A Krasg12d-Driven Genetic Mouse Model of Pancreatic Cancer Requires Glypican-1 for Efﬁcient Proliferation and Angiogenesis

Oncogene (2012) 31, 2535–2544 & 2012 Macmillan Publishers Limited All rights reserved 0950-9232/12 www.nature.com/onc ORIGINAL ARTICLE A KrasG12D-driven genetic mouse model of pancreatic cancer requires glypican-1 for efﬁcient proliferation and angiogenesis

CA Whipple1,2, AL Young1,2 and M Korc1,2

1Departments of Medicine and Pharmacology and Toxicology, Dartmouth Medical School, Hanover, NH, USA and 2The Norris Cotton Cancer Center at Dartmouth-Hitchcock Medical Center, Lebanon, NH, USA

Pancreatic ductal adenocarcinomas (PDACs) exhibit Introduction multiple molecular alterations and overexpress heparin- binding growth factors (HBGFs) and glypican-1 (GPC1), Pancreatic ductal adenocarcinoma (PDAC) is a highly a heparan sulfate proteoglycan that promotes efficient metastatic malignancy that is the fourth leading cause signaling by HBGFs. It is not known, however, whether of cancer death in the United States, with an overall GPC1 has a role in genetic mouse models of PDAC. 5-year survival rate of o6% (Siegel et al., 2011). PDAC Therefore, we generated a GPC1 null mouse that exhibits a wide range of genetic and epigenetic altera- combines pancreas-specific Cre-mediated activation of tions including a high frequency (90–95%) of activating oncogenic Kras (KrasG12D) with deletion of a conditional Kras mutations, homozygous deletion (85%) and INK4A/Arf allele (Pdx1-Cre;LSL-KrasG12D;INK4A/ epigenetic silencing (15%) of the tumor suppressor Arflox/lox;GPC1À/À mice). By comparison with Pdx1- genes p16INK4A/p14ARF (INK4A), as well as an over- Cre;LSL-KrasG12D;INK4A/Arflox/lox mice that were wild abundance of heparin-binding growth factors (HBGFs) type for GPC1, the Pdx1-Cre;LSL-KrasG12D;INK4A/ (Korc, 2003; Schneider and Schmid, 2003). In spite of an Arflox/lox;GPC1À/À mice exhibited attenuated pancreatic improved understanding of these molecular alterations, tumor growth and invasiveness, decreased cancer cell there is a high incidence of treatment failure in PDAC, proliferation and mitogen-activated protein kinase activa- due, in part, to the resistance of pancreatic cancer cells tion. These mice also exhibited suppressed angiogenesis to chemotherapy and radiotherapy (Schneider and in conjunction with decreased expression of messenger Schmid, 2003; Kleeff et al., 2007; Korc, 2007). RNAs encoding several pro-angiogenic factors and mole- Many HBGFs, including fibroblast growth factor-2 cules, including vascular endothelial growth factor-A (FGF-2), vascular endothelial growth factor (VEGF), (VEGF-A), SRY-box containing gene (SOX17), chemo- platelet-derived growth factor and hepatocyte growth kine C-X3-C motif ligand 1 (CX3CL1) and integrin b3 factor, require the presence of the co-receptor glypican-1 (ITGB3). Moreover, pancreatic cancer cells isolated (GPC1) for efficient signaling (Berman et al., 1999; from the tumors of GPC1À/À mice were not as invasive Gengrinovitch et al., 1999; Chen and Lander, 2001). in response to fibroblast growth factor-2 (FGF-2) as GPC1 is a heparin sulfate proteoglycan (HSPG) that is cancer cells isolated from wild-type mice, and formed attached to the plasma membrane of epithelial and smaller tumors that exhibited an attenuated metastatic stromal cells via a glycosylphophatidylinositol anchor. potential. Similarly, VEGF-A and FGF-2 did not enhance GPC1, in conjunction with HBGFs, has a critical role in the migration of hepatic endothelial cells and immorta- cell growth, differentiation, adhesion and, possibly, lized murine embryonic fibroblasts isolated from GPC1 malignant transformation (Kleeff et al., 1998; Blackhall null mice. These data demonstrate in an oncogenic Kras- et al., 2001; Chen and Lander, 2001; Aikawa et al., driven genetic mouse model of PDAC that tumor growth, 2008). GPC1 is overexpressed in PDAC and is required angiogenesis and invasion are enhanced by GPC1, and for efficient HBGF signaling (Kleeff et al., 1998, 1999; suggest that suppression of GPC1 may be an important Aikawa et al., 2008). component of therapeutic strategies in PDAC. Using athymic mice in a GPC1 null background Oncogene (2012) 31, 2535–2544; doi:10.1038/onc.2011.430; and human pancreatic cancer cells in an orthotopic published online 26 September 2011 pancreatic cancer model, we have demonstrated that both host-derived and cancer cell-derived GPC1 contri- Keywords: pancreatic cancer; glypican-1; proliferation; bute to enhanced tumor growth and metastasis in angiogenesis; tumor microenvironment pancreatic cancer (Kleeff et al., 1998; Aikawa et al., 2008). It is not known, however, if GPC1 is important for both tumor initiation and progression in PDAC. Therefore, in the present study, we sought to investigate Correspondence: Dr M Korc, Department of Medicine, One Medical the impact of GPC1 on pancreatic tumor formation Center Drive, Dartmouth-Hitchcock Medical Center, Lebanon, NH and growth in a transgenic mouse model that combines 03756, USA. G12D E-mail: [email protected] the latent LSL-Kras knock-in allele, the conditional lox/lox Received 16 June 2011; revised 30 July 2011; accepted 18 August 2011; INK4A knockout allele, and either a GPC1 wild- published online 26 September 2011 type (Pdx1-Cre;LSL-KrasG12D;INK4A/Arflox/lox;GPC1 þ / þ ) Role of glypican-1 in pancreatic cancer progression CA Whipple et al 2536 or knockout (Pdx1-Cre;LSL-KrasG12D;INK4A/Arflox/lox; GPC1 þ / þ mice was significantly greater than in the GPC1À/À) allele. GPC1À/À mice (Figure 1b). At 65 days, 14 of 14 GPC1 þ / þ We now report that the absence of GPC1 results in a transgenic mice harbored large and invasive pancreatic significant decrease in tumor growth, invasiveness, tumors that adhered to and invaded surrounding organs, angiogenesis, pancreatic cancer cell proliferation and whereas 4 of 20 GPC1À/À mice did not have any tumors, mitogen-activated protein kinase (MAPK) activation and the remaining 16 mice exhibited significantly smaller in vivo. In vitro, pancreatic cancer cells isolated from tumors that were not grossly invasive or adherent tumors arising in GPC1À/À transgenic mice exhibited (Figures 1a and b). None of the mice exhibited grossly attenuated invasion in response to FGF-2 by compar- evident distant metastases. ison with pancreatic cancer cells established from Next, we carried out histological and immunohisto- transgenic GPC1 þ / þ mice. Additionally, primary hepa- chemical analysis of pancreata collected from GPC1 þ / þ tic endothelial cells and murine embryonic fibroblasts and GPC1À/À mice, using 10 pancreata per group from isolated from GPC1À/À mice were completely resistant to the 30-day-old mice, and 14 pancreata per group from FGF-2-mediated migration. These findings indicate that the 65-day-old mice. Half of the compound transgenic endogenous GPC1 contributes to pancreatic cancer mice developed pancreatic intraepithelial neoplasia progression, even when driven by oncogenic Kras in (PanIN) lesions by 30 days of age and most of these conjunction with loss of INK4A, two highly prevalent were PanIN-1, -2 lesions (Supplementary Table 1). By alterations in PDAC. contrast, 3 of the 10 GPC1 þ / þ mice exhibited PanIN-3 lesions (Supplementary Figure 1), and 1 mouse had 2 separate invasive PDACs (Supplementary Table 1), whereas only 1 of the 10 GPC1À/À mice displayed a Results PanIN-3 lesion, and none had a cancer. By day 65, 14 of 14 GPC1 þ / þ and 13 of 14 GPC1À/À mice harbored Creation of a novel transgenic mouse model invasive carcinomas. Although one GPC1À/À mouse did We generated a compound mouse model in which not have any PanIN or carcinoma, all of the transgenic pancreas-specific activation of endogenous KrasG12D and mice displayed focal areas of acinar to ductal metaplasia the loss of pancreatic INK4A expression were combined (ADM lesions). with the complete absence of GPC1. To maintain a PanIN lesions in both GPC1 þ / þ and GPC1À/À mice similar genetic background for both the GPC1 wild-type were histologically similar across all PanIN stages. (GPC1 þ / þ ) and knockout (GPC1À/À) transgenic mouse Moreover, PanIN-1A lesions in both groups of mice lines and to avoid lethality caused by homozygosity of displayed abundant mucin production as evidenced either the mutant Kras or the Pdx1-Cre alleles (Johnson by the presence of strong MUC1 immunoreactivity et al., 1997), generation of the mouse model was carried and Alcian blue staining (Figure 1c). In addition, out in a step-wise manner (Supplementary Figure 1). trichrome staining revealed that the stroma was First, to create the GPC1À/À transgenic mouse, the similarly abundant in the tumors of both groups of mouse line carrying either the conditionally expressed G12D lox/lox mice (Figure 1c). By contrast, Ki67 and CD34, which LSL-Kras mutation or the INK4A knockout are markers for proliferation and angiogenesis, respec- allele was bred separately with GPC1À/À mice. In tively (Vilar et al., 2007; Le Mercier et al., 2009), were parallel, the mice harboring the INK4Alox/lox knockout markedly decreased in tumors from GPC1À/À by com- allele and the Pdx1-Cre allele were bred to the GPC1À/À parison with GPC þ / þ mice (Figure 2). Immunofluo- mouse line. Two rounds of breeding for each line rescent staining for Ki67 indicated that proliferation was produced mice that were either Pdx1-Cre;INK4ALox/Lox; significantly reduced in both the PanIN (CK19-positive) GPC1À/À or LSL-KrasG12D;INK4ALox/Lox;GPC1À/À. The and stromal cells of tumors from the 65-day-old GPC1À/À offspring from these separate breeding strategies were transgenic mice (Figure 3). Immunofluorescent staining crossed to produce Pdx1-Cre;LSL-KrasG12D; for pMAPK was also decreased in the cancer cells within INK4ALox/Lox;GPC1À/À mice (Supplementary Figure 1). the tumors in GPC1À/À mice (Figure 4), whereas caspase-3 A similar breeding strategy was used to create the staining was similar in both groups (data not shown). GPC1 þ / þ transgenic mouse. Loss of INK4A and recom- Quantitative real-time RT–PCR (Q–PCR) of tumor bination of the mutant Kras allele were confirmed by RNA isolated from 65-day-old GPC þ / þ and GPC1À/À genotyping all the mice. mice revealed that the loss of GPC1 did not significantly alter glypican-2, -3 -4, -5 or -6 levels (Supplementary Loss of GPC1 impacts tumor development and metastasis Figure 2). By contrast, tumors arising from GPC1À/À To assess the impact of the loss of GPC1 on tumor pancreata expressed markedly reduced levels of VEGF-A development, we analyzed the pancreata of Pdx1- messenger RNA (mRNA) by comparison with GPC1 þ / þ Cre;LSL-KrasG12D;INK4ALox/Lox;GPC1À/À and the corre- tumors (Supplementary Figure 2). sponding GPC1 þ / þ mice (Figure 1). At 30 days of age, To further delineate the role of GPC1 in pancreatic by gross analysis, only 1 of 10 GPC1À/À mice had cancer progression, primary cancer cell lines were developed a small pancreatic tumor versus 7 of the 10 established from the tumors that arose in the GPC1 þ / þ GPC1 þ / þ mice. In addition, the GPC1À/À pancreata and GPC1À/À transgenic mice. Four murine pancreatic were visibly smaller than the GPC1 þ / þ pancreata cancer cell lines were established, two from GPC1 þ / þ (Figure 1a), and the wet weight of the pancreata in (F1015 and F1048) and two from GPC1À/À (J444 and

Oncogene Role of glypican-1 in pancreatic cancer progression CA Whipple et al 2537

Figure 1 Effects of GPC1 loss on pancreas size and tumor morphology. (a) Representative pancreata from 30- and 65-day-old Pdx1- Cre;LSL-KrasG12D;INK4A/Arflox/lox;GPC1 þ / þ or GPC1À/À transgenic mice. (b) Pancreatic wet weight. Data are an average wet weight of 10 GPC1 þ / þ and 10 GPC1À/À 30-day-old mice, and 15 GPC1 þ / þ and 20 GPC1À/À 65-day-old transgenic mice. Data are the means±s.e.m.; *Po0.05, when compared with the corresponding GPC1 þ / þ group. (c) Pancreatic morphology. MUC1, Alcian blue and Mason’s-trichrome staining were performed on parafﬁn sections (5 mm thick) prepared from indicated pancreatic tumors. Magniﬁcation 10 Â . Bar: 50 mM.

Figure 2 Effects of GPC1 loss on Ki67 and CD34 immunoreactivity. (a) Ki67 immunoreactivity. Staining was performed on paraffin sections (5 mm thick) prepared from indicated pancreatic tumors (six mice in each group). (b) CD34 immunoreactivity. CD34 immunostaining was analyzed in the indicated pancreatic tumors (seven mice in each group). Magnification 20 Â .(c) Morphometric analysis. Quantitative morphometry was performed after scoring 12 areas on each slide prepared from the 26 tumors described in a and b. Magnification 10 Â . Data are the means±s.e.m. *Po0.05 and **Po0.002 by comparison with the respective values in the GPC1 þ / þ tumors. Bar: 50 mM.

J1032) mice. Following subcutaneous injection in shorter doubling times (15 and 20 h) than J444 and GPC1 þ / þ nude mice, all four cell lines formed tumors. J1032 cells (30 and 24 h), whereas J444 and J1032 cells In vitro, F1015 and F1048 cells displayed signiﬁcantly exhibited decreased invasiveness in response to FGF-2

Oncogene Role of glypican-1 in pancreatic cancer progression CA Whipple et al 2538

Figure 3 Evaluation of cellular proliferation in cancer cells and adjoining stroma. (a) Ki67 immunofluorescence. Immunofluorescent staining was performed on paraffin sections (5 mm thick) prepared from pancreatic tumors isolated from two tumors/group of 65-day- old Pdx1-Cre;LSL-KrasG12D;INK4A/Arflox/lox;GPC1 þ / þ or GPC1À/À mice. Ki67-positive cells (stained red) are seen in the cancer cells (arrow) in which the cells are also CK19-positive (stained green), and in the adjacent stromal cells (asterisks). Nuclei are stained with 4,6-diamidino-2-phenyl indole in blue. Magnification 20 Â . Bar: 50 mM.(b) Quantitative morphometry. Ki67-positive cancer cells and stromal fibroblasts were scored in five areas from each slide, and expressed as a percentage of the total number of cancer cells and stromal fibroblasts, respectively. Data are means±s.e.m. from nine GPC1 þ / þ and nine GPC1À/À transgenic mice. *Po0.05 when compared with the corresponding GPC1 þ / þ group.

by comparison with either F1015 or F1048 cells The implanted tumor fragments contained cancer- (Figure 5). associated fibroblasts and endothelial cells, which may Given that loss of GPC1 reduced doubling times and help to promote a pro-metastatic profile. Therefore, we invasion in vitro, we next sought to determine whether next sought to assess the impact of GPC1 loss on GPC1 also modulated the metastatic potential of fibroblast and endothelial cell function. In a first set of pancreatic cancer cells. Accordingly, small (2 mm3) experiments, mouse embryo fibroblasts (MEFs) were fragments were prepared from 65-day-old GPC1 þ / þ isolated from GPC1 þ / þ or GPC1À/À mice at embryonic and GPC1À/À mice, and implanted, respectively, into day 14.5, and immortalized by stable infection with a the pancreata of athymic GPC1 þ / þ and GPC1À/À mice. retroviral vector containing a dominant negative p53 Tumor growth arising from transplanted fragments mutation. The immortalized MEFs were readily main- reflects the influence of both transplanted cancer cells tained in culture. In contrast to wild-type MEFs, and associated stroma, and yields metastatic lesions. GPC1À/À MEFs failed to respond to FGF-2 in a Dramatically, 2 weeks following fragment implant- migration assay, even when the ligand concentration ation, only 2 of 14 GPC1À/À mice developed metastases, was increased 5-fold (Figure 6). In a second set of which were confined to the mesentery, whereas 9 of experiments, the effects of VEGF-A on primary 15 GPC1 þ / þ mice developed numerous (over 100 per endothelial cells isolated from GPC1 þ / þ or GPC1À/À mouse) mesenteric metastases and 3 of these mice also mice were studied in a transwell migration assay. exhibited multiple renal metastases (Table 1). VEGF-A significantly increased the migration of

Oncogene Role of glypican-1 in pancreatic cancer progression CA Whipple et al 2539

Figure 4 Effects of GPC1 on pMAPK immunofluorescence. pMAPK immunostaining was performed on paraffin sections (5 mm thick) prepared from pancreatic tumors isolated from 65-day-old Pdx1-Cre;LSL-KrasG12D;INK4A/Arflox/lox mice that were either GPC1 þ / þ or GPC1À/À. The slides were also stained for CK19 to identify the epithelial cells. Negative controls were stained in the absence of the respective primary antibody. The lower right panel reveals morphology of negative control samples following staining with 40,6-diamidino-2-phenylindole. Magnification 20 Â . Bar: 50 mM.

Table 1 Effects of the loss of GPC1 on tumor metastasis Mouse WT KO

Mice with mesenteric metastases (4100 per mouse) 9/15 2/14 Mice with renal metastases (3–5 per mouse) 3/15 0/14

Abbreviations: GPC1, glypican-1; KO, knockout; WT, wild type. Tumor fragments isolated from pancreatic cancers arising in GPC1 þ / þ or GPC1À/À mice were inserted into the pancreas of GPC1 þ / þ or GPC1À/À athymic mice, respectively. Metastasis was assessed 3 weeks later.

endothelial cells from these mice, we next sought to determine whether the loss of GPC1 altered the expression of angiogenic genes. Microarray analysis of Figure 5 Effects of GPC1 on primary cancer cell invasion. RNA extracted from transgenic tumors isolated from F1015 and F1048 cells, derived from GPC1 þ / þ tumors, and J444 65-day-old mice indicated that six other pro-angiogenic and J1032 cells, derived from GPC1À/À tumors, were incubated for genes were also expressed at low levels in the absence 22 h in serum-free medium, in the absence or presence of 75 ng/ml of GPC1. Thus, SRY-box containing gene (SOX17), of FGF-2. Data are expressed as the total number of cells that invaded in response to 0.1% bovine serum albumin (control) or chemokine C-X3-C motif ligand 1 (CX3CL1), ephrin- FGF-2, and are the means±s.e.m. from three separate experi- A1 (EFNA1), guanine nucleotide-binding protein alpha ments. *Po0.05, when compared with the corresponding control. 13 (GNA13), integrin beta 3 (ITGB3), protein tyrosine kinase 2 beta (PTK2B) and VEGF-A mRNA levels were expressed at a signiﬁcantly lower level in the GPC1À/À GPC1 þ / þ -derived endothelial cells, but did not alter the tumors (Table 2 and Supplementary Figure 3). By migration of GPC1À/À-derived endothelial cells (Figure 6). contrast, the expression of angiostatic genes, such as Thus, both ﬁbroblasts and endothelial cells derived from tissue inhibitor of metalloproteinases (TIMP)-1 and -2, GPC1À/À mice were less responsive to HBGFs. nm23, angiopoietin-2, thrombospondin-2 or prolactin, In view of the lower level of VEGF-A in GPC1À/À- was similar in both groups of pancreata (data not derived tumors and the decreased migration of shown).

Oncogene Role of glypican-1 in pancreatic cancer progression CA Whipple et al 2540

Figure 6 Effects of GPC1 on cell migrations. (a) Migration of immortalized MEFs. Cells that were originally derived from GPC1 þ / þ or GPC1À/À mice were serum starved overnight and then incubated for 8 h with 75 ng/ml FGF-2. Representative photos are shown at 0 h and 8 h following FGF-2 addition. Bar graphs on the right are the means±s.e.m. of duplicate determinations from three separate experiments. *Po0.05 by comparison with the respective control in the absence of FGF-2. (b) Migration of primary hepatic endothelial cells. Endothelial cells were incubated overnight in the absence or presence of 75 ng/ml of VEGF-A. Representative photos are of endothelial cells derived from GPC1 þ / þ or GPC1À/À mice. Bar graphs on the right are the means±s.e.m. of duplicate determinations from three separate experiments. *Po0.05 by comparison with the respective control in the absence of FGF-2.

Table 2 Effects of GPC1 loss on the expression of pro-angiogenic cancer cells in PDAC, and in vitro studies with human genes pancreatic cancer cell lines have shown that GPC1 is Gene Microarray fold change Q–PCR fold change readily released by these cells (Matsuda et al., 2001), pointing to a possible role for GPC1 within the tumor EFNA1 À5.4 À3.6 microenvironment. Fourth, studies with athymic CX3CL1 À6.9 À3.2 GPC1À/À mice have demonstrated that host-derived PTK2B À4.7 À3.1 GNA13 À5.4 À2.3 GPC1, produced by stromal and endothelial cells, VEGFA À4.3 À3.5 contributes to pancreatic tumor growth, metastasis SOX17 À6.8 À7.8 and angiogenesis (Aikawa et al., 2008). Fifth, gene ITGB3 À4.9 À2.4 expression profiling in pancreatic intraductal papillary- mucinous tumors revealed that GPC1 is upregulated in Abbreviations: GPC1, glypican-1; Q–PCR, Quantitative real-time RT– PCR. these lesions, but nearly exclusively in invasive intra- RNA isolated from pancreatic tumors derived from five GPC þ / þ and ductal papillary-mucinous tumors, suggesting a poten- five GPC1À/À mice was analyzed using an Agilent 4 Â 44K Whole tial role in tumor invasion (Terris et al., 2002). Mouse Genome Oligo Microarray. Pro-angiogenesis genes down- The LSL-KrasG12D mouse, which we combined with À/À regulated in the GPC1 mice are shown above (full names are listed the loss of both GPC1 and INK4A, carries an oncogenic in Supplementary Figure 3). Microarray data were validated by Q– G12D PCR, which was performed in triplicate. All data were significantly Kras (Kras ) allele that has been knocked-in within different from corresponding values in GPC1 þ / þ tumors (Po0.05). its own locus but which is transcriptionally silenced by the insertion of a LoxP-Stop-LoxP element (LSL) located upstream of the transcriptional start site Discussion (Aguirre et al., 2003; Hingorani et al., 2003). Oncogenic Kras expression remains under the control of its own Several lines of evidence suggest that GPC1 may have an endogenous promoter, and the transcript is produced important role in PDAC. First, studies with human only in early pancreatic progenitor cells after excision of pancreatic cancer cell lines injected subcutaneously in the LSL sequence via the Pdx-1-driven Cre recombinase. athymic mice have demonstrated that GPC1 expression The Pdx1-Cre;LSL-KrasG12D mice develop low-grade is required for efficient pancreatic tumor growth and PanIN and exhibit areas of acinar to ductal metaplasia angiogenesis (Kleeff et al., 1998, 1999). Second, GPC1 is by 2 months of age (Aguirre et al., 2003). When the mice overexpressed in resected PDAC samples, and both the are 8 to 12 months old, they develop PDAC at low cancer cells and the cancer-associated fibroblasts express penetrance (Aguirre et al., 2003; Habbe et al., 2008). GPC1 at the mRNA and protein level (Kleeff et al., PanIN progression to pancreatic cancer is an impor- 1998), indicating that both cell types synthesize GPC1. tant feature of PDAC initiation in both human and Third, GPC1 is abundant in the stroma adjacent to the mouse models, and is dramatically accelerated in

Oncogene Role of glypican-1 in pancreatic cancer progression CA Whipple et al 2541 Pdx1-Cre;LSL-KrasG12D;INK4Alox/lox mice. In addition GPC1À/À pancreata displayed markedly reduced angio- to harboring oncogenic KrasG12D, the pancreata of these genesis. Thus, tumor angiogenesis is important in at mice have sustained a homozygous deletion of the least some genetic mouse model of PDAC, is driven by INK4A locus, resulting in large, highly invasive tumors GPC1-dependent mechanisms and can contribute to by 7–11 weeks of age (Jackson et al., 2001; Aguirre enhanced pancreatic cancer growth and metastasis. This et al., 2003). pre-clinical model could therefore be useful for assessing In the present study, both 30- and 65-day-old the effectiveness of specific antiangiogenic therapeutic Pdx1-Cre;LSL-KrasG12D;INK4ALox/Lox;GPC1À/À mice modalities in PDAC, and for investigating the mechan- displayed significantly smaller tumors than the corre- isms that contribute to the failure of antiangiogenic sponding GPC1 þ / þ mice. PanIN appeared at the same therapy in clinical trials (Preis and Korc, 2010). time in both groups of mice and were morphologically At least six mechanisms may underlie the ability of similar, exhibiting similar patterns of MUC1 and Alcian GPC1 to promote tumor angiogenesis in Pdx1-Cre;LSL- blue staining. Nonetheless, proliferation markers were KrasG12D;INK4A/Arflox/lox mice. First, the GPC1À/À increased in the PanIN lesions in GPC1 þ / þ mice by tumors exhibited decreased expression of VEGF-A, a comparison with GPC1À/À mice. Moreover, PanIN potent pro-angiogenic factor. Second, our array studies progression from PanIN-1 to PanIN-3 lesions and to revealed that there was concomitant downregulation of PDAC occurred more rapidly in the GPC1 þ / þ mice. six additional genes implicated in promoting angiogen- Thus, GPC1 is not required for PanIN initiation but esis, such as integrin b3 and ephrin-A1. By contrast, the acts to promote progression toward malignancy, most expression of angiostatic genes was similar in both likely by facilitating epithelial cell proliferation within groups of pancreata. Inasmuch as VEGF-A is known to these lesions. upregulate the expression of ephrin-A1 (Cheng et al., The cancer cells and the cancer-associated fibroblasts 2002), it is possible that loss of GPC1, by leading to in GPC1 þ / þ mouse tumors also exhibited increased decreased VEGF-A expression and function, results in proliferation and elevated phospho-MAPK levels, in- the downregulation of downstream genes such as dicating that GPC1 enhances mitogenic signaling in ephrin-A1, further impeding angiogenesis. Third, there both cell types. By contrast, Mason’s trichrome and was a marked decrease in Sox17 mRNA levels in tumors caspase-3 staining were similar in tumors from GPC1 þ / þ from GPC1À/À mice. Sox17 is a transcriptional regulator and GPC1À/À mice, suggesting that GPC1 promoted that is expressed in fetal and neonatal hematopoietic larger tumors by facilitating enhanced proliferation of stem cells (HSCs), but not in adult HSCs (Kim et al., the cancer cells rather than by promoting increased 2007). Inasmuch as HSCs have been implicated in collagen deposition or attenuating apoptosis. In support PDAC angiogenesis (Li et al., 2011), it is possible that of this conclusion, primary cancer cells derived from the GPC1 is also important for PDAC’s ability to recruit GPC1 þ / þ mice exhibited enhanced proliferation in vitro. HSCs to promote tumor angiogenesis. Fourth, by Inasmuch as GPC1 is devoid of intrinsic signaling contrast to endothelial cells isolated from GPC1 þ / þ activity, these observations suggest that GPC1 may act mice, endothelial cells from GPC1À/À mice did not by a cell-autonomous manner to facilitate autocrine migrate in response to VEGF-A, underscoring the growth stimulation elicited by cancer cell-derived HBGFs GPC1 dependence for efficient VEGF-A signaling. that act via their cognate high-affinity receptors. Fifth, GPC1 may also act to promote growth factor Recent studies underscore the importance of the sequestration within the tumor microenvironment, cancer-associated stroma in facilitating tumor spread, enhancing their stability and ability to act as morpho- invasion and metastasis, enhancing chemoresistance and gens (Vlodavsky et al., 1990; Lander et al., 2002). Thus, promoting apoptosis resistance (Hwang et al., 2008; the absence of GPC1 within the tumors of GPC1À/À mice Vonlaufen et al., 2008). PDAC is characterized by may decrease the pool of heparan sulfate side chains, an abundant desmoplastic reaction that is rich in thereby decreasing the amount of stored HBGFs, such collagen I, pancreatic stellate cells, fibroblasts and as VEGF-A (41, 42), that can be released by the actions inflammatory cells, which collectively promote cancer of the endoglycosidase heparanase, which is upregulated cell growth and invasiveness, and contribute to chemo- in PDAC (Zetser et al., 2006). Sixth, the absence of resistance (Kleeff et al., 2007). While orthotopic models GPC1 may also interfere with the shedding of active of PDAC have generally not been associated with MMP7 (Ding et al., 2005), which acts to release all desmoplasia, many genetic mouse models of PDAC, stored growth factors, irrespective of their heparin- including the Pdx1-Cre;LSL-KrasG12D;INK4A/Arflox/lox binding ability. Together, these observations suggest mice used in the current study, exhibit desmoplasia that GPC1 may be an important component of a pro- (Aguirre et al., 2003; Tuveson and Hingorani, 2005). angiogenic switch in PDAC. By contrast, orthotopic models exhibit aberrant angio- Using athymic mice, both host-derived and cancer genesis (Rowland-Goldsmith et al., 2002), whereas cell-derived GPC1 has been shown to contribute to genetic mouse models of PDAC are ostensibly not enhanced pancreatic cancer growth (Kleeff et al., 1998; angiogenesis dependent and consequently are more Aikawa et al., 2008). In the present study, pancreatic resistant to chemotherapy (Olive et al., 2009). It is cancer cell lines established from GPC1À/À mice demon- noteworthy, therefore, that there was abundant angio- strated a marked decrease in invasiveness in vitro, as well genesis in the tumors arising in Pdx1-Cre;LSL-KrasG12D; as attenuated growth and metastatic potential following INK4A/Arflox/lox;GPC1 þ / þ mice, and that tumors in intra-pancreatic implantation in GPC1 deficient nude

Oncogene Role of glypican-1 in pancreatic cancer progression CA Whipple et al 2542 mice. Thus, in the absence of GPC1, pancreatic cancer with sterile scissors, digested with collagenase IV (2.5 mg/ml) cells derived from Kras-driven genetic mouse model of for 1 h at 371C and resuspended in cancer cell growth media PDAC exhibit attenuated tumor progression, angio- (RPMI 1640 with 20% fetal bovine serum, 100 units per ml genesis, invasiveness and metastasis, suggesting that penicillin/streptomycin). The cell mixture was laid over an targeting GPC1 either alone or in combination with equal volume of histopaque, and spun for 15 min at 2000 rpm. mitogenesis and/or angiogenesis inhibitors may help to The pellet was isolated, resuspended in cancer cell growth medium and seeded on tissue culture plates. overcome the resistance to Kras-targeted therapeutic Primary MEFs were isolated from 13.5-day-old embryos strategies. (Xu, 2005), and placed in growth medium consisting of Dulbecco’s modified Eagle’s medium (DMEM) containing 8% fetal bovine serum, 100 units per ml penicillin/streptomycin Materials and methods and 25 ng/ml amphotericin B, plated on tissue culture plates and incubated at 37 1C with 5% CO2. At cell passage two, þ / þ À/À Mouse colony generation MEFs isolated from GPC1 or GPC1 mice were LSL-KrasG12D;INK4A/ArfLox/Lox;SMAD4lox/lox and Pdx1-Cre immortalized with a p53 dominant negative retroviral vector mice strains were kindly provided by Nabeel Bardeesy at (a gift from Robert Weinberg, Whitehead Institute, Cam- Harvard University (Bardeesy et al., 2006) and Guo Gu at bridge, MA, USA), utilizing standard retroviral infection Vanderbilt University (Gu et al., 2003), respectively. The protocols (Cepko and Warren, 1996). SMAD4lox/lox mutation was bred out of this line before Primary endothelial cells were isolated from adult livers as introducing the Pdx1-Cre and GPC1 mutations (Supplemen- described previously, using Dynalbeads (Invitrogen, Carlsbad, tary Figure 4). The GPC1À/À mice were produced by targeted CA, USA) labeled with anti-CD31 antibodies (Aikawa et al., mutagenesis as described previously (Aikawa et al., 2008; Jen 2008). Endothelial cells were similarly isolated from tumors G12D Lox/Lox þ / þ et al., 2009). All mice were genotyped twice by PCR to verify arising in Pdx1-Cre;LSL-Kras ;INK4A ;GPC1 or À/À the presence of each transgene. GPC1 transgenic mice that were 60 to 65 days old. Cells were then grown in DMEM supplemented with 20% fetal bovine serum, penicillin/streptomycin (100 U/ml) (Mediatech, Hern- Histology and immunohistochemistry of paraffin-embedded don, VA, USA), heparin (100 mg/ml) (Sigma, St Louis, MO, tissues USA), endothelial cell growth factor (100 mg/ml) (Biomedical Pancreatic tissues were fixed for 6–12 h in 10% formalin, Technologies, Stoughton, MA, USA), 1 Â non-essential amino embedded in paraffin, sectioned (5 mm thick), mounted on acids (Gibco, Grand Island, NY, USA), 2 mML-glutamine 1 poly-L-lysine-coated glass slides and incubated at 79 C for (Gibco), 1 Â sodium pyruvate, gentamicin (58 mg/ml) (Gibco) 30 min. After deparaffinization, antigen retrieval was per- and amphotericin B (25 ng/ml) (Mediatech), termed endothelial formed by using the 2100-Retriever (PickCell Laboratories, complete medium, and plated on fibronectin (1 mg/cm2)-coated Amsterdam, The Netherlands) and antigen unmasking solu- plates (BD Biosciences, San Jose, CA, USA). tion (pH 6.0). Sections were washed in phosphate-buffered saline, incubated for 10 min with 3% H2O2 to block endogenous peroxidase activity, washed in phosphate-buffered Cell culture and in vitro assays saline with 0.025% Triton-X, incubated for 1 h in blocking Anchorage-dependent growth was assessed by plating 3 Â 103 buffer (1% bovine serum albumin and 5% normal goat or primary cancer cells per well of a 96-well plate, serum starved horse serum) and overnight at 4 1C with the appropriate overnight, incubated for 48 h in serum-free media (DMEM, monoclonal antibody (Aguirre et al., 2003; Aikawa et al., 5 mg/ml apotransferrin, 5 ng/ml selenium, 0.01% BSA and 2008). Bound antibody was detected with the suitable 100 U/ml penicillin/streptomycin) and then utilizing the 3-(4,5- biotinylated secondary antibody (1:500) and VECTASTAIN methylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) ABC Reagent (Vector Laboratories, Burlingame, CA, USA) assay (Kleeff et al., 1998). Doubling times were determined complex, using diaminobenzidine tetrahydrochloride as the by plating 2 Â 104 cells/well in 6-well plates and counting cells substrate (Rowland-Goldsmith et al., 2002). with a hemocytometer every 24 h for 5 days. The following primary antibodies were used: anti-Ki67 BD Biocoat Matrigel Invasion Chambers (BD Biosciences) (Novocastra, Bannockburn, IL, USA), 1:100; anticleaved were utilized to assess cancer cell invasion. The chambers were caspase-3 (Cell Signaling Technology, Beverly, MA, USA), incubated at room temperature for 30 min and rehydrated with 1:200; anti-MUC1 (Abcam Inc, Cambridge, MA, USA), 1:1000; 0.5 ml of serum-free DMEM medium for 2 h at 37 1C. Cancer TromaIII (anti-CK19 antibody developed by Rolf Kemler and cells were seeded in each chamber at a cell density of 50 Â 103 obtained from the Developmental Studies HybridomaBank, in 500 ml of serum-free medium, and incubated for 22 h at 37 1C Iowa City, IA, USA), 1:10; anti-Cd34 (Abcam Inc), 1:25; and in a well containing 750 ml of serum-free medium in the absence antiphospho-MAPK (Cell Signaling Technology, Danvers, MA, or presence of 50 ng/ml of FGF-2. After washing, fixing and USA), 1:500. Negative controls without primary antibodies did staining, cells were counted on an inverted microscope (10 Â not exhibit immunoreactivity. Pictures were captured on an objective). Olympus BX60 microscope (Olympus, Center Valley, PA, USA) Wounding assays were performed to assess the ability of withaQImagingEXIBluecamerautilizing10Â and 20 Â immortalized MEFs to migrate in response to growth factors objectives. Quantitative morphometry was performed with (Rowland-Goldsmith et al., 2001; Arnold et al., 2004). Briefly, Image-Pro Plus 7 (Media Cybernetics, Silver Springs, MD, 75 Â 103 MEFs originally isolated from GPC1 þ / þ or GPC1À/À USA), as previously reported (Aikawa et al., 2008). mice were plated in each well of a 6-well plate and serum starved overnight. Confluent cells were washed with Hank’s Isolation of primary cell lines balanced salt solution; a p200 pipet tip was used to make two To isolate and establish primary cancer cell lines, pancreatic scratches across each well, the cells were gently washed again tumors were rapidly harvested from Pdx1-Cre;LSL-KrasG12D; and incubated in serum-free medium in the absence or presence of INK4ALox/Lox;GPC1 þ / þ or GPC1À/À transgenic mice that were 75 ng/ml of FGF-2 or VEGF-A. Photos taken at 0 and 8 h allowed 60 to 65 days old (Seeley et al., 2009). The tumors were minced for analysis of the migration distance (Suyama et al., 2003).

Oncogene Role of glypican-1 in pancreatic cancer progression CA Whipple et al 2543 Transwell migration assays were used to analyze the ability pancreatic tumors, GPC1 þ / þ (F1015 and F1048) and GPC1À/À of primary endothelial cells to migrate in response to HBGFs. (J444 and J446) cells (5 Â 106) were injected subcutaneously For these assays, transwell chambers (BD Biosciences) were into athymic nude mice and assessed for tumors daily. To coated by overnight incubation at 4 1C with 100 ml of the determine the consequences of GPC1 loss on the metastatic purified matrix proteins: 1.5 mg/ml collagen (Cohesion, Palo potential of the cancer cells, two small tumor fragments from Alto, CA, USA), 0.5 mg/ml vitronectin (Sigma), 5 mg/ml 65-day-old Pdx1-Cre;LSL-KrasG12D;INK4ALox/Lox;GPC1 þ / þ or fibronectin (BD Biosciences) and 0.5% gelatin. The primary GPC1À/À transgenic mice were sutured into the pancreata of endothelial cells were serum starved overnight. After removal 4–6 athymic GPC1 þ / þ or 4–6 GPC1À/À mice, respectively. of the matrix protein solution, the chambers were saturated Each orthotopic experiment was carried out in triplicate. with 100 ml of 1% bovine serum albumin in phosphate- buffered saline and incubated for 1 h at 37 1C. The bovine Microarray analysis serum albumin was removed, and 10 Â 104 endothelial cells Total RNA from five GPC1 þ / þ and five GPC1À/À tumors isolated from GPC1 þ / þ or GPC1À/À mice were added to each isolated from 65-day-old transgenic mice was extracted using chamber using serum-free endothelial growth medium. The Trizol (Invitrogen). Gene expression analysis and bioinfor- chambers were moved to a well containing 600 ml of serum-free matics was performed by Miltenyi Biotec (Bergisch Gladbach, medium and incubated for 12 h in the absence or presence of Germany), using two-color Agilent whole-mouse genome VEGF-A. The cells were then fixed with 4% paraformalde- arrays where each knockout sample was hybridized to a hyde, stained with 1% toluidine blue and 2% sodium borate, control sample on the same array and assignment of the Cy5/ and counted on an inverted microscope with the 20 Â Cy3 pairs was arbitrary. objective (Neupane and Korc, 2008).

Q–PCR analysis Statistics To assess the impact of GPC1 on the expression levels of For statistical group comparisons of array data between the VEGF-A and GPC2-6, total RNA was isolated from tumors knockout and control samples, unpaired t-tests with equal arising in 60-day-old GPC1 þ / þ and 5-day-old GPC1À/À mice, variance were employed utilizing the normalized intensity using the low-temperature guanidine isothiocyanate method data. All other data were analyzed by Student’s t test, with (Han et al., 1987; Chang et al., 1990). The concentration of Po0.05 defined as significant. RNA was determined spectrophometrically, and RNA integ- rity was verified on a 1% agarose gel. cDNA was synthesized using oligo-dT primers as part of the SuperScript RT kit Conflict of interest (Invitrogen) according to the manufacturer’s protocol. Q–PCR was performed using an ABI Prism 7300 sequence detection The authors declare no conflict of interest. system (Applied Biosystems, Carlsbad, CA, USA) utilizing the Applied Biosystems Taqman Assay with pre-validated murine probes and primer sets. All RNA samples were run in duplicate and each experiment was repeated in triplicate. A dilution Acknowledgements series was carried out for each gene, and the 18S ribosomal subunit was used as an internal control. Q–PCR data were We thank Dr Daniel Longnecker (Dartmouth Medical School) expressed as a relative quantity based on the ratio of for his expert assistance in the evaluation of pancreatic tumor fluorescent change observed with the 18S ribosomal subunit. pathology. This work was supported by two US Public Health Service Grant (CA-R37-075059) awarded to MK. Orthotopic mouse models CAW was supported by a Ruth L Kirschstein National To compare the tumorogenicity of the four primary mouse Research Service Award, F32 CA-144579 from the National cancer cell lines isolated from GPC1 þ / þ and GPC1À/À mice Cancer Institute.

References

Aguirre AJ, Bardeesy N, Sinha M, Lopez L, Tuveson DA, Horner J growth factor and the keratinocyte growth factor. J Biol Chem 274: et al. (2003). Activated Kras and Ink4a/Arf deficiency cooperate to 36132–36138. produce metastatic pancreatic ductal adenocarcinoma. Genes Dev Blackhall FH, Merry CL, Davies EJ, Jayson GC. (2001). Heparan 17: 3112–3126. sulfate proteoglycans and cancer. Br J Cancer 85: 1094–1098. Aikawa T, Whipple CA, Lopez ME, Gunn J, Young A, Lander AD Cepko CP, Warren P. (1996). Current protocols in molecular biology. et al. (2008). Glypican-1 modulates the angiogenic and metastatic Retrovirus Infection of Cells in vitro and in vivo. Wiley Interscience: potential of human and mouse cancer cells. J Clin Invest 118: 89–99. Indianapolis, IN, USA. pp 9.14.11–19.14.16. Arnold NB, Ketterer K, Kleeff J, Friess H, Buchler MW, Korc M. Chang SC, Brannon PM, Korc M. (1990). Effects of dietary (2004). Thioredoxin is downstream of Smad7 in a pathway that manganese deficiency on rat pancreatic amylase mRNA levels. promotes growth and suppresses cisplatin-induced apoptosis in J Nutr 120: 1228–1234. pancreatic cancer. Cancer Res 64: 3599–3606. Chen RL, Lander AD. (2001). Mechanisms underlying preferential Bardeesy N, Aguirre AJ, Chu GC, Cheng KH, Lopez LV, Hezel AF assembly of heparan sulfate on glypican-1. JBiolChem276: 7507–7517. et al. (2006). Both p16(Ink4a) and the p19(Arf)-p53 pathway Cheng N, Brantley DM, Liu H, Lin Q, Enriquez M, Gale N et al. constrain progression of pancreatic adenocarcinoma in the mouse. (2002). Blockade of EphA receptor tyrosine kinase activation Proc Natl Acad Sci USA 103: 5947–5952. inhibits vascular endothelial cell growth factor-induced angiogen- Berman B, Ostrovsky O, Shlissel M, Lang T, Regan D, Vlodavsky I esis. Mol Cancer Res 1: 2–11. et al. (1999). Similarities and differences between the effects Ding K, Lopez-Burks M, Sanchez-Duran JA, Korc M, Lander AD. of heparin and glypican-1 on the bioactivity of acidic fibroblast (2005). Growth factor-induced shedding of syndecan-1 confers

Oncogene Role of glypican-1 in pancreatic cancer progression CA Whipple et al 2544 glypican-1 dependence on mitogenic responses of cancer cells. J Cell Li A, Cheng XJ, Moro A, Singh RK, Hines OJ, Eibl G. (2011). Biol 171: 729–738. CXCR2-dependent endothelial progenitor cell mobilization in Gengrinovitch S, Berman B, David G, Witte L, Neufeld G, Ron D. pancreatic cancer growth. Transl Oncol 4: 20–28. (1999). Glypican-1 is a VEGF165 binding proteoglycan that acts Matsuda K, Maruyama H, Guo F, Kleeff J, Itakura J, Matsumoto Y as an extracellular chaperone for VEGF165. J Biol Chem 274: et al. (2001). Glypican-1 is overexpressed in human breast cancer 10816–10822. and modulates the mitogenic effects of multiple heparin-binding Gu G, Brown JR, Melton DA. (2003). Direct lineage tracing reveals growth factors in breast cancer cells. Cancer Res 61: 5562–5569. the ontogeny of pancreatic cell fates during mouse embryogenesis. Neupane D, Korc M. (2008). 14-3-3sigma modulates pancreatic cancer Mech Dev 120: 35–43. cell survival and invasiveness. Clin Cancer Res 14: 7614–7623. Habbe N, Shi G, Meguid RA, Fendrich V, Esni F, Chen H et al. Olive KP, Jacobetz MA, Davidson CJ, Gopinathan A, McIntyre D, (2008). Spontaneous induction of murine pancreatic intraepithelial Honess D et al. (2009). Inhibition of Hedgehog signaling enhances neoplasia (mPanIN) by acinar cell targeting of oncogenic Kras in delivery of chemotherapy in a mouse model of pancreatic cancer. adult mice. Proc Natl Acad Sci USA 105: 18913–18918. Science 324: 1457–1461. Han JH, Stratowa C, Rutter WJ. (1987). Isolation of full-length Preis M, Korc M. (2010). Kinase signaling pathways as targets for putative rat lysophospholipase cDNA using improved methods for intervention in pancreatic cancer. Cancer Biol Ther 9: 754–763. mRNA isolation and cDNA cloning. Biochemistry 26: 1617–1625. Rowland-Goldsmith MA, Maruyama H, Kusama T, Ralli S, Korc M. Hingorani SR, Petricoin EF, Maitra A, Rajapakse V, King C, (2001). Soluble type II transforming growth factor-beta (TGF-beta) Jacobetz MA et al. (2003). Preinvasive and invasive ductal receptor inhibits TGF-beta signaling in COLO-357 pancreatic pancreatic cancer and its early detection in the mouse. Cancer Cell cancer cells in vitro and attenuates tumor formation. Clin Cancer 4: 437–450. Res 7: 2931–2940. Hwang RF, Moore T, Arumugam T, Ramachandran V, Amos KD, Rowland-Goldsmith MA, Maruyama H, Matsuda K, Idezawa T, Ralli Rivera A et al. (2008). Cancer-associated stromal fibroblasts M, Ralli S et al. (2002). Soluble type II transforming growth factor- promote pancreatic tumor progression. Cancer Res 68: 918–926. beta receptor attenuates expression of metastasis-associated genes Jackson EL, Willis N, Mercer K, Bronson RT, Crowley D, Montoya and suppresses pancreatic cancer cell metastasis. Mol Cancer Ther 1: R et al. (2001). Analysis of lung tumor initiation and progression 161–167. using conditional expression of oncogenic K-ras. Genes Dev 15: Schneider G, Schmid RM. (2003). Genetic alterations in pancreatic 3243–3248. carcinoma. Mol Cancer 2: 15. Jen YH, Musacchio M, Lander AD. (2009). Glypican-1 controls brain Seeley ES, Carriere C, Goetze T, Longnecker DS, Korc M. (2009). size through regulation of fibroblast growth factor signaling in early Pancreatic cancer and precursor pancreatic intraepithelial neoplasia neurogenesis. Neural Dev 4: 33. lesions are devoid of primary cilia. Cancer Res 69: 422–430. Johnson L, Greenbaum D, Cichowski K, Mercer K, Murphy E, Siegel R, Ward E, Brawley O, Jemal A. (2011). Cancer statistics, 2011: Schmitt E et al. (1997). K-ras is an essential gene in the mouse with the impact of eliminating socioeconomic and racial disparities on partial functional overlap with N-ras. Genes Dev 11: 2468–2481. premature cancer deaths. CA Cancer J Clin 61: 212–236. Kim I, Saunders TL, Morrison SJ. (2007). Sox17 dependence Suyama E, Kawasaki H, Nakajima M, Taira K. (2003). Identification distinguishes the transcriptional regulation of fetal from adult of genes involved in cell invasion by using a library of randomized hematopoietic stem cells. Cell 130: 470–483. hybrid ribozymes. PNAS 100: 5616–5621. Kleeff J, Beckhove P, Esposito I, Herzig S, Huber PE, Lohr JM Terris B, Blaveri E, Crnogorac-Jurcevic T, Jones M, Missiaglia E, et al. (2007). Pancreatic cancer microenvironment. Int J Cancer 121: Ruszniewski P et al. (2002). Characterization of gene expression 699–705. profiles in intraductal papillary-mucinous tumors of the pancreas. Kleeff J, Ishiwata T, Kumbasar A, Friess H, Buchler MW, Lander AD Am J Pathol 160: 1745–1754. et al. (1998). The cell-surface heparan sulfate proteoglycan glypican- Tuveson DA, Hingorani SR. (2005). Ductal pancreatic cancer in 1 regulates growth factor action in pancreatic carcinoma cells and humans and mice. Cold Spring Harb Symp Quant Biol 70: 65–72. is overexpressed in human pancreatic cancer. J Clin Invest 102: Vilar E, Salazar R, Perez-Garcia J, Cortes J, Oberg K, Tabernero J. 1662–1673. (2007). Chemotherapy and role of the proliferation marker Ki-67 Kleeff J, Wildi S, Kumbasar A, Friess H, Lander AD, Korc M. (1999). in digestive neuroendocrine tumors. Endocr Relat Cancer 14: Stable transfection of a glypican-1 antisense construct decreases 221–232. tumorigenicity in PANC-1 pancreatic carcinoma cells. Pancreas 19: Vlodavsky I, Korner G, Ishai-Michaeli R, Bashkin P, Bar-Shavit R, 281–288. Fuks Z. (1990). Extracellular matrix-resident growth factors and Korc M. (2003). Pathways for aberrant angiogenesis in pancreatic enzymes: possible involvement in tumor metastasis and angiogen- cancer. Mol Cancer 2:8. esis. Cancer Metastasis Rev 9: 203–226. Korc M. (2007). Pancreatic cancer-associated stroma production. Am Vonlaufen A, Phillips PA, Xu Z, Goldstein D, Pirola RC, Wilson JS J Surg 194: S84–S86. et al. (2008). Pancreatic stellate cells and pancreatic cancer cells: an Lander AD, Nie Q, Wan FY. (2002). Do morphogen gradients arise by unholy alliance. Cancer Res 68: 7707–7710. diffusion? Dev Cell 2: 785–796. Xu J. (2005). Preparation, culture, and immortalization of mouse Le Mercier M, Fortin S, Mathieu V, Roland I, Spiegl-Kreinecker S, embryonic fibroblasts. Curr Protoc Mol Biol Chapter 28: Unit 28 21. Haibe-Kains B et al. (2009). Galectin 1 proangiogenic and Zetser A, Bashenko Y, Edovitsky E, Levy-Adam F, Vlodavsky I, Ilan promigratory effects in the Hs683 oligodendroglioma model are N. (2006). Heparanase induces vascular endothelial growth factor partly mediated through the control of BEX2 expression. Neoplasia expression: correlation with p38 phosphorylation levels and Src 11: 485–496. activation. Cancer Res 66: 1455–1463.

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

Oncogene