Int. J. Cancer: 97, 770 –774 (2002) © 2002 Wiley-Liss, Inc. DOI 10.1002/ijc.10144
Publication of the International Union Against Cancer
ROLE OF STROMAL COLLAGEN IN IMMUNOMODULATION AND PROGNOSIS OF ADVANCED GASTRIC CARCINOMA Satoshi OHNO*, Mitsuo TACHIBANA, Toshiyuki FUJII, Shuhei UEDA, Hirofumi KUBOTA and Naofumi NAGASUE Second Department of Surgery, Shimane Medical University, Shimane, Japan Although several hypotheses have been proposed explaining the mechanisms of the immune-privileged status of malignant tumors, the exact pathway is yet to be explored. Tumor stroma plays a vital role in the prognosis of cancer patients; however, the immunomodulatory impact of gastric cancer stroma has not been reported. We have evaluated the amount of stromal collagen and its impact on the infiltration of immune-competent cells into the tumor cell nest in gastric carcinoma. Tissue specimens from 84 advanced gastric carcinoma patients who had undergone a curative resection were evaluated for host immune status (CD8ⴙ T cells), tumor stromal reaction (AZAN staining), tumor Fas ligand expression and incidence of tumor cell apoptosis (by TUNEL). The number of apoptotic tumor cells (apoptotic index [AI]) increased proportionally with an increase in the number of CD8ⴙ T cells within the cancer cell nest (nest CD8) (p ⴝ 0.0001). Nest CD8 was inversely correlated with the amount of stromal collagen (p < 0.0001). Nest CD8 and AI became independent predictors of patient survival (p ⴝ 0.0023 and p ⴝ 0.044, respectively) in Cox’s multivariate analysis. The amount of stromal collagen was found to be a significant predictor of disease relapse in univariate analysis (p ⴝ 0.0010) but not in multivariate analysis (p ⴝ 0.4729). In conclusion, increased nest CD8 produced a survival advantage by inducing tumor cell apoptosis in gastric carcinoma patients. Increased tumor stromal collagen worked as a barrier for CD8ⴙ T-cell infiltration and might be one of the mechanisms of tumor escape from the host immune attack. © 2002 Wiley-Liss, Inc. Key words: gastric cancer; stromal collagen; immunomodulation; CD8⫹ T cell; apoptosis
Despite improved outcome after curative resection, a significant number of patients with gastric carcinoma still succumb to death due to primary treatment failure.1 One of the major reasons for this failure is that the cancers recur through evading the host immune surveillance in many ways2 and continue to grow. Among the several possibilities, the Fas/FasL pathway is considered to be a major mechanism:3,4 tumor cells express FasL, and tumor-infiltrating lymphocytes (TILs) express Fas. This enables the tumor to counterattack TILs, inducing Fas-mediated apoptosis. Previously, we have shown that tumor FasL expression had a negative correlation with the number of CD8⫹ T cells within the cancer cell nest (nest CD8), but no correlation with the number of CD8⫹ T cells along the invasive margin (margin CD8).5 At the present moment, it is still ill understood why the large numbers of armed T cells stay idle at the periphery of the tumor and are unable to exert an antitumor effect. Lieubeau et al.6 presented the unique hypothesis that tumor-activated myofibroblasts and extracellular contents may prevent contact between cancer cells and immune cells in in vitro and animal models Tumor tissues are composed of heterogeneous types of cellular and noncellular elements. A substantial part of their volume is extracellular space, which is largely filled by an intricate network of macromolecules secreted by the fibroblasts.7 Until recently, the extracellular matrix proteins such as collagen, fibronectin and proteoglycan were thought to serve mainly as a relatively inert scaffolding to stabilize the physical structure of tissues. However, it is now clear that the extracellular matrix protein plays a far more active and complex role in regulating the behavior of the cells that come into contact with it, influencing their development, migration, proliferation, shape and function.8 The stromal reaction seen
in many invasive carcinomas suggests that stromal cells and matrix play a role in cancer pathogenesis and progression.9 Although a large number of studies have been carried out to investigate the role of TILs and their survival benefits,10 –12 little is known about the relationship between the development of stroma and TILs in cancers. In our study, we measured the amount of collagen within tumor tissue using the Azocarmine and Aniline blue (AZAN) histochemical technique,13 and we performed quantitative measurement of stromal collagen (SC) using a computerassisted image analysis system. We also explored the correlation with survival data of apoptosis, CD8⫹ T cells and FasL expression in advanced gastric carcinoma. MATERIAL AND METHODS
Patients A total of 84 advanced gastric adenocarcinoma patients who had undergone curative gastrectomy at the Shimane Medical University Hospital from January 1990 to December 1997 were included in our study. There were 41 patients with invasion of the muscularis propria or subserosa (T2) and 43 patients with serosal invasion (T3). The criteria considered for curative resection were the complete removal of primary gastric tumor, dissection of the regional lymph nodes and no remaining macroscopic tumors. All selected patients had tumors with no known distant metastases at the time of surgery and had received no preoperative radiotherapy or chemotherapy. No other previous or concomitant primary cancer was present. The clinicopathologic variables based on the TNM classification of the malignant tumors14 were retrospectively reviewed by searching a specially designed database for the gastric carcinoma patients. All patients had a complete follow-up. CD8 expression Immunohistochemical staining was performed using a mouse monoclonal antibody against CD8 protein (Novocastra Laboratories, Newcastle, UK) at 1:30 dilution. The details of this method have been described previously.15 Lymphocytes stained by antiCD8 antibody within the cancer cell nests were considered nest CD8. We also classified CD8⫹ T cells distributed along the invasive margin as margin CD8.12 For evaluation of the number of CD8⫹ T cells in each group, 3 high-power fields (⫻400 magnification) with the most abundant distribution of CD8⫹ T cells were selected from each specimen and counted; the result was expressed as the mean number of CD8⫹ T cells per 3 high-power fields. Counting was done on a conference microscope by 2 observers (S.O. and T.F.) who did not know any details of the patient’s background Histochemical detection of apoptotic cells and bodies Apoptotic cells and bodies were detected by the terminal deoxynucleotidyl transferase-mediated (dUTP) biotin nick-end label*Correspondence to: Second Department of Surgery, Shimane Medical University, Enya-cho 89-1, Izumo 693-8501, Shimane, Japan. Fax: ⫹81-853-20-2229. E-mail:
[email protected] Received 4 July 2001; Revised 24 September 2001; Accepted 28 September 2001 Published online 9 November 2001
STROMAL COLLAGEN IN GASTRIC CANCER
771
ing (TUNEL) method. The DeadEnd™ Colorimetric Apoptosis Detection System Kit (Promega, Madison, WI) was used in our study. The counting procedures have been described in our previously published study.15 Briefly, positively stained tumor cells with morphologic characteristics of apoptosis were identified using the standard criteria, including chromatin condensation, nucleolar disintegration and formation of crescentic caps of condensed chromatin at the nuclear periphery.16,17 To determine the number of appoptotic cells, 15–20 representative areas without necrosis and comprising at least 1,000 cancer cells were counted from each sample with a light microscope (⫻400 magnification). The results were expressed as apoptotic index (AI), representing the number of apoptotic cells/1,000 counted cells, and given in percentages. All samples were analyzed by 2 observers (S.O. and T.F.) unaware of clinical data. Interobserver variability was found in less than 10% of the slides examined, and consensus was reached on further review. Quantification of amount of stromal collagen For quantitative analysis of extracellular matrix in cancer stroma, tissue collagen was highlighted by AZAN staining. The tumor was scanned by using a light microscope (BX50; OLYMPUS, Tokyo, Japan) at low magnification (⫻40) to search for the areas with strong AZAN staining. The stained areas were photographed at a magnification of ⫻200 (⫻20 objective lens and ⫻10 ocular lens). The images were visualized on a computer display (Power Macintosh G4; Apple Computer, Cupertino, CA) through a charge-coupled device (CCD) video camera module (HC-300/OL; OLYMPUS) and a color image freezer (Photograb™-300 SH-3; FUJIFILM, Tokyo, Japan) (Fig. 1a). The image files were then opened one at a time in image processing software (Adobe Photoshop4.0J; Adobe Systems, San Jose, CA) using an iMac (Apple Computer) without additional video hardware. AZAN staining appeared as bright blue for a thickened collagen band. To extract bright blue pixels from the RGB image, the other color pixels were erased using a wand tool. A series of these operations was repeated until all the required bright blue pixels remained in the image (Fig. 1b). Quantification of the pixel area was done with the public-domain NIH Image program (developed at the U.S. National Institutes of Health and available on the Internet at http://rsb.info.nih.gov/nih-image/) on a Macintosh computer. The amount of SC was expressed as percentage of total area of the image (square pixels ⫻ 100/1.28 million square pixels; Fig. 1c). For each tumor, the mean of the amount of SC for 5 representative fields with highest amount of collagen was used. FasL expression Sections were immunostained as described previously15 using rabbit polyclonal antibody (Santa Cruz Biochemistry, Santa Cruz, CA), which recognized FasL protein. The primary antibody was diluted to 1:200 and incubated on the slides overnight at 4°C. Immunohistochemical staining was done using a streptoavidinbiotin (SAB) kit (Nichirei, Tokyo, Japan) with the peroxidaselabeled streptoavidin method. Tumors were considered as positive for FasL when they had stronger expression than the corresponding normal gastric mucosa. All histologic slides were examined on a conference microscope by 2 observers (S.O. and T.F.) who were unaware of the clinical data or the disease outcome. Statistical analysis We used the chi-square test and Fisher’s exact test for 2 ⫻ 2 tables to compare the categorical data. Differences between the groups were compared using the Mann-Whitney U-test. Relations between continuous variables were investigated by means of Spearman’s rank correlation coefficient. The survival rates were estimated by the Kaplan-Meier method,18 and the statistical analysis was carried out by the log-rank test to test for equality of the survival curves. In the analysis of disease-free survival rates, those who died of causes unrelated to gastric cancer were considered to
FIGURE 1 – Amount of SC assessment within gastric cancer. (a) The image with strong AZAN staining (bright blue) was set to evaluate the thickened collagen band and stored for quantitative analysis. (b) Unnecessary color pixels were erased using a wand tool to extract bright blue pixels from the RGB image. (c) The results were expressed as percentage of total area of the image (the number of colored square pixels within the image ⫻ 100/1.28 million square pixels of total area). Original magnification ⫻200.
772
OHNO ET AL.
be disease-free at the end point. Cox proportional hazard analysis was used to determine the relative contribution of various factors to the risk of recurrence.19 A p-value of ⬍ 0.05 was considered to indicate statistical significance. All statistical analyses were performed on a personal computer with the statistical package StatView version 4.5 for Macintosh (Abacus Concepts, Berkeley, CA). RESULTS
Patient characteristics There were 57 male patients and 27 female patients, and their ages ranged from 29 – 85 years (mean, 64.8 years). Of the 84 patients, 49 were still alive at the time of this reporting. Recurrence occurred in 27 patients, and 8 patients died from diseases other than gastric cancer. The median follow-up time for all patients was 3.19 years (range, 0.24 –9.10 years). CD8 expression at different localization Most of the CD8⫹ T cells were distributed along the invasive margin of the tumor. This distribution pattern was similar to that of lymphocytes by conventional hematoxylin and eosin staining. CD8⫹ T cells were also detected within the cancer cell nest in 72 of 84 patients. The median values of nest and margin CD8 were 4.00 and 49.83, respectively. The median value was used as the cutoff point to stratify tumors into strong and weak for CD8⫹ T cells. The infiltration pattern of CD8⫹ T cells had no association with patient’s age, gender, tumor invasion, lymph node metastasis, tumor size or Lauren or Borrmann classification (data not shown). Apoptotic index Almost all of the positively stained cells and bodies considered as apoptotic cells corresponded morphologically to the standard criteria of apoptotic cells as mentioned previously. Nonspecific staining in necrotic foci was faint and diffuse and could be easily distinguished from the apoptotic nuclei by simple morphologic examination. The incidence of AI varied from 0 to 9.40%, with a median of 1.20%. A statistically significant positive correlation was observed between the AI and CD8⫹ T cells within cancer cell nests ( ⫽ 0.417, n ⫽ 84, p ⫽ 0.0001) (Fig. 2a). However, there was no correlation between AI and CD8⫹ T cells along the invasive margin ( ⫽ 0.179, n ⫽ 84, p ⫽ 0.1025; Fig. 2b). Stromal collagen The percentage of SC varied from 0.32 to 26.46%, with a median of 5.23%. There were significantly high percentages of SC in tumors with serosal invasion (p ⫽ 0.0001), in Borrmann III/IV tumors (p ⫽ 0.0031) and in the scirrhous type of tumor (p ⫽ 0.0003). Patient age, gender, lymph node metastasis, tumor size and Lauren classification were not associated with the amount of SC (Table I). The degree of infiltration of CD8⫹ T cells into the tumor nest was dependent on the amount of SC located at the tumor margin. To stratify patients into 2 groups, the cutoff value for SC was set at a median of 5.23%. The ratio of the nest to margin CD8 was significantly (p ⫽ 0.0024) lower in patients with high SC (median, 0.054 ) than in those with low SC (median, 0.202) (Fig. 3a). Also, there were significantly (p ⬍ 0.0001) fewer numbers of nest CD8⫹ T cells in the high SC group (median, 1.667) than in the low SC group (median, 9.000) (Fig. 3b). Differential expression of FasL and SC Twenty-six (31%) cases were found to be positive for FasL only (FasL ⫹/low SC), whereas in 11 (13%) cases high SC (FasL-/high SC) was noted. Both FasL and SC were positive in 31 (37%) cases, and both of them were negative in 16 (19%) cases. Although a strong inverse association was found between the amount of SC and nest CD8, there was no correlation between FasL expression and nest CD8 (Fig. 3c). This inverse association was found to be
FIGURE 2 – (a) A significant direct correlation ( ⫽ 0.417, n ⫽ 84, p ⫽ 0.0001) between the AI and nest CD8 was observed by using the Spearman rank correlation coefficient. (b) However, there was no correlation ( ⫽ 0.179, n ⫽ 84, p ⫽ 0.1025) between the AI and margin CD8. SC, stromal collagen.
stronger (p ⬍ 0.0001) in cases with both FasL-positive and high SC tumors (median, 1.667) than in others (median, 8.333; Fig. 3d). Survival analysis The median values of nest CD8, margin CD8 and AI, as well as the percentage of SC, were used as cutoff points to stratify patients into 2 groups in each case. The factors negatively influencing the disease-free survival rate by univariate analysis were serosal invasion, presence of lymph node metastasis, fewer nest CD8, low AI and increase in SC (Fig. 4, Table II). Patient age and gender, tumor size, Lauren or Borrmann classification, FasL expression and margin CD8 had no statistically significant impact on patient survival (data not shown). Cox proportional hazard analysis of all factors listed in Table II revealed that depth of invasion (p ⫽ 0.0129), nest CD8 (p ⫽ 0.0023) and AI (p ⫽ 0.0440) were independent risk factors related to patient survival rate but that lymph node metastasis and amount of SC had no independent significant effect (Table II). DISCUSSION
The desmoplastic reactions in cancer stroma and the wound healing process have close similarities. Both the tumor stroma and the granulation tissue in a healing wound are composed of 3 main
STROMAL COLLAGEN IN GASTRIC CANCER
773
TABLE I – AMOUNT OF STROMAL COLLAGEN (SC) AND CLINICOPATHOLOGIC CHARACTERISTICS OF GASTRIC CANCER1 Variable
Age (yr) ⬍65 (n ⫽ 32) ⱖ65 (n ⫽ 52) Gender Female (n ⫽ 27) Male (n ⫽ 57) T stage T2 (n ⫽ 41) T3 (n ⫽ 43) Lymph node metastasis Negative (n ⫽ 22) Positive (n ⫽ 62) Tumor size (cm) ⬍6.0 (n ⫽ 49) ⱖ6.0 (n ⫽ 35) Lauren classification Intestinal type (n ⫽ 29) Diffuse type (n ⫽ 55) Scirrhous type Nonscirrhous (n ⫽ 52) Scirrhous (n ⫽ 32) Borrmann classification I, II (n ⫽ 31) III, IV (n ⫽ 53)
Amount of SC
p-value2
4.994 (7.422) 5.229 (8.340)
0.8611
4.491 (8.645) 5.300 (7.550)
0.9427
2.496 (4.505) 9.192 (7.421)
0.0001
2.684 (8.023) 5.742 (7.565)
0.0784
5.228 (9.013) 5.230 (7.362)
0.9602
2.753 (8.112) 6.478 (7.720)
0.1800
2.654 (6.793) 8.409 (6.505)
0.0003
2.496 (7.546) 6.680 (7.631)
0.0031
1 Data are medians, with interquartile ranges in parentheses.–2The Mann-Whitney U-test was used for group comparisons.
elements: newly formed blood vessels, inflammatory/immune cells and connective tissues. During carcinogenesis and tumor progression, tumor stroma undergoes repeated degradation and remodeling.7 In our study, we found that variable amounts of collagen (stained by AZAN) existed in tumor stroma of gastric carcinoma. This indicates that during tumor progression stroma cause injury to the host’s tissue and activate the never ending host wound-healing response. We speculate that immature tumor stroma or a wound healing response that has gone awry provide malignant tissue with a suitable milieu for growth. The amount of SC or components of stroma is one of the most important prognostic factors in malignant tumors.20 In our study, we have shown that a dense SC had a negative impact on the survival of patients with gastric carcinoma. Recently, Sethi et al.21 reported that small cell lung cancer patients with extensive extracellular matrix (ECM) around their tumors had a significantly shorter survival than those with focal or no matrix. The authors further noted that the excessive ECM induces tumor resistance to chemotherapy by blocking apoptosis and inducing the enzyme caspase through protein tyrosine kinase activation. Among the several components of tumor stroma, tenascin and hyaluronan expression is thought to impact significantly on patient prognosis.22,23 Higher expression of tenascin, a well-known stromal marker, has been shown to be related to increased frequency of lymph node metastasis and poor outcome in human breast tumors.24 It has been thought that tenascin produced by the tumor mesenchyme actively participates in cancer development by promoting cancer cell proliferation and invasion. Hyaluronan, another component of the ECM produced by myofibroblasts, has been shown to produce an immunoprotective cocoon for cancer cells and block the lymphocyte-mediated cytolysis.25 Accordingly, a high level of stromal hyaluronan predicts poor patient survival, whereas the hyaluronidase enzyme has been successfully used as a chemosensitizer in several carcinomas including breast, melanoma and bladder cancers. Most of these studies, however, have not addressed the immunomodulatory impact of these factors in malignant tumors. In the present study, we have evaluated the interactions between the desmoplastic and immune reactions against tumor progression, with particular emphasis on the role of SC in infiltration of CD8⫹
FIGURE 3 – CD8⫹ T-cell analysis by Mann-Whitney U test in gastric carcinoma tissues. (a, b) The ratio of the nest to margin CD8 (a) and the nest CD8 (b) were dependent on the amount of SC (p ⫽ 0.0024 and p ⬍ 0.0001, respectively). (c) There was only a tendency (p ⫽ 0.2158) toward inverse association between FasL expression and nest CD8. (d) The nest CD8 was significantly (p ⬍ 0.0001) reduced in patients with both FasL positive and high SC. In the box and whisker plot, the median values are indicated by horizontal bars. The vertical bars indicate the 10th and 90th percentiles, and the horizontal boundaries of the boxes represent the 1st and 3rd quartiles.
FIGURE 4 – A significant difference between the 2wo groups was seen at each time point and when the whole period of follow-up was compared by using Kaplan-Meier analysis and the log rank test (p ⫽ 0.0010). A median value of 5.23% was applied to the cutoff point.
T cells into the tumor nest. Nest CD8 was inversely decreased with the increasing amount of SC and also had a direct correlation with AI. Furthermore, the ratio of the nest to margin CD8 was significantly lower in patients with high SC than those with low SC, indicating that the SC served as a mechanical barrier against the infiltration of immune-competent cells into the cancer nest. Similarly, in a coculture system, Lieubeau et al.6 have recently shown that the numbers of lymphocytes/macrophages penetrated
774
OHNO ET AL. TABLE II – STATISTICAL ANALYSIS OF PROGNOSTIC FACTORS Variable1
Tumor invasion (T2/T3) Lymph node metastasis (⫹/⫺) Apoptotic index (high/low) Nest CD8 (strong/ weak) Amount of SC (high/ low) 1
Kaplan-Meier p-value (log-rank test)
Cox proportional hazards model Hazard ratio
95% CI
p-value
0.0012
3.803
1.327–10.904 0.0129
0.0382
2.369
0.704–7.969
0.1635
⬍0.0001
2.964
1.030–8.535
0.0440
⬍0.0001
6.581
1.959–22.101 0.0023
0.0010
1.545
0.471–5.067
0.4729
SC, stromal collagen.
are significantly decreased with increasing concentrations of collagen gel. In addition, a report by Vaage et al.26 indicating that lymphocytes and macrophages penetrated progressive and regressive tumors differently in a mouse mammary model also supports our findings. Previously, we reported that nest CD8 might be a significant prognostic factor in gastric carcinoma patients. We also found an inverse association between the number of CD8⫹ T cells and tumor FasL expression in a set of both early and advanced gastric carcinomas, and we have postulated that a counterattack mecha-
nism by the tumor FasL against the host immune cells exists in gastric carcinoma patients.15 However, in our study of only advanced gastric carcinomas (T2 and T3), there was no significant association between these factors. This could be explained by an additional mechanism of tumor immune privilege by the formation of SC around the tumor nest in advanced gastric carcinomas. Indeed, the suppression of nest CD8 was stronger in cases with FasL⫹/high SC compared with the other cases. Nest CD8 had the highest predictive value (Hazard ratio ⫽ 6.581) for determining gastric carcinoma disease recurrence after curative resection. However, the amount of SC was not a significant predictor in the multivariate analysis. A possible explanation for this might be the incorporation of T stage and nest CD8 (which are strongly correlated) in the multivariate analysis along with the SC factor. Although it was not an independent predictor of patient survival, our study shows that SC might be one of the several limiting factors controlling the degree of CD8⫹ T-cell infiltration into the cancer cell nest. It seems that the ECM forms a barrier that both works against and supports cancer cells. The ECM is also involved in the storage of growth factors that induce tumor neovascularization and thus accelerate tumor growth. In this respect, for success in T-cell-based immunotherapy, complementary treatments against local barriers involving the tumor stroma will be required. Several elements of the cancer stroma could be selected as targets for investigative cancer therapy. ACKNOWLEDGEMENTS
The authors thank Ms. M. Asazu for her excellent technical assistance.
REFERENCES
1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14.
Hundahl SA, Phillips JL, Menck HR. The national cancer data base report on poor survival of U.S. gastric carcinoma patients treated with gastrectomy. Cancer 2000;88:921–32. Bodmer WF, Browning MJ, Krausa P, et al. Tumor escape from immune response by variation in HLA expression and other mechanizm. Ann NY Acad Sci 1993;690:42– 49. O’Connell J, Houston A, Bennett MW, et al. Immune privilege or inflammation? Insights into the Fas ligand enigma. Nat Med 2001;7:271–74. Bennett MW, O’Connell J, O’Sullivan GC, et al. Expression of Fas ligand by human gastric adenocarcinomas: a potential mechanism of immune escape in stomach cancer. Gut 1999;44:156 – 62. Shibakita M, Tachibana M, Dhar DK, et al. Prognostic significance of Fas and Fas ligand expressions in human esophageal cancer. Clin Cancer Res 1999;5:2464 – 69. Lieubeau B, Heymann MF, Henry F, Immunomodulatory effects of tumor-associated fibroblasts in colorectal-tumor development. Int J Cancer 1999;81:629 –36. Dvorak HF. Tumors: wounds that do not heal. Similarities between tumor stroma generation and wound healing. N Engl J Med 1986; 315:1650 –59. Sieweke MH, Bissell MJ. The tumor-promoting effect of wounding: a possible role for TGF--induced stromal alterations. Crit Rev Oncogen 1994;5: 297–311. Gr´egoire M, Lieubeau B. The role of fibroblasts in tumor behavior. Cancer Metastasis Rev 1995;14:339 –50. Underwood JCE. Lymphoreticular infiltration in human tumors: prognostic and biological implications. Br J Cancer 1974;30:537– 48. Ropponen KM, Eskelinen MJ, Lipponen PK, et al. Prognostic value of tumor-infiltrating lymphocytes (TILs) in colorectal cancer. J Pathol 1997;182:318 –24. Naito Y, Saito K, Shiiba K, et al. CD8⫹ T cells infiltrated within cancer cell nests as a prognostic factor in human colorectal cancer. Cancer Res 1998;58:3491–94. ¨ ber die Mallorysche Bindegewebsf¨arbung mit Heidenhain M. U Karmin und Azokarmin als Vorfarben. Z Wissenschaft Mikrosk Mikrosk Techn 1915;32:361–72. Sobin LH, Wittekind C, eds. TNM classification of malignant tumors, 5th ed. New York: Wiley-Liss, 1997.
15. Ohno S, Tachibana M, Shibakita M, et al. Prognostic significance of Fas and Fas ligand system-associated apoptosis in gastric cancer. Ann Surg Oncol 2000;7:750 –57 16. Gavrieli Y, Sherman Y, Ben-Sasson SA. Identification of programmed cell death in situ via specific labeling of nuclear DNA fragmentation. J Cell Biol 1992;119:493–501. 17. Ansari B, Cortes PJ, Greenstein BD, et al. In situ end-labeling detects DNA strand breaks in apoptosis and other physiological and pathological states. J Pathol 1993;170:1– 8. 18. Kaplan EL, Meier P. Nonparametric estimation for incomplete observation. J Am Stat Assoc 1958;53:457– 81. 19. Cox DR. Regression models and life tables. J R Stat Soc Ser B 1972;34:187–220. 20. Shimosato Y, Hashimoto T, Kodama T, et al. Prognostic implications of fibrotic focus (scar) in small peripheral lung cancers. Am J Surg Pathol 1980;4:365–73. 21. Sethi T, Rintoul RC, Moore SM, et al. Extracellular matrix proteins protect small cell lung cancer cells against apoptosis: a mechanism for small cell lung cancer growth and drug resistance in vivo. Nat Med 1999;5:662-68. 22. Auvinen P, Tammi R, Parkkinen J, et al. Hyaluronan in peritumoral stroma and malignant cells associates with breast cancer spreading and predicts survival. Am J Pathol 2000;156:529 –36. 23. Pirinen R, Tammi R, Tammi M, et al. Prognostic value of hyaluronan expression in non-small-cell lung cancer: increased stromal expression indicates unfavorable outcome in patients with adenocarcinoma. Int J Cancer 2001;95:12–17. 24. Ishihara A, Yoshida T, Tamaki H, et al. Tenascin expression in cancer cells and stroma of human breast cancer and its prognostic signifinance. Clin Cancer Res 1995;1:1035– 41. 25. McBride WH, Bard JBL. Hyaluronidase-sensitive halos around adherent cells. Their role in blocking lymphocyte-mediated cytolysis. J Exp Med 1979;149:507–15. 26. Vaage J, Pepin KG. Morphological observations during developing concomitant immunity against a C3H/He mammary tumor. Cancer Res 1985;45:659 – 66.
