Metastases are commonly found in the lymphatic system. The molecular mechanism of lymphatic metastasis is, however, poorly understood. Here we report that vascular endothelial growth factor (VEGF)-A stimulated lymphangiogenesis in vivo and that overexpression of VEGF-A in murine T241 fibrosarcomas induced the growth of peritumoral lymphatic vessels, which occasionally penetrated into the tumor tissue. As a result of peritumoral lymphangiogenesis, metastases in lymph nodes of mice were detected. VEGF-A–overexpressing tumors contained high numbers of infiltrating inflammatory cells such as macrophages, which are known to express VEGF receptor (VEGFR)-1. It seemed that in the mouse cornea, VEGF-A stimulated lymphangiogenesis through a VEGF-C/-D/VEGFR-3–independent pathway as a VEGFR-3 antagonist selectively inhibited VEGF-C–induced, but not VEGF-A–induced, lymphangiogenesis. Our data show that VEGF-A contributes to lymphatic mestastasis. Thus, blockage of VEGF-A–induced lymphangiogenesis may provide a novel approach for prevention and treatment of lymphatic metastasis.
Metastasis is a hallmark of malignancy that distinguishes invasive tumors from benign ones and constitutes a major cause of cancer mortality. In certain types of cancers, such as breast cancer, lymphatic metastasis is the major route for the spread of cancer cells (1). Spontaneous lymphatic metastases are the consequence of a complex metastatic process that includes (a) dissemination of malignant cells from a primary tumor to the lymphatics; (b) transport of tumor cells via the lymphatics to the local lymph nodes; (c) settlement of tumor cells in the lymph nodes; and (d) growth of metastatic tumors in the lymph nodes. The complexity of this process implies the existence of multiple steps controlling cancer metastasis. Malignant tumors stimulate blood- and lymphatic vessel growth by producing angiogenic factors (2–4). Although it is known that induction of tumor angiogenesis involves the interplay of a dozen tumor-derived growth factors, it is poorly understood how tumors induce lymphangiogenesis and what its role is in tumor spread.
Among the tumor-derived growth factors, vascular endothelial growth factor (VEGF)-A is a key angiogenic factor most frequently used by tumors and other tissues to switch on their angiogenic phenotypes. In fact, most, if not all, tumors express VEGF-A at high levels (5). VEGF-A is the prototype of a growth factor family that contains at least four additional structurally related members, including placenta growth factor, VEGF-B, VEGF-C, and VEGF-D (6). The angiogenic signals triggered by members of the VEGF family are mainly mediated by the activation of two homologous tyrosine kinase receptors, VEGF receptor (VEGFR)-1 and VEGFR-2, both of which are expressed on blood vessel endothelial cells (7). VEGF-A binds to VEGFR-1 and VEGFR-2, and induces vasculogenesis, angiogenesis, and vascular permeability. In contrast, placenta growth factor and VEGF-B only bind to VEGFR-1, which may participate in modulation of VEGF-A–induced angiogenesis (8, 9). In addition to VEGFR-1 and VEGFR-2, a lymphatic endothelial cell specific tyrosine kinase receptor, VEGFR-3, has been identified. VEGF-C and VEGF-D interact with both VEGFR-2 and VEGFR-3 and induce both blood angiogenesis and lymphangiogenesis (10, 11).
High expression levels of VEGF-C and VEGF-D in tumors have been correlated with enhanced lymphatic metastasis (11, 12). Interestingly, it has recently been found that VEGF-A also stimulates lymphangiogenesis (13–16). In this article, we show that in a murine xenograft tumor model, VEGF-A induces peritumoral lymphatic vessel growth and promotes lymphatic metastasis.
Materials and Methods
Reagents and animals. All animal studies done at the Karolinska Institutet were reviewed and approved by the animal care and use committee of the North Stockholm Animal Board. All handling of animals was done in accordance with University of California Los Angeles Animal Research Committee guidelines (see Supplementary data for details).
Vascular endothelial growth factor receptor-3/porcine aortic endothelial cell shape changes and motility assay. Analysis of VEGFR-3/porcine aortic endothelial (PAE) cell shape changes and motility was carried out according to our previously published methods, with the inclusion of the anti-mouse VEGFR-3 antibody (500 ng/mL; ref. 10; see Supplementary data for details).
Mouse corneal neovascularization assay. The mouse corneal angiogenesis assay was done according to procedures previously described (17).
Mouse corneal tumor model. Implantation of tumor tissues in the mouse cornea was carried out according to previously described procedures (ref. 18; see Supplementary data for details).
Retroviral vector design and tumor cell transduction. A murine fibrosarcoma cell line, T241, was transfected with complementary cDNAs coding for human VEGF (hVEGF)-A or hVEGF-C according to previously published procedures (19).
Tumor cell proliferation assay. Proliferation of nontransduced, empty vector–, hVEGF-A– and hVEGF-C–transduced T241 fibrosarcoma cells was done as previously published (19).
ELISA. Detection of endogenous mouse and human VEGF-A expression levels in conditioned media collected from wild-type (WT) and VEGF-A–transduced T241 cells were determined using the Quantikine ELISA system (R&D Systems, Inc., Minneapolis, MN) according to the instructions of the manufacturer and our previously published methods (19). A 72-hour conditioned media was prepared from cell cultures (15 mL/flask) originally seeded with 4 × 106 cells/flask.
Tumor growth assay. Approximately 1 × 106 tumor cells of WT, vector-, hVEGF-A– or hVEGF-C–transduced tumor cells were implanted s.c. on the back of 6- to 7-week-old female C57BL/6 mouse, and tumor volumes were measured as previously reported (20).
Lymphatic metastasis assay. The metastatic potential of s.c. implanted T241-VEGF-A primary tumors was studied according to previously described procedures (17). In another experiment, 0.5 × 106 Renilla luciferase–transduced VEGF-A T241 tumor cells (21) were implanted s.c. in the right upper back of C57BL/6 mice (n = 10). Anesthetized mice were imaged weekly using an optical imaging system immediately following i.v. injection of Renilla luciferase substrate coelenterazine. About 2 to 4 weeks after tumor implantation, the brachial lymph nodes from both sides were dissected and imaged ex vivo for the presence of Renilla luciferase as well as green fluorescent protein (GFP) signals according to previously described techniques (22). Optical signals were semiquantified using Living Image 2.5 (Xenogen) and IGOR (Wavemetrics, Lake Oswego, OR) image analysis softwares. Bioluminescence ex vivo signal was considered positive if the maximal signal in the region of interest exceeded 1 × 104 photons/s/cm2/steradian (see Supplementary data for details).
Histologic analysis of lymph nodes. Paraffin-embedded tissue sections (5 μm) of the brachial lymph nodes were stained with Surgipath eosin and Mayers' hematoxylin (HistoLab products AB, Gothenburg, Sweden). Further, GFP-positive tumor cells were detected in the lymph nodes using confocal microscopy (Zeiss Confocal LSM510 Microscope).
Whole-mount staining and confocal analysis. Growth factor– and tumor-implanted mouse eyes and primary tumors were double stained for CD31 and LYVE-1 using a whole-mount staining protocol recently published (ref. 17; see Supplementary data for details).
Detection of macrophages. Tumor sections of WT, VEGF-A, and VEGF-C were stained for macrophages according to standard immunohistochemical procedures (see Supplementary data for details).
In situ hybridization. Detection of mouse VEGF-C in tumor tissues by in situ hybridization was done according to our recent published procedures using an oligo probe complementary to VEGF-C (nucleotides 536-585 of murine VEGF-C mRNA; sequence reference no. NM-009506.1; ref. 17).
Statistical analyses. Statistical analyses of the in vitro and in vivo results were made by a standard two-tailed Student's t test using Microsoft EXCEL 5. P < 0.05 (*) and P < 0.001 (***) were deemed as significant and highly significant, respectively.
Vascular endothelial growth factor-A stimulated lymphatic vessel growth in vivo. We developed a mouse corneal lymphangiogenic model to evaluate the lymphangiogenic activity of VEGF-A. The newly formed lymphatic vessels were detected using a specific antibody against LYVE-1, a receptor almost exclusively expressed on lymphatic endothelial cells (23), with the exception of liver sinusoids (24). Implantation of a slow release polymer lacking growth factors did not result in the growth of new lymphatic vessels (Supplementary Fig. S1A, LYVE-1). In contrast, VEGF-A potently stimulated the growth of new lymphatic vessels sprouting from the existing limbal lymphatics (Fig. 1C). These newly formed lymphatic vessels were larger in diameter than newly formed blood vessels. The lymphatic vessels expressed only low, or undetectable, levels of platelet endothelial cell adhesion molecule 1 (PECAM-1; CD31), and their distribution did not generally overlap with blood vessels (Fig. 1E and G; CD31). Similar to VEGF-A, fibroblast growth factor (FGF)-2, a known lymphangiogenic factor, also induced a robust lymphangiogenic response in this model (Fig. 1D , F, and H). These findings show that VEGF-A is a potent lymphangiogenic factor.
Remodeling of newly formed lymphatic vessels. Similar to blood vessels, new lymphatic structures were detectable at day 5 after VEGF-A implantation (Supplementary Fig. S1B). At this time point, the early lymphatic sprouts lacked a vascular tree-like structure and the blood vessels appeared as disorganized vascular plexuses. At day 14, robust angiogenic and lymphangiogenic responses were detected, as the early primitive blood- and lymphatic vessels present at day 5 had undergone remarkable remodeling to form vascular tree-like structures mainly consisting of distinct vessels (Supplementary Fig. S1C). Unlike VEGF-A, FGF-2–induced lymphatic- and blood vessels appeared as well-organized vasculatures already by day 5, and they were further remodeled into mature vascular networks by day 14 after implantation (Supplementary Fig. S1D and E). Quantification analysis showed that VEGF-A and FGF-2 induced approximately equal amounts of blood- and lymphatic vessels at day 14 although FGF-2 was more potent at day 5 (Supplementary Fig. S1F and G).
Lymphangiogenesis in the absence of blood angiogenesis. In the corneal lymphangiogenesis model, a VEGF-A pellet was implanted into a micropocket created on one side of the circumferential corneal globe, as previously described (17). Immunohistologic analysis on this side of the cornea showed a delayed lymphangiogenic response as compared with blood angiogenesis, suggesting that blood vessels might be a prerequisite for lymphangiogenesis (Fig. 2A). Surprisingly, on the opposite side of the circumferential eye globe, VEGF-A stimulated lymphatic vessel growth with almost no blood neovascularization (Fig. 2B). This finding shows that VEGF-A–induced lymphangiogenesis can occur in the absence of blood vessels.
Anti–vascular endothelial growth factor receptor-3 neutralizing antibody did not block vascular endothelial growth factor A–induced lymphangiogenesis in the cornea. It is possible that VEGF-A induced lymphatic vessel growth via induction of VEGF-C/-D expression. To test this hypothesis, a neutralizing anti–VEGFR-3 antibody was examined for its potential effect on VEGF-A–induced lymphangiogenesis in the cornea. To first confirm the neutralizing activity, the anti–VEGFR-3 antibody was used to block VEGF-C–induced VEGFR-3/PAE cell activity in vitro. The anti–VEGFR-3 antibody significantly inhibited VEGF-C–induced morphologic changes (Fig. 3A-D) and cell migration (Fig. 3E), clearly demonstrating that the anti–VEGFR-3 antibody sufficiently blocked VEGF-C–induced endothelial cell activity in vitro.
To test the blocking capacities of the anti–VEGFR-3 antibody in vivo, VEGF-A or VEGF-C was coimplanted with the antibody in the mouse cornea. As expected, the anti–VEGFR-3 antibody almost completely blocked VEGF-C–induced lymphangiogenesis, but not blood angiogenesis (Fig. 3H-K). In contrast, neither VEGF-A–induced angiogenesis nor lymphangiogenesis was affected by this antibody treatment (Fig. 3F , G, J, and K), suggesting that the VEGFR-3 signaling pathway is not critical for the lymphangiogenic activity of VEGF-A.
Vascular endothelial growth factor-A stimulated peritumoral lymphangiogenesis in a xenograft tumor model. To test if VEGF-A was able to induce tumoral lymphangiogenesis, we transfected a murine T241 fibrosarcoma cell line with VEGF-A cDNA. As determined by a sensitive ELISA assay, the VEGF-A–transfected cell line showed an ∼285-fold increase in the expression of hVEGF-A (567 ng/mL) as compared with the endogenous levels of mVEGF-A (2.0 ng/mL) in these cells (19). The high expression levels of VEGF-A seemed to have no effect on tumor cell growth rate in vitro (Fig. 4A). However, implantation of VEGF-A T241 tumor cells in syngeneic mice resulted in accelerated tumor growth as measured by tumor volume and weight (Fig. 4B and C). Interestingly, VEGF-A T241 tumors contained a large number of peritumoral lymphatic vessels (Fig. 4E, LYVE-1). The peritumoral lymphatic vessels were of larger diameters than those of the blood vasculature. Similarly, overexpression of VEGF-C in the T241 fibrosarcoma cells also resulted in a remarkable growth of lymphatic vessels in the peritumoral area (Fig. 4F, LYVE-1). In addition to peritumoral lymphatic vessels, the VEGF-C T241 tumors also contained a high density of intratumoral lymphatic vessels (Fig. 4H, LYVE-1), whereas in the VEGF-A T241 tumors intratumoral lymphatic vessels were rarely detectable (Fig. 4G, LYVE-1). In the vector-transfected (Fig. 4D, LYVE-1) or nontransfected (data not shown) tumors, only a few peritumoral lymphatic vessels were detected around the border of the tumors. VEGF-C was more potent than VEGF-A in inducing lymphangiogenesis in the peritumoral as well as the intratumoral area of the tumors (Fig. 4I and J).
Vascular endothelial growth factor-A stimulated tumoral lymphangiogenesis in a corneal tumor model. To further study VEGF-A–induced tumoral lymphangiogenesis, we established a mouse corneal tumor model. Because of the corneal avascularity, tumor growth in the cornea excludes the involvement of any preexisting blood or lymphatic vessels. Implantation of WT or VEGF-A–transduced tumor tissues into the corneal micropockets resulted in the growth of tumors, expanding from the micropockets to the limbus. Corneal tumor neovascularization became visible by direct gross examination 2 weeks after implantation (Fig. 5A and F). The implanted VEGF-A–transfected tumor cells induced a robust angiogenic response, a pseudo-hemorrhagic phenotype, with a trimmed leading edge growing towards the implanted tumor (Fig. 5F). In contrast, WT tumors induced relatively well-defined tumor vasculatures in this model (Fig. 5A). Immunohistologic analysis revealed that VEGF-A tumors contained a honeycomb-like vascular network extending throughout the entire tumor tissue (Fig. 5G,, I-K, M, and N). LYVE-1 staining revealed the existence of gigantic lymphatic vessels growing throughout the entire VEGF-A tumor (Fig. 5H,-J and L-N). In the WT tumor, lymphatic vessels only grew in the surrounding area of the tumor, and they rarely penetrated into the tumor tissue (Fig. 5C-E). This corneal tumor model further shows that VEGF-A stimulates tumoral lymphatic vessel growth. As corneal tumors are relatively flat structures, as compared with those implanted s.c., the peritumoral and intratumoral lymphatic vessels might be difficult to distinguish completely in this model system.
Vascular endothelial growth factor-A promoted lymphatic metastasis. Stimulation of peritumoral lymphangiogenesis by VEGF-A raised the possibility that this factor might promote lymphatic metastasis. Indeed, resection of s.c. primary VEGF-A T241 tumors implanted in the middle dorsum of mice resulted in metastatic lesions of brachial lymph nodes in the majority of animals. Of 21 mice, 14 mice had metastases in both brachial lymph nodes, 6 mice had metastases in one brachial lymph node, and 1 mouse had no metastasis. These lymphatic tumor lesions were significantly greater in size than the lymph nodes of healthy animals or mice with WT or vector-transfected tumors as measured by volume and weight (Fig. 6D and E). Similarly, removal of VEGF-C T241 primary tumors also resulted in brachial lymphatic metastasis on both sides of all mice (n = 16). In contrast, resection of WT or vector-transfected tumors did not result in visible metastases in regional lymph nodes at the time point when these mice were sacrificed. Histologic examination of the lymph nodes confirmed the presence of GFP-positive VEGF-A or VEGF-C tumor cells in brachial lymph nodes (Fig. 6B and C, GFP), whereas lymph nodes of control mice lacked detectable GFP-positive tumor cells (Fig. 6A, GFP). H&E staining revealed the presence of invasive tumor cells in lymph nodes of VEGF-A and VEGF-C tumor–bearing mice but not in the healthy lymph nodes of control mice (Fig. 6A,-C, H, and E). To further validate these findings, VEGF-A T241 tumor cells were labeled in vitro by transducing with a lentivirus expressing Renilla luciferase reporter gene (21). Two weeks after implantation of the tumor cells in the right upper back of mice, bioluminescence signals were detectable in ex vivo dissected brachial lymph nodes and revealed the presence of lymphatic metastases (Fig. 6F and G). The optical signals were intensified more than 10-fold at 4 weeks after implantation (4 × 104 versus 5 × 105 photons/s/cm2/steradian), indicating the in situ growth of the metastatic lesions in the lymph nodes (Fig. 6H). The metastases were mainly confined to the ipsilateral side, although contralateral involvement was also observed after 4 weeks. The finding that VEGF-A promotes lymph node metastasis is further confirmed by our unpublished data from a prostate cancer xenograft model where ∼40% of the mice developed brachial lymph node lesions.
Vascular endothelial growth factor-A tumors contain high numbers of inflammatory macrophages. It has recently been reported that VEGF-A activates inflammatory cells that contribute to VEGF-A–induced lymphangiogenesis (13). To study if VEGF-A could also induce inflammatory cells in tumors, tissue sections of various tumors were stained with an antimacrophage specific antibody. Interestingly, VEGF-A induced higher numbers of intratumoral macrophages as compared with WT or VEGF-C tumors (Supplementary Fig. S2A, C, E, and G). However, VEGF-C was also able to significantly induce macrophage infiltration (Supplementary Fig. S2C and G). In situ hybridization analysis showed that VEGF-A did not up-regulate VEGF-C mRNA expression in VEGF-A–overexpressing tumors as compared with control tumors (Supplementary Fig. S2B, F, and H). As a positive control, we used a VEGF-C T241 cell line expressing high levels of VEGF-C mRNA (Supplementary Fig. S2D and H). This finding indicates that VEGF-A does not up-regulate VEGF-C expression and thus supports our conclusion that VEGF-A induces lymphangiogenesis via a VEGF-C–independent pathway.
Most solid malignant tumors metastasize to regional lymph nodes. However, at present, it remains controversial on the issue whether tumor lymphatics are functional or not (11, 12, 25). Tumor cells may either access into the lymphatic system by inducing intratumoral lymphangiogenesis or by co-opting preexisting lymphatics in the surrounding tissue (26). As the key function of the lymphatic system is immune surveillance, intratumoral lymphangiogenesis could be a part of our body's defense system against cancer. However, cancer cells have attained the ability to escape our immune surveillance system and have adopted growth advantages in different environments. As a consequence, the lymph node can become a site for tumor growth instead of a site for tumor elimination.
VEGF-A was previously thought to act as a specific blood angiogenic factor through activation of VEGFR-1 and VEGFR-2. However, a couple of recent studies have shown that VEGF-A is able to induce lymphangiogenesis in animal models (13, 15, 16). In a recent article, Cursiefen et al. (13) report that VEGF-A–activated inflammatory macrophages release VEGF-C/-D that contributes to lymphangiogenesis. In support of this finding, we observed significantly increased number of inflammatory macrophages in the VEGF-A tumors. However, we were not able to block VEGF-A–induced lymphangiogenesis with a neutralizing anti–VEGFR-3 antibody in the cornea assay, suggesting that VEGF-A stimulates lymphatic vessel growth via a VEGF-C/-D/VEGFR-3–independent pathway. In support of our finding, it has been reported that the VEGFR-2 receptor is expressed occasionally on lymphatic endothelial cells, suggesting a direct stimulatory effect of VEGF-A on lymphatic vessels (27). We found that on the opposite side of the implanted corneal globe, lymphatic vessels can grow without blood vessels, suggesting that lymphangiogenesis is not entirely dependent on the blood vessel system.
We observed that tumors overexpressing VEGF-A were highly infiltrated by macrophages. This finding is consistent with the fact that macrophages are responsive to VEGF-A–induced chemotaxis via the VEGFR-1 signaling pathway (28). Inflammatory cells, such as macrophages, are recruited to tumors by a wide range of tumor cell–derived cytokines and growth factors, including VEGF-A (29). Recent publications have provided experimental evidence that macrophages promote progression of solid tumors and metastasis (30, 31). If monocytes/macrophages play a critical role in mediating VEGF-A–induced lymphangiogenesis, what are the active lymphatic cytokines produced by these cells? Among known lymphangiogenic factors, VEGF-C and VEGF-D are the best-characterized growth factors that act directly on lymphatic endothelial cells via activation of VEGFR-3. Our data show that an anti–VEGFR-3 neutralizing antibody was unable to block VEGF-A–induced lymphangiogenesis in the cornea tissue, suggesting that VEGF-C/-D are not major players in mediating VEGF-A–stimulated lymphangiogenesis. The fact that tumors overexpressing VEGF-A do not display increased levels of VEGF-C expression further supports this finding.
Our results provide convincing evidence that VEGF-A induces a lymphangiogenic response in tumors expressing this factor at a high level. As a result of the lymphangiogenic effect, VEGF-A–expressing tumors metastasize to the local lymph nodes. Consistent with our findings, a recent report using a genetic skin tumor model showed that VEGF-A promotes metastasis to regional lymph nodes (29). Thus, our work, together with this report, provides novel insights for the understanding of the role of VEGF-A in promoting primary tumor growth and cancer metastasis. VEGF-A is probably the most frequently up-regulated angiogenic factor during tumor growth. Thus, it might also be the most frequently occurring lymphangiogenic factor contributing to lymphatic metastasis as nearly all solid tumors express VEGF-A at high levels. Recently, clinical evaluation of a VEGF-A antagonist, avastin, used in the treatment of human cancers, has provided encouraging results in the field of antiangiogenic therapy. Taken together with our work, further development of VEGF antagonists in the treatment of human cancers may improve the therapeutic outcome for cancer patients.
Note: M. Björndahl and R. Cao contributed equally to this work.
Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
Grant support: The Swedish Research Council, the Swedish Heart and Lung Foundation, the Swedish Cancer Foundation, the Söderberg Foundation, and the Karolinska Institute Foundation; National Cancer Institute Specialized Program of Research Excellence program P50 CA092131 (L. Wu); University of California at Los Angeles Research Training in Pharmacological Sciences Training program #T32-GM008652 (J. Burton); and the Swedish Wenner-Gren Foundations (E. Brakenhielm).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We thank Dr. David Jackson (John Radcliffe Hospital, Oxford, United Kingdom) for providing us valuable anti–LYVE-1 antibodies for our study. We thank Irvin Chen and John Colicelli (Department of Biological Chemistry, University of California, Los Angeles, CA) for providing materials for the generation of lentiviral constructs.