Liposomes Complexed to Plasmids Encoding Angiostatin and Endostatin Inhibit Breast Cancer in Nude Mice¹

Angiogenesis, the growth of new microvessels, is not only essential for a number of physiological processes but also occurs in many pathological conditions including tumor growth. Studies have shown that tumor growth is dependent on angiogenesis (1). This provides the rationale for antiangiogenic therapy in cancer. Several angiogenesis inhibitors that inhibit tumor growth are now in clinical trials. Angiostatin and endostatin are two recently reported potent endogenous angiogenesis inhibitors (2, 3). Both of these inhibitors are proteolytic polypeptides derived from their parent proteins. Angiostatin is a M_r 38,000 protein whose sequence is identical with the first four triple-loop (kringle) structures of plasminogen. Endostatin is a M_r 20,000 COOH-terminal fragment of collagen XVIII. However, neither plasminogen nor collagen XVIII inhibits angiogenesis. Angiostatin and endostatin specifically inhibit endothelial proliferation without a direct effect on the tumor cell or nonneoplastic cell growth. They cause human carcinomas to regress in mice to a microscopic dormant state, with inhibition of virtually all neovascularization (3, 4). Furthermore, there is no observed toxicity with these inhibitors.

Antiangiogenic therapy with angiostatin and endostatin in cancer requires prolonged administration of recombinant protein in vivo. In addition, production of these functional polypeptides has proved difficult, perhaps due to physical properties and variations in the purification procedure by different laboratories. Gene therapy transfer of these polypeptides represents an alternative method to deliver these agents (5, 6, 7). A few groups have shown that antiangiogenic gene therapy with viral vectors is a potentially useful approach to inhibit tumor growth in mouse models (7, 8, 9, 10, 11). Although viral vectors are commonly used as carriers for gene transfer due to their high in vitro transfection efficiency, safety issues and the toxicity of these viral vectors will likely preclude their i.v. use in humans in the near future. In contrast, systemically delivered liposome:DNA complexes have been shown to have low toxicity (12, 13, 14, 15). In this study, we investigated whether intratumoral and i.v. delivery of cationic liposomes complexed to angiostatin and endostatin plasmids inhibited angiogenesis and breast cancer in a nude mouse model.

The angiostatin coding sequence was obtained by amplifying murine plasminogen cDNA with primers 5′-TTAGGGTCGACATGAAGTGGGTAACCTTTCTCCTCCTCCTCTTCATCTCCGGTTCTGCCTTTTCTAGGGGTGTGTATCTGTCAGAATGTAAG-3′ and 5′-GAATGTCGACTCACCAAGGGCCCTTGTCACCATCC-3′. The amplified angiostatin fragment contained a genetically engineered ATG start site, a signal peptide, and a TAA stop codon. The amplified angiostatin cDNA was digested with SalI, purified, and inserted into the SalI site of the PCI vector (Promega, Madison, WI). The nucleotide sequence that encodes the signal peptide is derived from the human monocyte chemoattractant protein 1 (16). To construct the PCI-Endo, primers with SalI restriction sites were used to amplify nucleotides 2869–3399 of murine collagen XVIII. The 5′ primer also contained a signal peptide and an ATG start codon (TTAGGGTCGACATGAAGTGGGTAACCTTTCTCCTCCTCCTCTTCATCTCCGGTTCTGCCTTTTCTAGGGGTCATACTCATCAGGACTTTCAGC). The 3′ primer (TTAGGGTCGACTTATTTGGAGAAAGAGGTCATGAAGCTATTCTCAATGCACAGGACGATGTAGC) contained a TAA stop codon. The amplified endostatin product and the PCI vector were digested with SalI, purified on a 1% agarose gel, and then ligated to one another. The sequence and orientation of PCI-Angio and PCI-Endo were verified.

The CAT coding sequence was inserted in-frame into unique restriction sites of either PCI-Angio (NsiI) or PCI-Endo (ApaI). These restriction sites were placed on the 5′ and 3′ ends of the amplified CAT product, and the 3′ end of the CAT product also contained a stop codon site. The primers 5′-TTTAGGATGCATGGAGGAAAAAATCACTGGATATACC-3′ and 5′-TTATCCATGCATTTACGCCCCGCCCTGCCACTCATCGC-3′ amplified the CAT gene, and this amplified product was inserted into PCI-Angio. Similarly, the primers 5′-TTATCTGGGCCCTTACGCCCCGCCCTGCCACTCACGC-3′ and 5′-TTTAGTGGGCCCTAGAGAAAAAAATCACTGGATATACC-3 amplified the CAT gene, and this PCR product was inserted into PCI-Endo.

Breast cancer cell line MDA-MB-435 and CHO cells were maintained in DMEM containing 10% FCS and 20 mm glutamine.

Transfection of cells was done in 10-cm plates with 60 μl of Superfect (Qiagen, Chatworth, CA) mixed with 10 μg of plasmid as described by the Qiagen protocol. Thirty-six h after transfection, the cells were rinsed with PBS and incubated for 3 h in serum-free DMEM. The cells were then rinsed extensively and incubated in serum-free Iso-Cho medium (Irvine Scientific Inc., Santa Ana, CA) for 24 h. The conditioned media were collected, cleared of cell debris, and concentrated 20-fold with Amicon membranes (Amicon Inc., Beverly, MA) with a M_r 10,000 cut-off. Total protein was determined using a Coomassie Plus protein assay kit (Pierce, Rockford, IL).

The experiments were performed as described in the Promega TNT technical bulletin. In brief, 1 μg of each plasmid was mixed with the reaction buffer, wheat germ extract, T7 polymerase, amino acid mixture, and [³H]leucine (100–200 Ci/mmol) for 90 min at 30°C. The products of translation were denatured and separated by electrophoresis on a 12% polyacrylamide gel. Before gel drying, the gel was fixed for 30 min and then soaked with Amplify (Amersham) for 30 min. The gel was exposed to Kodak X-ray film for 7 days at −70°C.

Total protein (10 μg) from conditioned media was resuspended in loading buffer, boiled for 3–5 min, run on a 12% SDS-PAGE, and blotted to a polyvinylidene difluoride membrane (Hybond-PVDF; Amersham). The membrane was blocked overnight with 5% skim milk in PBS (−Mg, −Ca) containing 0.1% Tween at 4°C and further incubated for 2 h with polyclonal rabbit anti-CAT antibody (5′-3′ Inc., Boulder, CO) at a dilution of 1:750. After incubation with secondary anti-rabbit horseradish peroxidase-conjugated antibody (Amersham; 1:2000 dilution) for 1 h, immunoreactive bands were stained by an enhanced chemiluminescence Western blot analysis system (Amersham).

Endostatin in the culture media was measured with a murine endostatin enzyme immunoassay kit (Accucyte, College Park, MD). Unconcentrated media (100 μl) collected from CHO or MDA-MB-435 cells transfected with PCI or PCI-Endo were used for this assay.

Conditioned media from the CHO cells transfected with PCI-Angio, PCI-Endo, or PCI (control) were concentrated 20-fold. Matrigel (380 μl; Collaborative Biomedical Products, Bedford, MA) with or without 0.1 μg of bFGF (R&D Systems, Minneapolis, MN) were then mixed with 50 μl of concentrated conditioned media. A total of 0.4 ml of this Matrigel was injected into each mouse. There were six treatment groups with three mice in each treatment group, and the experiment was performed twice. One week after the initial injection, the Matrigel plug was removed and bisected. Hemoglobin in one half of the Matrigel plug was measured using the Drabkin method (Drabkin reagent kit 525; Sigma, St. Louis, MO) as described previously (17). The other half of the Matrigel plug was fixed in 4% formaldehyde and stained with H&E to determine blood vessel density. Similarly, we measured the hemoglobin and blood vessel density in Matrigel with media obtained from transfected MDA-MB-435 cells.

Preparation of the liposome:plasmid complexes has been described previously (12, 15). In brief, DH5α bacteria (Life Technologies, Inc., Gaithersburg, MD) containing the plasmids were grown in Superbroth to mid-log phase. The plasmids were then purified with Qiagen columns. An analytical gel of each plasmid (cut and uncut) was done to ensure that there was no contamination with other nucleic acids. Liposomes were composed of 1,2-dioleoyl-3-tri-methylammonium-propane and cholesterol (Avanti, Birmingham, AL) in a 9:1 ratio. After hydration of the lipids, the liposomes were sonicated until clear with a Branson 1210 bath sonicator in the presence of argon. The liposomes were then extruded through 50 nm of polycarbonate membranes with LipsoFast-Basic (Avestin Inc., Ottawa, Canada). Each intratumoral injection of the liposome:DNA complex consisted of 33 nmol of liposome and 2.9 μg of DNA. Each i.v. injection consisted of 200 nmol of liposome and 15 μg of DNA.

After administering anesthetic, female athymic nude mice received injections of 2.5 × 10⁵ MDA-MB-435 tumor cells bilaterally into the mammary fat pads with a stepper (Tridak) and a 27.5-gauge needle. Six days after injection of the tumor cells, the mice received injections of the liposome:plasmid complexes. The time interval between each injection was dependent on whether the mice received intratumoral or i.v. injections. The tumors were measured before each injection and 7 days after the last injection with skin calipers.

Six days after the injection of cells, the mice were randomly divided into four treatment regimens: (a) untreated; (b) empty vector; (c) liposome:PCI-Angio; and (d) liposome:PCI-Endo. Each treatment group contained 10 mice; each mouse received three intratumoral injections 7 days apart. Each injection consisted of liposomes (33 nmol) complexed to 2.9 μg of a plasmid that encoded angiostatin, endostatin, or a control plasmid.

After injection of 15 μg of PCI-Endo complexed to 200 nmol of a cationic liposome into mice, endostatin was measured in the serum 24–48 h later. Endostatin levels were measured in two tumor-bearing mice at each time point, whereas untreated mice were used as controls.

To evaluate the antitumor efficacy of systemically delivered PCI-Endo, 30 mice were inoculated with tumor cells and divided into three groups: (a) untreated; (b) liposome:PCI; and (c) liposome:PCI-Endo treatment groups. The first liposome:plasmid injection was given 6 days after the tumor cell inoculation and every 10 days thereafter for a total of six liposome:plasmid DNA treatments via the tail vein. Each injection consisted of liposomes (200 nmol) complexed to 15 μg of a plasmid that encoded endostatin or a control plasmid.

We initially cloned murine angiostatin and endostatin cDNAs into the PCI vector, generating PCI-Angio and PCI-Endo, respectively. In vitro transcription and translation analysis confirmed the expression of angiostatin or endostatin gene products from PCI-Angio or PCI-Endo (Fig. 1,A). Fusion proteins of CAT with angiostatin or endostatin were then constructed to ensure that the promoter, the start codon, and the signal peptide of the PCI-Angio and PCI-Endo constructs were functionally intact. Initially, we chose to transfect CHO cells with our constructs due to their high transfection efficiency. Western blotting analysis of conditioned media from CHO cells transfected with PCI-Angio-CAT and PCI-Endo-CAT showed clearly detectable levels of Angio-CAT and Endo-CAT proteins, respectively (Fig. 1 B). Cell lysates from transfected CHO cells expressing the CAT protein were used as a positive control. The gel mobility differences between Angio-CAT, Endo-CAT, and CAT corresponded to their predicted molecular weights. This experiment showed that CHO cells that have been transfected with the PCI-Angio-CAT or PCI-Endo-CAT plasmids efficiently secreted the fusion proteins into the culture media. Because construction of the fusion proteins did not disturb the signal peptide sequences of PCI-Angio or PCI-Endo vectors, this experiment suggests that angiostatin and endostatin are also secreted efficiently into the media.

To verify this, conditioned media from MDA-MB-435 or CHO cells transfected with PCI-Endo were analyzed for the presence of endostatin. Endostatin levels were 610 and 724 ng/mg cell protein in MDA-MB-435 (Fig. 1 C) and CHO cells, respectively. In cells transfected with PCI-Angio, we were unable to detect angiostatin directly due to the lack of a specific antimouse angiostatin antibody.

To ensure that angiostatin and endostatin derived from our plasmids are antiangiogenic, we tested their ability to reduce angiogenesis with an in vivo assay. Matrigel impregnated with bFGF was used to induce neovascularization. Measurement of the hemoglobin content of the gel indicated the formation of a functional vasculature at the site of angiogenesis in the presence of bFGF (17). Furthermore, sections of the gel examined with H&E showed endothelial-like cells and neovessels with RBCs in the lumen, indicative of angiogenesis. Addition of the media conditioned by PCI-Angio- or PCI-Endo-transfected CHO cells in the Matrigel/bFGF mixture showed marked inhibition of neovascularization as measured by both hemoglobin content (Fig. 2,A) and examination of the gel for infiltrating vessels (Fig. 2 B). Similarly, media from MDA-MB-435 cells transfected with PCI-Angio or PCI-Endo inhibited angiogenesis in an in vivo Matrigel assay (data not shown). These observations are consistent with the antiangiogenic effect of angiostatin and endostatin described previously (2, 3).

To investigate the effect of PCI-Angio or PCI-Endo on tumor growth, a liposome:DNA complex was injected into pre-established human MDA-MB-435 breast carcinomas grown in athymic mice. The liposome:DNA complex consisted of 33 nmol of liposome and 2.9 μg of DNA. As shown in Fig. 3, tumors from the PCI-Angio and PCI-Endo groups were smaller than those from the untreated group at days 21 and 28 (P < 0.05). The PCI-Endo group was significantly different from the PCI control at day 28 (P < 0.05). After the third injection (day 28), the mice treated with PCI-Angio and PCI-Endo showed a 36% and 49% reduction, respectively, in tumor size compared to the untreated group. These results indicate that local angiostatin and endostatin gene therapy with liposomes as a carrier effectively inhibits tumor growth in vivo. These treatment groups differed from the control groups, despite the finding that the transfection efficiency of an intratumoral injection of a liposome complexed to a CAT marker gene was 5% (data not shown).

Because the intratumoral injection with PCI-Endo appeared to be more effective than PCI-Angio, we investigated whether the i.v. administration of PCI-Endo would inhibit tumor growth. To ensure that endostatin was expressed in vivo, we measured serum levels 1 and 2 days after i.v. injections. The endostatin levels at 24 and 48 h after injection were 10.8 and 33 ng/ml, respectively. The ability to measure serum endostatin levels in vivo quickly without sacrificing the animal should facilitate the development of nonviral carriers and plasmids with increased expression. We then determined whether endostatin administered i.v. decreased tumor growth (Fig. 4). After the third injection, there was a significant reduction in tumor size of mice treated with plasmids encoding endostatin compared to untreated controls (P < 0.05). By the sixth injection, there was a 40% reduction in the liposome:PCI-Endo-treated group compared to either the liposome:PCI-treated or the untreated groups (P < 0.05).

To achieve further tumor reduction, further advancements in nonviral delivery systems will need to be accomplished. Our report demonstrates that the liposome:endostatin plasmid complexes inhibit the tumor when given i.v., although several laboratories have shown that the vast majority of the liposome:plasmid complexes target the lung (18). As a result, tumors and their vessels located outside the lung are transfected inefficiently by liposome:plasmid complexes (15). Increased specificity of these liposome:DNA complexes may allow more endostatin to be expressed in the peritumoral area and limit the effect of endostatin on normal physiological angiogenic processes. Coupling peptides, which target the tumor endothelial cells, to the liposomes and changing the liposome composition may further increase the specificity of this therapy (19).

In summary, gene therapy is one attractive option to deliver the angiogenic polypeptide inhibitors angiostatin and endostatin. We have demonstrated that cationic liposomes complexed to plasmids that encoded angiostatin and endostatin inhibited the growth of human breast cancer implanted in the mouse model with either intratumoral or i.v. administration. This report should provide a basis for further development of nonviral delivery of antiangiogenic genes.

We thank Drs. Antonino Passaniti and Lynne Abruzzo for careful reading and useful comments concerning the manuscript.