Cyclooxygenase-2 inhibition causes antiangiogenic effects on tumor endothelial and vascular progenitor cells

Authors

  • Chikara Muraki,

    1. Division of Oral Pathology and Biology, Graduate School of Dental Medicine, Hokkaido University, Sapporo, Japan
    2. Division of Vascular Biology, Graduate School of Dental Medicine, Hokkaido University, Sapporo, Japan
    3. Division of Oral and Maxillofacial Surgery, Graduate School of Dental Medicine, Hokkaido University, Sapporo, Japan
    Search for more papers by this author
  • Noritaka Ohga,

    1. Division of Vascular Biology, Graduate School of Dental Medicine, Hokkaido University, Sapporo, Japan
    Search for more papers by this author
  • Yasuhiro Hida,

    1. Department of Surgical Oncology, Graduate School of Dental Medicine, Hokkaido University, Sapporo, Japan
    Search for more papers by this author
  • Hiroshi Nishihara,

    1. Laboratory of Translational Pathology, Department of Pathology, Hokkaido University Graduate School of Medicine, Sapporo, Japan
    Search for more papers by this author
  • Yasutaka Kato,

    1. Laboratory of Cancer Research, Department of Pathology, Graduate School of Medicine, Hokkaido University, Sapporo, Japan
    Search for more papers by this author
  • Kunihiko Tsuchiya,

    1. Division of Oral Pathology and Biology, Graduate School of Dental Medicine, Hokkaido University, Sapporo, Japan
    2. Division of Vascular Biology, Graduate School of Dental Medicine, Hokkaido University, Sapporo, Japan
    3. Department of Renal and Genitourinary surgery, Graduate School of Medicine, Hokkaido University, Sapporo, Japan
    Search for more papers by this author
  • Kohei Matsuda,

    1. Division of Vascular Biology, Graduate School of Dental Medicine, Hokkaido University, Sapporo, Japan
    2. Division of Oral and Maxillofacial Surgery, Graduate School of Dental Medicine, Hokkaido University, Sapporo, Japan
    Search for more papers by this author
  • Yasunori Totsuka,

    1. Division of Oral and Maxillofacial Surgery, Graduate School of Dental Medicine, Hokkaido University, Sapporo, Japan
    Search for more papers by this author
  • Masanobu Shindoh,

    1. Division of Oral Pathology and Biology, Graduate School of Dental Medicine, Hokkaido University, Sapporo, Japan
    Search for more papers by this author
  • Kyoko Hida

    Corresponding author
    1. Division of Vascular Biology, Graduate School of Dental Medicine, Hokkaido University, Sapporo, Japan
    • Department of Oral Pathology and Biology, Division of Vascular Biology, Graduate School of Dental Medicine, Hokkaido University, N13 W7 Kita-ku, Sapporo 060-8586, Japan
    Search for more papers by this author

Abstract

Tumor angiogenesis is necessary for solid tumor progression and metastasis. Cyclooxygenase (COX)-2 is known to play an important role in cancer growth and invasion, and it activates the signaling pathways controlling cell proliferation, migration, apoptosis, and angiogenesis. COX-2 is reported to be expressed in many cancer cells. Several studies have reported successful treatment of cancer cells with COX-2 inhibitors (COX-2is). However, the effect of COX-2 inhibition on the tumor endothelium remains to be elucidated. Our study shows that COX-2 is expressed in the vasculature of surgically resected human tumors. To investigate the effects of COX-2 inhibition on the tumor endothelium in vitro, we isolated tumor endothelial cells (TECs) from human melanoma and oral carcinoma xenografts in mice, in which we confirmed that tumor growth was suppressed by inhibiting angiogenesis with the COX-2is NS398. COX-2 mRNA was upregulated in TECs compared to normal endothelial cells (NECs). Cell migration and proliferation were suppressed by NS398 in TECs but not in NECs. The effects of NS398 in vivo were consistent with the in vitro results. The number of CD133+/vascular endothelial growth factor receptor-2+ cells in circulation was significantly suppressed by COX-2 inhibition. In addition, the number of progenitor marker-positive cells decreased in the tumor blood vessels after COX-2i treatment, which suggests that the homing of progenitor cells into the tumor was also blocked. We conclude that NS398 specifically targets both TECs and vascular progenitor cells without affecting NECs.

The tumor microenvironment has recently become a target for cancer chemoprevention as it plays important roles in tumorigenesis and tumor progression.1 In the tumor microenvironment, angiogenesis is necessary for tumor progression and metastasis.2 Control of tumor angiogenesis is therefore a very important preventive strategy against invasive cancers.3 Many molecules regulating tumor angiogenesis have been identified and characterized recently, including vascular endothelial growth factor (VEGF). A couple of antiangiogenic drugs have been implemented in clinical practice for several years.

In the tumor microenvironment, endothelial cells (ECs) are exposed to hypoxic conditions that induce hypoxia-inducible factor-1α, which upregulates cyclooxygenase (COX)-2 in many cancers. COX-2 is a key enzyme in prostaglandin production and is also known to be upregulated during inflammation and in tumor tissues.4 Nonsteroidal anti-inflammatory drugs (NSAIDs) inhibit COX-2 activity, and chronic intake of NSAIDs is known to significantly reduce the risk of colon cancer development.5 In addition, COX-2 stimulates angiogenesis by inducing VEGF expression in tumor and stromal cells.6 COX-2-overexpressing tumor cells express increased levels of angiogenic factors and induce tube formation of normal human umbilical cord vein EC.7 Several studies have suggested that COX-2 blockade has therapeutic benefits in various cancer models.8, 9 ECs are also exposed to hypoxia in solid tumors,10 and COX-2 expression in human umbilical vein ECs is upregulated in hypoxia,11, 12 although the reported mechanisms for hypoxic induction of COX-2 vary. A few reports have confirmed COX-2 expression in tumor ECs (TECs) in human pathological specimens, but the effect of COX-2 inhibitors (COX-2is) on TECs with altered phenotypes in the tumor microenvironment is unknown.

It has been recently reported that TECs differ from normal ECs (NECs). We have shown that TECs have an altered phenotype and they express specific markers such as tumor endothelial markers,13 epithelial growth factor (EGF) receptor,14 CD1315 and Dkk-3.16, 17 TECs also show different biological activities from NECs, for example, they grow and migrate faster than NECs.18 We have reported that TECs are more sensitive to certain drugs, such as EGFR inhibitors14 and the green tea polyphenol epigallocatechin-3 gallate (EGCG).19 On the other hand, TECs are reported to be more resistant than NECs to chemotherapeutic drugs such as vincristine.20 Furthermore, we have shown that mouse21, 22 and human TECs18 harbor cytogenic abnormalities. It has therefore been speculated that TECs are different from NECs in terms of response to COX-2is. Meanwhile, endothelial progenitor cells (EPCs) play a critical role in tumor angiogenesis,23 and strategies that block the mobilization of EPCs into circulation may provide a new approach in inhibiting tumor angiogenesis. Thus, it is necessary to investigate whether COX-2is specifically acts on TECs and EPCs rather than NECs, although the effects of COX-2is on EPC mobilization have not been investigated.

In our study, we compared the effects of a COX-2i on TECs and NECs in vitro and in vivo, and we investigated its effects on EPC mobilization into the circulation.

Material and Methods

Antibodies and chemicals

The following antibodies and chemicals were purchased: rabbit antihuman COX-2 polyclonal antibody (Cayman Chemical, Ann Arbor, MI); mouse antihuman CD31 antibody (DAKO Japan Co., Kyoto, Japan); rat antimouse CD31 antibody and fluorescein isothiocyanate (FITC)-antimouse CD31 antibody (eBioscience, San Diego, CA); rat antimouse Ly6A/E antibody (Sca-1; BD Pharmingen, San Diego, CA); FITC-antimouse CD133 antibody (eBioscience); FITC-Bandeiraea simplicifolia lectin 1-B4 (BS1-B4; Vector Laboratories, Burlingame, CA); Alexa Fluor 488 goat antirat antibody (Molecular Probes); normal rat IgG (BD Pharmingen); monoclonal rabbit anti-VEGF receptor-2 (VEGFR2) antibody (BD Pharmingen); anti-β-actin-peroxidase antibody (Sigma, St. Louis, MO); P-Akt antibody and Akt antibody (Cell Signaling Technology, Boulder, CO) and 4,6-diamidino-2-phenylindole (DAPI; Roche, Indianapolis, IN). The PI3K inhibitor LY294002 was purchased from Calbiochem (San Diego, CA).

Immunohistochemistry and microvessel density

Formalin-fixed paraffin-embedded specimens, including four cases of oropharyngeal squamous cell carcinoma and two cases of malignant melanoma, were selected. The protocols were approved by the Institutional Ethics Committee, and written informed consent was obtained from each patient before surgery. The human specimens were sectioned and stained with hematoxylin and eosin using a standard protocol. Immunohistochemical analysis was performed using the primary antibodies for COX-2 (1:50 dilution) and CD31 (1:50 dilution) followed by detection of the antibody with a peroxidase-conjugated streptavidin-diaminobenzidine (DAB) readout system (DAKO). In the mouse model, tumors were dissected from mice after sacrifice and divided along the longest diameter using a surgical knife. Tumors were embedded in cryocompound (Tissue-Tek; Miles, Elkhart, IN) and immediately immersed in liquid nitrogen. The frozen specimens were sectioned using a cryotome at a thickness of 10 μm. Frozen sections were fixed in 100% ice-cold acetone for 10 min and blocked with 2% goat and 5% sheep serum in phosphate-buffered saline (PBS) for 30 min. All procedures for animal experimentation were approved by local animal research authorities, and animal care was in accordance with institutional guidelines.

Sections were incubated with primary antibody for 16 hr and Alexa Fluor 488-secondary goat anti-rat antibody for 2 hr and were counterstained with DAPI. Images were obtained randomly with an Olympus IX-71 microscope. For microvessel density (MVD) analysis, the number of vessels per area in CD31-stained sections was determined using MetaMorph software (Molecular Devices, Tokyo, Japan).

Cell lines and culture conditions

The A375SM, super-metastatic human malignant melanoma cell line, was kindly gifted by Dr. Isaiah J Fidler (M.D. Anderson Cancer Center Houston, TX). The cells were cultured in a humidified atmosphere of 5% CO2 and 95% air at 37°C in minimum essential medium (MEM; GIBCO, Grand Island, NY) supplemented with 10% heat-inactivated fetal bovine serum (FBS). The medium was changed every 3 days. The HSC-3, human oral carcinoma cell line, was supplied by the Japanese Cancer Research Bank (Tokyo, Japan). The cells were cultured in the same condition as described above in Dulbecco's modified Eagle's medium (DMEM; Sigma) supplemented with 10% heat-inactivated FBS.

Tumor xenografts in mice

Pathogen-free six-week-old (∼20 g) female athymic nude mice (nu/nu) were obtained from Sankyo Labo Service Corporation (Tokyo, Japan) and randomly divided into two groups (groups A and B; five animals/group). A375SM or HSC-3 cells (1 × 106 cells/mouse) were inoculated subcutaneously into the left flanks of nude mice. Mice received a single intraperitoneal injection of NS398 (5 mg/kg; Cayman Chemical) after tumor induction (day 0) and three times weekly thereafter for the duration of the experiment. NS398 was prepared as described previously.24 Control mice received the same amount (0.6%) of Dimethylsulfoxid (DMSO) in 750 μl of PBS. The shortest and longest diameters of the tumors were measured on the day of treatment, and tumor volume (mm3) was calculated using the following standard formula: (the shortest diameter)2 × (the longest diameter) × 0.5. No differences were observed in the body weight among the groups throughout the experiment. Mice were sacrificed under anesthesia 4–5 weeks after inoculation.

Isolation of TECs from melanoma and oral carcinoma xenografts

ECs were isolated as described previously.21 TECs were isolated from melanoma (A375SM) and oral carcinoma (HSC-3) xenografts in nude mice. It is considered that TECs derived from new blood vessels came from dermis, because our tumors were inoculated subcutaneously. Thus, we decided to isolate skin EC as control in these experimental models as described previously.21 NECs were isolated from the skin and served as a control. ECs were isolated by Magnetic cell separation (MACS) (Miltenyi Biotec, Tokyo, Japan) according to the manufacturer's instructions using FITC-anti-CD31 antibody. CD31-positive cells were sorted and plated onto 1.5% gelatin-coated culture plates and grown in Endothelial growth medium (EGM)-2 MV (Clonetics, Walkersville, MD) and 15% FBS. Diphtheria toxin (500 ng/ml; Calbiochem) was added to TEC cultures to kill any remaining human tumor cells21, 25 and to NECs to ensure that all ECs were consistently treated in the same manner. The isolated ECs were purified by a second round of purification using FITC-BS1-B4, and the purity was determined as described previously.21 For all in vitro experiments, different lots of TECs and NECs (i.e., isolated on different date) were used and these experiments were perfomed three times.

Isolation of RNA and quantitative real-time PCR

RNA was isolated using the RNeasy Micro Kit (Qiagen, Valencia, CA) according to the manufacturer's instructions. The RNase-Free DNase Set (Qiagen) treatment was incorporated into the RNA isolation protocol. The primers were as follows: COX-2 forward, 5′-CAGACAACATAAACTGCGCCTTTT-3′, reverse, 5′-GACTTCCTGCCCCACAGCAA-3′ and GAPDH forward, 5′-TCTGACGTGCCGCCTGGAG-3′, reverse, 5′-TCGCAGGAGACAACCTGGTC-3′. RNA (5 μg) was used in a 15-μl reaction, and quantitative real-time Polymerase chain reaction (PCR) was performed using the DyNAmo SYBR Green qPCR Kit (Finnzymes, Espoo, Finland) according to the manufacturer's instructions. Cycling conditions followed the manufacturer's instructions based on the use of Opticon Monitor version 3.0 (Bio-Rad, Hercules, CA). Cycling was performed as follows: activation of the polymerase for 15 min at 95°C, 30 cycles of 15 sec at 95°C, 15 sec at 57°C and 20 sec at 72°C. The expression levels of COX-2 mRNA were normalized to GAPDH.

Measurement of prostaglandin E2 level

ECs (4 × 103 cells/ml) were plated in a 96-well culture dish for 18 hr in EGM-2 and then treated by NS398 treatment (0, 10, 30 and 50 μM) for 4 hr. After NS398 treatment, supernatants were collected, and prostaglandin E2 (PGE-2) levels were measured by prostaglandin E2 kit (Cayman Chemical). The experiment was repeated three times with similar results.

Cell migration assay

Cell migration was measured in a Boyden chamber as described previously with modifications.19 In the upper chambers, 1.5 × 104 ECs in Endothelial basal medium (EBM)-2/0.5% FBS were seeded and then NS398 was added. Medium containing 10 ng/ml of VEGF was added in the lower chambers as a chemoattractant. After 4 hr at 37°C, the cells migrating through vitronectin (3 μg/ml)-coated polycarbonate membranes (8-μm pores, Corning Costar, Nagog Park, MA) were blocked with 1% Bovine serum albumin (BSA) MTS:3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium, fixed in 2% paraformaldehyde and stained with DAPI. The number of cells that migrated to the lower side of the filter was counted in three high-power fields. Similarly, after starvation for overnight in EBM-2/0.5% FBS, ECs were pretreated with LY294002 (20 μM) for 2 hr and were seeded in EBM-2/0.5% FBS, including LY294002. The data is presented as the average of three counts ± SD. The experiment was repeated three times with similar results.

In vitro proliferation assay

Cell proliferation was measured by the MTS assay as described previously.19, 26 Cultured cells were briefly seeded at 4 × 103 cells per well in 96-well flat-bottomed plates in MEM or DMEM for tumor cells or EGM-2 MV for ECs and allowed to adhere for 6 hr. After 16 hr of serum starvation, NS398 was added at final concentrations of 0, 10, 25 and 50 μM in EBM-2/5%FBS. After 72 hr of culture in the presence of NS398, proliferative activity was determined by the MTS assay (CellTiter-96 Non-Radioactive Cell Proliferation Assay; Promega, Madison, WI), which monitors the number of viable cells according to the manufacturer's instructions. The results represent the average of six wells per cell line. The experiment was repeated three times with similar results.

Western blotting analysis

Western blot analysis was also performed for the detection of COX-2 in nonstimulated EC and the phosphorylation of Akt by NS398 treatment in EC.

ECs were lysed as described previously.19 Total protein was assayed using the Bicinchoninate (BCA) protein assay kit (Thermo Scientific, Rockford, IL). After electrotransfer, the membranes were probed with COX-2 antibody (Cayman Chemical) and P-Akt antibody in Solution 1 (Can Get Signal Immunoreaction Enhancer Solution; Toyobo, Osaka, Japan) overnight at 4°C. After washing, the membrane was incubated with Horseradish peroxidase (HRP)-conjugated anti-rabbit IgG secondary antibody (Cell Signaling Technology) diluted in Solution 2. We used the Western Lighting Chemiluminescence reagent plus system™ (Perkin Elmer Life Sciences, Boston, MA) for detection. The membranes were stripped by washing with striping buffer (Thermo) and were reblotted with monoclonal anti-β-actin-peroxidase antibody and Total Akt antibody.19

For the detection of phosphorylated Akt in NS398 treatment, ECs were cultured in serum-free medium before NS398 treatment for 16 hr. ECs were then treated with NS398 (50 μM) in 0.5% FBS/EBM for 0, 30 and 120 min. After NS398 treatment, ECs were treated with or without VEGF (10 ng/ml) for 30 min. The level of COX-2 was normalized to β-actin, and the level of phosphorylated Akt was normalized to total Akt by scanning densitometry using Image J software from the National Institute of Health (NIH) (Bethesda, MD).27

Blood collection and mononuclear cell isolation from in vivo tumor model

Peripheral blood (700–1,000 μl) was collected from each mouse in both control and NS398-treated tumor groups (n = 5 in each group) before sacrificing the mice. Mononuclear cells were isolated by sucrose gradient centrifugation from peripheral blood as described previously with modifications,28 and the cell number was counted. Peripheral blood mononuclear cells (PBMCs) were incubated with FITC- conjugated monoclonal antibodies against mouse CD133 and Phycoerythrin (PE)-conjugated monoclonal antibodies against mouse VEGFR2 for Fluorescence-activated cell sorter (FACS) to measure the number of CD133+/VEGFR2+ cells. A minimum of 10,000 cells per sample were analyzed by flow cytometric analysis of cell surface protein (FACS) using a FACS Caliber flow cytometer.

Measuring the number of circulating mononuclear cells and endothelial progenitor cells

VEGF (300 ng) was injected intraperitoneally into the mice to mobilize EPCs from bone marrow as described previously with modifications.29 NS398 (5 mg/kg) or the vehicle was also injected once a day for 2 days. On day 3, peripheral blood was collected from each mouse (n = 3), and mononuclear cells were isolated as described above. After counting PBMCs, they were incubated with FITC-conjugated monoclonal antibodies against mouse CD133 and PE-conjugated monoclonal antibodies against mouse VEGFR2 for FACS, to measure the number of CD133+/VEGFR2+ double-positive cells in circulation.

Statistical analysis

Differences between the groups were statistically evaluated using the Mann–Whitney U test. The results are presented as means ± SD. All p values were two-tailed, and p < 0.05 was considered statistically significant.

Results

COX-2 is overexpressed in TECs in human cancers

We evaluated COX-2 expression in human tissue samples of oral squamous carcinoma and malignant melanoma by immunohistochemistry. In all four cases of squamous cell carcinoma and two cases of malignant melanoma, we found prominent expression of COX-2 in vascular ECs within tumors, which were also positive for the endothelial marker CD31 in serial sections and only weakly positive in vessel in normal tissue (Fig. 1).

Figure 1.

COX-2 is expressed in tumor blood vessels in human cancers. COX-2 expression was analyzed by immunostaining. CD31-positive endothelial cells were also stained with anti-COX-2 antibody in two cases of malignant melanoma (cases 1 and 2) and four cases of human oral squamous cell carcinoma (cases 3–6). These results suggest that COX-2 expression is induced in tumor endothelial cells in human tumors and only weakly expressed in vessel in normal tissue in vivo.

NS398 suppressed tumor growth in vivo with inhibition of angiogenesis

As COX-2 expression was found in TECs in human tumor sections, we investigated whether tumor angiogenesis in the mouse tumor model is inhibited by the COX-2i NS398. A375SM human melanoma cells or HSC-3 human oral carcinoma cells were inoculated subcutaneously into nude mice. Either NS398 or vehicle was administered intraperitoneally into the mice three times a week from day 0. After 35 days, the growth of the NS398-treated melanoma was suppressed compared to the control group (444.4 ± 512.7 mm3vs. 79.3 ± 47.1 mm3, mean ± SD, p < 0.05) (Fig. 2a). In the oral carcinoma model, the size of NS398-treated tumors was significantly smaller than the size of control tumors on day 29 (680.3 ± 264.2 mm3vs. 362.2 ± 125.4 mm3, mean ± SD, p < 0.05) (Fig. 2b). Drug treatment was well tolerated without apparent toxicity or body weight changes throughout the study. Snap-frozen tumor tissue specimens were processed for immunohistochemical study. CD31-stained vessel areas markedly decreased in NS398-treated tumors (Fig. 2c-b, d) compared to the controls (Fig. 2c-a, c). Morphometric analysis of 50 separate digital images per group showed that the MVD of the NS398-treated tumors was less than 50% of the MVD of control tumors (Fig. 2d), suggesting that NS398 markedly inhibited tumor angiogenesis.

Figure 2.

Tumor growth of melanoma and oral carcinoma xenografts was suppressed by NS398 treatment due to decreased vessel density. The tumor cells were inoculated subcutaneously into nude mice. Day 0 was considered as the first day of treatment. Two groups of animals were then treated with the vehicle (0.6% DMSO, n = 5) or the COX-2i NS398 (5 mg/kg, n = 5) intraperitoneally three times a week. Points; mean tumor volume over time. (a) Melanoma xenograft. (b) Oral carcinoma xenograft. Melanoma tumor growth was inhibited by NS398. In particular, oral carcinoma growth was significantly suppressed compared to control tumors (*p < 0.05). (c) Snap-frozen tumor tissue specimens were processed for the immunohistochemical study after mice were sacrificed. Cryosections were stained with a FITC-labeled anti-CD31 antibody. The sections were counterstained with DAPI. Representative images of melanoma (a, b) and oral carcinoma (c, d) are shown. The CD31-positive vessel area was decreased in NS398-treated tumors (b, d) compared to control tumors (a, c). Bar, 100 μm. (d) MVD was analyzed quantitatively by comparing the NS398-treated group with the control group. In both melanoma and oral carcinoma treated by NS398, MVD was significantly decreased (*p < 0.01). [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

NS398 inhibited cell proliferation of isolated TECs

Cell proliferation was analyzed in A375SM and HSC-3 cells that were used in tumor xenograft experiments. Tumor cell proliferation was suppressed modestly by 50 μM NS398, although the growth of tumors in vivo was inhibited using these cell lines (Fig. 3a). We isolated two different TECs, melanoma ECs and oral carcinoma ECs, to investigate the effects of NS398 on TECs. NECs from dermal tissue were also isolated as a normal counterpart for TECs. Cell proliferation was significantly inhibited by 10 μM NS398 in both TEC lines, whereas that of NECs was only slightly inhibited (Fig. 3b) (p < 0.01). COX-2 mRNA expression was analyzed by regular-PCR. Both TEC lines showed elevated expression of COX-2 compared to NECs (Fig. 3c). These results were confirmed by real-time PCR, which indicated that COX-2 mRNA expression was significantly higher in TECs than in NECs (Fig. 3d, p < 0.01). Western blotting of COX-2 also showed that COX-2 expression is upregulated in TECs (Figs. 3e and 3f).

Figure 3.

The effect of NS398 on cell proliferation and on PGE2 synthesis in isolated TECs and NECs. (a) Tumor cell proliferation was assessed by MTS assay. NS398 did not inhibit cell proliferation of either melanoma or oral carcinoma cells. (b) To investigate the effects of NS398 on TECs, TECs were isolated from human tumor (melanoma and oral carcinoma) xenografts in nude mice. NECs were isolated from skin tissue in nude mice and were used as a normal control. Cell proliferation was significantly inhibited by 10 μM NS398 in both TEC, whereas that of NECs was only slightly inhibited (Fig. 3b) (*p < 0.01). (c) RT-PCR analysis of COX-2 mRNA expression in tumor endothelial and normal endothelial cells. COX-2 was represented by a 118-bp product. (d) The COX-2 mRNA expression level was higher in tumor endothelial cells than in normal endothelial cells analyzed by real-time PCR (*p < 0.01). (e) Western blotting of COX-2 also showed that COX-2 expression was upregulated in TECs. (f) The levels of COX-2 were normalized for total β -actin by scanning densitometry using computed image analysis. The graph shows the relative ratio of COX-2/β-actin by densitometric analysis. (g) Prostaglandin E2 (PGE-2) levels of TECs were significantly higher than those of NECs [melanoma EC: 900 ± 23.1 pg/ml (p < 0.05 vs. skin EC), oral carcinoma EC: 1300.4 ± 115.7 pg/ml (*p < 0.01 vs. skin EC), skin EC: 208.0 ± 11.3 pg/ml]. NS398 inhibited PGE-2 synthesis in TEC in dose dependant manner, assessed by ELISA.

In addition, PGE-2 levels of TECs were significantly higher than those of NECs [melanoma EC: 900 ± 23.1 pg/ml (p < 0.05 vs. skin EC), oral carcinoma EC (p < 0.01 vs. skin EC): 1300.4 ± 115.7 pg/ml, skin EC: 208.0 ± 11.3pg/ml]. COX-2i, NS398 treatment decreased PGE-2 synthesis in dose dependant manner in TEC assessed by ELISA (Fig. 3g).

TEC migration was inhibited by NS398

VEGF-stimulated EC migration in the presence or absence of NS398 was analyzed on vitronectin-coated membranes using a Boyden chamber. NS398 significantly inhibited TEC migration toward VEGF (p < 0.01), whereas the migration of NECs was not affected (Fig. 4a). As cell migration is related to adhesion to the matrix in addition to chemotaxis, cell adhesion on vitronectin was analyzed. Cell adhesion by TECs or NECs was not suppressed by equal concentrations of NS398 (Fig. 4b). Suppression of TEC migration is considered to be due to inhibition of migration, not proliferation, as migration assay is a 4-hr experiment. To evaluate the signaling pathway involved in cell migration, Akt phosphorylation was analyzed by Western blotting. The phosphorylation of Akt was suppressed by NS398 with or without VEGF stimulation in TECs, while Akt phosphorylation was not changed in NECs (Fig. 4c). Both, basal phosphorylated Akt and Akt phosphorylation levels induced by VEGF were markedly suppressed by NS398 treatment, whereas Akt phosphorylation in NECs was not affected by NS398 (Fig. 4d). Furthermore, we analyzed the effect of the PI3K inhibitor LY294002 on VEGF-induced migration in ECs. LY294002 (20 μM) suppressed the VEGF-induced cell migration of ECs (Fig. 4e). These results suggest that cell migration of ECs is stimulated by VEGF through activation of Akt-PI3K signaling pathway. These results suggest that NS398 suppresses TEC migration toward VEGF through inactivation of Akt signaling at a concentration that does not affect NECs.

Figure 4.

NS398 suppressed tumor endothelial cell migration by blocking Akt phosphorylation. (a) Endothelial cell migration toward VEGF was analyzed on vitronectin-coated membranes using a Boyden chamber. After 4 hr of incubation either in the presence or absence of NS398, the membrane was fixed and stained with DAPI. The cells that migrated to the lower side of the membrane were counted in three different high-power fields. The cell migration toward VEGF by TECs, but not NECs, was inhibited significantly by NS398 (*p < 0.01). The results represent the average of three counts ± SD. Cell proliferation was assessed by the MTS assay. (b) Cells adhesion on vitronectin was analyzed. Cell adhesion by TECs or NECs was not suppressed by NS398 (50 μM). The results represent the average ± SD of triplicate experiments. (c) Endothelial cells were incubated with NS398 (50 μM) in 0.5% FBS/EBM-2 for 0, 30 or 120 min. After NS398 treatment, cells were stimulated with VEGF (10 ng/ml) for 30 min. Equal protein concentrations of cell lysates were loaded onto lanes. Phosphorylated Akt (P-Akt) levels were determined by Western blot analysis using an anti-phospho-Akt antibody. NS398 reduced Akt phosphorylation in TECs (melanoma EC and oral carcinoma EC) but not in NECs (skin EC). The blots were stripped and reprobed for total Akt as a loading control. (d) The levels of phosphorylated Akt were normalized by total Akt and measured quantitatively by densitometry using the Win ROOF version 5.0 Analysis software. It was shown that the phosphorylated Akt level in TECs was markedly decreased by NS398, whereas Akt phosphorylation was not changed in NECs. (e) The effect of PI3K inhibitor LY294002 on VEGF-induced migration in ECs. LY294002 (20 μM) suppressed the VEGF-induced cell migration of ECs.

NS398 suppressed VEGF-stimulated mobilization of CD133+/VEGFR2+ cells

We previously reported that TECs, either in isolated ECs or in xenograft sections, coexpress Sca-1 and CD31, suggesting a contribution of EPCs to TECs.26 We analyzed the number of CD133+/ VEGFR2+ cells in the blood of mice treated with or without NS398/VEGF. The number of circulating CD133+/VEGFR2+ cells was counted by FACS and compared in tumor-bearing mice between the NS398-treated group and the control group, which were analyzed in the in vivo tumor growth experiment described in Figure 2. The number of circulating CD133+/VEGFR2+ cells significantly decreased after NS398 administration in the VEGF-treated group (p < 0.05) (Fig. 5a), and they did not change with NS398 treatment without VEGF coadministration. This suggested that NS398 inhibited VEGF-stimulated EPC mobilization from the bone marrow (Fig. 5a). As the number of VEGFR2+ cells in circulation was not changed, it was presumed that the number of mature circulating ECs was not affected by NS398 (Fig. 5b). Cryosections of these tumors were stained with anti-CD133 and anti-VEGFR2 antibodies to detect the EPC populations. Representative images are shown (Fig. 5c). In NS398-treated tumors, the population of CD133-positive cells was less than that in the control group (Fig. 5c), and the number of CD133+/VEGFR2+ cells also decreased in the NS398-treated tumors (Figs. 5c and 5d). These results suggest that the homing of EPCs into tumor tissue was also inhibited by NS398. It has been reported that MMP-9 in the bone marrow plays an important role in the mobilization of bone marrow-derived VEGFR2+ cells into the peripheral circulation. To test the effects of NS398 on MMP-9 expression in bone marrow stromal cells (BMSCs), we isolated BMSCs from oral carcinoma-bearing mice. To investigate whether NS398 affects the expression of MMP-9 mRNA in BMSCs, these cells were treated with or without NS398 (50 μM) for 24 hr; MMP-9 mRNA expression in BMSCs was downregulated by NS398 (Fig. 5e).

Figure 5.

The EPC numbers in circulation and in tumors were decreased by NS398. (a) NS398 or vehicle was injected into mice once a day for 2 days. On day 3, peripheral blood was collected from each mouse prior to sacrifice. The number of CD133/VEGFR2 double-positive cells, determined as EPCs, was measured by FACS. VEGF-induced EPC mobilization into circulation was significantly decreased by NS398. (b) Peripheral blood was collected from each HSC-3 tumor-bearing mouse treated with the vehicle or NS398 (n = 5 in each group). CD133+, VEGFR2+ and CD133+/VEGFR2+ cell numbers were analyzed. Mature circulating EC (VEGFR2+) numbers were not changed by NS398, whereas CD133+ and CD133+/VEGFR2+ cell numbers were decreased by NS398. It was suggested that NS398 affects the mobilization of EPCs but not mature circulating ECs. The circulating EPC number was decreased in the NS398-treated group, suggesting that EPC mobilization from the bone marrow was also inhibited. (c) Cryosections of tumors were processed for immunohistochemistry using anti-mouse CD133 and PE-anti-mouse VEGFR2. Representative images are shown demonstrating CD133-positive cells (green), VEGFR2-positive cells (red) and the merged image (yellow). Nuclei were stained with DAPI (blue). CD133+/VEGFR2+ cell numbers were decreased in NS398-treated tumors (right panels) compared to control tumors (left panels), suggesting that EPC homing into the tumor was also decreased. (d) The CD133+VEGFR2+ area was compared between NS398-treated tumors and control tumors using immunostained in vivo tumor sections [as shown in (c)]. NS398 treatment decreased CD133+/VEGFR2+ progenitor cell numbers in tumors. (e) NS398 suppressed the mRNA expression of MMP-9 in bone marrow stromal cells (BMSCs) of oral carcinoma-bearing mice. The experiment was performed three times with similar results.

Discussion

In our study, we provided several new lines of evidence about the effects of COX-2i on tumor angiogenesis. We demonstrated COX-2 expression in TECs in human clinical tumor tissues. We found that COX-2 expression in TECs is rather homogenous, although it is heterogeneous in the tumor cells. TECs isolated from mouse xenograft tumors expressed higher levels of COX-2 than NECs. COX-2i specifically inhibits proliferation and migration of TECs via Akt phosphorylation but does not have this effect on NECs. NS398 had effects on TECs at concentrations as low as 10 μM, indicative of modest effects on the tumor cells. The COX-2i reduced the mobilization of CD133+/VEGFR2+ cells from bone marrow, partly due to the suppression of MMP-9 expression by BMSCs, resulting in diminished incorporation of CD133+/VEGFR2+ cells into the tumor vasculature.

In our previous studies with isolated TECs, we found that they are different from NECs in many aspects,22, 30 including sensitivity to growth factors and certain drugs such as EGF receptor kinase inhibitors14 and EGCG.19 In addition, TECs have cytogenetic abnormalities.18, 21

In our study, we demonstrated that COX-2 is overexpressed in TECs in human clinical cancers by investigating surgically resected squamous carcinoma and malignant melanoma specimens. All six investigated specimens expressed COX-2 homogeneously in TECs but heterogeneously in the tumor cells, which is consistent with the assumption that tumor angiogenesis is a primary target of COX-2is.

Next, we confirmed that melanoma and oral carcinoma growth and tumor angiogenesis were inhibited by the COX-2i. NS398 inhibited tumor growth in malignant melanoma and oral carcinoma xenografts in mice and markedly suppressed MVD. Although the antiangiogenic effect may be due to inhibition of tumor cell proliferation by the COX-2i, inhibition of tumor cell proliferation by NS398 in vitro was found to be modest. In contrast, proliferation and migration of TECs were inhibited at concentrations of NS398 as low as 10 μM. Inhibition by NS398 was significantly stronger in TECs than in NECs. This specificity appears to be due to higher expression of COX-2 in TECs. It appears that TECs are more dependent on COX-2 activity than NECs or tumor cells. We speculate that COX-2 secreted from TECs acts on themselves in autocrine manner, as in low serum culture condition, COX-2i decreased survival of TECs and apoptosis of TECs were inhibited by PGE-2 (data not shown).

COX-2i is reported to inhibit Akt phosphorylation, resulting in cell apoptosis.31 Akt signaling is involved in EC migration. Our results showed that Akt activation was blocked by the COX-2i in TECs but not in NECs. These data suggest that inactivation of Akt signaling is involved in the antiangiogenic effects of COX-2 inhibition on TECs. The specific inhibition of Akt phosphorylation in TECs suggests that the PI3K/Akt pathway may be a specific target of COX-2is. Our previous study also showed that a green tea polyphenol, EGCG, specifically inhibits proliferation and migration of TECs via Akt signaling.19

It was recently indicated that tumors mobilize bone marrow (BM)-derived EPCs in addition to recruiting neighboring blood vessels or ECs and that EPCs migrate to tumors and become incorporated into their developing vasculature.32 Moreover, EPCs are considered a novel target for antiangiogenic therapy as they play important roles in tumor metastasis.23 There is controversy about the identity and function of BM-derived EPCs. Some studies have not only questioned their relatively low contribution to tumor vasculature but also their functional significance in tumor growth. Broad variability ranging from a major contribution of EPC32, 33, 34 to a minor contribution35, 36 has been reported.

Cause of variability about EPC contribution to tumor vasculature may arise from the differences in tumor types,37 stage and insufficient functional analysis. Our previous reports suggest that EPCs contributed to tumor neovascularization and differentiate into TECs.19, 26, 38 Several antiangiogenic drugs have already been reported to have inhibitory effects on EPCs.39 However, the effect of COX-2 inhibition on EPCs has not been well documented. A recent study reported that COX-2 blockade decreased the number of circulating ECs that were positive for VEGFR2 or CD31, including the mature EC population.40 There is a lack of consensus regarding the precise panel of cell surface markers that uniquely define EPCs. Some studies have shown that subsets of VEGFR-2+ cells that coexpress the human progenitor marker CD133 have high proliferative capacity and comprise endothelial progenitors.41, 42, 34 In our study, we defined VEGFR2+/CD133+ cells as EPCs. In our results, COX-2 inhibition affected EPCs (VEGFR2+/CD133+) but not mature CECs (VEGFR2+/CD133). We are the first to provide evidence that systemic administration of a COX-2i suppresses EPC mobilization in tumor-bearing mice. In addition, the EPC population decreased in the tumors treated with the COX-2i, suggesting that EPC homing into the tumors was inhibited. Furthermore, to evaluate the effect of the COX-2i on EPC mobilization, circulating EPCs were counted in normal mice after treatment with VEGF and/or the COX-2i. VEGF-stimulated mobilization of EPCs was suppressed by COX-2 inhibition. Our results suggest that the homing of VEGFR2+/CD133+ cells might be inhibited by COX-2i. However, to prove this, further experiment using bone marrow transplantation model will be necessary. Anyway, given that EPCs are among the therapeutic targets for cancers as well as TECs, COX-2i would work on EPCs as well as TECs.

COX-2 inhibition has been evaluated in combination with chemotherapy in several cancers.43–45 COX-2is and rofecoxib improved the response rate and several quality-of-life measures in lung cancer patients including pain-related items and global quality of life, although it did not prolong patient survival. Severe cardiac ischemia was significantly more frequent with rofecoxib-treated patients.43 It merits mentioning that the dose of COX-2i is determined to inhibit TEC proliferation and migration and not for tumor cells. Lower doses may thus achieve lower cardiac toxicity without lowering the response rate. EPC number may be useful as a surrogate marker to evaluate and monitor the effects of COX-2is.

We conclude that TECs are more sensitive to COX-2is compared to NECs and tumor cells. An important consequence of this finding is that TECs could be targets for COX-2is and that COX-2is can be used to target the tumor vasculature directly at a lower dose. This may decrease the rate of undesirable side effects, such as myocardial infarction, observed with the use of COX-2is.46

Finally, EPCs from the bone marrow could be a target for anti-COX-2 therapy, leading to a strategy for antiangiogenic therapy for cancer prevention or for cancer therapy in established disease, either alone or in combination with other cancer therapies such as chemotherapy and radiation.

Acknowledgements

We thank Dr. I.J. Fidler for providing the A375SM super-metastatic human malignant melanoma cell line, and Ms. T. Takahashi for technical assistance. This work was supported by Grants-in-Aid for Scientific Research from the Ministry of Education, Science and Culture of Japan (K. Hida, F. Higashino and M. Shindoh), the Haraguchi Memorial Foundation for Cancer Research, the Akiyama Foundation and the Takeda Science Foundation (K. Hida).

Ancillary