Gene Therapy (2005) 12, 831–842 & 2005 Nature Publishing Group All rights reserved 0969-7128/05 $30.00 www.nature.com/gt RESEARCH ARTICLE Full kringles of plasminogen (aa 1–566) mediate complete regression of human MDA-MB-231 breast tumor xenografted in nude mice

A Galaup1, C Magnon1, V Rouffiac2, P Opolon1, D Opolon1, N Lassau3, T Tursz4, M Perricaudet1 and F Griscelli1 1Le Centre National de la Recherche Scientifique, Unite´ Mixte de Recherche (UMR) 8121, Institut Gustave Roussy, Villejuif Cedex, France; 2Laboratoire d’Imagerie du Petit Animal (LIPA), Institut Gustave Roussy, Villejuif Cedex, France; 3Imaging Department, Institut Gustave Roussy, Villejuif Cedex, France; and 4Department of Medical Oncology, Institut Gustave Roussy, Villejuif Cedex, France

Since kringle (K)5, not present in the angiostatin molecule, intratumoral injection of AdK1–5(1–566) into MDA-MB-231 was shown to be a key functional domain possessing potent breast human carcinoma tumors was followed by a total antiangiogenic activity, we have evaluated a new plasmino- regression of 40% of the tumor and by significant arrest of gen-derived fragment, consisting of the N-terminal part of tumor growth (90%), which was correlated with a drastic human plasminogen, that included the complete secondary decrease of functional neovascularization into the tumors. structure of K1–5 (aa 1–566). In contrast to other fragments Furthermore, systemic delivery of AdK1–5(1–566) in mice described to date, K1–5 includes cysteine residues at inhibited the lung invasion of melanoma B16-F10 cells by positions 543, 555 and 560 allowing the formation of the 87%. Our findings provide evidence that the full kringles three disulfide bonds lying within K5. Vascular endothelial of plasminogen (aa 1–566) may be much more potent than cell proliferation and migration assays revealed that a K1–3 (aa 1–354), for the suppression of angiogenesis, tumor replication-defective adenovirus (AdK1–5(1–566)), expres- growth and metastatic dissemination. sing K1–5 (aa 1–566), was dose dependently more potent Gene Therapy (2005) 12, 831–842. doi:10.1038/ that AdK1–3(1–354), an adenovirus that expresses only the sj.gt.3302474; Published online 24 March 2005 first three kringles. In contrast to AdK1–3(1–354), a single

Keywords: antiangiogenesis; K1–5 (aa 1–566); adenovirus; cancer

Introduction constitutes a new strategy to suppress tumor growth. Some angiogenic inhibitors are fragments, or cryptic Angiogenesis is defined as the formation of new blood domains, of larger proteins. This is the case for vessels that sprout from pre-existing ones. The different angiostatin,3 endostatin,4 and canstatin,5 which are steps of this process respond to stimuli that promote the proteolytic fragments of plasminogen, collagen XVIII proliferation, migration, and differentiation of endo- and collagen IV, respectively, and for the 16 kDa N- thelial cells, and the degradation of the blood vessel terminal fragment-derived prolactin6 or the N-terminal basement membrane and extracellular matrix.1 This fragment-derived platelet factor 4.7 process leads to the sprouting of new microtubes and The different fragments derived from the five kringles implies the presence of multiple controls, which can be (K) of plasminogen that are known to inhibit angiogen- turned on and off within a short period of time. esis have been characterized. Human angiostatin K1–4 Angiogenesis is required for physiological processes (aa 97–454), containing the first four disulfide-linked such as embryonic development, wound healing, tissue structures of plasminogen, which was the first protein and organ regeneration, and for pathological processes described, is generated following hydrolysis of plasmi- such as tumor growth, metastatic dissemination, cardio- nogen by a metalloelastase.3,8–10 In vitro, angiostatin K1–4 2 vascular diseases and diabetic retinopathy. (aa 97–454) inhibits endothelial cell growth with an ED50 The angiogenic switch depends upon the balance of 135 nM,11 and endothelial cell migration in a specific between angiogenic stimulators and inhibitors.2 There- manner.12 In vivo, angiostatin K1–4 (aa 97–454) inhibits fore, blocking the action of angiogenic factors, or neovascularization in chick chorioallantoic membrane utilizing angiogenic suppressors to inhibit angiogenesis, and mouse corneal assays, and suppresses subsequent tumor growth in different murine models.13 On the other hand, a smaller fragment of plasminogen, containing the Correspondence: Dr A Galaup, Le Centre National de la Recherche first three kringles (K1–3 (aa 97–354)), inhibits endothe- Scientifique, Unite´ Mixte de Recherche (UMR) 8121 PR2, Institut Gustave Roussy, 39 rue Camille Desmoulins, 94805 Villejuif Cedex, France lial cell proliferation with an ED50 of 70 nM, but does not Received 1 April 2004; accepted 16 November 2004; published display an inhibitory effect on cell migration. In vivo online 24 March 2005 studies have shown that K1–3 (aa 97–354) inhibits tumor K1–5 (aa 1–566) gene transfer A Galaup et al 832 growth in different murine models.14 Another fragment observed from the AdK1–3(1–354)- and AdK1–5(1–566)- of plasminogen containing only the K5 (aa 481–566) infected cell supernatants (Figure 1c), whereas no signal displays distinct inhibitory functions on endothelial cell could be detected following the infection with AdCO1.

proliferation with an ED50 of 50 nM, and on endothelial The immunoreactive peptides appeared with a molecu- cell migration.15,16 Furthermore, K5 (aa 481–566) and lar weight of 40.5 kDa for the K1–3 (aa 1–354) molecule K1–3 (aa 97–354) were shown to cause cell cycle arrest and 63.8 kDa for the K1–5 (aa 1–566) molecule. and apoptosis of endothelial cells.13,14 It was recently To assess in vitro the antiangiogenic effect of AdK1– shown that plasminogen can be converted by plasmin 5(1–566), and to compare its activity to AdK1–3(1–354), (a fibrinolytic serine protease)17 or by glioma cell BT325- human umbilical vein endothelial cells (HUVECs) were derived proteinase,18 into a new angiogenic inhibitor, infected for 96 h at various MOI (100–800 PFU/cell), and called K1–5 (aa 97–549), which contains the K1–4 and a then incubated in the presence of MTT reagent to part of K5 domains of plasminogen. The inhibitory effect determine the percentage of surviving cells. In this assay, of K1–5 (aa 97–549), also called angiostatin 4.5 by other transduction of HUVEC by AdK1–5(1–566) resulted in authors,17 on endothelial cell proliferation appears to be a strong and dose-dependent inhibition of endothelial at least 50-fold greater than that of angiostatin cell surviving (Figure 2a). AdK1–5(1–566) infection of K1–4. In vivo, low doses of K1–5 (aa 97–549) suppress HUVEC led to a significantly higher inhibitory effect angiogenesis in chicken embryo and in mouse corneal (Po0.001) as compared with cells that had been infected assays, whereas angiostatin has no effect at the same by AdK1–3(1–354). At an MOI of 100 PFU/cell, HUVEC dose. Furthermore, daily systemic injections of mice with survival of 46720.6 and 88711.9% was observed after low doses of K1–5 (aa 97–549) (2 mg/kg) over 18 days, infection of HUVEC with AdK1–5(1–566) and AdK1– during which K1–4 (aa 97–454) was ineffective, inhibited 3(1–354), respectively. Thus, compared to the infection of the growth of, and angiogenesis within, murine T241 HUVEC with AdK1–3(1–354), cells were found to be fibrosarcomas in mice. Thus, K1–5 (aa 97–549) appears to systematically more sensitive to infection with AdK1–

be a naturally occurring human isoform that is more 5(1–566), with an ED50 (defined as the MOI leading to potent than K1–4 (aa 97–454) in inhibiting the prolifera- 50% inhibition) in the range of 100 PFU/cell, whereas tion and migration of endothelial cells in vitro and the AdK1–3(1–354) was ineffective at this dose. The infection T241 fibrosarcoma in vivo.17 Furthermore, it was recently of HUVEC with 800 PFU/cell of AdK1–3(1–354) and demonstrated that K1–5 (aa 97–549) mediates apoptosis AdK1–5(1–566) resulted in cell survival of 5972.8 and of vascular endothelial cells by the activation of caspase- 1178.8%, respectively. 3, -8, and -9 activity.19 The quantities of K1–3 (aa 1–354) and K1–5 (aa 1–566) Interestingly, the K1–5 (aa 97–549) fragment lacks produced in supernatant of HUVECs 48 h or 96 h cysteine residues in positions 555 and 560, thus prevent- postinfection (p.i.) with AdK1–3(1–354) or AdK1–5(1– ing the formation of two out of three disulfide bonds 566) were evaluated with a specific ELISA. The levels of lying within K5. In order to optimize the antitumor proteins expressed by both adenoviruses were identical, effects associated with the secondary structures of showing that the higher inhibitory effect of AdK1–5(1– kringle domains, we chose to evaluate the antitumor 566) was related to the specific activity of the K1–5 (aa properties of a fragment that encompasses the first 566 1–566) molecule and not to a difference in transgene amino acids of plasminogen, allowing the complete production (Figure 2b). In contrast, infection of HUVEC secondary structure of the K5. with the control empty virus had only a marginal effect, We constructed a replication-defective adenovirus even at the highest tested dose of 800 PFU/cell. The directing the expression and secretion of the N-terminal dose-dependent inhibitory effect specifically observed fragment of human plasminogen, including K1–5 after HUVEC infection with AdK1–5(1–566) is thus a (AdK1–5(1–566)), and assessed its in vitro and in vivo direct consequence of peptide expression, and not due activity in different murine tumor models. We showed to infection with adenovirus per se. This proliferation that a single injection of AdK1–5(1–566) into human assay demonstrated that the AdK1–5(1–566) recombinant breast carcinomas resulted in a significant tumor regres- adenovirus generates a functional secreted peptide, sion, which was complete in 40% of treated animals. which exhibits a high inhibition of endothelial cell proliferation. To assess the paracrine potential of K1–3 (aa 1–354) Results and K1–5 (aa 1–566) molecules to exert these effects, virus-free culture media from virally infected MRC-5 Molecular and functional characterization were added to HUVECs. As illustrated in Figure 2c, after of AdK1–5(1–566) 2 or 4 days of exposure and in comparison to the control The recombinant adenoviruses AdK1–5(1–566) and (supernatant of AdCO1-transduced fibroblasts), we ob- AdK1–3(1–354) used in this study carry a cytomegalo- served a significantly higher inhibition of HUVECs’via- virus (CMV)-driven N-terminal fragment of human bility with the supernatant containing the K1–5 (aa 1– plasminogen that includes all five, or the first three, 566) molecule than with the supernatant containing the K domains, respectively, whereas AdCO1 is an ‘isogenic’ K1–3 (aa 1–354) molecule (P ¼ 0.027). control adenovirus that does not encode any expression To determine whether K1–3 (aa 1–354) and K1–5 (aa 1– cassette (Figure 1a and b). To characterize the expression 566) affect endothelial cell migration, HUVECs were of the K1–5 (aa 1–566) and K1–3 (aa 1–354) molecules allowed to migrate in fibronectin-coated Petri dishes in from the engineered adenovirus, we infected HMEC-1 response to vascular endothelial growth factor (VEGF) at cells at a multiplicity of infection (MOI) of 300 plate- a dose of 30 ng/ml after infection at 800 PFU/cell with forming units (PFU)/cell for 72 h and analyzed the AdCO1, AdK1–3(1–354) or AdK1–5(1–566). Infection of supernatants by Western blotting. Specific signals were HUVECs with AdK1–5(1–566) resulted in total inhibition

Gene Therapy K1–5 (aa 1–566) gene transfer A Galaup et al 833 a E1A E1B E3 3 and -9 levels was seen between AdCO1-treated cells and cells treated with either AdK1–5(1–566) or AdK1– 5' 3' ITR 3' 5' 3(1–354). As such, no significant difference in caspase 10 20 30 40 50 60 70 80 90 100 activation was observed between AdK1–5(1–566) and E2 E4 AdK1–3(1–354) following infection in HUVECs.

pCMV K1K2 K3 poly A AdK1-3(1-354) AdK1–5(1–566) inhibits tumor growth following pCMV K1 K2 K3 K4 K5 poly A AdK1-5(1-566) intratumoral administration Adenovirus-mediated delivery of K1–3 (aa 1–354) and b 1 K1–5 (aa 1–566) peptides was first studied in pre- K1-3 (aa 1-354) MEHKEVVLLLLLF established human breast (MDA-MB 231) and prostate YCKIPSCDS (PC3) models grown in athymic mice. To this end, 347 352 experimental xenografts were intratumorally injected with 2 Â 109 PFU of AdK1–5(1–566) or AdK1–3(1–354), 1 K1-5 (aa 1 -566) whereas control animals were intratumorally injected MEHKEVVLLLLLF with the AdCO1 control virus or PBS. We followed this WCYTTNPRKLYDYCDVPQCAAPSFD 543 555 560 protocol for five independent experiments in which mice were randomized and treated when the tumors had c 1 2 3 4 5 reached a mean volume of 62.5715 mm3 for the PC3 model and 3074mm3 for the MDA-MB 231 model. Plasminogen Tumor growth was then monitored until day 20 or 42 (90 KDa) postinjection, and the percentage inhibition of tumor K1-5 (aa 1-566) growth was calculated. As shown in Figure 4a and c, (63.8 KDa) the dose of 2 Â 109 PFU of AdK1–3(1–354) produced only a limited, or insignificant, growth inhibition compared K1-3 (aa1- 354) with control (AdCO1 or PBS), with a percentage of tumor (40.5 KDa) growth inhibition at day 10 of 873.4 and 27.8712.6% for PC3 and MDA-MB 231, respectively. In contrast, significant tumor growth inhibition was obtained at the end of treatment when PC3 and MDA-MB 231 tumors Figure 1 Construction and characterization of AdK1–3(1–354) and were treated with 2 Â 109 PFU of AdK1–5(1–566) (4473.4 AdK1–5(1–566). (a) AdK1–3(1–354) and AdK1–5(1–566) are isogenic 7 E1/E3-deleted adenoviruses. The expression cassettes of human K1–3 (aa and 90.2 1.9% of tumor growth inhibition at day 10, 1–354) and K1–5 (aa 1–566) are shown under the Ad5 genome; with these respectively). These results were correlated with the transgenes being driven by the CMV promoter, and transcriptionally detection of immunoreactive human K1–3 (aa 1–354) and terminated by an SV40 late polyadeylation signal (polyA). (b) Schematic K1–5 (aa 1–566) in PC3 (Figure 4b) and MDA-MB-231 representation of the structures of plasminogen and its K fragments (Figure 4d) tumors injected with AdK1–3(1–354) or contained in angiostatin and in the N-terminal fragment of plasminogen K1–3 (aa 1–354) and K1–5 (aa 1–566). The cysteine residues (C) are AdK1–5 (1–566) but not in tumors injected with AdCO1 shown for both molecules (c) analysis of K1–3 (aa 1–354) and K1–5 (aa 1– or PBS (Figure 4b and d). In these two models, tumors 566) secretion by Western blotting from uninfected HMEC-1 cells (lane 1) injected with AdCO1 or AdK1–3(1–354) grew rapidly or those infected at an MOI of 300 PFU/cell with AdCO1 (lane 2), AdK1– between days 8–10 and 20 postinjection. At the end of the 3(1–354) (lane 3) or AdK1–5(1–566) (lane 4). In all, 100 ng of human experiment, PC3 tumors had reached a mean volume plasminogen (lane 5) was run as control. of 996.37197.6 mm3 for the AdCO1 group and 1186.47 120 mm3 for the AdK1–3(1–354) group. In contrast, PC3 tumors treated with AdK1–5(1–566) exhibited slower of endothelial cell migration 72 h after infection, and in 7 3 significant inhibition (a distance of 0.0770.06 mm with tumor growth, reaching a volume of 542.5 106.1 mm AdK1–5(1–566), compared to 0.8570.15 mm observed (Po0.04) (Figure 4a). In the MDA-MB 231 model, at day 42 postinjection, tumors had reached a mean volume with AdCO1, Po0.001) at 96 h p.i. (Figure 3a). In 7 7 7 3 contrast, AdK1–3(1–354) was markedly less active. of 445 178.1, 199.5 72 and 18.7 12.3 mm for the Following these observations and considering that AdCO1, AdK1–3(1–354) and AdK1–5(1–566) groups, angiostatin 4.5 (aa 97–549) mediated apoptosis of respectively. Interestingly, in two independent experi- vascular endothelial cells,19 we wondered if the differ- ments, AdK1–5(1–566) significantly inhibited growth of MDA-MB 231-derived tumors until day 42 postinjection ence in activity between AdK1–5(1–566) and AdK3(1– 7 354) on endothelial cells could be explained by the (94 3.5%, Po0.05) and yielded total regression in 2/5 activity of caspase-3, -8 and -9. HUVECs were infected animals (Figure 4c). On hematoxylin–eosin–saffranin with 800 PFU/cell of AdCO1, AdK1–3(1–354) or AdK1– (HES)-stained sections, a total absence of tumor was 5(1–566) in the presence of VEGF at a dose of 3 ng/ml. observed in 40% of mice (Figure 4c), whereas all the After 48 h, protease activity in the cell lysates was animals treated with AdCO1 or AdK1–3(1–354) showed evaluated with caspase-3, -8 and -9 assay kits in two the presence of persistent tumor cells. independent experiments. A two-fold increase in cas- pase-8 was detected 48 h following treatment in AdK1– AdK1–5(1–566) inhibits angiogenesis following 5(1–566)- and in AdK3(1–354)-treated cells as compared intratumoral administration with cells infected with AdCO1 (P ¼ 0.031 and 0.017, A quantitative assessment of angiogenic inhibition in- respectively, Figure 3b). Caspase-3 and -9 did not appear duced by AdK1–5(1–566) and AdK1–3(1–354) was per- to be activated, since no significant difference in caspase- formed on the PC3 murine tumor model. Intratumoral

Gene Therapy K1–5 (aa 1–566) gene transfer A Galaup et al 834 a a 1.6 100 1.4 90 * 1.2 72h 80 70 1 96h 60 0.8 50 0.6 40 0.4 * * * *

Cell viability (%) 30 0.2 AdCO1 HUVEC migration (mm) 20 Ad K1-3(1- 354) 0 10 Ad K1-5(1-566) AdCO1 AdK1-3(1-354) AdK1-5(1-566) 0 0 200 400 600 800 1000 b Viral doses (pfu/cell) AdK1-5(1-566) * b 35 48h AdK3(1-354) * 30 96h

cells) 25 6 AdCO1 20

15 Staurosporine

10 0 0.1 0.2 0.3 0.4 0.5 5 OD (nm)/100 µg of protein K1-3 / K1-5 (ng/ 10 0 Figure 3 Functionality of AdK1–3(1–354) and AdK1–5(1–566). (a) AdCO1 Ad K1-3(1-354) Ad K1-5(1-566) Inhibition of endothelial cell migration. HUVECs were infected with 800 PFU/cell of AdCO1, AdK1–3(1–354) or AdK1–5(1–566), and the c 140 degree of cellular migration assessed every 24 h for 3 days (see Materials 48h and methods). (b) Activation of caspase-8 in response to AdK1–3(1–354) and AdK1–5(1–566). HUVECs were infected with 800 PFU/cell of 120 96h AdCO1, AdK1–3(1–354) and AdK1–5(1–566), and exposed to 50 nM of *** * staurosporine used as a positive control. After 48 h, cells were resuspended 100 in 50 ml of cell lysis buffer, and the protease activity of the lysates was evaluated with caspase-8 Fluorometric Assay Kits (see Materials and 80 methods).

60

Surviving cells (%) 566) is a more potent angiogenic inhibitor than K1–3 (aa 40 1–354) (score of 0.5570.081). Inhibition of the functional intratumoral vasculariza- 20 tion induced by AdK1–5(1–566) and AdK1–3(1–354) was evaluated on the MDA-MB 231 model by Contrast- 0 AdCO1 Ad K1-3(1-354) Ad K1-5(1-566) Enhanced Color Doppler High-Frequency Ultrasono- Figure 2 Functionality of K1–3 (aa 1–354) and K1–5 (aa 1–566). (a) graphy (CDHFUS). CDHFUS specifically detects the Inhibition of endothelial cell viability. HUVECs were infected with functional vessels, which is not possible with classical AdCO1, AdK1–3(1–354) or AdK1–5(1–566) at doses ranging from 100 to histological techniques. For this study, 120 assays were 800 PFU/ml and cell survival was assessed after 4 days. (b) Quantification performed using CDHFUS. The injection of the contrast of K1–3(1–354) and K1–5(1–566) in supernatants of AdK1–3(1–354)- and agent Levovist allowed a precise, weekly quantification AdK1–5(1–566)-HUVEC-infected cells by ELISA 48 and 96 h following of the intratumoral vascularization for the duration of infection. (c) Inhibition of endothelial cell viability by virus-free super- natants from AdCO1-, AdK1–3(1–354)- and AdK1–5(1–566)-infected the treatment period (42 days). The resulting data human fibroblasts (MRC-5 cells). HUVECs survival was assessed after 4 indicated a significant reduction of intratumoral vascu- days of exposure (see Materials and methods). larization (88.7711.3% at week 5 and 61710% at day 42, Po0.05) within the tumors from the AdK1–5(1–566)- treated group as compared with tumors from the AdCO1 vascularization was assessed by immunohistochemistry group (Figure 6a and b). Thus, we demonstrated, using using rat primary antibodies raised against the mouse an innovative method, that the antitumoral effect of the platelet endothelial cell adhesion molecule (PECAM-1). AdK1–5(1–566) is correlated with an antiangiogenic Quantification was performed by assessing the immu- activity. noreactive score per histological field for each animal (see Materials and methods). As shown in Figure 5a and Systemic administration of AdK1–5(1–566) b, the data indicate a significant reduction of intratumor- To test whether secreted K1–5 (aa 1–566) can function in al vascularization within the tumor sections from the a paracrine manner, we assessed the systemic adminis- AdK1–5(1–566)-treated group (score of 0.4470.138) as tration of AdK1–5(1–566) and AdK1–3(1–354) in order compared with tumor sections from the AdCO1-injected to infect a wide range of tissues that are able to group (score of 1.4370.26, Po0.05). Thus, K1–5 (aa 1– secrete angiostatic factors. After systemic administration,

Gene Therapy K1–5 (aa 1–566) gene transfer A Galaup et al 835 ab 1600 PBS 1 2 1400 AdCO1 AdK1-3(1-354)

) 1200 AdK1-5(1-566) 3

1000 3 4 800

600

Tumor volume (mm 400

200 5 6 0 0 5 10 15 20 25 Day

c 600 d PBS 1 2 AdCO1 500 AdK1-3(1-354) AdK1-5(1-566) ) 3 400

3 4 300

200 Tumor volume (mm

100

0 0 5 10 15 20 25 30 35 40 45 Day Figure 4 Comparison of the antitumor effect of AdK1–5(1–566) with that of AdK1–3(1–354) after local administration. (a) PC3 cells were s.c. implanted into athymic mice and when the tumor had reached a mean volume 460 mm3 (day 1), mice received an intratumoral injection of 2  109 PFU of AdCO1, AdK1–3(1–354) or AdK1–5(1–566). Mean values are represented with the s.e.m. (b) Immunostaining of K1–3 (aa 1–354) (3 and 5, respectively, magnification  100 and  400) and K1–5 (aa1–566) (4 and 6, respectively, magnification  100 and  400) on PC3-derived tumors injected with AdK1– 3(1–354) or AdK1–5(1–566). AdCO1- (1) and PBS- (2) treated tumors are used as control. (c) MDA-MB-231 cells were s.c. implanted into athymic mice, and when tumor had reached a mean volume 430 mm3 (day 1), the mice were given a local injection of 2  109 PFU of AdCO1, AdK1–3(1–354) or AdK1– 5(1–566). Mean values are represented with the s.e.m. A macroscopic view of a representative mouse treated with AdK1–5(1–566) shows at day 42 that the tumor has totally regressed, as does the corresponding histological section of the lesion. (d) Immunostaining of K1–3 (aa 1–354) (3) and K1–5 (aa1–566) (4) on MDA-MB-231-derived tumors injected with AdK1–3(1–354) or AdK1–5(1–566). AdCO1- (1) and PBS- (2) treated tumors are used as control. adenoviral vectors can efficiently transduce the host liver (172.8763 mm3 versus 519.67162 mm3). The AdK1–5(1– and use it as an endogenous factory to produce high 566)-treated mice exhibited tumor growth inhibition circulating levels of the antiangiogenic protein. of 67% (Po0.02) until day 35 (Figure 7a). These results In order to evaluate the efficiency of AdK1–5(1–566) are correlated with the detection of an equivalent and AdK1–3(1–354) in preventing tumor establishment, amount of K1–3(1–354) and K1–5(1–566) in the sera of 5  109 PFU of AdK1–5(1–566), AdK1–3(1–354) and mice (Figure 7b). AdCO1 were injected in the retro-orbital sinus 24 h In order to evaluate the inhibitory effect of AdK1–5(1– before subcutaneous injection of MDA-MB 231 cells. In 566) on the dissemination of lung metastases, total, 87.5% of mice from the AdCO1-treated group 5  109 PFU of AdK1–5(1–566), AdK1–3(1–354) and developed tumors, whereas 27.3% mice from the AdK1– AdCO1 were injected intravenously (i.v.) 24 h before i.v. 3(1–354)-treated group, and 36.4% mice from the AdK1– injection in the retro-orbital sinus of melanoma cells B16- 5(1–566)-treated group, remained tumor free 42 days F10 (day 1). Mice were killed at day 10 and the number of after implantation of MDA-MB 231 cells. The tumor lung metastases determined under a dissecting micro- volumes within the AdK1–5(1–566) group at day 35 scope. Metastases observed in AdK1–5(1–566)-treated were smaller than those from the AdCO1 group mice were considerably smaller than those observed in

Gene Therapy K1–5 (aa 1–566) gene transfer A Galaup et al 836 a a AdCO1 AdK1-3(1-354) AdK1-5(1-566) PBS Ad CO1

Day 15

Day 29 Ad K1-3(1-354) Ad K1-5(1-566)

Day 36

b 2 b Day 8 1.8 Day 15 1.6 4 Day 29 1.4 Day 36 1.2 3 Day 42 1 *** 0.8 2 * * * vessels per mouse

Angiogenic score 0.6

0.4 Number of intra-tumoral 1 0.2 0 0 PBS AdCO1 AdK1-3(1-354) AdK1-5(1-566) AdCO1 AdK3(1-354) AdK1-5(1-566) Figure 5 Inhibition of intratumoral vascularization in PC3-derived Figure 6 Inhibition of MDA-MB-231 tumor growth and functional tumors. (a) The extent of vascularization of a representative tumor vascularization. (a) The extent of intratumoral vascularization was injected with AdCO1, AdK1–3(1–354) and AdK1–5(1–566) is shown 8 assessed by CDHFUS at day 15, 29 and 36 for each group. Vessels were days after injection. (b) The importance of intratumoral vascularization observed in the tumors (arrow) treated with AdCO1 and AdK1–3(1–354). into PC3-treated tumors was assessed by PECAM-1 immunostaining after In contrast, for mice treated with AdK1–5(1–566), no vessels were AdK1–3(1–354) or AdK1–5(1–566) injection (2 Â 109 PFU). Quantifica- observed within the tumor. (b) Visualization of 120 recorded ultrasono- tion was performed by assessing the mean immunoreactive score per field graphy examinations enabled the quantification of functional intratumoral at day 8. vascularization and its weekly evolution at days 8, 15, 29, 36 and 42 for each group. AdCO1-treated mice. The number of lung metastases counted in the AdK1–5(1–566)- and AdK1–3(1–354)- treated groups present a significant decrease in compar- cDNA, or systemic delivery of the vector, may thus ison to the AdCO1-treated group (8776.3 and represent a more efficient method of achieving constant 44712.9%, respectively, Po0.05) (Figure 7c and d). long-lasting intratumoral or systemic concentrations of These results are correlated with the detection of an angiostatin. In this context, we previously reported that a equivalent amount of K1–3(1–354) and K1–5(1–566) in single, local injection of a replication-defective adeno- the sera of mice (Figure 7e). virus encoding a secretable angiostatin-like molecule (AdK1–3(1–354)) into small (20 mm3) xenografts could suppress subsequent tumor growth.22 On the other hand, Discussion we documented that the effects were only marginal when AdK1–3(1–354) was intratumorally or systemically Tumor-secreted angiostatin has been shown to be a injected into larger tumors (420 mm3).23,24 This is physiopathological inhibitor of angiogenesis, driving supported by a study in which bolus injections of the metastases into a dormant state. However, subcutaneous angiostatin protein were reported to exert very different and intraperitoneal bolus injections of human angiostatin antitumor effects depending on the developmental stage have highlighted pharmacological issues as angiostatin of experimental carcinogenesis in a pancreatic murine is rapidly cleared from the circulation. Furthermore, its tumor model.25 These authors documented that end- serum half-life is rather limited in immunocompetent stage pancreatic tumors were resistant to both angio- C57Bl/6 mice,20 raising the possibility that such a high statin and endostatin in mice. In contrast, TNP-470, a clearance rate may be a direct consequence of its low synthetic analogue of fumagillin, and the combination of molecular weight and efficient renal filtration. A pro- angiostatin and endostatin, produced significant tumor longed exposure of purified angiostatin at high doses has regression of large tumors, highlighting the fact that the been shown to be required to maintain cytostatic use of angiogenic inhibitors could become a plausible intratumoral concentrations of angiostatin.21 The direct and important component of anticancer treatment transduction of the tumor and the surrounding tissue strategies. Nevertheless, these observations emphasize with a recombinant virus encoding the angiostatin the need to discover new antiangiogenic molecules that

Gene Therapy K1–5 (aa 1–566) gene transfer A Galaup et al 837 a 700 b 35 PBS AdCO1 600 30 AdK1-3(1-354) AdK1-5(1-566) 25

) 500 3 20 400 15 300 10 protein concentration (nmol/mL)

Tumor volume (mm 200 5

100 0 PBS AdCO1 AdK1-3 AdK1-5 0 (aa1-354) (aa1-566) 0 5 10 15 20 25 30 35 40 Day

cde300 35

30 250

25 200 20 *** 150 15

100 10 protein concentration (nmol/mL)

Number of lung metastasis 5 50 AdK1-5(1-566) AdCO1 0 PBS AdCO1 AdK1-3 AdK1-5 0 (aa1-354) (aa1-566) AdCO1 AdK1-3 AdK1-5 (1-354) (1-566) Figure 7 Comparison of the antitumor effect of AdK1–5(1–566) with that of AdK1–3(1–354) after systemic administration. (a) Effect of systemic injection of AdK1–3(1–354) and AdK1–5(1–566) on the MDA-MB-231 mammary tumors model. PBS or 5 Â 109 PFU of AdK1–3(1–354), AdK1–5(1–566) and AdCO1 was i.v. injected into nude mice 24 h before injection of 4 Â 106 MDA-MB-231 cells. Tumor establishment and growth were followed until day 42 after the injection. Data represent the tumor volume (mean7s.e.m.) for each group. (b) In vivo quantification by ELISA in mice sera of K1–3 (aa 1–354) and K1–5 (aa 1–566) after systemic administration. PBS or 5 Â 109 PFU of AdK1–3(1–354), AdK1–5(1–566) and AdCO1 was i.v. injected into nude or C57Bl/6 mice, 24 h before s.c. injection of 4 Â 106 MDA-MB-231 cells. (c) B16F10 melanoma cell lung invasion in syngenic mice. C57Bl/6 mice were first i.v. injected with 5 Â 109 PFU of AdK1–3(1–354), AdK1–5(1–566) and AdCO1, and 24 h later, animals received an i.v. injection of 2 Â 105 B16F10 cells via the retro-orbital sinus. At 10 days after tumor cell injection, animals were killed, and their lungs excised. The number of metastases were counted under a dissecting microscope. Histograms represent the mean of the measurements. (d) Representative lung of an animal from the AdCO1-treated group and the AdK1–5(1–566)-treated group at day 10 after treatment. (e) In vivo quantification by ELISA in mice sera of K1–3 (aa 1–354) and K1–5 (aa 1–566) after systemic administration. PBS or 5 Â 109 PFU of AdK1–3(1–354), AdK1–5(1–566) and AdCO1 was i.v. injected into nude or C57Bl/6 mice, 24 h before s.c. injection of 2 Â 105 B16F10 cells. are more efficient than angiostatin and endostatin for the and comprises the three cysteine residues in positions treatment of well-established tumors. 543, 555 and 560, allowing the formation of the three Since the K5 (aa 481–566) of plasminogen, which is not disulfide bonds lying within the K5 (aa 481–566). In our present in the originally described isoforms (K1–3 (aa 97– study, we have assessed the antiangiogenic activity of the 354) or K1–4 (aa 97–454)), has been shown to be a key K1–5 (aa 1–566) molecule in vitro and in vivo in different functional domain, which possesses potent antiangio- murine tumor models in comparison to AdK1–3(1–354), genic activity, we have chosen to focus on new an adenovirus expressing the N-terminal fragment of plasminogen-derived isoforms consisting of K1–5 do- plasminogen containing the first three complete kringles mains including the C-terminal regions of K5 (aa 481– (K1–3 (aa 1–354)). 566). To this end, we have constructed a replication- Our findings demonstrate that the inclusion of the defective adenovirus (AdK1–5(1–566)) that expresses a K4 (aa 355–454) and the K5 (aa 481–566) domains in new N-terminal (aa 1–566) form of human plasminogen, the K1–3 (aa 1–354) molecule considerably increased which includes the preactivation peptide and full K1–5, its anti-angiogenic activity, as AdK1–5(1–566)-infected

Gene Therapy K1–5 (aa 1–566) gene transfer A Galaup et al 838 endothelial cells showed a marked dose-dependent nounced when AdK1–5(1–566) was injected, and 40% of arrest of proliferation in vitro, compared to infection of the tumors had totally regressed only in the AdK1–5(1– HUVECs with AdK1–3(1–354), under the same condi- 566)-treated group. Interestingly, the AdK1–5(1–566)- tions. This effect was related to the K1–5 (aa 1–566) transduced MDA-MB-231 tumors apparently failed to molecule and not to a difference in protein levels. Indeed, establish a vascular network within the tumor mass in a specific ELISA for the quantification of K1–3 (aa 1–354) order to support rapid and extensive growth, and this and K1–5 (aa 1–566) levels showed that the quantity of failure translated into a 490% inhibition of tumor both proteins was identical in supernatants harvested growth. For this model, the number of functional at 48 h or 96 h p.i. Furthermore, we have also shown intratumoral vessels was calculated by CDHFUS analysis that a virus-free supernatant from AdK1–5(1–566)- after injection of the D-galactose-based contrast agent infected MRC-5 cells allowed a higher antiangiogenic Levovist. Treatment efficacy was evaluated by monitor- effect compared to the use of the supernatant from ing tumor size and percentage of tumor growth inhibi- AdK1–3(1–354)-infected MRC-5 cells, illustrating the tion until day 42 postinjection of adenovirus. The higher paracrine effect of K1–5(1–566) secreted by resulting data indicated a significant reduction (460%, transduced primary fibroblasts. Po0.05) of intratumoral vascularization within the Fluorogenic peptide substrates specific for caspase-3, tumors from the AdK1–5(1–566)-treated group, which -8 and -9 revealed a more pronounced activation of was correlated with significant tumor growth inhibition caspase-8 showing that K1–3(1–354) and K1–5(1–566) (490%, Po0.05). PECAM-1 immunostaining of PC3 selectively induce endothelial cells to undergo a cell tumor sections also revealed a significant reduction of death that is biochemically identical to apoptosis. We neoangiogenesis in the AdK1–5(1–566)-treated tumors, also performed a cell viability assay with HUVECs which was more pronounced than in the case of tumors infected with 800 PFU/cell of AdK1–5(1–566), AdK1– treated with AdK1–3(1–354). Well-formed vessels with a 3(1–354) or AdCO1 for 96 h in the presence or absence of mature lumen were observed in control PC3 tumors, but caspase-3-, -8 or -9 inhibitors. This assay also demon- never in AdK1–5(1–566)-treated tumors. strated that the caspase-8 inhibitor blocks the K1–5 effect In addition, we showed that the systemic injection of on cell viability of the HUVECs more efficiently than the AdK1–5(1–566) delayed MDA-MD-231 tumor establish- caspase-3 and -9 inhibitors (data not shown). ment, which highlights the paracrine activity of secreted Despite their structural similarity and sequence K1–5 (aa 1–566). In contrast, the systemic injection of homology, K domains appear to have distinct inhibitory AdK1–3(1–354) had no effect on tumor growth, and was profiles in endothelial cell migration assays. We have similar in the AdCO1-treated mice, demonstrating that shown that K1–5 (aa 1–566) molecule is a potent inhibitor K1–5 (aa 1–566) yields stronger antiangiogenic activity of endothelial cell migration, implying that both K4 (aa than K1–3 (aa 1–354). Furthermore, we provide evidence 355–454) and/or K5 (aa 481–566) may be necessary for that AdK1–5(1–566), and not AdK1–3(1–354), could the antiendothelial cell activity of the protein. In contrast, strongly interfere with tumor cell dissemination in the AdK1–3(1–354)-infected HUVEC had only little activity B16-F10 C57Bl/6 syngenic model, by restricting tumor on endothelial cell migration. Many previous studies cell access to the vasculature. have suggested that the functional elements for the Our results clearly show that K1–5 (aa 1–566) is a very inhibition of migration may reside on both the K4 (aa potent therapeutic candidate to target tumor angiogen- 355–454) and K5 (aa 481–566) domains.12,16 esis. It has a more potent effect in inhibiting angiogenesis Considering that neovascularization is required for and tumor growth, because it displays a stronger effect tumor growth in vivo, we wondered whether the marked on endothelial cell growth and motility, and possesses a responsiveness of endothelial cells to AdK1–5(1–566) that longer half-life in vivo, than K1–3 (aa 1–354), due to its was observed in vitro would induce more pronounced increased molecular weight, as described previously.17 antitumoral effects than those exhibited by AdK1–3(1– To date, it has been shown that limited proteolysis of 354). Three different experimental models (human MDA- plasminogen leads to the generation of several isoforms MB-231 breast carcinoma, human PC3 prostatic carcino- of endothelial growth inhibitors, which have a different ma and murine B16 F10 melanoma) and two different range of activity including inhibition of endothelial cell routes of adenoviral delivery (ie intratumoral and proliferation, migration and induction of apoptosis. To intravenous) were employed. We have shown that these date, it seems that K1–3 (aa 97–354) is at least as potent as three models are quite different with regard to their angiostatin K1–4 (aa 97–454) in suppression of endothe- levels of VEGF production and growth when established lial cell growth,11 and that K5 (aa 481–566) and after subcutaneous (s.c.) injection into nude mice (data angiostatin 4.5 (aa 97–549) appear to be more potent not shown). than angiostatin K1–4 (aa 97–454) in vitro and in vivo.15–17 A single intratumoral injection of AdK1–5(1–566) was More recently, it was shown that angiostatin 4.5 (aa 97– shown to inhibit primary tumor growth in two well- 459) initiates a cascade of protease activation events established (420 mm3) xenograft murine models (hu- involving caspase-3, -8 and -9.19 In our study, we have man MDA-MB-231 breast carcinoma and human PC3 shown that K1–5 (aa 1–566) displays a more potent effect prostatic carcinoma), and this effect was not observed on suppression of endothelial cell growth and migration, with AdK1–3(1–354). In all cases, AdK1–5(1–566) dis- but displays no significant difference regarding caspase played a more potent effect on tumor growth than activation versus K1–3 (aa 1–354). It has been shown that AdK1–3(1–354), since the percentage of tumor growth an endothelial cell protein, angiomotin, has a role in inhibition was constantly increased in the AdK1–5(1– mediating migration, and that binding of K1–4 (aa 97– 566)-injected group. The antitumoral effect was more 454) to angiomotin inhibits endothelial cell migration.26 apparent in the case of the MDA-MB-231 model, since Thus, it is plausible that K1–5 (aa 1–566) binds marked inhibition of tumor growth became very pro- angiomotin more strongly than K1–3 (aa 1–354), and

Gene Therapy K1–5 (aa 1–566) gene transfer A Galaup et al 839 therefore inhibits the migration in vitro of the endothelial recombinant adenoviral genome bearing K1–5 was cells more potently. generated by homologous recombination between the With regard to molecular mechanisms underlying the shuttle plasmid and pXL3215 in Escherichia coli,as angiostatic actions of the K structures, there are still described previously.30 After recombination, the adeno- questions that remain to be clarified. In particular, it is viral genome is excised by PacI digestion. Recombinant not known whether there are one or more endothelial- adenoviruses (AdK1–5(1–566)) were obtained following specific receptors for the different isoforms. It appears Lipofectamine Plus-mediated transfection of 293 cells that K5 (aa 481–566) and other isoforms such as K1–4 (aa with PacI-digested DNA (Lipofectamine Pluss, Life 97–454) may act through separate pathways because it Technologies Inc.). AdCO1 is a control E1E3-deleted has been shown that K1–4 (aa 97–454) and K5 (aa 481– adenovirus that does not bear an expression cassette in 566) synergistically suppress endothelial cell growth in place of the E1 genes. AdK1–5(1–566), AdK1–3(1–354) vitro. Therefore, we believe that K1–5 (aa 1–566) and K1– and AdCO1 were further amplified in 293 cells and 3 (aa 1–354) could differentially activate both pathways purified by two-step CsCl gradient ultracentrifugation. and/or use other endothelial cell signals such as Viral titers were quantified both as viral particles integrins,27 focal adhesion kinases28 or ATP synthases29 (vp)/ml by DNA measurement by high-performance on the surface of endothelial cells. liquid chromatography analysis and as PFU/ml Angiostatic therapy using recombinant adenoviruses following infection of 911 cells. Typical titers were has been shown to be experimentally plausible and obtained for the viruses used in this work: AdK1–3 efficient. Nevertheless, the failure of both whole angios- (1–354), 2.4 Â 1011 PFU/ml (4.12 Â 1012 vp/ml); AdK1– tatin protein and a protein composed of only the first 5(1–566), 2.8 Â 1011 PFU/ml (4.86 Â 1012 vp/ml); AdCO1, three K domains to abolish tumorigenesis underlines the 3.6 Â 1011 PFU/ml (6.18 Â 1012 vp/ml). Thus, for the importance in discovering and evaluating new antian- viruses used in the following experiments, viral quanti- giogenic molecules. In our study, we have clearly shown fication showed that 1 PFU is equivalent to 17.2, 17.4 and that the K1–5 (aa 1–566) is a better candidate than K1–3 17.1 vp for AdK1–3(1–354), AdK1–5(1–566) and AdCO1, (aa 1–354) for such a role, as K1–5 (aa 1–566) has been respectively. shown to display a potent effect on the suppression of endothelial cell growth, tumor growth and metastatic Cell lines dissemination. It would be interesting to compare the HUVECs were cultured in EBM-2 medium (Clonetics) new fragment K1–5 (aa 1–566) with other plasminogen- as recommended by the manufacturer. PC3 cells were derived fragments such as angiostatin 4.5 (aa 97–549) or cultured in Ham’s F12 with 10% fetal calf serum (FCS), K5 (aa 481–566). Adenovirus-mediated delivery of K1–5 1mML-glutamine and 1.5 g/l sodium bicarbonate. The (aa 1–566), or purified K1–5 (aa 1–566) protein, may MDA-MB-231 human breast carcinoma and the B16-F10 become a potential clinical treatment of cancer patients, mouse melanoma cell lines were cultured in DMEM with or be used as prevention therapy for metastatic devel- 10% FCS, 1 mML-glutamine and 1.5 g/l sodium bicarbo- opment. nate. HEK 293 and HMEC-1 cells were harvested, as described previously.22,31 Materials and methods Western blot analysis Subconfluent HMEC-1 cells were infected with AdK1– Construction, propagation and infection 5(1–566), AdK1–3(1–354) and AdCO1 at MOI of of recombinant adenoviruses 300 PFU/cell. Cell culture supernatants were collected AdK1–5(1–566) and AdK1–3(1–354) are both replication- 72 h p.i.. In all, 100 mg of protein from each sample was defective recombinant E1E3-deleted adenoviruses direct- run on a NuPage MOPS 10% gel (Novex, San Diego, ing the expression and secretion of plasminogen-derived USA) before being transferred onto a 0.45 mm nitrocellu- factors (including the signal peptide), under the control lose membrane (Schleicher and Schuell). As a control, of the human CMV immediate-early promoter. AdK1– 100 ng of human plasminogen (Stago, Asnie`res, France) 5(1–566) expresses the N-terminal fragment of human were coelectrophoresed with all samples. After overnight plasminogen (up to residue 566) containing K1–5, and incubation in blocking buffer (Tris 1 mM (pH 8), 150 mM AdK1–3(1–354) expresses the N-terminal fragment NaCl, 0.05% Tween 20, 5% nonfat dry milk), the of human plasminogen (up to residue 354) containing membranes were incubated for 3 h with anti-human K1–3.22 plasminogen monoclonal antibody A1D12.23 Membranes DNA fragments encoding K4 and K5 were generated were washed in TBS–0.05% Tween 20 and incubated by PCR from plasminogen cDNA, and were inserted into for 1 h with a horseradish peroxidase-conjugated goat the shuttle plasmid pMP1320 between the N-terminal anti-mouse antibody (Biosys, Compiegne, France). After fragment of plasminogen (containing the 18 amino-acid washing, localization of antibody-hybridized complexes signal peptide of plasminogen followed by K1–3) and the on the membranes was performed with enhanced SV40 late polyadenylation signal. The resulting KmR- chemiluminescence ECL Plus kit (Amersham). SacB-ColE1 plasmid carrying the K1–5 expression cas- sette under the control of CMV promoter is bordered by Cell viability assay the left end of the Ad5 genome (nucleotides 1–386) and HUVECs were seeded at a density of 5 Â 103 cells in part of the Ad pIX gene (nucleotides 3446–4296), which 24-well tissue culture dishes. After 24 h, HUVECs were facilitates homologous recombination with plasmid starved in EBM-2 medium supplemented with 0.2% FCS pXL3215b.20 pXL3215 contains the Ad5 genome deleted for 24 h. An optimal stimulatory concentration of for E1 (nucleotides 1–386), E3 (nucleotides 28592–30470), 3 ng/ml for VEGF was used for all viability experiments. and bordered by two PacI sites. A plasmid encoding a To ascertain the AdK1–5(1–566) dose yielding 50% of

Gene Therapy K1–5 (aa 1–566) gene transfer A Galaup et al 840 inhibition (ID50), a survival test was performed after 96 h Caspase activity of treatment for two independent experiments. Endothe- To assess intracellular caspase activity, HUVECs seeded lial cells were infected with AdK1–5(1–566), AdK1–3(1– at 5 Â 105 cells in six-well tissue culture dishes were 354) or AdCO1 at an MOI of 100, 300, 600 and 800 PFU/ starved 24 h postseeding in EBM-2 media supplemented cell (n ¼ 6 for each condition) in 250 ml of medium with 0.2% FCS for a 24 h period. An optimal stimulatory supplemented with 2% FCS for 4 h, before the addition concentration of 3 ng/ml for VEGF was used for all of 500 ml of culture medium supplemented with 2% FCS. the caspase experiments. To compare the AdK1–5(1–566) In each experiment, an internal control (ie cells that were and AdK1–3(1–354), a caspase activity assay was mock treated) was included in each culture plate. Cell performed for two independent experiments. Endothe- survival was quantified by a cell viability assay using lial cells were infected with AdK1–5(1–566), AdK1–3(1– MTT (3-[4,5-dimethylthiazole-2-yl]-2,5-diphenyltetrazo- 354) or AdCO1 at 800 PFU/cell in 3 ml of medium 24 lium bromide – Sigma), as described previously. The supplemented with 2% FCS. In each experiment, 50 nM percentage of viable cells was calculated by comparing of staurosporine was included as a positive control. After the OD of the samples with those incubated with 48 h, cells were resuspended in 50 ml of cell lysis buffer, noninfected cell supernatants. and the protease activity of the lysates was evaluated by In addition, a cell viability assay was performed with caspase-3, -8, and -9 Fluorometric Assay Kits (R&D HUVECs using supernatant of AdK1–5(1–566), AdK1– Systems, Minneapolis, MN, USA). The lysates were 3(1–354) or AdCO1-transduced human fibroblasts (MRC- diluted 1:2 in reaction buffer and incubated with the 5 cells). MRC-5 cells were infected with a dose of caspase-3 (-8 or -9)-specific peptides conjugated to 7- 800 PFU/cell for 4 h. Viruses were then removed by amino-4-trifluromethyl coumarin at 371C. After 2 h of washing the cells three times with PBS. MEM supple- incubation, the optical density was determined using a mented with 5% of SVF was then added to the vials and microplate reader set at 405 nm. Bradford protein assays supernatants were collected 96 h after infection and were performed on each lysate to normalize the OD unit exposed for 30 min to ultra-violet light in order to value and results are expressed as OD (405 nm) for completely inactivate the virus. In this assay, HUVECs 100 mg of proteins. were seeded at a density of 5 Â 103 cells in 24-well tissue culture dishes and after 24 h were incubated with 400 ml of EBM-2 medium supplemented with 3 ng/ml of VEGF Tumor models and 400 ml of each supernatant. After 96 h, the cells Cultured MDA-MB 231 and PC3 cells were harvested, were quantified with a cell viability assay using MTT washed and resuspended in PBS at 20 Â 106 cells/ml, and compound. 200 ml of the cell suspension was injected s.c. into the flank of 8-week-old athymic mice. To evaluate the Human K1–3 (aa 1–354) and K1–5 (aa 1–566) in vitro antitumor effect of AdK1–5(1–566), five independent experiments were performed with MDA-MB-231 and quantification PC3 cells, and the animals were randomized into two To quantify the amount of K1–3 (aa 1–354) and K1–5 (aa groups when tumor volumes had reached mean values 1–566) in supernatants of AdK1–3(1–354)- and AdK1– of 62715 and 3074mm3, respectively, per group (n ¼ 5– 5(1–566)-infected cells, HUVECs were plated in 24-well 10/group). All tumors received one intratumoral injec- tissue culture dishes and infected with AdK1–3(1–354), tion of 2 Â 109 PFU AdK1–5(1–566), AdK1–3(1–354) or AdK1–5(1–566) or AdCO1 at an MOI of 300 PFU/cell in AdCO1 in a volume of 50 ml of PBS. One additional m m 250 l of medium for 2 h, before adding 500 l of culture group received 50 ml of PBS. In order to evaluate medium. Noninfected cells were included in each culture treatment efficacy, tumor sizes were monitored until plate as control. The culture supernatants were collected day 20 p.i. for the PC3 models, and up to 42 days p.i. and the cells counted 48 and 96 h p.i. For each analysis, injection for the MDA-MB-231 model. Tumor volumes samples were diluted to 1:10 and run with the standards were calculated with the following formula: (d3 Â p)/6, in duplicate. For the quantification, a solid-phase ELISA 24 where d is the mean diameter of the tumor. Tumor was used as described previously. volumes were converted to values relating to the initial

tumor volume and expressed by the ratio Vx/V0, where

Migration assays Vx is the mean relative volume of treated tumors on day

For the migration assay, HUVECs were seeded at a x and V0 the tumor volume at the start of the treatment density of 2 Â 105 cells in EBM-2 medium supplemented period. The percentage of growth inhibition (100–T/C%) with 2% FCS in fibronectin-coated (10 mg/ml) Petri was calculated to express treatment efficacy, where T is

dishes (35 mm diameter) the day before viral infection. the ratio of Vx/V0 of treated tumors on day x and C is the

Migration assays were performed in triplicate in two ratio of Vx/V0 of control tumors. independent experiments as described previously.32 AdK1–5(1–566) and AdK1–3(1–354) were also tested to HUVECs were infected with AdK1–5(1–566), AdK1– assess the effect of AdK1–5(1–566) and AdK1–3(1–354) 3(1–354) or AdCO1 at an MOI of 800 PFU/cell in 500 ml on tumor establishment and progression when adeno- of medium supplemented with 0.2% FCS for 4 h. The viruses were delivered i.v. In the ‘prevention’ cells were then washed and a ‘scratch wound’ was experiment, 5 Â 109 PFU of AdK1–5(1–566), AdK1–3(1– created by scraping the cell monolayer with a sterile 354) or AdCO1 were injected systemically in the retro- pipette tip, followed by the addition of culture medium orbital sinus of 8-week-old female nude mice (n ¼ 12) 24 h supplemented with 0.2% FCS and 30 ng/ml of VEGF. before s.c. injection of 4 Â 106 MDA-MB-231 cells, Every 24 h for a 4-day period, migration was quantified which were resuspended in a volume of 200 ml of PBS. by counting cell numbers at indicated migration dis- Tumor establishment and growth were followed until tances from the wound edge. day 42 p.i.

Gene Therapy K1–5 (aa 1–566) gene transfer A Galaup et al 841 To evaluate the antimetastatic activity of K1–5 (aa 1– Images were then analyzed with a Linux-based specially 566), mice (n ¼ 6–7) received one i.v. injection of developed program, producing a quantitative index 5  109 PFU of AdK1–5(1–566), AdK1–3(1–354) or from 0 to 255. Digitalized color images were transformed AdCO1 in a volume of 100 ml of PBS via the retro-orbital into various gray-level images. A theoretical image sinus. After 24 h, cultured B16-F10 melanoma cells were exclusively composed of red-brown vessels would harvested, washed and resuspended in PBS at 2  105 correspond to an index of 255, whereas an image totally cells/ml, of which 100 ml were injected i.v. via the retro- devoid of vessels (entirely stained blue) is associated orbital sinus of 8-week-old syngenic C57Bl/6 mice. At 10 with an index of 0. Each pixel of the image is associated days after tumor cell injection, animals were killed, their with a figure between 0 and 255 and a mean value of all lungs excised and the number of metastases counted pixels is obtained for each image. The final index (score) under a dissecting microscope. For all the experiments, for each animal is obtained by calculating the mean value a Student’s t-test was used for statistical analysis. of the eight contiguous fields. CDHFUS was used in order to evaluate inhibition of In vivo quantification of human K1–3 (aa 1–354) tumor angiogenesis for the MDA-MB-231 tumors model. Weekly examinations were performed on MDA-MB-231 and K1–5 (aa 1–566) tumors (five for control or AdK3(1–354)-treated mice and In order to detect K1–3 (aa 1–354) and K1–5 (aa 1–566) in six for AdCO1- or AdK1–5(1–566)-treated animals) after tumors after viral local administration, PC3- and MDA- the beginning of treatments (day 2) for a total of 6 weeks MB-231-derived tumors were collected and immuno- (day 42) with CDHFUS. Examinations were performed stained for human angiostatin, respectively, 6 and 42 with an ATL-HDI 5000 sonograph using a 12 MHz linear days after the injection of the adenovirus. For angiostatin probe. For all the 120 investigations, sonograph settings immunohistochemistry, samples were fixed in Finefix remained unchanged within the 6 weeks of examina- (Microm). Heat-induced epitope retrieval in 10 mmol/l tions. For each investigation, the protocol was: (1) in citrate buffer (pH 6.0) was performed. After quenching of color Doppler mode after contrast agent (Levovist, endogenous peroxidase activity, sections were incubated Schering) injection; (2) counting of the number of with goat anti-angiostatin primary antibody (R&D intratumoral vessels visualized by a color scale on the systems, 1:50) for 1 h and processed with biotinylated 2D maximal transversal scan of the tumor. Vessels are rabbit anti-goat secondary antibody (1:50) for 30 min, colored in red (inversely in blue) when the flow direction followed by streptavidin-horseradish peroxidase (1:100, inside the vessel arrives toward (respectively, moves Dako) 30 min. Slides were treated with DAB chromo- away from) the ultrasound beam. The color intensity, genic substrate (Microm) for 10 min and counterstained in each red or blue scale, depends on the mean with Mayer’s hematoxylin and mounted. To quantify the velocity module inside the vessel. After 6 weeks, all the amount of K1–3 (aa 1–354) and K1–5 (aa 1–566) in mice MDA-MB-231-derived tumors were collected and sera after viral systemic administration, blood samples stained with HES. were collected 8 days after adenovirus delivery, from AdK1–3(1–354)- and AdK1–5(1–566)-injected mice, sera from PBS- and AdCO1-injected mice were systematically included as controls. For each analysis, samples were Acknowledgements diluted to 1:100 and run with the standards in duplicate. We sincerely thank the SCEA and specially M Stanciu, For the quantification, a solid-phase ELISA was used, as D Challuau and P Ardouin for animal care, E Connault described previously.24 for technical assistance, I Chawi for echography analysis, C Bouquet for the kind gift of pMP13 plasmid and Quantification of angiogenesis N Lamande´ (Inserm U36 – Colle`ge de France) for the To evaluate the antiangiogenic effect of AdK1–5(1–566) kind gift of HUVECs. We warmly acknowledge B Mullan and AdK1–3(1–354), PC3 tumors with mean volumes of and M Mackenthun for critical reading. Le Centre 62.5715 mm3 or MDA-MB 231 tumors with mean National de la Sante´ et de la Recherche Scientifique volumes of 3074mm3 were injected at day 0 with a (CNRS), la ligue nationale contre le cancer and l’Associa- single injection of 2  109 PFU of AdK1–5(1–566), AdK1– tion pour la Recherche sur le Cancer (ARC) are acknowl- 3(1–354) or AdCO1 in a volume of 50 ml of PBS. edged for financial support. PC3-derived tumors were collected when their vo- lumes had reached mean values of 524724 mm3 in order to limit necrosis and to improve the image analysis for References vessel quantification. Following fixation of the tumors in absolute ethanol, paraffin sections (4 mm thick) were 1 Hanahan D, Folkman J. Patterns and emerging mechanisms 86 prepared and stained with HES. PC3-derived tumors of the angiogenic switch during tumorigenesis. Cell 1996; : 353–364. were immunostained for murine CD31 (PECAM-1) and 2 Folkman J. Angiogenesis in cancer, vascular, rheumatoid and analyzed using a Zeiss Axiophot microscope (Germany) 24 other disease. Nat Med 1995; 1: 27–31. and a Sony 3CCD camera as described previously. For 3 O’Reilly MS et al. Angiostatin: a novel angiogenesis inhibitor each animal, a representative histological sample im- that mediated the suppression of metastases by Lewis lung munostained for murine CD31 was submitted to image carcinoma. Cell 1994; 79: 315–328. analysis using a Zeiss Axiophot microsope (Germany) 4 O’Reilly MO et al. Endostatin: an endogenous inhibitor of and a Sony 3CCD camera (resolution 768  576 pixels). angiogenesis and tumor growth. Cell 1997; 88: 277–285. The choice of a  100 magnification enabled a digitiza- 5 Kamphaus GD et al. Canstatin, a novel matrix-derived tion of nearly the whole specimen in all cases (eight inhibitor of angiogenesis and tumor growth. J Biol Chem 2000; contiguous fields, excluding necrotic and fibrotic areas). 275: 1209–1215.

Gene Therapy K1–5 (aa 1–566) gene transfer A Galaup et al 842 6 Clapp C et al. The 16 kDa N-terminal fragment of human 20 Bouquet C, Frau E et al. Systemic administration of a prolactin is a potent inhibitor of angiogenesis. Endocrinology recombinant adenovirus encoding a HSA-angistatin kringle 1993; 133: 1292–1299. 1–3 conjugate inhibits MDA-MB-231 tumor growth and metas- 7 Gupta SK, Hassel T, Singh JP. A potent inhibitor of endothelial tasis in a transgenic model of spontaneous eye cancer. Mol Ther cell proliferation is generated by proteolytic cleavage of the 2003; 7: 174–183. chemokine platelet factor 4. Proc Natl Acad Sci USA 1995; 92: 21 Dong Z, Kumar R, Yang X, Fidler I. Macrophage-derived 7799–7803. metalloelastase is responsible for the generation of angiostatin 8 Gately S et al. Human prostate carcinoma cells express in Lewis lung carcinoma. Cell 1997; 88: 801–810. enzymatic activity that converts human plasminogen to the 22 Griscelli F et al. Angiostatin gene transfer: inhibition of tumor angiogenesis inhibitor, angiostatin. Cancer Res 1996; 56: growth in vivo by blockage of endothelial cell proliferation with 4887–4890. a mitosis arrest. Proc Natl Acad Sci USA 1998; 95: 6367–6372. 9 O’Mahony CA et al. Angiostatin generation by human pancreatic 23 Griscelli F et al. Combined effects of radiotherapy and cancer. J Surg Res 1998; 77: 55–58. angiostatin gene therapy in glioma tumor model. Proc Natl Acad 10 Westphal JR et al. Angiostatin generation by human tumor cell Sci USA 2000; 97: 6698–6703. lines: involvement of plasminogen activator. Int J Cancer 2000; 86: 24 Galaup A et al. Combined effects of docetaxel and angiostatin 760–767. gene therapy in prostate tumor model. Mol Ther 2003; 6: 731–740. 11 Cao Y et al. Kringle domains of human angiostatin. J Biol Chem 25 Bergers G et al. Effects of angiogenesis inhibitors on multistage 1996; 271: 29461–29467. carcinogenesis in mice. Science 1999; 284: 808–812. 12 Ji RW et al. Characterization of kringle domains of angiostatin as 26 Troyanovsky B et al. Angiomotin: an angiostatin binding protein antagonist of endothelial cell migration, an important process in that regulates endothelial cell migration and tube formation. angiogenesis. FASEB J 1998; 12: 1731–1738. J Cell Biol 2001; 152: 1247–1254. 13 Tanaka T, Cao Y, Folkman J, Fine HA. Viral vector-targeted 27 Brooks PC et al. Integrin alpha vs beta 3 antagonists promote antiangiogenic gene therapy utilizing an angiostatin comple- tumor regression by inducing apoptosis of angiogenic blood mentary DNA. Cancer Res 1998; 58: 3362–3369. vessels. Cell 1994; 79: 1157–1164. 14 McDonald NJ et al. The tumor-suppressing activity of angiosta- 28 Claesson-Welsh L et al. Angiostatin induces endothelial cell tin protein resides within kringles 1 to 3. Biochem Biophys Res apoptosis and activation of focal adhesion kinase independently Commun 1999; 264: 469–477. of the integrin-binding motif RGD. Proc Natl Acad Sci USA 1998; 15 Cao Y et al. Kringle 5 of plasminogen is a novel inhibitor of 95: 5579–5583. endothelial cell growth. J Biol Chem 1997; 272: 22924–22928. 29 Moser TL et al. Angiostatin binds ATP synthase on the surface 16 Ji RW et al. Selective inhibition by kringle 5 of human of human endothelial cells. Proc Natl Acad Sci USA 1999; 96: plasminogen on endothelial cell migration, an important process 2811–2816. in angiogenesis. Biochem Biophys Res Commun 1998; 247: 414–419. 30 Crouzet J et al. Recombinational construction in Escherichia coli of 17 Cao R et al. Suppression of angiogenesis and tumor growth by infectious adenoviral genomes. Proc Natl Acad Sci USA 1997; 18: the inhibitor K1–5 generated by plasmin-mediated proteolysis. 1414–1419. Proc Natl Acad Sci USA 1999; 96: 5728–5733. 31 Stratford-Perricaudet L, Maked I, Perricaudet M, Briand P. 18 Li F et al. Human glioma cell BT325 expresses a proteinase that Widespread long-term gene transfer to mouse skeletal muscles converts human plasminogen to kringle 1–5-containing frag- and heart. J Clin Invest 1992; 90: 626–630. ments. Biochem Biophys Res Commun 2000; 278: 821–825. 32 Tamura M et al. Inhibition of cell migration spreading, 19 Handford AH et al. Angiostatin 4.5-mediated apoptosis of focal adhesions by tumor suppressor PTEN. Science 1998; 280: vascular endothelial cells. Cancer Res 2003; 63: 4275–4280. 1614–1617.

Gene Therapy