Functional implication of BMP4 expression on angiogenesis ... - Nature

Oncogene (2007) 26, 4158–4170

& 2007 Nature Publishing Group All rights reserved 0950-9232/07 $30.00 www.nature.com/onc

ORIGINAL ARTICLE

Functional implication of BMP4 expression on angiogenesis in malignant melanoma T Rothhammer1, F Bataille1, T Spruss2, G Eissner3 and A-K Bosserhoff1 1

Institute of Pathology and Molecular Pathology, University of Regensburg, University Hospital, Regensburg, Germany; Institute of Pharmacy, University of Regensburg, Regensburg, Germany and 3Department of Hematology and Oncology, University of Regensburg, University Hospital, Regensburg, Germany 2

Analyses of malignant melanomas revealed a strong expression of bone morphogenic proteins (BMPs) and their autocrine effect in promoting cell invasion and migration. Here, we report a paracrine effect of BMPs on the vascular network. Both BMP2 and BMP4 induced tube formation as well as the migratory efficiency of microvascular endothelial cells. Melanoma cells with reduced BMP activity attracted less endothelial cells in invasion assays than control cells. Furthermore, reduction of BMPs in melanoma cells had a strong effect on vasculogenic mimicry. Tube formation on matrigel was analysed for melanoma cells as well as in co-cultures of endothelial and melanoma cells. Melanoma cells with reduced BMP activity were not capable of forming cordlike structures by themselves and additionally inhibited tube formation of the endothelial cells. Genes involved in angiogenesis turned out to be strongly downregulated in these cell clones. Tumors derived from cells with impaired BMP activity showed reduced tumor growth or large necrotic areas owing to lack of angiogenesis in in vivo analyses. Oncogene (2007) 26, 4158–4170; doi:10.1038/sj.onc.1210182; published online 18 December 2006 Keywords: malignant melanoma; endothelium; migration; invasion; angiogenesis

Introduction Bone morphogenetic proteins (BMPs) are secreted growth factors that belong to the transforming growth factor beta (TGFb) superfamily. BMPs exert their effects via two different types of serine/threonine kinase receptors, BMP type I (BMPR-IA and BMPR-IB) and type II (BMPR-II) receptors. BMP receptor type I is activated upon ligand binding and subsequently phosphorylates receptor-activated smad proteins (R-smads, Correspondence: Dr A-K Bosserhoff, Institute of Pathology, University of Regensburg, Franz-Josef Strauss Allee 11, D-93053 Regensburg, Germany. E-mail: [email protected] Received 1 June 2006; revised 6 October 2006; accepted 31 October 2006; published online 18 December 2006

Smad-1, Smad-5 and Smad-8). The phosphorylated R-smads then bind to a common mediator Smad (Co-smad, Smad-4), translocate into the nucleus and activate transcription of target genes (ten Dijke et al., 2003; Yamamoto and Oelgeschlager, 2004). The first BMPs were purified from bone extract and showed the ability to induce ectopic bone and cartilage formation in vivo (Wozney et al., 1988; Wang et al., 1990). Today, more than 30 members of the BMP superfamily have been identified in various tissues and species, and it is known that they control a broad range of biological activities in many cell types like epithelial, mesenchymal or neuronal cells (Mehler et al., 1997; Turgeman et al., 2001). BMPs affect cellular processes like proliferation, differentiation, chemotaxis, motility or cell death (Hogan, 1996). Furthermore, they are involved in early embryonic development, and they control the formation and patterning of various tissues, the left–right asymmetry and the morphogenesis of a number of organs (Balemans and Van Hul, 2002; Yamamoto and Oelgeschlager, 2004). Besides their function in embryonic development and normal cellular processes, BMPs are also linked to tumor formation and progression. Several groups have shown that BMP transcripts are expressed in various carcinoma cell lines from gastric, ovarian, prostate, pancreatic and breast origin (Hatakeyama et al., 1993, 1997; Ide et al., 1997; Kleeff et al., 1999; Kiyozuka et al., 2001). Recently, our group identified BMP4 and BMP7 as new important molecules in melanoma development (Rothhammer et al., 2005). In this study, we found that BMPs are important for melanoma migration and invasion. However, we wondered about additional functions of BMPs as we also expected an impact on the tumor microenvironment. Therefore, we concentrated on analysis of tumor angiogenesis to define a role of melanoma-derived BMPs in this process. It is generally known that tumors require a system of blood supply for growth and metastasis. For melanoma cells, two ways of establishing new microcirculation have been described. On the one hand, the tumor can recruit new blood vessels from the existing circulation network (angiogenesis) by releasing angiogenic factors from the tumor cells themselves or the surrounding stroma cells. On the other hand, melanoma cells are able

BMP4 in angiogenesis in melanoma T Rothhammer et al

4159

to de novo generate microcirculatory channels that are composed of extracellular matrix and lined by tumor cells. This process of formation of non-endothelial cell-lined channels has been called ‘vasculogenic mimicry’ (Maniotis et al., 1999; Folberg et al., 2000; Folberg and Maniotis, 2004). Several genes like VE-cadherin (vascular endothelial cadherin), laminin 5 g2 chain, TIE-1 (tyrosine kinase with immunoglobulin and EGF factor homology domains) and EphA2 (epithelial cell receptor protein-tyrosine kinase), known to be important in angiogenesis, were also identified to play a role in vasculogenic mimicry (Hendrix et al., 2003).

Results Studies by Langenfeld and Langenfeld (2004) revealed that BMP2 has a positive effect on angiogenesis. As we recently found expression of BMPs with strongest expression of BMP4 in malignant melanoma (Rothhammer et al., 2005), we asked whether BMP expression has an influence on angiogenesis in the microenvironment of melanomas or on vascular mimicry of melanoma cells themselves. Induction of tube formation and migration of human microvascular endothelial cells (HMEC) by BMP2 and BMP4 First, we performed tube formation assays using HMEC and treated them with recombinant BMP2 or BMP4 (100 ng/ml), respectively. Both BMP2 and BMP4 clearly induced tube formation whereas fewer vessels were visible without BMP treatment (Figure 1a). Quantification of the total tube length (TTL) revealed a strong increase in BMP2- and BMP4- stimulated HMEC (Figure 1b). Furthermore, we analysed the effects of BMPs on endothelial cell migration. In Boyden Chamber models, the ability of BMP2 and BMP4 to induce endothelial cell migration and to act as chemo-attractant was assayed (Figure 2a). In fact, we observed a significant effect on HMEC migration after treatment with BMP2 or BMP4. To analyse direct effects of melanoma cells on endothelial cell migration and invasion, matrigel-coated invasion chambers were used with different melanoma cell clones seeded into the lower compartment and the HMEC onto the matrigel-coated microporous membrane. Melanoma cell clones with reduced BMP activity generated by either overexpression of the BMP inhibitor chordin or stable antisense BMP4 transfection were described previously (Rothhammer et al., 2005). The earlier generated cell clones Mel Im þ CHRD K1, Mel Im þ asBMP4 K7 and Mel Im þ asBMP4 K9 were chosen for further experiments. The migratory and invasive potential of endothelial cells was significantly reduced in the invasion chambers using the chordinexpressing and the asBMP4 melanoma cell clones compared to the control cells, where strong induction of migration and invasion was measurable (Figure 2b).

These results show that both BMP2 and BMP4 can act as a chemo-attractant and are involved in the induction of tube-like structures of endothelial cells. In conclusion, the secretion of BMPs could be a mechanism by which melanoma cells enhance the migratory and the vessel forming capacity of endothelial cells. Tube formation capacity of melanoma cells We next asked whether vascular mimicry of the melanoma cells is also influenced by or even dependent on BMP expression. The melanoma cell line Mel Im and the mock control were able to efficiently form tubelike structures (Figure 3a). In contrast, cell clones with reduced BMP4 expression or chordin overexpression (Figure 3b) did not form tubes. It is further known that melanoma cells are able to interact with endothelial cells (Ding et al., 2005; Saalbach et al., 2005). So, we co-cultured melanoma and endothelial cells in matrigel and, indeed, co-cultures of Mel Im cells with HMEC resulted in tube formation (Figure 3c). In this assay, HMEC were labeled with fluorescent dye to prove the cooperative manner of vessel formation. Again, the cell clones with reduced BMP4 expression or chordin overexpression (Figure 3d) were not capable of forming vessels. Moreover, they inhibited vessel formation of the endothelial cells. Studies by Peng et al. (2005) and Vogt et al. (2006) showed that either BMP2 or BMP4, respectively, induced vascular endothelial growth factor (VEGF)mediated angiogenesis during bone healing or in retinal pigment epithelial cells (Peng et al., 2005; Vogt et al., 2006). According to these findings, we analysed whether recombinant VEGF could restore the tube-forming capacity of endothelial cells in co-culture with antisense BMP4 or chordin-overexpressing cell clones. The tube formation capacity of endothelial cells and melanoma cells was assayed supplementing 100 ng/ml recombinant VEGF. VEGF-treated cells were able to form chord-like structures whereas co-cultures of HMEC and antisense BMP4 or chordin-overexpressing melanoma cells without VEGF did not form tube-like structures as already shown in Figure 3d (Figure 3e). This suggests that the action of BMPs is mediated through VEGF. Analysis of key molecules involved in melanoma vasculogenic mimicry Hendrix et al. (2003) have shown that epithelial cell kinase (EphA2), VE-cadherin and laminin 5 g2 chain play an important role in the formation of vasculogeniclike networks by melanoma cells (Hendrix et al., 2003). Furthermore, it was demonstrated by Valdimarsdottir et al. (2002) that ectopic Id-1 expression was sufficient to induce tube formation of bovine aortic endothelial cells. In addition to these molecules, several groups showed that the endothelial cell-specific tyrosine kinase receptor (TEK/Tie-2) is a key factor for the subsequent remodelling and stabilization of the developing vasculature and seems to be involved in tumor angiogenesis (Cavallaro and Christofori, 2000; Ward and Dumont, 2002; Peters et al., 2004). Oncogene


4160

a

HMEC control

HMEC control 5x

5x

HMEC + BMP2 (100 ng/ml)

HMEC + BMP4 (100 ng/ml) 5x

Microvascular density [%]

b

5x

p = 0.0004 ***

300 p < 0.0001 ***

250 200 150 100 50 0 HMEC control

HMEC + BMP2 HMEC + BMP4 (100 ng/ml) (100 ng/ml)

Figure 1 Induction of tube formation of HMEC by BMP2 and BMP4. (a) Human microvascular endothelial cells (5 104) were seeded onto matrigel and tube formation was assayed after 16 h by phase contrast microscopy. Cells were treated with MCDB-131 culture medium supplemented with 100 ng/ml BMP2 or BMP4, respectively, control cells received MCDB-131 medium alone. Cells that were treated with BMP2 or BMP4 clearly showed tube formation on matrigel, whereas the control cells poorly built tube-like structures. (b) TTL was assessed by measuring the TTL per area using the ImageJ software. The TTL of the control cells without BMP treatment was set as 100%. BMP2 and BMP4- treated cells displayed a significantly higher TTL, which reached 225% for BMP2 or 270% for BMP4- treated cells over the control. TTL of eight wells per treatment was analysed, bars display mean þ s.d., statistical analysis was performed using Student’s unpaired t-test.

Based on these findings, the expression levels of epithelial cell kinase (EphA2), VE-cadherin, laminin 5 g2 chain, Id-1 and the endothelial cell-specific tyrosine kinase receptor (TEK/Tie-2) were analysed by quantitative real-time reverse transcriptase-PCR (RT–PCR) in the parental cell line Mel Im, the mock control, the chordin-expressing and the antisense BMP4 melanoma cell clones, respectively. A marked downregulation of EphA2 expression was detectable in chordin-expressing as well as in antisense BMP4 cell clones compared to the Oncogene

control cells (Figure 4a). Also VE-cadherin expression was reduced in the cell clones with modified BMP levels (Figure 4b). In contrast, laminin 5 g2 chain expression levels were not altered in any of the cell clones (data not shown). Id-1 was hardly measurable in the chordinexpressing cell clone and only 30–40% of the expression levels of the control cells (set as 100%) were detectable in the antisense BMP4 cell clones (Figure 4c). The results for TEK/Tie-2 expression were consistent with the other genes analysed; a strong downregulation was


4161

a

p < 0.0001 ***

250

p = 0.0019 **

Migration [%]

200

150

100

50

0 DMEM

DMEM + BMP2 (100 ng/ml)

DMEM + BMP4 (100 ng/ml)

Invasion [%]

b 120 110 100 90 80 70 60 50 40 30 20 10 0

p = 0.0024 ** p = 0.0004 *** p < 0.0001 *** Mel Im

Mel Im + pCMX1

Mel Im + Mel Im + Mel Im + CHRD K1 asBMP4 K7 asBMP4 K9

Figure 2 Induction of migration of HMEC by BMP2 and BMP4. (a) Analysis of the ability of BMP2 and BMP4 to either induce endothelial cell migration or act as chemo-attractant was assayed in the Boyden Chamber model. Both BMP2 and BMP4 (100 ng/ml) strongly induced the migratory potential of endothelial cells compared to non-treated control cells. (b) Analysis of the capacity of melanoma cell clones to induce migration and invasion of HMEC was assayed using matrigel-coated transwell systems. Mock controls, antisense BMP4 and chordin-overexpressing cell clones were seeded into the lower compartment to attract migration and invasion of endothelial cells from the upper compartment. Migration efficiency of HMEC was strongly reduced in cell clones with lower BMP activity compared to controls. All experiments were performed in triplicate and repeated three times with similar results, bars display mean þ s.d., statistical analysis was performed using Student’s unpaired t-test.

found in the chordin-expressing and the antisense BMP4 cell clones compared to the controls (Figure 4d). Analysis of BMP target genes It is known that Id-1 is a target gene of BMP/Smad signaling (Miyazawa et al., 2002). Therefore, we treated the parental melanoma cell line Mel Im with recombinant BMP4 (100 ng/ml) and analysed Id-1 gene expression at different time points by quantitative real-time RT–PCR. We were able to detect a strong upregulation of Id-1 after 1 h of BMP4 treatment, reaching a maximum after 3 h (Figure 5). Expressions of EphA2,

VE-cadherin and TEK were also analysed after BMP4 treatment, however with no significant changes in gene expression levels after 6 h (data not shown). These data lead to the conclusion that EphA2, VE-cadherin and TEK are indirect target genes of BMP4 signaling. Tumor growth and angiogenesis of melanoma cell clones in nude mice For in vivo analyses of the melanoma cell clones, the mock control, a chordin-overexpressing cell clone and two antisense BMP4 cell clones were injected subcutaneously into nude mice. Between day 12 and day 20 at least nine out of 10 animals in each group developed tumors, but mice from the Mel Im þ CHRD K1 group showed later tumor onset (d20) than the other groups (d13) (Table 1). Furthermore, tumors derived from chordin-overexpressing melanoma cell clones were significantly smaller than tumors from the Mel Im þ pCMX1 control group at day 35. However, there was no detectable difference in tumor size between the mock control (Mel Im þ pCMX1) and tumors that arose from antisense BMP4 melanoma cell clones (Table 1). Interestingly, detailed histological analyses of similar size tumors revealed larger necrotic areas in tumors derived from antisense BMP4 cell clones compared to the control tumors (Figure 6a). Immunohistochemical staining against the endothelial cell marker CD31 (PECAM-1) revealed a higher number of vessels in the Mel Im þ pCMX1 control tumors compared to the antisense BMP4 tumors (Figure 6b). Analysis of the same tumors for tumor cell proliferation did not reveal any differences between the antisense BMP4 and the control tumors (data not shown). Analysis for apoptotic cells using the terminal deoxynucleotidyl transferase dUTP nick-end labeling (TUNEL) system showed hardly any apoptotic cells in Mel Im þ pCMX1-derived tumors, whereas the Mel Im þ asBMP4 tumors displayed large apoptotic areas (Figure 6c). Discussion Our group recently showed that BMP4 and BMP7 are strongly expressed in primary and metastatic malignant melanomas and their involvement in the metastatic process. BMPs seem to be inevitable factors for enhanced migration and invasion of melanoma cells (Rothhammer et al., 2005). In this study, we now focused on the role of BMPs on the tumor microenvironment, especially the vascular network. BMP4 has a positive effect on tube formation in vitro as well as on endothelial cell migration. It, therefore, seems to act as an angiogenic factor as well as a chemoattractant. Additionally, we analysed the direct influence of melanoma cells on endothelial cell migration and invasion. The melanoma cell clones with reduced BMP activity attracted much less endothelial cells, which leads to the conclusion that BMPs are important factors to establish the vascular network to supply the growing tumor. Oncogene


4162

a

Mel Im

Mel Im + pCMX1 5x

b

Mel Im + CHRD K1

5x

Mel Im + asBMP4 K7

Mel Im + asBMP4 K9

5x

c

5x

HMEC* + Mel Im

HMEC* + Mel Im + pCMX1 10x

d

HMEC* + Mel Im + CHRD K1

5x

10x

HMEC* + Mel Im + asBMP4 K7 10x

HMEC* + Mel Im + asBMP4 K9 10x

10x

Figure 3 Continued

Several groups have shown that BMP family members have an impact on the development of the vascular system (Deckers et al., 2002; Valdimarsdottir et al., 2002; Langenfeld and Langenfeld, 2004; Raida et al., Oncogene

2005). Studies by Raida et al. (2005) analysed the impact of BMP2 on angiogenesis of a breast cancer xenograft model. MCF-7 cells were stably transfected with a BMP2 expression construct or an empty vector control


4163

e

HMEC* + Mel Im + CHRD K1

HMEC* + Mel Im + asBMP4 K7

HMEC* + Mel Im + asBMP4 K9

no VEGF

+ VEGF 100 ng/ml

Figure 3 Tube formation capacity of melanoma cells. (a) 7 104 cells of the melanoma cell line Mel Im and the transfection control Mel Im þ pCMX1 were seeded onto matrigel and tube formation was assayed after 16 h by phase contrast microscopy. Both Mel Im cells and the mock control show tube formation on the matrigel matrix. (b) 7 104 cells of the chordin-expressing melanoma cell clones (Mel Im þ CHRD K1) and of two of the antisense BMP4 cell clones (Mel Im þ asBMP4 K7, Mel Im þ asBMP4 K9) were seeded onto matrigel and tube formation was assayed as described in (a). All three cell clones with reduced BMP expression were not able to form tubes on matrigel compared to the parental cell line Mel Im and the mock control. (c) Human microvascular endothelial cells (4 104), which were labeled with a fluorescent dye and 4 104 of the melanoma cell line Mel Im or the transfection control Mel Im þ pCMX1, respectively, were seeded onto matrigel and tube formation was assayed after 16 h by phase contrast microscopy. Both Mel Im cells and the mock control showed cooperative manner of vessel formation with HMEC. (d) Human microvascular endothelial cells (4 104), which were labeled with a fluorescent dye and 4 104 of the chordin-expressing or the asBMP4 melanoma cell clones, respectively, were seeded onto matrigel and tube formation was assayed as described in (c). The three cell clones were not able to form tubes on matrigel and even prevented endothelial cells from forming tube like structures. (e) Human microvascular endothelial cells (4 104), which were labeled with a fluorescent dye and 4 104 of the chordin-expressing or the asBMP4 melanoma cell clones, respectively, were seeded onto matrigel and were treated either without or with 100 ng/ml recombinant VEGF. Tube formation was assayed as described in (c). Untreated endothelial and melanoma cell co-cultures were not able to form chord-like structures whereas VEGF treated cells displayed tube forming capacity. All experiments were performed in duplicate and repeated three times with consistent results.

and subcutaneously injected into nude mice. All mice bearing the MCF-7/BMP2-transfected cells developed tumors showing vascularization from the tumor stroma, whereas the control group did not develop any tumors within 6 weeks (Raida et al., 2005). Moreover, Langenfeld and Langenfeld (2004) demonstrated a proangiogenic effect of BMP2 in in vivo studies of a lung cancer xenograft model. Co-injection of lung carcinoma cells with recombinant BMP2 into nude mice resulted in tumors with three times more and larger blood vessels compared to the controls (Langenfeld and Langenfeld, 2004). Valdimarsdottir et al. (2002) described an impact of BMP6 on tube formation of endothelial cells in vitro via BMP/Smad signaling. Studies by Vogt et al. (2006) showed a role of BMP4 in the regulation of ocular angiogenesis by stimulating VEGF release from retinal pigment epithelial cells. Earlier investigations focused on the role of BMPs during bone development and repair. Deckers et al. (2002) demonstrated that BMPs regulate the expression of VEGF A and might thus stimulate angiogenesis

during fracture healing. Peng et al. (2005) confirmed these findings; they showed that endogenous VEGF activity is one of the main mediators of angiogenesis during BMP2 -induced bone formation. Moreover, a synergistic effect of BMP4 and VEGF on bone formation and bone healing by enhancing angiogenesis and cartilage resorption was demonstrated by the same group (Peng et al., 2002). The ability of BMPs to promote angiogenesis in tumor development as well as in normal cellular processes like bone healing was demonstrated in several different studies. Here, we analysed the effects of BMPs on angiogenesis in malignant melanoma and found that, in line with other studies, BMPs are important factors to establish the vascular system of the tumor. Angiogenesis induced by endothelial cell proliferation is not the only way in which melanomas ensure their blood and nutrient supply. In addition, they can even contribute to the microvasculature by themselves, a process called vasculogenic mimicry. Several studies revealed that networks built by melanoma cells can Oncogene


4164

a

EphA2 1.4 1.2

1.0 x-fold to Mel Im + pCMX1

x-fold to Mel Im + pCMX1

VE-Cad

b

1.0 0.8 0.6 0.4 0.2 0.0

0.8 0.6 0.4 0.2 0.0

Mel Im + pCMX1

Mel Im + Mel Im + Mel Im + Mel Im + pCMX3 CHRD K1 asBMP4 K7 asBMP4 K9

Id-1

c 1.2

Mel Im + pCMX1

Mel Im + Mel Im + Mel Im + Mel Im + pCMX3 CHRD K1 asBMP4 K7 asBMP4 K9

TEK

d

1.0

1.6 x-fold to Mel Im + pCMX1

x-fold to Mel Im + pCMX1

1.8

0.8 0.6 0.4 0.2

1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0

0.0 Mel Im + pCMX1

Mel Im + pCMX3


Mel Im + pCMX1

Mel Im + pCMX3


Figure 4 Key molecules involved in melanoma vasculogenic mimicry. (a) Quantitative real-time RT–PCR analysis of epithelial cell kinase (EphA2) expression in the parental cell line Mel Im, the mock transfected control, the chordin-expressing and the antisense BMP4 melanoma cell clones. (b) Quantitative real-time RT–PCR analysis of VE-cadherin. (c) Quantitative real-time RT–PCR analysis of Id-1. (d) Quantitative real-time RT–PCR analysis of endothelial cell-specific tyrosine kinase receptor (TEK/Tie-2). All the melanoma cell clones with reduced BMP expression showed a strong reduction of EphA2, VE-cadherin, Id-1 and TEK expression compared to the control cells, Mel Im and the mock transfected control. All experiments were repeated five times with similar results, bars display mean þ s.d.

distribute plasma and red blood cells in vivo and in vitro, enabling the tumor to become, at least in part, independent of angiogenesis (Folberg et al., 2000; Hendrix et al., 2003; Folberg and Maniotis, 2004). Vasculogenic mimicry, however, is controversially discussed. McDonald and Foss (2000) raised several questions as to whether tumor cells really have the ability to form vessels. Vessel- like structures built by tumor cells should be able to transport erythrocytes and should be connected to ‘normal’ endothelial cell-lined blood vessels (McDonald and Foss, 2000). Maniotis et al. (1999) identified vascular channels by the presence of erythrocytes and lumen-like structures. However, McDonald et al. (2000) raised the question if extravasated erythrocytes in the extracellular matrix can be misinterpreted as vascular channels. Moreover, no Oncogene

direct evidence shows that junctions between the lumen of endothelial cell vessels and tumor cell-lined tube structures exist. So, there is still no absolute proof that tumor cells and not endothelial cells form the vascular network in tumors. In our study, we used a highly aggressive metastatic melanoma cell line that efficiently formed tube-like networks on matrigel in vitro. In contrast, melanoma cells with reduced BMP activity were not able to form tube-like structures at all. Additionally, we showed that melanoma and endothelial cells form chord-like networks in a cooperative manner on matrigel. Co-cultures of endothelial cells and different melanoma cell clones revealed the formation of tubular structures only in those cell clones with unchanged BMP levels. The antisense BMP4 and chordin-overexpressing cell clones


4165

were not able to form networks and even prevented the tube formation of endothelial cells. However, VEGF treatment restored the tube-forming capacity of endothelial cells in co-culture with the antisense BMP4 or chordin-overexpressing melanoma cells. Our findings are consistent with the analyses of Vogt et al. (2006) that showed a role of BMP4 in ocular angiogenesis by stimulating VEGF release from retinal pigment epithelial cells. Altogether, these data provide evidence that BMPs are not only important to attract endothelial cells to the tumor site and enforce angiogenesis but also to support vasculogenic mimicry in malignant melanoma. Microarray analyses of highly versus poorly invasive melanoma cell lines revealed different genes to be upregulated in the highly invasive melanoma cells. Among those genes EphA2 and VE-cadherin are most important for vasculogenic mimicry (Hendrix et al., 2003). Both EphA2 and VE-cadherin, are co-expressed in vasculogenic mimicry and transient downregulation of either VE-cadherin or EphA2 resulted in the loss of the network-forming capacity (Seftor et al., 2002). In our analyses, we found EphA2 and VE-cadherin to be downregulated in the antisense BMP4 as well as in the chordin-overexpressing cell clones, suggesting that downregulation of BMP activity leads to a reduced expression of genes involved in vasculogenic mimicry,

Id-1

20 18 x-fold to Mel Im control

16 14 12 10 8 6 4 2 0 Mel Im control

Mel Im + BMP4, 1h

Mel Im + BMP4, 3h

Mel Im + BMP4, 6h

Figure 5 Analysis of BMP target genes. Analysis of Id-1 by quantitative real-time RT–PCR analysis in the melanoma cell line Mel Im after BMP4 treatment (100 ng/ml). Id-1 expression strongly increased in melanoma cells treated with BMP4 after 1 h and reached a maximum at 3 h. Experiments were repeated five times with similar results, bars display mean þ s.d.

such as EphA2 and VE-cadherin. As a consequence, melanoma cell clones do not form tubular networks anymore. Id-1 and the endothelial cell specific tyrosine kinase receptor (TEK/Tie-2) both seem to be crucial for embryonic and tumor angiogenesis (Jones and Dumont, 2000; Benezra, 2001; Benezra et al., 2001; Ward and Dumont, 2002; Alani et al., 2004). Benezra et al. showed that tumor xenografts failed to grow and metastasize in Id1 þ / and Id3/ knockout mice. Furthermore, blood vessels were not able to branch and sprout normally in these animals. Id1 þ / and Id3/ animals showed severe angiogenic defects (Benezra et al., 2001). Another study by Pache et al., (2006) showed that Id-1 expression was induced by BMP2 in mouse podocytes, followed by an increase of reactive oxygen species (ROS). ROS were shown to play an important role in VEGF-induced angiogenesis (Ushio-Fukai et al., 2002). These data prompted us to analyze the melanoma cell clones with reduced BMP activity for Id-1 expression. A strong downregulation of Id-1 was found to be comparable to the downregulation seen for EphA2 and VE-cadherin. Furthermore, Lin et al. (1998) showed that expression of a soluble form of the TEK receptor resulted in a reduced growth rate of murine mammary tumors and melanoma in mice. Analyses of the melanoma cell clones for TEK expression revealed a diminished expression of TEK in antisense BMP4 and chordin-overexpressing cell clones. In summary, these data clearly indicate that BMPs have an impact on genes involved in vasculogenic mimicry like EphA2 and VE-cadherin as well as genes known to be relevant during neo-angiogenic processes like Id-1 and TEK. The fact that these genes are downregulated in melanoma cells with reduced BMP activity prompted us to investigate whether BMPs have a direct influence on EphA2, VE-cadherin, Id-1 or TEK expression. It is known that Id-1 is a direct target gene of the BMP/smad signaling cascade (ten Dijke et al., 2003). Consequently, we could show that stimulation with BMP4 leads to an augmentation in Id-1 gene expression. Stimulation of antisense BMP4 or chordin-overexpressing cells with exogenous BMPs, however, did not lead to re-expression of EphA2, VE-cadherin or TEK. We suggest an indirect effect that might function via Id-1 or the BMP/MAPK signaling cascade. Interestingly, Straume and Akslen (2005) have shown that the expression of EphA2 is associated with strong Id-1 expression in malignant melanomas. However, the

Table 1 Table displays mean tumor size+s.e.m., 10 animals per group

Mel Mel Mel Mel

Im+pCMX1 Im+CHRD K1 Im+asBMP4 K7 Im+asBMP4 K9

Day 0

Day 5

Day 12

Day 20

Day 28

Day 35

0.0+0.00 0.0+0.00 0.0+0.00 0.0+0.00

0.0+0.00 0.0+0.00 0.0+0.00 0.0+0.00

0.9+0.18 0.0+0.00 0.6+0.40 0.9+0.41

12.3+2.49 0.3+0.15 11.3+3.60 17.8+6.33

41.1+6.58 6.7+6.7 40.0+10.82 51.8+18.90

70.3+19.61 13.2+6.60* 71.1+18.62 58.1+21.92

*P ¼ 0.0103. Statistical analysis was performed using Student’s unpaired t-test. Oncogene


4166

a

b

Mel Im + pCMX1

Mel Im + pCMX1 20x

p < 0.0001 ***

45 40

p < 0.0001 ***

35 30 25 20

Mel Im + asBMP4 K7

15

Mel Im + asBMP4 K9 20x

10

20x

c

Mel lm + asBMP4 K9

0

Mel lm + asBMP4 K7

5 Mel lm + pCMX1

necrotic area per total turn or area [%]

20x

Mel Im + pCMX1

Mel Im + pCMX1 20x

Mel Im + asBMP4 K7

20x

Mel Im + asBMP4 K9 20x

20x

Figure 6 Tumor growth of melanoma cell clones implanted into nude mice. (a) H&E-stained tumors were screened for necrotic areas within the tumor tissue. Area of necrotic zones and total tumor area was measured using the ImageJ software. Tumors derived from Mel Im þ asBMP4 displayed significantly larger areas of necrosis than the control tumors. For each group, three tumors were sectioned into 4 mm slices, stained with H&E and four sections were evaluated. Bars display mean þ s.d., statistical analysis was performed using Student’s unpaired t-test. (b) Tumors derived from nude mice were stained for CD31 (PECAM-1) expression. Strong expression of CD31 was detectable in Mel Im þ pCMX1 controls, whereas very weak staining was visible in tumors derived from antisense BMP4transfected melanoma cells. (c) Tumors derived from nude mice were analysed for apoptotic cells using a fluorometric TUNEL assay system. Cells showing DNA fragmentation are labeled with green fluorescent dye, nuclei are stained with DAPI (blue). Hardly any apoptotic cells were detectable in the control tumors (Mel Im þ pCMX1), whereas the Mel Im þ asBMP4 tumors displayed large areas of apoptotic cells.

regulation of VE-cadherin and TEK expression still needs further investigation. Moreover, we think that on the one hand long time downregulation of BMPs might lead to extensive Oncogene

changes within the cell, which cannot be restored by short-time exogenous BMP application. On the other hand, the localization of BMPs might play a role as well. Cells can target secretion of growth factors, whereas


4167

exogenously applied BMPs are in the cell culture supernatant and not at specific sites of the cell. In vivo analyses of the melanoma cell clones in athymic nude mice revealed that the control cells as well as the chordin-overexpressing and the antisense BMP4 melanoma cells were able to induce tumors. Tumors derived from chordin-overexpressing cell clones but not from the antisense BMP4 cell clones were significantly smaller than tumors in the control group. This might be due to the fact that chordin inhibits not only BMP4 but also BMP2 and BMP7 (Balemans and Van Hul, 2002; Botchkarev and Sharov, 2004). All these BMPs can be supplied by the tumor surrounding tissues and can be neutralized by chordin, which is secreted from the cell clones. The antisense BMP4 cell clones however only show a downregulation of endogenous BMP4 levels and can still be influenced by BMPs from the surrounding tumor stroma. It seems that downregulation of only one family member of the BMP family is not sufficient to negatively influence tumor growth in vivo, but targeting several members by using antagonists might be a useful tool. Further analyses of the tumors derived from Mel Im þ asBMP4, however, showed immunohistochemical differences to the control group. The control tumors showed more CD31 positive cells than the antisense BMP4 tumor group, suggesting that more endothelial structures are present. Additional histological analysis of all tumors revealed large necrotic areas in tumors derived from antisense BMP4 cell clones. Moreover, many apoptotic cells were detectable around the necrotic areas, whereas hardly any apoptotic cells were visible in the control tumors. This might be due to reduced endothelial structures and consequently poor blood supply. In summary, our studies show that BMPs are essential factors that enhance migration and invasion of melanoma cells and also have a strong and important influence on the tumor microenvironment. BMPs act as a chemo-attractant for endothelial cells migrating to the tumor site and are necessary factors to establish a vascular network for tumor nutrition. The establishment of tube-like structures is achieved by either directly or indirectly regulating key molecules involved in neoangiogenesis and vasculogenic mimicry.

Materials and methods Cell lines and tissue culture The melanoma cell line Mel Im was described in detail previously (Rothhammer et al., 2005). It was derived from

Gene b-actin EphA2 Id-1 VE-cadherin TEK

a metastasis of malignant melanoma. Chordin-expressing cell clones were established by stable transfection of Mel Im melanoma cells with a sense chordin expression construct (gift from Theresa E Gratsch and K Sue O’Shea) (Rothhammer et al., 2005). For the generation of antisense BMP4 cell clones, a 670 bp fragment of the BMP4 coding region was cloned in antisense orientation into the pCMX-PL1 vector and Mel Im cells were stably transfected. Plasmids were cotransfected with pcDNA3 (Invitrogen, NV Leek, The Netherlands) containing the selectable marker for neomycin resistance. Transfections were performed using Lipofectamine Plus (Invitrogen). Cells were grown at 371C/5% CO2 in Dulbecco’s modified Eagle medium (DMEM); (Gibco, Eggenstein, Germany) supplemented with penicillin (100 U/ml), streptomycine (10 mg/ml) (both Sigma, Deisenhofen, Germany) and 10% fetal calf serum (FCS) (Gibco). The human microvascular endothelial cell line CDC/ EU-HMEC-1 (HMEC) was kindly provided by the Centers for Disease Control and Prevention (Atlanta, GA, USA) and has been described previously (Ades et al., 1992). HMEC were cultured in MCDB131 (Sigma) medium, supplemented with 15% fetal calf serum, 1 mg/ml hydrocortisone (Sigma), 10 ng/ml epidermal growth factor (Collaborative Biochemical Products, Bedford, MA, USA), penicillin (100 U/ml) and streptomycine (10 mg/ml). All cells were split 1:5 at confluence. HMEC were labeled with 5-(and 6)-carboxyfluorescein diacetate, succinimidyl ester (CFSE, Invitrogen Molecular Probes Inc., Eugene, USA). Cells were harvested, counted and washed twice with phosphate-buffered saline (PBS) to remove traces of FCS. 5 106 cells were incubated in a 2 mM solution of CFSE in PBS for 4 min at room temperature. Staining was terminated by washing the cells twice with PBS þ 10% FCS. Cells were seeded onto Matrigel in co-culture with melanoma cell clones to assay tube formation efficiency.

Isolation and quantification of RNA Total RNA was isolated using the RNeasy Mini Kit following the manufacturer’s manual (Qiagen, Hilden, Germany). 1/10 of the RNA was used for reverse transcription as described previously (Rothhammer et al., 2005). Reverse transcription was performed using Superscript II (Invitrogen) according to the manufacturer’s manual. For quantitative real-time RT–PCR 1 ml cDNA preparation, 2.4 ml 25 mM MgCl2, 0.5 mM of forward and reverse primer and 2 ml of SybrGreen LightCycler Mix (Roche Diagnostics, Mannheim, Germany) in a total of 20 ml were applied. The following PCR program was performed: 60 s 951C, 201C/s temperature transition rate up to 951C for 15 s, 10 s 581C, 22 s 721C, 10 s 821C acquisition mode single, repeated 40 times (amplification). MgCl2 concentration and annealing temperatures were optimized for each primer set.

Forward primer

Reverse primer

50 -CTACGTCGCCCTGGACTTCGA 50 -CACCCGTATGGCAAAGGGTGGGAC 50 -TGTTACTCACGCCTCAAGGAG 50 -CAAGTATGGGCAGTTTGACC 50 -TTAGCTTAGGAGGCACACCCTACTGCGG

50 -GATGGAGCCGCCGATCCACAC 50 -CCGACTCGGCATAGTAGAGGTGAA 50 -TTCAGCGACACAAGATGCG 50 -AGGAAGATGAGCAGGGTGAT 50 -GCAGTTGTCAAGGGTCTCCCATGCCAGT Oncogene


4168 PCR reactions were evaluated by melting curve analysis and checking the PCR products on 1.8% agarose gels. Each quantitative PCR was performed in triplicate for two sets of RNA preparations. Migration assays For migration assays Boyden chambers containing polycarbonate filters with 8 mm pore size (Costar, Bodenheim, Germany) were used, as described previously (Rothhammer et al., 2005). Filters were coated with gelatine (5 mg/l) to improve cell attachment. The lower compartment was filled with DMEM supplemented with 100 ng/ml BMP2 or BMP4. Cells were harvested, resuspended in DMEM without FCS at a density of 2 105 cells/ml and placed in the upper compartment. After incubation at 371C for 4 h, the filters were collected and cells adhering to the lower surface were fixed, stained and counted. Experiments were carried out in triplicate and were repeated three times with consistent results. Invasion assays Invasion assays were performed using the BD BioCoatMatrigel Cell Environments, invasion chambers (BD Biosciences, Discovery Labware, Bedford, MA, USA). 6 104 cells of each melanoma cell clone were seeded into 24-well plates (BD Biosciences). After 24 h invasion chambers were inserted into each well according to the manufacturer’s protocol and 5 104 HMEC where seeded into the upper chamber to measure their invasive capacity. After 24 h, the matrigel layer was removed, and cells adhering to the lower surface were fixed, stained and counted. Experiments were carried out in triplicate and were repeated three times with consistent results. Matrigel tube formation assays Growth factor reduced matrigel (200 ml) (BD Biosciences) was added to eight-chamber polystyrene vessel tissue culturetreated glass slides (BD Falcon, Discovery Labware, Bedford, MA, USA) and allowed to gelatinize for 20 min at 371C. Human microvascular endothelial cells (5 104) were seeded into each well of the culture slides in MCDB-131 medium. Cells were treated with 100 ng/ml BMP2 or BMP4. Tube formation was assessed by phase contrast microscopy after 16 h and recorded with a digital camera (AxioCam, MR Grab, Carl Zeiss, Mu¨nchen-Hallbergmoos, Germany). Total tube length (TTL) was analysed using the ImageJ software (Image processing and analysis in Java) freely available from the National Institutes of Health at http:// rsb.info.nih.gov/ij. Experiments were repeated three times; for each experiment five pictures were taken and total vessel length was measured in mm from each picture area. To assay vasculogenic mimicry, either 7 104 melanoma cells alone or 4 104 melanoma cells together with 4 104 CFSE labeled HMEC, respectively, were seeded onto matrigelcoated culture slides. To analyze the effects of VEGF, 4 104 melanoma cells together with 4 104 CFSE labeled HMEC, respectively, were seeded onto matrigel-coated culture slides and either cultured in DMEM or in DMEM supplemented with 100 ng/ml VEGF. Tube formation was assessed by phase contrast as well as fluorescence microscopy and an image overlay was generated using Adobe Photoshop 6.0 software.

mouse (cf. www.felasa.org). Drinking water and laboratory animal chow (V1244; ssniff Spezialdiaeten, Soest, Germany) were provided ad libitum. Water and food as well as caging and bedding materials (H1505-01; ssniff Spezialdiaeten, Soest, Germany) were autoclaved before use. The physical data of the animal room were temperature 22721C, relative humidity 5575%, 15 air changes per hour, and 12 h/12 h light/dark cycles. Tumor cell inoculation and measurement of tumor growth The melanoma cell clones Mel Im þ pcMX1, Mel Im þ CHRD K1, Mel Im þ asBMP4 K7 and Mel Im þ asBMP4 K9 were harvested after incubation with PBS containing 0.05% trypsin and 0.04% ethylenediaminetetraacetic acid (Sigma). Tumor cells were washed twice with serum- free DMEM at room temperature and resuspended in DMEM at a concentration of 1 107 cells/ml. For each of the five tumor cell lines, a group of 10 NMRI (nu/ nu) mice with a mean body weight of 32 g was formed. All mice were injected subcutaneously with a cell suspension of 0.1 ml containing 1 106 cells of a single line. Tumor growth kinetics was recorded by weekly measurement of tumor diameters at the inoculation site (region of the thoracic mammary fat pad) with an electronic caliper. Tumor areas were calculated as the product of two perpendicular diameters, one measured across the greatest width of the tumor. For ethical reasons the mice were killed after the tumors underwent ulceration or reached a maximum size of approximately 300 mm2, respectively. At day 35 all mice were killed, and the tumors were taken out and fixed for histopathological evaluation. Immunohistochemistry Formalin-fixed, paraffin-embedded mouse tumor tissue samples were sectioned at 4 mm thickness. Tissue sections were stained against CD31 (PECAM-1) (Dako Cytomation, Hamburg, Germany) or KI67 (clone MIB-1, Dako Cytomation), respectively, on a Dako Autostainer (Dako Cytomation). For histological analysis, sections were stained with hematoxylin and eosin (H&E staining). For each group three tumors were sectioned into 4 mm slices, stained with H&E and four sections were evaluated. TUNEL assay Formalin-fixed, paraffin-embedded mouse tumor tissue samples were sectioned at 4 mm thickness and stained for apoptotic cells using the DeadEnd fluorometric TUNEL system (Promega Corporation, Madison, USA) according to the manufacturer’s protocol. DNA strand breaks which occur in apoptotic cells were labeled with fluorescein-12-dUTP, and intact nuclei were stained with 4,6-diamidino-2-phenylindole (DAPI) by using Vectashield mounting medium with DAPI (Vector Laboratories, Burlingame, USA). Statistical analysis Results are expressed as mean þ s.d. (range) or percent. Comparison between groups was made using the Student’s unpaired t-test. A P-value o0.05 was considered to be statistically significant. All calculations were made by using the GraphPad Prism Software (GraphPad Software, Inc., San Diego, USA). Acknowledgements

Animals and housing conditions Naval Medical Research Institute (NMRI) (nu/nu) mice from the universities breeding stock were kept in type III filter top cages (5 mice per cage; Tecniplast, Hohenpeissenberg, Germany) under specified pathogen free (SPF) conditions according to the FELASA recommendations for the species Oncogene

We thank Sibylla Lodermeyer, Silvia Haffner and Verena Burger for excellent technical assistance; Dr J Johnson (University of Munich, Germany) for providing the melanoma cell lines; The Centers for Disease Control and Prevention (Atlanta, Georgia, USA) for providing the human


4169 microvascular endothelial cell line CDC/EU.-HMEC-1; Theresa E Gratsch and K Sue O’Shea (University of Michigan Medical School) for providing the chordin expression

vector. This work was supported by grants from the Deutsche Krebshilfe to AKB and the Deutsche Forschungsgemeinschaft to GE.

References Ades EW, Candal FJ, Swerlick RA, George VG, Summers S, Bosse DC et al. (1992). HMEC-1: establishment of an immortalized human microvascular endothelial cell line. J Invest Dermatol 99: 683–690. Alani RM, Silverthorn CF, Orosz K. (2004). Tumor angiogenesis in mice and men. Cancer Biol Ther 3: 498–500. Balemans W, Van Hul W. (2002). Extracellular regulation of BMP signaling in vertebrates: a cocktail of modulators. Dev Biol 250: 231–250. Benezra R, Rafii S, Lyden D. (2001). The Id proteins and angiogenesis. Oncogene 20: 8334–8341. Benezra R. (2001). Role of Id proteins in embryonic and tumor angiogenesis. Trends Cardiovasc Med 11: 237–241. Botchkarev VA, Sharov AA. (2004). BMP signaling in the control of skin development and hair follicle growth. Differentiation 72: 512–526. Cavallaro U, Christofori G. (2000). Molecular mechanisms of tumor angiogenesis and tumor progression. J Neurooncol 50: 63–70. Deckers MM, van Bezooijen RL, van der HG, Hoogendam J, van Der BC, Papapoulos SE et al. (2002). Bone morphogenetic proteins stimulate angiogenesis through osteoblastderived vascular endothelial growth factor A. Endocrinology 143: 1545–1553. Ding Z, Liu Z, Bi Y, Tian H, Li G, Song T. (2005). Morphological study of the interaction between M21 melanoma and lymphatic endothelium. Lymphology 38: 87–91. Folberg R, Maniotis AJ. (2004). Vasculogenic mimicry. APMIS 112: 508–525. Folberg R, Hendrix MJ, Maniotis AJ. (2000). Vasculogenic mimicry and tumor angiogenesis. Am J Pathol 156: 361–381. Hatakeyama S, Gao YH, Ohara-Nemoto Y, Kataoka H, Satoh M. (1997). Expression of bone morphogenetic proteins of human neoplastic epithelial cells. Biochem Mol Biol Int 42: 497–505. Hatakeyama S, Ohara-Nemoto Y, Kyakumoto S, Satoh M. (1993). Expression of bone morphogenetic protein in human adenocarcinoma cell line. Biochem Biophys Res Commun 190: 695–701. Hendrix MJ, Seftor EA, Hess AR, Seftor RE. (2003). Molecular plasticity of human melanoma cells. Oncogene 22: 3070–3075. Hogan BL. (1996). Bone morphogenetic proteins in development. Curr Opin Genet Dev 6: 432–438. Ide H, Yoshida T, Matsumoto N, Aoki K, Osada Y, Sugimura T et al. (1997). Growth regulation of human prostate cancer cells by bone morphogenetic protein-2. Cancer Res 57: 5022–5027. Jones N, Dumont DJ. (2000). Tek/Tie2 signaling: new and old partners. Cancer Metastasis Rev 19: 13–17. Kiyozuka Y, Nakagawa H, Senzaki H, Uemura Y, Adachi S, Teramoto Y et al. (2001). Bone morphogenetic protein-2 and type IV collagen expression in psammoma body forming ovarian cancer. Anticancer Res 21: 1723–1730. Kleeff J, Maruyama H, Ishiwata T, Sawhney H, Friess H, Buchler MW et al. (1999). Bone morphogenetic protein 2 exerts diverse effects on cell growth in vitro and is expressed in human pancreatic cancer in vivo. Gastroenterology 116: 1202–1216.

Langenfeld EM, Langenfeld J. (2004). Bone morphogenetic protein-2 stimulates angiogenesis in developing tumors. Mol Cancer Res 2: 141–149. Lin P, Buxton JA, Acheson A, Radziejewski C, Maisonpierre PC, Yancopoulos GD et al. (1998). Antiangiogenic gene therapy targeting the endothelium-specific receptor tyrosine kinase Tie2. Proc Natl Acad Sci USA 95: 8829–8834. Maniotis AJ, Folberg R, Hess A, Seftor EA, Gardner LM, Pe’er J et al. (1999). Vascular channel formation by human melanoma cells in vivo and in vitro: vasculogenic mimicry. Am J Pathol 155: 739–752. McDonald DM, Foss AJ. (2000). Endothelial cells of tumor vessels: abnormal but not absent. Cancer Metastasis Rev 19: 109–120. McDonald DM, Munn L, Jain RK. (2000). Vasculogenic mimicry: how convincing, how novel, and how significant? Am J Pathol 156: 383–388. Mehler MF, Mabie PC, Zhang D, Kessler JA. (1997). Bone morphogenetic proteins in the nervous system. Trends Neurosci 20: 309–317. Miyazawa K, Shinozaki M, Hara T, Furuya T, Miyazono K. (2002). Two major Smad pathways in TGF-beta superfamily signalling. Genes Cells 7: 1191–1204. Pache G, Schafer C, Wiesemann S, Springer E, Liebau M, Reinhardt HC et al. (2006). Upregulation of Id-1 via BMP-2 receptors induces reactive oxygen species in podocytes. Am J Physiol Renal Physiol 291: F654–F662. Peng H, Usas A, Olshanski A, Ho AM, Gearhart B, Cooper GM et al. (2005). VEGF improves, whereas sFlt1 inhibits, BMP2-induced bone formation and bone healing through modulation of angiogenesis. J Bone Miner Res 20: 2017–2027. Peng H, Wright V, Usas A, Gearhart B, Shen HC, Cummins J et al. (2002). Synergistic enhancement of bone formation and healing by stem cell-expressed VEGF and bone morphogenetic protein-4. J Clin Invest 110: 751–759. Peters KG, Kontos CD, Lin PC, Wong AL, Rao P, Huang L et al. (2004). Functional significance of Tie2 signaling in the adult vasculature. Recent Prog Horm Res 59: 51–71. Raida M, Clement JH, Leek RD, Ameri K, Bicknell R, Niederwieser D et al. (2005). Bone morphogenetic protein 2 (BMP-2) and induction of tumor angiogenesis. J Cancer Res Clin Oncol 131: 741–750. Rothhammer T, Poser I, Soncin F, Bataille F, Moser M, Bosserhoff AK. (2005). Bone morphogenic proteins are overexpressed in malignant melanoma and promote cell invasion and migration. Cancer Res 65: 448–456. Saalbach A, Wetzel A, Haustein UF, Sticherling M, Simon JC, Anderegg U. (2005). Interaction of human Thy-1 (CD 90) with the integrin alphavbeta3 (CD51/CD61): an important mechanism mediating melanoma cell adhesion to activated endothelium. Oncogene 24: 4710–4720. Seftor EA, Meltzer PS, Schatteman GC, Gruman LM, Hess AR, Kirschmann DA et al. (2002). Expression of multiple molecular phenotypes by aggressive melanoma tumor cells: role in vasculogenic mimicry. Crit Rev Oncol Hematol 44: 17–27. Oncogene


4170 Straume O, Akslen LA. (2005). Strong expression of ID1 protein is associated with decreased survival, increased expression of ephrin-A1/EPHA2, and reduced thrombospondin-1 in malignant melanoma. Br J Cancer 93: 933–938. ten Dijke P, Korchynskyi O, Valdimarsdottir G, Goumans MJ. (2003). Controlling cell fate by bone morphogenetic protein receptors. Mol Cell Endocrinol 211: 105–113. Turgeman G, Pittman DD, Muller R, Kurkalli BG, Zhou S, Pelled G et al. (2001). Engineered human mesenchymal stem cells: a novel platform for skeletal cell mediated gene therapy. J Gene Med 3: 240–251. Ushio-Fukai M, Tang Y, Fukai T, Dikalov SI, Ma Y, Fujimoto M et al. (2002). Novel role of gp91(phox)-containing NAD(P)H oxidase in vascular endothelial growth factorinduced signaling and angiogenesis. Circ Res 91: 1160–1167. Valdimarsdottir G, Goumans MJ, Rosendahl A, Brugman M, Itoh S, Lebrin F et al. (2002). Stimulation of Id1 expression by bone morphogenetic protein is sufficient and necessary

Oncogene

for bone morphogenetic protein-induced activation of endothelial cells. Circulation 106: 2263–2270. Vogt RR, Unda R, Yeh LC, Vidro EK, Lee JC, Tsin AT. (2006). Bone morphogenetic protein-4 enhances vascular endothelial growth factor secretion by human retinal pigment epithelial cells. J Cell Biochem 98: 1196–1202. Wang EA, Rosen V, D’Alessandro JS, Bauduy M, Cordes P, Harada T et al. (1990). Recombinant human bone morphogenetic protein induces bone formation. Proc Natl Acad Sci USA 87: 2220–2224. Ward NL, Dumont DJ. (2002). The angiopoietins and Tie2/ Tek: adding to the complexity of cardiovascular development. Semin Cell Dev Biol 13: 19–27. Wozney JM, Rosen V, Celeste AJ, Mitsock LM, Whitters MJ, Kriz RW et al. (1988). Novel regulators of bone formation: molecular clones and activities. Science 242: 1528–1534. Yamamoto Y, Oelgeschlager M. (2004). Regulation of bone morphogenetic proteins in early embryonic development. Naturwissenschaften 91: 519–534.