Hedgehog Promotes Neovascularization in Pancreatic Cancers by Regulating Ang-1 and IGF-1 Expression in Bone-Marrow Derived Pro-Angiogenic Cells

Background The hedgehog (Hh) pathway has been implicated in the pathogenesis of cancer including pancreatic ductal adenocarcinoma (PDAC). Recent studies have suggested that the oncogenic function of Hh in PDAC involves signaling in the stromal cells rather than cell autonomous effects on the tumor cells. However, the origin and nature of the stromal cell type(s) that are responsive to Hh signaling remained unknown. Since Hh signaling plays a crucial role during embryonic and postnatal vasculogenesis, we speculated that Hh ligand may act on tumor vasculature specifically focusing on bone marrow (BM)-derived cells. Methodology/Principal Findings Cyclopamine was utilized to inhibit the Hh pathway in human PDAC cell lines and their xenografts. BM transplants, co-culture systems of tumor cells and BM-derived pro-angiogenic cells (BMPCs) were employed to assess the role of tumor-derived Hh in regulating the BM compartment and the contribution of BM-derived cells to angiogenesis in PDAC. Cyclopamine administration attenuated Hh signaling in the stroma rather than in the cancer cells as reflected by decreased expression of full length Gli2 protein and Gli1 mRNA specifically in the compartment. Cyclopamine inhibited the growth of PDAC xenografts in association with regression of the tumor vasculature and reduced homing of BM-derived cells to the tumor. Host-derived Ang-1 and IGF-1 mRNA levels were downregulated by cyclopamine in the tumor xenografts. In vitro co-culture and matrigel plug assays demonstrated that PDAC cell-derived Shh induced Ang-1 and IGF-1 production in BMPCs, resulting in their enhanced migration and capillary morphogenesis activity. Conclusions/Significance We identified the BMPCs as alternative stromal targets of Hh-ligand in PDAC suggesting that the tumor vasculature is an attractive therapeutic target of Hh blockade. Our data is consistent with the emerging concept that BM-derived cells make important contributions to epithelial tumorigenesis.


Introduction
Pancreatic ductal adenocarcinoma (PDAC) is the fourth leading cause of cancer-related death in the United States and fifth in Japan, and its overall 5-year survival rate is only 5.0-9.7% [1,2]. Gemcitabine-based chemotherapy for locally advanced or metastatic PDAC has only modest activity with a small survival benefit [3]. The identification of new molecular targets for PDAC to overcome the dismal prognosis is therefore necessary. Thus far, a range of targeted therapies against EGFR [4], Ras/MEK [5] and VEGF [6] have failed to improve survival significantly in clinical trials.
Aberrant activation of the hedgehog (Hh) pathway has more recently been recognized as one of the mediators of PDAC development and is therefore considered to be a promising target for therapy [7,8]. Hh is a morphogen required for proper pattern formation during embryogenesis. Elevated expression of Hhrelated proteins, sonic hedgehog (Shh) or indian hedgehog (Ihh), impairs pancreatic morphogenesis with increases in mesenchyme and decreases in the epithelial compartment [9], indicating that tight regulation of the overall level of Hh activity is crucial for normal pancreatic development. Since Shh is misexpressed not only in PDAC but also in the precursor lesions pancreatic intraepithelial neoplasm (PanIN) [8] and intraductal papillary mucinous neoplasia [10], dysregulation of this pathway may play a role during early stages of tumorigenesis. A number of Hh-target genes, such as Bmp [11], FOXM1 [12], and PMP22 [13], are overexpressed in PDAC, and there is cross-talk between oncogenic pathways including MAPK/ERK, PI3K/Akt [14], Wnt [15], TGFb/BMP [16] and Hh signaling. Among the molecules that participate in the Hh signaling, Smoothened (Smo) has been recognized as a key mediator of the signaling; it is one of the receptor components for Hh ligand and converts Gli2 into a transcriptional activator [17]. Thus, Smo is a particularly promising target for anti-cancer therapy [18]. Cyclopamine, a natural steroid alkaloid, has strong inhibitory effects against Smo [18] and treatment with cyclopamine in the experimental setting inhibits growth of many cancers [19,20,21]. Cyclopamine derivatives with various structural motifs have been generated and show promise for possible clinical use [22].
Specifically, several studies have demonstrated significant growth inhibition of PDAC in vivo by Hh pathway blockade [7,23]. The tumor growth inhibitory mechanism has been thought to be primarily mediated through blocking an autocrine loop that might be required for proliferation [7], survival [8] and motility [21] of cancer cells. However, Smo deletion in the pancreatic epithelium does not impair tumorigenesis in a mouse model of PDAC [24], and Hh ligands do not induce transcriptional reporters of the pathway in PDAC lines. Moreover, transgenic expression of an activated Smo allele in the pancreatic epithelium fails to induce Smo pathway target genes or to induce neoplastic change [25]. Consequently, it has been speculated that Hh signaling might play a paracrine role within the tumor microenvironment, in epithelial cancers that lack mutations in Hh signaling components. Indeed, studies performed by us and others have defined crucial roles of cancer cell derived Hh on cell types within tumor stromal compartments [26,27,28]. Despite these findings, the precise mechanisms by which some tumor cells are insensitive to Hh inhibition and the stromal cell type(s) that are responsive to Hh signaling remained unknown. We therefore undertook studies to elucidate these mechanisms.

Methods
Cell Culture, Protein Analysis, and RNA Analysis Human PDAC cell lines were obtained from ATCC, Health Science Research Resources Bank (Osaka, Japan) and Cell Resource Center for Biochemical Research (Sendai, Japan). BMMNC isolation, proliferation assays, co-culture, migration assays, immunoblotting, and qPCR are described in Methods S1. Primer sequences are summarized in Table S1.

Animals, Bone Marrow Transplantation (BMT) and Immunohistochemistry
Therapeutic studies with cyclopamine were performed with nude mice xenografts as described [7]. BMT procedures are provided in Methods S1. Protocols for animal experiments were approved by the Asahikawa Medical Collage Institutional Animal Care and Use Committee.
Tissues were processed as reported previously [26]. Antibodies and conditions for immunohistochemistry are presented in Methods S1.

Cellular, Molecular, and Statistical Analyses
Detailed descriptions of procedures are provided in Methods S1.

In Vivo Effects of Cyclopamine on KP-1N Xenografts
In order to elucidate the mechanisms by which Hh signaling might support tumor progression in a tumor cell nonautonomous fashion, we evaluated the role of Hh signaling on growth of PDAC in vivo. We selected a cell line, KP-1N, that exhibited high expression of Shh but was insensitive to growth inhibition by cyclopamine in vitro (Figure S1, S2 and Table  S2). Treatment of KP-1N xenografts with cyclopamine daily for 7 days caused a 46-percent reduction in tumor weight as compared with vehicle control-treated tumors ( Figure 1A). Histopathological analysis revealed that cyclopamine treated tumors had abundant areas of necrosis resulting in a 59% reduction in viable tumor weight. In line with these findings, staining for Ki-67 and TUNEL demonstrated that cyclopamine treatment reduced proliferation by 14% and increased cell death by 9% in vivo ( Figure 1B); in contrast, there was no significant effect of cyclopamine on proliferation/cell death kinetics in vitro ( Figure S2).
In order to determine which cells were affected by cyclopamine treatment, we measured expression of Ptch1 and Gli1 mRNA, transcriptional targets of Hh signaling, in both the xenografted human tumor cells as well as the mouse stromal compartment. To do so, we made use of species-specific probes. We observed that human Ptch1 and Gli1 in xenograft tissues were only modestly decreased by cyclopamine treatment ( Figure 1C). Comparable, minor decreases in expression were observed in vitro ( Figure S3). In comparison, mouse Ptch1 and Gli1 mRNA levels showed a markedly more pronounced downregulation, indicating that stromal cells are significant targets of Hh blockade. Collectively, these results indicate that the anti-tumor effect of cyclopamine on xenograft growth is primarily mediated through a downregulation of Hh signaling in tumor-associated stroma.

Effects of Cyclopamine on the Tumor Vasculature
Among the stromal components in PDAC xenografts, we were particularly curious about the tumor vasculature, since we have demonstrated that BM-derived pro-angiogenic (precursor) cells are responsive to Hh ligand during neovascularization [26,29]. Therefore, we investigated the effect of cyclopamine on the tumor vasculature in KP-1N xenografts. Immunostaining for the endothelial cell marker CD31 revealed that microvascular density was significantly reduced (42.6%) when mice were treated with cyclopamine ( Figure 2A). The resulting tumor vasculature was narrowed and fragmented, and the reduction in mean vessel area was dramatic (64.8%). The inhibition of angiogenesis by cyclopamine was also observed in Suit-2 xenografts ( Figure S4).
Our results suggested that blockade of Hh signaling ''destabilizes'' the tumor vessels. We wished to confirm this hypothesis by immunohistochemical staining for a-smooth muscle actin (SMA), NG2, and CD31. SMA and NG2 are expressed by pericytes, which are cells that closely associate with CD31-positive endothelial cells to stabilize blood vessels [30,31]. We observed 50.0% or 62.1% fewer CD31 + microvessels covered with SMA + or NG2 + pericytes in the cyclopamine treated tumors ( Figure 2B). Thus, Hh signaling is important for maturation and/or stabilization of the tumor vasculature.

Tumor-Derived Hh Promotes Incorporation of BM-Derived Cells into the Neovasculature
Bone marrow (BM)-derived cells are thought to play a role in tumor development [32,33]. For instance, various types of BMderived hematopoietic cells have been observed to closely associate with the tumor neovasculature [34,35], and indeed, a small number of BM-derived progenitor cells were demonstrated to incorporate into the lumen of a growing vasculature where they differentiate into endothelial cells in a mouse lung cancer model [36]. Various factors in the tumor microenvironment can modulate the recruitment and retention of BM-derived cells during the development of the tumor vasculature [37]. Therefore, we sought to determine whether Hh signaling contributes to this process in PDAC. In order to trace BM-derived cells, we implanted KP-1N xenografts in chimeric mice with GFP-labeled BM (GFP-bone marrow transplant model; see Methods). Cyclopamine significantly decreased the recruitment of BM-derived cells into the xenografts by 60% as compared with control tumors ( Figure 3A).
Among Gli family of transcription factors, control of the degradation of Gli2 and the processing into its repressor form define the Hh signaling response [38]. Gli2 protein is processed and degraded in the absence of Hh ligand, whereas the presence of Hh ligand promotes Smo enrichment in primary cilia [39], resulting in nuclear translocation of full length, transcriptionally competent Gli2 protein [40]. We therefore evaluated the effect of cyclopamine on Gli2 protein via immunofluorescence analysis utilizing an antibody against full length form of Gli2 [41] ( Figure  S5A). BM-derived cells express full length Gli2 protein and cyclopamine decreased both the absolute number and the relative proportion of Gli2 positive BM-derived cells ( Figure 3B), indicating that Hh signaling is activated in BM-derived cells in a cyclopamine-sensitive (Smo-dependent) manner. Ptch1 protein was also observed in BM-derived cells as well as in cancer cells and cyclopamine reduced the Ptch1 expression in stromal compartment ( Figure S5B). We failed to demonstrate strong Gli1 protein expression in the xenograft tissue (data not shown).
In control tumors, GFP-positive BM-derived cells are closely associated with CD31 + tumor vessels but they are not incorporated to the lumen of the vasculatures ( Figure 3C). On the other hand, a subset of BM-derived cells co-expressed VE-cadherin demonstrating that BM-derived cells are not only recruited to the perivascular area but also give rise directly to the tumor endothelium. The NG2 proteoglycan is expressed by nascent pericytes during the early stages of angiogenesis, and a subset of BM-derived cells in xenograft tumors co-expressed NG2. These results indicate that BM-derived cells contribute to the development of tumor vasculature not only by giving rise to specific type of cells but also by associating in different manners. Of note, cyclopamine treatment significantly decreased the integration of the GFP + BMderived pro-angiogenic cells (BMPCs) into CD31 + , VE-cadherin + , or NG2 + tumor vasculatures. We confirmed that BMPCs were responsive to Hh since in vitro co-cultures with KP-1N tumor cells resulted in a cyclopamine-sensitive induction of Gli1/Ptch1 mRNA in BMPCs, but not in the mature endothelial cell line MS-1 ( Figure S6 and data not shown). Taken together, these results indicate that the Hh pathway plays a role in neovascularization by modulating the incorporation of BM-derived proangiogenic (precursor) cells into the tumor vessels.

Tumor-Derived Hh Promotes Migration of BM-Derived Pro-Angiogenic Cells In Vitro and In Vivo
We speculated that the decreased numbers of BM-derived cells in the cyclopamine-treated tumors might be the result of attenuated mobilization of cells from the BM. However, the number of circulating BMPCs, as measured by labeling cultured peripheral blood MNCs with acLDL/isolectin B4 [42], was not significantly reduced in the presence of cyclopamine ( Figure S6). This finding suggests that Hh might directly regulate recruitment, differentiation, or migration of precursors for BMPCs within the tumor microenvironment rather than affecting mobilization of the precursors from the marrow.
Since inhibition of Hh signaling did not affect mobilization of the BMPCs, we sought to investigate the mechanisms by which Hh signaling might enhance their incorporation into the tumor vasculature. To do so, we examined the effects of Hh signaling on migration of the precursors. The migration of BMPCs through a transwell membrane in vitro was measured in the presence of KP-1N in the lower well. Cultured BMMNCs expressing c-Kit, CD11b and VE-cadherin were utilized as the precursors for BMPCs ( Figure S6). The number of the BMPCs migrating to the bottom side of the membrane was significantly increased when cocultured with KP-1N as compared with the pro-angiogenic cells without KP-1N, implying that secreted factors from the PDAC cells can enhance the migration of the precursors ( Figure 4A). The role of tumor-derived Shh in migration of the precursors was also confirmed by using a Shh neutralizing antibody ( Figure 4C).
In order to elucidate the effects of Smo pathway modulation on migration of the BMPCs in vivo, we performed matrigel plug assays. When concentrated supernatant from KP-1N cells mixed with matrigel was implanted subcutaneously into nude mice, the number of host cells migrating into the matrigel was significantly increased, and the cells formed a cord-like structure, mimicking capillary morphogenesis in vivo ( Figure 5A). In order to distinguish BM-derived cells from other stromal cells, the experiment was repeated by utilizing other chimeric mice that underwent BMT. Since we observed significant Tie2 mRNA expression in the BM- derived pro-angiogenic cells utilized in in vitro assays (data not shown), Tie2/LacZ mice were used as donor for the BMT [42]. A large fraction of the cells within matrigel originated from BM as reflected by positive immunostaining for b-gal ( Figure 5B). The migration of host cells into matrigel plugs was attenuated significantly when cyclopamine was added to the matrigel. A similar result was also obtained using an anti-Shh antibody (data not shown).
Collectively, our results demonstrate that inhibition of tumorderived Hh signaling leads to inhibition of tumor angiogenesis and destabilization of vasculature with a concomitant decrease in recruitment of BMPCs into the tumor stroma.

Regulation of Pro-and Anti-Angiogenic Factors by Hh Signaling
To identify the molecular basis of the vascular disruption by Hh blockade, we quantified the mRNA expression levels of pro-and anti-angiogenic factors in PDAC xenografts using species-specific primer/probe sets for qPCR. Specifically, we were curious to know the levels of VEGF and SDF-1, which are considered proangiogenic factors due to their demonstrated effects on the recruitment and retention of the pro-angiogenic precursor cells at sites of tumor formation [34,43].
We found that human VEGF and SDF-1 mRNA expression were decreased by 31% and 33%, respectively, in the cyclopamine-treated tumors ( Figure S7). These results are in contrast to the modest downregulation of VEGF and SDF-1 mRNA in cyclopamine-treated KP-1N cells in vitro, either in normoxic or hypoxic conditions ( Figure S8). Likewise, HIF-1a stability was not affected under hypoxic conditions in vitro. Moreover, blocking Shh by neutralizing antibody did not cause a reduction in VEGF mRNA in KP-1N cells, suggesting that autocrine Shh signaling does not impart an angiogenic phenotype directly on the KP-1N ( Figure S8). Instead, these collective results suggest that Hh signaling impinges upon the stroma and that stromal-derived factors might contribute to the expression of VEGF and SDF-1 in the tumor cells in vivo.
It has been demonstrated that cancer associated fibroblasts secrete SDF-1 to recruit bone marrow-derived endothelial precursor cells into the tumor stroma in order to mediate angiogenesis [44]. Therefore, we next quantified the expression of these pro-/anti-angiogenic factors by the murine tumor stroma. Mouse VEGF and SDF-1 levels were not affected by cyclopamine in vivo ( Figure 6A), indicating that Hh is not involved in the modulation of these pro-angiogenic factors by tumor stroma.
We next examined the expression of factors that modulate vascular stability. In the tumors of cyclopamine-treated mice, we noted a marked reduction in murine Ang-1, IGF-1, and PDGF-B mRNA, factors known to stabilize the neovasculature. Ang-1 plays a crucial role during angiogenesis to stabilize neovessels [30], and its cognate Tie2 receptor is expressed in ECs, pro-angiogenic monocytes as well as pericyte precursors of mesenchymal origin [45]. The reduction of Ang-1 was coupled with 86.3% upregulation of Ang-2, a natural antagonist for Tie2, that leads to dissociation of the mural cell coating to initiate angiogenesis [46] ( Figure S9). In an effort to understand the regulation of these factors by Hh signaling, we compared their mRNA levels in BMPCs that were co-cultured with KP-1N cells. Ang-1 and IGF-1 but not PEDF-B mRNA was upregulated by more than 2-fold in a cyclopamine-sensitive manner ( Figure 6B), and simultaneous downregulation of Ang-2 mRNA in the precursors was observed. The altered expression of Ang-1 and Ang-2 was canceled by cyclopamine, and the effect on Ang-1/Ang-2 ratio was dramatic (54.3% reduction; Figure S9). Consistent results were observed when we targeted Smo in the BM-derived precursors by lentiviral shRNA, confirming that the alteration of Ang-1 gene expression by cyclopamine was due to blockade of Smo-dependent pathway ( Figure 6C). Thus, Hh-dependent Ang-1/Tie2 signaling appears to be one of the potential mechanisms by which BMPCs stabilize neovessels.

IGF-1 Is a Target of Hh Signaling in BM-Derived Pro-Angiogenic Cells
The Reduction in IGF-1 mRNA that we observed in cyclopamine treated xenografts ( Figure 6A) was also recapitulated in the co-culture assays; i.e., the strong induction of IGF-1 in BMPCs by KP-1N-derived factors was blocked significantly either by cyclopamine or shRNA against Smo ( Figure 6B, C). To verify the role of IGF-1 during neovascularization, a tube formation assay was performed using MS-1 (mouse mature EC line) with GFP-labeled BM-derived pro-angiogenic precursors ( Figure 6D). Capillary morphogenesis by MS-1 was induced by the BMPCs, and the effect was enhanced when they were preconditioned in the presence of KP-1N cells. Strikingly, the effect on capillary formation was significantly reduced upon treatment with an anti-IGF-1 neutralizing antibody. Thus, IGF-1 is an important target of Hh signaling in BM-derived cells, and the downregulation of IGF-1 in tumor xenografts by cyclopamine treatment could contribute to the disruption of angiogenesis.

Discussion
In the current study, we identified the BM-derived proangiogenic cells (BMPCs) as one of the paracrine targets of Hh, which plays a critical role to promote tumor angiogenesis and tumor progression. The destabilized tumor vasculature that we observed upon cyclopamine treatment first lead us to speculate that we might observe a reduction in stromal Ang-1 and IGF-1, factors known to be regulated by Hh pathway [47,48].
Ang-1 develops and stabilizes the primitive vasculature under Hh signaling in lung branching morphogenesis [49] and Shh increased the mRNA levels of Ang-1 in fibroblasts, but decreased Ang-2 [47,50]. Our co-culture experiment demonstrated Ang-1 mRNA in BMPCs was upregulated by soluble factors from KP-1N in a Smo-dependent manner, suggesting a role of the precursors as reservoir of Ang-1 in the tumor microenvironment to stabilize Tie2-expressing ECs as well as subsets of pro-angiogenic cells. Indeed, a certain fraction of BM-derived cells also express Tie-2 receptor [51], supporting an alternative autocrine or paracrine effect on the their angiogenic activity by the surrounding stroma via Ang-1 ligand.
There are various cellular sources of IGF-1 in the tumor stroma, and the reduction in stroma-derived IGF-1 that we observed may explain the growth inhibition of PDAC cells in vivo when Hh signaling was blocked [27]. IGF-1 can also enhance the activity of BM-derived cells to promote neovascularization through PI3K/ Akt signaling [52]. In the current study, the enhancement of capillary morphogenesis of MS-1 cells by the BMPCs stimulated with PDAC cells was clearly attenuated by anti-IGF-1 ( Figure 5D). Thus, IGF-1 appears to be one of the crucial targets of Hh signaling in the BMPCs and mediates its effect on neovascularization. There was a significant reduction in PDGF-B expression in the tumor stroma in vivo as well as in BMPCs in vitro; however, our co-culture assay of BMPCs with KP-1N did not demonstrate upregulation of PDGF-B mRNA, suggesting that cancer-derived Hh ligand is unlikely to play a major role in the induction of PDGF-B. However, the downregulation of PDGF-B by cyclopamine may partially account for the decreased recruitment of pericytes into newly formed vascular network in xenografts.
The molecular mechanisms that we uncovered here clearly demonstrate that Hh signaling plays a paracrine role in tumor progression. It was recently reported that transcriptional induction of Gli can be maintained independent of Smo [24], and the noncanonical regulation of Gli-target genes is mediated in part through TGF-b and Kras signaling. Indeed, most PDAC cells, including KP-1N, express oncogenic Kras and it has been demonstrated that TGF-b-Smad pathway is activated in KP-1N cells [53]. These pathways might account for the modest reduction of Gli1 transcription by cyclopamine that we observed. It is also possible that the in vivo growth inhibition by Hh blockade that we observed might be mediated through an effect on self-renewal property of rare PDAC initiating cells [54]. However, the significant change in proliferation/death kinetics that we observed in a large proportion of cancer cells in the xenografts treated with cyclopaminewas clearly accompanied by the inhibition of angiogenesis.
Cyclopamine attenuated not only the recruitment of BMderived cells to the tumor but also their incorporation into the tumor vasculature, supporting an involvement of Hh in the tumorassociated angiogenesis. Our data do not support the notion that autocrine signaling of Hh in cancer cells contributes to tumor angiogenesis, and rather suggest that Hh acts exclusively on the stroma. Shh has been shown to directly control the activity of BMPCs cells by stimulating their proliferation, migration and inducing angiogenic factors to promote vascular remodeling [26,29,55]. We demonstrated that factors secreted by PDAC cells can enhance migration of the precursor cells at least in part in a Smo-dependent manner. There was an additional effect of Hh signaling on EC assembly through an activation of the BMPCs together with sprouting mature ECs. We have previously shown that Shh can enhance the activity of the precursors to promote capillary morphogenesis by HUVEC [26]. This is probably mediated through factors ''imported'' by the pro-angiogenic precursors that can prime neovascularization. BM-derived cells can differentiate into various cell types in the tumor microenvironment and the anti-tumor effect of Hh blockade may be mediated through multiple players in the cancer stroma [33]. Mesenchymal stem cells and immune cells are alternative targets of Hh [56,57]. In addition, fibroblasts in PDAC desmoplasia could be important Hh reactive cells [25]. Bailey et al. demonstrated that fibroblasts in human PDAC were positive for Gli1, and Shh promoted the proliferation of pancreatic stellate cells and induced their differentiation into myofibroblasts [28]. A similar conclusion was reached based on experiments showing that wild type, but not Smo knockout, mouse embryonic fibroblasts promote the growth of PDAC xenografts [27]. More recent study by Olive et al. indicated that IPI-926, a semisynthetic derivative of cyclopamine, can inhibit proliferation of SMA-positive stromal myofibroblasts, resulting in a depletion of desmoplastic PDAC stroma [22].
Although they described a transient 'enhancement' of tumor angiogenesis by Smo inhibition, the reduction in SMA + myofibroblasts may limit the maturation of neovasculature through a depletion of vascular pericytes, which in turn consequently 'destabilize' tumor-associated neovessels as we and others [58] observed. Further studies are required to monitor the tumor vessels during longer periods of Hh signaling inhibition in types of tumor with diverse histological architectures.
Our proposed model for Hh-mediated tumor angiogenesis is illustrated in Figure S10. The paracrine effect of Hh emerges at late stages of tumorigenesis and regulates neovascularization by acting through BM-derived cells. Hh ligand stimulates migration/interaction of the precursors to neovessels, which in turn act as a reservoir of cytokines to stabilize/maintain the tumor vasculature. The inhibition of angiogenesis by targeting Smo is consistent with earlier studies demonstrating a pro-angiogenic role for Hh on ischemic tissues  [29,47], embryonic vasculogenesis [59,60], as well as in tumor [58]. Hh could also contribute to the activation of stromal myofibroblasts (potentially derived from the BM) that promote tumorigenesis through supporting angiogenesis or other mechanisms [22]. However, so far the possibility that Hh signaling may affect on pre-existing vascular cells as well still exists. In addition, the potential early cell-autonomous role needs further investigation since it is not easily reconciled with the dispensability of epithelial Smo expression for PanIN initiation and progression [24].
Together with recent studies [24,27,28], it is becoming clear that a paracrine effect of Hh is crucial for tumorigenesis. Activation of the Hh pathway is important for the stromal cells, rather than aberrant proliferation of cancer cells. We have identified a novel aspect of Hh as a late mediator to support the tumor microenvironment, which promotes the development of the tumor vasculature in a paracrine manner, and this is mediated by a significant effect on BM-derived pro-angiogenic cells.

Supporting Information
Methods S1 Supplementary methods.