Abstract
Understanding the complex process of tumor metastasis is a problem which has challenged both clinician and scientist for well over 100 years. Defining molecular markers which reflect the metastatic potential of a tumor has also proved elusive. Recently, members of the vascular endothelial growth factor (VEGF) family of glycoproteins have been demonstrated to be potent mediators of both blood vessel and lymphatic vessel formation in the context of tumor biology. Experimental studies in animal models combined with extensive clinicopathological data provide a compelling case indicating that members of the VEGF family play a key role in the formation of metastases in a broad range of solid tumors. The question of whether VEGF signaling pathways can now serve as therapeutic targets alone, or in combination with other forms of anti-cancer agents, needs to be addressed.
The concept that tumors metastasize to other organ systems in the body, not by chance, but because of specific influences, was first speculated by the English surgeon Stephen Paget in his now famous “soil and seed” hypothesis (1, 2). Since this time others have proposed that factors such as the anatomical structure or mechanical function of the vascular system could be influential in tumor spread (2, 3). More recent intensive study of the cellular and molecular control of metastasis has allowed the elucidation of many defined stages of tumor metastasis (2, 4). With this has come the definition of a range of molecules, including receptors, growth factors, enzymes, matrix proteins and transcription factors, which control the growth, invasion, migration and establishment of metastatic tumor cells (5–7).
The discovery and analysis of the molecular systems which control the formation of blood vessels and lymphatics have provided great insight into the potential for angiogenesis and lymphangiogenesis (the growth of lymphatic vessels) to influence tumor growth and spread. The VEGF/VEGF receptor (VEGFR) families of ligands and cognate cell surface receptors are important mediators of both angiogenesis and lymphangiogenesis in tumors (8, 9). Mounting evidence from experimental data, clinicopathological correlations and clinical trials indicates that these pathways are critical components for the development of primary and metastatic cancer (8, 9). This review will summarise the key recent findings which demonstrate that the lymphangiogenic subfamily of the VEGF ligands, consisting of VEGF-C and VEGF-D, are markers of tumor growth and metastasis, and should be considered important potential targets for anti-metastatic therapy.
THE VEGF AND VEGF RECEPTOR SYSTEM
The VEGF family of ligands consists of a group of secreted glycoproteins which activate cell surface receptors that regulate angiogenesis and lymphangiogenesis (10). The original family member VEGF (also called vascular permeability factor, VPF (11) or VEGF-A) is a potent mitogen for endothelial cells, an angiogenic growth factor (12), and has the capacity to rapidly induce the release of fluid from blood vessels to the interstitial space (13). VEGF activates two cell surface receptor tyrosine kinases, VEGFR-1 (also known as Flt1) (14) and VEGFR-2 (also known as Flk1 in mouse and KDR in human) (15), that are critical for the formation of the vascular system in mammals (16–18). VEGF can also bind the receptors neuropilin-1 (NP-1) and NP-2 that appear to act as co-receptors for VEGFR-2 (19, 20).
A large body of experimental data demonstrates the critical role of VEGF in blood vessel formation (reviewed in (10, 21, 22)). Perhaps the most compelling pieces of evidence for the role of VEGF in blood vessel formation come from the study of two biological models: the deletion of the mouse Vegf gene in vivo and the inhibition of VEGF action in solid tumors. Studies which examined the consequences of Vegf gene deletion in the mouse demonstrated that elimination of one Vegf allele is sufficient to generate an embryonic lethal phenotype due to failure of the blood vascular system to develop normally (23, 24). Therefore, the absolute protein levels of VEGF are so critical to the development of the animal that small variations in concentration are sufficient to abort the developmental process. A second approach involved the generation of inhibitory monoclonal antibodies to VEGF and the demonstration that they have the capacity to inhibit the growth of solid tumours in vivo by preventing the formation of new blood vessels driven by tumor–derived VEGF (25). Initial studies in mouse models have led the way to the generation of a humanised version of the anti-VEGF antibodies, the efficacy of which has been demonstrated in phase III clinical trials in patients with metastatic colorectal cancer (26). Thus, a critical role for VEGF in driving tumor angiogenesis and solid tumor growth has been conclusively demonstrated in animal models and validated by clinical trials in cancer patients.
New members of the mammalian VEGF family have been defined which have been designated placenta growth factor (PlGF) (27), VEGF-B (28), VEGF-C (29) and VEGF-D (30), each encoded by a separate gene. In addition, novel VEGFs have been described from viral sources (31–33) and in snake venom (34). VEGF-C and VEGF-D have been defined as a subgroup within the VEGF family due to their distinct structural features and receptor specificity (29, 30). These growth factors share, with the other VEGFs, a highly conserved central structural domain, the so-called VEGF homology domain (VHD) containing receptor binding sites and a cystine knot motif (10, 35, 36), but in addition have long N- and C-terminal propeptides that are cleaved from the VHD to generate the mature form of the growth factors that bind receptors with high affinity (37, 38). Recent studies have demonstrated that this cleavage is mediated at least in part by enzymes such as plasmin and furins (39, 40). VEGF-C and VEGF-D are the only members of the VEGF family that bind VEGFR-3 (also known as Flt4) (29, 30), a cell surface receptor tyrosine kinase expressed on the endothelium of lymphatic vessels (41). Activation of VEGFR-3 by VEGF-C or VEGF-D is sufficient to induce lymphangiogenesis (42) and VEGF-C is critical for development of the lymphatics as VEGF-C null mice exhibit edema and die before birth whereas mice heterozygous for VEGF-C deficiency develop chylous fluid in the abdomen after birth due to dysfunction of the lymphatic vasculature (43). The proteolytically activated forms of human VEGF-C and VEGF-D can also activate VEGFR-2 (29, 30) and thereby promote angiogenesis (44–49). However, in contrast to human VEGF-D, mature mouse VEGF-D does not bind VEGFR-2 (50–52).
EXPERIMENTAL DEMONSTRATION OF VEGF-C AND VEGF-D ACTION IN CANCER
The important role of VEGF in promoting tumor angiogenesis has been known for some years, however, studies of VEGF-C and VEGF-D in tumor angiogenesis and lymphangiogenesis have been more recent. The direct demonstration of these properties was shown in a series of studies in which these growth factors were expressed in transgenic mouse tumor models or in tumor cell lines that were used to generate tumors in immunocompromised mice. Skobe and co-workers demonstrated that expression of VEGF-C in the breast cancer cell line MBA-MD-435 induced lymphangiogenesis, but not angiogenesis, in tumors grown in mice (53). The tumors expressing VEGF-C metastasized to local lymph nodes and lung whereas control tumors lacking VEGF-C did not. This finding suggested that tumor lymphangiogenesis induced by VEGF-C can promote lymphatic metastasis. Analysis of the effect of VEGF-C in tumors generated in SCID mice with another breast cancer cell line, MCF-7, revealed that VEGF-C promoted formation of tumor-associated lymphatics, but not angiogenesis (54). Inhibition of this lymphangiogenesis using a soluble form of VEGFR-3 demonstrated the potential of inhibiting metastasis with reagents that block the VEGFR-3 signaling pathway (54). In another study, Rip1Tag2 transgenic mice were used to analyse the effect of VEGF-C in a pancreatic β cell tumor model (55). In this model, VEGF-C induced development of peri-tumoral lymphatics which correlated with an increased rate of metastatic spread to draining pancreatic lymph nodes. The effect of VEGF-C was also studied in tumors established with the MeWo melanoma cell line (56). In this system VEGF-C promoted both lymphangiogenesis and angiogenesis, in contrast to the other VEGF-C tumor models in which only lymphangiogenesis was observed. This discrepancy may be due to different degrees of proteolytic processing of VEGF-C in the various tumor models, an important consideration because it seems that only the mature, fully processed form of VEGF-C can readily activate VEGFR-2 (38), and thereby induce angiogenesis, in vivo. VEGF-C also promoted recruitment of macrophages in the MeWo tumor model, suggesting a potential function of this growth factor in immune modulation (56). Finally, overexpression of soluble VEGFR-3 (VEGFR-3-Ig) by stably transfected LNM35 cells, a human lung cancer cell line selected for a highly metastatic phenotype and expressing high levels of endogenous VEGF-C, inhibited tumor lymphangiogenesis and lymph node metastasis when cells were grown as tumors in immunodeficient mice, effects that were also achieved by adenoviral delivery of VEGFR-3-Ig (54, 57). A similar approach led to inhibition of lymph node metastasis in a syngeneic mammary tumor model in immunocompetent rats (58).
A study employing a tumor xenograft model also indicated an important role for VEGF-D in tumor biology (46). When 293EBNA cells expressing VEGF-D were grown as tumors in SCID/NOD mice, the tumors contained lymphatic vessels and exhibited metastatic spread to lymph nodes whereas control tumors lacking VEGF-D did not. Other controls expressing VEGF exhibited enhanced angiogenesis but did not contain lymphatic vessels and did not spread to lymph nodes. The dependence of the observed lymphangiogenesis and lymphatic metastasis on VEGF-D was conclusively demonstrated by treatment of tumors with a neutralizing antibody that blocked the binding of VEGF-D to both VEGFR-2 and VEGFR-3 (59). This antibody potently inhibited the metastatic spread of the tumors. VEGF-D also promoted angiogenesis and more rapid tumor growth in this model, processes that were both blocked by the neutralizing antibody demonstrating the potential for an anti-metastatic approach (46, 60).
CLINICOPATHOLOGICAL CORRELATIONS
Since the definition of VEGF-C and VEGF-D as activating ligands for the lymphangiogenic receptor VEGFR-3 in 1996 and 1998, respectively (29, 30), it has been speculated that these growth factors might stimulate vessel formation in the context of tumor biology. Prior to the generation of evidence in animal models that these growth factors are involved in lymphogenous metastasis (discussed in the previous section), researchers had already begun to determine if the expression of these glycoproteins correlated with the growth or spread of solid tumors in humans. These studies, which have spanned the past 6 years, provide a broad analysis of human carcinomas derived from various tissues of origin (9, 61). These studies have been summarized in Table 1 and a number of observations are worthy of note. Firstly, these studies have come from a broad range of research and clinical laboratories in different regions throughout the world and therefore encompass tumors derived from different genetic backgrounds and exposed to diverse environmental factors. In addition, the studies involved tumors derived from cells of many different origins (e.g. lung, breast, colorectal, thyroid, gastric, prostate, cervical, endometrial, esophageal, melanoma and head & neck cancers have been analysed), which effectively covers the broad range of cell types from which most carcinomas are derived. Therefore, the breadth of these studies now allows us to make more confident predictions about the utility of VEGF-C and VEGF-D as prognostic markers of solid tumor metastasis.
| Tumor type | Technique | Correlations | Reference |
|---|---|---|---|
| Thyroid | RT-PCRa | VEGF-C expression correlates with lymph node invasive tumors | (71) |
| Thyroid | RT-PCR | VEGF-C expression is increased in papillary thyroid cancer | (72) |
| Thyroid | RT-PCR | Higher VEGF-C mRNA detected in papillary carcinoma with nodal involvement than in tumors without nodal involvement | (73) |
| Gastric | RT-PCR | Correlation between VEGF-C expression pattern and lymph node status | (74) |
| Gastric | IHCb | VEGF-C expression higher in lymphatic invasion–positive early gastric cancer | (75) |
| Gastric | IHC | VEGF-C expression correlated with lymphatic and venous invasion and had negative impact on prognosis in patients who did not express VEGF | (76) |
| Gastric | RT-PCR, IHC | Positive correlation between incidence of VEGF-C and VEGFR-3 mRNA in primary tumors; number of VEGFR-3-positive vessels was related to grade of lymphatic invasion | (77) |
| Gastric | IHC | VEGF-C and VEGF-D correlated with lymphatic invasion in adenocarcinomas of undifferentiated type | (78) |
| Gastric | IHC | VEGF-C was associated with tumor invasion, lymphatic invasion and lymph node metastasis | (79) |
| Pancreas | IHC | VEGF-C associated with increased lymphatic vessel invasion and lymph node metastasis | (80) |
| Prostate | ISHc | Higher expression of VEGF-C mRNA in lymph node positive patients | (81) |
| Gallbladder | IHC | VEGF-C correlated with lymphatic vessel involvement, lymph node metastasis and poor outcome | (82) |
| Lung | RT-PCR | Lymph node metastasis associated with high VEGF-C expression and low VEGF-D expression | (83) |
| Lung | RT-PCR | Expression of both VEGF and VEGF-C in primary tumor was significantly associated with nodal microdissemination | (84) |
| Lung | IHC | VEGF-C expression significantly associated with lymph node metastasis, lymphatic vessel invasion and poor outcome for patient | (85, 86) |
| Colorectal | RT-PCR | VEGF-C mRNA correlated with lymph node metastasis, lymphatic involvement and depth of invasion | (87) |
| Colorectal | RT-PCR, IHC | VEGF mRNA, but not VEGF-C and VEGF-D mRNA, correlated with lymphatic metastases | (88) |
| Colorectal | IHC | VEGF-D is an independent prognostic factor for both disease-free and overall survival and its association with disease outcome may be via the promotion of lymphatic involvement/metastases | (63) |
| Colorectal | RT-PCR, IHC | VEGF-C mRNA correlated with lymph node metastasis, lymphatic involvement and invasion depth | (89, 90) |
| Colorectal | IHC | VEGF-C correlated with poorer histologic grade, lymphatic invasion, depth of invasion, lymph node metastasis, venous invasion, liver metastasis and Dukes' stage | (91) |
| Colorectal | RT-PCR | VEGF-D more highly expressed in tumor tissues | (92) |
| Cervical | RT-PCR, IHC | VEGF-C expression was the sole independent factor influencing pelvic lymph node metastases; patients with VEGF-C expression had poorer prognosis | (66, 93, 94) |
| Esophageal | IHC | VEGF-C expression correlated with depth of tumor invasion, tumor stage, venous invasion, lymphatic invasion and lymph node metastasis | (95) |
| Esophageal | RT-PCR, IHC | VEGF-C correlates with depth of tumor, lymph node metastasis, lymphatic involvement, pathological stage and poor prognosis | (96) |
| Endometrial | IHC | VEGF-C correlated with depth of invasion, vascular invasion, lymphatic vessel invasion and lymph node metastasis | (97) |
| Endometrial | IHC | Presence of VEGF-D and VEGFR-3 may predict myometrial invasion, lymph node metastasis and patients at increased risk for poor survival | (98) |
| Breast | RT-PCR, IHC | VEGF-C correlated with lymphatic invasion; 5-year disease-free survival of VEGF-C-positive group was poorer | (99) |
| Breast | IHC | VEGF-D correlates with lymph node metastasis and is associated with disease-free and overall survival | (64) |
| Breast | RT-PCR, IHC | VEGF-C expression correlated with lymphatic vessel invasion; disease-free survival of VEGF-C-positive group was significantly poorer | (100) |
| Ovary | IHC | VEGF-D is associated with lymph node metastasis and peritoneal metastasis outside the pelvis and is an independent predictor of poor outcome | (65) |
| Melanoma | RT-PCR, IHC | VEGF-C correlates with lymph node localisation | (101) |
| Melanoma | IHC | VEGF-C did not correlate with lymphatic vessel density | (102) |
| Head and neck | RT-PCR | Increased expression of all four isoforms of VEGF and of VEGF-C, but decreased expression of VEGF-D, in tumors versus normal epithelium; levels of VEGF-C had predictive value for cervical node metastases | (103) |
| Neuro- blastoma | RT-PCR | No correlation between VEGF-C and lymph node metastasis |
- a Reverse transcription polymerase chain reaction used to detect mRNA in tissue samples.
- b Immunohistochemistry used to detect protein in tissue sections.
- c In situ hybridisation used to detect mRNA in tissue sections.
The majority of these studies (approximately 80%) show a strong positive relationship between the expression of VEGF-C and VEGF-D protein or mRNA and a range of clinicopathological parameters which reflect the prognosis of the patient (e.g. lymph node metastasis, tissue invasion, patient survival and pathological stage). Moreover, in some of these studies the expression of VEGF-C and VEGF-D were shown to be independent prognostic factors. For example, White and co-workers have reported that VEGF-D, which is encoded by an X-linked gene (62), is an independent prognostic factor for both disease-free and overall survival in colorectal cancer and that association of VEGF-D with disease outcome may be via the promotion of lymphatic involvement/metastases (63). VEGF-D expression has also been reported to be associated with lymph node metastasis in breast carcinoma and may be a novel prognostic factor for this disease (64) and to be a novel prognostic factor in epithelial ovarian carcinoma (65). Likewise, VEGF-C expression in cervical cancer was reported to be an independent prognostic factor influencing pelvic lymph node metastasis and patient outcome (66). In summary, the clinicopathological data implicating VEGF-C and VEGF-D as drivers of lymphangiogenesis and lymphogenous metastasis in human cancer are compelling.
Mode of detection of VEGF-C and VEGF-D in cancer
Prior to the development of specific antibodies to VEGF-C and VEGF-D research groups had used PCR to determine the level of gene expression in clinical tissue samples. Although an acceptable technique in terms of allowing the specific detection of mRNA, it has the disadvantage that it does not always reflect the actual protein levels within a tumor. It is also impossible to assess the location of the mRNA being detected, whether this is in tumor cells, tumor-associated stroma or immediately adjacent normal tissue. The development of specific antisera and monoclonal antibodies to VEGF-C and VEGF-D has facilitated the examination of protein levels in human tumors using immunohistochemistry (see Table 1). The most applicable antibodies for this purpose are those that recognize the central VEGF homology domain containing the cystine-knot motif and receptor binding sites (59). Such antibodies will detect both fully and partially processed forms of VEGF-C and VEGF-D containing the VHD. In contrast, use of antibodies targeting the N- or C-terminal propeptides of VEGF-C or VEGF-D is problematic because these reagents can monitor the propeptides after they have been cleaved from the VHD and therefore give misleading results as to the location of bioactive growth factor. Furthermore, antibodies to the propeptides do not detect the mature, fully processed forms of VEGF-C and VEGF-D.
Although use of antibodies that target the VHDs of VEGF-C and VEGF-D is currently the preferred approach for assessing the localization of bioactive growth factors in tumor tissue, even these reagents do not discriminate between the unprocessed and fully processed forms. Development of antibodies targeting the VHD that bind specifically to the mature forms would be of benefit in interpreting the effects of VEGF-C and VEGF-D on tumor angiogenesis and lymphangiogenesis.
THERAPEUTIC STRATEGIES
Evidence arising from both experimental animal models and clinicopathological correlations strongly suggests that both VEGF-C and VEGF-D are key players in mediating steps of the metastatic process. Therefore, inhibiting the angiogenic and lymphangiogenic activities of these two growth factors by antagonizing their capacity to interact with VEGFR-2 and VEGFR-3 provides attractive strategies for therapeutic intervention designed to block metastatic spread (60, 67). Our initial work using an experimental mouse model in which tumors expressing VEGF-D were treated with intraperitoneal injections of a neutralizing VEGF-D monoclonal antibody (designated VD1) gave proof-of-principle that inhibition of VEGF-D-mediated signaling could reduce the rate of growth of primary tumors and the development of lymph node metastasis (46, 60). The recent success using the humanized anti-VEGF monoclonal antibody (AvastinTM, bevacizumab) in phase III clinical trial with patients suffering from metastatic colorectal cancer (26), and subsequent FDA approval (February 2004), has already indicated the potential for blocking angiogenic signaling in the clinic with antibody-based reagents. Therefore, inhibition of VEGF-C and VEGF-D using neutralizing antibodies directed against these glycoproteins is one attractive approach for clinical development of anti-lymphangiogenic/angiogenic agents.
Another promising approach for blocking tumor lymphangiogenesis would be to use a soluble version of VEGFR-3 to sequester all forms of VEGF-C and VEGF-D and thereby render them unable to activate receptors on the endothelium. This approach has shown great promise in a range of animal models. For example, adenoviral delivery of a construct consisting of the first three immunoglobulin homology domains of VEGFR-3 fused to the Fc-domain of the human immunoglobulin γ chain blocked the growth of peritumoral lymphatic vessels in a mouse breast cancer model (54). Alternative approaches could involve use of peptidomimetics (MGA, SAS and Tony Hughes unpublished data) that block the binding of VEGF-C and VEGF-D to their receptors, neutralizing antibodies directed to VEGFR-2 (68) and VEGFR-3 or orally active small molecule inhibitors of the VEGFR-2 and VEGFR-3 tyrosine kinases (69, 70).
CONCLUSIONS
The data from an extensive range of animal models and clinicopathological studies have established that VEGF-C and VEGF-D, signalling via VEGFR-2 and VEGFR-3, can promote tumor angiogenesis, lymphangiogenesis, tumor growth and metastatic spread. The challenge is now to develop inhibitors of these signalling pathways that are appropriate for clinical development. Such reagents will surely include soluble receptors, neutralizing antibodies directed at the growth factors/receptors and small molecule inhibitors of the tyrosine kinases associated with the receptors. The next few years will see much energy devoted to both reagent development and clinical testing that will provide clear direction as to the importance of VEGF-C/D signalling as a target for cancer therapeutics. The recent success in advanced clinical trials with AvastinTM (a neutralizing VEGF antibody) for treating patients with metastatic colorectal cancer is the first clinical success with an anti-angiogenic agent and provides optimism that blocking the action of other members of the VEGF family will ultimately bring benefit to cancer patients.
SAS and MGA are supported by a Program Grant from the National Health and Medical Research Council of Australia (NHMRC) and Senior Research Fellowships from the Pharmacia Foundation and the NHMRC, respectively.
