The Dickkopf-homolog 3 is expressed in tumor endothelial cells and supports capillary formation



A critical event in tumor growth and progression is the upregulation of angiogenesis. Thus, targeting angiogenesis has become an attractive treatment modality in cancer medicine. Our study analyzed various solid tumor types for the expression of Dkk-3, a cystein-rich, N-glycosylated secreted member of the dickkopf protein family that has been proposed as a tumor suppressor gene. Tissue microarrays of gliomas (n = 30), high-grade non-Hodgkin's lymphomas (NHL, n = 80), colorectal cancer (n = 35) and melanoma (n = 30) were immunohistochemically analyzed for Dkk-3 and CD31 expression. Moreover, the effects of Dkk-3 were studied in vitro in primary endothelial colony-forming cells (ECFC) and in vivo in a mouse melanoma model. In comparison to normal tissue, the number of blood vessels expressing Dkk-3 was increased in glioma, high-grade NHL, melanoma and colorectal carcinoma. Confocal laser scanning microscopy revealed that Dkk-3 vesicles localized also in Weibel Palade bodies. In vitro cell proliferation and migration of ECFC was not significantly affected by adenoviral overexpression or siRNA-mediated downregulation of Dkk-3. Interestingly, tube formation in matrigel decreased after downregulation of Dkk-3 and increased after adenoviral overexpression. Stable overexpression of murine Dkk-3 in B16F10 cells significantly increased microvessel density in the C57/BL6 melanoma model. Thus, we postulate a novel function of Dkk-3 in endothelial cells as a differentiation factor involved in remodeling the tumor vasculature. © 2007 Wiley-Liss, Inc.

Growth, progression and metastasis of solid tumors are strongly dependent on the supply of oxygen and nutrients.1 Without adequate blood supply tumors become hypoxic, acidotic and necrotic. Therefore, growing tumors induce an angiogenic switch2 and produce a variety of growth factors that are responsible either for remodeling of the pre-existing vascular network by means of angiogenic sprouting or intus-susceptive growth or for neovascularization by recruiting circulating endothelial cells.3 Capillaries of the tumor vascular bed show a variety of structural abnormalities, such as the lack of a basement membrane, low numbers of surrounding pericytes, mosaic vessel structure and altered expression patterns of leukocyte adhesion molecules.4 On the molecular level, tumor endothelial cells show a distinct gene expression profile5 that differs from that of normal endothelial cells.6 Thus, markers specifically expressed or overexpressed on tumor endothelial cells or secreted into the perivascular space are potential candidates for vascular targeting.7

Among the dickkopf family of secreted modulators of wnt signalling, Dkk-3 is the most divergent member with still unknown signal transduction and function.8 Dkk-1 and Dkk-2 have been shown to bind to Kremen receptors, resulting in down-regulation of the Wnt-coreceptor lipoprotein-related protein from the cell surface.9–11Dkk-1 knock-out mice show severe defects in head and limb formation,12 whereas Dkk-2 knock-outs are blind and have defects in osteoblast differentiation and bone mineralization.13 Interestingly, Dkk-3 knock-out mice show no major alterations in organ morphology, but abnormalities in hematological and immunological parameters.14 The Dkk-3 gene is strongly expressed in the developing heart and blood vessel system of mice and chicken.15 Interestingly, in humans the gene is downregulated in a variety of tumor cell lines by CpG island methylation.16, 17 Since tumor angiogenesis in many aspects resembles the early stages of primitive embryonic vasculogenesis,18 we investigated the expression of Dkk-3 in human tumor vessels. In our study we demonstrate Dkk-3 protein expression in blood vessels of highly vascularized neoplasms, such as malignant glioma, high-grade non-Hodgkin's lymphoma (NHL), melanoma and colorectal carcinoma. Moreover, we show that peripheral blood derived endothelial colony-forming cells (ECFC) contain this protein in their secretory apparatus and in Weibel Palade bodies and release it upon TNFα stimulation and capillary tube formation. Finally, we show that overexpression of murine Dkk-3 increased microvessel density in B16/F10 melanoma in the C57/BL6 mouse model.

Material and methods

Tissue samples

The glioma tissue microarray (TMA) consisted of duplicate cores of brain biopsies of glioma tissue (30 cases) and normal brain (n = 3). The lymphoma TMA was composed of 80 cases. Samples from normal human lymph node biopsies (n = 5) and additional samples of normal human brain (n = 2) were kindly provided by Dr. Alexandar Tzankov (Department of Pathology, University of Basel, Switzerland) and were used as normal controls. The melanoma TMA consisted of triplicate cores of human primary melanoma (n = 30) and normal skin tissues (n = 9) and the colon carcinoma TMA of duplicate cores of colonic adenocarcinoma (n = 35) and cases of normal tissue from autopsy (n = 5). All TMAs were obtained from US Biomax. Detailed information about all tumor samples can be found on the supplier's web site (

Tissue culture

All procedures were carried out in accordance with the guidelines of the institutional review board. Written informed consent was obtained from all blood donors. ECFC clones were generated from cord blood-derived mononuclear cells (CBMNC, n = 5) using Ficoll density gradient centrifugation (Lymphoprep). Isolated cells were resuspended in EGM-2 (Cambrex), placed in a 6-well plate coated with type I collagen (from kangaroo, Sigma-Aldrich) and incubated at 37°C. After 24 hr, nonadherent cells were removed by changing the medium. Thereafter, medium was changed every 3 days. Expanded ECFC cell clones were passaged on collagen type-I (from calf skin, Sigma) coated tissue culture flasks (Costar, Corning) using collagenase-I (Sigma) and Trypsin-EDTA (Gibco-Life Technologies). ECFC were characterized by FACS for the expression of endothelial surface markers and for angiogenic features as described elsewhere.19

Dkk-3 protein expression and secretion was studied after stimulation of ECFC for 24 hr with 1, 10 and 100 ng/ml VEGF, bFGF, G-CSF, SDF-α1, or IFN-y, TNF-α, IL-1α and IL-6 (Immunotools) in basal EGM2 medium containing only 1% FCS and no further supplements. Supernatants were concentrated by the use of 10 KDa centrifugal filter devices (Microcon YM-10, Millipore). ECFC proliferation and migration was analyzed after stimulation with 1, 10 and 100 ng/ml recombinant hDkk-3 (R&D Systems). For regulation of Dkk-3 gene expression mRNA was extracted from log-phase growing ECFC, contact inhibited (growth arrest) cells and capillary tubes isolated from matrigel by the use of Cell Recovery Solution (BD-Biosciences).

Western blot analysis

Cells were harvested and lysed in a buffer containing 10 mM Tris-HCl pH 7, 0.2% Triton X-100, and protease inhibitors (Complete Mini EDTA-free; Roche Applied Science); 20 μg total protein was denaturated and transferred to an Immuno-Blot™ polyvinylidene difluoride membrane (Bio-Rad). After blocking the membrane in 3% skim milk powder dissolved in PBS, it was probed with primary antibodies directed against alpha tubulin (Sigma-Biochemicals, mouse monoclonal), Dkk-3 (R&D systems, mouse monoclonal) or murine Dkk-3 (R&D systems, goat polyclonal) and vWF (Dako Cytomation, rabbit polyclonal), for 1–2 hr and incubated with a 1:2,500 dilution of an HRP-conjugated goat anti-mouse IgG or a 1:2,500 dilution of an HRP-conjugated goat anti-rabbit IgG (Promega) for 1 hr. After washing, a chemoluminescent substrate (Super Signal West dura extended, Pierce) was added to the membrane, which was then exposed to the ECL Hyperfilm (Amersham Biosciences).

Immunofluorescence and confocal microscopy

Cells were plated on 8-well culture slides (Falcon BD Labware). After being fixed in acetone/methanol (1:1) and permeabilized with 0.2% Triton-X-100 cells were blocked with PBS containing 3% BSA for 45 min at room temperature (RT). vWF (1 μg/ml rabbit polyclonal anti-human vWF, DakoCytomation) and Dkk-3 (1 μg/ml goat polyclonal anti-human Dkk-3, R&D Systems) antibodies were applied at RT for 2 hr. Direct competition with a 50-fold excess of recombinant Dkk3 (R&D Systems) on the slide served as specificity control. After washing in PBS cells were incubated with secondary fluorochrome-labeled antibodies (polyclonal swine anti-rabbit TRITC, followed by polyclonal rabbit anti-goat FITC, DakoCytomation) and nuclei counterstained with TOP-RO-3 iodide (Molecular Probes). Cells were embedded in fluorescent mounting medium (DakoCytomation) and viewed by CLSM (Zeiss Axiophot, μ-Radiance Scanning System, Carl Zeiss Laser Optics/Laser Sharp Software, Bio-Rad Laboratories). For colocalization studies tissue sections of colorectal carcinoma were incubated for 2 hr with antisera directed against Dkk-3 (4 μg/ml) and vWF (1 μg/ml, Dako Cytomation). Primary antisera were detected in 2 steps by a first incubation with swine anti-rabbit IgG/FITC followed by additional 45 min of incubation with rabbit anti-goat IgG TRITC. Immunofluorescence was detected by CLSM.


Paraffin-embedded tissue sections were deparaffinized and hydrated in xylene and graded alcohol series. Thereafter, antigen retrieval was performed by microwave treatment in citrate-buffer (10 mM, pH 6.0) and endogenous peroxidase activity was blocked with 3% H2O2/methanol. Sections were incubated in blocking solution containing 10% bovine calf serum (Dako Cytomation) for 45 min and then stained for 1 hr with primary antiserum (goat anti-human Dkk3 polyclonal 1 μg/ml, R&D Systems). Moreover, serial sections were incubated with a monoclonal mouse anti-human CD31 (1:40, Dako Cytomation). Primary antiserum was detected after incubation with a biotinylated secondary antibody (biotinylated horse anti-goat IgG, biotinylated rabbit anti-mouse IgG, Vector Laboratories) using the Vectastain Elite ABC Kit (Vector Laboratories) and the FAST DAB Tablet Set (Sigma). Sections were counterstained with Meyer's Haemalaun and mounted with Pertex. Specificity controls of the Dkk-3 polyclonal antibody were performed by blocking experiments with an excess of recombinant hDkk-3 (R&D Systems). Positive-stained vessels of the cores of each tumor sample were counted by 2 independent observers and the mean MVD per mm2 was subsequently calculated.

For colocalization studies tissue sections of colorectal carcinoma were incubated for 2 hr with antisera directed against Dkk-3 (4 μg/ml) and vWF (1 μg/ml, Dako Cytomation). Primary antisera were detected according to a 2 step protocol by incubation with swine anti-rabbit IgG/FITC followed by additional 45 min of incubation with rabbit anti-goat IgG TRITC. Immunofluorescence was detected by confocal image scanning.

Transfection of siRNA into primary endothelial cells

ECFC were transfected using the Matra reagent (magnetic nanoparticles, IBA Biotechnology). Dkk-3 siRNA sense 5′-UUCACAAGAUAACCAACAA-3′ (corresponding to nucleotide 340–359 relative to the start codon) and Dkk-3 antisense 5′-UUGUUGGUUAUCUUGUGAA-3′ oligonucleotides were purchased from MWG-Biotech and transfected into cells by means of magnetic nanoparticles (IBA Biotechnology). Control siRNA directed against KIAA1036 was purchased from MWG-Biotech (sense-5′ AUGCCCUGGAAGCUGUGAUC and antisense 5′ GAUCACAGCUUCCAGGCAU). RNA was extracted 6 hr after transfection, and protein after 48 hr. ECFC proliferation (30,000 ECFC/well; 24-well format) was analyzed 96 hr post-transfection. For matrigel and cell migration assays cells were used 48 hr after transfection with siRNA.

Cell proliferation assays

Approximately 30,000 ECFC were seeded into each well of a 24-well plate in culture medium and allowed to adhere overnight before the tissue culture medium was changed. Cells were detached and cell numbers determined after 3 and 5 days using the Buerker counting chamber.

Tube formation assay

To analyze tube formation, 24-well plates were coated with 200 ml growth factor-reduced matrigel (BD Biosciences). ECFC were resuspended in 200 ml EGM-2 medium (1 × 105 cells) and placed on top of the polymerized matrix; tube formation was observed after 6 hr. Tubes were viewed under an inverted transmission microscope (Zeiss Axiovert 200M) and documented with a digital imaging system (Axiovision Software, Zeiss).

Cell migration assay, Scratch assay

ECFC were seeded onto 6-well cell culture plates. Once at confluence, cells were serum-starved overnight in medium containing 0.5% FBS. Scratch injury was then administered using a disposable pipette tip (1-mm width). After injury, the monolayer was gently washed with PBS, and the medium was replaced with medium containing 1% FBS. Endothelial cells sprouting from the edge of the injured monolayer were examined and photographed at 2, 4 and 6 hr after scratching. Migrated endothelial cells were counted in 10 randomly selected high-power fields (HPF) adjacent to the scratch injury and are expressed as cells/mm2.

Quantitative PCR

RNA was purified by cell lysis and nucleic acid extraction using the RNeasy Kit (Qiagen). Thereafter, viral and genomic DNA in the RNA samples were digested with the RQ1 DNAse (Promega). Extracted total RNA was transcribed into cDNA by oligo-dT- und hexanucleotide-random primers and the AMV-Reverse Transcriptase (all Promega). For analysis in the quantitative PCR 20 ng of each cDNA was used; 5 μl Sybr-green Mix (Bio-Rad), and 10 pMol of each primer were mixed with the cDNA sample. The primers used for quantification were elongation factor 1α (for: 5-cacacggctcacattgca; rev: 5-cacgaacagcaaagcgacc) and human Dkk-3 (for: 5-tcatcacctgggagctagag; rev: 5-caacttcatactcatcgggg).

B16F10 melanoma cDNA was analyzed by primers specific for murine actin beta (mACTB, for: 5-aagagctatgagctgcctga; rev: 5-tacggatgtcaacgtcacac), dickkopf 3 (mDkk-3, for: 5-atgggaccatctgtgacaac; rev: 5-gcttgcacatgtacaccaga), vascular growth factor (mVEGF, for: 5-ttactgctgtacctccacc; rev: 5-acaggacggcttgaagatg), angiopoietin 1 (mANG1, for: 5-aaatgcgctctcatgctaac; rev: 5-cagctttctttgcagctttc) and angiopoietin 2 (mANG2, for: 5-gagcaaaccaccttcagaga; rev: 5-atcttctcggtgttggatga).

Analysis was performed within 50 cycles in the Bio-Rad iCycler (Bio-Rad). Data were collected and analyzed with iCycler Software. Specificity of the amplified PCR product was checked by melting curve analyses, gel electrophoresis and sequencing.


Replication-defective adenoviruses were generated with the Ad Easy Adenoviral vector system (Stratagene) according to the manufacturer's instructions. In brief, the hDKK-3 cDNA was subcloned into the pShuttle CMV GFP vector. Recombinant adenoviral DNA was generated in BJ5183 bacteria cells using a double-recombination event between cotransfected adenoviral backbone plasmid vector, pAdEasy-1, and a shuttle vector carrying the gene of interest. For generation of replication-defective adenovirus recombinant DNA was transfected into HEK293 cells by means of Lipofectamin 2000 (Invitrogen), after which cytosolic extracts were prepared. All viral titers were determined by qPCR for the gene coding for the encapsulation signal (for:5-cgacggatgtggcaaaagt, rev: 5-cctaaaaccgcgcgaaaa) and the respective viral plasmid DNA standards. ECFC were transfected with an MOI of 100 viruses/cell and tested for gene and protein expression 24 or 48 hr after transfection. All cell proliferation, migration and tube formation assays were performed 48 hr after adenoviral transfection.

Generation of transgenic B16 F10

Murine Dkk-3 (clone IRAVb968A0291D, RZPD, Berlin, Germany) was subcloned in a transposon vector with a strong EF1α promoter (pT2 neo EF1α). Thereafter, nucleotide sequence was confirmed by double-strand sequencing. B16F10 cells were transfected with the empty pT2 neo EF1α vector and the vector containing mDkk-3. Cotransfection was performed with the pCMV SB11 vector encoding the sleeping beauty (SB) transposase. Clones resistant to neomycin were selected and propagated from the ClonaCell medium (Stem Cell Technologies) and then analyzed by Western Blot for mDkk3 protein expression.

Mouse melanoma model

Animal experiments were performed after review and approval by the Austrian Government (Ministry of Culture, Science and Education) according to the local committee for animal studies and established standards for human handling. Female C57/BL6 mice, 5- to 6-weeks-old, were purchased from Harlan (Birchen, Germany). Mice were inoculated subcutaneously in the right flank with 1 × 106 B16F10 cells dissolved in PBS. All clones (pT2 mDkk#1, pT2 mDkk #2, pT2 mDkk #3) and controls (pT2 co.#1, pT2 co.#7, pT2 co.#8) used were analyzed for mDkk-3 gene expression and in vitro proliferation before in vivo use. Experiments involved 2 mice per group (total 12 animals). After 1 week tumors were measured daily using a caliper. After 14 days animals were sacrificed, tumor weight determined and cryosections prepared for CD31 (PECAM-1) immunofluorescence. Moreover, mRNA and protein were extracted from all tumors to verify mDkk-3 gene/protein expression in vivo.

Immunofluorescent staining of microvessels

Cryosections of B16 melanoma were prepared after freezing the removed tumor tissue in Tissue-Tec (Sakura Finetech) using methylbuthanol and liquid nitrogen. Thereafter, cryosections of 5 μm were cut and fixed in acetone. Sections were blocked in 3% BSA/PBS and incubated with a 1 μg/ml rat anti-mouse CD31 (PECAM-1, BD Biosciences). Detection was performed using a goat anti-rat Alexa-488 antibody (Molecular Probes, 4 μg/ml). Tumors were examined with the Axiovert 200 M fluorescence microscope and Axiovert Software (Zeiss). Microvessel density was determined by 2 independent observers counting 10 different high-power fields (HPF, 200×) for each tumor.

Statistical analyses

Statistical analyses were performed with the GraphPad Prism™ software for Windows. All tests of statistical significance were 2-sided. Student's t test and the Mann–Whitney U test were used to study differences between 2 groups. The Kruskal–Wallis H test was applied to study differences between the 3 glioma groups. Statistical analyses of quantitative PCR data were performed according to the delta Ct method described by Pfaffl20 and p values were calculated with Student's t test.


Dkk-3 is expressed in tumor vessels

Whereas tumor vessels from malignant glioma specimens as well as blood vessels from the surrounding healthy brain tissue stained positive for CD31, only tumor vessels stained positive for Dkk-3 (Fig. 1). Notably, with a median of 83 Dkk-3-positive microvessels per mm2, aggressive gliomas (i.e., grade IV) displayed a significantly higher MVD as compared to grade II (43 microvessels per mm2) or grade III (37 microvessels per mm2) glioma specimens (p < 0.01). To explore whether Dkk-3 expression by tumor vessels is also detectable in other types of cancer, 80 high-grade NHL samples and normal lymph nodes (n = 5) were examined. In parallel to the results obtained with normal brain tissue, vessels from normal lymph nodes were virtually Dkk-3-negative (Fig. 2b). Despite strong staining for CD31 (37 microvessels per mm2) normal lymph nodes exhibited only a mean of 3 microvessels per mm2 that were positive for Dkk-3 (p < 0.01). NHL showed a strong staining for Dkk-3 (38 microvessels per mm2, Fig. 2a). Assessment of MVD by Dkk-3 staining was comparable to CD31 (35 microvessels per mm2, p = 0.96).

Figure 1.

Immunohistochemical detection of Dkk-3 in blood vessels from normal brain and aggressive glioma Blood vessels from healthy brain tissue stain positive for CD31 (a) and negative for Dkk-3 (b). Tumor blood vessels of high-grade glioma stain positive for CD31 (c) and Dkk-3 (d). Blood vessels are marked by arrows. Magnification ×400.

Figure 2.

Immunohistochemical detection of Dkk-3 in high-grade NHL, colorectal cancer, melanoma and control tissue High-grade NHL (n = 80), melanoma (n = 35), colorectal carcinoma (n = 30) and normal control tissues (n = 19) were stained for Dkk-3. High-grade NHL specimens stain positive for Dkk-3 (a), normal lymph nodes do not (b). The same was observed for colorectal carcinoma (c) and normal colon mucosa (d), melanoma (e) and normal skin (f). Magnification ×400. Note: The normal epidermis (keratinocytes) and enterocytes of the colon show immunoreactivity for Dkk-3.

Dkk-3 expression was also detected in human colorectal carcinoma and primary melanoma (Figs. 2c/2e). Normal colon mucosal tissue (n = 5) exhibited a weak immunoreactivity and only few dispersed vessels that stained positive for Dkk-3 (Fig. 2d). In contrast, tumor blood vessels were strongly positive in 69% of adenocarcinoma specimens analyzed (n = 35). Similar observations were made in primary melanoma (70% positive, n = 30). Microvessels in melanoma specimens showed strong staining for Dkk-3 (Fig. 2e), whereas in normal skin only few Dkk-3 positive vessels were detectable. Keratinocytes of the epidermis stained positive for Dkk-3 (Fig. 2f), whereas tumor cells were consistently negative.

Dkk-3 is released upon capillary tube formation and stimulation with TNFα

Dkk-3 gene expression was analyzed in whole blood samples (PBMNC) and ECFC established from CD34+ cells of cord blood by means of qPCR. ECFC displayed an approximately 120-fold higher Dkk-3 gene expression/20 ng total RNA (p < 0.01). Dkk-3 gene expression was not significantly (2 times) upregulated by contact inhibition or tube formation in matrigel (Fig. 3a). Interestingly, Dkk-3 protein decreased in the cytoplasm after tube formation (Fig. 3b) and was increased in the respective supernatant (Fig. 3c). Having observed that Dkk-3 is strongly expressed in tumor blood vessels, we hypothesized that distinct angiogenic cytokines produced by tumor cells might upregulate Dkk-3 expression. Thus, ECFC were stimulated with various angiogenic cytokines. VEGF at concentrations of 1, 10 and 100 ng/ml (Fig. 3d) and basic FGF, GM-CSF, and SDF1α did not affected Dkk-3 protein expression in vitro (Fig. 3f). Secretion of Dkk-3 by ECFC into cell culture supernatants as assessed by Western Blot was not influenced by any of the angiogenic factors tested (data not shown). Interestingly, Dkk-3 protein expression declined 48 hr after stimulation with the proinflammatory cytokine TNFα (Figs. 3e/3f). This observation led us to test whether Dkk-3 protein is released into culture supernatants after proinflammatory stimulation. Indeed, TNFα stimulation was reponsible for secretion of Dkk-3 into the supernatant (Fig. 3g).

Figure 3.

Regulation of Dkk-3 expression and secretion (a) Dkk-3 gene expression was assessed in log-phase growing (prol.), contact-inhibited (arrested) endothelial cells and after tube formation in matrigel (tubes) by qPCR. (b) Intracellular Dkk-3 protein content was analyzed by Western Blots. Tubulin alpha served as loading control. (c) Supernatants of proliferating, arrested cells and tubes were concentrated and analyzed for secreted Dkk-3 by Western Blot. Tube formation was responsible for a decrease of intracellular Dkk-3 and an increase of Dkk-3 protein in the supernatant. (d) Intracellular Dkk-3 protein content was analyzed by Western Blot 48 hr after stimulation with proangiogenic factors (VEGF,1-100 ng/ml) or (e) proinflammatory cytokines (TNF-α 1-100 ng/ml). TNF-α decreased intracellular Dkk-3 protein. (f) Thereafter, a densitometric analysis was performed after normalization with tubulin alpha. Note: The figure shows data from the 1 ng/ml concentration series. (g) Supernatants of TNF-α stimulated cells (48 hr) were analyzed for secretion of Dkk-3 and vWF by Western Blot. TNF-α (1 ng/ml) increased Dkk-3 and vWF secretion.

Dkk-3 is localized to endothelial cell-specific Weibel Palade bodies

Dkk-3 protein expression was localized in secretory vesicles by immunofluorescence and confocal laser scanning microscopy (Fig. 4a). This signal was efficiently blocked by recombinant Dkk-3 (Fig. 4b). Costaining experiments with vWF (Weibel Palade bodies, Fig. 4c) clearly demonstrated that a large part of the Dkk-3 vesicles colocalized in Weibel Palade bodies (WPB, Fig. 4d). In situ analysis of blood vessels of colorectal carcinoma by immunofluorescence and confocal image scanning confirmed a colocalization with vWF (Fig. 4e).

Figure 4.

Immunofluorescence analysis of Dkk-3 in human endothelial colony-forming cells (ECFC) by confocal laser scanning microscopy Dkk-3 is produced and stored in vesicles of ECFC (a, Dkk-3, green, FITC signal). This signal can be blocked by simultaneous incubation with an excess of recombinant antigen (b, Dkk-3 + antigen). Weibel Palade bodies in ECFC were stained with an anti-von Willebrand factor antibody (vWF, red, TRITC signal, c). In part, Dkk-3 vesicles localized to Weibel Palade bodies (d, vWF+Dkk-3). All cell nuclei were counterstained with To-Pro-3 (inverted blue signal). (e) Tissue section of colorectal carcinoma stained for Dkk-3 and vWF. Note: Yellow signals show colocalization of Dkk-3 and vWF in situ. Magnification: ×630.

Knock-down of Dkk-3 inhibits endothelial tube formation in vitro

The effects of Dkk-3 on endothelial cell proliferation, migration and tube formation capacity were studied by adenoviral overexpression of the Dkk-3 gene and by endogenous downregulation of Dkk-3 by siRNA. Adenoviral overexpression of Dkk-3 mRNA and protein was verified after 24 hr by RT-PCR and after 48 hr by Western Blot, respectively (Fig. 5a). However, Dkk-3 overexpression affected neither cell proliferation (Fig. 5b) nor in vitro cell migration (Fig. 5c). These findings were also confirmed by exogenous addition of recombinant Dkk-3 (data not shown).

Figure 5.

Adenoviral overexpression of Dkk-3 in ECFC (a) ECFC were transfected with GFP and Dkk-3 adenovirus (MOI of 100). Gene expression was analyzed 24 hr after transfection using a semi quantitative RT-PCR with 25 cycles of amplification (upper panel). EF1A served as house keeping gene. Protein expression was assessed 48 hr after transfection using Western Blot analysis for Dkk-3 (lower panel). Tubulin alpha served as internal loading control. Transfection efficiency was between 90 and 95%, as determined by GFP-positive cells after 48 hr. (b) ECFC cell number was determined 5 days after adenoviral transfection and overexpression of GFP or Dkk-3. Cells were counted in triplicate in the Buerker Chamber. Bars indicate mean ± SD. (c) Cell migration was assessed in a scratch assay 48 hr after adenoviral transfection. In comparison to GFP controls (67.4 ± 14.2 cells/mm2), ECFC overexpressing Dkk-3 (75.0 ± 15.7 cells/mm2) did not significantly increase cell migration within 6 hr of observation. (d) Tube formation was assessed 48 hr after adenoviral transfection in a matrigel assay. In comparison to GFP-transfected cells (85.7 ± 8.6 capillaries/mm2) adenoviral overexpression of Dkk-3 resulted in increased tube formation after 6 hr of incubation (100.3 ± 4.0 capillaries/mm2, p < 0.05).

In comparison to GFP-transfected cells (85.7 ± 8.6 capillaries/mm2), adenoviral overexpression of Dkk-3 resulted in increased tube formation (100.3 ± 4.0 capillaries/mm2, p < 0.05, Fig. 5d).

Specific knock-down of Dkk-3 mRNA and protein was done by transfection of ECFC with Dkk-3-specific siRNA. Transfection caused a significant decrease in Dkk-3 mRNA after 6 hr (Fig. 6a upper panel) and protein after 48 hr (Fig. 6a, lower panel). SiRNA-mediated downregulation of Dkk-3 in ECFC inhibited neither cell proliferation (Fig. 6b) nor cell migration in the scratch assay (Fig. 6c). However, downregulation of Dkk-3 gene expression by siRNA resulted in a significant reduction in tube formation capacity in matrigel (32.2 ± 8.5 capillaries/mm2vs. 18.4 ± 6.7 capillaries/mm2, p < 0.05, Fig. 6d).

Figure 6.

Knock down of Dkk-3 gene expression in ECFC by siRNA (a) ECFC were transfected with siRNA for downregulation of Dkk-3 gene expression. Downregulation of Dkk-3 mRNA due to siRNA was analyzed after 6 hr using a semi quantitative RT-PCR with 25 cycles of amplification. EF1A served as house keeping gene. Dkk-3 protein levels were determined 48 hr after transfection with Dkk-3-specific siRNA. Tubulin alpha served as loading control for Western Blot analysis. (b) 48 hr after transfection with siRNA and downregulation of Dkk-3 protein ECFC were analyzed for cell proliferation for further 3 days. Bars indicate mean ±SD. (c) Cell migration was analyzed 48 hr after transfection with Dkk-3-specific and control siRNA. The confluent cell monolayer was wounded by mechanical scratching and cell migration assessed after 6 hr. Migration of ECFC into the wounded area (dotted rectangle) was not significantly altered between control siRNA- (87.0 ± 4.3 cells/mm2) and Dkk-3 siRNA- (88.0 ± 8.5 cells/mm2) transfected cells. (d) ECFC were transfected with Dkk-3 or control siRNA and incubated for 48 hr. Then, cells were analyzed for tube formation in matrigel for 6 hr. Downregulation of Dkk-3 (siRNA Dkk3) resulted in a significant loss of tube formation in comparison to control-transfected cells (32.2 ± 8.5 capillaries/mm2vs. 18.4 ± 6.7 capillaries/mm2, p < 0.05).

Effects of Dkk-3 on in vivo tumor growth and angiogenesis

Murine Dkk-3 was subcloned into a Sleeping Beauty transposon (PT2 EF1A) to generate B16/F10 cell clones stably expressing and secreting mDkk-3 (Fig. 7a, upper and middle panels). In vitro proliferation of these clones was not significantly reduced as compared to controls (data not shown). Control clones (n = 3) and Dkk-3-overexpressing clones (n = 3) were injected into C57/BL6 mice (n = 12) and tumor weight was analyzed after 14 days. Dkk-3 overexpression was retained in the tumors even without neomycin selection (Fig. 7a, lower panel). Tumor weight was markedly increased although statistical significance was not reached (median 0.40 g vs. 0.25 g; p = 0.15, n = 12, Fig. 7c). Moreover, MVD as assessed by CD31 revealed significantly increased neovascularization in Dkk-3-overexpressing B16/F10 melanoma as compared to controls (55 microvessels per mm2vs. 44/mm2, p = 0.03, n = 12, Fig. 7d). To evaluate a potential impact of Dkk-3 overexpression on critical angiogenic mediators, RNA was isolated from each murine tumor and mVEGF, mANG1 and mANG2 gene expression was subsequently subjected to qPCR analysis. Notably, Dkk-3-overexpressing tumors showed a significant reduction in mANG1 gene expression as compared to controls (p < 0.05), whereas mVEGF and mANG2 were not significantly changed (Fig. 7b).

Figure 7.

Analysis of Dkk-3 overexpression in the murine B16F10 melanoma model (a) B16F10 cells with genomic integration of murine Dkk-3 were generated by Sleeping Beauty transposon technology. Expression of murine Dkk-3 was analyzed by semi-quantitative PCR (25 cycles) and Western Blot of in vitro propagated cell clones (Dkk-3# 1-3, controls: co.# 1, 7, 8; upper and middle panels). B16F10 cell clones were injected subcutaneously into C57/BL6 mice and melanomas isolated after 14 days of in vivo growth. Overexpression of mDkk-3 was verified by Western Blot analysis (lower panel). (b) Gene expression analysis of B16F10 melanomas (n = 12): Melanoma overexpressing Dkk-3 (n = 6) were compared with tumors lacking Dkk-3 expression (n = 6) for the relative expression of VEGF, angiopoietin 1 (ANG1) and angiopoietin 2 (ANG2). Beta actin (ACTB) served as internal housekeeping gene for relative quantification. Stars indicate p values < 0.05. (c) Tumor weight of B16F10 melanoma after 14 days of in vivo growth. Three different control and 3 mDkk-3-overexpressing clones were subcutaneously injected into female B57/CL6 mice. In total 12 tumors were analyzed. (d) Microvessel density (MVD) was determined by means of immunofluorescence and a CD31 (PECAM-1) specific mab. MVD was determined per mm2 by counting 10 high-power fields for each tumor.


Dkk-3 has been isolated according to its sequence homology to the cystein-rich dickkopf family of secreted canonical Wnt-inhibitors.8 Its gene expression has been shown to be downregulated in immortalized cells as well as in tumor cells.16 Moreover, Dkk-3 has been identified as a gene upregulated in prostate basal epithelial cell senescence and downregulated in prostate cancer tissue.21 There is evidence that Dkk-3 might influence acinar differentiation and morphogenesis.22

Apart from its postulated function as a “tumor suppressor” in epithelial cell compartments,17, 23Dkk-3 gene expression has been demonstrated in endothelial cells freshly isolated from advanced colorectal carcinomas by SAGE analysis.5 However, there is no data on the functional role of Dkk-3, neither in endothelial cell biology nor in tumor-stroma interaction.

Here, we demonstrate that Dkk-3 is strongly expressed in tumor endothelial cells. A strong Dkk-3 expression was detectable almost exclusively in tumor vessels, whereas vessels from normal tissues exhibited only very few scattered endothelial cells that stained positive for Dkk-3. These findings indicate that Dkk-3 might be a marker for endothelial cell activation during tumor angiogenesis. In fact, we were able to show that Dkk-3 protein expression is detectable in ex vivo propagated ECFC, an endothelial cell type that has been shown to be involved in tumor angiogenesis, but also in various other pathological conditions associated with angiogenesis such as ischemia and wound healing.19, 24–27

Whether Dkk-3 overexpression in blood vessels of the tumor stroma is a reaction to the loss of Dkk-3 expression in the epithelium or a sign of disturbed epithelial homeostas remains an open question. Immortalization and malignant transformation are responsible for dedifferentiation of epithelial cells and might induce loss of expression due to methylation of the Dkk-3 promotor, as shown for all dickkopf genes in colorectal cancer.28 In our IHC analysis of several cancer types we were not able to find any tumor cells expressing Dkk-3.

A large amount of Dkk-3 protein was localized in the Weibel Palade bodies of endothelial cells. These specialized endothelial organelles serve mainly as a storage pool for von Willebrand factor (vWF) and other factors such as P-selectin that are critical to rapid alteration of the vascular bed in response to a broad range of stimuli.29 Accordingly, the release of proteins stored in Weibel Palade bodies is an initial step in the transition from a resting, quiescent endothelial layer to an activated endothelium. This also holds true for angiopoietin-2 (ANG-2), that is overexpressed in tumor vasculature, stored in Weibel Palade bodies and upregulated by inflammatory TNF-α.30 Therefore, localization of Dkk-3 in Weibel Palade bodies suggests a role during the modulation of vascular function by leucocytes and their cytokines.

Release of Dkk-3 protein by endothelial cells after tube formation in matrigel implies that Dkk-3 may be involved in the process of capillary tube formation. Indeed, adenovirally induced overexpression of Dkk-3 resulted in significantly increased tube formation capacity. Conversely, siRNA-mediated downregulation of Dkk-3 significantly inhibited tube formation in matrigel.

In vivo effects of Dkk-3 were analyzed by the use of the syngeneic B16F10 melanoma model in C57/BL6 mice, i.e., genetically modified tumor cells that secrete murine Dkk-3 into the tumor microenvironment. This model has the limitation that not the endothelial cells themselves secrete Dkk-3. Therefore, Dkk-3 could also act in an autocrine way on tumor cells by affecting angiogenic growth factor release or extracellular matrix composition and deposition. Furthermore, increased microvessel density associated with Dkk-3 overexpression might result from paracrine effects on stromal fibroblasts/myofibroblasts, i.e., altered angiopoietin production. Future analyses are needed to discern auto- and paracrine effects of Dkk-3 in this mouse model. Interestingly, overexpression of Dkk-3in vivo caused a significantly higher vascularization of the tumors, although there was merely a trend towards greater tumors, probably due to the small sample of animals. In view of the rapid kinetics of our murine melanoma model we refrained for ethical reasons from allowing the tumors to grow longer. Analysis of the gene expression pattern of distinct angiogenic factors in vivo revealed a pronounced downregulation of ANG-1, whereas ANG-2 and VEGF were not affected. Thus, loss of angiopoietin-1 might be the reason for vessel destabilization, thereby facilitating vascular remodeling and angiogenic sprouting, leading to increased MVD.31 In accordance to the data of Kuphal et al.17 on human melanoma cell lines, mDkk-3 overexpression in B16F10 cells did not affect in vitro proliferation (data not shown). However, we did not observe effects of murine Dkk-3 overexpression on in vitro cell migration like Kuphal et al.,17 but instead a significant increase of microvessel density in vivo. This might indicate that Dkk-3 functions only in a complex microenvironment, thereby interfering in differentiation processes of the tumor stroma.

A possible role of Dkk-3 in vessel morphogenesis/maturation is supported by recent bioinformatic data (accessible through the NCBI server in the Conserved Domain Database; Next to the dickkopf domain, Dkk-3 contains a cystein-rich prokineticin domain also found in endocrine gland-derived vascular endothelial growth factor (EG-VEGF). EG-VEGF has been discovered to be a mitogenic and chemoattractive growth/differentiation factor able to induce fenestration in endothelial cells of steroidogenic organs.32 A crosstalk of Dkk-3 with TGF-beta signaling was recently shown during mesoderm induction in xenopus.33 Thus, Dkk-3 might support TGF-β signaling, stromal reactivation and angiogenic remodeling processes in tumors.11

In summary, our studies show a significant upregulation of Dkk-3 expression in tumor blood vessels as compared to vessels from healthy tissues. The mechanism for upregulating this embryonic differentiation factor in the tumor endothelium remains to be investigated. With regard to the function in endothelial cells, Dkk-3 does not interfere with cell proliferation, but supports tube/capillary formation in vitro and increases microvessel density in vivo. Thus, Dkk-3 might be an interesting candidate for novel therapeutic interference in tumor angiogenesis.


The authors are grateful to Cornelia Heiss and Monika Bauer for their excellent technical support.