Differential in Vivo and in Vitro Expression of Vascular Endothelial Growth Factor (VEGF)-C and VEGF-D in Tumors and Its Relationship to Lymphatic Metastasis in Immunocompetent Rats¹

Clinically, the most critical point in tumor progression is the acquisition of metastatic potential. The majority of human carcinomas show a predilection for metastasizing at least initially via the lymphatic system (reviewed in Ref. 1). Whether or not a tumor has disseminated to regional lymph nodes has a strong influence on the prognosis of cancer patients, and assessment of regional lymph node status is very important in determining the course of cancer therapy. Despite this, most experimental work addressing systemic tumor dissemination has focused on hematogenic spread. Little is known about the mechanism by which tumor cells enter, interact with, and are transported within lymphatic vessels.

Study of the lymphatic system has recently been promoted by the identification of molecules that act as markers of the lymphatic endothelium (reviewed in Ref. 2). One of these molecules is a fms-like tyrosine kinase receptor called VEGFR-3.³ VEGFR-3 is activated in response to its ligands, VEGF-C and VEGF-D, causing stimulation of lymphangiogenesis, the new growth of lymphatic vessels.

VEGF-C and VEGF-D are progressively processed during their biosynthesis to remove the NH₂- and COOH-terminal ends of the protein, leaving the central VEGF homology domain (3, 4). This processing progressively increases the affinity of VEGF-C for VEGFR-3. The fully processed forms are also able to activate VEGFR-2 and therefore have the potential to induce both angiogenesis and lymphangiogenesis. However, mutation of a cysteine residue in the central VEGF homology domain ablates the ability of fully processed VEGF-C to activate VEGFR-2, making this mutated protein a specific activator of VEGFR-3 (5, 6).

Whereas the presence or absence of lymphatic vessels inside tumors remains controversial (7), many studies have demonstrated that an increased density of lymphatic vessels is found in the stromal tissue at the periphery of tumors (8, 9, 10, 11, 12). Recent studies stress the importance of functional lymphatics at the periphery of tumors (13). The invasive external margin of tumors is the site where invasion of lymphatic vessels is likely to occur. Thus, the entry of tumor cells into the lymphatic system may be potentiated by the growth of lymphatic vessels in the vicinity of the tumor, which would serve to increase the number of potential entry sites for disseminating tumor cells. If tumors are able to produce VEGF-C and/or VEGF-D, they are therefore likely to be able to promote their entry into the lymphatic system by stimulating lymphangiogenesis in their vicinity. In support of this notion, some studies of clinical material have shown that expression of VEGF-C and/or VEGF-D in primary tumors correlates with their ability to form lymph node metastases [see, for example, the review by Pepper (7)]. However, some of these studies based their conclusions on cultured tumor cells. It remains unclear whether expression of VEGF-C and VEGF-D in culture accurately represents the in vivo situation.

A number of animal studies using cell lines (14, 15, 16, 17, 18, 19, 20) and transgenic mice (21) have been conducted in an attempt to demonstrate functionally that VEGF-C or VEGF-D overexpression is able to promote tumor metastasis. Whereas these studies indeed show that in some cases expression and activity of these growth factors can be required for metastasis to lymph nodes, a number of open questions remain to be addressed. The forms of VEGF-C and VEGF-D used in the published animal studies had the potential to activate VEGFR-2 in addition to VEGFR-3, and thus it could not be determined whether or not VEGFR-3 activation was specifically responsible for the observed effects on metastasis. Furthermore, virtually all of the animal studies to date used immunocompromised animals, and thus these experiments lacked a major modifier of tumor growth and metastasis that would be found in human patients. From the point of view of possible applications in cancer therapy, it is vital to determine whether blocking the activity of endogenous VEGF-C and VEGF-D in naturally metastatic tumors can inhibit metastasis formation in immunocompetent animals. The aim of the study presented in this paper was to address these issues.

The culture conditions and properties of the following cell lines are described in Refs. 22 and 23: MTLN2; MTLN3; MTC; MTLy; MTPA; AT1; AT2.1; AT3.1; AT6.1; MatLu; MatLyLu; G; ASML; 10AS; 1AS; CREF and its derivatives; NM-081; MT-450; RFS; and BDX2. ARIP, 13762 MAT B III, AR42J, RBA, BRL3A, and L6 were obtained from American Type Culture Collection. RG2 is a rat glioma cell line (24), RTE2 is a rat tongue epithelial cell line (25), and NID2 is a neu-transformed rat Schwannoma cell line (26). Certain passages of tumor cells used in this paper showed altered metastatic potential compared with previously published data (22, 23), and this metastatic potential is indicated in Fig. 1. PAE cells expressing VEGFR-3 have been described previously (6).

Syngeneic rats received s.c. injection with 5 × 10⁵ cells in PBS. Animals were monitored regularly until their tumors grew to the German legal limit (50 mm in one dimension), or until they became moribund, at which time they were killed, and an autopsy was performed. For the MT-450 VEGFR-3-Rg experiment in Fig. 5, all of the animals in the different experimental groups were sacrificed and analyzed on the same day. At autopsy, the size of the primary tumor was measured, and the number, size, and location of metastases in the draining lymph nodes and other organs were assessed. Where required, primary tumors were excised carefully and trimmed to remove all of the surrounding tissues, and then chunks of tumor were snap frozen for RNA preparation or cryosectioning. Alternatively, primary tumor material, together with the surrounding tissue, was fixed in 4% paraformaldehyde for paraffin embedding.

Metastasis was defined as lymph nodes of ≥10 mm in diameter or lung nodules of ≥1 mm in diameter. When these conditions were not met, the lymph nodes and/or lungs were disaggregated and placed into tissues culture in selection medium. If tumor cells grew out of these cultures, the corresponding tissue was classified as having metastases. The lymph nodes and lungs were also analyzed histologically in parallel to determine the presence of metastases. Statistical significance of the results was determined using the Mann-Whitney rank test.

Tumor-bearing animals were killed by cervical dislocation. All subsequent processing of the tumors was performed at 4°C using ice-cold solutions. Tumors were excised and trimmed, cut into small pieces, and then washed thoroughly in PBS. The tumor pieces were then passed through a stainless steel tissue sieve (Sigma) into PBS. The resulting brei was then passed through a nylon cell strainer with a 40-μm pore size (Falcon). The cell suspension was then washed once in PBS and twice in PEB. The PEB was degassed extensively before use. The cells were then examined and counted under a light microscope. More than 97% of the cells were single cells. The cells were >90% viable, as determined by trypan blue exclusion. The single cell suspension was used immediately for sorting.

Single cell suspensions of tumor cells from a single tumor were resuspended in 10 ml of PEB (1 × 10⁷ cells/ml) and labeled on ice with primary antibody (10 μg/ml). After washing in ice-cold PEB, the cells were incubated at 4°C with antimouse IgG MACS beads (Miltenyi) and phycoerythrin-labeled antimouse IgG antibodies (Dako) and then separated on a VS+ MACS column (Miltenyi) as described in the manufacturer’s instructions. The bulk of sorted and nonsorted cells were immediately snap frozen in liquid nitrogen and stored until nonfrozen aliquots were subjected to FACS analysis using a Becton Dickinson FACStarPlus machine to check the sorting efficiency. The frozen cells were then used for RNA preparation and RT-PCR analysis.

Sorted and nonsorted tumor cells and corresponding cultured tumor cells were used for total RNA preparation and first-strand cDNA synthesis using Superscript II reverse transcriptase (Life Technologies, Inc.) according to the manufacturer’s instructions. Preliminary RT-PCR experiments were performed to determine linear amplification conditions for the different targets. For VEGF-C, the final cycle parameters used were 94°C, 15 s and 68°C, 4 min (35 cycles). For VEGF-D, the cycle parameters were 95°C, 5 s; 68°C, 15 s; and 72°C, 1 min (32 cycles). For GAPDH, the cycle parameters were 94°C, 1 min; 62°C, 1.5 min; and 72°C, 3 min (30 cycles). The following primer pairs were used: (a) RVEGFCF (GCGAATTCGGACCGGCCTCCTCGCTCCC) and RVEGFCR (GCACCGGTGTTCAGATGTGGTCTTTTCCAATATG) [amplifies nucleotides 54–1332 of rat VEGF-C, GenBank accession number AY032729]; (b) RVEGFDSF (CATCCCTACTCAATTATCAGAAGATCC) and RVEGFDSR (GGCAACAGCTTTCCAGACTTTCTTTGC) [amplifies nucleotides 651–963 of rat VEGF-D, GenBank accession number AY032728]; and (c) GAPDHF (AGACAGCCGCATCTTCTTGTGC) and GAPDHR (CTCCTGGAAGATGGTGATGG) [amplifies nucleotides 23–302 of rat GAPDH, GenBank accession number X02231].

Tissue culture cells were harvested, washed in PBS, pelleted, and then snap frozen. Polyadenylated RNA was prepared from snap-frozen cells and tumors (27). Northern blots using 1% agarose-formaldehyde gels and 5 μg of polyadenylated RNA were performed as described previously (27). Blots were probed sequentially at high stringency using VEGF-C (partial rat VEGF-C cDNA, GenBank accession number AF010302), VEGF-D (complete coding sequence, GenBank accession number AY032728), and GAPDH [PstI fragment of plasmid pGAPDH (28)] probes. In between each probing, the blots were stripped in 0.1% SDS at 100°C and then exposed to film to ensure complete removal of the probe. To permit the equilibration of signal intensity from one blot to another, identical positive control samples (RNA from BRL3A cells and mammary tissue) were included on each blot.

The extracellular portion of rat VEGFR-3 was amplified by PCR from rat spleen cDNA using primers encompassing bases 1–2310 of full-length rat VEGFR-3 (GenBank accession number AF402785).⁴ The forward and reverse primers used were 5′-atgcagccgggcgctgcgctgaaccggc-3′ and 5′-tcttccgagccttctacggccacgctggcggagg-3′. The resulting PCR product was flanked by BamHI and NheI sites for subsequent cloning. The IgG2a heavy chain fragment was amplified by PCR from rat spleen cDNA with the following primers: 5′-ggactgtgaagagttcagagaacc-3′ and 5′-ggatccttggccaggaagaggggtaaggg-3′. The resulting PCR product was flanked by NheI and EcoRI sites and encompassed the hinge and the CH2 and CH3 constant heavy chain domains of rat IgG2a. The two fragments were then ligated to the BamHI and EcoRI sites of the pcDNA 3.1+ (Invitrogen) vector in a three-fragment ligation. The resulting rat VEGFR-3/IgG2a heavy chain fusion construct (VEGFR-3-Rg) was sequenced to ensure no PCR artifacts had been introduced.

The ΔNΔC/VEGF-C/Cys152Ser insert was cut out of plasmid pAc5.1/His/BiP/ΔNΔC/VEGF-C/Cys152Ser (6) and cloned into pSecTag (Invitrogen).

MT-450 and NM-081 cells were stably transfected with the appropriate expression constructs using Tfx50 (Promega) according to the manufacturer’s instructions. For controls, cells were stably transfected with the pcDNA3.1 or pSecTag vectors as appropriate. Stably transfected clones were checked for expression of the transfected expression construct by Western blotting. Clones showing the highest levels of expression were selected for additional experiments.

The third and fourth Ig domains of mouse VEGFR-3 were cloned by PCR using mouse spleen cDNA as a template with the following primers: 5′-gcgggatcccagaatgacctgggcccc-3′ and 5′-gcgaattcgacgaccacacacttgtac-3′. The resulting PCR fragment corresponding to amino acids 301–530 of SwissProt sequence P35917 was cloned into the pGEX-2T vector, and the construct was used to make VEGFR-3-glutathione S-transferase fusion protein (29). This fusion protein was used to immunize rabbits according to standard protocols. The resulting bleeds (2622 and 2631) were shown to cross-react with rat and mouse VEGFR-3 in a number of immunoassays.

PAE cells expressing VEGFR-3 were stimulated either in the presence of 200 ng/ml VEGF-C in conditioned medium from mock-transfected cells or in the presence of 200 ng/ml VEGF-C in conditioned medium from cells expressing VEGFR-3-Rg protein. Ligand-induced phosphorylation of VEGFR-3 was then assessed as described previously (6).

Tumor cells (1 × 10⁶) were injected into the skin of the flank between the inguinal and axillary lymph nodes. After the tumors had grown to between 0.5 and 2 cm in diameter, the rats were sacrificed, and the tumors in the skin of the flank were exposed. Evan’s Blue dye (2.5 mg/ml in PBS) was injected at multiple sites into the skin on the inguinal side of the tumor. Dye injection sites were approximately 1.5 cm removed from the tumor, and 20–50 μl of dye were injected per site. Dye injection was continued along the span of the afferent lymphatics entering the tumor until lymphatic channels draining from the inguinal to the axillary region were marked that did not impinge on the tumor. Lymphatic vessels impinging on the tumor were counted, and the tumor volume was measured. Statistical significance of the results was determined using the Mann-Whitney rank test.

Polyclonal anti-Prox1 antibodies were used to stain frozen tumor sections as described previously (30). Briefly, 20-μm-thick cryosections fixed for 10 min in 100% methanol were incubated for 1 h with Prox1 antiserum diluted 1:500. The secondary Cy3-conjugated goat antirabbit antibody (Dianova, Hamburg, Germany) was applied at a dilution of 1:200 for 1 h. Sections were viewed under a fluorescence microscope.

To determine whether constitutive VEGF-C or VEGF-D expression in tumor cells in culture reflects the ability of tumor cells to metastasize via the lymphatics in vivo, the expression of VEGF-C and VEGF-D mRNA was examined in a panel of 34 rat tumor cell lines grown in tissue culture (Fig. 1). VEGF-C is constitutively expressed at relatively high levels in more than 50% of the cell lines. The expression of VEGF-C in the cell lines did not correlate with their metastatic proclivity as assessed by in vivo passaging. VEGF-D was expressed strongly in only one cell line, the pancreatic carcinoma ARIP. Weak VEGF-D expression was observed in a few other cell lines such as BRL3A and CREF. For both VEGF-C and VEGF-D, the size of the transcript expressed by the tumor cells corresponds to the size of the major transcript expressed in nonneoplastic tissues (data not shown). These data demonstrate that lymphangiogenic factors are constitutively expressed in tissue culture by numerous tumor cells of diverse origins, although this expression does not always correlate with the metastatic behavior of the cells. Furthermore, it should be borne in mind that the processing status and thus the activity of the VEGF-C/D protein produced by cells expressing the corresponding mRNA are not addressed in these studies.

An unexpected finding was that certain tumor cell lines such as MT-450 and MTLN3 that are highly metastatic via the lymphatics in vivo did not express significant amounts of VEGF-C or VEGF-D in culture. However, in situ hybridization analysis demonstrated that tumors derived from these tumor cells exhibit enhanced numbers of VEGFR-3-positive lymphatic vessels at their periphery (data not shown). These observations prompted us to examine whether the expression profile of VEGF-C and VEGF-D in tumor cells grown in tissue culture is maintained in tumors derived from these cells. To this end, a number of the tumor cell lines analyzed in Fig. 1 were s.c. injected into syngeneic animals. Primary tumors from these animals were then used for RNA preparation and Northern blot analysis. Surprisingly, we observed up-regulated VEGF-C and VEGF-D expression in some of these tumors. For example, Fig. 2 shows VEGF-C and VEGF-D mRNA expression in a selection of mammary carcinoma cell lines passaged in vivo. MTPa and MTC express VEGF-C in tissue culture, but VEGF-D is not significantly expressed (see Fig. 1), and this pattern is maintained in vivo (Fig. 2). Similarly, MTLy expresses barely detectable levels of VEGF-C and VEGF-D in tissue culture, and these factors are only weakly detectable in tumors derived from these cells (Figs. 1and 2). However, MT-450 cells express almost no VEGF-C or VEGF-D in tissue culture, but in MT-450 tumors, VEGF-C expression is moderately up-regulated, and VEGF-D expression is strongly up-regulated (Fig. 2). Moreover, MTLN3 tumors express VEGF-C and VEGF-D, despite the fact that MTLN3 cells express virtually none of these factors in tissue culture (Figs. 1and 2).

The VEGF-C and VEGF-D we observed to be expressed in tumors derived from cell lines that do not express these factors at significant levels in culture could be derived from up-regulation in the tumor cells themselves or from host-derived stromal cells that constitute part of the tumor. To differentiate between these possibilities, we performed in situ hybridization analysis to look at the cells expressing VEGF-C/D in tumors. Variable levels of hybridization to both tumor cells and stromal components could be observed, with focal areas of the tumors showing stronger expression than others (data not shown). To substantiate these results, we disaggregated MT-450 and MTLN3 tumors into single cell suspensions and used antibodies specific for the tumor cells to separate the tumor cells from the host-derived stromal cells using MACS. The anti-CD44v6 antibody 1.1ASML binds specifically to MTLN3 cells and not to stromal components (Ref. 31; data not shown), whereas the M-N#1 antibody binds specifically to MT-450 tumor cells and not to stromal cells (23). Aliquots of sorted cells were checked by FACS analysis to determine the sorting efficiency (Fig. 3,A). RNA was extracted from the remaining cells and subjected to semiquantitative RT-PCR analysis to determine in which compartment VEGF-C and VEGF-D are expressed (Fig. 3 B). These data demonstrate that both tumor and stromal components contribute to the VEGF-C/D expression seen in the tumors, in stark contrast to the lack of expression observed in the corresponding cultured tumor cells. Similar results were obtained in three independent experiments.

Several studies show that fusion of the extracellular portion of growth factor receptors with IgG heavy chains enforces dimerization of the receptor, resulting in a soluble protein that binds with high affinity to its cognate ligand. These Rg molecules can therefore be used to block activation of the corresponding cellular receptor by sequestering the ligand (e.g., Ref. 32). Such an approach with VEGFR-3 has recently been published, and it shows that blockade of VEGFR-3 activation inhibits lymphangiogenesis during embryogenesis (33) and in tumors grown on the avian chorioallantois (34).

We took a similar approach to block VEGF-C/D-induced activation of VEGFR-3 in the context of tumors to determine what effect this has on tumor-induced lymphangiogenesis and tumor metastasis in immunocompetent rats. We created a construct in which the extracellular portion of rat VEGFR-3 is fused with the rat IgG2a heavy chain. Expression of this construct in cells resulted in the secretion of a dimeric molecule that was very effectively in blocking VEGF-C/D-induced activation of VEGFR-3 in cellular assays (Fig. 4).

Stably transfected clones of MT-450 cells expressing the VEGFR-3-Rg protein were s.c. injected into the flank of syngeneic animals, and tumors were allowed to develop. Animals were sacrificed after 37 days, and the tumor volume and number of metastases were analyzed in comparison with those of animals receiving vector control-transfected MT-450 cells (Fig. 5). The growth of the primary tumor was not affected by expression of VEGFR-3-Rg. However, the size of the tumors in the draining axillary lymph nodes (P < 0.001) and the number of metastatic nodules in the lung (P < 0.001) were significantly reduced.

One explanation for the data in Fig. 5 would be that expression of VEGFR-3-Rg neutralized the activity of tumor-produced VEGF-C/D and that tumor-induced lymphangiogenesis was thereby suppressed, decelerating the rate of onset of lymphatic metastasis. To determine whether this is indeed the case, we looked directly at the number of lymphatic vessels in the periphery of MT-450 tumors expressing or not expressing VEGFR-3-Rg.

Due to the possible unreliability of certain markers of the lymphatic endothelium in the context of tumors (2), we chose first to directly visualize lymphatic vessels afferent to the tumors. The rationale of this approach is that the number of these vessels will increase in proportion to the activity of lymphangiogenic factors in and around the tumor. To this end, we s.c. injected MT-450 cells expressing or not expressing VEGFR-3-Rg into the flank between the inguinal and axillary lymph nodes. After they had grown, the tumors in the skin of the flank were exposed. Evan’s Blue dye was injected at multiple sites into the skin on the inguinal side of the tumor to delineate the lymphatic vessels impinging upon the tumor and enable them to be counted. Dye injection sites were approximately 1.5 cm removed from the tumor, forming a line spanning the complete afferent lymphatics able to impinge upon the tumor. This was ensured by continuing injecting above and below the tumor until lymphatic channels draining from the inguinal to the axillary region that did not impinge on the tumor were marked (see Figs. 6and 8). Several observations demonstrate that lymphatic vessels and not blood vessels were delineated by this method: (a) dye always entered vessels draining in the inguinal-axillary direction and not in the reverse direction, as would be expected for blood vessels; (b) blood vessels around the tumor could be clearly seen and were never filled with dye, and dye-filled vessels often ran along side and parallel to blood vessels (see Fig. 8); and (c) dye-filled vessels drained directly to the axillary lymph node and stained it blue (see Fig. 8).

Preliminary experiments showed that the number of lymphatic vessels draining tumors is proportional to the size of the tumor. We therefore carefully matched the sizes of the tumors measured in the different groups to ensure that no bias was introduced into the measurements. In experiments with larger numbers of tumors, expression of VEGFR-3-Rg in MT-450 tumors clearly reduced the number of afferent lymphatic vessels (Fig. 6, A, C, and E), suggesting that the VEGFR-3-Rg had indeed neutralized lymphangiogenic factors in the peritumoral region (P < 0.001).

To verify the data obtained in the experiments with Evan’s Blue dye, we took sections of MT-450 tumors expressing or not expressing VEGFR-3-Rg and immunohistochemically stained them with antibodies against Prox1, a marker of the lymphatic endothelium (30, 35). In tumors derived from empty vector-transfected MT-450 cells, a high density of Prox1-positive lymphatic vessels was observed at the periphery of the tumors (Fig. 6,B). It was often possible to see Prox1-positive lymphatic vessels filled with tumor cells (data not shown). Prox1-positive lymphatic vessels were only observed at the periphery of the tumors and were not observed intratumorally. Expression of VEGFR-3-Rg depressed the number of lymphatic vessels that could be detected in the vicinity of MT-450 tumors (Fig. 6 D), consistent with findings from the Evan’s Blue dye injections.

We have previously reported a fully processed form of rat VEGF-C (ΔNΔC/VEGF-C/Cys152Ser) that has a cysteine to serine mutation and is able to activate VEGFR-3 but not VEGFR-2 (6). To address whether specific activation of VEGFR-3 is sufficient in the context of a tumor to promote tumor metastasis, we set out to overexpress rat ΔNΔC/VEGF-C/Cys152Ser in weakly metastatic NM-081 mammary carcinoma cells and determine what effect this had on metastasis formation. These cells do not endogenously express VEGF-C or VEGF-D (Fig. 1).

We stably expressed rat ΔNΔC/VEGF-C/Cys152Ser in NM-081 cells. Clones expressing ΔNΔC/VEGF-C/Cys152Ser were injected s.c. into syngeneic animals, and tumors were allowed to develop. Animals were sacrificed when they became moribund, or when the size of their primary tumor reached the legal limit. Autopsy was then performed, and the number of metastases was analyzed in comparison with animals receiving vector control-transfected NM-081 cells. All animals bearing tumors expressing ΔNΔC/VEGF-C/Cys152Ser had lymph node and lung metastases. Only one animal from the control groups had a lymph node metastasis, and no animal had lung metastases. Correspondingly, expression of ΔNΔC/VEGF-C/Cys152Ser dramatically increased the size of lymph node metastases (P < 0.001) and the number of lung nodules (P < 0.001; Fig. 7).

To determine whether ΔNΔC/VEGF-C/Cys152Ser expression also increases the number of peritumoral lymphatic vessels, we performed Evan’s Blue dye injection experiments and immunohistochemistry for Prox1 as described above. Indeed, the number of afferent lymphatic vessels is strongly promoted by the expression of ΔNΔC/VEGFC/Cys152Ser in NM-081 tumors (Fig. 8, A, B, and E; P < 0.001). Whereas it was rare to see Prox1-positive vessels at the periphery of control NM-081 tumors, significant numbers of Prox1-positive lymphatic vessels were observed at the periphery of ΔNΔC/VEGF-C/Cys152Ser-expressing tumors (Fig. 8, C and D). Again, Prox1-positive lymphatic vessels were only observed at the periphery of the tumors and were not observed intratumorally. VEGF-C overexpression has been reported in some instances to induce lymphatic hyperplasia, resulting in the formation of lymphatic lacunae (21, 36). However, we saw little evidence of this in the Prox1-stained histological sections; rather, we saw increased numbers of capillaries. Thus, specific activation of VEGFR-3 alone by tumor-derived ΔNΔC/VEGF-C/Cys152Ser suffices to induce lymphangiogenesis and lymphatic metastasis.

The lymphatic endothelium that forms or lines lymphatic vessels has loose intercellular junctions that readily permit the passage of large biological macromolecules, pathogens, and migrating cells, and lymphatic capillaries have no or at best only an incomplete basement membrane (reviewed in Ref. 2). These features make the lymphatic system an easily available target for tumor cell entry into the circulatory system compared with blood capillaries. Tumor-induced lymphangiogenesis would increase the density of potential entry sites for tumor cells to access the lymphatic system and is thus one mechanism by which the metastasis of tumor cells could be potentiated. In support of this notion, we show here using immunocompetent rat tumor models that the activity of the lymphangiogenic factors VEGF-C and VEGF-D in tumors promotes tumor metastasis to regional lymph nodes and at the same time stimulates an increase in peritumoral lymphatic vessels. Specifically, expression of a VEGF-C mutant that is only capable of activating VEGFR-3 is sufficient to promote both peritumoral lymphangiogenesis and lymphatic metastasis, whereas neutralization of endogenous VEGF-C and VEGF-D activity in tumors inhibits both tumor metastasis and the formation of peritumoral lymphatic vessels. We provide evidence that tumor cell-host interactions can determine the expression levels of VEGF-C and VEGF-D in tumors.

In the animal experiments presented in this paper, blockade of VEGF-C/D activity reduced the size of lymph nodes metastases, whereas ectopic expression of ΔNΔC/VEGF-C/Cys152Ser had the reverse effect. It is unlikely that the modulation of metastatic proclivity caused by changes in VEGF-C/D activity is caused by differences in the growth rates of the tumor cells. The matched cell clones used in these experiments, which either expressed recombinant protein or were transfected with empty vector, proliferated equivalently in culture, and we observed no consistent changes in onset of tumor formation or size of tumors developing from these matched cell clones. For example, in Fig. 5, expression or nonexpression of VEGFR-3-Rg in MT-450 cells had no significant effect on the growth of primary tumors derived from these cells. Thus, the observation that VEGFR-3-Rg reduced the size of metastases in the draining lymph node suggests that as reduction in VEGF-C/D activity by VEGFR-3-Rg decreased peritumoral lymphatic vessel density, disseminating MT-450 tumor cells were at least temporally retarded from entering the lymphatics, and thus the onset of metastasis was slowed, resulting in smaller metastases at autopsy. Conversely, the enhanced peritumoral lymphatic vessel density caused by expression of ΔNΔC/VEGF-C/Cys152Ser in NM-081 tumors suggests that the greatly increased numbers and size of regional lymph node metastases were facilitated by enhanced entry of tumor cells into the lymphatics.

Modulation of VEGF-C/D activity also influenced the number of metastatic lung colonies that developed. Lymph node metastases grow to a significant size before lung nodules can be detected macroscopically or histologically in the MT-450 and NM-081 models (data not shown), suggesting that lung metastases are seeded from lymph node metastases that shed tumor cells into the blood. This notion is supported by two other observations: (a) NM-081 tumors did not give rise to lung metastases in the absence of lymph node metastases; and (b) MT-450 lung metastasis formation was blocked by VEGFR-3-Rg expression. We therefore suggest that the number of lung metastases that develop in these animal models is directly related to the ability of the primary tumor to give rise to lymph node metastases, which in turn is dependent on the activity of VEGF-C/D in the tumors.

While this paper was under review, He et al. (20) published results similar to ours in which they showed that VEGFR-3-Rg was able to reduce the metastasis of a human lung carcinoma cell line from a s.c. site to regional lymph nodes in immunocompromised mice. However, in contrast to our results, lung metastasis formation was not affected by VEGFR-3-Rg. A possible explanation for this difference is as follows. In the case of our experiments, we propose that lymph node metastases act as bridgeheads for the formation of lung metastases (see Ref. 1). For a variety of reasons, it is easier for tumor cells to form metastases in lymph nodes than in organs such as the lung. Primary tumors are comprised of a heterogenous population of tumor cells, few of which have the ability to successfully form metastases. Lymph nodes metastases represent expanded populations of tumor cells selected on the basis of their ability to metastasize. These cells therefore have at least some of the properties required for further successful metastasis to organs such as the lung. Most of the cells directly entering the bloodstream from the primary tumor do not have these properties, and therefore lung metastasis is much more likely to occur when lymph node metastases have already formed and are seeding cells into the bloodstream. Thus, in our experiments, inhibition of lymph node metastasis formation also inhibited lung metastasis formation. On the other hand, He et al. (20) used lung carcinoma cells in their experiments. The cellular origin of the tumor cells may have meant that circulating tumor cells were able to grow selectively in their native environment of the lung, and thus lung metastases formed without the need for preceding lymph node metastases. Thus, in this case, VEGFR-3-Rg-mediated inhibition of lymph node metastasis did not inhibit lung metastasis formation.

He et al. (20) also ectopically expressed VEGF-C in weakly metastatic lung carcinoma cells. Whereas this resulted in increased peritumor lymphangiogenesis, metastasis to lymph nodes was not enhanced, as was also reported by Karparnen et al. (17) using mammary carcinoma cells. However, several other publications agree with our finding that tumor-induced lymphangiogenesis is able to promote metastasis to lymph nodes (14, 15, 16, 17, 18, 19). One explanation for these differences is that some tumor lines need to acquire additional cellular properties before being able to metastasize via an expanded lymphatic capillary bed.

Several studies have examined VEGF-C or VEGF-D expression in cultured human cell lines and made conclusions on this basis about the relevance of these factors to metastasis formation and patient prognosis. Our data clearly show, however, that expression of these factors in tissue culture often has no bearing on their in vivo expression or on the metastatic competence of the tumor cells. Furthermore, host cells may contribute to or are responsible for VEGF-C/D production in some tumors. Therefore, expression of these growth factors in tissue culture cannot be reliably extrapolated to the in vivo situation.

In addition to stimulating lymphangiogenesis, VEGF-C and VEGF-D may also contribute to angiogenesis. The fully processed forms of VEGF-C and VEGF-D have the ability to activate VEGFR-2, although they are bound with lower affinity by VEGFR-2 compared with VEGFR-3 (3, 4). VEGFR-2 plays an important role in tumor angiogenesis (37, 38), and the secretion by tumors of VEGF-C and VEGF-D could therefore also play a part in tumor angiogenesis. However, VEGF-C and VEGF-D do not always induce angiogenesis. For example, transgenic expression of VEGF-C in the skin of mice stimulated lymphangiogenesis in the dermis but did not result in new blood vessel growth (39). Moreover, when recombinant VEGF-C was placed on the chick chorioallantoic membrane, lymphangiogenesis was stimulated far more efficiently than hemangiogenesis (40). On the other hand, VEGF-C induces angiogenesis in the context of tissue ischemia (41), in the cornea (42), and in an in vitro angiogenesis assay (43). Thus the context of expression of VEGF-C and VEGF-D may be critical in determining the angiogenic response, in addition to the processing status.

VEGFR-3 can be expressed on blood capillaries in the tumor context (44). Furthermore, it has been shown that VEGF-C/D expression promotes angiogenesis in the context of some tumors (e.g., Refs. 16 and 45), but not others (e.g., Ref. 21). Using PECAM-1 staining, we were unable to observe any significant difference in intratumor blood vessel density between tumors expressing VEGFR-3-Rg (MT-450) or VEGF-C/Cys152Ser (NM-081) and the respective parental tumors (data not shown).

Like VEGF-C and VEGF-D, VEGF, a pivotal inducer of tumor angiogenesis, is also expressed in tumors by both tumor and stromal cells (reviewed in Ref. 46). The expression of these factors in stromal cells suggests that the cellular makeup of the tumor stroma has an influence on the lymphangiogenic (and angiogenic) response of the host to the tumor. Because fibroblasts constitute a major component of the intratumor stromal cell population, our finding that stromal cells can contribute to the expression of VEGF-C and VEGF-D in tumors is perhaps not surprising. VEGF-D was first identified in murine fibroblasts as figf, a gene induced by the proto-oncogene c-fos (47). Moreover, we see constitutive VEGF-C expression in rat embryonic fibroblasts (CREF, Fig. 4) and in murine 3T3 cells (data not shown). Furthermore, it was recently shown that tumor-associated macrophages can express VEGF-C/D (48). Our observation that tumor cells may not express VEGF-C/VEGF-D in culture but do so within tumors also suggests that the molecular constituents of the tumor microenvironment may be critical in determining the expression of these factors. It has been shown, for example, that the pro-inflammatory cytokine interleukin 1β, which is expressed in certain tumors (49), can up-regulate VEGF-C expression (50).

Our data and that of others suggest that expression of VEGF-C and VEGF-D promotes tumor metastasis. We have demonstrated with our VEGFR-3-Rg experiments in immunocompetent animals that blockade of the activity of these factors might have therapeutic potential for cancer. Such therapies would have the potential to reduce the incidence of metastasis after the onset of treatment, although it is unlikely that the growth of preexisting metastases would be affected. Chronic blockade of VEGFR-3 activation may also be useful for several types of slow-growing, relatively benign tumors, where no clinical intervention is made in cases in which it is not clear whether the potential dangers and costs associated with intervention are justified. The risk of this type of decision is that the tumor progresses and metastasizes, which has a serious implication for the prognosis of the patient. In the case of stage T1a prostate tumors, for example, progression rates of 16–25% at 8–10 years have been reported in cases where no treatment is given (51). Chronic administration of inhibitors of tumor metastasis may therefore be of use in treating these and similar tumor types.

In conclusion, our data demonstrate that the secretion of the lymphangiogenic factors VEGF-C and VEGF-D can induce lymphangiogenesis and that this is one mechanism whereby metastasis via the lymphatic system can be potentiated. Other mechanisms that could also contribute to lymphatic metastasis include chemotactic or chemokinetic stimulation of tumor cells to enter the lymphatics and the passive transport of tumor cells into the lymphatics through interstitial fluid flow. Future work will be directed at functionally determining the contribution of each of these mechanisms to lymphatic metastasis.

We thank Monika Pech for excellent technical assistance, Norma Howells and Andrea D’ercole for animal care, and Hartmut Richter for the kind gift of RTE2 cells. The assistance of Markus Breig with animal photography is gratefully acknowledged.