Donnem, Tom MD*†; Al-Shibli, Khalid MD‡§; Al-Saad, Samer MD‡∥; Busund, Lill-Tove MD, PhD‡∥; Bremnes, Roy M. MD, PhD*†
Lung cancer is the leading cause of cancer-related mortality in both men and women.1 The most important prognostic variable for survival in non-small cell lung cancer (NSCLC) has been tumor stage, primarily because early stage disease is amenable to complete surgical resection and hitherto only patients who undergo curative surgery have a significant potential for cure.2,3 Several biochemical and clinical characteristics have been investigated to assess their prognostic and/or predictive relevance. In the era of new targeted therapies, identifying the patients most likely to benefit from such treatment is becoming increasingly important.
Angiogenesis, the process of new blood vessel formation from preexisting ones, plays a key role in tumor growth.4 The fibroblast growth factor (FGF) family represents a group of heparin-binding, multifunctional polypeptides with mitogen activity which also is involved in angiogenesis.5 Fibroblast growth factor 2 (FGF2; basic fibroblast growth factor, b-FGF) is considered a potent stimulator of angiogenesis and binds with high affinity mainly to fibroblast growth factor receptor-1 (FGFR-1), a tyrosine kinase receptor.4 FGF2 may exert its effect on endothelial cells via a paracrine mode as a consequence to its release by tumor and stromal cells. It is also suggested that FGF2 plays an autocrine role in endothelial cells.4,6
Previous data on FGF2’s prognostic impact in NSCLC has been conflicting.7–12 Some studies report tumor cell FGF2 expression to correlate with poor survival8,10,11 although other studies find no such association.9,12 In one study, however, an inverse correlation between stromal FGF2 expression and lymph node metastasis was observed.13 Several NSCLC studies have explored the prognostic role of elevated serum-FGF2, but no consensus has been reached.7,14–18
Although the activity of individual angiogenic factors is relatively well studied, less is known about the interplay between various tumor-produced angiogenic factors and their cooperative efforts in promoting tumor neovascularization. Interestingly, a recent study using murine fibrosarcomas reports a reciprocal interaction between FGF2 and platelet-derived growth factor-B (PDGF-B) through their tyrosine kinase receptors, FGFR-1 and PDGFR-β.19,20
An intimate cross-talk between FGF2 and different members of the VEGFs during hemangiogenesis and lymphangiogenesis has been proposed. Kubo et al.21 reported that blockade of vascular endothelial growth factor-3 (VEGFR-3) signaling inhibits FGF2-induced lymphangiogenesis in mouse cornea. We have previously reported on the importance of VEGFs and PDGFs and their receptors in both tumor cells and stroma.22–24 In this study, our aim was, based on appealing preclinical results, to explore the (1) prognostic significance of FGF2 and FGFR-1 expression in both tumor cells and stroma of resected NSCLC patients and (2) their coexpression with PDGF-B and VEGFR-3.
PATIENTS AND METHODS
Patients and Clinical Samples
Primary tumor tissues from anonymized patients diagnosed with NSCLC pathologic stage I to IIIA at the University Hospital of Northern Norway and Nordland Central Hospital from 1990 through 2004 were used in this retrospective study. In total, 371 patients were registered from the hospital database. Of these, 36 patients were excluded from the study due to: (i) Radiotherapy or chemotherapy before surgery (n = 10); (ii) Other malignancy within 5 years before NSCLC diagnosis (n = 13); (iii) Inadequate paraffin-embedded fixed tissue blocks (n = 13). Adjuvant chemotherapy was not introduced in Norway during this period (1990–2004). Thus, 335 patients with complete medical records and adequate paraffin-embedded tissue blocks were eligible.
This report includes follow-up data as of September 30, 2005. The median follow-up was 96 (range, 10–179) months. Complete demographic and clinical data were collected retrospectively. Formalin-fixed and paraffin-embedded tumor specimens were obtained from the archives of the Departments of Pathology at University Hospital of Northern Norway and Nordland Central Hospital. The tumors were staged according to the International Union Against Cancer’s tumor, node, metastasis classification and histologically subtyped and graded according to the World Health Organization guidelines.25 The National Data Inspection Board and The Regional Committee for Research Ethics approved the study.
All lung cancer cases were histologically reviewed by 2 pathologists (S.A.S.) and (K.A.S.) and the most representative areas of tumor cells (neoplastic epithelial cells) and tumor stroma were carefully selected and marked on the hematoxylin and eosin (H/E) slide and sampled for the tissue microarray blocks (TMAs). The TMAs were assembled using a tissue-arraying instrument (Beecher Instruments, Silver Springs, MD). The Detailed methodology has been previously reported.22 Briefly, we used a 0.6 mm diameter stylet, and the study specimens were routinely sampled with two replicate core samples (different areas) of neoplastic tissue and two of tumor stroma. Both normal lung tissue localized distant from the primary tumor, and one slide with normal lung tissue samples from 20 patients without a cancer diagnosis, were used as negative controls.
To include all core samples, eight tissue array blocks were constructed. Multiple 5-μm sections were cut with a Micron microtome (HM355S) and stained by specific antibodies for immunohistochemistry (IHC) analysis.
The applied antibodies were subjected to in-house validation by the manufacturer for IHC analysis on paraffin-embedded material. The antibodies used in the study were FGF2 (rabbit polyclonal; AB1458; Chemicon; 1:200) and FGFR-1 (rabbit polyclonal; sc-121; Santa Cruz; 1:50). The IHC procedures for VEGFR-3 and PDGF-B have been described earlier.22–24
Sections were deparaffinised with xylene and rehydrated with ethanol. Antigen retrieval was performed by placing the specimen in 0.01M citrate buffer at pH 6.0 and exposed to microwave heating of 10 minutes at 250W (FGF2) or heated by pressure boiler of 2 minutes (FGFR-1). The DAKO EnVision + System-HRP (DAB) kit was used as endogen peroxidase blocking. As negative staining controls, the primary antibodies were replaced with the primary antibody diluent. For antibodies where the pathologists where uncertain about the specificity, based on morphologic criteria, isotype control was done (VEGFR-3). Primary antibodies were incubated for 30 minutes (FGF2) or 60 minutes (FGFR-1) in room temperature. The kit DAKO EnVision + System-HRP (DAB) was used to visualize the antigens. This was followed by application of liquid diaminobenzidine and substrate-chromogen, yielding a brown reaction product at the site of the target antigen. Finally, all slides were counterstained with hematoxylin to visualize the nuclei. For each antibody, included negative staining controls, all TMA staining were performed in a single experiment.
Scoring of Immunohistochemistry
By light microscopy, representative and viable tissue sections were scored semiquantitatively for cytoplasmic staining. The dominant staining intensity in both tumor cells and stromal cells was scored as: 0 = negative; 1 = weak; 2 = intermediate; 3 = strong (Figure 1). The cell density of the stroma was scored as: 1 = low density; 2 = intermediate density; 3 = high density (Figure 1). All samples were anonymised and independently scored by 2 pathologists (S.A.S. and K.A.S.). In case of disagreement the slides were reexamined and a consensus was reached by the observers. In most tumor cores as well as in some stromal cores there is a mixture of stromal cells and tumor cells. However, by morphologic criteria we have only scored staining intensity of tumor cells in tumor cores and intensity and density of stromal cells in stromal cores. When assessing a variable for a given core, the observers were blinded to the scores of the other variables and to outcome. The interobserver scoring agreement has previously been found valid in the same TMA-blocks for one ligand and one receptor with similar cytoplasmic staining.22 After categorizing into high and low expression group, the percentage discordance among the pathologists was: Tumor cell ligand 8%, stromal ligand 8%, tumor cell receptor 2%, and stromal receptor 4%. Mean score for duplicate cores from each individual was calculated separately in tumor cells and stroma. High expression in tumor cells was defined as score >1 (FGFR-1) or = 3 (FGF2). Stromal expression was calculated by summarizing density score (1–3) and intensity score (0–3) before categorizing into low and high expression. High stromal expression was defined as score ≥4.5 (FGF2) or >4 (FGFR-1). The predefined cutoff values for VEGFR-3 and PDGF-B,22–24 were used to estimate the coexpressions with FGF2 and FGFR-1.
All statistical analyses were done using the statistical package SPSS (Chicago, IL), version 15. The χ2 test and Fisher’s exact test were used to examine the association between molecular marker expression and various clinicopathological parameters. Univariate analysis was done by using the Kaplan-Meier method, and statistical significance between survival curves was assessed by the log-rank test. Disease-specific survival (DSS) was determined from the date of surgery to the time of lung cancer death. To assess the independent value of different pretreatment variables on survival, in the presence of other variables, multivariate analysis was carried out using the Cox proportional hazards model. Only variables of significant value from the univariate analysis were entered into the Cox regression analysis. Probability for stepwise entry and removal was set at 0.05 and 0.10, respectively. The significance level was set at p < 0.05.
Demographic, clinical, and histopathologic variables are shown in Table 1. The median age was 67 (range, 28–85) years and the majority of patients were male (75%). The NSCLC tumors comprised 191 squamous cell carcinomas (SCCs), 95 adenocarcinomas, 31 large-cell carcinomas (LCCs), and 18 bronchioalveolar carcinomas (BACs). Due to nodal metastasis or nonradical surgical margins, 59 patients (18%) received postoperative radiotherapy.
Expression of FGF2 and FGFR-1 and their Correlations
FGFR-1 and FGF2 were expressed in the cytoplasm of tumor cells. Based on morphologic criteria, FGFR-1 showed primarily moderate staining intensity in pneumocytes in control cores. Lymphocytes showed all degrees of staining intensity, approximately 1/3 was negative, 1/3 was weakly positive, and 1/3 moderately to strongly positive. Macrophages and plasma cells were stained strongly positive in both control and stromal cores. Bronchial epithelium showed moderate or positive staining intensity although endothelial blood vessel cells in control cores were mostly negative and tumor cells weakly positive. Fibroblast-like cells were weakly stained in both control cores and tumor tissue.
There was a moderate FGF2 expression in pneumocytes. Lymphocytes were approximately 50% weakly positive and 50% negative in control cores, while near all lymphocytes showed moderate to strong staining in tumor stroma. Plasma cells and macrophages showed moderate to strong staining in both control cores and tumor stroma. Bronchial epithelium showed weak staining intensity although the endothelium was primarily weakly positive in both control cores and tumor stroma. Fibroblast-like cells were generally weakly stained in both control cores and tumor tissue.
Tumor or stromal cell FGF2 or FGFR1 expression did not correlate with age, performance status, tumor differentiation, or vascular infiltration. Tumor cell FGF2 was more frequently expressed in node positive patients (high expression; N0 6%, N1 15%, N2 15%, p = 0.024). Besides, stromal expression of FGF2 was significant reduced in LCC (high expression; LCC 7%, BAC 28%, SCC 29%, adenocarcinomas 20%, p < 0.029), T2-tumors (high expression; T1 32%, T2 20%, T3 33%, p = 0.044) and tumors with positive surgical margins (high expression; margins free 26%, margins not free 7%, p = 0.016).
Tumor cell FGFR-1 was more frequently expressed in females (high expression; females 82%, males 64%, p = 0.004), in patients without weight loss (high expression; weight loss 50%, no weight loss 70%) and in BAC and adenocarcinomas (high expression; LCC 58%, BAC 89%, SCC 62%, adenocarcinomas 82%, p < 0.001). Stromal FGFR-1 was more often expressed in tumors with T1-status (high expression; T1 33%, T2 21%, T3 18%, p = 0.049).
Among the clinical variables, shown in Table 1, performance status (p = 0.04), differentiation (p = 0.001), surgical procedure (p = 0.0009), pathologic stage (p < 0.0001), T-stage (p = 0.002), N-stage (p < 0.0001), vascular infiltration (p = 0.0005), and postoperative radiotherapy (p = 0.002) were all significant prognostic indicators for DSS. The influence on survival by tumor cell and stromal expression of FGF2 and FGFR-1 are shown in Table 2 and Figure 2. In univariate analyses, tumor cell FGF2 expression (p = 0.015; Figure 2A) and stromal FGF2 expression (p = 0.024; Figure 2B) were prognostic, but opposite indicators for DSS.
Multivariate Cox Proportional Hazards Analysis
Results of the multivariate analysis are presented in Table 3. Including significant clinicopathological and angiogenic variables from the univariate analysis, tumor cell FGF2 (p = 0.038), stromal FGF2 expression (p = 0.015), performance status (p = 0.012), pathologic T-stage (p = 0.02), N-stage (p < 0.001), histologic differentiation (p = 0.042), and vascular infiltration (p = 0.005) seemed as independent prognostic factors.
Coexpression of FGF2/FGFR1 with VEGFR-3 and PDGF-B
Table 2 and Figure 3 show the DSS rates of the patients stratified into four groups according to the basis of high or low FGF2 or FGFR1 expression versus a high or low VEGFR-3 or PDGF-B expression. The coexpression of tumor cell FGF2/VEGFR-3 (p < 0.001), FGF2/PDGF-B (p = 0.002), FGFR-1/VEGFR-3 (p = 0.001), and FGFR-1/PDGF-B (p = 0.002), were all significant prognostic indicators for DSS.
Examining the same combinations of stromal coexpressions, there were no significant associations with survival (stromal FGF2/VEGFR-3, p = 0.10; stromal FGF2/PDGF-B, p = 0.052; stromal FGFR-1/VEGFR-3, p = 0.73; stromal FGFR-1/PDGF-B, p = 0.24).
The coexpression of tumor cell FGF2/VEGFR-3 (high FGF2/high VEGFR-3: N0 2,2%; N1 9,3%; N2 14,8%, p < 0.001) and FGFR-1/VEGFR-3 (high FGFR-1/high VEGFR-3: N0 25%; N1 30%; N2 70%, p = 0.001) correlated significantly with lymph node metastasis, whereas the coexpression of tumor cell FGF2/PDGF-B (p = 0.07) and FGFR-1/PDGF-B (p = 0.09) tended to, but did not reach a significant level.
We present a large-scale study in an unselected population of surgically resected NSCLC patients using high-throughput TMA methodology to examine the prognostic impact of FGF2 and FGFR-1 and their coexpressions with VEGFR-3 and PDGF-B in both tumor cells and stroma. High tumor cell FGF2 expression is an independent negative prognostic indicator for DSS, although high stromal FGF2 expression correlates with a good prognosis. Interestingly, tumor cell coexpressions of both FGF2/VEGFR-3 and of FGFR-1/PDGF-B correlated significantly with a poor prognosis.
The prognostic impact of FGF2 in NSCLC is still controversial. Corroborating our results, some previous studies have found tumor cell FGF2 expression to be a negative prognostic factor.8,10,11 In a cohort of 119 resected NSCLC patients, Shou et al.,10 reported FGF2 as a negative prognosticator, though only in the univariate analysis. However, in a study of 167 stage I-IV NSCLC adenocarcinomas,11 FGF2 seemed as an independent indicator of poor prognosis while FGFR-1 had a negative prognostic impact in the univariate analysis. Using frozen tissue and enzyme-linked immunosorbent assay technique, in a cohort of 71 resected NSCLC patients, Iwasaki et al.8 observed that FGF2 had an independent negative impact on survival. In contrast, other studies revealed no correlation between tumor cell FGF2 expression and survival.9,12 In a relatively large study, involving 206 resected NSCLC tumors, Volm et al.12 found FGFR-1, but not FGF2, in univariate analysis to correlate with a poor prognosis. In addition, Kojima et al.9 did not observe a negative prognostic impact of tumor FGF2 expression in a cohort of 132 stage I NSCLC patients.
Studies on FGF2 serum levels in NSCLC patients have been contradictory with respect to prognostic relevance.14–18 One study reported high serum level of FGF2 to indicate a favorable prognosis.14 The latter may be explained by our finding of high stromal FGF2 expression as a favorable prognostic indicator. It can be argued that both stromal and tumor cell FGF2 may contribute to the serum level of FGF2. To our knowledge, this is the first study reporting stromal FGF2 expression to correlate with a good prognosis in NSCLC. Nevertheless, in 84 stage I-IIIA NSCLC patients, Guddo et al.13 reported stromal FGF2 to inversely correlate with lymph node metastasis, indicating an inhibitory role in NSCLC progression. Corroborating the findings by Guddo et al.,13 we have previously reported stromal VEGFs and VEGFRs to be correlated with increased survival,22 though the mechanisms behind these findings is not fully understood.
It has to be noted that the stromal expression of each marker is the total expression of the different stromal components, including lymphocytes, macrophages, granulocytes and fibroblast-like cells. Thus, the stromal FGF2 expression may be linked to one or more stromal cell types. Almost all lymphocytes were moderately to strongly FGF2 positive in tumor stroma and activation of the adaptive immune system may suppress malignant cell proliferation.26 Hence, high stromal FGF2 may to some extent reflect activation of the adaptive immune system, which corroborate our previous results in this cohort.27
This is the first study investigating the prognostic impact of the coexpression of FGF2 and VEGFR-3 in a large cohort of cancer patients. Beyond being expressed in lymphatics, VEGFR-3 is also up-regulated in blood vessels in several cancers.28,29 FGF2 is well established as an important mediator in angiogenesis, but also considered of importance in lymphangiogenesis. Actually, FGF2 pellets implanted in mouse cornea demonstrated the lymphatics to be more responsive to FGF2 than the blood vessels.30 It has been demonstrated that cross-talk between VEGFs and FGFs may occur in both hemangiogenesis and lymphanigiogenesis.4 In a study by Chang et al.31 it was proposed that the lymphatic activity of FGF2 is mediated by endogenous VEGF-C and VEGF-D, leading to VEGFR-3 activation. Hence, in the study by Kubo et al.21 administration of anti-VEGFR-3 antibodies inhibited the FGF2 lymphangiogenesis in mouse cornea.
Tumor lymphangiogenesis has been associated with lymphatic metastasis, although the precise mechanism is not fully understood.21,32 Albeit only 16 patients were in the subgroup of high FGF2/high VEGFR-3 expression, this coexpression of FGF2 and VEGFR-3 seems strongly associated with poor survival. In our previous reports, tumor cell VEGFR-3 expression correlated with both nodal status and survival.22,23 Herein, we find FGF2 alone and the coexpression of FGF2/VEGFR-3 and FGFR-1/VEGFR-3 to be significantly associated with lymph node metastasis. Actually, in the group of N2 patients, 70% of the patients (19 of 27) had high FGFR-1/high VEGFR-3 expression, indicating lymphangiogenesis as a plausible contributor to poor survival.
Of interest, Nissen et al.20 recently reported a reciprocal interaction between FGF2 and PDGF-B in a murine tumor model, leading to neovascularization and metastasis. The simultaneous overexpression of FGF2 and PDGF-B resulted in a formation of high-density primitive vascular plexuses, which were poorly coated with pericytes and vascular smooth muscle cells (VSMCs). The underlying mechanisms of this reciprocal interaction involve FGF2 associated up-regulation of PDGF receptors in endothelial cells and PDGF-B associated up-regulation of FGFR-1 in VSMCs. In our study, there is coexpression of high FGF2/high PDGF-B in only 8 patients. But these patients had a significantly shortened survival (37% 5-year survival) when compared with the low FGF2/low PDGF-B group (62% 5-year survival, p = 0.002). The fact that 86% of patients (24 of 28) with high tumor cell FGF2 expression also had high tumor cell FGFR-1 expression indicates an autocrine loop in the tumor cells. Although these findings are related to tumor cell expression, we may speculate that high production of these angiogenic factors in the tumor cells may also act in a paracrine fashion to stimulate pericytes, VSMCs and endothelial cells.
The VEGFs, PDGFs, and FGFs are all essential in tumor development and different novel targeted therapies aim to inhibit one ore more of these angiogenic markers. Herein, tumor cell FGF2 expression emerged as an independent negative prognostic factor for stage I-IIIA NSCLC while high stromal FGF2 expression favors a good prognosis. Supporting previous preclinical findings, we have for the first time shown that coexpressions of FGF2/VEGFR-3 and FGFR-1/PDGF-B seem to be significant prognosticators in NSCLC. Based on these results, a multitargeted antiangiogenic approach may be more promising than inhibiting single targets in the treatment of NSCLC patients.
1. Jemal A, Siegel R, Ward E, Murray T, Xu J, Thun MJ. Cancer statistics, 2007. CA Cancer J Clin
2. Flehinger BJ, Kimmel M, Melamed MR. The effect of surgical treatment on survival from early lung cancer. Implications for screening. Chest
3. Mountain CF. Revisions in the International System for Staging Lung Cancer. Chest
4. Presta M, Dell’Era P, Mitola S, Moroni E, Ronca R, Rusnati M. Fibroblast growth factor/fibroblast growth factor receptor system in angiogenesis. Cytokine Growth Factor Rev
5. Berger W, Setinek U, Mohr T, et al. Evidence for a role of FGF-2 and FGF receptors in the proliferation of non-small cell lung cancer cells. Int J Cancer
6. Gualandris A, Rusnati M, Belleri M, et al. Basic fibroblast growth factor overexpression in endothelial cells: an autocrine mechanism for angiogenesis and angioproliferative diseases. Cell Growth Differ
7. Bremnes RM, Camps C, Sirera R. Angiogenesis in non-small cell lung cancer: the prognostic impact of neoangiogenesis and the cytokines VEGF and bFGF in tumours and blood. Lung Cancer
8. Iwasaki A, Kuwahara M, Yoshinaga Y, Shirakusa T. Basic fibroblast growth factor (bFGF) and vascular endothelial growth factor (VEGF) levels, as prognostic indicators in NSCLC. Eur J Cardiothorac Surg
9. Kojima H, Shijubo N, Abe S. Thymidine phosphorylase and vascular endothelial growth factor in patients with Stage I lung adenocarcinoma. Cancer
10. Shou Y, Hirano T, Gong Y, et al. Influence of angiogenetic factors and matrix metalloproteinases upon tumour progression in non-small-cell lung cancer. Br J Cancer
11. Takanami I, Tanaka F, Hashizume T, et al. The basic fibroblast growth factor and its receptor in pulmonary adenocarcinomas: an investigation of their expression as prognostic markers. Eur J Cancer
12. Volm M, Koomagi R, Mattern J, Stammler G. Prognostic value of basic fibroblast growth factor and its receptor (FGFR-1) in patients with non-small cell lung carcinomas. Eur J Cancer
13. Guddo F, Fontanini G, Reina C, Vignola AM, Angeletti A, Bonsignore G. The expression of basic fibroblast growth factor (bFGF) in tumor-associated stromal cells and vessels is inversely correlated with non-small cell lung cancer progression. Hum Pathol
14. Brattstrom D, Bergqvist M, Larsson A, et al. Basic fibroblast growth factor and vascular endothelial growth factor in sera from non-small cell lung cancer patients. Anticancer Res
15. Brattstrom D, Bergqvist M, Hesselius P, et al. Elevated preoperative serum levels of angiogenic cytokines correlate to larger primary tumours and poorer survival in non-small cell lung cancer patients. Lung Cancer
16. Brattstrom D, Bergqvist M, Hesselius P, Larsson A, Wagenius G, Brodin O. Serum VEGF and bFGF adds prognostic information in patients with normal platelet counts when sampled before, during and after treatment for locally advanced non-small cell lung cancer. Lung Cancer
17. Joensuu H, Anttonen A, Eriksson M, et al. Soluble syndecan-1 and serum basic fibroblast growth factor are new prognostic factors in lung cancer. Cancer Res
18. Ueno K, Inoue Y, Kawaguchi T, Hosoe S, Kawahara M. Increased serum levels of basic fibroblast growth factor in lung cancer patients: relevance to response of therapy and prognosis. Lung Cancer
19. Arbiser JL. Why targeted therapy hasn’t worked in advanced cancer. J Clin Invest
20. Nissen LJ, Cao R, Hedlund EM, et al. Angiogenic factors FGF2 and PDGF-BB synergistically promote murine tumor neovascularization and metastasis. J Clin Invest
21. Kubo H, Cao R, Brakenhielm E, Makinen T, Cao Y, Alitalo K. Blockade of vascular endothelial growth factor receptor-3 signaling inhibits fibroblast growth factor-2-induced lymphangiogenesis in mouse cornea. Proc Natl Acad Sci U S A
22. Donnem T, Al-Saad S, Al-Shibli K, et al. Inverse Prognostic impact of angiogenic marker expression in tumor cells versus stromal cells in non small cell lung cancer. Clin Cancer Res
23. Donnem T, Al-Shibli K, Al-Saad S, Delghandi MP, Busund LT, Bremnes RM. VEGF-A and VEGFR-3 correlate with nodal status in operable non-small cell lung cancer: inverse correlation between expression in tumor and stromal cells. Lung Cancer
24. Donnem T, Al-Saad S, Al-Shibli K, Andersen S, Busund LT, Bremnes RM. Prognostic impact of platelet-derived growth factors in NSCLC tumor and stromal cells. J Thorac Oncol
25. World Health Organization. Histological Typing of Lung and Pleural Tumours, 3 Ed. Geneva, Switzerland: Springer-Verlag, 1999.
26. de Visser KE, Eichten A, Coussens LM. Paradoxical roles of the immune system during cancer development. Nat Rev Cancer
27. Al-Shibli K, Donnem T, Al-Saad S, Persson M, Bremnes RM, Busund LT. Prognostic impact of epithelial and stromal lymphocyte infiltration in non-small cell lung cancer. Clin Cancer Res
28. Saaristo A, Karpanen T, Alitalo K. Mechanisms of angiogenesis and their use in the inhibition of tumor growth and metastasis. Oncogene
29. Valtola R, Salven P, Heikkila P, et al. VEGFR-3 and its ligand VEGF-C are associated with angiogenesis in breast cancer. Am J Pathol
30. Chang L, Kaipainen A, Folkman J. Lymphangiogenesis new mechanisms. Ann N Y Acad Sci
31. Chang LK, Garcia-Cardena G, Farnebo F, et al. Dose-dependent response of FGF-2 for lymphangiogenesis. Proc Natl Acad Sci U S A
32. Alitalo K, Tammela T, Petrova TV. Lymphangiogenesis in development and human disease. Nature