282 Thymidine Phosphorylase in Human Esophageal Squamous Cell Carcinoma Yuji Takebayashi, M.D.1 Shoji Natsugoe, M.D.1 Masamichi Baba, M.D.1 Suminori Akiba, M.D.2 Toshitaka Fukumoto, M.D.1 Kazutaka Miyadera, Ph.D.3 Yuji Yamada, Ph.D.3 Sonshin Takao, M.D.1 Shin-ichi Akiyama, M.D.4 Takashi Aikou, M.D.1 1 First Department of Surgery, Faculty of Medicine, Kagoshima University, Kagoshima, Japan. 2 Department of Public Health, Faculty of Medicine, Kagoshima University, Kagoshima, Japan. 3 Taiho Pharmaceutical Company Ltd., Saitama, Japan. 4 Department of Cancer Chemotherapy, Institute for Cancer Research, Faculty of Medicine, Kagoshima University, Kagoshima, Japan. BACKGROUND. Experimental evidence has shown that thymidine phosphorylase (dThdPase) is identical to platelet-derived endothelial cell growth factor (PDECGF) and has angiogenic activity. The enzymatic activity of dThdPase was needed for the angiogenesis by the enzyme. These observations were catalysts for the current study. METHODS. The authors examined retrospectively the expression of the angiogenic factor dThdPase in 163 primary esophageal squamous cell carcinomas and its association with angiogenesis and clinicopathologic findings. To determine whether dThdPase expression was a prognostic factor after adjustment for the established prognostic factors and microvessel count, the authors conducted a survival analysis using the Cox proportional hazards model. RESULTS. dThdPase was expressed significantly more frequently (P ⬍ 0.001) in esophageal carcinomas (83 of 163, 50.9%) than in adjacent nonneoplastic esophageal tissue samples (20 of 163, 12.3%). Microvessel counts were significantly higher (P ⬍ 0.001) in dThdPase positive carcinomas (18.3 ⫾ 6.2) than in dThdPase negative carcinomas (8.2 ⫾ 7.5). Significant correlations were observed between dThdPase expression and numerous clinicopathologic findings, including pT, pN, pM categories; lymphatic invasion; venous invasion; and residual tumors. Prognostic variables studied using a Cox hazard regression model confirmed that dThdPase expression was an independent prognostic factor in esophageal squamous cell carcinoma, although pN category was the best predictor of patient survival. CONCLUSIONS. This study indicated that in esophageal squamous cell carcinoma, dThdPase expression is associated with angiogenesis and is an unfavorable prognostic factor. These findings implied that the inhibition of dThdPase would improve the prognoses of some patients with dThdPase positive esophageal tumors. Cancer 1999;85:282–9. © 1999 American Cancer Society. KEYWORDS: thymidine phosphorylase, angiogenesis, esophageal carcinoma, metastasis. T Address for reprints: Shin-ichi Akiyama, M.D., Department of Cancer Chemotherapy, Institute for Cancer Research, Faculty of Medicine, Kagoshima University, Sakuragaoka 8-35-1, Kagoshima 890, Japan. Received March 9, 1998; revision received July 1, 1998; accepted July 2, 1998. © 1999 American Cancer Society hymidine phosphorylase (dThdPase; EC 126.96.36.199) catalyzes the reversible phosphorolysis of thymidine, deoxyuridine, and their analogs to their respective bases and 2-deoxyribose-1-phosphate.1–3 dThdPase also catalyzes the transfer of deoxyribose from one deoxynucleoside to another base to form a second deoxynucleoside.4 – 6 In mammals, dThdPase consists of two identical subunits with a molecular weight of 55 kilodaltons (kD).7 dThdPase is identical to platelet-derived endothelial cell growth factor (PD-ECGF).8,9 PDECGF stimulates chemotaxis and [3H] thymidine incorporation by endothelial cells in vitro and has angiogenic activity in vivo.10 –12 Recently, we demonstrated that the enzymatic activity of dThdPase is indispensable for its angiogenic activity.12,13 Among the degradation Thymidine Phosphorylase in Esophageal Carcinoma/Takebayashi et al. products of thymidine by dThdPase, 2-deoxy-D-ribose, a dephosphorylated product derived from 2-deoxy-D-ribose-1-phosphate, had chemotactic activity in vitro and angiogenic activity in vivo. These findings suggested that the enzymatic products may stimulate the chemotaxis of endothelial and possibly other cells, causing angiogenesis. Transfection of PD-ECGF/dThdPase into transformed fibroblasts in nude mice resulted in increased tumor vascularization.11 Overexpression of dThdPase in MCF-7 cells transfected with dThdPase cDNA conferred a growth advantage to these cells when xenografted into nude mice, but it did not affect their growth in vitro, suggesting angiogenic advantage of the dThdPase-expressing cells.14 Increased dThdPase expression in human breast carcinoma cells was shown to correlate with microvessel density in the tumor.15 Areas of high blood velocity in ovarian tumors were associated with increased expression of PD-ECGF/dThdPase.16 These observations suggest that dThdPase actually is involved in angiogenesis in some human tumors. Compared with adjacent normal tissues, higher levels of dThdPase were observed in a variety of malignant tumors.15–25 Experimental evidence has shown that tumor growth is dependent on angiogenesis.26,27 When tumors reach a size of a few millimeters, new capillaries penetrate, allowing for rapid tumor growth. These new vessels facilitate the entry of tumor cells into the vasculature and their subsequent metastasis, so that angiogenesis correlates with the probability of metastases.28 –30 Angiogenesis in human solid tumors is a risk factor for metastasis and recurrence.31–38 The aims of this study were to examine the expression of dThdPase in esophageal squamous cell carcinomas and the correlation between dThdPase expression and angiogenesis, clinicopathologic findings, or clinical outcome. 283 TABLE 1 Relation between Thymidne Phosphorylase Expression and Clinicopathologic Variables in 163 Patients with Esophageal Carcinoma dThdPase evaluation (%) Variables Sex Female Male Histologic type Well Moderate Poor Residual tumor R0 R1 pT category pT1 pT2 pT3 pT4 pN category N0 N1 pM category M0 M1 Lymphatic invasion Negative Positive Venous invasion Negative Positive Stage I II III IV Negative (n ⴝ 80) Positive (n ⴝ 83) 8 (66.7) 72 (47.7) 4 (33.3) 79 (52.3) 33 (45.8) 36 (54.5) 11 (44.0) 39 (54.2) 30 (45.5) 14 (56.0) 69 (54.3) 11 (30.6) 58 (45.7) 25 (69.4) 35 (83.3) 7 (33.3) 30 (40.0) 8 (32.0) 7 (16.7) 14 (66.7) 45 (60.0) 17 (68.0) 49 (72.1) 31 (32.6) 19 (27.9) 64 (67.4) 64 (56.6) 16 (32.0) 49 (43.4) 34 (68.0) 45 (68.2) 35 (36.1) 21 (31.8) 62 (63.9) 68 (54.4) 12 (31.6) 57 (45.6) 26 (68.4) 29 (77.7) 24 (55.8) 12 (31.6) 15 (30.6) 4 (12.3) 19 (44.2) 26 (68.4) 34 (69.4) P value 0.241 0.394 0.019 ⬍0.001 ⬍0.001 0.006 ⬍0.001 0.023 ⬍0.001 dThdPase: thymidine phosphorylase. MATERIALS AND METHODS Patients and Tumors The characteristic features of the 163 patients with esophageal carcinoma investigated in this study are summarized in Table 1. We obtained informed consent from all the patients in this study. All patients who underwent surgical resection in the First Department of Surgery of Kagoshima University Hospital between January 1987 and December 1991 were evaluated. None had received prior chemotherapy or radiation therapy. Patients received postoperative therapy; however, there was no difference in outcome among the various treatment modalities. Of the 163 patients, 151 were males, and 12 were females; the average age was 63.5 years, ranging from 45 to 82 years. Histologic typing and grading of tumors was performed according to the criteria established by the World Health Organization.39 pT category, pN cate- gory, and pM category were determined according to TNM classification.40 The M category was determined from clinical data, including intraoperative findings, chest and bone radiography, ultrasonography, computed tomography, and laboratory tests reflecting bone and liver metastasis. Tumor specimens were collected after obtaining informed consent in accordance with institutional guidelines. The deepest invading sites were selected for immunohistochemistry of factor VIII and dThdPase. Immunohistochemical Detection and Evaluation of dThdPase and Factor VIII The sections were deparaffinized with xylene and dehydrated with 98% ethanol. Endogenous peroxidase was blocked by immersing the slides in 0.3% hydrogen 284 CANCER January 15, 1999 / Volume 85 / Number 2 peroxide in absolute methanol for 20 minutes at room temperature. After washing three times with phosphate-buffered saline (PBS) for 5 minutes, the sections were blocked by soaking in PBS containing 1% bovine serum albumin for 20 minutes at room temperature. The blocked sections were incubated overnight at 4°C with 0.5 mg/mL monoclonal antibody against dThdPase, which has been described previously.22,23,25 This antibody was specific for detecting dThdPase in human esophageal carcinoma tissues.25 After 12 hours, the slides were incubated for 30 minutes with biotinylated antimouse immunoglobulin (Ig) G diluted 100fold with PBS at room temperature, washed three times in PBS for 15 minutes, and incubated with avidin-biotin-peroxidase complex diluted with PBS for 30 minutes.40 After three washes with PBS for 15 minutes, the sections were incubated with 0.5 mg/mL diaminobenzidine and 0.03% (volume/volume) H2O2 in PBS for 7 minutes and finally counterstained with hematoxylin prior to mounting. Serial sections also were incubated with rabbit antiserum against human von Willebrand factor (DAKO Corporation, Santa Barbara, CA) diluted 1:200 with PBS containing 5% goat serum. Antibody binding was detected by sequential incubation with biotinylated goat antirabbit serum. Other procedures were the same as those described for dThdPase immunohistochemical staining. Representative sections were made from nonneoplastic tissues and carcinoma tissues, including the most invasive portions of carcinoma cells. For microscopic analysis, we examined the whole field of each section whether or not it was stained with the monoclonal antibody against dThdPase. Tumor samples were considered to be dThdPase positive when more than 5% of the carcinoma cells were stained, because less than 5% of cells were stained in 85.9% of normal tissues (Figs. 1, 2A,B). For microvessel counts, we screened areas of intense neovascularization at low power magnification (⫻40 and ⫻100), then areas with the highest number of factor VIII positive microvessels were examined at higher power magnification (⫻400) to obtain accurate microvessel counts (Fig. 2C,D). Both the evaluation of dThdPase expression and the microvessel count were assessed by two investigators (Y.T. and S.N.) without their knowledge of the clinicopathologic findings. Statistical Analysis Demographic and clinicopathologic characteristics were compared between patients with dThdPase positive and negative tumors by using the chi-square test or Student‘s t test.41 The Cox proportional hazards model was used for the multivariate survival analy- FIGURE 1. Percent of cells expressing thymidine phosphorylase (dThdPase) in normal esophageal tissues and in esophageal carcinomas obtained from 163 patients on study. sis.42 Maximum likelihood parameter estimates and likelihood ratio statistics (LRS) in Cox proportional hazards models were obtained by using a statistical package, EPICURE (HiroSoft International Corporation, Seattle, WA).43 We calculated Wald-type confidence intervals. Tests for statistical interaction were conducted by including a cross-product term of the two variables of interest in a model. All P values presented are two-sided. RESULTS Expression of dThdPase in Normal Mucosa or Esophageal Carcinoma Most normal esophageal mucosal cells were not stained with the anti-dThdPase antibody (Fig. 2A). When nonneoplastic esophageal squamous cell epithelium was reacted with this antibody, positive dThdPase immunostaining usually was not detected, but it was detected sometimes in the basal cells. In contrast, the cytoplasm of many esophageal carcinoma cells was strongly stained (Fig. 2B). The percentage of cells expressing dThdPase was less than 5% in 85.9% of nonneoplastic tissues. The positivity of dThdPase in esophageal carcinomas (83 of 163, 50.9%) was significantly higher (P ⬍ 0.001) than that of dThdPase in normal tissues (23 of 163, 14.1%). Thymidine Phosphorylase in Esophageal Carcinoma/Takebayashi et al. 285 FIGURE 2. Immunostaining for dThdPase in normal esophageal mucosal cells (A) and in esophageal carcinoma cells (B). Microvessel staining with antifactor VIII antibody on a ⫻100 field (C) and on a ⫻400 field (D). ⫻400 field, 84.6% of samples were dThdPase positive. Figure 4 shows that the mean microvessel count in dThdPase positive esophageal carcinomas (18.3 ⫾ 6.2) was significantly higher (P ⬍ 0.001) than that in dThdPase negative tumors (8.2 ⫾ 7.5). Correlationship between Clinicopathologic Findings and dThdPase Expression FIGURE 3. Microvessel count in a ⫻400 field as a function of the percent of cells expressing dThdPase. Overlap accounts for less than 166 points appearing on the graph. Correlation between dThdPase Expression and Microvessel Count Figure 3 shows that the percentage of dThdPase positive cells increased with increasing microvessel count. In tumors with microvessel counts of more than 20 per Figure 4 and Table 1 summarize the relation between clinicopathologic features and dThdPase expression in esophageal carcinomas. Although no significant correlation was found between dThdPase expression and age (Fig. 4A), sex, and histologic subtype (Table 1), significant correlations with other clinicopathologic findings were observed. Table 1 demonstrates that dThdPase positive tumors invaded deeper and had a higher frequency of lymph node metastasis than negative tumors (P ⬍ 0.001 and P ⬍ 0.01, respectively). Expression of dThdPase was significantly higher in carcinomas with lymphatic invasion (P ⬍ 0.001) and venous invasion (P ⬍ 0.001), and patients with dThd- 286 CANCER January 15, 1999 / Volume 85 / Number 2 FIGURE 4. Distribution of age, tumor size, and microvessel count as a function of dThdPase expression. In both graphs, the box corresponds to the interquartile ranges, with the lower boundary of the box representing the 25th percentile and the upper boundary representing the 75th percentile. The dotted line inside the box represents the median value. The vertical lines represent the 5th and 95th percentiles, and the open circles represent the outliers. P values were calculated by using Student’s t test. Pase positive carcinomas had more advanced stages than those with negative tumors (P ⬍ 0.001). Prognostic Relevance and Proportional Hazards Analysis To determine independent prognostic values on patient survival, a Cox regression model was constructed using dThdPase positivity, microvessel count, and several clinicopathologic findings. The analysis using model 1, which included microvessel count but not dThdPase positivity, showed that pT category, pN category, pM category, and residual tumor were independent prognostic factors. Microvessel count was not an independent prognostic factor (hazard ratio, 1.03; 95% confidence interval [CI], 1.00 –1.06; P ⫽ 0.084). The analysis using model 2, which included dThdPase positivity in addition to all of the covariables in model 1, showed that dThdPase was an independent prognostic factor (hazard ratio, 2.07; 95% CI, 1.04 – 4.10; P ⫽ 0.033), although pN category was the best predictor of survival for patients with esophageal carcinomas (hazard ratio, 3.60; 95% CI, 1.82– 6.97; P ⬍ 0.001). DISCUSSION We found complete sequence identity between 120 amino acids of human dThdPase and the sequence of PD-ECGF, and we also demonstrated that rPDECGF has dThdPase activity.8,9 These observations and similar reports from other laboratories suggest that human dThdPase is identical to PD-ECGF.44 – 46 PD-ECGF reportedly stimulated chemotaxis of endothelial cells in vitro and angiogenesis in vivo.10 In accordance with this, we have demonstrated that dThdPase has angiogenic activity and that its enzymatic activity is needed for angiogenesis.10 –13 In human tissues, several investigators reported the correlation between dThdPase expression and angiogenesis. In ovarian tumors, areas of increased blood flows as a measure of angiogenesis were associated with elevated dThdPase expression.16 Expression of thrombomodulin, a marker protein for an endothelial cell, was significantly correlated with dThdPase activity in colorectal carcinomas, suggesting that this enzyme may be important in angiogenesis in human colorectal carcinomas.20 Similar findings were reported in breast,15 gastric,23,24 and colorectal carcinomas.25 In this study, the correlation between dThdPase expression and microvessel count as an indicator of angiogenesis also was observed in esophageal squamous cell carcinoma. Retrospective study demonstrated significant correlations between dThdPase expression and sev- Thymidine Phosphorylase in Esophageal Carcinoma/Takebayashi et al. 287 TABLE 2 Results from Cox Proportional Hazard Regression Analysis for Survivals of 163 Patients with Esophageal Carcinomas Model 1 Model 2 Variables (category)a Hazard ratio 95% CI P value Hazard ratio 95% CI P value pT category (T1, T2 ⫽ 0; T3, T4 ⫽ 1 pN category (N0 ⫽ 0, N1 ⫽ 1 pM category (M0 ⫽ 0, M1 ⫽ 1 Residual tumor (R0 ⫽ 0, R1 ⫽ 1 Microvessel count (no.) dThdPase positivity Negative ⫽ 0, positive ⫽ 1 1.00 1.02 1.00 4.04 1.00 1.70 1.00 1.95 1.03 Reference 1.18–2.30 Reference 2.10–7.80 Reference 1.17–2.47 Reference 1.06–3.60 1.00–1.06 0.002 1.00 1.65 1.00 3.65 1.00 1.60 1.00 1.86 1.04 1.00 2.07 Reference 1.18–2.30 Reference 1.82–6.97 Reference 1.10–2.31 Reference 1.01–3.42 0.97–1.05 Reference 1.04–4.10 0.002 ⬍0.001 0.006 0.035 0.084 ⬍0.001 0.015 0.049 0.682 0.033 dThdPase: thymidine phosphorylase; CI: confidence interval. a The following codes were used: 0, negative; 1, positive, with the following hazard function at N time: W(t) ⫽ h0(t)exp(¥␤iXi), where h0 was a base line hazard rate, and X1 . . . Xn represented covariable measurements. i␤1 . . . ␤N were regression parameters to be estimated. In model 1, covariables were age, sex, histologic type, tumor location, lymphatic invasion, venous invasion, pT category, pN category, pM category, residual tumor, and microvessel count. Model 2 included dThdPase positivity in addition to the covariables in model 1. eral clinicopathologic findings as well as clinical outcome in primary esophageal carcinomas. The proportion of dThdPase positive cells was higher in carcinomas with deep invasion and/or lymph node metastases. Angiogenesis is needed for rapid tumor growth, and the vascularized tumor extends vertically into the deep tissues beneath the basement membrane. The new proliferating capillaries disrupt the basement membrane and are more penetrable by tumor cells than mature vessels. During the vascular phase, tumor cells may be shed into the circulation.28 The clinical significance of these findings has been documented in studies of cutaneous melanoma31 and of breast,32,33 prostate,34 lung,35,36 gastric,37 and colorectal39 carcinomas in which microvessel density has been shown to be correlated significantly with the occurrence of metastases and recurrence of the diseases. With this observation as the model, the current study suggests that increased dThdPase in esophageal carcinomas may enhance invasiveness and the ability to metastasize of the tumors by its angiogenic properties. dThdPase, but not microvessel count, proved to be a significant prognostic factor according to multivariate analysis with the Cox hazard regression model (Table 2). We previously reported that only Dukes stage and dThdPase expression, but not microvessel count, were statistically significant prognostic factors for mortality of patients with colorectal carcinoma.25 Furthermore, dThdPase expression was not correlated with microvessel counts in breast tumors and MCF-7 cells.14,48 Recently, O’Brien et al. reported that expression of PD-ECGF/dThdPase in invasive bladder carcinomas was 33-fold higher than in superficial bladder carcinomas and that expression of vascular endothelial cell growth factor in superficial bladder carcinomas was higher than in invasive bladder carcinomas.21 These findings suggest that dThdPase has capabilities that are concerned with the progression of carcinomas other than angiogenic activity. dThdPase expression in esophageal squamous cell carcinomas was higher than that in adjacent nonneoplastic tissues. dThdPase expression in esophageal squamous cell carcinomas also was higher than that in gastric and colorectal adenocarcinomas in the current study.24,25 This may explain in part the reason why patients with esophageal squamous cell carcinoma showed poorer prognosis than those with adenocarcinomas of the stomach or colorectum. This study showed that dThdPase expression in esophageal carcinoma was correlated significantly with several clinicopathologic findings and was an independent prognostic factor in a multivariate Cox proportional hazards regression model. A dThdPase inhibitor, 6 amino-5-chrolouracil, inhibited the angiogenesis by dThdPase in a CAM assay or a mouse gelatin sponge implanted assay.12,13 Although further studies are needed to clarify the mechanism of dThdPase-mediated angiogenesis, invasiveness, and its ability to metastasize, our current and previous studies10,13 suggest that inhibitors of this enzyme and prodrugs that are activated by this enzyme may suppress the growth of dThdPase expressing tumors and may be valuable in the therapy of patients with esophageal carcinoma. 288 CANCER January 15, 1999 / Volume 85 / Number 2 REFERENCES 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. Iltzssch MH, Kouni MH, Cha, S. Kinetic studies of thymidine phosphorylase from mouse liver. Biochemistry 1985;24: 6799 – 807. Friedkin D, Roberts D. The enzymatic synthesis of nucleosides. J Biol Chem 1954;207:245–56. Krenistsky TA, Koszalka GW, Tuttle JV. Purine nucleoside synthesis, an efficient method employing nucleoside phosphorylase. Biochemistry 1981;20:3615–21. Zimmerman M, Seidenberg J. Deoxyribosil transfer. J Biol Chem 1964;230:2618 –21. Gallo RC, Perry S, Breitman TR. The enzymatic mechanisms for deoxythymidine synthesis in human leukocytes. J Biol Chem 1967;242:5059 – 68. Krenitsky TA. Pentosyl transfer mechanisms of the mammalian nucleoside phosphorylases. J Biol Chem 1968;243: 2871–5. Desgranges C, Razaka G, Rabaud H. Catabolism of thymidine in human blood platelets: purification and properties of thymidine phosphorylase. Biochim Biophys Acta 1981; 654:211– 8. Furukawa T, Yosimura A, Sumizawa T, Haraguchi M, Akiyama S, Fukui K, Ishizawa M, Yamada Y. Angiogenic factor. Nature 1992;356:668. Sumizawa T, Furukawa T, Haraguchi M, Akiyama S, Fukui K, Ishizawa M, Akiyama S. Thymidine phosphorylase activity associated with platelet-derived endothelial cell growth factor. J Biochem (Tokyo) 1993;114:9 –14. Haraguchi M, Miyadera K, Uemura K, Sumizawa T, Furukawa T, Yamada K, Akiyama S, Yamada Y. Angiogenic activity of enzymes. Nature1994;368:198. Ishikawa F, Miyazono K, Hellman U, Drexler M, Wernstedt C, Hagiwara K, Usuki K, Takaku F, Risau W, Heldin CH. Identification of angiogenic activity and the cloning and expression of platelet-derived endothelial cell growth factor. Nature 1989;338:556 – 62. Miyazono K, Okabe T, Urabe A, Takaku F, Heldin CH. Purification and properties of an endothelial cell growth factor from human platelets. J Biol Chem 1987;262:4098 – 103. Miyadera K, Sumizawa T, Haraguchi M, Yoshida H, Konstanty W, Yamada Y, Akiyama S. Role of thymidine phosphorylase activity in the angiogenic effect of platelet-derived endothelial cell growth factor/thymidine phosphorylase. Cancer Res 1995;55:1687–90. Moghaddam A, Zhang HT, Fan TPD, Hu DE, Lees VC, Turley H, Fox SB, Gatter KC, Harris AL, Bickneel R. Thymidine phosphorylase is angiogenic and promotes tumor growth. Proc Natl Acad Sci USA 1995;92:998 –1002. Toi M, Hoshina S, Taniguchi T, Yamamoto Y, Ishituka H, Tominaga T. Expression of platelet-derived endothelial cell growth factor/thymidine phosphorylase in human breast cancer. Int J Cancer 1995;14:79 – 82. Reynold K, Farzaneh F, Collins WP, Harris AL. Association of ovarian malignancy with expression of platelet-derived endothelial cell growth factor. J Natl Cancer Inst 1994;86:1234 – 8. Pauly JL, Schuller MG, Zelcer AA, Kris TA, Gore SS. Identification and comparative analysis of thymidine phosphorylase in the plasma of healthy subjects and cancer patients: brief communication. J Natl Cancer Inst 1997;58:1587–90. Kono A, Hara Y, Sugata S. Substrate specificity of a thymidine phosphorylase in human liver tumor. Chem Pharmacol Bull 1984;32:1919 –21. 19. Yoshimura A, Kuwazuru Y, Furukawa T, Yoshida H, Yamada K, Akiyama S. Purification and tissue distribution of human thymidine phosphorylase; high expression in lymphocytes, reticulocytes and tumors. Biochim Biophys Acta 1990;1034: 107–13. 20. Takebayashi Y, Yamada K, Maruyama I, Fujii R, Akiyama S, Aikou T. The expression of thymidine phosphorylase and thrombomodulin in human colorectal carcinomas. Cancer Lett 1995;92:1–7. 21. O’Brien T, Cranston D, Fuggle S, Bicknell R, Harris AL. Different angiogenic pathways characterize superficial and invasive bladder cancer. Cancer Res 1995;55:510 –3. 22. Takebayashi Y, Yamada K, Miyadera K, Sumizawa T, Furukawa T, Kinoshita F, Aoki D, Okumura H, Yamada Y, Akiyama S, Aikou T. The activity and expression of thymidine phosphorylase in human solid tumors. Eur J Cancer 1996;32:1227–32. 23. Takebayashi Y, Miyadera K, Akiyama S, Hokita S, Yamada Y, Akiba S, Yamada Y, Sumizawa T, Aikou T. Expression of thymidine phosphorylase in human gastric carcinoma. Jpn J Cancer Res 1996;87:288 –95. 24. Maeda K, Chung YS, Ogawa Y, Takatsuka S, Kang SM, Ogawa M, Sawada T, Onda N, Kato Y, Sowa M. Thymidine phosphorylase/platelet-derived endothelial cell growth factor expression associated with hepatic metastasis in gastric carcinoma. Br J Cancer 1996;73:884 – 8. 25. Takebayashi Y, Akiyama S, Akiba S, Yamada K, Miyadera K, Sumizawa T, Yamada Y, Murata F, Aikou T. Clinicopathologic and prognostic significance of an angiogenic factor, thymidine phosphorylase, in human colorectal carcinoma. J Natl Cancer Inst 1996;88:1110 –7. 26. Folkman J, Klagsbrun M. Angiogenic factor. Science 1987; 235:442–7. 27. Folkman J. What is the evidence that tumors are angiogenesis dependent? J Natl Cancer Inst 1990;82:4 – 6. 28. Liotta L, Kleinerman J, Saidel G. Quantitative relationships of intravascular tumor cells, tumor vessels, and pulmonary metastases following tumor implantation. Cancer Res 1974; 34:997–1004. 29. Srivastava A, Laidler P, Hughes LE, Woodcock J, Shedden EJ. Neovascularization in human cutaneous melanoma; a quantitative morphological and Doppler ultrasound study. Eur J Cancer Clin Oncol 1986;22:1205–9. 30. Srivastava A, Laidler P, Davis RP, Horgan K, Hughes LE. The prognostic significance of tumor vascularity in intermediate-thickness (0.76 – 4.0 mm thick) skin melanoma: a quantitative histologic study. Am J Pathol 1988;133:419 –23. 31. Herlyn M, Clark WH, Rodeck U, Mancianti ML, Jambrosic I, Koprowski H. Biology of tumor progression in human melanocytes. Lab Invest 1987;56:461–74. 32. Weidner N, Semple JP, Welch WR, Folkman J. Tumor angiogenesis and metastasis— correlation in invasive breast carcinoma. N Engl J Med 1991;324:1– 8. 33. Horak ER, Leek R, Klenk N, Lejeune S, Smith K, Stuart N, Harris AL. Angiogeneis, assessed by platelet-derived cell adhesion molecular antibodies, as indicator of node metastases and survival in breast cancer. Lancet 1992;340:1120 –14. 34. Weidner N, Carroll PR, Flax J, Blumenfeld W, Folkman J. Tumor angiogenesis correlates with metastasis in invasive prostate carcinoma. Am J Pathol 1993;143:401–9. 35. Macchiarini P, Fontanini G, Hardin MJ, Squartini F, Angeletti CA. Relation of neovascularization to metastasis of non-small-cell lung cancer. Lancet 1992;340:145– 6. Thymidine Phosphorylase in Esophageal Carcinoma/Takebayashi et al. 36. Yamazaki K, Abe S, Takekawa H, Sudou N, Watanabe N, Ogura S. Tumor angiogenesis in human lung adenocarcinoma. Cancer 1995;74:2245–50. 37. Maeda K, Chung YS, Takatsuka S, Ogawa Y, Sawada T, Kondo Y, Sowa M. Tumor angiogenesis as a predictor of recurrence in gastric carcinoma. J Clin Oncol 1995;13:477– 81. 38. Takebayashi Y, Akiyama S, Yamada K, Akiba S, Aikou T. Angiogenesis as an unfavorable prognostic factor in human colorectal carcinoma. Cancer 1996;78:226 –31. 39. World Health Organization. International histologic classification of tumors: histological typing of esophageal and gastric tumors. 2ne ed. Berlin: Spring Verlag, 1990. 40. Hermanek P, Sobin LH, editors. UICC TNM classification of malignant tumors. 4th ed, 2nd rev. Berlin: Spring-Verlag, 1992. 41. Hsu SM, Raine L, Fanger AH. A comparative study of the peroxidase method and an avidin-biotin complex method for studying polypeptide hormones with radioimmunoassay antibody. Am J Clin Pathol 1981;75:734 – 8. 42. Breslow NE, Day NE. Statistical methods in cancer research. 1987. 43. Zar JH. Biostatistical analysis. 2nd ed. Totowa, NJ: Prentice Hall Press, 1996. 289 44. Preston DL, Lubin JH, Pierce DA. EPICURE: risk regression and data analysis software. Seattle: HiroSoft International Corporation, 1990. 27. Kaplan EL, Meier P. Nonparametric estimation from incomplete observation. J Am Stat Assoc 1997;16:95–101. 45. Usuki K, Saras J, Waltenberger J, Miyazono K, Pierce G, Thomason A, Heldin CH. Platelet-derived endothelial cell growth factor has thymidine phosphorylase activity. Biochem Biophys Res Commun 1992;84:1311– 4. 46. Moghaddam A, Bicknell R. Expression of platelet-derived endothelial cell growth factor in Escherichia coli and confirmation of its thymidine phosphorylase activity. Biochemistry 1992;31:12141– 6. 47. Nishimoto K, Miyadera K, Takebayashi Y, Fukuda K, Haraguchi M, Furukawa T, Yamada, Y, Akiyama S. Thymidine phosphorylase activity required for tumor angiogenesis and growth. Oncol Rep 1997;4:55– 8. 48. Fox SB, Westwood M, Moghaddam A, Comely M, Turley H, Whitehouse RM, Bicknell R, Gatter KC, Harris AL. The angiogenic factor platelet-derived endothelial cell growth factor/thymidine phosphorylase is up-regulated in breast cancer epithelium. Br J Cancer 1996;73:275– 80.