy numbers were <5% (median 1%) in all of the cases (data not shown). Thus, knocking down β-catenin mRNA levels leads to an incomplete decrease of uPA mRNA levels indicating again that β-catenin plays an important but not exclusive role in the regulation of the human uPA gene (16) in human colorectal cancer cells. Finally, ChIP experiments were done using a β-catenin-specific antibody and fragmented chromatin from the colorectal carcinoma cell lines SW480 and DLD-1, both of which express endogenously uPA (data not shown). It was possible to precipitate chromatin of the c-myc and uPA promoter/enhancers spanning the respective TBEs but not the chromatin of genomic β-actin or melanoma growth stimulatory activity α fragments that are unrelated to β-catenin binding (Fig. 4C ⇓ ; ChIP), although a very weak signal was obtained for c-myc in DLD-1 cells. Input controls showed that comparable amounts of chromatin were used in all of the experiments and that PCR efficiencies of the different reactions were comparable (Fig. 4C ⇓ , input). Thus, β-catenin binds to the human uPA enhancer under physiological conditions in a native chromatin context without any exogenous disturbance.
Taken together, our data show that uPA is a direct target gene of β-catenin in the context of other transcription factors (Fig. 3B) ⇓ , which was described for other human β-catenin target genes, like MMP7 or laminin-5γ2 (Fig. 1C) ⇓ . This might explain the high uPA expression at the invasive front (Fig. 1A) ⇓ displaying high amounts of β-catenin (Fig. 1B ⇓ ; Refs. 7 , 11 ). Together with the β-catenin target genes MMP-7, MT1-MMP, laminin-5γ2, and CD44, uPA may promote invasion and metastasis, a hallmark of cancer (19) , in colorectal tumors especially at the invasion front (8) . Moreover, these tumor cells, which are in a resting state indicated by high expression of the cell cycle inhibitor p16INK4A (12) , might be protected from dormancy (10) by the expression of uPA. Interestingly other integral parts of the dormancy regulation system, the uPA receptor and fibronectin, have already been shown to be transactivated via β-catenin (8) . Thus, it is attractive to speculate that the integrins, another essential component of dormancy regulation (10) , may also be regulated by β-catenin.
Although most colorectal adenomas and carcinomas have many genetic alterations in common, especially with respect to the deregulation of the Wnt signaling pathway, it remains open why adenomas are not growing invasively. It seems that the microenvironment, in which colorectal tumor cells are embedded, might modify the characteristic of genetic mutations. Thus, β-catenin is not necessarily found in all colorectal tumor cells in the nucleus, although they display adenomatous polyposis coli mutations (7 , 8 , 20) .
Material was generously provided by Francesco Blasi (pEMBL8-uPA), Howard Crawford [pets-1 (PU.1)], Paul Polakis (stably β-catenin transfected 293 cells), Edgar Serfling (pfos and pjun), and Kenneth Kinzler and Bert Vogelstein [pBV-luc, pΔ45-β-catenin, pGST-TCF4 (DBD), pTCF-4, and pdnTCF-4 ]. Ludger Klein-Hitpass did parts of the microarray analysis, and Johannes Bode gave important hints for the ChIP assay. We thank Kerstin Amann for providing digital photographic equipment (AnaLysis), and Stephen Köver and Gerald Niedobitek for critical reading and discussion.
Grant support: Wilhelm-Sander-Stiftung Az. 99.065.2 (A. Jung and T. Kirchner) and by Bundesministerium für Bildung und Forschung; Nationales Genomforschungsnetz KR-S05T02 (T. Brabletz, A. Jung, and T. Kirchner).
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.
Notes: E. Hiendlmeyer and S. Regus contributed equally to this work. S. Regus is currently at Klinikum Bamberg, Abt. Chirugische Klinik, 96049 Bamberg, Germany. Part of this work was presented at the 91st Annual Meeting of the American Association for Cancer Research and was supported by a travel grant sponsored by Aventis Inc. (A. Jung). Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org).
Requests for reprints: Andreas Jung, Pathologisches Institut, Universität Erlangen-Nürnberg, Krankenhausstr. 8-10, 91054 Erlangen, Germany. Phone: 49-9131-8522610; Fax: 49-9131-8524745; E-mail:
↵4 F. Hlubek, T. Brabletz, A. Jung, J. Budczies, S. Pfeiffer, S. Spaderna, and T. Kirchner. Gene expression profiles of colorectal cancer invasion, manuscript in preparation.
- Received December 15, 2003.
- Revision received December 23, 2003.
- Accepted December 31, 2003.
- ©2004 American Association for Cancer Research.