Cyclooxygenase, involved in tumor growth and angiogenesis, converts arachidonic acid to prostaglandin (PG)H2, which is immediately converted to bioactive prostanoids including PGE2, PGD2, thromboxane (TX)A2 and PGI2. To test the hypothesis that changes in the prostanoid profile alter cancer growth, we transduced the retroviral vectors carrying TXA2 synthase cDNA or PGI2 synthase cDNA to colon-26 adenocarcinoma cells and subsequently inoculated each transformant to syngeneic BALB/c mice. Tumors derived from TXA2 synthase transformants grew faster (280%, day 8, versus null-vector control; P < 0.05) and showed more abundant vasculature (204%, versus null-vector control; P < 0.01), whereas tumors from PGI2 synthase transformants presented opposite effects. These effects by the transgenes were reversed by administration of specific inhibitors. These results suggest that the profile of downstream metabolites of cyclooxygenase in cancer cells can be a determinant for tumor development.

NSAIDs3 inhibit both COX-1, a constitutively expressed isozyme implicated in maintaining normal cellular functions, and COX-2, an inducible isozyme expressed in inflammatory lesions and in many types of cancers including colon, stomach, esophagus, lung, breast, prostate, skin, and melanoma (1, 2). Epidemiological, animal, and clinical studies have established that NSAIDs are effective for the prevention and size-reduction of colon cancers and have suggested that they may also be effective for other types of cancers (3, 4, 5). Studies using COX-2-specific inhibitors have demonstrated that the anticancer effect of NSAIDs is likely attributable to the inhibition of COX-2 activity (6), although the contribution of a COX-independent mechanism has also been suggested (7).

COXs convert arachidonic acid to PGH2, a common precursor to a variety of prostanoids (summarized in Fig. 1). PGH2, which by itself has no known physiological functions, is immediately catalyzed to bioactive prostanoids PGE2, PGD2, TXA2, PGF2α, and PGI2. The effects of COX expression in cancer cells are considered to be related to the fractional amounts of these prostanoids (i.e., the prostanoid profile; Refs. 8, 9). However, little information is available for the relationship of prostanoid profile and cancer growth. We hypothesize that: (a) changes in the prostanoid profile alter cancer growth; and (b) inhibitors of the procancer prostanoids retard cancer growth and vice versa.

Prostanoids exert a wide range of biological functions on a variety of cells. Some of these actions are in opposition to each other, for example, TXA2 promotes platelet aggregation and vasoconstriction, whereas PGIS inhibits platelet aggregation and promotes vasodilatation (10). Because of the fact that a variety of the host cells are involved in tumor growth, the effects of an individual prostanoid on tumor growth are hard to predict and must be studied in vivo. Therefore, to test our hypotheses, we have investigated the effects of TXA2 and PGI2 on tumor growth using a mouse model system. We chose these prostanoids because they have biological functions opposing each other and may provide us with a clear view of the relationship between prostanoid function and tumor growth. In this study, TXAS and PGIS were introduced into murine colon-26 adenocarcinoma cell line (C26) to alter the prostanoid balance. The resulting cells were then inoculated into syngeneic BALB/c mice, tumor growth and animal survival were monitored, and tumor histology was examined. Comparing the results of PGIS and TXAS gene transduction, we could test our hypotheses and infer the underlying mechanisms that resulted from differential prostanoid profiles.

Cell Lines and Animals.

Murine colon-26 adenocarcinoma cell line (C26) was maintained in RPMI 1640 (Life Technologies) with 10% FCS. The ψCRIP cells, a packaging cell line that produces replication-incompetent retrovirus (11), were maintained in DMEM (LifeTechnologies) with 10% calf serum (CS). Female BALB/c mice at 6–8 weeks of age were obtained from Charles River Japan.

Retroviral Vector Construction and Transduction into the C26 Cells.

The human TXAS cDNA and the human PGIS cDNA were gifts from Dr. Lee-Ho Wang. A 1.8-kb BamHI fragment that contained the entire coding sequence of TXAS and a 1.6-kb BamHI fragment that contained the entire coding sequence of PGIS were each blunt-ended and inserted into the HpaI site of the retroviral vector pLNCX (Clontech) to generate pLNCX-TXAS and pLNCX-PGIS, respectively (see Fig. 2 a). The three retroviral vector constructs, pLNCX-TXAS, pLNCX-PGIS, and pLXIN (a retroviral vector carrying the neoR gene driven by 5′ Moloney murine leukemia virus LTR, used as a control null vector; Clontech) were individually transfected into ψCRIP. Neo-resistant, retrovirus-producing cells were selected with 400 μg/ml G418 (Life Technologies, Inc.), and named ψCRIP-TXAS, ψCRIP-PGIS, and ψCRIP-LXIN. C26 cells were incubated with each viral supernatant in the presence of 8 μg/ml Polybrene (Aldrich Chemical), and the transduced cells were selected with 600 μg/ml of G418. The resultant colonies (>200 colonies) were collected as a mass population culture and designated as C26-TXAS, C26-PGIS, and C26-neo.

Northern Blot.

Total cellular RNA was extracted using the Isogen kit (Nippon Gene). RNA (10 μg) from each cell line was electrophoresed on a 1.2% agarose gel containing 2.2 m formaldehyde and transferred to Hybond nylon membrane (Amersham Pharmacia Biotech). TXAS, PGIS, and GAPDH cDNAs, labeled with [α-32P]dCTP (DuPont) using Prime-It kit (Stratagene), were used as probes. Hybridization was performed in Quikhyb solution (Stratagene) at 68°C overnight. Filters were washed three times in 2× SSC, 0.1% SDS at 68°C, three times in 0.2× SSC, 0.1% SDS at 68°C, and then exposed to XA-R film (Kodak) at −70°C overnight.

EIA.

The activities of TXAS or PGIS were indirectly estimated by measuring their stable metabolites, TXB2 or 6-keto PGF1α, by EIA, respectively. The TXB2 EIA kit and the 6-keto PGF1α EIA kit were purchased from Cayman Chemical. For TXB2, 3 × 105 cells were plated in 2 ml of RPMI 1640 2 h prior to the assay. For 6-keto PGF1α, 3 × 105 cells were plated in 2 ml of growth medium 24 h prior to the assay. The media were collected and subjected to EIA.

Cell Growth Assay.

C26-TXAS, C26-PGIS, C26-neo, and wild-type C26 cells were plated in 35-mm dishes (1 × 105 cells/well, in 2 ml of RPMI medium containing 10% FCS). The number of cells was counted after 24, 48, and 72 h of seeding.

Tumor Growth Assay.

C26-TXAS, C26-PGIS, C26-neo, and wild-type C26 cells (5 × 105 cells) were s.c. inoculated into the left flanks of BALB/c mice that are syngeneic with C26 cells. Two perpendicular diameters of the resultant tumors were measured daily using calipers. Tumor volumes were calculated as described previously (12).

Immunohistochemical Staining of the Tumor Tissue.

When tumors reached 1 cm in the longer diameter, they were resected, embedded in Tissue-Tek OCT embedding medium (Sakura Finetechnical) and stored at −80°C until use. Thin sections of the tumor tissues were prepared by cryostat and placed on glass slides. Sections were then fixed in 1% paraformaldehyde at room temperature for 30 min, washed three times with PBS, and incubated overnight with a 1:100 dilution of biotin-conjugated rat antimouse CD31 (platelet endothelial cell adhesion molecule-1; PharMingen) to detect the vascular endothelial cells. The bound antibody was coupled with streptavidin-peroxidase complex (Histofine; Nichirei Corporation) and visualized by 3,3′-diaminobenzidine tetrahydrochloride (DAB). The sections were then counterstained with methylgreen for 1 min and observed under a microscope. Four high-power fields (×400) from the tumor region were arbitrarily selected, and two pathologists (M. M. and M. T.) independently counted the number of the vessels stained.

Inhibitors.

Tranylcypromine, a PGIS inhibitor (13), was obtained from Aldrich. Tranylcypromine (0.7 mg/kg/day) was dissolved in water and administered to animals daily through gavage tubes.

Seratrodast, a TXA2 receptor inhibitor (14), was from Takeda Pharmaceutical, Japan. Seratrodast (3 mg/kg/day) was suspended in 5% arabic gum solution and administered daily through gavage tubes.

Statistical Analysis.

Significant differences in the means were examined by Student’s unpaired, two-tailed t test. Survival curves were analyzed by the method of Kaplan and Meier (15).

The retroviral constructs used in this study are schematically shown in Fig. 2,a. The constructs, pLNCX-TXAS, pLNCX-PGIS, and pLXIN, gave comparable numbers of colonies after the corresponding retroviruses were transduced into C26 cells. All of these colonies were collected and were grown as mass population cultures to reduce the artifacts caused by the differences in growth rates among colonies attributable to the position effects of random insertion of retroviral constructs. Both C26-TXAS, and C26-PGIS express high amounts of mRNAs derived from the transduced cDNAs (Fig. 2 b).

TXAS converts PGH2 to TXA2, and PGIS converts PGH2 to PGI2 (Fig. 1). Both TXA2 and PGI2 have extremely short half-lives and cannot easily be measured quantitatively. Instead, we measured the amounts of TXB2 and 6-keto PGF1α, stable compounds derived from TXA2 and PGI2, respectively. As expected, C26-TXAS and C26-PGIS produced significantly higher amounts of TXB2 and 6-keto PGF1α, respectively (Fig. 2,c). Thus, transduced TXAS and PGIS cDNAs both produced functioning enzymes in the cells. We next studied the effect of TXAS and PGIS expression on the cell growth in vitro. C26-TXAS and C26-PGIS did not show any significant differences in their growth rates (Fig. 2 d). These results indicate that exogenous expression of TXAS and PGIS did not affect the cell growth in vitro.

The cell growth rate in vitro does not predict the tumor growth rate in vivo. The latter is affected by host factors, such as migration of endothelial cells into tumors to generate blood vessels, involvement of fibroblasts to form tumor interstitium, or reactions of the immune cells against tumors. Mice carrying C26-TXAS and C26-PGIS transformants exhibited contrasting effects in tumor characteristics (Fig. 3, a–d). Tumors established from C26-TXAS grew more rapidly than C26-neo or C26 (wild type; P < 0.05 at days 7, 8, and 9), and resulted in the death of all of the mice at day 13 (P < 0.01). On the other hand, tumors from C26-PGIS grew more slowly (P < 0.05 after day 12), and resulted in longer survivals (P < 0.05) in these mice (Fig. 3, a and b). These results indicate that TXA2 and PGI2 exerted antithetical effects on tumor growth through their actions on the host cells.

To identify the TXAS or PGIS target(s) that modified the tumor growth, we conducted the immunohistological analyses on the established tumors. H&E staining or immunostaining using lineage-specific antibodies did not show any differences in the numbers or subsets of the invading immune cells (data not shown). Therefore, the immune system was probably not the main target. In contrast, the staining of these sections using vascular endothelial cell-specific antibody (antimurine CD31) revealed a marked difference in the density of the tumor vasculature. Tumors established from C26-TXAS had significantly richer vasculature (204%, versus C26-neo; P < 0.01), and tumors from C26-PGIS had much poorer vasculature (52%, versus C26-neo; P < 0.01), than tumors from C26-neo or C26 (wild type). The density of the tumor vasculature was correlated with the tumor growth rate: the denser the tumor vasculature, the faster the tumors grew (compare Fig. 3,a with Fig. 3, c and d). This indicates that TXA2 and PGI2 modified tumor growth through tumor angiogenesis. A recent study showed that COX-2 overexpression in colon cancer cells stimulated angiogenesis and that the stimulation was inhibited by NSAIDS or COX-2 inhibitors (16). TXA2, a downstream metabolite of PGH2, might be involved in the angiogenic activity observed in those studies (17, 18).

We next tested whether the effects of TXA2 and PGI2 on tumor angiogenesis and tumor growth could be reversed by the specific inhibitors. Administration of seratrodast, a TXA2 receptor inhibitor, reduced the vasculature and tumor growth in C26-TXAS-derived tumors. Administration of tranylcypromine, a PGIS inhibitor, increased the vasculature and tumor growth in C26-PGIS-derived tumors (Fig. 4, a and b). These results confirmed that the changes in the tumor growth and angiogenesis actually resulted from the changes in the prostanoid profile in the tumors.

In this study, our hypotheses were: (a) changes in the prostanoid profile would alter cancer growth; and (b) inhibitors of procancer prostanoid(s) would retard cancer growth and vice versa. Our results support both hypotheses. Accumulated knowledge about cancer has established the clonal expansion scheme in carcinogenesis: (a) gene mutations or alterations in gene expression patterns produce clones with a selective growth advantage, and these clones expand more rapidly than the other cells; (b) step a is repeated many times to cause a full-blown cancer to arise. The prostanoid profile of the cancer cells is likely to have been altered to benefit their growth. Our results indicate that TXA2 is a proangiogenic and procancer prostanoid, and PGI2 is an anticancer prostanoid for C26 cells. By examining other prostanoids in other cancer cells, the prostanoid profile shared by many cancers could be determined. In addition to TXA2 and PGI2 investigated in this study, PGE2 is of special interest, because increased PGE2 levels were reported in intestinal adenoma and colon cancer (8), and a recent study in which the EP2 gene (a PGE2 receptor) was disrupted in APCΔ716 mice showed that PGE2 is involved in tumor angiogenesis (19). Studies on prostanoid profiles will enable us to select candidate prostanoids that can be molecular targets for future cancer treatment.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

      
1

Supported in part by a grant for cancer research from The Sagawa Foundation for Promotion of Cancer Research and Grant-in-Aid 10770261 for scientific research from the Ministry of Education, Science, Sports, Culture and Technology of Japan.

            
3

The abbreviations used are: NSAID, nonsteroidal anti-inflammatory drug; COX, cyclooxygenase; TX, thromboxane; PG, prostaglandin; PGIS, PGI2 synthase; TXAS, TXA2 synthase; GAPDH, glyceraldehyde 3-phosphate dehydrogenase; EIA, enzyme immunoabsorbent assay; LTR, long terminal repeat.

Fig. 1.

Schematic presentation of the metabolic pathways of prostanoids. Phospholipase A2 (PLA2) converts phospholipid localized in cell membrane to arachinonic acid, which in turn is converted to PGH2 by COXs. PGH2 is immediately catalyzed to various prostanoids. In ellipses, enzymes. Black arrows, catalytic pathways. Opposite effects of TXA2 and PGI2 on platelet aggregation and vasoconstriction are shown: gray arrow, stimulation; gray T-bar, inhibition.

Fig. 1.

Schematic presentation of the metabolic pathways of prostanoids. Phospholipase A2 (PLA2) converts phospholipid localized in cell membrane to arachinonic acid, which in turn is converted to PGH2 by COXs. PGH2 is immediately catalyzed to various prostanoids. In ellipses, enzymes. Black arrows, catalytic pathways. Opposite effects of TXA2 and PGI2 on platelet aggregation and vasoconstriction are shown: gray arrow, stimulation; gray T-bar, inhibition.

Close modal
Fig. 2.

Transduction of TXAS gene and PGIS gene to C26 cells. a, schematic presentation of the retroviral constructs. The aminoglycoside phosphotransferase gene (neoR) that confers neomycin resistance to the cells is transcribed from the LTR of the Moloney murine leukemia virus (Mo-MuLV). TXAS or PGIS cDNAs are transcribed from an internal cytomegalovirus (CMV) immediate early gene promoter (pLNCX-TXAS; pLNCX-PGIS). Ψ+, Mo-MuLV retroviral packaging signal. b, Northern blots. Total RNAs from C26-TXAS or C26-PGIS as well as C26-neo and C26 (wild type) were subjected to Northern blot analyses to monitor the expression levels of TXAS (left panel) and PGIS (right panel). Expression of the GAPDH gene is shown to confirm equal loading of RNA. Arrow, position of each transcript. c, production of TXB2 and 6-keto PGF. Quantitative analyses of TXB2 (left panel) and 6-keto PGF (right panel) produced by C26-TXAS and C26-PGIS, respectively, were determined as described under “Materials and Methods.” As controls, prostanoids from C26-neo and C26 (wild type) cells under the same growth conditions were quantitatively determined. Data, the mean ± SD of four independent experiments. d, growth rate of the cell lines C26-TXAS, C26-PGIS, C26-neo, and C26 (wild type). Cell number of each clone was determined at the indicated growth time. Results were presented in a semilog plot. No significant differences were observed in the cell growth rates.

Fig. 2.

Transduction of TXAS gene and PGIS gene to C26 cells. a, schematic presentation of the retroviral constructs. The aminoglycoside phosphotransferase gene (neoR) that confers neomycin resistance to the cells is transcribed from the LTR of the Moloney murine leukemia virus (Mo-MuLV). TXAS or PGIS cDNAs are transcribed from an internal cytomegalovirus (CMV) immediate early gene promoter (pLNCX-TXAS; pLNCX-PGIS). Ψ+, Mo-MuLV retroviral packaging signal. b, Northern blots. Total RNAs from C26-TXAS or C26-PGIS as well as C26-neo and C26 (wild type) were subjected to Northern blot analyses to monitor the expression levels of TXAS (left panel) and PGIS (right panel). Expression of the GAPDH gene is shown to confirm equal loading of RNA. Arrow, position of each transcript. c, production of TXB2 and 6-keto PGF. Quantitative analyses of TXB2 (left panel) and 6-keto PGF (right panel) produced by C26-TXAS and C26-PGIS, respectively, were determined as described under “Materials and Methods.” As controls, prostanoids from C26-neo and C26 (wild type) cells under the same growth conditions were quantitatively determined. Data, the mean ± SD of four independent experiments. d, growth rate of the cell lines C26-TXAS, C26-PGIS, C26-neo, and C26 (wild type). Cell number of each clone was determined at the indicated growth time. Results were presented in a semilog plot. No significant differences were observed in the cell growth rates.

Close modal
Fig. 3.

Characterization of the tumors of BALB/c mice inoculated with C26-TXAS, C26-PGIS, C26-neo, and C26 (wild type). a, growth of the tumors established from the individual cell line. Tumor volumes are presented as the mean ± SE of five tumors. ∗, the statistical significance, analyzed using Student’s unpaired, two-tailed t test. b, survival rates of the BALB/c mice bearing tumors established from the individual cell line (n = 5). The statistical significance was analyzed by the method of Kaplan and Meier. c, immunohistochemical staining of the tumors established from individual cell line (×100). Brown stain, vascular endothelial cells. C26-TXAS tumor has denser and C26-PGIS tumor has sparser blood vessels than C26-neo and C26 (wild type) tumors. d, numerical comparison of the blood vessels formed in individual tumor. Blood vessels were visualized by immunohistochemical staining against antimouse CD31. Data, the mean ± SD of four tumors. The statistical significance was analyzed using Student’s unpaired, two-tailed t test.

Fig. 3.

Characterization of the tumors of BALB/c mice inoculated with C26-TXAS, C26-PGIS, C26-neo, and C26 (wild type). a, growth of the tumors established from the individual cell line. Tumor volumes are presented as the mean ± SE of five tumors. ∗, the statistical significance, analyzed using Student’s unpaired, two-tailed t test. b, survival rates of the BALB/c mice bearing tumors established from the individual cell line (n = 5). The statistical significance was analyzed by the method of Kaplan and Meier. c, immunohistochemical staining of the tumors established from individual cell line (×100). Brown stain, vascular endothelial cells. C26-TXAS tumor has denser and C26-PGIS tumor has sparser blood vessels than C26-neo and C26 (wild type) tumors. d, numerical comparison of the blood vessels formed in individual tumor. Blood vessels were visualized by immunohistochemical staining against antimouse CD31. Data, the mean ± SD of four tumors. The statistical significance was analyzed using Student’s unpaired, two-tailed t test.

Close modal
Fig. 4.

Effect of TXA2 receptor inhibitor and PGIS inhibitor on tumor growth and tumor blood vessel formation. a, C26-TXAS-inoculated mice were administered with seratrodast or vehicle. Tumor volumes (left panel) and numbers of blood vessel (right panel) were determined as described under “Materials and Methods.” Tumor volumes are presented as the mean ± SE of five tumors. ∗, statistical significance. b, C26-PGIS-inoculated mice were administered with tranylcypromine or vehicle. Tumor volumes (left panel) and numbers of blood vessel (right panel) were determined as described under “Materials and Methods.” Tumor volumes are presented as the mean ± SE of five tumors. ∗, statistical significance.

Fig. 4.

Effect of TXA2 receptor inhibitor and PGIS inhibitor on tumor growth and tumor blood vessel formation. a, C26-TXAS-inoculated mice were administered with seratrodast or vehicle. Tumor volumes (left panel) and numbers of blood vessel (right panel) were determined as described under “Materials and Methods.” Tumor volumes are presented as the mean ± SE of five tumors. ∗, statistical significance. b, C26-PGIS-inoculated mice were administered with tranylcypromine or vehicle. Tumor volumes (left panel) and numbers of blood vessel (right panel) were determined as described under “Materials and Methods.” Tumor volumes are presented as the mean ± SE of five tumors. ∗, statistical significance.

Close modal

We thank Dr. L-H. Wang (University of Texas Medical School, Houston, TX) for critical reading of the manuscript and for providing us with TXAS and PGIS cDNAs.

1
FitzGerald G. A., Loll P. COX in a crystal ball: current status and future promise of prostaglandin research.
J. Clin. Investig.
,
107
:
1335
-1337,  
2001
.
2
Prescott S. M., Fitzpatrick F. A. Cyclooxygenase-2 and carcinogenesis.
Biochim. Biophys. Acta
,
1470
:
M69
-M78,  
2000
.
3
Thun M. J., Namboodiri M. M., Calle E. E., Flanders W. D., Heath C. W., Jr. Aspirin use and risk of fatal cancer.
Cancer Res.
,
53
:
1322
-1327,  
1993
.
4
Williams C. S., Mann M., DuBois R. N. The role of cyclooxyneases in inflammation, cancer and development.
Oncogene
,
18
:
7908
-7916,  
1999
.
5
Fournier D. B., Gordon G. B. COX-2 and colon cancer: potential targets for chemoprevention.
J. Cell Biochem.
,
77
:
97
-102,  
2000
.
6
Steinbach G., Lynch P. M., Phillips R. K., Wallace M. H., Hawk E., Gordon G. B., Wakabayashi N., Saunders B., Shen Y., Fujimura T., Su L. K., Levin B. The effect of celecoxib, a cyclooxygenase-2 inhibitor, in familial adenomatous polyposis.
N. Engl. J. Med.
,
342
:
1946
-1952,  
2000
.
7
Zhang L., Yu J., Park B. H., Kinzler K. W., Vogelstein B. Role of BAX in the apoptotic response to anticancer agents.
Science (Wash. DC)
,
290
:
989
-992,  
2000
.
8
Yang V. W., Shields J. M., Hamilton S. R., Spannhake E. W., Hubbard W. C., Hylind L. M., Robinson C. R., Giardiello F. M. Size-dependent increase in prostanoid levels in adenomas of patients with familial adenomatous polyposis.
Cancer Res.
,
58
:
1750
-1753,  
1998
.
9
Barnes P. J., Belvisi M. G., Newton R., Mitchell J. A. Cyclooxygenase-2 expression in airway cells Szczeklik A. Gryglewski R. J. Vane J. R. eds. .
Eicosanoids, Aspirin, and Asthma
,
:
111
-128, Marcel Dekker New York  
1998
.
10
Ullrich V., Zou M. H., Bachschmid M. New physiological and pathophysiological aspects on the thromboxane A2-prostacyclin regulatory system.
Biochim. Biophys. Acta
,
1532
:
1
-14,  
2001
.
11
Danos O., Mulligan R. C. Safe and efficient generation of recombinant retroviruses with amphotropic and ecotropic host ranges.
Proc. Natl. Acad. Sci. USA
,
85
:
6460
-6464,  
1988
.
12
Tanaka M., Saijo Y., Sato G., Suzuki T., Tazawa R., Satoh K., Nukiwa T. Induction of antitumor immunity by combined immunogene therapy using IL-2 and IL-12 in low antigenic Lewis lung carcinoma.
Cancer Gene Ther.
,
7
:
1481
-1490,  
2000
.
13
Bicknell R., Vallee B. L. Angiogenin stimulates endothelial cell prostacyclin secretion by activation of phospholipase A2.
Proc. Natl. Acad. Sci. USA
,
86
:
1573
-1577,  
1989
.
14
Fujimura M., Abo M., Kamio Y., Myou S., Ishiura Y., Hashimoto T., Matsuda T. Effect of leukotriene and thromboxane antagonist on propranolol-induced bronchoconstriction.
Am. J. Respir. Crit. Care Med.
,
160
:
2100
-2103,  
1999
.
15
Kaplan E. L., Meier P. Nonparametric estimation from incomplete observations.
J. Am. Stat. Assoc.
,
53
:
457
-481,  
1958
.
16
Tsujii M., Kawano S., Tsuji S., Sawaoka H., Hori M., DuBois R. N. Cyclooxygenase regulates angiogenesis induced by colon cancer cells.
Cell
,
93
:
705
-716,  
1998
.
17
Nie D., Lamberti M., Zacharek A., Li L., Szekeres K., Tang K., Chen Y., Honn K. V. Thromboxane A2 regulation of endothelial cell migration, angiogenesis, and tumor metastasis.
Biochem. Biophys. Res. Commun.
,
267
:
245
-251,  
2000
.
18
Daniel T. O., Liu H., Morrow J. D., Crews B. C., Marnett L. J. Thromboxane A2 is a mediator of cyclooxygenase-2-dependent endothelial migration and angiogenesis.
Cancer Res.
,
59
:
4574
-4577,  
1999
.
19
Sonoshita M., Takaku K., Sasaki N., Sugimoto Y., Ushikubi F., Narumiya S., Oshima M., Taketo M. M. Acceleration of intestinal polyposis through prostaglandin receptor EP2 in ApcΔ716 knockout mice.
Nat. Med.
,
7
:
1048
-1051,  
2001
.