Endostatin Therapy Reveals a U-Shaped Curve for Antitumor Activity

ORIGINAL ARTICLE Endostatin therapy reveals a U-shaped curve for antitumor activity RM Tjin Tham Sjin1, J Naspinski2,5, AE Birsner1,CLi2, R Chan1, K-M Lo3, S Gillies3, D Zurakowski4, J Folkman1, J Samulski2 and K Javaherian1 1Vascular Biology Program, Department of Surgery, Children’s Hospital Boston, Harvard Medical School, Boston, MA, USA; 2Gene Therapy Center and Department of Pharmacology, University of North Carolina, Chapel Hill, NC, USA; 3EMD Lexington Research Center, Billerica, MA, USA and 4Departments of Orthopaedic Surgery and Biostatistics, Children’s Hospital Boston, Harvard Medical School, Boston, MA, USA

Developing continuous systemic delivery of endostatin has been a goal of many laboratories. We have employed a method of gene therapy utilizing different viral constructs. Here, we report that a new serotype of adeno-associated viruses, which incorporates canine endostatin, provides dose-dependent transgene expression in the circulation after intramuscular injection in mice. Elevated levels of endostatin remained stable in the circulation for at least 4 months. In vitro assays determined that the protein expressed was biologically active. Antitumor activities of the above construct demonstrated a U-shape curve, where the maximum activity was observed within a certain critical concentration range. These data suggest that an optimum dose range may be required to achieve therapeutic efficacy in large animal models. Cancer Gene Therapy (2006) 13, 619–627; published online 3 February 2006; doi:10.1038/sj.cgt.7700938 Keywords: endostatin; AAV-1; antitumor; angiogenesis; U-shaped activity

Introduction Endostatin inhibits phosphorylation of focal adhesion kinase via binding to a5b1 integrin.9 Cell surface Angiogenesis, the sprouting of new capillaries from pre- glypicans have also been shown to be low-affinity 10 existing vascular beds, is critical for many pathological endostatin receptors. Endostatin has been implicated disorders, such as tumor growth, development and in several signaling pathways, such as downregulation of 11 12 13 metastasis.1 An extensive network of stimulators and c-myc, RhoA activity and cyclin-D1, blockage of 14,15 inhibitors regulates angiogenesis, and the net balance VEGF signaling, and inhibition of the wnt-signaling 16 between these positive and negative regulators determines pathway. Furthermore, endostatin has been shown 17–19 the angiogenic phenotype.2–4 Antiangiogenic therapies to bind and inactivate metalloproteinases and to 20 have therefore been used to target the tumor vascular regulate a spectrum of genes that suppress angiogenesis. system in cancer therapy due to their high degree of Recently, we have shown that an NH2-terminal 27-amino- specificity for endothelial cells,3 low toxicity2 and relative acid synthetic peptide of endostatin, corresponding to lack of acquired drug resistance.5 the zinc-binding domain, is responsible for its anti- 21 Endostatin is a 20 kDa COOH-terminal proteolytic tumor properties. fragment of collagen type XVIII, first isolated from Gene therapy is a feasible approach to combat cancer, murine hemangioendothelioma cell conditioned media with a sustained systemic therapeutic release of anti- as an antiangiogenic agent capable of inhibiting angiogenic substances and a highly localized delivery. tumor growth without displaying toxic side effects.3,6–8 Sustained elevated blood levels obtained by continuous administration from an implanted pump have been demonstrated to be significantly more effective against Correspondence: Dr K Javaherian, Vascular Biology Program, tumors in animals than intermittent peak levels from Department of Surgery, Children’s Hospital Boston, Karp Family bolus dosing.7 There are now many studies demonstrating Research Laboratories, Rm. 11.213, One Blackfan Circle, Boston, the efficacy of antiangiogenic gene therapy in tumor- MA 02115, USA. bearing animal models. Adenoviruses are the most E-mail: [email protected] 22–29 5Current address: Merck Research Laboratories, Merck & Co., Inc., commonly used vectors for this strategy. However, West Point, PA, USA. expression of antiangiogenic molecules mediated by Received 8 August 2005; revised 18 November 2005; accepted 16 adenovirus-based vectors is limited by an effective host December 2005; published online 3 February 2006 immune response and transient nature of expression. Endostatin’s U-shaped antitumor activity RM Tjin Tham Sjin et al 620 Adeno-associated virus (AAV)-based vectors are emer- Construction and purification of adeno-associated ging as potential alternative vectors for cancer therapy viruses because AAV is a non-pathogenic vector with limited host pTR/canine FcES was constructed by inserting cFcES immune response, and allows for long-term and stable DNA fragment into the region of green fluorescence transgene expression. Furthermore, the inhibition of protein (GFP) in the plasmid pTRUF.37 Recombinant tumor growth was possible using AAV-based vectors AAV vectors were produced, purified and titered as expressing appropriate levels of endostatin.30–34 described previously.38 Initially, we set out to treat spontaneous tumors in canines using AAV-mediated canine FcEndostatin ELISA determination of canine FcEndostatin (cFcES). Canines develop spontaneous tumors and serve Serum samples were obtained every 2 weeks by retro- as a suitable model for human cancer. There are several orbital puncture with non-heparinized capillary tubes advantages of using Fc-encoded endostatin: (1) the half- (Fisher, Pittsburgh, PA) after anesthesia. Samples were life is longer; (2) higher expression is obtained; and (3) the placed at 41C overnight and serum was collected after purification process is facilitated. Furthermore, human centrifugation at 10 000 r.p.m. for 10 min. Canine FcEndo- and murine FcEndostatin have been successfully used in 35 statin concentrations were determined by competition tumor inhibition studies. Among 11 serotypes of AAV ELISA (Cytimmune Sciences, College Park, MD) follow- isolated, AAV serotype 1 (AAV-1) has the most potential ing the manufacturer’s protocol. Final endostatin transduction in muscle than other serotypes. Therefore, in concentration was determined by taking the average of this study we tested the efficacy of AAV-1/cFcES the last time points that showed maximum expression. transduction on muscles in SCID mice before initiating canine clinical trials. We show that AAV-1-mediated cFcES gene therapy offers efficient and systemic thera- Western blot determination of canine FcEndostatin peutic release of cFcES for human pancreatic tumor expression inhibition in SCID mice, and that optimum dosing Serum from infected mice was collected as described follows a biphasic curve that is U-shaped. above and analyzed by Western blot for canine endostatin. A detailed description of this procedure was given previously.39 Serum samples (100 ml) of AAV-1/cFcES- and AAV-1/GFP-infected mice were compared with Materials and methods different concentrations of recombinant cFcES. The Cell culture and reagents immunoblot was probed with rabbit anti-canine endo- Human pancreatic adenocarcinoma BxPC-3 cells, which statin antibody (1:2000). Development was carried are p53À/À, were grown and maintained as described out with goat anti-rabbit IgG-horseradish peroxidase previously.2,7 Human umbilical cord endothelial cells secondary antibody (Sigma, St Louis, MO, 1:1000) and (HUVEC; Clonetics, Walkersville, MD) were cultured in chemiluminescence (Pierce, Rockford, IL). endothelial cell growth medium (EGM-2; Clonetics). BxPC-3 cells were obtained from the American Type Matrigel tube formation assay Culture Collection (ATCC, Rockville, MD). Tube formation assay using HUVEC cells was performed Purified rat anti-mouse CD31 (PECAM) was obtained as described previously.39 Serum samples (250 ml) of from BD PharMingen (San Diego, CA). Rabbit anti- AAV-1/cFcES- and AAV-1/GFP-infected mice were used canine endostatin antibodies were produced by immuniz- and compared to recombinant cFcES (10 mg/ml). ing rabbits with recombinant canine endostatin. Animals were exsanguinated after four immunizations. Animal studies Cloning of canine endostatin All animal procedures were performed in compliance with Canine endostatin was cloned from a canine liver cDNA Boston Children’s Hospital guidelines, and protocols were library (Clontech, Palo Alto, CA) using degenerate PCR approved by the Institutional Animal Care and Use primers based on known human and murine endostatin Committee. Male (6–8 weeks) SCID mice (Massachusetts DNA sequences. Forward primer (50-GCAGATGACA General Hospital, Boston, MA) were used. Animals TCCTGGCCA-30) and reverse primer (50-TGGCTCAT were anesthetized via inhalation of isoflurane (Baxter, CTTYGTGG-30) were synthesized by GIBCO BRL Deerfield, IL) before all surgical procedures and observed (Gaithersburg, MD). The reaction was allowed to until fully recovered. Animals were euthanized by a lethal denature at 941C for 1 min, anneal at 551C for 1 min, dose of CO2 asphyxiation. and polymerize at 721C for 1.5 min for 40 cycles using QIAGEN Taq polymerase (QIAGEN, Valencia, CA) and Tumor model and adeno-associated virus injections Q buffer. Human pancreatic BxPC-3 tumor cells were implanted as The canine immunoglobin Fcg chain was fused at the described previously.7 Tumors were measured every 3–5 N-terminus of canine endostatin to make cFcES. This days in two diameters with a dial caliper. Volumes were construct was inserted into the pdCs expression vector,36 determined using the formula (Width)2 Â Length Â 0.52. where cFcES is under the control of human cytomegalo- Data are represented as volume of treated tumor over virus (CMV) enhancer and promoter. control (T/C).

Cancer Gene Therapy Endostatin’s U-shaped antitumor activity RM Tjin Tham Sjin et al 621 Mice were infected with the indicated particles intra- all three histidines (at positions 1, 3 and 11) and aspartic muscularly (i.m.). BxPC-3 cells were implanted into acid (at position 76) residues important for Zn binding are infected mice at 45 days post-infection. conserved.41

Immunohistochemistry Dose-dependent expression of endostatin Tumors were fixed in 10% buffered formalin overnight We and others have determined that, among the new at 41C. The next day, tumors were washed three times serotypes of AAV, AAV-1 has better transduction in PBS and paraffin-embedded. Sections (5 mm) were efficiency in muscle.42,43 Therefore, AAV-1/cFcES vectors permeabilized with 40 mg/ml proteinase K (Roche Diag- were produced in which cFcES is under the control of the nostics Corp., Indianapolis, IN) in 0.2 M Tris-HCl buffer human CMV enhancer and promoter (Figure 2a). Before (pH 7.6) for 25 min at 371C, and washed with PBS, and initiating canine clinical trials, we evaluated the efficacy of then incubated with PECAM (1:250) overnight at 41C. the AAV-1/cFcES construct in SCID mice. Stainings were amplified using tyramide signal amplifica- Different doses of AAV-1/cFcES particles were injected tion direct and indirect kits (NEN Life Science Products i.m. into SCID mice (Figure 2b). As control, AAV-1/GFP Inc., Boston, MA). Sections were photographed at Â 200 vector was used (5.5 Â 1011 particles). Peak levels of magnification using a NIKON TE300 microscope cFcES were reached at approximately 8 weeks post- (Melville, NY). Vessel density (average of three fields) infection and were stable up to the end of the experiment was determined by IPLab software. (15 weeks). Different serum levels of endostatin were observed, with the highest level of B4225 ng/ml (2 Â 1012 Statistical methods particles of AAV-1/cFcES) and the lowest level of Data are expressed as mean7s.d. Statistical significance 9 B19 ng/ml (AAV-1/GFP and 4 Â 10 particles of AAV-1/ was assessed by analysis of variance (ANOVA) with a cFcES). There was an excellent correlation between serum significant overall F-test followed by treatment compar- levels and the doses of AAV-1/cFcES particles injected isons relative to AAV-1/GFP control by Dunnett’s post- (Pearson r ¼ 0.9977; Figure 2c). These data suggest that hoc tests in order to protect against Type I errors due to stable and different endostatin levels could be achieved by multiple pairwise group comparisons against a single injecting different doses of AAV-1/cFcES. control.40 Two-tailed Po0.05 was considered statistically significant. Statistical analysis was performed using the SPSS software package (version 14.0, SPSS Inc., Characterization of biological activity of adeno- Chicago, IL). associated virus-1-produced canine FcEndostatin To determine the biological activity of AAV-1-produced cFcES from mouse serum, first cFcES expression was analyzed by Western blot to identify the non-degraded Results protein (Figure 3a). The cFcES expression in sera Cloning of canine endostatin from mice infected with AAV-1/cFcES (2 Â 1012 particles, Canine endostatin was cloned from a canine liver cDNA n ¼ 2) for 15 weeks was compared to that in sera from a library using degenerate primers based on conserved mouse with AAV-1/GFP and different concentrations DNA sequences obtained from human and murine of purified recombinant cFcES protein. No detectable endostatin. Amino-acid comparison showed that canine cFcES was observed from serum of the AAV-1/GFP- endostatin is 84.8 and 83.7% identical to human and infected mouse, whereas serum samples from the AAV-1/ murine endostatin, respectively (Figure 1). Furthermore, cFcES-infected mice revealed a protein of size similar to

Figure 1 Alignment of human, murine, and canine endostatin. The protein sequences of human, murine and canine endostatin were compared using the ESPript 2.2 program. Identical amino-acid residues are indicated by red overlap.

Cancer Gene Therapy Endostatin’s U-shaped antitumor activity RM Tjin Tham Sjin et al 622

Figure 2 Expression levels of cFcES after injections with several doses of AAV-1/cFcES. (a) Schematic diagram of the AAV-1/cFcES construct. The cFcES cDNA was inserted into the region of GFP in the plasmid pTRUF. Canine FcES expression is under the control of the CMV immediate-early promoter and enhancer. ITR is inverted Figure 3 Characterization of AAV-1-produced cFcES. (a) Western terminal repeat, pA1 is the SV40 polyadenylation signal from SV40 analysis of cFcES after AAV-1/cFcES infection (2 Â 1012 particles). genome, pTK is the thymidine kinase promoter of herpes simplex Serum samples were collected from two mice (M1 and M2) 15 weeks virus, pA2 is the bovine growth hormone polyadenylation signal of after infection. Control serum from one AAV-1/GFP-infected mouse pRc/CMV (Invitrogen, Carlsbad, CA). (b) Serum levels of cFcES was also used. Serum samples (100 ml) were run on SDS-PAGE. As after AAV-1/cFcES infection. Adeno-associated virus-1/cFcES par- controls, different amounts of recombinant cFcES were used. kDA ticles at several doses were injected i.m. into SCID mice. As control, markers are indicated. Arrow indicates cFcES. (b) Inhibition of 11 AAV-1/GFP (5.5 Â 10 particles) was used. Final serum levels were HUVEC tube formation by AAV-1-produced cFcES. The same serum determined by averaging levels of 8, 10, 12 and 15 weeks. Doses of samples used for the Western analysis were tested for this assay. particles injected with corresponding serum levels are indicated. Recombinant cFcES (10 mg/ml) was used as a positive control for 11 9 Group sizes: n ¼ 10, except for 4 Â 10 and 4 Â 10 infected tube formation inhibition. Bar indicates 1 cm. (c) Quantification of particles (n ¼ 9). (c) Correlation between endostatin serum levels tube formation. Branch points per cm2 were counted from three and amount of AAV-1/cFcES particles injected into SCID mice. different fields (3 Â 4 cm) for each treatment. *Po0.001, ANOVA with Pearson r was determined by the Microsoft Excel program. post-hoc Dunnett’s tests.

recombinant cFcES (Figure 3a). This analysis shows that These data show that recombinant cFcES could the cFcES protein secreted from muscles transduced with inhibit HUVEC tube formation and suggest that the AAV-1 vectors is not degraded. AAV-1-produced cFcES is biologically active. To test the biological activity of the AAV-1-produced cFcES, Matrigel tube formation assay was performed. Sera from the AAV-1/cFcES- and AAV-1/GFP- Systemic inhibition of human pancreatic cancer by infected mice (15 weeks) were tested and compared adeno-associated virus-1/canine FcEndostatin follows to purified recombinant cFcES for their ability to a biphasic dose curve inhibit HUVEC tube formation (Figure 3b). The number To determine the optimal endostatin serum level for of branch points/cm2 were significantly lower for human pancreatic cancer inhibition, pancreatic BxPC-3 recombinant cFcES (Po0.001) and AAV-1/cFcES tumor cells (5 Â 106) were injected s.c. into dorsa of the (Po0.001) treatments compared to AAV-1/GFP control AAV-1-infected mice 45 days after infection. Tumor (Figure 3c). growth was followed up for 59 days and measurements

Cancer Gene Therapy Endostatin’s U-shaped antitumor activity RM Tjin Tham Sjin et al 623 were taken every 3–4 days. Human pancreatic cancer was Antiangiogenic effect of adeno-associated inhibited by B33% (P ¼ 0.03) only in mice infected with virus-1/canine FcEndostatin 4 Â 1010 AAV-1/cFcES particles, whereas doses higher We also determined the antiangiogenic effect of AAV-1/ or lower had no effect (Figure 4a). At the completion of cFcES treatment. Therefore, pancreatic cancers treated the experiment, pancreatic BxPC-3 tumors were excised with AAV-1/cFcES and AAV-1/GFP were analyzed for (Figure 4b) from the mice and tumor volumes were vessel density (CD31) (Figure 5a and b). Human determined by weight. Human pancreatic cancer was pancreatic tumor sections from day 59 were formalin inhibited by B32% (P ¼ 0.03; Figure 4b) only at serum fixed and paraffin embedded and then stained with CD31 levels of approximately 126 ng/ml, when tumor volumes antibodies. There were no significant differences in (based on weight) were plotted against serum cFcES %CD31/hpf between serum level groups compared to levels. This is comparable to the data from the earlier AAV-1/GFP (all P40.80) except for 126 ng/ml level measurements (see Figure 4a). When cFcES serum (P ¼ 0.02, Figure 5b). These data further corroborate the levels were higher or lower than 126 ng/ml, antitumor U-shaped tumor response to endostatin that was observed efficacy was lower. A slight tumor inhibition was in the previous experiment (see Figure 4). observed for a serum level of 1040 ng/ml (B11%, P ¼ NS). However, this inhibition was not statistically Optimal endostatin serum level for human pancreatic significant. These data indicate that there is a U-shaped cancer inhibition is between 126 and 173 ng/ml response of human pancreatic tumor growth to To determine more precisely the optimal endostatin endostatin concentrations, and that the optimal endosta- serum levels for human pancreatic tumor inhibition, tin level for tumor growth inhibition is approximately 1 Â 1011,2Â 1010 and 1 Â 1010 AAV-1/cFcES particles 126 ng/ml in SCID mice. were injected i.m. into SCID mice (Figure 6a). Stable and long-term cFcES expression (175 ng/ml) was observed in mice injected with 1 Â 1011 AAV-1/cFcES particles beginning at 7 weeks post-infection (Figure 6a). For 2 Â 1010 and 1 Â 1010 AAV-1/cFcES particles, stable and long-term cFcES expression was observed between 8–10 weeks (Figure 6a), with maximum expressions of 50 and 32 ng/ml respectively. Human pancreatic BxPC-3 tumor cells (5 Â 106) were injected s.c. into the infected mice 45 days after infection to determine the effect on tumor inhibition. Pancreatic tumor growth was inhibited by 31% (P ¼ 0.02) at serum cFcES levels of 175 ng/ml compared to AAV-1/GFP (Figure 6b). Serum levels of 32 ng/ml (P ¼ 0.99) and 50 ng/ml (P ¼ 0.43) were not significantly different than AAV-1/GFP (Figure 6b). Treatment of human pancreatic cancer with 175 ng/ml AAV-1/cFcES reduced vessel density by 73% (P ¼ 0.009) compared to AAV-1/GFP (3 ng/ml), while the remaining doses of 32 (P ¼ 0.99) and 50 ng/ml (P ¼ 0.82) had no effect (Figure 6c). These results combined with prior data suggest that the optimal endostatin serum level for pancreatic cancer inhibition is between 126 and 175 ng/ml, while levels higher or lower did not inhibit growth of pancreatic cancer (Figure 7).

Discussion

We have previously shown that continuous delivery of Figure 4 U-shaped endostatin inhibition of human pancreatic endostatin obtained by an implanted osmotic pump is cancer. (a) Inhibition of human pancreatic tumor growth by AAV-1/ 10-fold more effective in inhibiting pancreatic tumor 6 cFcES. Human pancreatic BxPC-3 tumor cells (5 Â 10 ) were growth than daily bolus injections.7 Therefore, antiangio- implanted s.c. 45 days after AAV-1/cFcES infection. Tumor growth genic gene therapy is a feasible approach to obtain was followed up for 59 days. (b) Pancreatic tumor growth based on sustained systemic therapeutic release of angiogenesis weight versus cFcES serum levels. At the completion of the experiment (day 59), tumors were resected and weighed inhibitors. An added benefit of antiangiogenic gene (1 g ¼ 1000 mm3). Representative pictures of the tumors are shown. therapy is that it may reduce the cost of injecting This experiment was performed three different times. However, recombinant protein on a daily basis. Furthermore, the previous experiments were performed in smaller trials (n ¼ 3–5). use of AAV vectors is advantageous because of the P ¼ 0.03, ANOVA with post-hoc Dunnett’s tests. relative lack of vector-associated toxicity and immunity

Cancer Gene Therapy Endostatin’s U-shaped antitumor activity RM Tjin Tham Sjin et al 624

Figure 5 Determination of vessel density (CD31). (a) Pancreatic tumor sections stained for PECAM. Tumor sections from AAV-1/cFcES- and AAV-1/GFP-infected mice at the completion of the experiment (day 59 after treatment) were stained. Endostatin serum levels are indicated. (b) Inhibition of vessel density. The average of three fields was taken. hpf, high-power field. *P ¼ 0.02, ANOVA with Dunnett’s test.

with sustained long-term transgene expression. Here, we tumor-dependent and is not restricted to the endostatin have shown that an AAV-1 vector can provide stable, range of 126–175 ng/ml. It is highly likely that more systemic and long-term expression of cFcES. SCID mice angiogenic (aggressive) tumors will require a higher are not entirely devoid of an immune response. Therefore, endostatin range for effective treatment, while the reverse we examined the production of antibodies against cFcES is true for less angiogenic tumors. Also, using a colon liver in serum from 15-weeks-infected mice. No antibodies metastasis model, Chen et al.24 showed that blood were detected at any of the infected doses (data not endostatin levels for optimal survival were between 200 shown). and 300 ng/ml, while levels higher or lower had a reduced An optimum dose range of endostatin was required for survival rate.24 Similar U-shaped responses have been effective pancreatic tumor growth inhibition. Canine reported for other antiangiogenic treatments, such as FcEndostatin serum levels of 126–175 ng/ml demon- angiostatin,45 interferon-a46 and rosiglitazone,47 where a strated significant inhibition of pancreatic tumor growth. low dose was more effective than a high dose. Therefore, No inhibition was observed at higher or lower serum it is possible that a U-shaped response curve is not levels, suggesting that a U-shaped curve seems to be uncommon for antiangiogenic therapies. Thus, determi- associated with the antitumor activity of endostatin as a nation of optimum endostatin concentration may be an function of protein concentration (see Figure 7). Similar important factor for long-term tumor treatment. results were observed for recombinant human endostatin Several groups have reported that at high serum using the human pancreatic BxPC-3 and ASPC-1 tumor endostatin levels no or insignificant tumor inhibition models, where the more aggressive ASPC-1 tumor was observed. Employing adenoviral-mediated expression required more endostatin for effective treatment, thereby of endostatin, it was shown that at high endostatin serum shifting the endostatin U-shaped response.44 These results levels of 15–20 mg/ml, no T241 fibrosarcoma and BxPC3 show that the U-shaped endostatin response will be pancreatic tumor inhibition was observed.25 Similarly,

Cancer Gene Therapy Endostatin’s U-shaped antitumor activity RM Tjin Tham Sjin et al 625

Figure 7 Systemic inhibition of human pancreatic cancer by AAV-1/cFcES follows a biphasic dose curve. Endostatin serum levels versus percentage BxPC-3 pancreatic tumor growth of all individual mice are indicated. Adeno-associated virus-1/GFP control of each experiment was considered as 100% (dotted line). Data obtained from Figures 4 and 6 were combined by expressing BxPC-3 growth as a percentage of AAV-1/GFP control within each experiment.

Figure 4). It is possible that receptor desensitization (such as integrins) occurs at sustained high serum endostatin levels, which would impair endostatin activity in vivo. Also, it has been reported that sustained high levels of adenoviral endostatin elicit toxicity and immunogenicity. Wen et al.28 have shown that high-dose intravenous delivery of adenoviral endostatin was associated with severe acute toxicity. However, no adenoviral-mediated endostatin toxicity was observed by other groups. Besides toxicity, potential immunogenicity against endostatin could limit its antitumor activity. Antibodies were detected in the serum and tumor tissues of a patient with multifocal glioblastoma, suggesting that endostatin over- expression could induce a humoral immune response, reducing the effectiveness of viral-mediated endostatin Figure 6 Determination of optimal endostatin serum levels for 49 human pancreatic tumor inhibition. (a) Serum levels of cFcES after gene therapy when given at high doses. AAV-1/cFcES infection. SCID mice were injected with 1 Â 1011, Our data suggest that there is a narrow range for 2 Â 1010 and 1 Â 1010 AAV-1/cFcES particles per mouse. Final a therapeutic dose of endostatin. However, it is highly serum levels were determined by averaging levels of the last five likely that the optimum therapeutic dose will be time points for 1 Â 1011 particles and last three time points for tumor-dependent. Therefore, significant testing will 2 Â 1010 and 1 Â 1010 particles. Adeno-associated virus-1/GFP be required in animal tumor models to determine the (1 Â 1011 particles) was used as control. (b) Inhibition of pancreatic dose-dependent efficacy. Furthermore, our results indi- BxPC-3 tumor growth after AAV-1/cFcES infection. Human pan- 6 cate that this treatment only delayed tumor incidence creatic BxPC-3 cells (5 Â 10 ) were injected s.c. 45 days post- because all animals developed tumors even at the infection and tumor growth was followed for an additional 59 days. optimal doses. This suggests that this treatment would T/C is indicated. *P ¼ 0.02, ANOVA with Dunnett’s tests. (c) Determination of vessel density. Tumor sections from AAV-1/ be more suitable as adjuvant therapy or in combina- cFcES- and AAV-1/GFP-infected mice at the completion of the tion with other antiangiogenic therapies, chemotherapy experiment were stained with PECAM. The average of three fields or radiation. was taken. *P ¼ 0.009, ANOVA with Dunnett’s tests. There is already clinical precedence in designing a clinical trial for an angiogenic inhibitor, INF-a, which has been reported to have a narrow range of antitumor Pawliuk et al.48 reported that systemic endostatin levels of efficacy, that also follows a biphasic U-shaped curve.46 750 ng/ml failed to inhibit T241 fibrosarcoma delivered Therefore, accurate knowledge of the U-shaped para- through retroviral endostatin-transduced HSCs. Here, meters of endostatin’s antitumor activity and in under- we have also shown that no pancreatic tumor inhibition standing that the dosing is not linear will allow for better was observed at high concentrations of endostatin (see and more effective cancer treatments.

Cancer Gene Therapy Endostatin’s U-shaped antitumor activity RM Tjin Tham Sjin et al 626 Acknowledgements 13 Hanai J, Dhanabal M, Karumanchi SA, Albanese C, Waterman M, Chan B et al. Endostatin causes G1 arrest We thank Dr David Resnick for his assistance with of endothelial cells through inhibition of cyclin D1. J Biol cloning of canine endostatin; Dr Arja Kaipainen for Chem 2002; 277: 16464–16469. immunohistochemistry assistance; Drs Ilhan Celik and 14 Hajitou A, Grignet C, Devy L, Berndt S, Blacher S, Oliver Kisker for their assistance with the BxPC-3 tumor Deroanne CF et al. The antitumoral effect of endostatin and angiostatin is associated with a down-regulation of model; Gustave Alberti for his assistance with mice; vascular endothelial growth factor expression in tumor cells. Susan Park for determining endostatin serum levels and Faseb J 2002; 16: 1802–1804. helpful discussions; Kristen Gullage for photography. 15 Km YM, Hwang S, Pyun BJ, Kim TY, Lee ST, Gho YS et al. This work was supported by NIH Grants R01 CA064481, Endostatin blocks vascular endothelial growth factor- HL051818, HL066973 and the Breast Cancer Research mediated signaling via direct interaction with KDR/Flk-1. Foundation. J Biol Chem 2002; 277: 27872–27879. 16 Hanai J, Gloy J, Karumanchi SA, Kale S, Tang J, Hu G et al. Endostatin is a potential inhibitor of Wnt signaling. Conflict of Interest J Cell Biol 2002; 158: 529–539. We declare that there are no conflicts of interest. 17 Kim YM, Jang JW, Lee OH, Yeon J, Choi EY, Kim KW et al. Endostatin inhibits endothelial and tumor cellular invasion by blocking the activation and catalytic activity of matrix metalloproteinase. Cancer Res 2000; 60: 5410–5413. References 18 Nyberg P, Heikkila P, Sorsa T, Luostarinen J, Heljasvaara R, Stenman UH et al. Endostatin inhibits human tongue 1 Folkman J. Angiogenesis in cancer, vascular, rheumatoid carcinoma cell invasion and intravasation and blocks the and other disease. Nat Med 1995; 1: 27–31. activation of matrix metalloprotease-2, -9, and -13. J Biol 2 O’Reilly MS, Holmgren L, Shing Y, Chen C, Rosenthal RA, Chem 2003; 278: 22404–22411. Moses M et al. Angiostatin: a novel angiogenesis inhibitor 19 Lee SJ, Jang JW, Kim YM, Lee HI, Jeon JY, Kwon YG that mediates the suppression of metastases by a Lewis lung et al. Endostatin binds to the catalytic domain of matrix carcinoma. Cell 1994; 79: 315–328. metalloproteinase-2. FEBS Lett 2002; 519: 147–152. 3 O’Reilly MS, Boehm T, Shing Y, Fukai N, Vasios G, 20 Abdollahi A, Hahnfeldt P, Maercker C, Grone HJ, Debus J, Lane WS et al. Endostatin: an endogenous inhibitor of angio- Ansorge W et al. Endostatin’s antiangiogenic signaling genesis and tumor growth. Cell 1997; 88: 277–285. network. Mol Cell 2004; 13: 649–663. 4 Parangi S, O’Reilly M, Christofori G, Holmgren L, Grosfeld 21 Tjin Tham Sjin RM, Satchi-Fainaro R, Birsner AE, J, Folkman J et al. Antiangiogenic therapy of transgenic Ramanujam VMS, Folkman J, Javaherian K. A 27-amino- mice impairs de novo tumor growth. Proc Natl Acad Sci USA acid synthetic peptide corresponding to the NH2-terminal 1996; 93: 2002–2007. zinc-binding domain of endostatin is responsible for its 5 Boehm T, Folkman J, Browder T, O’Reilly MS. Antiangio- antitumor activity. Cancer Res 2005; 65: 3656–3663. genic therapy of experimental cancer does not induce 22 Feldman AL, Restifo NP, Alexander HR, Bartlett DL, Hwu acquired drug resistance. Nature 1997; 390: 404–407. P, Seth P et al. Antiangiogenic gene therapy of cancer 6 Dhanabal M, Ramchandran R, Volk R, Stillman IE, utilizing a recombinant adenovirus to elevate systemic Lombardo M, Iruela-Arispe ML et al. Endostatin: yeast endostatin levels in mice. Cancer Res 2000; 60: 1503–1506. production, mutants, and antitumor effect in renal cell 23 Sauter BV, Martinet O, Zhang WJ, Mandeli J, Woo SL. carcinoma. Cancer Res 1999; 59: 189–197. Adenovirus-mediated gene transfer of endostatin in vivo 7 Kisker O, Becker CM, Prox D, Fannon M, D’Amato R, results in high level of transgene expression and inhibition of Flynn E et al. Continuous administration of endostatin by tumor growth and metastases. Proc Natl Acad Sci USA 2000; intraperitoneally implanted osmotic pump improves the 97: 4802–4807. efficacy and potency of therapy in a mouse xenograft tumor 24 Chen CT, Lin J, Li Q, Phipps SS, Jakubczak JL, Stewart DA model. Cancer Res 2001; 61: 7669–7674. et al. Antiangiogenic gene therapy for cancer via systemic 8 Folkman J, Kalluri R. Tumor angiogenesis. In: Kufe DW, administration of adenoviral vectors expressing secretable Pollock RE, Weichselbaum RR et al. (eds). Cancer Medicine, endostatin. Hum Gene Ther 2000; 11: 1983–1996. 6th edn. Hamilton, Ontario: BC Decker Inc., 2003, 25 Kuo CJ, Farnebo F, Yu EY, Christofferson R, Swearingen pp. 161–194. RA, Carter R et al. Comparative evaluation of the antitumor 9 Sudhakar A, Sugimoto H, Yang C, Lively J, Zeisberg M, activity of antiangiogenic proteins delivered by gene transfer. Kalluri R. Human tumstatin and human endostatin exhibit Proc Natl Acad Sci USA 2001; 98: 4605–4610. distinct antiangiogenic activities mediated by alpha v beta 3 26 Jin X, Bookstein R, Wills K, Avanzini J, Tsai V, LaFace D and alpha 5 beta 1 integrins. Proc Natl Acad Sci USA 2003; et al. Evaluation of endostatin antiangiogenesis gene 100: 4766–4771. therapy in vitro and in vivo. Cancer Gene Ther 2001; 8: 10 Karumanchi SA, Jha V, Ramchandran R, Karihaloo A, 982–989. Tsiokas L, Chan B et al. Cell surface glypicans are 27 Calvo A, Yokoyama Y, Smith LE, Ali I, Shih SC, Feldman low-affinity endostatin receptors. Mol Cell 2001; 7: 811–822. AL et al. Inhibition of the mammary carcinoma angiogenic 11 Shichiri M, Hirata Y. Antiangiogenesis signals by endo- switch in C3(1)/SV40 transgenic mice by a mutated form of statin. Faseb J 2001; 15: 1044–1053. human endostatin. Int J Cancer 2002; 101: 224–234. 12 Wickstrom SA, Alitalo K, Keski-Oja J. Endostatin associ- 28 Wen XY, Bai Y, Stewart AK. Adenovirus-mediated human ates with lipid rafts and induces reorganization of the actin endostatin gene delivery demonstrates strain-specific anti- cytoskeleton via down-regulation of RhoA activity. J Biol tumor activity and acute dose-dependent toxicity in mice. Chem 2003; 278: 37895–37901. Hum Gene Ther 2001; 12: 347–358.

Cancer Gene Therapy Endostatin’s U-shaped antitumor activity RM Tjin Tham Sjin et al 627 29 Schmitz V, Wang L, Barajas M, Gomar C, Prieto J, Qian C. genic properties of the collagen XVIII endostatin domain. Treatment of colorectal and hepatocellular carcinomas by J Biol Chem 2002; 277: 45211–45218. adenoviral mediated gene transfer of endostatin and 40 Glantz SA. Primer of Biostatistics, 6th edn. New York: angiostatin-like molecule in mice. Gut 2004; 53: 561–567. McGraw-Hill, 2005, pp. 73–125. 30 Shi W, Teschendorf C, Muzyczka N, Siemann DW. Adeno- 41 Ding YH, Javaherian K, Lo KM, Chopra R, Boehm T, associated virus-mediated gene transfer of endostatin inhibits Lanciotti J et al. Zinc-dependent dimers observed in crystals angiogenesis and tumor growth in vivo. Cancer Gene Ther of human endostatin. Proc Natl Acad Sci USA 1998; 95: 2002; 9: 513–521. 10443–10448. 31 Shi W, Teschendorf C, Muzyczka N, Siemann DW. Gene 42 Rabinowitz JE, Rolling F, Li C, Conrath H, Xiao W, Xiao X therapy delivery of endostatin enhances the treatment et al. Cross-packaging of a single adeno-associated virus efficacy of radiation. Radiother Oncol 2003; 66: 1–9. (AAV) type 2 vector genome into multiple AAV serotypes 32 Hong SY, Lee MH, Kim KS, Jung HC, Roh JK, Hyung WJ enables transduction with broad specificity. J Virol 2002; 76: et al. Adeno-associated virus mediated endostatin gene 791–801. therapy in combination with topoisomerase inhibitor 43 Chao H, Liu Y, Rabinowitz J, Li C, Samulski RJ, Walsh CE. effectively controls liver tumor in mouse model. World Several log increase in therapeutic transgene delivery by J Gastroenterol 2004; 10: 1191–1197. distinct adeno-associated viral serotype vectors. Mol Ther 33 Ponnazhagan S, Mahendra G, Kumar S, Shaw DR, Stockard 2000; 2: 619–623. CR, Grizzle WE et al. Adeno-associated virus 2-mediated 44 Celik I, Surucu O, Dietz C, Heymach JV, Force J, antiangiogenic cancer gene therapy: long-term efficacy of a Hoschele I et al. Therapeutic efficacy of endostatin vector encoding angiostatin and endostatin over vectors exhibits a biphasic dose–response curve. Cancer Res 2005; encoding a single factor. Cancer Res 2004; 64: 1781–1787. 65: 11044–11050. 34 Noro T, Miyake K, Suzuki-Miyake N, Igarashi T, Uchida E, 45 Lalani AS, Chang B, Lin J, Case SS, Luan B, Wu-Prior WW Misawa T et al. Adeno-associated viral vector-mediated et al. Anti-tumor efficacy of human angiostatin using liver- expression of endostatin inhibits tumor growth and metas- mediated adeno-associated virus gene therapy. Mol Ther tasis in an orthotropic pancreatic cancer model in hamsters. 2004; 9: 56–66. Cancer Res 2004; 64: 7486–7490. 46 Slaton JW, Perrotte P, Inoue K, Dinney CP, Fidler IJ. 35 Bergers G, Javaherian K, Lo KM, Folkman J, Hanahan D. Interferon-alpha-mediated down-regulation of angiogenesis- Effects of angiogenesis inhibitors on multistage carcinogen- related genes and therapy of bladder cancer are dependent on esis in mice. Science 1999; 284: 808–812. optimization of biological dose and schedule. Clin Cancer 36 Lo KM, Sudo Y, Chen J, Li Y, Lan Y, Kong SM et al. High Res 1999; 5: 2726–2734. level expression and secretion of Fc-X fusion proteins in 47 Panigrahy D, Singer S, Shen LQ, Butterfield CE, Freedman mammalian cells. Protein Eng 1998; 11: 495–500. DA, Chen EJ et al. PPARgamma ligands inhibit primary 37 Zolotukhin S, Potter M, Hauswirth WW, Guy J, Muzyczka tumor growth and metastasis by inhibiting angiogenesis. N. A ‘humanized’ green fluorescent protein cDNA adapted J Clin Invest 2002; 110: 923–932. for high-level expression in mammalian cells. J Virol 1996; 48 Pawliuk R, Bachelot T, Zurkiya O, Eriksson A, Cao Y, 70: 4646–4654. Leboulch P. Continuous intravascular secretion of endosta- 38 Xiao X, Li J, McCown TJ, Samulski RJ. Gene transfer by tin in mice from transduced hematopoietic stem cells. adeno-associated virus vectors into the central nervous Mol Ther 2002; 5: 345–351. system. Exp Neurol 1997; 144: 113–124. 49 Ratel D, Nasser V, Dupre I, Benabid AL, Berger F. 39 Javaherian K, Park SY, Pickl WF, LaMontagne KR, Tjin Antibodies to endostatin in a multifocal glioblastoma Tham Sjin R, Gillies S et al. Laminin modulates morpho- patient. Lancet 2000; 356: 1656–1657.

Cancer Gene Therapy