Secretion of vascular endothelial growth factor in adenocarcinoma and squamous cell carcinoma of the uterine cervix

Secretion of Vascular Endothelial Growth Factor in Adenocarcinoma and Squamous Cell Carcinoma of the Uterine Cervix ALESSANDRO D. SANTIN, MD, PAUL L. HERMONAT, PhD, ANTONELLA RAVAGGI, PhD, SERGIO PECORELLI, MD, PhD, MARTIN J. CANNON, PhD, AND GROESBECK P. PARHAM, MD Objective: To determine whether major differences in vascular endothelial growth factor secretion exist between adenocarcinomas of the uterine cervix compared with squamous cell carcinomas. Methods: The secretion of vascular endothelial growth factor by eight fresh cervical cancer cell preparations (four adenocarcinomas and four squamous cell carcinomas) and four established squamous cell lines was evaluated using a sensitive enzyme-linked immunosorbent assay in vitro. Results: All cervical tumors secreted significant amounts of vascular endothelial growth factor, and no significant differences between fresh and established squamous cell lines were detectable. In contrast, a highly significant difference in vascular endothelial growth factor secretion was noted between fresh adenocarcinomas (mean 5 2712, range between 1700 to 3500 pg/mL/105 cells/48 hours) when compared with fresh squamous (mean 5 575, range between 200 to 950 pg/mL/105 cells/48 hours) or established squamous cervical carcinoma cell lines (mean 5 712, range between 400 to 1000 pg/mL/105 cells/48 hours) (F-test, P < .001). Conclusion: These data strongly suggest that major differences in the secretion of vascular endothelial growth factor exist between squamous cell carcinoma and adenocarcinomas of the uterine cervix. Therefore, at least some of the differences in the natural biologic behavior of these two histologic types of cervical cancer, including the propensity for earlier lymphatic and hematogenous metastasis as well as the lower response to radiation treatment, could be related to major differences in the secretion of this powerful angiogenic and immunosuppressive cytokine. (Obstet Gy-

From the Division of Gynecologic Oncology, Department of Obstetrics and Gynecology, and the Department of Microbiology and Immunology, University of Arkansas, Little Rock, Arkansas; and the Division of Gynecologic Oncology, Department of Obstetrics and Gynecology, University of Brescia, Brescia, Italy. Supported in part by grants from the Camillo Golgi Foundation, Brescia, Italy to A. S., and from the Arkansas Science and Technology Authority to G. P.

78 0029-7844/99/$20.00 PII S0029-7844(99)00282-3

necol 1999;94:78 – 82. © 1999 by The American College of Obstetricians and Gynecologists.)

Human papillomavirus (HPV) infection represents the most important risk factor for developing cervical cancer.1 Recent accumulating evidence suggests that the majority of cervical squamous cell carcinomas and a large proportion of adenocarcinomas share a common pathogenesis involving oncogenic HPV types 16 and 18.1,2 Several studies have reported a significantly less favorable prognosis, stage-for-stage, for patients harboring adenocarcinomas when compared with squamous cell cervical cancers.3– 6 Given that stage of the disease cannot explain these difference in survival, it is likely that differences in the biology of these two major histologic cancer types and/or in their response to treatment may be the reason. Indeed, it has been suggested that adenocarcinomas are less sensitive to radiation than are squamous cell carcinomas.7,8 It has also been reported that a higher proportion of adenocarcinomas than squamous cell carcinomas show lymph node metastasis at the time of diagnosis.9 However, the reasons for such differences in biologic behavior remain poorly understood. Growth and metastasis of solid tumors depend on the establishment of new blood vessels from preexisting vasculature.10 Recently, several angiogenic factors have been identified,11 and among these peptides, including basic and acidic fibroblast growth factor, transforming growth factor a, transforming growth factor b, plateletderived endothelial cell growth factor, vascular endothelial growth factor has emerged as a prominent tumor angiogenesis factor in vivo.12 High vascular endothelial growth factor secretion by primary tumors significantly

Obstetrics & Gynecology

correlates with a more aggressive biologic behavior (ie, increased vessel involvement, lymph node metastastis, and liver metastasis) and a worse prognosis compared with low vascular endothelial growth factor secretion in several human cancers.13,14 Furthermore, vascular endothelial growth factor secretion from tumor cells has been correlated with a powerful immunosuppressive activity.15 Indeed, Gabrilovich et al15 have reported that vascular endothelial growth factor may have an inhibitory effect on the maturation of dendritic cells, which are the most powerful professional-antigen presenting cells known in humans. Such findings raise the possibility that vascular endothelial growth factor may act both as a powerful angiogenic factor and as a growth regulator with broader activities including the ability to affect tumor–host immune system interaction dramatically. Previous reports have demonstrated overexpression of vascular endothelial growth factor using immunochemistry in adenocarcinomas compared with squamous cell cervical carcinomas16,17; however, no studies have quantitatively analyzed and compared the production of vascular endothelial growth factor in cervical cancer cells. Because of this, we examined and quantified the production of vascular endothelial growth factor by several fresh tumors and established cervical cancer cell lines with different histotypes using a sensitive enzyme-linked immunosorbent assay (ELISA).

Materials and Methods Cervical punch biopsies from four squamous cell carcinomas and four adenocarcinomas of the cervix were taken from patients with gross tumors at the time of diagnosis through the Pathology Department at the University of Arkansas for Medical Sciences, Little Rock, Arkansas, under approval of the institutional review board. The age range of patients harboring adenocarcinoma was 26 to 51 years (40 6 10 years; mean 6 standard deviation), and the age range of patients harboring squamous cell carcinomas was 27 to 48 years (39 6 5 years). Human papillomavirus typing was performed on all eight fresh tissue biopsies and on the derived fresh cultures by polymerase chain reaction (PCR) using sequence specific primers for HPV-16, -18, -31, -33, -52b, and -58.18 Stage of the tumors from which the cervical cancer cells were extracted is described in Table 1. All four fresh squamous carcinomas (CVX-1, CVX-2, CVX-3, CVX-4), were positive for HPV-16, whereas the four fresh cell lines derived from adenocarcinomas (CVX-AD-1, CVX-AD-2, CVX-AD-3, CVXAD-4) were positive for HPV-18 (data not shown). Established squamous cell lines (SiHa, CaSki [both HPV 161] and MS751 and C-4I [both HPV181])19 were purchased from the American Type Culture Collection

VOL. 94, NO. 1, JULY 1999

Table 1. Vascular Endothelial Growth Factor Production in Fresh Cervical Cancer Cells and Established Cervical Carcinoma Cell Lines Tumor cells

pg/mL

Stage

CVX-SQ-1* CVX-SQ-2 CVX-SQ-3 CVX-SQ-4 CVX-AD-1† CVX-AD-2 CVX-AD-3 CVX-AD-4 CaSKI SiHA MS751 C-4 I

650 950 500 200 1700 2900 3500 2750 1000 800 650 400

IB2 IB2 IIB IB2 IB2 IVa IB2 IB2 ‡ ‡ ‡ ‡

* Fresh squamous cell lines. † Fresh adenocarcinoma cell lines. ‡ Data not available.

(Rockville, MD) and were maintained at 37C, 5% CO2 in complete media containing Roswell Park Memorial Institute (RPMI) medium 1640 (GIBCO; Life Technologies, Grand Island, NY), supplemented with 10% fetal bovine serum (Gemini Bioproducts, Calabasas, CA). All fresh tumors were maintained at 37C, 5% CO2 in complete media containing keratinocyte serum-free medium (GIBCO BRL), supplemented with 5 ng/mL epidermal growth factor and 35 to 50 mg/mL bovine pituitary extract (GIBCO) and 1% penicillin/streptomycin sulfate (Irvine Scientific, Santa Ana, CA). Briefly, single cell suspensions were obtained by processing solid tumor samples under sterile conditions at room temperature. Viable tumor tissue was mechanically minced in RPMI 1640 to portions no larger than 1–3 mm3 and washed twice with RPMI 1640. The portions of minced tumor were then placed into 250-mL trypsinizing flasks containing 30 mL of enzyme solution (0.14% Collagenase Type I [Sigma Chemical Co., St. Louis, MO) and 0.01% DNAse (Sigma; 2000 KU/mg)] in RPMI 1640, and incubated on a magnetic stirring apparatus overnight at 4C. Enzymatically dissociated tumor was then filtered through 150-mm nylon mesh to generate a single cell suspension. The resultant cell suspension was then washed twice in RPMI 1640. Experiments were performed only with suspensions that had at least 90% viability and contained more than 99% tumor cells. The percentage of tumor cells was determined by differential counts of Giemsa-stained cytospin slides, and viability was confirmed by trypan blue exclusion. Tumor cells were seeded at a density of 1 3 105 cells/mL in tissue culture dishes (Corning, Costar Corp., Boston, MA) containing complete medium. After a 48-hour incubation at 37C, supernatants were aspirated, rendered cell free by centrifugation at 1500 3 g for 10 minutes, and stored at 220C. Vascular endothe-

Santin et al

Cervical Cancer Growth Factor

79

lial growth factor concentration was determined by ELISA, using a commercially available kit (Research & Diagnostic Systems, Minneapolis, MN) with vascular endothelial growth factor standards ranging from 10 to 2000 pg/mL. Samples containing more then 2000 pg/mL of vascular endothelial growth factor were reanalyzed with 5:1 or 10:1 dilutions. All samples were assayed in duplicate. Standard regression curves were generated by plotting log10 concentration versus log10 optical density, creating correlation coefficients greater than 0.98 in all cases. The maximal allowed sample duplicate error was 10%. Duplicates falling outside this margin were reanalyzed. Significance analysis to test for mean differences in vascular endothelial growth factor secretion between the three different groups of tumor cells (ie, adenocarcinomas [CVX-AD], fresh squamous [CVX-SQ], and established squamous cancer cell lines) was performed using a one-way analysis of variance, followed by a Duncan multiple range test. Only P values , .05 were considered significant.

Results Cell-free supernatants from eight freshly isolated cervical cancer (four adenocarcinomas and four squamous cell carcinomas) and four established squamous cell lines were collected and analyzed for the levels of vascular endothelial growth factor by ELISA. All vascular endothelial growth factor levels were analyzed in duplicate and had a coefficient of variation of less than 10%. Because prolonged passages in vitro are known to alter the physiology and phenotype of primary tumor cells, we performed all our experiments with fresh tumor cells grown for less than 3 passages (less than 2 weeks) in vitro. Growth control medium was always analyzed at the same time. In this regard, keratinocyte serum-free medium as well as RPMI 1640 containing 10% fetal bovine serum had no detectable endogenous levels of vascular endothelial growth factor activity (data not shown). As shown in Table 1, all cervical tumor cells secrete measurable amounts of vascular endothelial growth factor. No significant differences in secretion between fresh squamous cell carcinomas (range of secretion 200 to 950 pg/mL/105 cells/48 hours [mean 575]) and established squamous cell lines (range of secretion 400 to 1000 pg/mL/105 cells/48 hours [mean 712]) were detectable. In contrast, all four fresh adenocarcinomas were shown to produce significantly higher levels of vascular endothelial growth factor (range of secretion 1700 to 3500 pg/mL/105 cells/48 hours [mean 2712]) when compared with both fresh and established squamous carcinoma cell lines (F test, P # .001) (Figure 1). Thus, mean vascular endothelial

80 Santin et al

Cervical Cancer Growth Factor

Figure 1. Vascular endothelial growth factor secretion in established squamous cervical cancer cell lines and fresh squamous cell carcinomas and adenocarcinomas. Data are shown as mean 6 standard deviation. 1 5 established squamous cell lines; 2 5 fresh adenocarcinoma cell lines; 3 5 fresh squamous cell lines. Adenocarcinomas compared with established squamous cell lines and fresh squamous tumor cells, P , .0003; established squamous cell lines compared with fresh squamous tumor cells, P not significant.

growth factor secretion by fresh adenocarcinomas was shown to be 4.7 and 3.8 times higher when compared with fresh or established squamous tumor cells, respectively.

Discussion The fate of tumor– host interaction is thought to depend on the balance between the intrinsic aggressiveness (ie, ability to metastasize) of the tumor and the strength of the host immunity. Several comparative studies between squamous cell cervical cancers and adenocarcinomas have shown a decreased survival, stage-forstage, for patients with the latter histologic subtype.3– 6 The reasons for these differences in survival remain poorly understood but are believed to be related to a propensity for earlier lymphatic and hematogenous metastases by adenocarcinomas.7 In addition, other reports have shown differences in the outcome of patients treated with primary radiotherapy, suggesting that adenocarcinomas are also less radiosensitive than squamous cell cervical cancers.8,9 Vascular endothelial growth factor is a wellcharacterized angiogenic factor that has been reported to be expressed in several human tumors and is known to play a crucial role in neovascularization.10 –12 Differential splicing of a single gene transcript gives rise to four messenger RNA (mRNA) and, in turn, four protein isoforms of 121, 165, 189, and 206 amino acids, which

Obstetrics & Gynecology

differ in efficiency of secretion, affinity for heparin, and relative potency of vascular permeabilizing and mitogenic activities.12 Over the past decade, numerous studies have demonstrated that neovascularity of tumors correlates with aggressiveness and metastatic potential.13,14,20 –22 Relapse-free survival rate of patients with vascular endothelial growth factor–rich tumors has been reported to be significantly worse than that of vascular endothelial growth factor–poor tumors,13,14 and high circulating vascular endothelial growth factor levels before cancer treatment have been correlated with poorer prognosis in breast cancer13,22 and colorectal cancer.21 Furthermore, vascular endothelial growth factor has recently been reported to have an inhibitory effect on the maturation of professional-antigen presenting cells such as dendritic cells.15 In this regard, substantial evidence has recently established that bone marrow– derived antigen presenting cells and not tumor cells play the most important role in the presentation of tumor antigens to the immune system.23,24 These findings lead to the provocative hypothesis that tumors producing high vascular endothelial growth factor levels may facilitate their growth also by avoiding the induction of an effective immune response.15 This could be particularly important in human tumors such as cervical cancer that express immunogenic HPV antigens. Indeed, several lines of evidence suggest that cell-mediated immunity may be particularly important in HPV-associated malignancy such as cervical cancer. First, there is an increased incidence of associated genital cancer in immunosuppressed patients, whereas only a minority of genital HPV infections result in the development of cancer in otherwise healthy individuals.25–27 Second, infiltrating CD41 (T helper cells) and CD81 (cytotoxic/suppressor T cells) T cells have been observed in spontaneously regressing warts. Third, studies on animals have demonstrated that immunized animals are protected from papillomavirus infections and from transplanted tumor cells expressing HPV E6/E7 viral proteins.28,29 Expression of vascular endothelial growth factor using immunochemistry staining has previously been reported in cervical carcinomas.16,17 However, it has been previously shown that using immunohistochemical staining, antibodies have different affinities and can give only semiquantitative estimates of the protein levels. The same is true with the levels of mRNA as well, which do not invariably correlate with the pattern or intensity of protein expression.30,31 In this study we have confirmed the purity of the tumor cells in fresh tumor specimens by differential counts of Giemsa-stained cytospin slides as well as by

VOL. 94, NO. 1, JULY 1999

cytokeratin expression using immunohistochemical techniques (data not shown). Our fresh tumor samples contained more than 99% tumor cells. The results demonstrated that vascular endothelial growth factor is actively secreted by all fresh human cervical tumors as well as the established cervical cancer cell lines evaluated. Fresh squamous tumors secreted similar levels of vascular endothelial growth factor when compared with established squamous cell lines (P not significant). In contrast, vascular endothelial growth factor secretion by fresh adenocarcinomas was always significantly (ie, three to four times) higher than that by both fresh squamous cell carcinomas or established squamous carcinoma cell lines (F test, P # .0003). These data are in agreement with reports showing a significant increase in microvessel density in adenocarcinomas compared with squamous cell carcinomas,16,17 as well as an increase in vascular endothelial growth factor expression by immunocytochemistry in the same cancer histotypes.17 The prognosis of patients harboring adenocarcinoma is widely known to be poorer than that for patients harboring squamous cell cervical cancers, and adenocarcinomas have also been described to be less radiosensitive than squamous cell cervical cancers to primary radiation treatment. Recently, we have reported that high doses of irradiation can induce a significant and long-lasting upregulation of E6/E7 oncogenes32 and major histocompatibility complex class I restriction elements33 on HPV-positive cervical cancer cell lines. It is likely that these effects combined with a lower vascular endothelial growth factor secretion at the tumor site by squamous cell cervical cancers could favor local immune reactions within the tumor microenvironment during radiation treatment. Indeed, it is well known that inflammatory infiltrates rich in T and plasma cells are commonly seen accompanying squamous cell cervical cancers. In contrast, significantly higher secretion of vascular endothelial growth factor by adenocarcinomas could hamper recognition and response by the immune system. In this regard, it has been reported that dendritic cells cultured in the presence of tumor supernatant containing vascular endothelial growth factor present a concentration-dependent reduction in expression of major histocompatibility complex class II and costimulatory molecules, lack the morphological features typical of dendritic cells, and have a reduced ability to take up soluble antigens and activate T-cell lymphocytes.15,34 Therefore it is likely that high vascular endothelial growth factor secretion by cervical adenocarcinomas could more strongly inhibit immune recognition of tumor cells.

Santin et al

Cervical Cancer Growth Factor

81

References 1. Bosh FX, Manos MM, Munoz N, Sherman M, Jansen AM, Peto J, et al. Prevalence of human papillomavirus in cervical cancer: A worldwide perspective. J Natl Cancer Inst 1995;87:796 – 802. 2. Brinton LA. Epidemiology of cervical cancer: Overview. In: Munoz N, Bosh FK, Shah V, Meheus A, eds. The epidemiology of human papillomavirus and cervical cancer. Lyon, France: IARC Science Publications, 1992;119:3–23. 3. Eide TJ. Cancer of the uterine cervix in Norway by histologic type, 1970 – 84. J Natl Cancer Inst 1987;79:199 –205. 4. Silcocks PBS, Thornton-Jones H, Murphy M. Squamous and adenocarcinoma of the uterine cervix: A comparison using routine data. Br J Cancer 1987;55:321–5. 5. Kleine W, Rau K, Schwoeorer D, Pfleiderer A. Prognosis of the adenocarcinoma of the cervix uteri: A comparative study. Gynecol Oncol 1989;35:145–9. 6. Hopkins MP, Morley GW. A comparison of adenocarcinoma and squamous cell carcinoma of the cervix. Obstet Gynecol 1991;77: 912–7. 7. Korhonen MO. Adenocarcinoma of the uterine cervix: Prognosis and prognostic significance of histology. Cancer 1984;53:1760 –3. 8. Randall ME, Constable WC, Hahn SS, Kim JA, Mills SE. Results of the radiotherapeutic management of carcinoma of the cervix with emphasis on the influence of histologic classification. Cancer 1988;62:48 –53. 9. Milsom I, Friberg LG. Primary adenocarcinoma of the uterine cervix: A clinical study. Cancer 1983;52:942–7. 10. Folkman J. Angiogenesis in cancer, vascular, rheumatoid and other disease. Nat Med 1995;1:27–31. 11. Folkman J, Klagsburn M. Angiogenic factors. Science 1987;235: 442–7. 12. Ferrara N, Davis-Smyth T. The biology of vascular endothelial growth factor. Endocr Rev 1997;18:4 –25. 13. Toi M, Hoshima S, Takayanagi T, Tominaga T. Association of vascular endothelial growth factor expression with tumor angiogenesis and with early relapse in primary breast cancer. Jpn J Cancer Res 1994;85:1045–9. 14. Maeda K, Chung YS, Ogawa Y, Takatsuka S, Kang SM, Ogawa M, et al. Prognostic value of vascular endothelial growth factor expression in gastric carcinoma. Cancer 1996;77:858 – 63. 15. Gabrilovich DJ, Chen HL, Girgis KR, Cunningham T, Meny GM, Nadaf S, et al. Production of vascular endothelial growth factor by human tumors inhibits the functional maturation of dendritic cells. Nat Med 1996;2:1096 –103. 16. Guidi AJ, Abu-Jawdeh G, Berse B, Jackman RW, Tognazzi K, Dvorak HF, et al. Vascular permeability factor (vascular endothelial growth factor) expression and angiogenesis in cervical neoplasia. J Natl Cancer Inst 1995;87:12137– 45. 17. Tokumo K, Kodama J, Seki N, Nakanishi Y, Miyagi Y, Kamimura S, et al. Different angiogenic pathways in human cervical cancers. Gynecol Oncol 1998;68:38 – 44. 18. Fujinaga Y, Shimada M, Okazawa K, Fukushima M, Kato I, Fujinaga K. Simultaneous detection and typing of genital human papillomavirus DNA using the polymerase chain reaction. J Gen Virol 1991;72:1039 – 44. 19. Yee C, Krishhnan-Hewlett I, Backer CC, Schlegel R, Howley PM. Presence and expression of human papillomavirus sequences in human cervical carcinoma cell lines. Am J Pathol 1985;119:361– 6. 20. Takahashi Y, Clearly KR, Mai M, Kitadai Y, Bucana CD, Ellis LM. Significance of vessel count and vascular endothelial growth factor and its receptor (KDR) in intestinal-type gastric cancer. Clin Cancer Res 1996;2:1679 – 84. 21. Takahashi Y, Kitadai Y, Bucana CD, Clearly KR, Ellis LM. Expres-

82 Santin et al

Cervical Cancer Growth Factor

22.

23.

24.

25.

26. 27. 28.

29.

30.

31.

32.

33.

34.

sion of vascular endothelial growth factor and its receptor, KDR, correlated with vascularity, metastasis, and proliferation of human colon cancer. Cancer Res 1995;55:3964 – 8. Weidner N, Semple JP, Welch WR, Folkman J. Tumor angiogenesis and metastasis-correlation in invasive breast cancer. N Engl J Med 1991;324:1– 8. Huang AYC, Golumbek P, Ahmadzadeh M, Jaffee E, Pardoll D, Levitsky H. Role of bone-marrow– derived cells in presenting MHC class I–restricted tumor antigens. Science 1994;264:961–5. Young JW, Inaba K. Dendritic cells as adjuvants for class I major hystocompatibility complex–restricted antitumor immunity. J Exp Med 1996;183:7–11. Rellihan MA, Dooley DP, Burke TW, Berkland ME, Longfield RN. Rapidly progressing cervical cancer in a patient with human immunodeficiency virus infection. Gynecol Oncol 1990;36:435– 8. Penn I. Cancers of the anogenital region in renal transplant recipients. Analysis of 65 cases. Cancer 1988;53:539 – 45. Nasiell K, Roger V, Nasiell M. Behavior of mild cervical dysplasia during long term follow-up. Obstet Gynecol 1986;67:665–9. Chen L, Thomas EK, Hu SL, Hellstrom I, Hellstrom KE. Human papilloma virus type 16 nucleoprotein E7 is a tumor rejection antigen. Proc Natl Acad Sci USA 1991;88:110 – 4. Chen L, Mizuno MT, Singal MC, Hu SL, Galloway DA, Hellstrom I, et al. Induction of cytotoxic T lymphocytes specific for a syngeneic tumor expressing the E6 oncoprotein of human papilloma virus type 16. J Immunol 1992;148:2617–21. Akhurst RJ, Fitzpatrick DR, Gatherer D, Lehnert SA, Millan FA. Transforming growth factor-bs in mammalian embryogenesis. Prog Growth Factor Res 1991;2:153– 68. Berger DP, Herbstritt L, Dengler WA, Marme D, Mertelsmann R, Fiebig HH. Vascular endothelial growth factor (VEGF) mRNA expression in human tumor models of different histologies. Ann Oncol 1995;6:817–25. Santin AD, Hermonat PL, Ravaggi A, Chiriva-Internati M, Pecorelli S, Parham GP. Radiation-enhanced expression of E6/E7 transforming oncogenes of human papillomavirus 16 in human cervical cancer. Cancer 1998;83:2346 –52. Santin AD, Hermonat PL, Chiriva-Internati M, Hiserodt JC, Woodliff J, Barclay D, et al. Effects of irradiation on the expression of major histocompatibility complex class I antigen and adhesion/costimulation molecules ICAM-1 in human cervical cancer. Int J Radiat Oncol Biol Phys 1997;39:737– 42. Oyama T, Ran S, Ishida T, Nadaf S, Kerr L, Carbone DP, et al. Vascular endothelial growth factor affects dendritic cell maturation through the inhibition of nuclear factor-kB activation in hemopoietic progenitor cells. J Immunol 1998;160:1224 –32.

Address reprint requests to:

Alessandro D. Santin, MD Division of Gynecologic Oncology UAMS Medical Center 4301 West Markham Little Rock, AR 72205-7199

Received October 7, 1998. Received in revised form December 30, 1998. Accepted January 13, 1999. Copyright © 1999 by The American College of Obstetricians and Gynecologists. Published by Elsevier Science Inc.

Obstetrics & Gynecology