The ‘angiogenic switch’ in the progression from Barrett's metaplasia to esophageal adenocarcinoma

EJSO 2003; 29: 890–894 doi:10.1016/j.ejso.2003.07.002

The ‘angiogenic switch’ in the progression from Barrett’s metaplasia to esophageal adenocarcinoma C. Mo¨bius*, H. J. Stein*, I. Becker†, M. Feith*, J. Theisen*, P. Gais‡, U. Ju¨tting§ and J. R. Siewert* Departments of *Surgery, †Pathology, Klinikum rechts der Isar der TU Munchen, Munchen, Germany; ‡Department of Pathology, GSF National Research Center of Environment and Health, and §Institute of Biomathematics and Biometry, GSF National Research Center of Environment and Health, Neuherberg, Germany

Aims: We investigated VEGF expression and neovascularisation in the metaplasia –dysplasia – carcinoma sequence of Barrett’s esophagus and 47 shades of adenocarcinoma. Method: Slides of 27 cases of Barrett’s metaplasia and high grade dysplasia were immunostained for VEGF, CD 31 and a-sm actin to discriminate between mature and immature vessels. VEGF stained slides were quantitatively evaluated measuring optical density with a computer based program. The neovascularisation coefficient was estimated with an interactive analytic computer program. Results: The median VEGF expression increased from metaplasia to advanced carcinoma. VEGF expression and the neovascularisation coefficient reached statistical significance between Barrett’s metaplasia and high grade dysplasia ðp , 0:001Þ; but were not statistically different between high grade dysplasia and microinvasive carcinoma (p ¼ 0:421; p ¼ 0:146). Comparing microinvasive to advanced carcinoma the difference was significant for both parameters ðp , 0:001Þ: Conclusions: Based on a quantitative computer based evaluation program, the present study suggests, that an angiogenic switch might exist and that it is an early event in the metaplasia –dysplasia –carcinoma sequence of Barrett’s carcinoma. The neovascularisation phase in Barrett’s carcinoma may precede tumour growth. Q 2003 Elsevier Ltd. All rights reserved. Key words: metaplasia; esophageal; adenocarcinoma.

INTRODUCTION Angiogenesis plays a major role in tumour growth and metastasis. It has been associated with poor prognosis in esophageal, breast, renal and lung cancer.1 – 4 Little is known about the role of angiogenesis in the precursor lesions of invasive carcinoma. Genetically engineered mouse models of cancer may be used to investigate the occurrence of neovascularisation in malignant progression.5 In three different murine models, RIP-Tag transgenic mice, BPV1.69 transgenic mice and K14-HPV16 transgenic mice, it neovascularisation arises already in hyperplastic tissue.6 There is elevated microvessel density and angiogenic factor expression. A similar pattern has been described for a subset of premalignant lesions in human tissue, including Correspondence to. Dr med C. Mobius, Chirurgische Klinik und Poliklinik der Technischen Universita¨t Mu¨nchen, Klinikum r. d. Isar, Ismaningerstr. 22, 81675 Mu¨nchen, Germany. Tel.: þ 49-89-4140-4747; Fax: þ 49-89-4140-4940 0748–7983/$30.00

dysplastic melanocytic lesions as precursor of malignant melanoma and moderate to high-grade dysplasia in cervix carcinoma.7,8 Adenocarcinoma of the distal esophagus usually arises from metaplastic glandular epithelium, as in Barrett’s esophagus, through a well-characterized metaplasia – dysplasia – carcinoma sequence. 9 This sequence is accompanied by a variety of molecular changes. Alterations of the transcripts of FHIT, a tumour suppressor gene, is one of the early events in the sequence and is followed by p53 mutations.10 Inactivation of other tumour suppressor genes by mutation (APC, p16) or hypermethylation (p16) as well as amplification of oncogenes such as c-erb B2 are relatively late events in the development of adenocarcinoma.11,12 Little is known about the onset of the transition to an angiogenic state (‘angiogenic switch’) in the metaplasia –dysplasia – carcinoma sequence of Barrett’s carcinoma. Recently, Auvinen reported VEGF A expression on goblet cells.13 Angiogenesis is usually measured by assessing the microvessel density.14 This method cannot distinguish Q 2003 Elsevier Ltd. All rights reserved.

‘ANGIOGENIC SWITCHES’ IN BARRETT’S METAPLASIA between newly formed and pre-existing vessels. We investigated the neovascularisation coefficient, the relation between newly formed and pre-existing capillaries, and expression of vascular endothelial growth factor (VEGF) in the metaplasia –dysplasia – carcinoma sequence, to characterize the angiogenic phenotype of each stage of malignant progression.

MATERIALS AND METHODS Patients Tissue samples of esophageal adenocarcinoma were obtained from 47 patients who underwent esophagectomy between 1991 and 1999. None of the patients had neo-adjuvant or adjuvant chemo- or radiation treatment. The pathology reports and clinical records were reviewed to determine patient’s age, sex and tumour stage. Follow-up was completed for all 47 patients with a median follow-up of 47.5 months (range 0– 111 months). T1N0 carcinomas were classified as microinvasive carcinoma, T2-4Nx carcinomas were classified as advanced carcinoma.

Tumour samples Based on initial review of all available hematoxylin – eosin stained slides of the surgical specimen sections, we selected one or two representative paraffin blocks from each of the 47 cases. The specimens were sectioned according to a standard protocol. In 27 of the selected cases, areas with intestinal metaplasia and dysplasia were present in association with the invasive carcinoma. These samples allowed assessment of the entire metaplasia – dysplasia –carcinoma sequence.

Immunohistochemistry The paraffin-embedded tumour sections were stained for VEGF (polyclonal antibody, Zymed), CD 31 (monoclonal antibody, DAKO) to mark endothelial cells and a-smooth muscle actin (a-SMA, monoclonal antibody, DAKO) to detect mural cells (smooth muscle cells and pericytes).15 VEGF expression was analyzed using the standard avidinbiotin method. Sections (2 mm) were deparaffinized and pretreated with citrate buffer using a heat-induced epitope retrieval protocol. Endogenous peroxidase was blocked with 20% hydrogen peroxide for 15 min at room temperature followed by incubation with VEGF antibody (dilution 1:100) for 30 min. A biotinylated goat antimouse immunoglobulin G secondary antibody (Dako) was then applied to each slide for 30 min. After washing in Tris –hydrochloric acid buffer (TBS), the slides were incubated with peroxidase-conjugated-streptavidin complex reagent (DAKO) and developed with 3,30 -diaminobenzidine for 5 min. Between each step skimmed milk

891 powder and goat serum (dilution 1:100, DAKO) were used for 10 min to reduce nonspecific staining. The slides were counterstained and dehydrated. Labelled neurons of myenteric plexus and capillaries of the mucosa were used as internal positive controls for VEGF and CD 31.14 Negative control slides were obtained by omitting the primary antibody. To visualize the antibodies against a-smooth muscle actin the alkaline phosphatasis/anti-alkaline phosphatases (APAAP) technique was used as described previously.14

Analysis VEGF expression The slides were analyzed quantitatively by measuring the optical density in areas with metaplasia, dysplasia and at the invasive edge of the carcinoma. Optical density was measured using orthoplan microscope (Leica, Bensheim, Germany) and a CCD camera (KY-F30, JVC, Tokyo, Japan).

Analysis of neovascularisation The capillary vessels (CD 31 positive) and mature vessels (a-smooth muscle actin positive) were counted in Barrett’s metaplasia, dysplasia and at the invasive edge of the tumour, where six separated areas were selected. The number of microvessels was assessed on a £ 200 field using an interactive, strong quantitative computerbased program. In brief, the stained slides were placed under the microscope (Leitz Orthoplan, Leican, Bensheim, Germany) and a representative area was digitized (CCD camera, KY-F30, JVC, Tokyo, Japan) and analyzed with picture analytic system SAMBA (Systeme d’Analyses Microscopiques a Balayage Automatique, UNILOG, Meylan, France). The vessels were highlighted by setting a threshold that marked only the endothelial cells. The computer counted each point that was set interactively by the investigator. Any brown-staining endothelial cell or endothelial cell cluster, clearly separate from adjacent microvesssls, tumour cells, and other connective-tissue elements, was considered a single, countable microvessel. A vessel lumen was not necessary for a structure to be counted as a microvessel, and red cells were not used to define a vessel lumen (2). Neovascularisation coefficient was calculated as CD 31 positive vessels over a-sm-actin positive vessels.

Statistical analysis Data are presented as mean with standard deviation. The comparison between the various histological groups was done using the Mann – Whitney U test. The relation between VEGF expression and the neovascularisation coefficient was determined with Pearson’s coefficient of correlation. p # 0:05 was considered to indicate statistical significance.

¨ BIUS ET AL. C. MO

892

RESULTS Cohort characteristics Forty-seven cases with adenocarcinoma of the distal esophagus were evaluated for VEGF expression and the neovascularisation coefficient. In 27 of these patients areas with metaplasia and dysplasia were present close to the invasive carcinoma. These areas were assessed separately. The mean age of the study population was 66 (range 47 –79) and sex ratio was 38 men to 9 women. Pathological classifications and referring VEGF expression/neovascularisation coefficient are shown in Table 1.

Staining pattern The squamous epithelium of the esophagus did not stain for VEGF. Slight staining was observed in the smooth muscle cells of muscularis mucosa, muscularis propria and the muscular walls of vessels. A weak staining was regularly observed in the endothelial cells. Barrett’s mucosa did stain very slightly for VEGF in the cytoplasm. In high grade dysplasia VEGF was also localized in the cytoplasm. The expression was always stronger in the distal part of the glandular epithelium. In cancer, the VEGF expression was mainly at the invasive edge of the tumour and localized in the cytoplasma.

VEGF expression In Barrett’s metaplasia the median value of VEGF expression was 20.7% (range 14.2 – 74.1%). In high grade dysplasia median value of VEGF expression was 52.4% (range 10.1 – 92.2%) and in microinvasive carcinoma the median value of VEGF expression was 49.6% (10.4 – 85.5%). In advanced carcinoma the VEGF expression was 89.8% (range 57.8 – 100.3%). The VEGF expression reached statistical significance between Barrett’s metaplasia and high grade dysplasia ðp , 0:001Þ; but did not between high grade dysplasia and microinvasive carcinoma ðp ¼ 0:421Þ: Comparing microinvasive to advanced carcinoma the difference was significant for VEGF expression ðp , 0:001Þ: Results are shown in Figure 1(a). Table 1

Figure 1 (a) This plot shows the VEGF expression in Barrett’s metaplasia (BE), in high grade dysplasia (HGD), microinvasive carcinoma (EC) and advanced carcinoma (AC). The VEGF expression increases in a statistically significant fashion from BE to HGD ðp , 0:001Þ and EC to AC ðp , 0:001Þ: There is no statistical difference between HGD and EC. (b) This plot shows the neovascularisation coefficient (ratio) in Barrett’s metaplasia (BE), in high grade dysplasia (HGD), in microinvasive carcinoma (EC) and advanced carcinoma (AC). The ratio increases in a statistical significant fashion from BE to HGD ðp , 0:001Þ and EC to AC ðp , 0:001Þ: There is no statistical difference between HGD and EC.

Neovascularisation coefficient In Barrett’s metaplasia the neovascularisation coefficient was 3.25 (range 1.78 –5.38) and in high grade dysplasia the neovascularisation coefficient was 11.42 (8.68 – 16.27). In microinvasive carcinoma the neovascularisation coefficient was 11.68 (range 8.00 – 19.08). In advanced carcinoma the neovascularisation coefficient was 30.17 (16.98– 43.64). The neovascularisation coefficient reached statistical significance between Barrett’s

Pathological classifications and corresponding VEGF expression/neovascularisation coefficient are shown

Pathological characteristics

VEGF expression

Neovascularisation coefficient

Metaplasia High grade dysplasia Tumour pT1 21 pT2 – pT4 24

25.77% (^ 14.68) 51.57% (^ 18.16)

3.35 (^ 0.92) 11.41 (^ 2.03)

67.72% (^ 31.84) 75.05% (^ 47.72)

12.05% (^ 2.84) 27.32% (^ 8.87)

‘ANGIOGENIC SWITCHES’ IN BARRETT’S METAPLASIA metaplasia and high grade dysplasia ðp , 0:001Þ; but did not between high grade dysplasia and carcinoma ðp ¼ 0:146Þ: Comparing microinvasive (pT1) to advanced carcinoma (pT2-Tx) the difference was significant for neovascularisation coefficient ðp , 0:001Þ: Results are shown in Figure 1(b).

DISCUSSION The incidence of adenocarcinoma of the distal esophagus is increasing.10 Most tumours seem to arise through a multistep process from intestinal metaplasia (Barrett mucosa) to dysplasia (classified histologically as low and high grade dysplasia) towards invasive carcinoma.11 This malignant progression is accompanied by a variety of molecular changes.11 The results of the present study demonstrate that neovascularisation, the induction of new capillary blood vessels, occurs already during the initial step of this process. This may be mediated by VEGF, a major growth factor for blood vessel formation. In Barrett cancers angiogenesis has been assessed as microvessel density by staining with different antibodies (CD 31, CD 34, von Willebrand factor) against endothelial cells.13,14,16,17 Millikan et al. correlation between microvessel density and lymph node metastases, suggested finding. Newly formed and preexisting vessels were usually not distinguished from each other. There is a fraction of immature vessels lacking periendothelial cells, namely pericytes and smooth muscle cells, during human angiogenesis.15 These two cell types can be visualized by a-sm actin antibodies. Therefore, we used CD 31 to mark capillary vessels, but also used antibodies against a-sm actin to identify the fraction of mature vessels. The relation between CD 31 positive cells over a-sm actin positive cells represents the neovascularisation coefficient and the activity of angiogenesis. Our computer based evaluation program is a semiautomated counting program. Computer based analysis may be a more objective method18 but this computer program is not able to overcome the problem of heterogeneity in the tumour mass, because the selected areas do not represent the whole tumour. Tumour growth divides into two different phases: the prevascular phase and the vascular phase regulated mainly by the diffusion limit of oxygen. The prevascular phase was associated with tumour growth not larger than 1 – 2 mm3 and few or no metastases. The vascular phase is linked to rapid tumour growth, bleeding and metastasis.2 This might be true for some tumour entities. But, in accordance with the observations in transgenic mouse models of carcinogenesis,6 an ‘angiogenic switch’ may be an early event in the metaplasia –dysplasia – carcinoma sequence of the Barrett carcinoma. Angiogenesis is activated by the imbalance between pro- and antiangiogenic molecules.19 VEGF is one of the

893 most potent and critical vascular regulatores.20 It was shown for different tumour entities,21 that mutation of p53 is associated with an increase of VEGF expression in tumour cells and vascular density. p53 mutations have also been reported to be an early event in the malignant progression from metaplasia to carcinoma in Barrett’s cancer.22 Our findings may be explained by VEGF increase due to p53 mutations in dysplastic cells. Another explanation for the early trigger to the ‘angiogenic switch’ in esophageal carcinoma may be the correlation of this disease with gastroesophageal reflux and esophageal inflammation. Interleukin 6 induces VEGF expression in stroma cells and granulocytes.20 This mechanism could also support the development of new vessels in Barrett’s metaplasia. In summary, we show, that VEGF expression and neovascularisation increases before invasive carcinoma develops. These findings might offer new strategies for early detection and prevention of the malignant progression of Barrett’s esophagus.

REFERENCES 1. Ogawa Y, Chung YS, Nakata B, Takatsuka S, Maeda K, Sawada T et al. Microvessel quantitation in invasive breast cancer by staining for factor VIII-related antigen. Br J Cancer 1995; 71: 1297–301. 2. Weidner N, Semple JP, Welch WR, Folkman J. Tumor angiogenesis and metastasis-correlation in invasive breast carcinoma. N Engl J Med 1991; (1): 1–8. 3. Macchiarini P, Fontanini G, Hardin MJ, Chuanchieh H, Bigini D, Vignati S et al. Relation of neovasculature to metastasis of nonsmall-cell lung cancer. Lancet 1992; 340: 145 –6. 4. Cascinu S, Staccioli MP, Gasparini G, Giordani P, Catalano V, Ghiselli R et al. Expression of vascular endothelial growth factor can predict event-free survival in stage 2 colon cancer. Clin Can Res 2000; 6: 2803–7. 5. Bergers G, Javaherian K, Lo KM, Folkmann J, Hanahan D. Effects of angiogenesis inhibitors on multistage carcinogenesis in mice. Science 1999; 284: 808– 12. 6. Hanahan D, Christofori G, Naik P, Arbeit J. Transgenic mouse models of tumor angiogenesis: the angiogenic switch, its molecular controls, and prospects for preclinical therapeutic models. Eur J Cancer 1996; 32(14): 2386–93. 7. Barnhill RL, Fandrey K, Levy MA, Mihm MC, Hyman B. Angiogenesis and tumor progression of melanoma. Quantification of vascularity in melanocytic nevi and cutaneous malignant melanoma. Lab Invest 1992; 67(3): 331 –7. 8. Smith-McCune K, Zhu YH, Hanahan D, Arbeit J. Cross-species comparison of angiogenesis during the premalignant stages of squamous carcinogenesis in the human cervix and K14-HPV16 transgenic mice. Cancer Res 1997; 57(7): 1294– 300. 9. Stein HJ, Feith M, Siewert JR. Malignant degeneration of Barrett’s esophagus: clinical point of view. Recent Results Cancer Res 2000; 155: 42–53. 10. Werner M, Mueller J, Walch A, Ho¨fler H. Molecular pathology of Barrett’s esophagus. Histo Histopathol 1999; 14: 553 –9. 11. Mueller J, Werner M, Siewert JR. Malignant progression in Barrett’s esophagus: pathology and molecular biology. Recent Res Cancer Res 2000; 155: 29–41. 12. Jankowski JA, Wright NA, Meltzer SJ, Triadafilopoulos G, Geboes K, Casson AG et al. Molecular evolution of the

¨ BIUS ET AL. C. MO

894

13.

14.

15.

16.

17.

metaplasia–dysplasia–adenocarcinoma sequence in the esophagus. Am J Pathol 1999; 154(4): 965–73. Auvinen MI, Sihvo EI, Ruohtula T, Salminen JT, Koivistoinen A, Siivola P et al. Incipient angiogenesis in Barrett’s epithelium and lymphangiogenesis in Barrett’s adenocarcinoma. J Clin Oncol 2002; 20(13): 2971– 9. Couvelard A, Paraf F, Gratio V, Scoazec J, Henin D, Degott C et al. Angiogenesis in the neoplastic sequence of Barrett’s esophagus. Correlation with VEGF expression. J Pathol 2000; 192: 14–18. Benjamin LE, Golijanin D, Itin A, Pode D, Keshet E. Selective ablation of immature blood vessels in established human tumors follows vascular endothelial growth factor withdrawal. J Clin Invest 1999; 103: 159–65. Millikan KW, Mall JW, Hollinger JA, Doolas A, Saclarides TJ. Do angiogenesis and growth factor expression predict prognosis of esophageal cancer? Am Surg 2000; 66: 401–5. Torres C, Wang H, Shahsafei A, Odze RD. Prognostic significance and effect of chemoradiotherapy on microvessel density (angiogenesis) in esophageal Barrett’s esophagus-associated adenocarcinoma and squamous cell carcinoma. Hum Pathol 1999; 30: 753–8.

18. Vermeulen PB, Gasparini G, Fox SB, Toi M, Martin L, McCulloch P et al. Quantification of angiogenesis in solid human tumours: an international consensus on the methodology and criteria of evaluation. Eur J Can 1996; 32A: 2474–84. 19. Yancopoulos GD, Davis S, Gale NW, Rudge JS, Wiegand SJ, Holash J. Vascular-specific growth factors and blood vessel formation. Nature 2000; 407: 242–8. 20. Ferrara N. VEGF: an update on biological and therapeutic aspects. Curr Opin Biotechnol 2000; 11(6): 617 –24. 21. Wallace-Brodeur RR, Lowe SW. Clinical implications of p53 mutations. Cell Mol Life Sci 1999; 55(1): 64 –75. 22. Casson AG, Manolopoulos B, Troster M, Kerkvliet N, O’Malley F, Inculet R et al. Clinical implications of p53 gene mutations in the progression of Barrett’s epithelium to invasive esophageal cancer. Am J Surg 1994; 167: 52 –7. Accepted for publication 7 July 2003