Oncogene (2008) 27, 6834–6844 & 2008 Macmillan Publishers Limited All rights reserved 0950-9232/08 $32.00 www.nature.com/onc ORIGINAL ARTICLE Glioblastoma-secreted factors induce IGFBP7 and angiogenesis bymodulating Smad-2-dependent TGF- b signaling

A Pen1,2, MJ Moreno1, Y Durocher3, P Deb-Rinker4 and DB Stanimirovic1,2

1Cerebrovascular Research Group, Neurobiology Program, Institute for Biological Sciences, National Research Council of Canada, Ottawa, Ontario, Canada; 2Department of Cellular and Molecular Medicine, Faculty of Medicine, University of Ottawa, Ottawa, Ontario, Canada; 3Animal Cell Technology Group, Biotechnology Research Institute, National Research Council of Canada, Montreal, Quebec, Canada and 4Neurogenesis and Brain Repair Group, Institute for Biological Sciences, National Research Council of Canada, Ottawa, Ontario, Canada

Insulin-like -binding 7 (IGFBP7) is a Introduction selective biomarker of glioblastoma (GBM) vessels, stronglyexpressed in tumor endothelial cells and vascular Gliomas are the most common tumors of the central basement membrane. IGFBP7 regulation and its nervous system (CNS) derived from glial cells and potential role in tumor angiogenesis remain unclear. classified into four clinical grades (World Health Mechanisms of IGFBP7 induction and its angiogenic Organization (WHO) grades I–IV; Kleihues et al., capacitywere examined in human brain endothelial cells 1995; Dai and Holland, 2001). Glioblastoma multiforme (HBECs) exposed to tumor-like conditions. HBEC (GBM; WHO grade IV) is the most malignant type with treated with GBM cell (U87MG)-conditioned media a mean survival time of approximately 9–12 months (-CM) exhibited fourfold upregulation of IGFBP7 mRNA (Vajkoczy and Menger, 2000). Common features and protein compared to control cells. IGFBP7 gene of GBMs include fast cell proliferation, invasion regulation in HBEC was methylation independent. into normal brain parenchyma, intratumoral necrosis, U87MG-CM analysed by enzyme-linked immunosorbent hypoxia and high angiogenic activity (Vajkoczy et al., assaycontained B5pM transforming growth factor 1999; Louis, 2006). (TGF)-b1, a concentration sufficient to stimulate IGFBP7 The microenvironment of the CNS is maintained by in HBEC to similar levels as U87MG-CM. Both pan- the blood–brain barrier (BBB) formed by specialized TGF-b-neutralizing antibody(1D11) and the TGF- b1 endothelial cells, which are distinguished from those in receptor (activin receptor-like kinase 5, ALK5) antago- the periphery by high mitochondrial content (Oldendorf nist, SB431542, blocked U87MG-CM-induced IGFBP7 et al., 1977), lack of fenestrations, minimal pinocytic expression in HBEC, indicating that TGF-b1isan activity and the presence of tight junctions that restrict important tumor-secreted effector capable of IGFBP7 paracellular permeability of hydrophilic molecules induction in endothelial cells. HBEC exposed to either (Kniesel and Wolburg, 2000). GBM vessels lose many U87MG-CM or IGFBP7 protein exhibited increased BBB properties, leading to clinical signs of brain edema capillary-like tube (CLT) formation in Matrigel. Both due to increased vessel permeability (Long, 1970). TGF-b1- and U87MG-CM-induced Smad-2 phosphoryla- Morphologically, GBM vessels are characterized by tion and U87MG-CM-induced CLT formation in HBEC fenestrations, increased number of caveolae, wide were inhibited bythe ALK5 antagonist, SB431542. These intercellular junctions, abnormal pericytes and a dis- data suggest that proangiogenic IGFBP7 maybe induced continuous basement membrane (Hirano and Matsui, in brain endothelial cells byTGF- bs secreted byGBM, 1975; Dinda et al., 1993). GBM vessels express unique most likelythrough TGF- b1/ALK5/Smad-2 pathway. gene and protein biomarkers (Pen et al., 2007), often Oncogene (2008) 27, 6834–6844; doi:10.1038/onc.2008.287; associated with angiogenesis and/or increased perme- published online 18 August 2008 ability, including aquaporin 4 (Davies, 2002), plasma- lemmal vesicle associated protein-1 (Madden et al., Keywords: glioblastoma; angiogenesis; IGFBP7; 2004) and TEM7 (PLXDC1; Beaty et al., 2007). Recent TGF-b; Smad transcriptomic analysis of laser-captured microdissected vessels from GBM tumors identified -like growth factor-binding protein 7 (IGFBP7) as a highly selective biomarker of tumor vessels (Pen et al., 2007). IGFBP7 was expressed by GBM endothelial cells and extensively Correspondence: Dr DB Stanimirovic, Institute for Biological deposited into the vascular basal lamina (Pen et al., Sciences, National Research Council of Canada, 1200 Montreal Road, 2007). Building M-54, Ottawa, Ontario, Canada K1A 0R6. IGFBP7 is a related member of the IGFBP family E-mail: [email protected] Received 25 February 2008; revised 5 June 2008; accepted 1 July 2008; (Hwa et al., 1999), cloned by several groups (Murphy published online 18 August 2008 et al., 1993; Akaogi et al., 1994; Yamauchi et al., 1994), Tumor cells induce IGFBP7 in endothelium APenet al 6835 which, unlike the other family members (IGFBP1–6), exposed to U87MG-CM responded with a time-depen- exhibits a low affinity for IGF but a high and specific dent upregulation of IGFBP7 mRNA (Figure 1a, upper affinity for insulin (Oh et al., 1996). IGFBP7 is a cell panel) and IGFBP7 protein (Figure 1a, lower panel), adhesive glycoprotein of about 30kDa (Akaogi et al., reaching maximal levels after 3–6 days of incubation in 1994), which is regulated by proteolytic cleavage into a U87MG-CM. The pronounced upregulation of IGFBP7 two-chain form by a membrane-bound serine proteinase protein at 6 days of exposure to U87MG-CM was matriptase (Ahmed et al., 2006). Varied IGFBP7 detected by western blot in both HBEC lysates expression patterns have been reported in different (Figure 1c) and in the secreted fractions (Figure 1a, tumor types (Ruan et al., 2007). IGFBP7 transcription lower panel, d). Conditioned media (CM) from another in tumor cells is modulated by DNA methylation human GBM cell line, T98G, induced IGFBP7 mRNA (Komatsu et al., 2000), retinoic acid and transforming and protein in HBEC in a similar fashion as that growth factor (TGF)-b1 (Hwa et al., 1999). IGFBP7 has observed with U87MG-CM (Supplementary Figure S1). been implicated in tumor growth suppression (Burger IGFBP7 protein was not detected in either U87MG-CM et al., 1998; Sprenger et al., 1999) through induction of (Figure 1d) or T98G-CM (Supplementary Figure S1, apoptosis or (Wilson et al., 2002; lower panel). Mutaguchi et al., 2003). Recently, Wajapeyee et al. Heat-inactivated U87MG-CM lost the ability to (2008) identified a mechanism by which BRAF proto- induce IGFBP7 mRNA (Figure 1b) and protein oncogene triggers cellular senescence in melanocytes (Figures 1c and d) expression in HBEC, suggesting that through induction of IGFBP7. heat-liable mediator(s), most likely , produced In contrast to the tumor suppressor role of IGFBP7 in by GBM cells were responsible for IGFBP7 induction. cancer cells, IGFBP7 is highly upregulated in vessels of several tumors, including GBM (Akaogi et al., 1996; Pen IGFBP7 gene methylation in HBEC is not affected et al., 2007), and interacts in vitro with various by glioblastoma-conditioned media extracellular matrix (ECM) proteins to stimulate adhe- To investigate whether IGFBP7 induction in HBEC sion and migration/invasion of vascular endothelial cells treated with U87MG-CM was the result of DNA on plastic substrates (Sato et al., 1999; Kishibe et al., demethylation, the methylation status of IGFBP7 gene 2000); hence, the original name ‘angiomodulin’ (Akaogi in HBEC treated with either Dulbecco’s modified Eagle’s et al., 1996). Although literature evidence suggests that medium (DMEM; control) or U87MG-CM was ana- IGFBP7 might exhibit angiogenesis-modulating proper- lysed (as described in Supplementary Materials and ties (Akaogi et al., 1996; van Beijnum et al., 2006), the Methods; Miller et al., 1998; Deb-Rinker et al., 2005) role of IGFBP7 in tumor angiogenesis remains poorly using primer sets (Supplementary Table 1) designed to understood. amplify cytosine-phosphate-guanine (CpG) sites present As IGFBP7 is selectively expressed in GBM vessels, in the promoter region (primer sets A and B) or exon 1 the goal of this study was to identify the components of and exon 1/intron 1 (primer sets C and D, respectively) the glial tumor microenvironment that induce IGFBP7 of IGFBP7 gene (Chen et al., 2007). The methylation expression in brain endothelial cells and to determine status of DMEM- and U87MG-CM-treated HBEC in the potential role of IGFBP7 in angiogenesis. IGFBP7 the exon 1 region ( þ 73 to þ 472; primer pair IGFBP7- expression in brain endothelial cells was found to be C) of IGFBP7 spanning 45 CpG sites could not be unresponsive to hypoxia, but upregulated by secreted determined, possibly due to the high density of CpG factors from GBM cells through TGF-b1/ALK5/Smad- islands present in this region resulting in unspecific PCR 2 signaling pathway. products. The PCR-amplified fragment spanning 19 CpG sites ( þ 464 to þ 649; primer pair IGFBP7-D) showed a higher (but not statistically significant) number of methylation sites in control HBEC DNA compared to Results U87MG-CM-treated HBEC DNA (Supplementary Fig- ure S2, graph). No differences in methylation state were Glioblastoma-conditioned media induces IGFBP7 detected within the 5 CpG (À941 to À766; primer pair in human brain endothelial cells IGFBP7-A) and 6 CpG (À789 to À577; primer pair To examine whether tumor-like microenvironment IGFBP7-B) sites of IGFBP7 promoter regions (Supple- affects the expression of IGFBP7 in brain endothelial mentary Figure S2). As no specific CpG sites were cells, cultured human brain endothelial cells (HBECs) consistently and significantly methylated in control and were exposed to either hypoxia (Vaupel et al., 1998) or demethylated in U87MG-CM-treated HBEC, we media conditioned by the human GBM cell line, concluded that IGFBP7 induction in U87MG-CM U87MG (U87MG-CM). Under control conditions, exposed HBEC was most likely regulated by mechanisms HBEC exhibited low abundance of IGFBP7 transcripts other than IGFBP7 DNA methylation. (two to threefold lower than b-actin in the same cells) (Figure 1a, upper panel) and protein (Figure 1a, lower panel, c, d), in agreement with the observed absence of TGF-b1induces IGFBP7 expression in HBEC IGFBP7 protein immunoreactivity in nonmalignant The nature of IGFBP7-inducing mediators present in brain vessels in vivo (Pen et al., 2007). Whereas hypoxia U87MG-CM was examined next. From our previous failed to induce IGFBP7 (A Pen, unpublished), HBEC studies (Moreno et al., 2006) analysing U87MG cell

Oncogene Tumor cells induce IGFBP7 in endothelium APenet al 6836

Figure 1 Insulin-like growth factor-binding protein 7 (IGFBP7) expression in human brain endothelial cells (HBECs) exposed to glioblastoma cell-conditioned media (U87MG-CM). (a) IGFBP7 mRNA (upper panel) and protein (lower panel) expression, determined by Q-PCR and western blot, respectively, in HBEC exposed to serum-free Dulbecco’s modified Eagle’s medium (DMEM) or U87MG-CM for indicated periods. (b) IGFBP7 mRNA expression in HBEC exposed to serum-free DMEM, U87MG-CM, or heat- inactivated U87MG-CM (HI-U87MG-CM) for 6 days. (c) IGFBP7 protein expression in total HBEC lysates. (d) In the HBEC media before and 6 days after the addition of U87MG-CM or HI-U87MG-CM. Relative expression of IGFBP7 mRNA was normalized to that of b-actin in same cell extracts. Each bar represents the mean±s.e.m. of 4–11 experiments. Asterisks indicates a significant (Po0.05; analysis of variance (ANOVA) followed by Newman–Keuls posttest) difference between DMEM- and U87MG-CM-treated HBEC. Actin was used as a loading control for the western blot bands in cell lysates. Gels are representative of at least three separate experiments showing similar results.

transcriptome, we selected and examined two secreted was B5pM. HBEC exposed to 5 pM TGF-b1 responded angiogenic proteins identified by enzyme-linked immu- with a time-dependent upregulation of IGFBP7 mRNA nosorbent assay (ELISA) in U87MG-CM, VEGF-A (Figure 2a, upper panel) and protein (Figure 2a, lower and SPARC, for their ability to induce IGFBP7 panel), reaching maximal levels 3–6 days after IGFBP7 expression in HBEC. Either VEGF-A or SPARC addition. In HBEC exposed to TGF-b1 concentrations (20ng/ml) demonstrated minimal effects on IGFBP7 ranging from 5 to 100 pM for 6 days, IGFBP7 mRNA expression in HBEC (A Pen, unpublished). On the basis was upregulated in a dose-dependent manner, with of previous studies showing that TGF-b1 induces EC50 ¼ 12 pM (Figure 2b). The IGFBP7 induction with IGFBP7 expression in bovine retinal capillary endothe- 5pM TGF-b1 was similar (2.5- to 4-fold) to that lial cells (Hata et al., 2000), we examined whether TGF- observed with U87MG-CM (Figures 1 and 2). b1 is released in U87MG-CM. The average amount of To confirm that secreted TGF-b1 is responsible for TGF-b1 secreted in U87MG-CM determined by ELISA the IGFBP7 upregulation, U87MG-CM were pretreated

Oncogene Tumor cells induce IGFBP7 in endothelium APenet al 6837

Figure 2 Transforming growth factor (TGF)-b1 induces insulin-like growth factor-binding protein 7 (IGFBP7) expression in human brain endothelial cell (HBEC). (a) IGFBP7 mRNA (upper panel) and protein (lower panel) expression, determined by Q–PCR and western blot, respectively, in HBEC exposed to Dulbecco’s modified Eagle’s medium (DMEM) or 5 pM TGFb-1 for indicated periods. (b) The expression of IGFBP7 mRNA in HBEC exposed to either DMEM or the indicated concentrations of TGF-b1 for 6 days. (c) IGFBP7 mRNA expression in HBEC exposed to 5 pM TGF-b1 or U87MG-CM in the absence or presence of 1.25 mg/ml of the pan- TGF-b-neutralizing antibody 1D11 (R&D System). (d) IGFBP7 mRNA expression in HBEC exposed to 5 pM TGF-b1 or U87MG- CM in the absence or presence of 20or 200ng/ml of TGF- b1-specific neutralizing antibody (R&D System). Relative expression of IGFBP7 mRNA was normalized to that of b-actin in same cell extracts. Each bar represents the mean±s.e.m. of at least three experiments. Asterisks indicate significance between DMEM-treated and TGF-b1- or U87MG-CM-treated HBEC; # and þþ indicate significance between 5 pM TGF-b1- or U87MG-CM-treated HBEC, respectively, in the absence or presence of the neutralizing antibodies. Significance (Po0.001) was determined by analysis of variance (ANOVA) followed by Newman–Keuls posttest comparison among means. with an antibody that neutralizes the activities of TGF- TGF-b1modulates IGFBP7 expression through ALK5 b1, TGF-b2 and TGF-b3 (pan-TGF-b antibody 1D11); receptor and the signal transduction adapter proteins the 1D11 antibody-treated U87MG-CM failed to induce Smad-2 and Smad-1/-5 IGFBP7 in HBEC (Figure 2c). In contrast, an antibody TGF-b1 has been reported to regulate angiogenesis in specific for TGF-b1 ligand (R&D System, Minneapolis, endothelial cells through two Smad pathways: the MN, USA) only partially (B50%) inhibited the canonical Smad-2/-3/-4 pathway by its predominant IGFBP7-inducing activity of U87MG-CM (Figure 2d), type I receptor activin receptor-like kinase (ALK) 5, and suggesting that other TGF-b isoforms participate in the Smad-1/-5/-4 pathway by ALK1 (Lebrin et al., IGFBP7 induction by U87MG-CM. Two other TGF-b 2005). To investigate which signaling cascade triggers isoforms, TGF-b2 and TGF-b3, also induced IGFBP7 TGF-b1-mediated induction of IGFBP7, HBEC were expression in HBEC in a concentration-dependent exposed to TGF-b1 or U87MG-CM in the presence of manner; the IGFBP7 induction was similar in magni- SB431542, a selective antagonist of ALK5, but not of tude for all three TGF-b isoforms (Supplementary ALK1 (Inman et al., 2002). Both IGFBP7 mRNA Figure S3). However, only TGF-b2 isoform was (Figure 3a) and protein induction and release detected in U87MG-CM by ELISA (B2–5 pM) suggest- (Figure 3b) stimulated by TGF-b1 or U87MG-CM ing that TGF-b2, in addition to TGF-b1, may partici- were completely inhibited in the presence of 10 mM pate in U87MG-CM-mediated induction of IGFBP7 SB431542 (Figures 3a and b). The levels of phosphory- expression in HBEC. lated Smad-2 and Smad-1/-5 were examined in nuclear

Oncogene Tumor cells induce IGFBP7 in endothelium APenet al 6838

Figure 4 Transforming growth factor (TGF)-b1 induces insulin- like growth factor-binding protein 7 (IGFBP7) through ALK5/ Smad-2 pathway. Human brain endothelial cells (HBECs) were treated with DMEM, 5 pM TGF-b1 or U87MG-CM alone or in the presence of 10 mM of the ALK5 antagonist SB431542 for 30min at 37 1C before lysis. SB431542 was added to HBEC 20min before TGF-b1 or U87MG-CM. Nuclear and cytosolic extracts were isolated as described in ‘Materials and methods’ section. The phosphorylation of Smad-2 and Smad-1/-5 was analysed by western blot using respective anti-phospho-Smad antibodies. Cytosolic fraction was probed by respective anti-Smad antibodies. NuMA and actin were used as controls for nuclear and cytosolic protein loading, respectively. Gels shown are representative of three separate experiments showing similar results.

Figure 3 Insulin-like growth factor-binding protein 7 (IGFBP7) induction by U87MG-CM in human brain endothelial cells (HBECs) is inhibited by ALK5 antagonist. HBECs were exposed to either a serum-free Dulbecco’s modified Eagle’s medium Smad-1/-5 phosphorylation were inhibited in the presence (DMEM), 5 pM transforming growth factor (TGF)-b1, U87MG- of SB431542. As SB431542 has no effect on ALK1/Smad CM in the absence or presence of 10 mM ALK5 antagonist phosphorylation (Goumans et al., 2003; Watabe et al., (SB431542) for 6 days. (a) IGFBP7 mRNA expression was 2003), these observations suggest that ALK1-mediated determined by real-time Q–PCR. The expression of IGFBP7 was phosphorylation of Smad-1/-5 in HBEC cells is probably normalized to that of b-actin in same cell extracts. Each bar represents the mean±s.e.m. of at least three experiments. Asterisks induced secondary to the activation of ALK5. Findings indicate significance between DMEM-treated and 5 pM TGF-b1or by Goumans et al. (2003) demonstrating that ALK5 U87MG-CM-treated HBEC; # and þþ indicate significance kinase activity is required for ALK1 activation by TGF-b between 5 pM TGF-b1 and U87MG-CM-treated HBEC without support this conclusion. These results suggest that or with ALK5 antagonist, respectively. Significance (Po0.05) was determined by analysis of variance (ANOVA) followed by New- U87MG-CM induces IGFBP7 in HBEC preferentially man–Keuls posttest comparison among means. (b) The expression through ALK5 pathway leading to Smad-2 protein of IGFBP7 protein in cell lysates or secreted into the media was phosphorylation. evaluated by western blot. Actin was used as a loading control. Gels shown are representative of at least three separate experiments showing similar results. IGFBP7-induced capillary formation To evaluate the effect of IGFBP7 protein on endothelial cells, a recombinant IGFBP7 was produced and purified and cytosolic extracts of cells treated with TGF-b1or (as described in Supplementary Materials and Methods; U87MG-CM in the absence or presence of SB431542; Durocher et al., 2002). Recombinant IGFBP7 was whereas the levels of Smad-2 and Smad-5 in the detected by SDS–polyacrylamide gel electrophoresis cytosolic extracts remained unchanged in differently (PAGE) Coomassie blue staining (Supplementary treated cells (Figure 4, lower panel), a strong Smad-2 Figure S4A) and by western blot using a penta-His phosphorylation and a weaker Smad-1/-5 phosphoryla- horseradish peroxidase (HRP) antibody (Supplementary tion in response to both TGF-b1 and U87MG-CM was Figure S4B) as a band of approximately 37 kDa. Mass observed in the nuclear extracts of HBEC (Figure 4, spectrometry analysis indicated that IGFBP7 purity was upper panel). TGF-b1 was previously shown to induce B90–95% and that the contaminant band observed at both Smad-2 and Smad-1/-5 phosphorylation in wild- B200 kDa in SDS–PAGE gel was collagen II a-precursor type mouse embryonic endothelial cells and bovine (A Pen, unpublished). IGFBP7 functionality/activity aortic endothelial cells (Goumans et al., 2002). Both was determined by its ability to bind rh6Ckine/CCL21 TGF-b1- and U87MG-CM-induced Smad-2 and (Nagakubo et al., 2003; Supplementary Figure S4C).

Oncogene Tumor cells induce IGFBP7 in endothelium APenet al 6839

Figure 6 Transforming growth factor (TGF)-b1 antagonist inhibits U87MG-CM-induced capillary-like tube (CLT) formation. Human brain endothelial cells (HBEC) were grown in Matrigel in Figure 5 Effect of insulin-like growth factor-binding protein 7 Dulbecco’s modified Eagle’s medium (DMEM) (a), U87MG-CM (IGFBP7) on capillary-like tube (CLT) formation by human brain alone (b) or supplemented with 10 mM SB431542 (c) for 24 h. Scale endothelial cells (HBEC) grown in Matrigel. Representative bar ¼ 500 mm. (d) The total length of CLTs and the number of photomicrographs of CLTs formed by HBEC grown in Matrigel branching points were quantified using Northern Eclipse v.5.0. and exposed to Dulbecco’s modified Eagle’s medium (DMEM) (a), Each bar represents the mean±s.e.m. of 11 experiments. Asterisks 50n M IGFBP7 (b) and U87MG-CM (c) for 24 h are shown. Scale indicate significance between DMEM- and U87MG-CM-treated bar ¼ 500 mm. (d) The total length of CLTs and the number of HBEC; # indicates significance between U87MG-CM in the absence branching points were quantified using Northern Eclipse v.5.0. or presence of 10 mM SB431542. Significance (P 0.05) was Each bar represents the mean±s.e.m. of triplicate experiments. o determined by analysis of variance (ANOVA) followed by New- Asterisk indicates a significant (P 0.05; analysis of variance o man–Keuls posttest comparison among means. (ANOVA) followed by Newman–Keuls posttest) difference be- tween HBEC treated with DMEM and IGFBP7 or U87MG-CM.

To assess whether IGFBP7 has angiogenic activity, proangiogenic activity of IGFBP7 in the examined HBECs grown in Matrigel, a biological basement concentration range. membrane matrix, were exposed to DMEM or To examine if the proangiogenic activity of U87MG- U87MG-CM or to different concentrations of IGFBP7. CM is associated with TGF-b signaling pathway, HBEC HBEC exposed to U87MG-CM for 24 h (Figure 5c) were exposed to U87MG-CM in the presence of 10 mM formed an extensive network of capillary-like tubes SB431542. U87MG-CM-induced CLT formation in (CLTs) compared to cells cultured in DMEM alone HBEC (Figures 6a and b) was reduced to control levels (Figure 5a), quantified as significant increases in both in the presence of 10 mM SB431542 (Figures 6c and d). The total length of capillary cords and the number of same concentration of SB431542 solvent, DMSO, did not branching nodes (Figure 5d). IGFBP7 also induced affect CLT formation (A Pen, unpublished). These results CLT formation in HBEC (Figure 5b) with a maximal imply that TGF-b signaling through ALK5 is necessary induction observed at 50n M (Figure 5d), suggesting a for the proangiogenic activity of U87MG-CM.

Oncogene Tumor cells induce IGFBP7 in endothelium APenet al 6840 Discussion invasion, metastasis (Massague, 1998) and vascular remodeling (Pepper, 1997), acting as either an inhibitor The selective upregulation of IGFBP7 in GBM vessels or stimulator of angiogenesis depending on experimen- compared to nonmalignant brain vessels (Pen et al., tal conditions (Goumans et al., 2002). TGF-b signaling 2007) suggests a role for IGFBP7 in tumor vascular is initiated by binding of the ligand to its constitutively function. However, the mechanisms of IGFBP7 induc- active cell surface receptor, TbR-II. Upon binding, tion and its role in tumor angiogenesis are poorly TbR-II activates and phosphorylates TbR-I, also known understood. The present study provides the first evidence as ALK (de Caestecker, 2004). Subsequently, the signal that IGFBP7 is induced in HBEC by GBM-cell-secreted is propagated to the nucleus by phosphorylated Smad mediators, including TGF-b1, through TGF-b1/ALK5/ proteins where they participate in transcriptional Smad-2 pathway and by a DNA methylation-independent regulation of target (Derynck et al., 1998). mechanism. Furthermore, IGFBP7 is shown to stimulate Although activated ALK5 induces phosphorylation of HBEC angiogenic phenotype and may be involved in Smad-2/-3, ALK1 induces the phosphorylation of angiogenesis induced by tumor microenvironment. Smad-1/-5 (Chen and Massague, 1999). Prefer- IGFBP7 stimulation by tumor microenvironment in ential induction of ALK1 or ALK5 signaling activity brain endothelial cells was investigated by probing both (Goumans et al., 2002) is dependent on TGF-b potential mechanisms of IGFBP7 gene regulation in concentrations: Smad-2 phosphorylation is initiated at endothelial cells and the nature and signaling pathways 1pM and is maximal at 10p M TGF-b, whereas Smad-5 induced by secreted mediators (proteins in CM) of GBM phosphorylation is triggered by 10–20 pM of TGF-b cells. (Goumans et al., 2002). We propose that TGF-bs Compelling literature evidence indicates that hyper- (-1 and -2) secreted by GBM cells mediate IGFBP7 methylation of CpG islands of IGFBP7 is responsible induction in HBEC by preferentially activating ALK5/ for downregulation of IGFBP7 in tumor cells contribut- Smad-2 signaling pathway based on the following ing to tumor progression (Komatsu et al., 2000; Chen observations: (1) measured levels of TGF-b1in et al., 2007; Lin et al., 2007). In contrast to colon (Lin U87MG-CM were B5pM, (2) the U87MG-CM induc- et al., 2007) and lung (Chen et al., 2007) IGFBP7- tion of IGFBP7 was completely blocked by a selective negative cancer cell lines in which sites þ 490to þ 610 inhibitor of ALK5, SB431542 and (3) increases in of the 50 CpG island of IGFBP7 gene were all phosho-Smad-2 detected in nuclear extracts of HBEC methylated, neither these sites ( þ 490to þ 610) nor in response to both TGF-b1 and U87MG-CM were also promoter regions of IGFBP7 were highly methylated in inhibited by SB431542. The observed weak increases in either control or U87MG-CM-treated HBEC. These phosphorylated Smad-1/-5 after TGF-b1 or U87MG- results suggest that IGFBP7 gene regulation by DNA CM addition indicate that IGFBP7 expression can also methylation is likely different between cancer cells be induced by activation of ALK1/Smad-1/-5 pathway. and endothelial cells, and that DNA demethylation However, as SB431542 partially blocked these re- was not involved in regulating IGFBP7 gene induction sponses, the phosphorylation of Smad-1/-5 appears to in HBEC. be in part secondary to the activation of ALK5. TGF-b1 IGFBP7 expression in peripheral endothelial cells can also act through receptor-dependent, but Smad- can be modulated by cell confluence, hypoxia, and independent pathways, including the activation of cytokines including VEGF, bFGF and TGF-b1(Hata mitogen-activated protein kinase (Blanchette et al., et al., 2000). In this study, IGFBP7 expression in HBEC 2001), c-Jun N-terminal kinase (Engel et al., 1999) and was potently induced by U87MG-CM and TGF-b1and protein kinase C (Yakymovych et al., 2001). was not affected by VEGF or hypoxia. TGF-b is a 25 kDa In this study, the IGFBP7 induction was observed as homodimeric protein present in three isoforms: TGF-b1, early as 24 h after HBEC treatment with either TGF-b1 -b2and-b3(Lebrinet al., 2005); the three isoforms of TGF- or U87MG-CM, reaching maximal levels of mRNA b interact with the same cell surface serine/threonine kinase and/or protein upregulation after 6 days. Given this receptor type I (TbR-I) and type II (TbR-II) (ten Dijke prolonged time course, autocrine stimulation by other et al., 1988; de Caestecker, 2004) to induce several common TGF-b1-induced mediators might in part be responsible but also distinct responses (Lebrin et al., 2005). GBM for IGFBP7 induction. tumors express all three TGF-b isoforms (Constam et al., The study further provides evidence of proangiogenic 1992; Olofsson et al., 1992; Weller et al., 1995; Yamada function of IGFBP7 in HBEC. Whereas IGFBP7 failed et al., 1995; Wick et al., 2001). Although all three TGF-b to stimulate HBEC proliferation (A Pen, unpublished), isoforms can induce IGFBP7 expression in HBEC, only it induced a dose-dependent elaboration of CLT net- TGF-b1and-b2 were detected in U87MG-CM. As works in HBEC grown in Matrigel. In agreement with the specific TGF-b1-neutralizing antibody only partially this observation, a recent study demonstrated that blocked U87MG-CM-induced IGFBP7 expression, in IGFBP7 antibody inhibits endothelial tube formation contrast to the complete inhibition observed with the pan- in an in vitro collagen gel-based sprout formation assay TGF-bs antibody (1D11), it is suggested that TGF-b1 (van Beijnum et al., 2006). HBEC CLT formation and -b2 isoforms are key U87MG-released mediators that induced by U87MG-CM was inhibited by the ALK5 stimulate IGFBP7 in HBEC. antagonist, SB431542, implicating TGF-b/ALK5 TGF-b has been implicated in various biological signaling pathway in U87MG-CM-induced angiogenic functions including cell cycle control, carcinogenesis, responses. As the same pathway was responsible for

Oncogene Tumor cells induce IGFBP7 in endothelium APenet al 6841 IGFBP7 induction in HBEC, it is tempting to speculate that autocrine actions of IGFBP7 could participate in U87MG-CM-induced angiogenic responses in HBEC. TGF-b mediates the transcription of several genes related to the ECM remodeling, including SPARC and collagen I (Reed et al., 1994), MMP-2, TIMP-1 (Kwak et al., 2006), plasminogen activator inhibitor-1 (Keeton et al., 1991) and collagen IV (Poncelet and Schnaper, 1998). Secreted IGFBP7 is known to interact with ECM components (Akaogi et al., 1996; St Croix et al., 2000), and could thus participate in the TGF-b1-induced ECM turnover and angiogenesis. Whereas inhibition of CLT formation was observed by TGF-b1 alone, in combina- tion with VEGF and 1, CLT formation is rescued and stabilized, suggesting that the process is regulated by the interaction of angiogenic and anti- angiogenic growth factors (Ramsauer and D’Amore, 2007). The process of angiogenesis is characterized by two different phases (Lebrin et al., 2005): an activation phase accompanied by increased vascular permeability, degra- dation of the basement membrane, endothelial cell proliferation and migration, and the late phase typified by the inhibition of endothelial cell proliferation and migration, reforming of the basement membrane and stabilization and maturation of the vessel. A recent Figure 7 Proposed mechanism of U87MG-CM and transforming model proposed that the activation phase of angiogen- growth factor (TGF)-b-mediated insulin-like growth factor-bind- esis is promoted by TGF-b1/ALK1/Smad-1/-5 signaling ing protein 7 (IGFBP7) induction in human brain endothelial cell (HBEC). U87MG-CM secretes several factors including TGF-bs. pathway, whereas the late phase is induced by TGF-b1/ TGF-bs bind to TGF-b type II receptor (TbR-II) at the surface of ALK5/Smad-2/-3 pathway (Goumans et al., 2002); endothelial cells, which in turn recruits TGF-b type I receptor given that both Smad-2 and Smad-1/-5 can be activated (TbR-I) ALK1 or ALK5. Activated TbR-II and TbR-I complexes by TGF-b in endothelial cells, the overall angiogenesis stimulate both Smad-1/-5 and Smad-2/-3 phosphorylation; outcome is likely dependent on the balance between IGFBP7 transcription in this system appears to be preferentially induced by Smad-2/-3 complexing with Smad-4 (bold arrows). TGF-b-induced ALK1 or ALK5 activation (Goumans ALK1-mediated phosphorylation of Smad-1/-5 could be secondary et al., 2002). As (1) endothelial IGFBP7 induction by to the activation of ALK5 (dash arrows). IGFBP7, in concert with tumor microenvironment is predominantly mediated by other TGF-b-induced genes/proteins, participates in promoting ALK5/Smad-2 pathway, (2) IGFBP7 accumulates in the late-phase angiogenesis. basal lamina of GBM vessels in vivo (Pen et al., 2007) and (3) IGFBP7 does not stimulate endothelial cell important in developing appropriate therapies for GBM proliferation, we propose that IGFBP7 might be tumors as well as other tumors characterized by high involved in the late phase of angiogenesis in the context vascular expression of IGFBP7 such as colon cancer of GBM tumors. Proposed mechanisms of GBM- and (Umeda et al., 1998; van Beijnum et al., 2006). TGF-b-mediated induction of IGFBP7 in HBEC are shown schematically in Figure 7. The proposed angiogenic role of IGFBP7 may appear Materials and methods contradictory to literature evidence of tumor-suppresor actions of IGFBP7 (Sprenger et al., 1999; Wilson et al., Reagents and antibodies 2002; Mutaguchi et al., 2003; Wajapeyee et al., 2008). TGF-b1 recombinant protein, TGF-b1 polyclonal antibody The first line of explanation is that IGFBP7 could (AF-101-NA) and pan-TGF-b-neutralizing monoclonal anti- trigger differential signaling pathways in tumor and body (1D11) were purchased from R&D System (Minneapolis, endothelial cells; for example, in melanocytes, BRAF MN, USA). The TGF-b type I receptor (ALK5) inhibitor, oncogene induces IGFBP7 that causes cellular senes- SB431542 was purchased from Sigma (St Louis, MO, USA). cence through inhibition of BRAF-MEK-ERK signal- Primary antibodies for western blot included mouse mono- clonal antibodies to Smad-2 (BD Biosciences, Bedford, MA, ing (Wajapeyee et al., 2008), whereas in endothelial cells USA) and to actin (Chemicon, Temecula, CA, USA), IGFBP7 is induced through TGF-b/ALK5/Smad-2 polyclonal antibody to P-Smad-2, P-Smad-1/-5 and Smad-5 and participates in late-stage angiogenesis. As there is (Cell Signaling, Beverly, MA, USA) and polyclonal antibody currently no clear understanding whether late-stage to IGFBP7 (R&D System). angiogenesis contributes to tumor progression or tumor growth stabilization, the overall effect of IGFBP7 Cell cultures in tumor milieu cannot be easily generalized. Further HBECs were obtained from small intracortical microvessels understanding of the regulation and functions of this (20–112 mm) harvested from temporal cortex from patients highly selective biomarker of tumor vessels will be treated surgically for idiopathic epilepsy as previously

Oncogene Tumor cells induce IGFBP7 in endothelium APenet al 6842 Table 1 Primers designed to amplify selected genes by Q-PCR Gene Sense primers for Q-PCR Antisense primers for Q-PCR

IGFBP7 50-GCGAGCAAGGTCCTTCCATA-30 50-GGGATTCCGATGACCTCACA-30 b-Actin 50-TGTCCACCTTCCAGCAGATGT-30 50-AGTCCGCCTAGAAGCATTTGC-30

Each set of forward and reverse primers was designed from NCBI published sequences using the Primer Express Software v.2.0.

described (Stanimirovic et al., 1996). Tissues were obtained Superscript II reverse transcriptase (Invitrogen, Burlington, with approval from the Office of Research Ethics Boards. ON, Canada). Quantitative Q-PCR reactions were performed HBEC were grown at 37 1C in media containing Earle’s salts, with iTaq SYBR Green Supermix (Bio-Rad, Mississauga, 25 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid, Ontario, Canada) using an ABI Prism 7700 Sequence Detector 4.35 g/l sodium bicarbonate, and 3 mML-glutamine, 10% fetal System (Applied Biosystems, Foster City, CA, USA). Primers bovine serum (FBS), 5% human serum, 20% of media used for real-time Q-PCR were designed according to conditioned by murine melanoma cells, 5 mg/ml insulin, sequences in GenBank (Table 1). The thermal PCR conditions 5 mg/ml transferrin, 5 ng/ml selenium and 10 mg/ml endothelial were 15 s denaturation at 95 1C and 1 min annealing extension cell growth supplement. HBEC cultures were routinely at 60 1C for 40cycles. Fluorescence was detected at the end of characterized for purity and the expression of endothelial cell each 60 1C phase. The transcript of IGFBP7 was normalized to markers (Stanimirovic et al., 1996). Passages 4–6 were used in the transcript levels of the housekeeping gene b-actin. this study. The human glioma cell lines U87MG and T98G were established from surgically removed type III glioma/GBM and Western blot GBM, respectively, and were both obtained from the Amer- HBEC were lysed in radioimmuno precipitation assay buffer ican Type Culture Collection. U87MG cells were grown at (1% IGEPAL, 0.5% deoxycholic acid, 0.1% SDS, protease 37 1C in DMEM supplemented with 100 U/ml penicillin, inhibitor capsule (Sigma)) and CM were collected. The cell 100 mg/ml streptomycin, 0.25 mg/ml amphotericin B with lysates were centrifuged at 20 000 g for 10min at 4 1C and 0.085 g/l NaCl and 10% heat-inactivated FBS (HyClone, protein content in the supernatant was measured using BCA Logan, UT, USA) in humidified atmosphere of 5% CO2/ protein assay (Pierce, Rockford, IL, USA). For detection of 95% air. T98G cells were cultured in minimum essential Smad proteins, nuclear extracts were isolated using a Nuclear medium Eagle (MEM; Sigma) supplemented with 10% FBS. Extraction Kit (Panomics, Redwood City, CA, USA) as Media for both cell lines were changed every 2–3 days. described by manufacturer’s protocol. Total protein per sample (10–20 mg) or equal volumes of CM were subjected to Preparation of U87MG- and T98G-conditioned media 12% SDS–PAGE; proteins were then transferred to U87MG cells and T98G cells (5  104 cells per ml) were plated Immobilon membrane (Millipore, Bedford, MA, USA). The in poly-L-lysine (25 mg/ml) precoated and noncoated 75 cm2 membrane was blocked with 5% skim milk in TBST (20mmol/ Falcon tissue culture flasks, respectively (Becton Dickinson l Tris–HCl (pH 8.0), 137 mmol/l NaCl and 0.1% Tween 20) Labware, Franklin Lakes, NJ, USA), and incubated at 37 1C for 1 h at room temperature (RT) and incubated with the in growth medium. U87MG and T98G growth media were primary antibodies overnight at 4 1C. After washing replaced with serum-free DMEM and MEM, respectively, 4 with TBST, blots were probed with HRP-conjugated second- days after plating. Serum-free media were collected after 3 ary antibodies (anti-mouse immunoglobulin G (IgG), anti- days and filtered (Millex-GV sterilizing filter membrane, rabbit IgG or anti-goat IgG) (Sigma) for 1 h at RT. 0.22 mm). Immunoblots were visualized using Amersham Biosciences enhanced chemiluminescence detection system (Piscataway, NJ, USA). ELISA The levels of secreted human TGF-b1, TGF-b2 and TGF-b3in serum-free CM from U87MG cells were determined using DuoSet ELISA Development System kits (R&D System) Capillary-like tube formation according to manufacturer’s protocol. In vitro angiogenesis was assessed by endothelial tube formation in growth factor reduced Matrigel (BD Biosciences). To allow the Matrigel solution to polymerize, 24-well plates Cell treatments were coated with 300 ml of diluted Matrigel solution and HBEC grown in 35 mm dishes were washed with DMEM and incubated at 37 1C for at least 1 h. HBEC (40 000 cells) were growth medium was replaced by DMEM alone or supple- harvested and suspended in one of the following: 500 ml mented with different concentration of TGF-b1 recombinant DMEM alone, DMEM supplemented with 10, 50 or 200 nM of protein, U87MG-CM either inactivated by boiling at 95 1C for recombinant IGFBP7, serum-free U87MG-CM alone or 10min or fresh, in the absence or presence of a pan-TGF- b- supplemented with 10 mM of SB431542. HBEC were then neutralizing antibody (1D11), an ALK5 inhibitor (SB431542) plated on top of the solidified Matrigel solution and or a TGF-b1 polyclonal antibody for various periods of time. maintained at 37 1C. Endothelial tube formation was digitally The treatments were refreshed at day 3. The expression of photographed under a phase contrast Olympus 1  50micro- IGFBP7 was evaluated by real-time Q-PCR and western blot scope (Olympus U-CMT) at  4 magnification, 24 h after analyses. treatment. Captured images were analysed using Northern Eclipse v.5.0software. Images were thresholded, converted to Quantitative real-time Q-PCR binary images and skeletonized. The total length of the CLT Total RNA from either untreated or treated HBEC was networks and the number of nodes (branching points) formed extracted using TRIzol Reagent (Gibco BRL, Gaithersburg, by HBEC were quantified. Experiments were performed with MD, USA) and reverse-transcribed into cDNA using at least three different HBEC isolations.

Oncogene Tumor cells induce IGFBP7 in endothelium APenet al 6843 Statistical analysis Acknowledgements Results are expressed as mean±s.e.m. The unpaired t-test was used for single comparisons of groups with equal variance and We thank Gilles St-Laurent for technical assistance for the normal distribution. Analysis of variance (ANOVA) followed production of IGFBP7, Mrs Debbie Callaghan for technical by Newman–Keuls posttest was used to compare multiple data assistance in nuclear protein extraction, Dr Anne Lenferink for to each group. valuable discussions and Mr Tom Devecseri for image processing.

References

Ahmed S, Jin X, Yagi M, Yasuda C, Sato Y, Higashi S et al. (2006). Goumans MJ, Lebrin F, Valdimarsdottir G. (2003). Controlling the Identification of membrane-bound serine proteinase matriptase as angiogenic switch: a balance between two distinct TGF-b receptor processing enzyme of insulin-like growth factor binding protein- signaling pathways. Trends Cardiovasc Med 13: 301–307. related protein-1 (IGFBP-rP1/angiomodulin/mac25). FEBS J 273: Goumans MJ, Valdimarsdottir G, Itoh S, Rosendahl A, Sideras P, 615–627. ten Dijke P. (2002). Balancing the activation state of the Akaogi K, Okabe Y, Funahashi K, Yoshitake Y, Nishikawa K, endothelium via two distinct TGF-beta type I receptors. EMBO J Yasumitsu H et al. (1994). Cell adhesion activity of a 30-kDa major 21: 1743–1753. secreted protein from human bladder carcinoma cells. Biochem Hata Y, Clermont A, Yamauchi T, Pierce EA, Suzuma I, Kagokawa H Biophys Res Commun 198: 1046–1053. et al. (2000). Retinal expression, regulation, and functional Akaogi K, Okabe Y, Sato J, Nagashima Y, Yasumitsu H, Sugahara K bioactivity of prostacyclin-stimulating factor. J Clin Invest 106: et al. (1996). Specific accumulation of tumor-derived adhesion factor 541–550. in tumor blood vessels and in capillary tube-like structures of Hirano A, Matsui T. (1975). Vascular structures in brain tumors. Hum cultured vascular endothelial cells. Proc Natl Acad Sci USA 93: Pathol 6: 611–621. 8384–8389. Hwa V, Oh Y, Rosenfeld RG. (1999). The insulin-like growth factor- Beaty RM, Edwards JB, Boon K, Siu IM, Conway JE, Riggins GJ. binding protein (IGFBP) superfamily. Endocr Rev 20: 761–787. (2007). PLXDC1 (TEM7) is identified in a genome-wide expression Inman GJ, Nicolas FJ, Callahan JF, Harling JD, Gaster LM, screen of glioblastoma endothelium. J Neurooncol 81: 241–248. Reith AD et al. (2002). SB-431542 is a potent and specific inhibitor Blanchette F, Rivard N, Rudd P, Grondin F, Attisano L, Dubois CM. of transforming growth factor-beta superfamily type I activin (2001). Cross-talk between the p42/p44 MAP kinase and Smad receptor-like kinase (ALK) receptors ALK4, ALK5, and ALK7. pathways in transforming growth factor beta 1-induced furin gene Mol Pharmacol 62: 65–74. transactivation. J Biol Chem 276: 33986–33994. Keeton MR, Curriden SA, van Zonneveld AJ, Loskutoff DJ. (1991). Burger AM, Zhang X, Li H, Ostrowski JL, Beatty B, Venanzoni M Identification of regulatory sequences in the type 1 plasminogen et al. (1998). Down-regulation of T1A12/mac25, a novel insulin-like activator inhibitor gene responsive to transforming growth factor growth factor binding protein related gene, is associated with disease beta. J Biol Chem 266: 23048–23052. progression in breast carcinomas. Oncogene 16: 2459–2467. Kishibe J, Yamada S, Okada Y, Sato J, Ito A, Miyazaki K et al. Chen Y, Pacyna-Gengelbach M, Ye F, Knosel T, Lund P, (2000). Structural requirements of heparan sulfate for the binding to Deutschmann N et al. (2007). Insulin-like growth factor binding the tumor-derived adhesion factor/angiomodulin that induces cord- protein-related protein 1 (IGFBP-rP1) has potential tumour- like structures to ECV-304 human carcinoma cells. J Biol Chem 275: suppressive activity in human lung cancer. J Pathol 211: 431–438. 15321–15329. Chen YG, Massague J. (1999). Smad1 recognition and activation by Kleihues P, Soylemezoglu F, Schauble B, Scheithauer BW, Burger PC. the ALK1 group of transforming growth factor-beta family (1995). Histopathology, classification, and grading of gliomas. Glia receptors. J Biol Chem 274: 3672–3677. 15: 211–221. Constam DB, Philipp J, Malipiero UV, ten Dijke P, Schachner M, Kniesel U, Wolburg H. (2000). Tight junctions of the blood–brain Fontana A. (1992). Differential expression of transforming growth barrier. Cell Mol Neurobiol 20: 57–76. factor-beta 1, -beta 2, and -beta 3 by glioblastoma cells, astrocytes, Komatsu S, Okazaki Y, Tateno M, Kawai J, Konno H, Kusakabe M and microglia. J Immunol 148: 1404–1410. et al. (2000). Methylation and downregulated expression of mac25/ Dai C, Holland EC. (2001). Glioma models. Biochim Biophys Acta insulin-like growth factor binding protein-7 is associated with liver 1551: M19–M27. tumorigenesis in SV40T/t antigen transgenic mice, screened by Davies DC. (2002). Blood–brain barrier breakdown in septic restriction landmark genomic scanning for methylation (RLGS-M). encephalopathy and brain tumours. J Anat 200: 639–646. Biochem Biophys Res Commun 267: 109–117. de Caestecker M. (2004). The transforming growth factor-beta Kwak HJ, Park MJ, Cho H, Park CM, Moon SI, Lee HC et al. (2006). superfamily of receptors. Cytokine Growth Factor Rev 15: 1–11. Transforming growth factor-beta1 induces tissue inhibitor of Deb-Rinker P, Ly D, Jezierski A, Sikorska M, Walker PR. (2005). metalloproteinase-1 expression via activation of extracellular Sequential DNA methylation of the Nanog and Oct-4 upstream signal-regulated kinase and Sp1 in human fibrosarcoma cells. Mol regions in human NT2 cells during neuronal differentiation. J Biol Cancer Res 4: 209–220. Chem 280: 6257–6260. Lebrin F, Deckers M, Bertolino P, Ten Dijke P. (2005). TGF-beta Derynck R, Zhang Y, Feng XH. (1998). Smads: transcriptional receptor function in the endothelium. Cardiovasc Res 65: 599–608. activators of TGF-beta responses. Cell 95: 737–740. Lin J, Lai M, Huang Q, Ma Y, Cui J, Ruan W. (2007). Methylation Dinda AK, Sarkar C, Roy S, Kharbanda K, Mathur M, Khosla AK patterns of IGFBP7 in colon cancer cell lines are associated with et al. (1993). A transmission and scanning electron microscopic levels of . J Pathol 212: 83–90. study of tumoral and peritumoral microblood vessels in human Long DM. (1970). Capillary ultrastructure and the blood–brain barrier gliomas. J Neurooncol 16: 149–158. in human malignant brain tumors. J Neurosurg 32: 127–144. Durocher Y, Perret S, Kamen A. (2002). High-level and high- Louis DN. (2006). Molecular pathology of malignant gliomas. Annu throughput recombinant protein production by transient transfec- Rev Pathol 1: 97–117. tion of suspension-growing human 293-EBNA1 cells. Nucleic Acids Madden SL, Cook BP, Nacht M, Weber WD, Callahan MR, Jiang Y Res 30: E9. et al. (2004). Vascular gene expression in nonneoplastic and Engel ME, McDonnell MA, Law BK, Moses HL. (1999). Inter- malignant brain. Am J Pathol 165: 601–608. dependent SMAD and JNK signaling in transforming growth Massague J. (1998). TGF-beta signal transduction. Annu Rev Biochem factor-beta-mediated transcription. J Biol Chem 274: 37413–37420. 67: 753–791.

Oncogene Tumor cells induce IGFBP7 in endothelium APenet al 6844 Miller SA, Dykes DD, Polesky HF. (1988). A simple salting out Sprenger CC, Damon SE, Hwa V, Rosenfeld RG, Plymate SR. (1999). procedure for extracting DNA from human nucleated cells. Nucleic Insulin-like growth factor binding protein-related protein 1 (IGFBP- Acids Res 16: 1215. rP1) is a potential tumor suppressor protein for prostate cancer. Moreno MJ, Ball M, Andrade MF, McDermid A, Stanimirovic DB. Cancer Res 59: 2370–2375. (2006). Insulin-like growth factor binding protein-4 (IGFBP-4) is a St Croix B, Rago C, Velculescu V, Traverso G, Romans KE, novel anti-angiogenic and anti-tumorigenic mediator secreted by Montgomery E et al. (2000). Genes expressed in human tumor dibutyryl cyclic AMP (dB-cAMP)-differentiated glioblastoma cells. endothelium. Science 289: 1197–1202. Glia 53: 845–857. Stanimirovic D, Morley P, Ball R, Hamel E, Mealing G, Durkin JP. Murphy M, Pykett MJ, Harnish P, Zang KD, George DL. (1993). (1996). Angiotensin II-induced fluid phase endocytosis in human Identification and characterization of genes differentially expressed cerebromicrovascular endothelial cells is regulated by the inositol- in meningiomas. Cell Growth Differ 4: 715–722. phosphate signaling pathway. J Cell Physiol 169: 455–467. Mutaguchi K, Yasumoto H, Mita K, Matsubara A, Shiina H, Igawa ten Dijke P, Hansen P, Iwata KK, Pieler C, Foulkes JG. (1988). M et al. (2003). Restoration of insulin-like growth factor binding Identification of another member of the transforming growth factor protein-related protein 1 has a tumor-suppressive activity through type beta gene family. Proc Natl Acad Sci USA 85: 4715–4719. induction of apoptosis in human prostate cancer. Cancer Res 63: Umeda F, Ono Y, Sekiguchi N, Hashimoto T, Masakado M, 7717–7723. Nakamura K et al. (1998). Increased mRNA expression of a novel Nagakubo D, Murai T, Tanaka T, Usui T, Matsumoto M, Sekiguchi prostacyclin-stimulating factor in human colon cancer. J Gastro- K et al. (2003). A high endothelial venule secretory protein, enterol 33: 213–217. mac25/angiomodulin, interacts with multiple high endothelial Vajkoczy P, Goldbrunner R, Farhadi M, Vince G, Schilling L, venule-associated molecules including chemokines. J Immunol 171: Tonn JC et al. (1999). Glioma cell migration is associated with 553–561. glioma-induced angiogenesis in vivo. Int J Dev Neurosci 17: 557–563. Oh Y, Nagalla SR, Yamanaka Y, Kim HS, Wilson E, Rosenfeld RG. Vajkoczy P, Menger MD. (2000). Vascular microenvironment in (1996). Synthesis and characterization of insulin-like growth gliomas. J Neurooncol 50: 99–108. factor-binding protein (IGFBP)-7. Recombinant human mac25 van Beijnum JR, Dings RP, van der Linden E, Zwaans BM, protein specifically binds IGF-I and -II. J Biol Chem 271: Ramaekers FC, Mayo KH et al. (2006). Gene expression of tumor 30322–30325. angiogenesis dissected: specific targeting of colon cancer angiogenic Oldendorf WH, Cornford ME, Brown WJ. (1977). The large apparent vasculature. Blood 108: 2339–2348. work capability of the blood–brain barrier: a study of the Vaupel P, Thews O, Kelleher DK, Hoeckel M. (1998). Oxygenation mitochondrial content of capillary endothelial cells in brain and of human tumors: the Mainz experience. Strahlenther Onkol 174 other tissues of the rat. Ann Neurol 1: 409–417. (Suppl 4): 6–12. Olofsson A, Miyazono K, Kanzaki T, Colosetti P, Engstrom U, Wajapeyee N, Serra RW, Zhu X, Mahalingam M, Green MR. (2008). Heldin CH. (1992). Transforming growth factor-beta 1, -beta 2, and Oncogenic BRAF induces senescence and apoptosis through path- -beta 3 secreted by a human glioblastoma cell line. Identification of ways mediated by the secreted protein IGFBP7. Cell 132: 363–374. small and different forms of large latent complexes. J Biol Chem 267: Watabe T, Nishihara A, Mishima K, Yamashita J, Shimizu K, 19482–19488. Miyazawa K et al. (2003). TGF-beta receptor kinase inhibitor Pen A, Moreno MJ, Martin J, Stanimirovic DB. (2007). Molecular enhances growth and integrity of embryonic stem cell-derived markers of extracellular matrix remodeling in glioblastoma vessels: endothelial cells. J Cell Biol 163: 1303–1311. microarray study of laser-captured glioblastoma vessels. Glia 55: Weller M, Malipiero U, Groscurth P, Fontana A. (1995). T cell 559–572. apoptosis induced by -2 deprivation or transforming Pepper MS. (1997). Transforming growth factor-beta: vasculogenesis, growth factor-beta 2: modulation by the phosphatase inhibitors angiogenesis, and vessel wall integrity. Cytokine Growth Factor Rev okadaic acid and calyculin A. Exp Cell Res 221: 395–403. 8: 21–43. Wick W, Platten M, Weller M. (2001). Glioma cell invasion: regulation Poncelet AC, Schnaper HW. (1998). Regulation of human mesangial of metalloproteinase activity by TGF-beta. J Neurooncol 53: cell collagen expression by transforming growth factor-beta1. Am J 177–185. Physiol 275: F458–F466. Wilson HM, Birnbaum RS, Poot M, Quinn LS, Swisshelm K. (2002). Ramsauer M, D’Amore PA. (2007). Contextual role for Insulin-like growth factor binding protein-related protein 1 inhibits and TGF-beta1 in blood vessel stabilization. J Cell Sci 120: proliferation of MCF-7 breast cancer cells via a senescence-like 1810–1817. mechanism. Cell Growth Differ 13: 205–213. Reed MJ, Vernon RB, Abrass IB, Sage EH. (1994). TGF-beta 1 Yakymovych I, Ten Dijke P, Heldin CH, Souchelnytskyi S. (2001). induces the expression of type I collagen and SPARC, and enhances Regulation of Smad signaling by protein kinase C. FASEB J 15: contraction of collagen gels, by fibroblasts from young and aged 553–555. donors. J Cell Physiol 158: 169–179. YamadaN,KatoM,YamashitaH,NisterM,MiyazonoK,HeldinCH Ruan W, Xu E, Xu F, Ma Y, Deng H, Huang Q et al. (2007). IGFBP7 et al. (1995). Enhanced expression of transforming growth factor-beta plays a potential tumor suppressor role in colorectal carcinogenesis. and its type-I and type-II receptors in human glioblastoma. Int J Cancer Biol Ther 6: 354–359. Cancer 62: 386–392. Sato J, Hasegawa S, Akaogi K, Yasumitsu H, Yamada S, Sugahara K Yamauchi T, Umeda F, Masakado M, Isaji M, Mizushima S, et al. (1999). Identification of cell-binding site of angiomodulin Nawata H. (1994). Purification and molecular cloning of prostacyclin- (AGM/TAF/Mac25) that interacts with heparan sulfates on cell stimulating factor from serum-free conditioned medium of human surface. J Cell Biochem 75: 187–195. diploid fibroblast cells. Biochem J 303(Part 2): 591–598.

Supplementary Information accompanies the paper on the Oncogene website (http://www.nature.com/onc)

Oncogene