Endostatin: Yeast Production, Mutants, and Antitumor Effect in Renal Cell Carcinoma¹

Thirty thousand cases of RCC³ were diagnosed in the United States in 1996 (1). The prognosis for metastatic RCC remains highly unfavorable. Despite advances in radiation therapy and chemotherapy, the long-term survival of treated patients has shown only marginal improvement over the past few decades (1). Because RCC is a highly vascularized tumor, VEGF is likely to play an important role in the formation of tumor-associated angiogenesis. Moreover, the VHL gene, lost in >99% of sporadic RCC cases along with alterations of the remaining allele in ∼70% of cases (2), represses VEGF (3, 4, 5, 6). In a nude mouse model, introduction of a wild-type VHL gene into 786-O cells, a RCC tumor cell line, inhibited tumor growth (7) and angiogenesis. The lack of significant treatment options available for RCC emphasizes the need to focus on the development of novel therapeutic strategies. In this regard, targeting tumor vasculature of solid tumors has recently shown promising results in several animal model systems (8, 9, 10, 11, 12).

The growth of solid tumors beyond a few mm³ depends on the formation of new blood vessels (13). Numerous studies have shown that both primary tumor and metastatic growth are angiogenesis dependent (13, 14, 15, 16). A number of angiogenesis inhibitors have been identified. Certain ones, such as platelet factor-4 (17, 18), IFN-α, IFN-inducible protein-10. and PEX (19, 20, 21), are not “associated with tumors,” whereas two others, angiostatin and endostatin, are “tumor associated” (22, 23). Angiostatin, a potent endogenous inhibitor of angiogenesis generated by tumor-infiltrating macrophages that up-regulate matrix metalloelastase (24), inhibits the growth of a wide variety of primary and metastatic tumors (25, 26, 27, 28, 29).

Recently, O’Reilly, et al. (23) have isolated endostatin, an angiogenesis inhibitor from a murine hemangioendothelioma cell line (EOMA). Circulating levels of a fragment of human endostatin have been detected in patients with chronic renal insufficiency with no detectable tumor (30). The NH₂-terminal sequence of endostatin corresponds to the COOH-terminal portion of collagen XVIII. Endostatin is a specific inhibitor of endothelial proliferation and angiogenesis. Systemic administration of non-refolded precipitated protein expressed in Escherichia coli caused growth regression of Lewis lung carcinoma, T241 fibrosarcoma, B16 melanoma, and EOMA (23) cells in a xenograft model. Moreover, no drug resistance was noted in three of the tumor types studied. Surprisingly, repeated cycles of administration with endostatin resulted in tumor dormancy (31).

The results from this study open new avenues for treatment of cancer and provide a promising route for overcoming drug resistance often seen during chemotherapy. However, in all of these investigations, a non-refolded precipitated form of endostatin was administered in the form of a suspension to tumor-bearing animals. In addition, large amounts of protein were required to cause tumor regression and to lead to tumor dormancy. As pointed out by Kerbel (32), an oral drug equivalent of endostatin is needed. Clearly, mechanistic investigations could be undertaken if recombinant protein were available in soluble form. Moreover, initial testing could be done in vitro with soluble protein before studying its efficacy under in vivo conditions. In the present study, we have obtained the expression of mouse endostatin in the Pichia pastoris system. The yeast expression system was selected because of its ability to express heterologous protein in a large scale and to process posttranslational modifications (33, 34). Our studies show for the first time that it is possible to express biologically active mouse endostatin with a yield of 15–20 mg/l of yeast culture. Biological activity was demonstrated in vitro by effects on endothelial proliferation and migration (the latter not described previously for endostatin) and in the CAM assay, along with growth inhibition in a RCC tumor xenograft model. For the first time, mutants of the endostatin protein (EM 1 and EM 2) were created, and one mutant (EM 2) showed loss of function in a RCC model.

786-O, a renal clear cell carcinoma line; C-PAE, a bovine pulmonary arterial endothelial cell line; and ECV304, a human endothelial cell line, were all obtained from American Type Culture Collection. The cell lines were maintained in DMEM (786-O and C-PAE) and M199 (ECV304) supplemented with 10% FCS, 100 units/ml of penicillin, 100 μg/ml of streptomycin, and 2 mm l-glutamine. The cDNA clone for mouse endostatin pBACPak 8 was kindly provided by B. R. Olsen (Department of Cellular Biology, Harvard Medical School, Boston, MA). The prokaryotic expression vector pET17b was purchased from Novagen (Madison, WI). The yeast expression system, P. pastoris (pPICZαA). was purchased from Invitrogen (San Diego, CA). Restriction enzymes and Vent DNA polymerase were purchased from New England Biolabs (Beverly, MA).

The sequence encoding the COOH-terminal portion of mouse collagen XVIII was amplified by PCR using Vent DNA polymerase, and the endostatin pBACPak 8 vector was used as a template. The primers used were 5′-GGC ATA TGC ATA CTC ATC AGG ACT TT-3′ and 5′-AAC TCG AGA TTT GGA GAA AGA GGT-3′.

PCR was carried out for 30 cycles with the following parameters: 94°C denaturation, 60°C for annealing, and 72°C extension, each for 1 min. The amplified DNA fragment (555 bp) was purified using a QIAquick PCR purification kit, digested with NdeI and XhoI, and ligated into the expression vector pET17bhis (35). Initial transformation was carried out with the HMS 174 host strain. Positive clones were sequenced on both strands. The desired clones were finally transformed into BL21(DE3) for expression. The expression of recombinant protein in the pET system was carried out as described by the manufacturer.

Primers were designed such that 9 and 17 amino acids were deleted from the COOH terminus of endostatin for EM 1 and EM 2, respectively. The amplified DNA fragments (528 bp for EM 1 and 504 bp for EM 2) were purified, digested with NdeI and NotI, and ligated into a predigested pET28(a) expression vector. The rest of the protocol was carried out as described above. Induction conditions and processing of the bacterial pellet were as described elsewhere (23). The purification of recombinant protein was performed using a Ni-NTA column in the presence of 8 m urea as described in the QIAexpressionist manual. Briefly, the bacterial pellet was solubilized in “equilibration buffer” (8 m urea, 10 mm Tris, and 100 mm sodium phosphate buffer, pH 8.0) for 1 h at room temperature. The suspension was sonicated three to four times and centrifuged at 10,000 × g, and the soluble fraction was loaded on a Ni-NTA column preequilibrated with the above buffer at a flow rate of 10–20 ml/h. The column was washed extensively with equilibration buffer. Bound proteins were eluted by lowering the pH of the buffer (from 8.0 to 6.3, 4.2, and 3.0). For the in vivo experiments using endostatin mutants, nonspecific proteins binding to the column were removed by an equilibration buffer wash, followed by 10 mm and 25 mm imidazole washes. Bound proteins were eluted in equilibration buffer containing 0.2 m acetic acid. The purified fractions were analyzed by SDS-PAGE, and the fractions containing purified endostatin (pH 4.2 and 3.0 for wild-type endostatin and equilibration buffer containing 0.2 m acetic acid for endostatin mutants) were pooled and refolded slowly. The final dialysis was carried out against PBS (pH 7.4) at 4°C. During dialysis, the protein precipitated out of solution. It was further concentrated and stored at −70°C in small aliquots. The concentration of protein was determined by the BCA assay (Pierce).

The sequence encoding mouse endostatin was further modified by PCR using Vent DNA polymerase. The amplified fragment containing EcoRI and NotI restriction sites was subcloned into a predigested yeast expression vector. The pPICZαA vector carries an α factor secretion signal sequence with a Zeocin marker for antibiotic selection. Initial transformation was done in the Top 10′ host strain. The resultant clones were screened for insert, and positive clones were sequenced. The plasmid was then linearized with SacI and used for homologous recombination into the yeast host strain GS115. The transformation was carried out by the lithium chloride method as described in the Pichia expression manual. Recombinants were selected by plating on YPD plates containing 100 μg/ml of Zeocin. Clones that grew on the YPD/Zeocin plate were tested for expression.

The expression of mouse endostatin in large scale was carried out in 2-liter baffled shaker flasks. The overnight culture (A₆₀₀, 2–6) was used to inoculate 2-liter flasks, with the addition of 500 ml of buffered glycerol medium. Cells were grown at 250 rpm at 30°C until A₆₀₀, 16–20 (2 days). Subsequently, cells were centrifuged at 5000 rpm for 10 min, and the yeast were resuspended in 300–400 ml of buffered methanol induction medium. The supernatant containing the secreted recombinant protein was harvested on the second, third, and fourth day after induction. After the final harvest, the cell-free supernatant was processed immediately.

The crude supernatant containing recombinant protein was concentrated by ammonium sulfate precipitation (70%). The precipitated protein was dissolved in 10 mm Tris buffer (pH 7.4) containing 150 mm NaCl and dialyzed overnight at 4°C (three changes at 6–8-h intervals). The dialyzed sample was further concentrated by ultrafiltration using an Amicon concentrator (YM 10). A disposable Polyprep column (Bio-Rad) was packed with heparin-agarose resin and equilibrated with 10 mm Tris, 150 mm NaCl, pH 7.4. The concentrated sample was loaded on the column at a flow rate of 20 ml/h using a peristaltic pump. The column was washed with equilibration buffer until the A₂₈₀ was <0.001. Bound proteins were eluted by a step-wise gradient of NaCl (0.3, 0.6, 1, and 2 m NaCl). The peak fractions from 0.6 to 1 m were pooled and dialyzed against PBS (pH 7.4). Protein concentration was measured by the BCA assay (Pierce). The purification process was performed in the cold room (4°C). Recombinant soluble endostatin expressed from the Pichia system was used in all of the in vitro assays.

The coding region of the mouse endostatin construct in the pET expression vector is preceded by a His.Tag (10 histidine residues). By DNA-PCR, the coding region including the His.Tag sequence was shuttled into pPICZαA vector. Linearization and recombination into the yeast host strain GS115 were done as described before. The cell-free medium was precipitated with ammonium sulfate (70% saturation). Precipitated proteins were dissolved in 50 mm sodium phosphate buffer (pH 8.0) containing 300 mm NaCl and dialyzed in the same buffer at 4°C after three changes at 6–8-h intervals. A Ni-NTA column was used for purification of the His.endostatin recombinant protein, as described in the QIAexpressionist manual. Bound proteins were eluted with a step-wise gradient of imidazole (10, 25, 50, and 100 mm). The peak fractions from 50 and 100 mm imidazole elutions were pooled and dialyzed against PBS buffer (pH 7.4).

The purified protein from the yeast expression system was further characterized by NH₂-terminal microsequencing for seven cycles. In addition, a polyclonal antiserum to mouse recombinant endostatin was raised by immunizing a rabbit with 10 μg of purified protein derived from the Pichia expression system. Recombinant endostatins expressed from bacteria and yeast systems were separated on 12% SDS-PAGE. The proteins were transferred to polyvinylidene difluoride membrane by semidry transfer (Trans-blot; Bio-Rad). The primary antiserum was diluted to 1:4000 in 1× TBS buffer containing 5% nonfat dry milk. Goat anti-rabbit IgG/horseradish peroxidase conjugate was used as a secondary antibody (1:5000). Immunoreactivity was detected by chemiluminescence (Pierce).

The antiproliferative effect of endostatin produced in the yeast system was tested using bovine pulmonary artery endothelial cells (C-PAE). The cells were plated in 24-well fibronectin (10 μg/ml)-coated plates at 12,500 cells/well in 0.5 ml of DMEM containing 2% FBS. After a 24-h incubation at 37°C, the medium was replaced with fresh DMEM and 2% FBS containing 3 ng/ml of bFGF (R & D systems) with or without recombinant mouse endostatin. The cells were pulsed with 1 μCi of [³H]thymidine for 24 h. Medium was aspirated, cells were washed three times with PBS, and then solubilized by addition of 1.5 n NaOH (100 μl/well) and incubated at 37°C for 30 min. Cell-associated radioactivity was determined with a liquid scintillation counter.

To determine the ability of recombinant endostatin to block migration of ECV304 cells toward bFGF, a migration assay was performed using 12-well Boyden chemotaxis chambers (Neuro Probe, Inc.) with a polycarbonate membrane (25 × 80-mm, PVD free, 8-μm pores; Poretics Corp., Livermore, CA). The nonspecific binding of growth factor to the chambers was prevented by coating the chambers with a solution containing 0.5% gelatin, 1 mm CaCl₂, and 150 mm NaCl at 37°C overnight. ECV304 cells were grown in 10% FBS containing 5 ng/ml 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate (DiIC18; Molecular Probes, Eugene, OR) overnight and washed with PBS containing 0.5% BSA. After trypsinization, the cells were counted using Coulter Counter Z1 (Luton, United Kingdom) and diluted to 300,000 cells/ml in Medium 199 containing 0.5% FBS. The lower chamber was filled with Medium 199 containing 25 ng/ml bFGF. The upper chamber was seeded with 15,000 cells/well with different concentrations of recombinant endostatin. Cells were allowed to migrate for 4 h at 37°C. At that time, the cells on the upper surface of the membrane were removed with a cell scraper, and the (migrated) cells on the lower surface were fixed in 3% formaldehyde and washed in PBS. Images of the fixed membrane were obtained using fluorescence microscopy at 550 nm with a digital camera, and the number of cells on each membrane was determined using the OPTIMAS (version 6.0) software.

The ability of mouse endostatin to block bFGF-induced angiogenesis in vivo was tested using the CAM assay. Fertilized white Leghorn chicken eggs (SPAFAS, Inc., Norwich, CT) were opened on 100-mm² Petri dishes and allowed to grow until day 11 in a humidified incubator at 38°C. Pellets containing Vitrogen (Collagen Biomaterials, Palo Alto, CA) at a concentration of 0.73 mg/ml and supplemented with: (a) vehicle alone; (b) VEGF (250 ng/pellet); (c) VEGF and endostatin (20 to 0.5 μg/pellet); (d) bFGF (50 ng/pellet); and (e) bFGF and endostatin (20 to 0.5 μg/pellet) were allowed to polymerize at 37°C for 2 h. The pellets were placed on a nylon mesh and oriented on the periphery of the CAM. Embryos were returned to the incubator for 24 h. Invasion of new capillaries on the collagen mesh was assessed by injection of FITC-dextran into the circulation of the chicken embryo. At the end of the experiment, the meshes were dissected, and evaluation of vascular density was done using the program NIH Image 1.59, as described previously (36). Assays were performed in triplicate, and four independent experiments were conducted.

The specificity of the inhibitory effect of endostatin was demonstrated by neutralization studies using endothelial proliferation and CAM assays. Briefly, in the endothelial proliferation assay, the endostatin was preincubated with polyclonal antiserum or purified antibody (IgG) and then added to the C-PAE cells. Preimmune serum was used as negative control. In addition, purified IgG and endostatin antibody alone were also used as a control. The cells were then pulsed with [³H]thymidine for 24 h, and cell-associated radioactivity was measured as described before. For the CAM assay, endostatin (10 μg) and antiserum (50 μg) were preincubated overnight end-over-end at 4°C prior to preparation of the pellets. Controls for these experiments included IgG alone and preimmune serum alone. Evaluation of the angiogenic response was determined as indicated above.

Male nude mice, 6–8 weeks of age, received injections s.c. in the right flank with 2 million 786-O cells in a 100-μl volume. Tumors appeared ∼2 weeks after implantation. Tumor size was measured using calipers, and tumor volume was calculated using a standard formula (22). The tumor volume ranged from 350 to 400 mm³. The animals were randomized, and each group had five mice with comparable tumor size within and among the groups. Treatment was started with recombinant endostatin (bacterial or yeast versions), with each mouse receiving 10 mg/kg body weight of recombinant protein daily, administered for a period of 10 days via i.p. injection. Control animals received PBS each day. Tumor size in all groups was measured on alternate days, and tumor volume was calculated. The treatment was terminated on day 10, and animals were sacrificed; tumors from each mouse were removed and fixed in 10% buffered formalin.

For the mutant study, each mouse received 20 mg/kg body weight of the protein daily for 2 weeks i.p. The initial tumor volume was 150–200 mm³. Wild-type endostatin, also produced in the pET28(a) vector, was given at 20 mg/kg body weight for the experiment as a positive control, and PBS was given as a negative control.

The gene encoding mouse endostatin was amplified from the pBACPak8 plasmid and expressed initially in the pET expression system. A Ni-NTA agarose column was used to purify the recombinant protein (Fig. 1,A). Protein present in inclusion bodies was solubilized in 8 m urea and purified under denaturing conditions as described by O’Reilly et al. (23). SDS-PAGE analysis showed a discrete band at M_r 22,000–24,000 under nonreducing conditions (Fig. 1 B). In addition, higher molecular complexes were also observed, which upon reduction resulted in a discrete band at M_r 22,000–24,000. The peaks at different pH elutions (pH 4.2 and 3.0) were pooled and dialyzed against decreasing concentrations of urea, and final dialysis was performed in PBS buffer (pH 7.4), at which time most of the proteins precipitated out of solution. Because non-refolded precipitated protein expressed from a similar system had shown biological activity in vivo, we followed the exact procedure used for “protein refolding” [as described by O’Reilly et al. (23)]. The precipitated protein was used in suspension form for in vivo experiments only, with the concentration of protein measured by the BCA method (solubilized in urea with a suitable blank) and stored at −70°C in small aliquots. Because mouse and human endostatin are strikingly conserved at the COOH terminus, we made two small deletions, reasoning that larger deletions may destroy activity. EM 1, a M_r 19,000 protein, was generated with an 9-amino acid deletion from the COOH terminus, leaving all of the four cysteine residues intact. EM 2 was an additional 8-amino acid deletion that omitted the most COOH-terminal cysteine.

P. pastoris, a methanotropic yeast strain, has many advantages of a higher eukaryotic expression system: (a) the presence of α factor signal sequence facilitates secretion of the expressed protein into the medium; (b) the yeast strain (GS115) secretes only very low levels of endogenous host protein, which further simplifies the purification process; (c) endotoxin contamination is not an issue; and (d) glycosylation can occur. The pPICZαA vector was selected for expression of mouse endostatin. Initial screening was used to identify yeast clones with high levels of expression. Endostatin was expressed as a soluble protein (M_r 20,000), with a peak level of expression noted on the second day after induction.

A heparin-agarose column was used for purification, based on data of O’Reilly et al. (23). Fig. 2 shows the elution profile and SDS-PAGE analysis of purified protein. Two distinct peaks were obtained with increasing concentration of NaCl (Fig. 2,A). The first peak at 0.3 m NaCl was small when compared with the major peak at 0.6 m NaCl. Most of the endostatin protein bound to the column as shown by the lack of the protein in the flow-through fraction (Fig. 2 B). The recombinant protein bound tightly, and washing with the low-salt Tris buffer removed other yeast-derived proteins. Protein eluted from the 0.3 m NaCl fraction had a trace amount of endostatin but was contaminated with other host-derived, high molecular weight protein. The purified protein migrated at M_r 20,000, which upon reduction migrated at M_r 22,000. The protein fractions eluted at 0.6 m and 1 m NaCl were pooled, concentrated, and dialyzed against PBS (pH 7.4). The purified protein was further separated by FPLC using a Superose 12 size separation column. The elution profile from this column showed a single peak (data not shown). SDS-PAGE analysis showed the presence of single discrete band corresponding to endostatin (data not shown). The level of expression was estimated to be in the range of 15–20 mg/l culture.

To further characterize the recombinant protein, NH₂-terminal microsequencing was carried out at the Harvard microsequencing facility. It showed that the yeast α factor signal peptide was processed and cleaved at alanine. The first seven residues (EFHTHQD) of the purified protein after signal peptide cleavage matched exactly the published sequence of endostatin protein (data not shown), with the first two residues (EF) derived from linker sequence.

The elution profile of His.endostatin from the Ni-NTA column showed that the recombinant protein bound tightly (Fig. 3). The yeast-derived host proteins in the culture supernatant did not bind to the column and were removed during the wash. Bound proteins were eluted by a stepwise gradient of imidazole (Fig. 3,A). The nonspecifically bound, host-derived proteins eluted with the addition of 10 mm imidazole. At 25 mm imidazole, a small fraction of the recombinant protein was eluted, along with proteins of higher molecular weight. Final elution with 50 and 100 mm imidazole showed a distinct peak. SDS-PAGE analysis (Fig. 3 B) of the eluted proteins showed that the flow-through fraction did not contain any endostatin, indicating that most of the protein bound to the column. Increasing the concentration of imidazole to 10 and 25 mm resulted in the elution of nonspecific protein. A protein with M_r 22,000 was seen at 100 mm, along with a smaller amount of protein corresponding to M_r 44,000–46,000. The concentration of purified protein was determined by the BCA method. The level of expression was estimated at 15 mg/l culture.

A polyclonal antibody was raised against purified mouse endostatin derived from the yeast expression system. The purified endostatin expressed from the bacterial and yeast expression systems were run under reducing and nonreducing conditions. Fig. 4 shows immunoreactive bands corresponding to endostatin. The size of the protein estimated from the Western blot ranges from M_r 22,000–24,000. In addition, the recombinant His.endostatin from yeast and bacteria was probed with a Penta His.monoclonal antibody (Qiagen, Santa Clarita, CA). The monoclonal antibody showed positive response only with the His.endostatin, whereas native endostatin did not show any immunoreactivity (data not shown). This data confirmed the presence of the His.Tag in the recombinant protein. The antiserum did not show any cross-reactivity to human or mouse angiostatin, demonstrating some degree of immunoreactivity specific to endostatin (data not shown). Immunoreactivity of the polyclonal antibody was also observed with EM 1 and EM 2 proteins (data not shown).

The antiproliferative effect of endostatin produced in the yeast system was tested using C-PAE cells. We initially experimented with different endothelial cell types and tested various parameters [time of “starvation,” serum concentration, concentration and type of mitogenic stimulus (VEGF versus bFGF)]. C-PAE cells gave the most reproducible response. A dose-dependent inhibition of bFGF-induced proliferation was seen (Fig. 5,A). The inhibition range (30–94% of control) was seen with increasing concentrations of endostatin (0.1–10 μg/ml), with an ED₅₀ in the range of 600–700 ng/ml. A similar inhibitory effect on C-PAE cells was seen when His.endostatin from yeast was tested in the above assay (Fig. 5,A). The recombinant protein did not inhibit the proliferation of the RCC cells (786-O and A498) at concentrations ranging from 0.5 to 10 μg/ml (Fig. 5 B), nor did it have an effect on IMR90 and NIH3T3 fibroblasts (data not shown).

Because C-PAE cells do not migrate in response to bFGF and VEGF, ECV304 cells were used with different concentrations of endostatin using bFGF as a stimulus. Addition of endostatin resulted in a dose-dependent inhibition of migration (Fig. 6). At a concentration <1 μg/ml, marginal inhibition of migration was noted, whereas at 10 μg/ml, 60% inhibition of endothelial cell migration was observed. These studies are the first to show the effect of endostatin on cell migration. The action of endostatin on migration of two non-endothelial cell lines was also assessed. No effect was seen on inner medullary collecting duct renal cells, and some effect (15% at 5 μg/ml and 50% at 20 μg/ml) was noted in the IC-21 macrophage precursor cell line (data not shown), suggesting that at high concentration, endostatin may block cell migration in some cell types.

Endostatin was able to suppress the angiogenic response mediated by both bFGF and VEGF (Fig. 7). The inhibition was dose dependent. Blocking of the VEGF response was somewhat more effective (47%) than the suppression of the bFGF response (39%), both at 20 μg/mesh.

The ability of our polyclonal antiserum to neutralize the biological activity of endostatin was tested in both endothelial proliferation and CAM assays. Fig. 8 demonstrates that the inhibitory effect of endostatin can be suppressed by incubation with specific antiserum. Anti-endostatin antiserum blocked the suppressive effect by 95% (data not shown). The preimmune serum and endostatin antibody alone did not have stimulatory effect nor did normal rabbit IgG.

Recombinant endostatin was administered daily at 10 mg/kg/day when the tumor size was ∼350–400 mm³. On the fifth day after treatment, there was a difference between control (963 mm³) and treated (Endo yeast, 405 mm³; endo bacteria, 442 mm³; and His.endo, 462 mm³) groups. A 2.5-fold decrease in tumor volume was observed on the fifth day after treatment between control and treated groups (Fig. 9, A and B). The growth of the tumor was suppressed in all of the treatment groups: a slower growth rate was seen compared with the control group. Bacterial (His.Tag) or yeast-derived (with or without His.Tag) endostatin at a dose of 10 mg/kg all worked equally well. On the tenth day after treatment, the tumor volume in the control animals was 1490 mm³, whereas in the treated group, it was in the range of 480–570 mm³ (P < 0.005). Endostatin administration did not inhibit tumor growth completely: the growth of the tumors slowed, with a marginal increase in volume during the treatment period.

A second set of experiments with endostatin and mutants EM 1 and EM 2 at a dosage of 20 mg/kg body weight were conducted in a RCC model, as a first step in exploring structure-function relationships. Nine days after treatment, the difference between groups was apparent (Fig. 9 C). On the eleventh day after treatment, the tumor volume in the control group (397 mm³) was approximately twice that of the two treated groups: endostatin (182 mm³) or EM 1 (259 mm³). However, on the same day, the tumor volume of the EM 2-treated group (389 mm³) was similar to that of the control group (397 mm³). Significance was at the 90% confidence level between the EM 2 and endostatin groups and 95% confidence level between endostatin and control groups. Dropping the value of the largest and smallest tumors on day 11 in each group increased the confidence level to 95% between EM 2 and EM 1 and between EM 2 and endostatin. Therefore, EM 1 protein retained the native biological activity of endostatin, whereas EM 2 with an additional 8 amino acids deleted did not. Also, of note, two of the five mice in the endostatin group and one of five in the EM 1 group had no detectable tumor at the end of the treatment period.

In the present study, we have shown that biologically active mouse endostatin can be expressed at high levels in the P. pastoris yeast expression system. This system has all of the advantages of an eukaryotic expression system and generally gives higher expression levels (34, 37). It has the added advantage of inducible expression, leading to 10–100-fold higher heterologous protein expression levels compared with other eukaryotic systems (baculovirus or mammalian) with values as high as 1–10 g/l culture when optimized in a fermenter (38). Moreover, endostatin expressed from the Pichia system was secreted into the medium with a molecular weight of M_r 20,000, in agreement with the size obtained from other sources (23), and was biologically active in vitro and in vivo.

The recombinant protein bound to a heparin-agarose column and was eluted at 0.6–1 m NaCl concentration. These data suggested that the yeast expressed protein was folded properly, because the ED₅₀ was comparable with that obtained with baculovirus-expressed protein (ED₅₀ ∼600–700 ng/ml; Ref. 23). Also, endostatin at high doses (>100 μg/ml) did not inhibit the growth of the 786-O cell line. Microsequence analysis confirmed the processing of the signal peptide, and the NH₂-terminal amino acids matched with published sequence. In the CAM assay, endostatin at 20 μg/disc inhibited angiogenesis induced by bFGF. Similar findings have also been reported as data not shown using baculovirus-expressed endostatin (23). The effect of endostatin on the CAM assay showed more potency on a molar basis than the inhibitors thrombospondin-1 (36), fumagillin (AGM 1470), or antibodies against the integrin α_vβ₃.⁴ Our studies on endothelial cell migration provide an additional mechanism of action for endostatin and a new in vitro assay for its efficacy.

In vivo, both E. coli-derived protein and yeast-produced soluble protein gave comparable inhibitory profiles. Moreover, the presence of a His.Tag sequence in the yeast-derived protein did not affect biological activity. O’Reilly et al. (23) reported that endostatin at 10 mg/kg inhibited tumor growth by 97%. In our first study, we failed to see such a dramatic response. Although there was significant difference in tumor volume between control and treated mice, the tumors in the treated group continued to grow slowly. On the tenth day after treatment, when administration of endostatin was stopped, the tumor grew rapidly, and within a week, the average size of the tumors was comparable with controls (data not shown). Subsequently, we tested the effects of endostatin (and its mutants) on tumors ranging from 150 to 200 mm³ and also increased the endostatin dosage to 20 mg/kg body weight. Three of 10 mice in the combined endostatin and EM 1 groups showed regression. Possible explanations for our differences with O’Reilly et al. (23) include: (a) protein was given i.p. (our data) versus s.c. (23); and (b) the response of this tumor (RCC) to antiangiogenic treatment may be different from that of other tumor types (23) in that RCC may secrete either higher levels or different types of angiogenic proteins. To obtain greater efficacy, it may be possible in the future to tailor the treatment option (e.g., type of antiangiogenic therapy), depending upon the angiogenic stimulator secreted by the tumor. Additionally, other modalities, such as chemotherapy and radiation (not efficacious as single agents in RCC), or biologicals (IL-2, IFN) could be combined with endostatin to provide synergism, as has been demonstrated recently for angiostatin (39).

Our data with neutralizing antibody and the two mutant endostatin proteins (with markedly differing efficacy) provide further evidence that endostatin and not a possible contaminant is the active molecule giving an antiangiogenic effect. Moreover, the mutant protein data point to the importance of eight residues surrounding and including the last cysteine as critical for endostatin activity. Given our present lack of knowledge of the protein conformation of the E. coli-generated proteins used in this study, it is not possible to conclude whether disulfide bonding (between cysteines 1 and 4; Ref. 40) is critical. Additional studies are in progress to address this issue.

At present, the origin of endostatin is not known. Collagen XVIII reported as a member of a family of collagen-like proteins is localized mainly in the perivascular portion of blood vessels (41, 42). Collagen XVIII by itself is not inhibitory to endothelial cells but when processed by as yet unknown mechanisms may lead to the release of the COOH-terminal portion (23). The proteases involved in the generation of endostatin is not clear, and how these may be regulated is also not known. The mechanism of endostatin-mediated tumor regression is unknown. Is there a receptor for endostatin on endothelial cells, and how does binding of endostatin to a putative endothelial cell-specific receptor initiate a cascade of events resulting in the inhibition of endothelial cell proliferation and migration? It is conceivable that endostatin may compete with binding of angiogenic stimulators such as bFGF and VEGF to its appropriate receptors? Alternatively, it is possible that proliferating endothelial cells up-regulate α_vβ₃, an endothelial integrin, and endostatin may act by disrupting the interaction of proliferating endothelial cells to matrix protein, thus driving endothelial cells to undergo apoptosis. It has been well documented that lack of attachment of endothelial cells to matrix protein may result in programmed cell death (43). Such mechanistic issues can now be addressed with our biologically active soluble version of yeast endostatin.