Soluble Recombinant Endostatin Purified from Escherichia coli: Antiangiogenic Activity and Antitumor Effect¹

Generally, tumor growth is critically dependent on blood supply(1). Folkman showed that suppression of tumor angiogenesis leads to tumor starvation and tumor regression. Therefore,the tumor vascular system has become an important target for cancer therapy. In contrast to malignant cells that are able to become resistant to conventional chemotherapy or radiotherapy, a body of data indicates that endothelial cells do not develop resistance to antiangiogenic agents (1). An increasing number of antiangiogenic agents have been discovered. Among these, endostatin was identified by O’Reilly et al. (2).

Endostatin is a M_r 20,000 protein that specifically and strongly inhibits tumor angiogenesis. It was originally isolated from the supernatant of a cultured murine hemangioendothelioma cell line and represents a COOH-terminal fragment of collagen XVIII. The endostatin gene has been cloned and expressed as a recombinant protein in Escherichia coli and yeast expression systems (2, 3). Animal studies demonstrated that recombinant endostatin strongly inhibits the growth of a variety of murine and xenotransplanted human tumors (4, 5).

Purification of bioactive recombinant endostatin from E. coli, however, has proved challenging because the endostatin is purified under denaturing condition (e.g., urea) but invariably becomes insoluble under physiological conditions. The most likely biochemical explanation for this precipitation phenomenon is that endostatin becomes misfolded during the process of renaturation. It is believed that injection of precipitated endostatin into animal results in in vivo conversion into a soluble form, thus enabling it to exert its antiangiogenic and antitumor activities (2). However, this refolding process likely depends on a variety of in vivo factors that are biologically variable, thus rendering this in vivorenaturation unpredictable. In addition, accurately assaying insoluble endostatin is problematic. For this and other reasons, the efficacy and activity of endostatin are difficult to reproduce accurately. Because of these issues, insoluble endostatin is unlikely to become clinically useful.

In recognition of the shortcomings of insoluble endostatin, a soluble form of recombinant endostatin was produced via a yeast expression system (3). However, the relatively low yield of the system has made it difficult to produce this agent in quantities sufficient for extensive clinical evaluation. In addition, its production is very expensive and represents a large step up in cost over bacterial recombinant methods. These barriers have hampered the widespread translation of endostatin research to clinical practice. There is therefore a great need to increase the yield and to reduce the cost of the production of recombinant endostatin that is suitable for clinical use. Our strategy centered on the optimization of the E. coli expression system because of its higher efficiency in expressing foreign proteins as compared with the yeast system. In this study, we demonstrate generation of soluble recombinant murine endostatin from E. coli and evaluate its antiangiogenic and antitumor activities. We have developed a purification protocol by which recombinant endostatin can be expressed with high efficiency,purified as inclusion bodies, and finally refolded into a soluble conformation. Our in vivo experiments showed that the SRE³possesses antiangiogenic and antitumor activity comparable with the originally published insoluble form.

Plasmid expressing murine endostatin with a histidine tag (pTB01#4) was provided by Dr. Judah Folkman (Harvard Medical School, Boston, MA; Ref.2). The expression of endostatin is driven by IPTG-inducible T7lac promoter elements (6). The bacteria strain used for endostatin expression is the BL21(DE3)pLysS strain from Promega [Madison, WI (7, 8)].

For small-scale expression screening, pTB01#4 was introduced into competent BL21(DE3)pLysS bacterial cells as described by the manufacturer. Ten colonies were randomly picked from the obtained transformants on a Kan⁺LB plate (10 grams/liter tryptone, 5 grams/liter yeast extract, 10 grams/liter NaCl, and 50 mg/liter kanamycin), and each of these was inoculated into 1 ml of Luria-Bertani medium containing 50 mg/liter kanamycin. These 10 inoculates were cultured in a 37°C shaker overnight at 280 rpm, and 50 μl of bacteria from each overnight culture were then transferred to a new tube containing 1 ml of the same broth, followed by continuous shaking culture at 37°C until A_{600 nm}reached 0.8. At this point, IPTG was added to the culture at a final concentration of 0.3 mm to induce the expression of endostatin. After culturing in a 37°C shaker for another 3-h period,10 μl of bacteria sample were collected from each 1-ml culture,followed by microcentrifugation at 1000 rpm for 5 min at 4°C. The bacterial pellets were resuspended with 10 μl of 1× sampling buffer[50 m[scap]m Tris-HCl (pH 6.8), 100 mm DTT, 2% SDS, 0.1% bromphenol blue, and 10%glycerol] and analyzed by SDS-PAGE.

The clone with the highest expression efficiency was identified from the 1-ml expression screening experiment by SDS-PAGE and subsequently applied to a 1-liter scale expression. The culture and induction condition for 1-liter scale expression were the same as that used for the 1-ml culture modified by simply scaling up each component by 1000 times.

To evaluate the biological activity of the soluble endostatin produced in this study, we used IRE as a reference for comparative studies. A purification procedure for recombinant endostatin from E. coli has been described previously (2). Briefly,bacteria pellet was collected with low-speed centrifugation, followed by lysis with 8 m urea. The lysate was then applied to a Ni₂-NTA column (Qiagen). After washing with 8 m urea containing 10 mm imidazole, endostatin was eluted with 8 m urea containing 250 mmimidazole. The dialysis product was subject to an endotoxin level determination (Limulu Amebocyte Lysate Progent/plus; Biowhittaker,Inc., Walkersville, MD). Quantification of the endostatin protein before dialysis was performed using the Bio-Rad protein dye method as described by the manufacturer. Finally, the endostatin product was dialyzed against 1× PBS (molecular weight cutoff, 6000–8000)at 4°C. During the dialysis, the purified protein precipitates to form IRE.

Bacteria were collected from the IPTG-induced 1-liter culture by centrifugation at 2500 × g for 10 min at 4°C. The pellet was then resuspended in 100 ml of buffer A [0.1 m Tris-HCl (pH 8.0) and 5 mm EDTA], followed by incubation at room temperature for 15 min, with the addition of lysozyme at a final concentration of 50 μg/ml. The suspension was then sonicated with a VibraCell VC50 sonicator (Sonic & Materials, Inc.) in the presence of 0.1% sodium deoxycholate, followed by centrifugation at 8000 × g for 10 min. The supernatant was discarded, and the pellet was resuspended in 100 ml of buffer A containing 0.1% sodium deoxycholate. The centrifugation/resuspension procedure was repeated twice. The resultant pellet from the last centrifugation was dissolved in 30 ml of buffer B [0.05 m Tris (pH 8.0), 1%SLS, and 1 mm DTT] and centrifuged at 8000 × g for 10 min at 4°C. The clear supernatant obtained was then transferred to dialysis tubing with a molecular weight cutoff of 8000 and dialyzed twice against 1500 ml of buffer C [0.05 m Tris-HCl, (pH 8.0) and 0.1 mm DTT] at 4°C for 4 h. The recombinant protein was then further dialyzed twice against 1500 ml of buffer D[0.05 m Tris-HCl (pH 8.0)] and 1000 ml of buffer E [0.05 m Tris-HCl (pH 8.0), 0.01 mm oxidized glutathione, and 1 mm reduced glutathione] at 4°C for 4 h/dialysis cycle, respectively. A final dialysis (4°C, 5 h)against 0.05 m Tris-HCl (pH 8.0) was performed twice to eliminate the glutathione redox pair. The dialysis product was subject to endotoxin level determination. The purified protein was finally quantitated by the Bio-Rad protein dye method as described by the manufacturer, aliquoted, and stored at −20°C.

SDS-PAGE was performed according to a standard procedure under reducing condition (9). Bacterial lysates from the small-scale expression screening procedure and 10 μg of SRE obtained as the final purification product were used as SDS-PAGE samples, respectively. Briefly, samples suspended in 10 μl of 1× sampling buffer were loaded on a 15% SDS-PAGE gel and run until the prestained molecular weight standard (Bio-Rad) revealed a good resolution, followed by staining with Coomassie Blue and destaining with methanol and acetic acid.

The SLS concentration in purified SRE was determined using a turbidimetric assay (10). Briefly, the SRE sample was serially diluted in distilled water to a final volume of 1 ml, and then 0.1 ml of 1 m HCl was added to each dilution and mixed by vortexing. After a 5-min incubation at room temperature, the A_{450 nm} values of the mixtures were measured. The amount of SLS present in the SRE sample was calculated from a standard curve made by using a series SLS with known concentrations ranging from 0.0025–0.025%. In this method, the A_{450 nm} value is a linear function of SLS concentration between 0.005% (0.05 mg/ml) and 0.05% (0.5 mg/ml).

In vivo Matrigel assay was used to test the antiangiogenic effect of soluble endostatin as described previously (11). Phenol red-free, growth factorreduced Matrigel (Collaborative Biomed and Cell Products, Bedford, MA) was mixed with 50 ng/ml recombinant human FGF-1 prepared as described previously(12) and 1 unit/ml heparin (Elkins-Sinn, Cherry Hill, NJ). C57BL/6 mice received s.c. injection in each flank of 0.4 ml of the prepared Matrigel. One day after injection, mice were divided into three groups (five mice/group). A control group was treated s.c. with PBS, and the rest of the groups were treated s.c. with either SRE or IRE (20 mg/kg, twice a day). Endostatin was injected at sites distant from the site of the Matrigel injection. Mice were sacrificed after 10 days of treatment. Harvested Matrigel plugs were fixed overnight in 10% buffered formalin at 4°C and then placed into 70%ethanol before being processed for paraffin sections. Immunohistochemical staining for factor VIII was carried out using Vectastain ABC according to the manufacturer’s instructions. Primary rabbit-derived antibody to human factor VIII was obtained from Dako. All photomicrographs were taken under identical conditions.

Confluent cultures of human umbilical vein endothelial cells were used to assess the ability of endostatin to inhibit endothelial cell migration and growth. These cells were grown with DMEM/10% FCS on 60-mm Falcon tissue culture dishes. Once the culture became confluent,half of the cell monolayer was sharply denuded using a sterile razor,and the demarcation line was etched onto the tissue culture vessel. The culture was washed with sterile PBS and incubated in DMEM/10%FCS/FGF-1 (10 ng/ml) containing either 0, 5, 10, or 30 ng/ml SRE. Cells were fed daily and fixed with 5% buffered formaldehyde 96 h after monolayer denudation. A calibrated ocular micrometer measured the distance traveled by the endothelial monolayer from the original denudation line. Measurements were taken from a point 0.5 cm proximal and distal from the center of the dish. Three cultures were used to test each concentration of endostatin.

The 3LL Lewis lung carcinoma model has been used extensively in preclinical testing of endostatin (2, 4, 5, 13). In this study, we used 3LL-C75, a cell clone isolated from 3LL Lewis lung carcinoma irradiated with UV light (14), to assess the antitumor effect of recombinant endostatin. In a separate set of animal experiments performed in our laboratory, this clone exhibited higher sensitivity to the growth-inhibitory effect of endostatin than the original 3LL tumor.⁴Tumor cells were cultured in RPMI 1640 supplemented with 10% FCS, 2 mm glutamine, and antibiotics. 3LL-C75 cells (1 × 10⁶) were inoculated into C57BL/6 mice. When tumors reached about 0.5 cm in diameter, mice were divided into three groups (five mice/group). Two groups of mice were treated s.c. with either SRE or IRE (20 mg/kg) twice a day. The control group received PBS injections. Tumor size was monitored by measuring the longest dimension (length) and shortest dimension three times/week with a dial caliper, and the tumor volume was calculated as width² × length × 0.52(15). All data are presented as mean ± SE. The experiments were terminated when tumors in control groups reached 2.0 cm in diameter or induced undue morbidity as per the protocol approved by the University of Pittsburgh Institutional Animal Care and Use Committee.

Our small-scale expression screening procedure identified a bacterial colony possessing the high expression efficiency (data not shown), from a cohort of 10 pTB01#4-transformed colonies. This clone was subsequently used for large-scale (1 liter) expression and purification.

In our protocol, the yield of purified SRE is approximately 150 mg/liter culture. This yield is similar to the IRE production made in our laboratory with the published insoluble protocol (2). The final dialysate in the soluble protocol did not contain any visible precipitants or fine particles and remained solubilized either at 4°C or after melting from −20°C storage. The levels of endotoxin in purified SRE and IRE were lower than 8 endotoxin unit/ml. When 10 μg of protein were analyzed by reducing SDS-PAGE, a single band of M_r 20,000 corresponding to endostatin was seen (Fig. 1). The identity of this homogeneous preparation was further confirmed by an endostatin-specific ELISA test (Cytimmune Science, College Park,MD). Almost all of the protein is reactive to the antiendostatin antibody provided in the kit (data not shown). Turbidimetric assessment of SLS showed that after serial dialysis, the amount of residual detergent in our final product was less than 0.005% (0.05 mg/ml),indicating that more than 99.5% of SLS used to solubilize the inclusion bodies was removed. The solubility remains stable even after the redox pair is eliminated from the system, indicating that this approach is effective in making soluble recombinant murine endostatin.

Matrigel-FGF from mice treated with saline shows a plethora of vascular endothelial cells, as indicated by factor VIII positivity and by the ability of these cells to form vessel-like or lumenal structures,within the Matrigel (Fig. 2, A and B). In marked contrast, there is a paucity of cells within Matrigel harvested from mice treated with insoluble(Fig. 2,C) or soluble (Fig. 2 D) endostatin (20 mg/kg/day, 9 days). No inflammation was evident in any of these sections. These results indicate that at this dose level, the antiangiogenic effects of SRE and IRE were equivalently potent.

The ability of the SRE to inhibit monolayer migration of human umbilical vein endothelial cells after partial denudation was tested. Our result showed that 5 and 10 ng/ml endostatin resulted in a 42% and 55% inhibition, respectively. This is significantly different from control cultures (P = 0.037 and 0.030,respectively). This effect was particularly dramatic at 30 ng/ml endostatin because none of these endothelial monolayers was able to traverse the original denudation line. Thus, no migration occurred at this concentration of endostatin. This result indicates that the E. coli-derived SRE has inhibitory activity on endothelial cell migration similar to that of the endostatin purified from the recombinant yeast system (4).

To assess the antitumor activity of the obtained SRE, we used the 3LL-C75 lung carcinoma model in this study. C57BL/6 mice bearing a tumor of about 0.5 cm in diameter were divided into three groups (five mice/group). Two groups of mice were treated twice with 20 mg/kg/day of either SRE or IRE. Control mice received injection of PBS in parallel. A substantial inhibitory effect was observed in mice treated with either IRE or SRE, and the degree of inhibition appeared to be similar(Fig. 3). These results suggest that at this dose, the bioactivity of the soluble form of recombinant endostatin is similar to that of insoluble endostatin. Thus, an expression and purification system for SRE protein from E. coli has been successfully established in a laboratory setting.

Here we report a protocol for purification of murine endostatin protein from E. coli as a soluble form. As a result, the bioactivity and solubility of the endostatin are very well maintained, whereas a high yield (150 mg/liter culture) and a high purity (>99%) have been achieved. The in vivoantiangiogenic activity and antitumor effect of the purified protein are comparable to those of the IRE prepared under denaturing condition. The yield and purity of this antiangiogenic protein produced from the reported procedure allow its virtual application at different laboratory levels. The established protocol also has the potential to be adapted to a larger scale production.

The IPTG-inducible T7lac promoter used in our system has previously been shown to be highly efficient in expressing heterologous proteins,including endostatin (2, 16). With the conventional purification protocol described previously (2), the cultured bacteria are lysed under denaturing condition (8 murea), and the dissolved endostatin is then subject to affinity chromatography for His tag-specific purification. However, the purified protein misfolds and precipitates during the dialysis that eliminates urea from the solvent system (2). In this study, we used a different purification approach that takes the advantages of inclusion body formation in the endostatin-expressing E. coli cells. Many proteins expressed in the described system aggregate and form intrabacterial inclusion bodies. The inclusion bodies can be easily collected by simple procedures such as centrifugation so that the target protein can be highly enriched at early stages of purification. The enrichment, followed by a rinsing procedure, could virtually be optimized to achieve high purity without further purification procedures such as affinity chromatography. In addition, inclusion bodies also provide the aggregated protein molecules with protection against degradation by intracellular proteases, which is not an uncommon posttranslational event for proteins heterologously expressed in bacteria. In most cases, however, protein molecules aggregated in inclusion bodies are not correctly folded, resulting in the generation of insoluble and/or biologically inactive molecules. Therefore,recovery of a recombinant protein from inclusion bodies requires an effective refolding process with optimized conditions that might vary from protein to protein. In our approach, detergent SLS and reducing agent DTT were used to dissolve the inclusion bodies that are partly but highly purified by repeated rinsing with a mild detergent-containing buffer. The dissolved endostatin molecules are then subject to a refolding procedure using a redox pair and oxidized and reduced glutathiones to facilitate formation of a stable conformation that makes and keeps the endostatin soluble.

This study outlines a strategy for the isolation of a soluble form of endostatin. We have focused this strategy on exhibiting that SRE is equivalent to IRE at a typical dose level (20 mg/kg/12 h) previously shown to be active in a large number of studies (4, 13). The true potency of SRE will require more extensive dose ranging studies. However, our preliminary data show that the same dose of SRE given once daily gives a similar angiogenesis inhibition as shown here(data not shown). Thus, we anticipate that this soluble material will have potency over a large dose range.

Endostatin is a potent antiangiogenic protein and an antitumor factor. However, the insoluble nature of the published version of E. coli recombinant endostatin hampered its clinical application. Although soluble endostatin prepared from a yeast system is being used in ongoing Phase I clinical trials, the low yield (approximately 20 mg/liter culture) and high cost of the system have made it difficult to produce in quantities that are realistic for comprehensive clinical evaluation and application. Our results presented in this report offer an alternative method that will prove valuable in helping to determine the clinical activity of endostatin. Obviously, it will be of great interest and importance to adopt a similar method in preparation of soluble human endostatin. This subject is being explored in our laboratory.

We thank Drs. Judah Folkman, Micheal O’Reilly, Jie Lin, and Tomas Boehm for providing the endostatin expression plasmid and extensive discussion on this work and other relevant subjects.