Soluble Eph a Receptors Inhibit Tumor Angiogenesis and Progression in Vivo

Soluble Eph A receptors inhibit tumor angiogenesis and progression in vivo

Dana M Brantley1, Nikki Cheng2, Erin J Thompson1, Qing Lin1, Rolf A Brekken3, Philip E Thorpe4, Rebecca S Muraoka2, Douglas Pat Cerretti5, Ambra Pozzi1,2, Dowdy Jackson6, Charles Lin2,7 and Jin Chen*,1,2

1Department of Medicine, Vanderbilt University School of Medicine, Nashville, Tennessee, TN 37232, USA; 2Cancer Biology, Vanderbilt University School of Medicine, Nashville, Tennessee, TN 37232, USA; 3Department of Vascular Biology, The Hope Heart Institute, Seattle, Washington, WA 98122, USA; 4Department of Pharmacology, University of Texas-Southwestern Medical Center, Dallas, Texas, TX 95235, USA; 5Immunex Corporation, Seattle, Washington, WA 98101, USA; 6Amersham-Pharmacia Corporation, St. Louis, Missouri, MO 63198, USA; 7Department of Radiation Oncology, Vanderbilt University School of Medicine, Nashville, Tennessee, TN 37232, USA

The Eph family of receptor tyrosine kinases and their to be a critical step in the progression and metastasis of ligands, known as ephrins, play a crucial role in vascular solid tumors (Folkman, 1990, 1994; Weidner, 1996). development during embryogenesis. The function of these Recruitment of new blood vessels by tumors enables molecules in adult angiogenesis has not been well tumor survival and growth via delivery of oxygen and characterized. Here, we report that blocking Eph A class host nutrients to the tumor, thus promoting malignant receptor activation inhibits angiogenesis in two indepen- progression. Tumor vascular density has been corre- dent tumor types, the RIP-Tag transgenic model of lated with malignant progression and a poor prognosis angiogenesis-dependent pancreatic islet cell carcinoma for patients suffering from breast and pancreatic and the 4T1 model of metastatic mammary adenocarcino- carcinoma (Goede et al., 1998; Kuehn et al., 1999; ma. Ephrin-A1 ligand was expressed in both tumor and Weidner et al., 1991, 1992). Indeed, newly formed endothelial cells, and EphA2 receptor was localized tumor vasculature often displays breaches in basement primarily in tumor-associated vascular endothelial cells. membrane integrity, which may permit invasion by Soluble EphA2-Fc or EphA3-Fc receptors inhibited tumor tumor cells and thereby facilitate metastasis (Dankort angiogenesis in cutaneous window assays, and tumor and Muller, 1996; Weidner, 1996). Therefore, under- growth in vivo. EphA2-Fc or EphA3-Fc treatment resulted standing the molecular mechanisms that regulate tumor in decreased tumor vascular density, tumor volume, and angiogenesis will enhance our understanding of tumor cell proliferation, but increased cell apoptosis. However, progression, and provide novel targets for therapeutic EphA2-Fc had no direct effect on tumor cell growth or intervention in cancer. apoptosis in culture, yet inhibited migration of endothelial Receptor tyrosine kinases (RTKs) have emerged as cells in response to tumor cells, suggesting that the soluble critical mediators of angiogenesis (reviewed in Cheng et receptor inhibited blood vessel recruitment by the tumor. al., 2002; Gale and Yancopoulos, 1999; Yancopoulos These data provide the first functional evidence for Eph A et al., 2000). For example, the vascular endothelial class receptor regulation of pathogenic angiogenesis growth factor (VEGF) family of ligands and their induced by tumors and support the function of A class RTKs function in endothelial cell differentiation and Eph receptors in tumor progression. blood vessel patterning during embryonic development, Oncogene (2002) 21, 7011 – 7026. doi:10.1038/sj.onc. and are also required for tumor angiogenesis. The Eph 1205679 family of RTKs and their ligands, known as ephrins, were originally identified as critical determinants of Keywords: tumor angiogenesis; Eph receptor tyrosine embryonic patterning and neuronal targeting (Holder kinase; ephrin; soluble Eph receptor; vascular window and Klein, 1999). These molecules also regulate model embryonic vascular development (Gale and Yancopou- los, 1999; Yancopoulos et al., 2000). Targeted Introduction disruption of, ephrinB2, EphB2/EphB3 or EphB4, results in embryonic lethality due to defects in primary Angiogenesis, the process by which new blood vessels capillary network remodeling and subsequent pattern- are formed from existing vasculature, has been shown ing defects in the embryonic vasculature (Adams et al., 1999a; Gerety et al., 1999; Wang et al., 1998b), suggesting that Eph RTKs and their ligands are critical for vascular development during embryogenesis. The A *Correspondence: J Chen, Vanderbilt University School of Medicine, class ligand, ephrin-A1, was originally identified as a A-4323 MCN, 1161 21st Avenue South, Nashville, Tennessee, TN 37232, USA; E-mail: [email protected] TNF-a-inducible gene in human umbilical vein Received 5 February 2002; revised 15 May 2002; accepted 20 May endothelial cells (HUVEC) (Holzman et al., 1990), 2002 and is expressed in the developing vasculature during Eph A receptors and tumor angiogenesis DM Brantley et al 7012 embryogenesis (McBride and Ruiz, 1998). Moreover, arrowhead), EphA2 protein appeared to be specifically ephrin-A1 induces endothelial cell migration and localized in tumor associated vasculature and was not capillary assembly in vitro, and angiogenesis in the detected in carcinoma cells (Figure 1c, arrowhead), corneal pocket assay in vivo (Daniel et al., 1996; Myers consistent with the data obtained from immunoblot et al., 2000; Pandey et al., 1995). These studies indicate analysis (Figure 1a). Localization of ephrin-A1 and that Eph signaling is critical for normal blood vessel EphA2 in tumor endothelial cells was also confirmed development, and suggest that these molecules may by dual immunofluorescence expression analysis using also play a role in angiogenesis caused by pathologic the endothelial cell specific marker CD31 (Figure 1h – m). stages in the adult. As RIP-Tag tumor angiogenesis is dependent on In this study, we report complementary expression of signaling by vascular endothelial growth factor ephrin-A1 ligand in tumor cells and EphA2 receptor in (VEGF; Bergers et al., 2000), and as soluble EphA2- tumor associated blood vessel endothelium. Soluble Fc receptor can inhibit VEGF-induced angiogenesis in EphA-Fc receptors inhibited tumor angiogenesis in corneal assays (Cheng et al., submitted), we analysed cutaneous window assays and tumor progression in expression of EphA2 and ephrin-A1 relative to VEGF vivo. Moreover, soluble Eph receptors inhibited tumor- in adjacent tumor sections (Figure 1n – p). VEGF and induced endothelial cell migration in culture. These ephrin-A1 were abundantly expressed in islet cell data provide the first evidence for A class Eph receptor carcinomas in overlapping patterns (Figure 1o,p). The signaling in tumor angiogenesis, and suggest that complementary patterns of ephrin-A1 and EphA2 targeting these molecules may be an effective ther- proteins in tumor cells and associated blood vessels apeutic strategy in cancer treatment. suggest that functions of EphA ligand/receptor may be required to promote tumor neovascularization. Results Soluble EphA2-Fc receptor inhibits RIP-Tag tumor- Ephrin-A1 and EphA2 are expressed in RIP-Tag tumors induced angiogenesis and associated vasculature To determine if class A Eph receptor/ligand regulates To determine if Eph A class receptors and ligands play tumor angiogenesis, we used EphA2-Fc as an inhibitor a role in tumor angiogenesis, we first examined the in an in vivo cutaneous tumor vascular window assay expression and localization of ephrin-A1 and EphA2 (Figure 2a). In this assay system, a fold is secured in proteins in vascularized tumor tissue. We focused on the dorsal skin of a mouse and a circular portion of this ligand and receptor pair as previous studies have skin is removed from one side of the fold. This creates correlated ephrin-A1 and EphA2 overexpression in a ‘window’ that exposes the subcutaneous blood vessels several types of human tumors (Ogawa et al., 2000). on the inner flap of skin on the opposite side of the The RIP-Tag transgenic model of multi-stage carcino- fold. This window preserves the exposed opposite genesis, in which the rat insulin promoter drives surface for observation through a glass coverslip, expression of the SV40 large T antigen oncogene in which is fastened on top of the subcutaneous skin pancreatic b-islet cells, was chosen for this study since surface. Previous studies have demonstrated that small these animals undergo tumorigenesis in a reproducible tumors placed within the dorsal window chamber series of stages, one of which involves the initiation of become vascularized by subcutaneous host blood angiogenesis (Folkman et al., 1989; Hanahan, 1985). vessels and undergo rapid growth within 10 – 14 days We first examined the expression of ephrin-A1 and (Huang et al., 1999; Li et al., 2000; Papenfuss et al., EphA2 protein in the bTC cell line, a pancreatic islet 1979). This system enabled us to monitor the cell carcinoma cell line that was originally derived from angiogenic response of the subcutaneous host blood RIP-Tag tumors (Radvanyi et al., 1993) and in the vessels initiated by tumor-derived signals in live MS-1 pancreatic microvascular endothelial cell line animals by photomicroscopy. (Arbiser et al., 1997) by immunoblot analysis. We In order to ascertain the role of Eph A class receptor detected abundant ephrin-A1 expression in both tumor function in tumor angiogenesis, we utilized a soluble and endothelial cell lines, though EphA2 expression chimeric protein (EphA2-Fc) in which the extracellular was restricted to endothelial cells (Figure 1a). domain of the EphA2 receptor is fused to the human To verify these expression patterns within intact IgG1 Fc chain. We have shown that EphA2-Fc RIP-Tag tumors in vivo, anti-ephrin-A1 antibodies prevents the interaction of multiple ephrin A class were used to stain RIP-Tag islet cell carcinoma ligands with endogenous receptors, effectively blocking sections. Ephrin-A1 protein was expressed predomi- A class Eph receptor activation in cell culture, and nantly within the carcinoma cells (Figure 1b,e). inhibiting corneal angiogenesis in vivo (Cheng et al., Expression of ephrin-A1 was also localized within the submitted). endothelium of some tumor-associated blood vessels by As shown in Figure 2, small islet cell carcinomas comparison with adjacent sections stained for CD31, were isolated from RIP-Tag mice and placed in the an endothelial cell-specific marker (Figure 1e,f, arrow- window chamber (outlined in blue). Hydron pellets head). Ephrin-A1 expression was not detected in (outlined in yellow) impregnated with EphA2-Fc, adjacent exocrine tissue (Figure 1b,e, arrows). In human IgG (Fc control), or PBS were implanted contrast, compared with CD31 expression (Figure 1d, adjacent to islet tumors in window chambers mounted

Oncogene Eph A receptors and tumor angiogenesis DM Brantley et al 7013 on syngeneic animals (Figure 2a – i). After 10 – 14 days pellets elicited a robust angiogenic response (Figure the islet tumors co-implanted with control IgG or PBS 2b – e). To determine if tumor associated vessels were

Figure 1 Ephrin-A1 and EphA2 are expressed in pancreatic islet cell carcinoma and/or associated tumor vasculature. (a) Total cellular protein (60 mg) from the RIP-Tag tumor-derived bTC cell line and from the MS-1 pancreatic microvascular endothelial cell line was subjected to immunoblot analysis for expression of ephrin-A1 and EphA2. Expression of ephrin-A1 was detected in both tumor and endothelial cells, while EphA2 expression was restricted to endothelial cells. (b – g) Adjacent pancreatic sections (10 mm) from tumor-bearing Rip-Tag mice were subjected to immunohistochemistry to detect the expression of ephrin-A1, EphA2, and CD31. Ephrin-A1 is expressed in RIP-Tag islet tumor cells (b,e) and in tumor associated endothelial cells (e, arrowhead), and EphA2 is expressed in blood vessel endothelial cells (c, arrowheads), as compared to an adjacent section stained with CD31 (d, arrowheads). Expression of ephrin-A1 was not observed in adjacent exocrine tissue (b,e, arrows). (h – m) Expression of ephrin-A1 and EphA2 relative to CD31 endothelial cell marker was assessed by dual immunoﬂuorescence analysis. (h) Ephrin-A1 expression (red) was predominantly in tumor cells surrounding (i) CD31 (green) expressing endothelial cells, although there was some co-localization within endothelial cells (j, merged image, arrows). (k) EphA2 expression (red) co-localized with (l) CD31 expression (green) in endothelial cells (m, merged image, arrowheads). (n – p) Ephrin-A1 expression in RIP-Tag islet cell tumors co-localizes with VEGF expression in 5 mm adjacent sections. n=3 to 6 independent samples. Scale bars: (b) and (n), 100 mm, (e), (h) and (k), 50 mm

Oncogene Eph A receptors and tumor angiogenesis DM Brantley et al 7014 functional and to visualize microvessels associated with visualized under the digital camera. In contrast, no the tumor, FITC-conjugated dextran was given visible, functional blood vessels were observed in islet intravenously. The vessels associated with RIP-Tag tumors adjacent to EphA2-Fc pellets (Figure 2f – i). tumors adjacent to PBS and IgG-impregnated pellets Vessel density within the tumor was quantified using were perfused well, as they were efficiently labeled with Scion Image (Figure 2j). A 28-fold decrease in blood fluorescent dextran (Figure 2c,d) and blood flow was vessel density within the tumor was observed in

Figure 2 Soluble EphA2-Fc receptor inhibits RIP-Tag tumor angiogenesis in vivo.(a) Angiogenic islet cell tumors were isolated from RIP-Tag mice and implanted into the dorsal window chamber of syngeneic mice. Hydron pellets impregnated with control human IgG or EphA2-Fc (outlined in yellow) were implanted into the chambers adjacent to tumors (arrowheads b and f; outlined in blue). (b – e) Tumors adjacent to IgG pellets elicited a robust angiogenic response 12 days after implantation. The tumor blood vessels connected with the host vasculature, as demonstrated by FITC-dextran (green) administered intravenously (c,26Green Field panel; d, higher magniﬁcation of area outlined in blue in c). (f – i) In contrast, a greatly reduced angiogenic response was observed in tumors implanted adjacent to EphA2-Fc pellets. Arrows indicate normal subcutaneous host blood vessels peripheral to the tumor (c, g). (j) A 28-fold decrease in blood vessel density within the vicinity of the tumor was observed in EphA2-Fc treated mice (n=8) versus PBS/IgG-treated controls (n=8) (P50.05, two-tailed, paired Student’s t-test). Scale bars: b, 1 mm (26); d, 200 mm, (106)

Oncogene Eph A receptors and tumor angiogenesis DM Brantley et al 7015 EphA2-Fc-treated mice versus controls (P50.01). To sion was detected in tumor cells surrounding blood verify that this effect was specific for EphA2-Fc protein vessels (Figure 3s,t, arrowheads indicate location of and not potential co-purified contaminants, two tumor blood vessels). We compared the expression of batches of independently purified soluble receptors EphA2 with activated VEGF receptor (VEGF : R) were tested. Both batches efficiently blocked tumor using an antibody that specifically recognizes VEGF angiogenesis compared to control proteins. These data in complex with VEGFR-1 or VEGFR-2 (Bergers et suggest a requirement for EphA receptor/ligand al., 2000; Brekken et al., 1998). EphA2 expression co- function in tumor angiogenesis. localized with activated VEGF : R in tumor microvessels, as demonstrated by comparison of immunostaining in 5 mm adjacent sections (Figure Ephrin-A1 and EphA2 are expressed in 4T1 tumors and 3v,w, arrowheads indicate location of tumor blood associated vasculature vessels). These overlapping expression patterns of To determine whether our observations are relevant for ephrin-A1 and VEGF ligands, as well as EphA2 and other tumor types, we investigated the role of class A activated VEGF receptors, suggest that class A Eph Eph receptors and ligands in tumor angiogenesis in ligand/receptor may also function in concert with 4T1 mammary adenocarcinoma. We selected the 4T1 VEGF to regulate 4T1-induced tumor angiogenesis. model due to the ability of transplanted 4T1 cells to rapidly and reproducibly generate mammary adeno- Soluble EphA2-Fc inhibits 4T1 tumor cell-induced carcinoma in syngeneic animals in vivo (Aslakson and angiogenesis Miller, 1992; Lin et al., 1998). Expression of ephrin-A1 and EphA2 in EMT6 cells was also analysed, as these We examined the function of A class Eph receptors/ tumor cells also produce malignant adenocarcinomas in ligands in 4T1 tumor cell-mediated blood vessel vivo, though with a slower progression than 4T1 cells recruitment and tumor progression by using EphA2- (Rockwell, 1977, 1978, 1981). As shown in Figure 3a, Fc in the cutaneous tumor vascular window assay. 4T1 ephrin-A1 was barely detectable in normal primary cells have been shown to elicit a robust angiogenic mouse mammary epithelial cell (PMEC) lysates, but response in this assay system (Huang et al., 1999; Li et was expressed at high levels in 4T1 and EMT6 tumor al., 2000). Hydron pellets (outlined in yellow) impreg- cells. EphA2 overexpression was observed in 4T1 and nated with PBS vehicle, control human IgG, or EphA2- EMT6 tumor cell lysates relative to cultured PMECs. Fc were co-implanted into the dorsal skin fold of mice Overexpression of both ephrin-A1 and EphA2 was also along with 4T1-GFP cells (green). The stable expression detected in 4T1-GFP cells, a 4T1 subline expressing of GFP within this 4T1 subline provides a means of green fluorescent protein (Li et al., 2000), used in tracking tumor cells in the window chamber (Li et al., subsequent functional analyses. 2000). After 10 – 14 days, rhodamine-conjugated Next, we examined expression and localization of dextran was given intravenously to visualize host blood ephrin-A1 and EphA2 proteins in tissue sections from vessel perfusion and small microvessels by fluorescence 4T1 tumors produced by injection of 4T1 cells into the microscopy. Abundant new vessel growth was detected mammary gland fat pad of female syngeneic Balb/c adjacent to 4T1 tumor cells in control windows (Figure mice. In 4T1 tumor sections, ephrin-A1 was detected in 4h,l), but not in windows treated with EphA2-Fc, even tumor cells throughout the course of tumor progres- in regions containing viable cells (see Figure 4c,d). sion from 3 to 9 days post-transplantation (Figure 3b – Intensity of rhodamine fluorescence was quantified as a i, arrowheads indicate tumor blood vessels). Expression measure of vessel density using Scion Image software. of ephrin-A1 appeared to increase steadily from day 3 A 120-fold decrease in blood vessel density within the to day 9 (Figure 3b – d). Although EphA2 expression vicinity of viable tumor cells was observed in EphA2-Fc was detected in tumor cells at lower levels (Figure 3h), treated mice versus controls (P50.01 Figure 4m). When it was predominantly expressed in tumor-associated tumor cell density was quantified, a fourfold reduction peripheral blood vessels and in internal microvessels 6 in 4T1-GFP tumor cell density in windows treated with days after injection of 4T1 cells (Figure 3f,i, arrow- EphA2-Fc, as compared with those treated with IgG or heads), and expression persisted up to 9 days post- PBS controls (Figure 4n) (P50.01). These data suggest injection (Figure 3g, arrowhead). Co-localization of that 4T1 tumor angiogenesis requires Eph A class ephrin-A1/EphA2 expression with endothelial cells was receptor/ligand function. confirmed by dual immunofluorescent staining with CD31 (Figure 3j – r). Soluble EphA2 receptors inhibit tumor progression in vivo As our previous studies indicate that VEGF- mediated angiogenesis can be inhibited by soluble To determine the impact of EphA class receptor EphA2-Fc receptor (Cheng et al., submitted), and as activation on tumor progression in vivo, we monitored blocking VEGF signaling abrogates tumor 4T1 tumor tumor angiogenesis, as well as tumor cell proliferation angiogenesis (Li et al., 2000; Prewett et al., 1999), we and apoptosis, in 4T1 tumors treated with EphA2-Fc examined the expression patterns of ephrin-A1 and or EphA3-Fc versus control IgG. EphA3-Fc also EphA2 relative to VEGF and activated VEGF receptor inhibits EphA class receptor activation in response to in adjacent 4T1 tumor sections (Figure 3s – x). Co- ephrin stimulation (data not shown), but mutation of localization of ephrin-A1 and VEGF protein expres- the Fc portion of the chimeric protein blocks

Oncogene Eph A receptors and tumor angiogenesis DM Brantley et al 7016 interaction with Fc receptors and complement (WC suspended in liquid Matrigel in the presence of 10 mg Fanslow, personal communication). 4T1 cells were EphA2-Fc, EphA3-Fc, or control human IgG, and the

Figure 3 Ephrin-A1 and EphA2 are overexpressed in metastatic mammary adenocarcinoma cell lines. (a) Total cellular protein lysates (40 mg) were fractionated and subjected to Western blot analysis. Ephrin-A1 and EphA2 expression levels were elevated in 4T1 and EMT6 cells relative to primary mammary epithelial cells (PMEC) from syngeneic Balb/c female mice. Data are a representation of five independent samples per cell line. (b – i) Sections from 4T1 tumors isolated 3, 6, and 9 days post-injection into Balb/c female mammary fat pads were probed for expression of ephrin-A1 and EphA2. (b – d) Ephrin-A1 expression was observed within tumor cells at all time points, including tumor cells surrounding tumor blood vessels (arrowheads). (e – g) EphA2 expression was detected as early at 6 days post-injection within peripheral tumor blood vessels (arrowheads), and expression persisted up to 9 days. (h and i) EphA2 expression was also detected within tumor cells (h) and in microvessels within the tumor (i, arrowheads) at 6 days post-injection. (j – r) Cryosections (7 mm) from 6 day 4T1 tumors were subjected to dual immunofluorescence analysis for expression of ephrin-A1 or EphA2 and the endothelial specific marker CD31. Ephrin-A1 (j) protein expression (red) was predominantly expressed in tumor cells, though some co-localization with CD31 (k) protein expression (green) in associated tumor blood vessels was detected (l, yellow in merged image, arrowheads). EphA2 (m) protein expression (red) localized to tumor cells, and also co-localized with CD31 (n) protein expression (green) in associated tumor blood vessels in the same section (o, yellow in merged image, arrowheads). (p – r) Staining for ephrin-A1, EphA2, and CD31 was specific, as fluorescence conjugated secondary antibodies displayed minimal background. (s – x) Adjacent sections (5 mm) from 4T1 tumors 6 days post-injection were probed for expression of ephrin-A1, VEGF, EphA2, and activated VEGF receptor (VEGF : R) in complex with VEGF. Ephrin-A1 expression (s) colocalized with VEGF (t) expression in tumor cells surrounding blood vessels (arrowheads), and EphA2 expression (v) co-localized with activated VEGF receptor (w). Specificity of staining was confirmed by probing adjacent sections with control rabbit IgG (u, rIgG) or mouse IgG (x, mIgG). Arrowheads indicate blood vessels (n=3 independent samples per condition). Scale bars: b and s, 100 mm; h, 50 mm; i, j, and v,5mm

Oncogene Eph A receptors and tumor angiogenesis DM Brantley et al 7017 mixture was injected into female mice subcutaneously. and in microvessel density within the tumor, as The mice were treated daily with 10 mg soluble factors determined by CD31 staining (Figure 5e versus f, by subcutaneous injection at the site of the tumor for arrowheads indicate microvessels). We observed a 2.5- 14 days, and tumors were then ressected for analysis. fold decrease in microvascular density in EphA2-Fc- Treatment with EphA2-Fc or EphA3-Fc resulted in a treated 4T1 tumors versus IgG-treated controls, as threefold reduction in tumor volume relative to 4T1 quantified by CD31 fluorescence intensity (P50.05, tumors treated with IgG (P50.05; Figure 5i). Figure 5h). Treatment of Matrigel plugs with EphA2- Furthermore, 4T1 cells elicited a robust angiogenic Fc resulted in significantly reduced tyrosine phosphor- response in Matrigel plugs, whereas Matrigel by itself ylation of endogenous EphA2 relative to IgG, did not induce any angiogenic response (Figure 5a suggesting that these soluble receptors disrupt EphA2 versus d). Consistent with data derived from the signaling in vivo (Figure 5j). Taken together, these data window assays, treatment with EphA2-Fc resulted in suggest that class A Eph ligands and receptors are a marked reduction in both surface vascular density required in 4T1 tumor angiogenesis and progression in (Figure 5a versus b, arrowheads indicate blood vessels) vivo.

Figure 4 Soluble EphA2-Fc receptor inhibits 4T1 cell-induced angiogenesis in vivo.(a – l) 4T1-GFP cells (green, approximately 1000) were injected into cutaneous window chambers harboring pellets (yellow lines) impregnated with EphA2-Fc or IgG and PBS as controls. After 10 to 14 days, the mice were injected intravenously with rhodamine-dextran (Red) to assess blood vessel perfusion. Windows treated with control IgG (e – h) or PBS (i – l) displayed abundant 4T1-GFP tumor density (f and j), and a robust angiogenic response (h and l). Windows treated with EphA2-Fc (a – d) displayed a decrease in tumor cell density (b) and a marked inhibition of tumor angiogenesis in regions with viable tumor cells (d). White boxes in 26green field panel indicate regions photographed at higher magnification in 106green and red fields, showing tumor cells and blood vessels, respectively. Arrowheads indicate normal subcutaneous host blood vessels (a, e, and i). (m) A 120-fold decrease in blood vessel density within the vicinity of the tumor cells was observed in EphA2-Fc treated mice (n=7) versus PBS/IgG-treated controls (n=7) (P50.01, two-tailed, paired Stu- dent’s t-test). (n) The decrease in tumor cell density in windows treated with EphA2-Fc versus IgG and PBS controls was quantified by GFP fluorescence intensity. A fourfold decrease in 4T1-GFP fluorescence intensity was observed in windows treated with EphA2- Fc (P50.007, two-tailed, paired Student’s t-test). Scale bars: b, 1 mm; c, 200 mm

Oncogene Eph A receptors and tumor angiogenesis DM Brantley et al 7018 The decrease in vascular density and tumor volume inhibitory effect of soluble EphA receptor was specific in EphA2-Fc/EphA3-Fc treated 4T1 Matrigel plugs to tumor angiogenesis, rather than growth or survival may be due to direct inhibition of tumor angiogenesis. of tumor tissue, we compared levels of cell growth and Alternatively, EphA-Fc may affect tumor cell growth cell death in 4T1 tumors with that of cultured 4T1 cells and/or induce tumor cell death thereby indirectly in vitro. 4T1 cell growth in vivo was assessed by inhibiting tumor angiogenesis. To determine if the quantifying expression of proliferating cell nuclear

Figure 5 Soluble EphA2-Fc and EphA3-Fc receptors inhibit 4T1 tumor progression in vivo.(a – d) Growth factor reduced Matrigel impregnated with 4T1 tumor cells in the presence of IgG control, EphA2-Fc, or EphA3-Fc was injected subcutaneously into the dorsal flank of Balb/c female mice. After 2 weeks, tumors were isolated and analysed for vascular density and tumor volume. 4T1 tumors treated with control IgG (a) displayed abundant surface vessels as compared to growth factor reduced Martrigel alone (d). In contrast, surface vessel density was reduced in tumors treated with EphA2-Fc (b) or EphA3-Fc (c). Microvascular density in tumors was visualized by immunofluorescence analysis for expression of the endothelial marker CD31 in tumor sections (e – g). Tu- mors treated with EphA2-Fc (f) or EphA3-Fc (g) displayed a reduction in microvessel density (green) compared to IgG-treated tumors (e; arrowheads indicated microvessels). (h) A 2.5-fold decrease in CD31 fluorescence intensity was observed in tumors treated with EphA2-Fc or EphA3-Fc (P50.05, two-tailed, paired Student’s t-test). (i) Treatment with EphA2-Fc or EphA3-Fc resulted in a threefold decrease in tumor volume compared to IgG-treated tumors (P50.05, two-tailed, paired Student’s t-test). (j) EphA2 was immunoprecipitated from IgG and EphA2-treated tumors and tyrosine phosphorylation was assessed by immunoblot. EphA2-Fc treatment substantially inhibited tyrosine phosphorylation of endogenous EphA2 relative to IgG control. Scale bars: a, 3.5 mm, d, 100 mm(n=5 to 10 animals per treatment group)

Oncogene Eph A receptors and tumor angiogenesis DM Brantley et al 7019 antigen (PCNA), a marker for actively dividing cells, We also compared levels of 4T1 cell apoptosis in including breast carcinoma cells (Aaltomaa et al., 1993; Matrigel plugs versus 4T1 cells cultured in the presence Preziosi et al., 1995). Tumors treated with EphA2-Fc or absence of EphA2-Fc or EphA3-Fc by TUNEL or EphA3-Fc displayed a 2.2-fold reduction in PCNA assay. 4T1 tumors treated with EphA2-Fc or EphA3- expression relative to control IgG-treated tumors Fc displayed a 3.2-fold increase in TUNEL positive (Figure 6b versus a, arrowheads indicate PCNA nuclei; data for EphA3-Fc not shown), as determined positive nuclei; data for EphA3-Fc not shown), as by quantification of TUNEL positive nuclei relative to determined by quantification of PCNA positive nuclei control IgG-treated tumors (Figure 6f versus e, arrow- relative to total nuclei/field (P50.01; Figure 6c). In heads indicate TUNEL positive nuclei relative to total contrast, treatment with EphA2-Fc did not affect nuclei/field (P50.05, Figure 6g). We then measured serum-induced 4T1 cell growth (Figure 6d), suggesting apoptosis in 4T1 cells cultured in serum-free media as a that blocking EphA-Fc does not directly affect tumor positive control for the induction of apoptosis in these cell proliferation. cells, or in full-serum media in the presence or absence

Figure 6 Soluble EphA-Fc receptors affect tumor cell proliferation and apoptosis in vivo.(a and b) Proliferation of 4T1 tumor cells in vivo was assessed by expression of proliferating cell nuclear antigen (PCNA). Fewer PCNA positive nuclei were observed in tumors treated with EphA2-Fc (b) versus control IgG (a; arrowheads indicate PCNA positive nuclei). The percentage of PCNA positive nuclei relative to total nuclei was determined, and a 2.2-fold reduction was observed in EphA2-Fc versus IgG-treated tumors (P50.01, two-tailed, paired Student’s t-test; c). EphA2-Fc did not affect the growth of 4T1 tumor cells in culture (d). Apoptosis in 4T1 tumor cells in vivo was assessed by TUNEL analysis. More TUNEL positive nuclei (red) were observed in tumors treated with EphA2-Fc (f) versus control IgG (e; arrowheads indicate TUNEL positive nuclei). Total nuclei in tumor sections were visualized by DAPI staining (blue). The number of pixels in TUNEL positive nuclei relative to the number of DAPI pixels (total nuclei) was quantified by Scion Image software analysis. A 3.2-fold increase in TUNEL positive nuclei was observed in EphA2-Fc treated tumors versus controls, as determined by quantification of TUNEL positive nuclei relative to total nuclei/field (P50.05, two-tailed, paired Student’s t-test; g). EphA2-Fc did not affect 4T1 tumor cell apoptosis in culture (h). Scale bars: a and e, 100 mm. (n=five independent samples/condition for in vivo analysis and three independent experiments/condition for cell culture analysis)

Oncogene Eph A receptors and tumor angiogenesis DM Brantley et al 7020 of EphA2-Fc. In contrast, treatment with EphA2-Fc at the level of endothelial cell migration, providing did not aﬀect basal apoptosis in these cells (Figure 6h), further support for a pro-angiogenic role for Eph A class suggesting soluble EphA-Fc did not aﬀect tumor cell receptors in tumor progression. survival in vitro.

Discussion Soluble EphA2-Fc inhibits endothelial cell migration in response to tumor tissue The functions of Eph receptor tyrosine kinases and To determine how EphA2-Fc affects endothelial cell their ligands during embryonic development have been function, we investigated the effect of EphA2-Fc on intensively studied. In addition to critical roles in axon endothelial cell proliferation, apoptosis, and migration. guidance and tissue boundary formation, certain Eph Consistent with previous studies (reviewed by Gale and receptors and their ligands are also required for Yancopoulos, 1999; blocking EphA receptor activation vascular development during embryogenesis. However, did not affect endothelial cell proliferation (data not the role of the Eph family in neovascularization shown). Ephrin A-1 ligand has been shown to regulate during pathologic states in adults remains unclear. endothelial cell migration (Daniel et al., 1996; Pandey et In this study, we provide the first functional data al., 1995), suggesting that overexpression of ephrin-A1 showing that class A Eph RTKs and ephrins play a in tumor tissue could promote migration and infiltration critical role in tumor angiogenesis. Ephrin-A1 ligand of endothelial cells in tumor tissue. To test this, we first is expressed predominantly in tumor cells, and EphA2 examined migration of microvascular endothelial cells in receptors are expressed primarily in a complementary response to soluble ephrin-A1 in a modified Boyden fashion in the tumor-associated vasculature. Further- chamber assay. Ephrin-A1 induced migration of bovine more, treatment with soluble EphA2-Fc chimeric pulmonary microvascular endothelial cells (BPMECs), receptor results in decreased neovascularization in consistent with previous studies performed using other two different tumor types in vascular window assays, endothelial cell types (Daniel et al., 1996; Pandey et al., and EphA2-Fc or EphA3-Fc treatment results in 1995). Excess EphA2-Fc inhibited ephrin-A1 induced decreased tumor vascular density and impaired tumor endothelial cell migration (P50.01, Figure 7a), and the progression in vivo. These data indicate that EphA inhibition was specific, as excess Fc control protein, IgG, signaling is a critical component for tumor progres- did not inhibit ephrin-A1-induced migration. Soluble sion, and that regulation of tumor neovascularization EphA2-Fc alone did not influence endothelial cell may play a significant role in EphA mediated tumor migration (data not shown; Cheng et al., submitted). progression. Since our previous studies indicate that VEGF-mediated Our expression data demonstrated that ephrin-A1 angiogenesis can be inhibited by soluble EphA2-Fc is elevated in RIP-Tag islet cell carcinoma and in (Cheng et al., submitted), and as blocking VEGF 4T1 mammary adenocarcinoma, and to a certain signaling abrogates RIP-Tag and 4T1 tumor angiogen- extent, is also expressed in tumor associated vessels. esis (Li et al., 2000; Prewett et al., 1999), we also tested In contrast, EphA2 receptor is expressed in both the ability of EphA2-Fc to inhibit VEGF induced endothelial cells and in tumor cells, depending on endothelial cell migration. VEGF-induced BPMEC tumor type. Specificity of immunoreaction was migration was inhibited by EphA2-Fc (P50.01, Figure confirmed by probing adjacent sections with normal 7a), suggesting that VEGF and ephrin signaling path- mouse IgG (Figure 1g) or rabbit IgG. Moreover, ways cooperate to mediate endothelial cell migration. anti-ephrin-A1 antisera specificity was confirmed We next assessed BPMEC migration in response to using two different antibodies and competition tumor cells in Boyden chamber co-culture assays experiments using purified ephrin-A1 protein (data (Laferriere et al., 2001). 4T1 and bTC cells were plated not shown). The EphA2 antibody has been used on the lower surface of Boyden chambers and labeled previously to detect expression in tumor tissue with FITC (Figure 7b – f). BPMECs were labeled with (Walker-Daniels et al., 1999), and specificity was Texas-red and added to the upper chamber, and confirmed by immunoblot (data not shown). A migration and intercalation of endothelial cells was number of studies reported overexpression of EphA2 quantified by fluorescence intensity (Figure 7g – k). Both receptor and/or ephrin-A1 in tumors, but in most bTC and 4T1 cells stimulated migration of BPMECs cases the cellular distribution in vivo has not been (Figure 7g,j, arrowheads indicate endothelial cells; characterized (Easty et al., 1995, 1999; Walker- Figure 7l,m). The induction of migration appeared to Daniels et al., 1999; Zantek et al., 1999). Consistent be specific to tumor cells, as non-malignant NMuMg with our observations, more recent studies report that (Owens et al., 1974; Van den Broecke et al., 1996) cells did expression of ephrin-A1 and EphA2 co-localized with not induce BPMEC migration (Figure 7i,m). Addition of tumor tissue and/or associated tumor vasculature in EphA2-Fc inhibited migration of BPMECs in response human tumor xenografts and biopsies from primary to both 4T1 and bTC (P50.01, Figure 7k,h,l,m), tumors (Ogawa et al., 2000; Zelinski et al., 2001). demonstrating that Eph A class receptor/ligand function Taken together, these studies and our findings suggest is necessary for endothelial cell migration in response to that the functional data presented in this study may tumor cells. These data suggest that EphA2-Fc specifi- represent a broad mechanism by which EphA class cally affects the angiogenic response initiated by tumors receptor/ligand regulates tumor progression.

Oncogene Eph A receptors and tumor angiogenesis DM Brantley et al 7021 Elucidation of the in vivo functions of Eph ligands we have utilized a soluble chimeric receptor fusion and receptors in adult angiogenesis is complicated by protein, EphA2-Fc, in which the extracellular ligand- functional redundancy within the Eph family and binding domain of the EphA2 receptor is fused to the embryonic lethality resulting from targeted disruption Fc portion of human IgG1. This soluble receptor of some individual Eph family members (Adams et al., eﬀectively blocks endogenous receptor phosphorylation 1999b; Chen and Ruley, 1998; Gerety et al., 1999; by competitively binding available ligand and inhibits Wang et al., 1998a). To circumvent these diﬃculties, ephrin-A1-induced corneal angiogenesis in vivo (Cheng

Figure 7 Soluble EphA2-Fc receptor inhibits endothelial cell migration in response to tumor cells. (a) BPMEC endothelial cell migration was assessed using a modified Boyden chamber assay. Endothelial cells were stimulated with ephrin-A1 or VEGF in the presence or absence of excess EphA2-Fc. EphA2-Fc inhibited endothelial cell migration in response to both ephrin-A1 and VEGF. Control human IgG did not affect cell migration in response to ephrin-A1. (b – m) FITC-labeled tumor cells were cultured with Texas-red labeled BPMECs in Boyden chambers in the presence or absence of EphA2-Fc or control hIgG, and endothelial cell migration in response to tumor cells was quantified. bTC, 4T1, or non-malignant NMuMg control cells were seeded on the underside of transwell filters and labeled with FITC-ovalbumin (b – f). Texas-red ovalbumin-labeled BPMECs were then added to the upper chamber and migration was assessed after 5 h (g – k). Both bTC and 4T1 cells induced migration of endothelial cells (g and j), arrowheads indicate endothelial cells; l and m), and the migration was specific to malignant cells as NMuMg cells did not stimulate endothelial cell migration (d and i). Migration of BPMECs in response to bTC and 4T1 cells was inhibited by EphA2-Fc (h and k; l and m, P50.01, two-tailed, paired Student’s t-test), but not by IgG. Scale bar: 100 mm, b (n=six to nine independent samples/condition)

Oncogene Eph A receptors and tumor angiogenesis DM Brantley et al 7022 et al., submitted). The inhibitory effects of soluble conditioned media from HUVECs after stimulation EphA2-Fc is specific to class A Eph receptor with TNF-a (Holzman et al., 1990). In addition, activation, as EphA2-Fc has no effect on B class Eph cleavage of cell surface ephrin-A2 to a soluble form receptor activation or phosphorylation of VEGF by Kuzbanian metalloprotease has been reported in the receptor (data not shown). The use of Fc chimeric developing central nervous system during embryogen- fusion proteins, however, raise the possibility that the esis (Hattori et al., 2000). We do not currently favor binding of EphA2-Fc or EphA3-Fc to membrane this hypothesis as a possible mechanism of ephrin-A1 tethered ephrin-A1 ligand may initiate an immune mediated angiogenesis in our tumor models, however, response leading to complement and/or cell-mediated since soluble ephrins have not been found to be cytotoxicity and clearance of tumor cells and/or biologically active in the absence of artificial clustering endothelial cells. We do not favor this hypothesis as (Davis et al., 1994; Gale and Yancopoulos, 1997). the EphA3-Fc fusion protein, which contains muta- Furthermore, we have not detected soluble ephrins in tions in the Fc region that inhibit interaction with culture media from RIP-Tag-derived bTC cells or from complement and Fc receptor, produces effects compar- 4T1 cells (data not shown). able to EphA2-Fc on vascular density and tumor Based on the expression patterns of ephrin-A1 ligand progression in vivo. and EphA2 receptor in tumors and on the ability of What is the mechanism of EphA-Fc-mediated EphA2-Fc to inhibit many of the angiogenic activities inhibition of tumor progression in vivo? In principle, of VEGF, we propose a cooperative model of ephrinA/ inhibition of tumor growth/survival in vivo could be EphA-mediated tumor angiogenesis in the context of achieved through direct inhibition of proliferation and the VEGF-pathway. In such a model, ephrin-A1 could survival of tumor cells, as 4T1 tumor cells express function to induce tumor angiogenesis through both EphA2 receptor. Alternatively, blocking EphA receptor paracrine and juxtacrine mechanisms. As shown in activation could indirectly affect tumor cell growth and Figure 1p, and Figure 3t, both RIP-Tag and 4T1 viability via regulation of blood vessel recruitment. To tumors express high levels of VEGF. Thus, when distinguish between these two possibilities, we tumor cells and endothelial cells are not in direct compared tumor cell proliferation and apoptosis both contact, VEGF expressed by tumors could induce within the intact tumor and in cell culture. Our data ephrin-A1 expression in adjacent endothelial cells demonstrate that EphA-Fc does not affect tumor cell (Cheng et al., submitted), as suggested by co-localiza- proliferation and survival in culture, though a decrease tion of ephrin-A1 with CD31 (Figure 1). Engagement in tumor cell proliferation and an increase in tumor cell of ephrin-A1 to EphA2 receptor may then activate death was observed in tumor-bearing animals treated juxtacrine signaling to regulate changes in endothelial with EphA2-Fc or EphA3-Fc. Treatment of EphA-Fcs cell morphology, cell – cell, and cell – matrix contact to in tumor-bearing animals did result in a significant promote angiogenesis. This hypothesis is in consistent reduction in tumor vascular density in intact tumors in with VEGF expression in the tumors (Figures 1 and 3) vivo, consistent with the reduction in vascular density and that VEGF-induced endothelial cell migration can observed in cutaneous window assays. In addition, be blocked by soluble EphA2-Fc receptor (Figure 7a). blocking soluble EphA-Fc receptor also inhibited Once endothelial cells are in direct contact with tumor endothelial cell migration in response to tumor cells cells in vivo, tumor cells may direct new blood vessel in transwell co-culture assays. Taken together, these growth through paracrine signaling between ephrin-A1 data suggest that soluble EphA-Fc receptor modulate expressed on tumor cells and EphA2 expressed on tumor progression by inhibiting tumor angiogenesis, surrounding endothelial cells. This is supported by our which deprives the tumors of blood-borne oxygen, tumor cell-endothelial cell co-culture data. As shown in nutrients, and growth/survival factors. These data are Figure 7b – m, in a transwell assay that stimulates consistent with known functions of Eph receptors and tumor-endothelium interaction in vitro, endothelial cells ligands, including ephrin-A1, in mediating endothelial migrate towards malignant bTC or 4T1 tumor cells, cell migration and assembly (Daniel et al., 1996), as but not non-malignant NMuMg cells, and this well as the ability of a truncated, dominant negative migration was blocked by EphA2-Fc. Thus, VEGF EphA2 receptor mutant to inhibit endothelial cell provides a relatively long-range signal while ephrin-A1 assembly into capillary-like structures in vitro (Ogawa provide a contact signal, both of which act in concert et al., 2000). to regulate tumor angiogenesis. One unique feature of the Eph family is that their Given recent reports of reverse signaling through ligands are membrane-bound (Holder and Klein, 1999). ephrin ligands upon stimulation with Eph receptors, it It is therefore intriguing to hypothesize how ephrin-A1 is also possible that EphA2-Fc or EphA3-Fc could expressed on tumor cells could stimulate an angiogenic initiate ephrin-mediated reverse signaling in ephrin-A1 response from vessels that are not in direct contact with expressing tumor cells and/or endothelial cells. Indeed, the tumor. One possibility is that ephrin-A1 ligand may stimulation of ephrin-A2 and ephrin-A5 expressing be cleaved from the surface of tumor cells, and that cells with soluble EphA receptors results increased soluble ephrin-A1 may then act as a long-range signal cellular adhesion involving downstream signaling to activate EphA2 receptor on local endothelial cells to targets such as b1 integrin and src/fyn intracellular initiate angiogenesis. Indeed, ephrin-A1 was originally tyrosine kinases (Davy et al., 1999; Davy and Robbins, characterized as a secreted protein, detected in 2000; Huai and Drescher, 2001). Thus, activation or

Oncogene Eph A receptors and tumor angiogenesis DM Brantley et al 7023 modulation of signaling cascades downstream of wild-type C57BL/6 and Balb/c mice were maintained in ephrins expressed on the surface of tumor and/or accordance with AAALAC and Vanderbilt University guide- endothelial cells by soluble EphA receptors might lines. RIP1-Tag2 transgenic animals (Folkman et al., 1989; therefore initiate changes in adhesion and migration Hanahan, 1985) were maintained on a sucrose-enriched diet that could contribute to decreased angiogenesis (Teklad Test Diets, Madison, WI, USA) and 4% glucose water. Animals positive for the RIP-Tag transgene were observed in the window assays and in Matrigel plugs. identified by PCR analysis of genomic DNA from tail biopsy We do not currently favor this hypothesis, however, as using the following primers: 5’-GGACAAACCACAACTA- EphA2-Fc stimulation alone does not induce angiogen- GAATG-3’ and 5’-CAGAGCAGAATTGTGGAGTGG-3’. esis in mouse corneal assays (Cheng et al., submitted), nor does EphA2-Fc treatment alone alter endothelial 4T1 cell culture and transplantation cell migration, sprouting, proliferation, or apoptosis (data not shown). Moreover, reduced expression of 4T1 (Aslakson and Miller, 1992; Lin et al., 1998) and 4T1- EphA2 receptor in endothelial cells by transfection of GFP (a generous gift from Chuan-Huan Li, Duke University, antisense oligonucleotides mimics the inhibitory effects (Li et al., 2000)) cell lines were maintained in DMEM with of soluble receptor on endothelial cell migration, 10% fetal calf serum. For transplantation experiments, 26105 cells in 0.1 ml media were injected into the number favoring the interpretation of disrupted EphA class four inguinal mammary fat pad of 10 – 15-week-old syngeneic receptor function by soluble receptors (Cheng et al., Balb/c female mice using a 1 ml syringe and a 30 gauge submitted). Though we cannot rule out modulation of needle. Tumors derived from these injected cells were ephrin-A signaling by treatment with soluble Eph surgically collected from the mice 3, 6, and 9 days post- receptors, we have shown that EphA2-Fc treatment implantation. blocks tyrosine phosphorylation of endogenous EphA2 For tumor progression studies, 16105 4T1 cells were receptor in endothelial cells stimulated with soluble suspended in 300 ml of liquid growth factor reduced Matrigel ephrin-A1, as well as corneal angiogenesis induced by (Becton-Dickinson, Franklin Lakes, NJ, USA) in the soluble ephrin-A1 (Cheng et al., submitted). Moreover, presence or absence of 10 mg EphA2-Fc (R&D Systems, EphA2-Fc treatment reduces endogenous EphA2 Minneapolis, MN, USA), EphA3-Fc (Immunex Corpora- tion), or human IgG (Sigma-Aldrich Corporation, St. Louis, receptor phosphorylation in 4T1 Matrigel plugs relative MO, USA). The suspension was injected subcutaneously on to controls, (Figure 5j) supporting the inhibitory the dorsal flank of 10- to 15-week-old syngeneic Balb/c function of soluble EphA receptors on EphA class female mice using a 1 ml syringe and a 30 gauge needle signaling. (Passaniti et al., 1992). The mice were injected daily with In summary, we provide the first evidence that 10 mg EphA2-Fc, EphA3-Fc or IgG in 100 ml of PBS engagement of class A Eph receptors and ligands subcutaneously at the site of Matrigel plug implantation. regulate tumor angiogenesis. In addition, treatment of The plugs were collected and the length and width of each tumor-bearing mice with soluble EphA class receptors tumor was measured using a caliper. Tumor volume was calculated by the following formula: Tumor volume=0.526 results in reduced tumor volume and vascular density 2 without directly affecting proliferation or apoptosis, width 6length (Bergers et al., 1999). Data are a representation of five to 10 independent samples per suggesting that these molecules may modulate angio- condition, and statistical significance was assessed by two- genesis-dependent tumor progression. Thus, A class tailed, paired Student’s t-test. Eph receptors and ligands may be effective new therapeutic targets for cancer treatment. Immunoblot and immunoprecipitation analyses bTC pancreatic carcinoma (Radvanyi et al., 1993), 4T1, and EMT6 (Rockwell, 1977, 1978, 1981) mouse mammary Materials and methods adenocarcinoma cell lines were maintained in DMEM with 10% FCS. Primary mouse mammary epithelial cells were Soluble EphA-Fc receptors isolated and maintained as described previously (Brantley et The EphA2-Fc soluble receptor cDNA construct was al., 2001; Muraoka et al., 2001). An MS-1 endothelial cell provided by Regeneron Inc. (Tarrytown, NY, USA) and line was maintained in endothelial cell growth (EGM) media subcloned into episomal expression vector pCEP4 (Invitro- from Clonetics (Cambrex Corporation, East Rutherford, NJ, gen, San Diego, CA, USA). pCEP4/ephrin-A1-Fc expression USA). Total cellular proteins from these cells were isolated vector was provided by Dr A Pandy (University of Michigan, by lysis in ice-cold RIPA buffer supplemented with protease Ann Arbor, MI, USA). Recombinant EphA2-Fc and ephrin- inhibitors [PMSF (100 mg/ml), aprotinin (40 mg/ml), and A1-Fc proteins were either purified from culture supernatant leupeptin (2 mg/ml)]. Sixty mg of protein lysate from bTC of stable 293T clones expressing these factors using protein A and MS-1 cells, or 40 mg protein lysate from 4T1, EMT6, and sepharose column, or purchased from R&D Systems PMECs, were fractionated on 8 – 10% SDS-polyacrylamide (Minneapolis, MN, USA). Recombinant EphA3-Fc protein gels. The proteins were then transferred to nitrocellulose was provided by Immunex Inc. (Seattle, WA, USA). The Fc membranes and probed with anti-EphA2 antibodies (Upstate region in EphA3-Fc was mutated to inhibit binding to Fc Biotechnology, Lake Placid, NY, USA; 1 mg/ml) and anti- receptors and complement. ephrin-A1 antibodies (Immunex, Seattle, WA, USA; 1 : 500). Specific immunoreaction was detected using anti-IgG antibodies conjugated to horseradish peroxidase (Promega, Mouse strains and maintenance Madison, WI, USA) and ECL plus chemiluminescence RIP1-Tag2 C57B1/6 mice (generously provided by Doug detection kit (Amersham Pharmacia Biotechnology). The Hanahan, University of California at San Francisco) and blots were stripped and re-probed with anti-b-tubulin (Sigma-

Oncogene Eph A receptors and tumor angiogenesis DM Brantley et al 7024 Aldrich, 1 : 750) antibodies to confirm uniform loading. Data diameter) isolated from eleven-week-old transgenic mice. In are a representation of five independent samples per cell line. addition, window assays were performed using cultured 4T1- For immunoprecipitation of EphA2 from Matrigel plugs, GFP cells (approximately 1000 cells) under the same tumor plugs were collected, frozen on dry ice and pulverized conditions. The chambers were sealed with glass coverslips. by mortar and pestle. Lysates were prepared as described Ten to 14 days after implantation, fluorochrome-conjugated above (RIPA buffer supplemented with 1 mM sodium dextran (2% in PBS, Sigma-Aldrich: FITC-conjugated orthovanadate phosphatase inhibitor (Sigma-Aldrich)), and dextran, 40 kDa, for RIP-Tag windows and rhodamine EphA2 was immunoprecipitated from 500 mg lysate using conjugated dextran, 65 kDa, for 4T1-GFP windows) was anti-EphA2 antibodies (Santa Cruz Biotechnology; Upstate injected intravenously, and tumors in window chambers were Biotechnology) plus A/G-sepharose beads (Santa Cruz photodocumented using an Olympus BX60 microscope and Biotechnology). Products were fractionated as above and digital camera. blots probed with anti-phosphotyrosine antibodies pY99 and Density of blood vessels within the window chambers was pY20 (Santa Cruz Biotechnology). Blots were stripped and quantified by fluorescence intensity of FITC or rhodamine- re-probed with anti-EphA2 antibodies (Upstate Biotechnol- dextran using Scion Image software. The density of ogy). fluorescent pixels within each 106tumor field was determined and compared in control IgG or PBS treated windows versus EphA2-Fc treated windows, and statistical significance was Histological analyses determined by two-tailed, paired Student’s t-test. Density of RIP-Tag pancreatic tumors (isolated from 11-week-old mice) 4T1GFP cells was also quantified by fluorescence intensity and 4T1-derived tumors were isolated and fixed in 10% using Scion Image software (version 1.62c). The density of neutral buffered formalin overnight at 48C. The tissue was fluorescent pixels 7 days post-implantation relative to the then paraffin-embedded and 7 – 10 mm sections prepared. total pixels within each 26image was calculated, and Alternatively, the tissue was fixed in 4% paraformaldehyde normalized for initial cell density at 3 days post-implantation. in PBS for 1 h, equilibrated into 30% sucrose in PBS Data are a representation of 6 – 8 independent samples per overnight, and embedded in O.C.T. for preparation of condition with standard error of the mean, and statistical cryosections. Immunohistochemical detection was performed significance was assessed by two-tailed, paired Student’s t-test. as described previously (Brantley et al., 2000, 2001), using primary antibodies against ephrin-A1 (Immunex, 1 : 200), Proliferation and apoptosis assays EphA2 (Upstate Biotechnology, 10 mg/ml), VEGF (Neomar- kers, Freemont, CA, USA; 10 mg/ml), and activated VEGF For cell culture proliferation assays, 16105 bTC or 4T1 cells receptor (GV39M, 1 : 500; Bergers et al., 2000; Brekken et al., were seeded on 48-well plates and serum-starved for 24 h. 1998)). Sections were counterstained with Mayer’s hematox- The cells were then stimulated with media supplemented with ylin, mounted, and photographed using an Olympus BX60 2% serum, 10% serum, 10% serum plus EphA2-Fc (10 mg/ microscope. ml), or 10% serum plus EGF (20 ng/ml) for 3 days. The cells For dual immunofluorescence analysis, cryosections (7 mm) were fixed in 2% glutaraldehyde in PBS, stained with 0.5% were probed with primary antibodies for CD31 (BD crystal violet (Sigma-Aldrich) in 0.2 M boric acid, and washed Pharmingen, San Diego, CA, USA) and ephrin-A1 or extensively with deionized water. The stained cells were lysed EphA2, followed by secondary antibodies conjugated to in 1% SDS and the absorbance of the lysate at 570 nm Alexa 488 green fluorochrome (anti-rat IgG, 1 : 2000; determined. Proliferation of 4T1 tumor cells in vivo was Molecular Probes, Eugene, OR, USA) for CD31 staining or quantified by PCNA immunohistochemistry. Tumor sections Cy3 red fluorochrome (anti-rabbit IgG-Cy3, 1 : 2000; Jackson from 4T1 Matrigel plugs (10 mm) were probed with an anti- Immunoresearch Laboratories, West Grove, PA, USA; anti- PCNA antibody (Neomarkers, 1 : 150) as described above, mouse IgG-Cy3, Molecular Probes) for ephrin-A1 and and the percentage of PCNA positive nuclei relative to total EphA2 staining, respectively. Sections probed with mouse nuclei in four random 406tumor fields per sample was anti-EphA2 antibodies were treated with mouse-on-mouse calculated. Data are a representation of three to five (M.O.M.) blocking reagents to reduce background staining independent experiments/condition with standard deviation, from endogenous mouse IgGs (Vector Laboratories, Burlin- and statistical significance was assessed by two-tailed paired game, CA, USA). Sections were counterstained with DAPI Student’s t-test. (Sigma-Aldrich) to visualize nuclei. Fluorescence images were For TUNEL assays 26105 bTC or 4T1 cells were seeded captured using an Olympus BX60 microscope and digital on chamber slides (Nalgene-Nunc, Rochester, NY, USA) camera. Digital images were processed and merged using coated with collagen (Cohesion, Palo Alto, CA, USA). The Adobe Photoshop 5.5 software. Data are representations of cells were then cultured in serum-free media, media three to six independent samples per condition for immuno- supplemented with 10% serum, or 10% serum plus EphA2- histochemical analyses. Fc for 48 h. TUNEL assays were performed using an Apoptag red in situ apoptosis detection kit (Intergen Company, Purchase, NY, USA) according to the manufac- Cutaneous window assays turer’s protocol. TUNEL staining was quantified by Window assays were performed as described previously calculating the number of fluorescent pixels in the rhodamine (Huang et al., 1999; Li et al., 2000). Briefly, a 5 mm diameter field (TUNEL positive nuclei) and normalizing based on flap of skin was dissected away from the dorsal skin flap of fluorescent pixels within the DAPI field (total nuclei) using anesthetized recipient mice (C57B1/6 recipient mice for RIP- Scion Image software. Apoptosis in 4T1 tumor cells in vivo Tag and Balb/c mice for 4T1 cells), leaving a fascial plane was also quantified. Tissue sections from 4T1 Matrigel with associated vasculature. A hydron pellet impregnated tumors (10 mm) were also subjected to TUNEL analysis as with normal human IgG (Sigma-Aldrich, 1 mg/pellet), described above. Data are a representation of three to five EphA2-Fc (1 mg/pellet, purified in house or from R&D independent experiments/condition with standard deviation, Systems), or PBS was implanted in the window chamber and statistical significance was assessed by two-tailed, paired adjacent to a RIP-Tag tumor (approximately 0.7 mm in Student’s t-test.

Oncogene Eph A receptors and tumor angiogenesis DM Brantley et al 7025 removed from the upper surface of the transwell filter using a Endothelial cell migration and co-culture assays cotton swab, and endothelial cell density on the lower surface Endothelial cell migration was assessed using a modified of the filter was quantified by counting red fluorescent pixels Boyden chamber assay as described previously (Cheng et al., using Scion Image software analysis. Data are a representa- submitted). Bovine pulmonary microvascular endothelial cells tion of six to nine independent samples per condition with (Clonetics, passage 3 – 8) were labeled for 1 h with Texas-red standard deviation, and statistical significance was assessed ovalubumin (Molecular Probes; 0.5 mg/ml in DMEM) and by two-tailed, paired Student’s t-test. detached from tissue culture plates using 0.5 M EDTA. Cells (16105) were added to upper transwell chambers and the chambers were placed in 24-well dishes supplemented with Acknowledgments DMEM/2% FCS (unstimulated) plus ephrin-A1-Fc (R&D We would like to thank Amanda Kizzee for excellent Systems, 2.5 mg/ml)+EphA2-Fc (12.5 mg/ml) or human IgG technical support, Nick Gale and George Yancopolous (Sigma-Aldrich, 12.5 mg/ml), VEGF (R&D Systems, 20 ng/ (Regeneron Inc.), Chuan-Huan Li and Mark Dewhirst ml)+EphA2-Fc (5 mg/ml). After 5 h, cells were removed (Duke University), and Doug Hanahan (UCSF) for from the upper surface of the transwell filter using a cotton providing EphA2-Fc construct, the 4T1-GFP subline, and swab, and endothelial cell density on the lower surface of the RIP-Tag transgenic mice, respectively. Special thanks to filter was quantified by counting fluorescent pixels using Laura Debusk for assistance with cutaneous window Scion Image software analysis. Data are a representation of assays, and to Rick Haselton for the transwell co-culture nine independent samples per condition with standard assay protocol and helpful discussions. We thank Ray deviation, and statistical significance was assessed by two- Dubois, Brigid LM Hogan, and Lynn M Matrisian for tailed, paired Student’s t-test. helpful discussions, and give special thanks to Carlos L For co-culture experiments, transwells were coated with Arteaga for his ongoing support. This work was supported growth factor-reduced Matrigel (1 : 20 dilution in DMEM) by National Institutes of Health grants HD36400 and and 26105 4T1, bTC, or NMuMg (Owens et al., 1974; Van DK47078, American Heart Association grant 97300889N, den Broecke et al., 1996) were plated on the lower surface of a JDF grant I-2001-519, and an ACS Institutional the transwell filter. Tumor cells on the lower filter surface Research Grant IN-25-38 to J Chen, Vascular Biology were labeled with FITC ovalbumin (Molecular Probes; Training Grant T32-HL-07751-06 and American Heart 0.5 mg/ml in DMEM). BPMECs labeled with Texas-red Association Fellowship 0120147B to D Brantley, and ovalbumin were added to upper transwell chambers as before Cancer training grant T-32 CA09592 to N Cheng. This in DMEM/2% FCS+EphA2-Fc (R&D Systems, 10 mg/ml) work was also supported by a core facilities grant or hIgG (R&D Systems, 10 mg/ml). After 5 h, cells were 2P30CA68485 to the Vanderbilt-Ingram Cancer Center.

References

Aaltomaa S, Lipponen P, Papinaho S and Syrjanen K. Dankort DL and Muller WJ. (1996). Transgenic models of (1993). J. Cancer Res. Clin Oncol., 119, 288 – 294. breast cancer metastasis. Kluwer Academic Publishers. Adams RH, Wilkinson GA, Weiss C, Diella F, Gale NW, Davis S, Gale NW, Aldrich TH, Maisonpierre PC, Lhotak V, Deutsch U, Risau W and Klein R. (1999a). Genes Dev., 3, Pawson T, Goldfarb M and Yancopoulos GD. (1994). 295 – 306. Science, 266, 816 – 819. Adams RH, Wilkinson GA, Weiss C, Diella F, Gale NW, Davy A, Gale NW, Murray EW, Klinghoffer RA, Soriano P, Deutsch U, Risau W and Klein R. (1999b). Genes Dev., 13, Feuerstein C and Robbins SM. (1999). Genes Dev., 13, 295 – 306. 3125 – 3135. Arbiser JL, Moses MA, Fernandez CA, Ghiso N, Cao Y, Davy A and Robbins SM. (2000). EMBO J., 19, 5396 – 5405. Klauber N, Frank D, Brownlee M, Flynn E, Parangi S, Easty DJ, Herlyn M and Bennett DC. (1995). Int. J. Cancer, Byers HR and Folkman J. (1997). Proc. Natl. Acad. Sci. 60, 129 – 136. USA, 94, 861 – 866. Easty DJ, Hill SP, Hsu MY, Fallowﬁeld ME, Florenes VA, Aslakson CJ and Miller FR. (1992). Cancer Res., 52, 1399 – Herlyn M and Bennett DC. (1999). Int. J. Cancer, 84, 494 – 1405. 501. Bergers G, Brekken R, McMahon G, Vu TH, Itoh T, Tamaki Folkman J. (1990). J. Natl. Cancer Inst., 82, 4–6. K, Tanzawa K, Thorpe P, Itohara S, Werb Z and Hanahan Folkman J. (1994). J. Clin. Oncol., 12, 441 – 443. D. (2000). Nat. Cell Biol., 2, 737 – 744. Folkman J, Watson K, Ingber D and Hanahan D. (1989). Bergers G, Javaherian K, Lo KM, Folkman J and Hanahan Nature, 339, 58 – 61. D. (1999). Science, 284, 808 – 812. Gale NW and Yancopoulos GD. (1997). Cell Tissue Res., Brantley DM, Chen CL, Muraoka RS, Bushdid PB, 290, 227 – 241. Bradberry JL, Kittrel F, Medina D, Matrisian LM, Kerr Gale NW and Yancopoulos GD. (1999). Genes Dev., 13, LD and Yull FE. (2001). Mol. Biol. Cell., 12, 1445 – 1455. 1055 – 1066. Brantley DM, Yull FE, Muraoka RS, Hicks DJ, Cook CM Gerety SS, Wang HU, Chen ZF and Anderson DJ. (1999). and Kerr LD. (2000). Mech. Dev., 97, 149 – 155. Molecular Cell, 4, 403 – 414. Brekken RA, Huang X, King SW and Thorpe PE. (1998). Goede V, Fleckenstein G, Dietrich M, Osmers RG, Kuhn W Cancer Res., 58, 1952 – 1959. and Augustin HG. (1998). Anticancer Res., 18, 2199 – 2202. Chen J and Ruley HE. (1998). J. Biol. Chem., 273, 24670 – Hanahan D. (1985). Nature, 315, 115 – 122. 24675. Hattori M, Osterﬁeld M and Flanagan JG. (2000). Science, Cheng N, Brantley D and Chen J. (2002). Cytokine and 289, 1360 – 1365. Growth Factor Reviews, 13, 75 – 85. Holder N and Klein R. (1999). Development, 126, 2033 – Daniel TO, Stein E, Cerretti DP, St John PL, Robert B and 2044. Abrahamson DR. (1996). Kidney Int. Suppl., 57, S73 – S81.

Oncogene Eph A receptors and tumor angiogenesis DM Brantley et al 7026 Holzman LB, Marks RM and Dixit VM. (1990). Mol. Cell. Prewett M, Huber J, Li Y, Santiago A, O’Connor W, King Biol., 10, 5830 – 5838. K, Overholser J, Hooper A, Pytowski B, Witte L, Bohlen P Huai J and Drescher U. (2001). J. Biol. Chem., 276, 6689 – and Hicklin DJ. (1999). Cancer Res., 59, 5209 – 5218. 6694. Preziosi R, Sarli G, Benazzi C and Marcato PS. (1995). J. Huang Q, Shan S, Braun RD, Lanzen J, Anyrhambatla G, Comp. Pathol., 113, 301 – 313. Kong G, Borelli M, Corry P, Dewhirst MW and Li CY. Radvanyi F, Christgau S, Baekkeskov S, Jolicoeur C and (1999). Nat. Biotechnol., 17, 1033 – 1035. Hanahan D. (1993). Mol. Cell Biol., 13, 4223 – 4232. Kuehn R, Lelkes PI, Bloechle C, Niendorf A and Izbicki JR. Rockwell S. (1977). Int. J. Radiat. Biol. Relat. Stud. Phys. (1999). Pancreas, 18, 96 – 103. Chem. Med., 31, 153 – 160. Laferriere J, Houle F, Taher MM, Valerie K and Huot J. Rockwell S. (1978). Br. J. Cancer Suppl., 37, 212 – 215. (2001). J. Biol. Chem., 276, 33762 – 33772. Rockwell S. (1981). Cancer Res., 41, 527 – 531. Li CY, Shan S, Huang Q, Braun RD, Lanzen J, Hu K, Lin P Van den Broecke C, Vleminckx K, De Bruyne G, Van and Dewhirst MW. (2000). J. Natl. Cancer Inst., 92, 143 – Hoorde L, Vakaet L, Van Roy F and Mareel M. (1996). 147. Clin.Exp.Metastasis,14, 282 – 296. LinP,BuxtonJA,AchesonA,RadziejewskiC,Maisonpierre Walker-Daniels J, Coffman K, Azimi M, Rhim JS, Bostwick PC, Yancopoulos GD, Channon KM, Hale LP, Dewhirst DG, Snyder P, Kerns BJ, Waters DJ and Kinch MS. MW, George SE and Peters KG. (1998). Proc. Natl. Acad. (1999). The Prostate, 41, 275 – 280. Sci. USA, 95, 8829 – 8834. Wang HU, Chen ZF and Anderson DJ. (1998a). Cell, 93, McBride JL and Ruiz JC. (1998). Nech. Dev., 77, 201 – 204. 741 – 753. Muraoka RS, Lenferink AE, Simpson J, Brantley DM, Weidner N. (1996). Angiogenesis and breast cancer.Kluwer Roebuck LR, Yakes FM and Arteaga CL. (2001). J. Cell Academic Publishers. Biol., 153, 917 – 932. Weidner N, Folkman J, Pozza F, Bevilacqua P, Allred EN, Myers C, Charboneau A and Boudreau N. (2000). J. Cell Moore DH, Meli S and Gasparini G. (1992). J. Natl. Biol., 148, 343 – 351. Cancer Inst., 84, 1875 – 1887. Ogawa K, Pasqualini R, Lindberg RA, Kain R, Freeman AL Weidner N, Semple JP, Welch WR and Folkman J. (1991). N. and Pasquale EB. (2000). Oncogene, 19, 6043 – 6052. Engl. J. Med., 324, 1–8. Owens RB, Smith HS and Hackett AJ. (1974). J. Natl. Yancopoulos GD, Davis S, Gale NW, Rudge JS, Wiegand SJ Cancer Inst., 53, 261 – 269. and Holash J. (2000). Nature, 407, 242 – 248. Pandey A, Shao H, Marks RM, Polverini PJ and Dixit VM. Zantek ND, Azimi M, Fedor-Chaiken M, Wang B, Bracken- (1995). Science, 268, 567 – 569. bury R and Kinch MS. (1999). Cell Growth Differ., 10, Papenfuss HD, Gross JF, Intaglietta M and Treese FA. 629 – 638. (1979). Microvasc. Res., 18, 311 – 318. Zelinski DP, Zantek ND, Stewart JC, Irizarry AR and Kinch Passaniti A, Taylor RM, Pili R, Guo Y, Long PV, Haney JA, MS. (2001). Cancer Res., 61, 2301 – 2306. Pauly RR, Grant DS and Martin GR. (1992). Lab. Invest., 67, 519 – 528.

Oncogene