Tumor-Induced Activation of Lymphatic Endothelial Cells via Vascular Endothelial Growth Factor Receptor-2 Is Critical for Prostate Cancer Lymphatic Metastasis

Lymphatic vessels are one of the major anatomic pathways for prostate cancer dissemination (1). Although the importance of intratumoral lymphatic vessels in mediating lymphatic metastasis is controversial (2, 3), peritumoral preexisting lymphatic vessels have been proven to be sufficient for lymphatic metastasis in several studies (2, 4, 5). However, because peritumoral lymphatic vessels exist before tumor development in nearly all tissues, why then do tumors of some origins preferentially metastasize to lymph nodes? Studies on metastasis patterns and lymphatic mapping of human cancers have shown that the dissemination of tumor cells to lymph nodes is nonrandom (1, 6, 7), suggesting that active molecular interactions between tumor and lymphatic endothelium are crucial for lymphatic metastasis.

The nature of interactions of tumor cells with lymphatic endothelium is poorly understood. Vascular endothelial growth factor (VEGF)-C and VEGF-D have been reported to activate lymphatic endothelium and promote lymph node metastasis via the VEGF receptor (VEGFR)-3 pathway (8–11). However, targeting the VEGF-C/VEGF-D/VEGFR-3 pathway to inhibit tumor lymph node metastasis has not been successful in some tumor models, including prostate cancer (2). In addition, the expression of VEGF-C or VEGF-D does not correlate with lymph node metastasis in all human tumors (12), implying that other factors participate in tumor lymphatic metastasis.

There remains much to be discovered about the mechanisms underlying tumor lymphatic metastasis. In the current study, we have used two-dimensional and three-dimensional in vitro coculture systems and in vivo orthotopic xenograft models to investigate the interactions of prostate cancer cells with prostate lymphatic endothelial cells (LEC) and show that the activation of LECs by cancer cells is associated with lymph node metastasis.

Reagents. The following antibodies were used: mouse α-human podoplanin (D2-40; Signet Laboratories, Dedham, MA), mouse α-human CD31 (DAKO, Glostrup, Denmark), mouse α-human CD34 (NeoMarkers, Fremont, CA), mouse α-human N-cadherin (Invitrogen, Carlsbad, CA), mouse α-human fibroblast antigen (Oncogene, San Diego, CA), mouse α-human cytokeratin (AE1/AE3, DAKO), mouse α-human pan-actin (Ab-5, NeoMarkers), mouse α-human mitochondria (Chemicon, Temecula, CA), mouse α-human VEGF-A (NeoMarkers), rabbit α-human VEGF-A (Genentech, San Francisco, CA), rabbit α-human Prox1 (Research Diagnostics, Flanders, NJ), goat α-mouse LYVE-1 (Santa Cruz Biotechnology, Santa Cruz, CA), rat α-mouse Ki-67 (DAKO), rabbit α-green fluorescent protein (α-GFP; Molecular Probes, Eugene, OR), Alexa Fluor 488–, Alexa Fluor 568–, and Alexa Fluor 680–conjugated secondary antibodies (Molecular Probes), 800 CW IRDye–conjugated secondary antibodies (Rockland, Gilbertsville, PA), and Dynabeads M-450 sheep α-mouse IgG (Invitrogen). Goat α-human VEGF-C, mouse α-human VEGF-D, goat α-mouse VEGFR-2, mouse α-human VEGFR-2 (used in neutralizing experiments), and goat α-human VEGFR-3 antibodies and recombinant human VEGF-A, VEGF-C, VEGF-C (Cys156Ser), and VEGF-D protein were purchased from R&D Systems (Minneapolis, MN).

Isolation of lymphatic and vascular endothelial cells from human prostate. Primary cultures were established using enzymatic digestion of fresh prostate tissue obtained from patients undergoing radical prostatectomy. The portions of tissue used for this study contained no prostate carcinoma as determined by histologic evaluation of immediately adjacent serial sections. Finely minced tissues were incubated with PBS containing 0.25% collagenase II (Worthington, Lakewood, NJ) and 0.01% DNase I (Worthington) for 30 to 45 minutes at 37°C followed by gently squeezing endothelial cells into DMEM (Invitrogen) containing 10% fetal bovine serum (FBS; JRH Biosciences, Brooklyn, Victoria, Australia). The resulting cell suspension was passed through a 100-μm nylon cell strainer, pelleted, resuspended, and plated in EGM-2 MV medium (Cambrex, Walkersville, MD) on fibronectin-coated (10 μg/mL; Sigma, St. Louis, MO) flasks at 37°C, 5% CO₂ in a humidified atmosphere. Cell selection was done as previously described (13) in the subconfluent primary culture that consists of endothelial cells, fibroblasts, and epithelial cells. Briefly, prostate CD34-positive blood vessel endothelial cells (BVEC) were isolated by immunomagnetic purification with anti-human CD34 antibody-conjugated immunomagnetic beads. Prostate LECs were then isolated by incubation of the remaining CD34-negative cells with anti-human CD31 antibody-conjugated immunomagnetic beads. Prostate LECs were propagated in LEC growth medium (EGM-2 MV medium supplemented with 50 ng/mL VEGF-C), and BVECs were cultured in EGM-2 MV without VEGF-C. LECs between passages 3 and 6 were used for the experiments described herein. All studies were conducted with the approval of the St. Vincent's Hospital (SVH) Human Research Ethics Committee and in accordance with Australian National Health and Medical Research Council (NHMRC) guidelines.

Cell culture, transfection, and vital dye labeling. Human prostate cancer cell lines LNCaP (UroCor, Oklahoma City, OK), PC-3, and DU145 [both from the American Type Culture Collection (ATCC), Rockville, MD] were cultured in DMEM supplemented with 10% FBS. Human prostate epithelial cell line RWPE-1 (ATCC) was cultured in keratinocyte serum-free medium (Invitrogen). Primary human umbilical vascular endothelial cells (HUVEC; Cambrex) were maintained in EGM medium (Cambrex). PC-3 and DU145 cells were transfected with either pEGFP-N1 (BD Biosciences, Bedford, MA) or pDsRed2-N1 plasmid (BD Biosciences) using LipofectAMINE 2000 (Invitrogen) as per the manufacturer's protocol, and stable transfectants were selected by G418 (800 μg/mL; Invitrogen) and sorted by fluorescence-activated cell sorting to obtain the brightest population (top 10%). Prostate LECs, LNCaP, and RWPE-1 cells were labeled with CM-DiI (CellTracker, Molecular Probes) as per the manufacturer's protocol.

Immunofluorescence and immunohistochemistry. Immunofluorescence analyses were done on isolated and Matrigel-embedded (BD Biosciences) cells as previously described (see Supplementary Data for details; ref. 14). Immunohistochemistry of human prostate and mouse xenograft tissues was done as described (15, 16). 4T1 mouse mammary tumor, a kind gift from A/Prof. Robin Anderson (Peter MacCallum Cancer Center, East Melbourne, Victoria, Australia; ref. 17), was used as a positive control for α-mouse Ki-67 staining.

Western blot analyses. Western blotting was done as previously described (see Supplementary Data for details; ref. 14). Signals were visualized using an Odyssey Infrared Imaging System (LI-COR Biosciences, Lincoln, NE).

Real-time quantitative reverse transcription-PCR. Quantitative reverse transcription-PCR (RT-PCR) was done as previously described (14). The primer sequences used were as follows: Prox1, 5′-CTGAAGAGCTGTCTATAACCAG-3′ (forward) and 5′-GGATCAACATCTTTGCCTGCG-3′ (reverse; 40 cycles); podoplanin, 5′-CGAGGATCTGCCAACTTCAGAAA-3′ (forward) and 5′-CAACCAGGGTCACTGTTGACAAA-3′ (reverse; 30 cycles); CEA-CAM, 5′-GTAGCAAAGCCCCAAATCAA-3′ (forward) and 5′-AACGGATGGAGATTCCAGTG-3′ (reverse; 30 cycles); VEGFR-2, 5′-GTCAAGGGAAAGACTACGTTGG-3′ (forward) and 5′-AGCAGTCCAGCATGGTCTG-3′ (reverse; 50 cycles); and VEGFR-3, 5′-CACTCCCGCCATACGCCACATCAT-3′ (forward) and 5′-CTGCTCTCTATCTGCTCAAACTCC-3′ (reverse; 30 cycles). The relative expression level of each target gene was normalized to the ribosomal housekeeping gene L32 as previously described (14). All quantitative PCR products were visualized on 2% agarose gel containing 1 μg/mL ethidium bromide (Bio-Rad, Hercules, CA).

Tube formation assay. To assess the tube-forming ability of isolated endothelial cells, standard Matrigel was allowed to polymerize in a 96-well plate and prostate LECs or BVECs were seeded at a density of 3 × 10⁴ per well in EGM-2 MV medium. For coculture experiments, a mixture of DiI-labeled LECs (1.5 × 10⁴ per well) with each prostate cell line (7.5 × 10³ per well) was seeded on growth factor–reduced (GFR) Matrigel in serum-free DMEM. In control cultures, prostate LECs were plated at two cell densities: low density (same as the LEC density in coculture) and high density (same as the total cell density in coculture). For VEGFR-2 blocking experiments, the mixed cells were resuspended in DMEM serum-free medium with 20 μg/mL of α-human VEGFR-2 or control IgG (R&D Systems) before seeding on GFR Matrigel. To assess growth factor effects, prostate LECs were seeded on GFR Matrigel at a density of 1.5 × 10⁴ per well in serum-free DMEM with or without specified growth factors. The concentrations of recombinant human growth factors used were as follows: VEGF-A, 100 ng/mL; VEGF-C, 100 ng/mL; VEGF-C156S, 500 ng/mL; and VEGF-D, 500 ng/mL. All tube formation experiments were observed using inverted fluorescence microscopy (Olympus, Tokyo, Japan), and images were digitally captured (Olympus) at 0, 3, 6, 9, 12, 24, and 48 hours after plating. The total length, area, and number of tube-like structures formed by LECs in each well were measured using Axiovision Rel 4.2 software (Carl Zeiss AG, Jena, Germany). All experiments were done using three different prostate LEC cell lines.

Wound-healing assay. DiI-labeled prostate LECs (10⁴ per well) were mixed with prostate cell lines (2.5 × 10³ per well), seeded in 96-well plates, and grown in EGM-2 MV medium overnight. A wound was made by scraping the confluent monolayer, and the cells were grown in EGM basal medium supplemented with 2% FBS. To assess conditioned medium (collected from subconfluent prostate cell cultures in serum-free DMEM for 48 hours) and growth factor effects, prostate LECs (10⁴ per well) were grown in a 96-well plate overnight. After wounding, cells were incubated with EGM basal medium/2% FBS with or without VEGFs or with 30% conditioned medium. The concentrations of recombinant human VEGFs used were same as those used in the tube formation assays. For the VEGFR-2 inhibition assay, cells were incubated with the corresponding growth medium containing 20 μg/mL of α-human VEGFR-2 or control IgG immediately after wounding. The migration of the cells was digitally recorded (Olympus) at 0, 9, 16, 20, 25, 32, 45, 54, 71, 96, 100, and 106 hours after wounding. The experiment was terminated when LEC wound was closed. The area of the uncovered wound gap was measured using Axiovision Rel 4.2 software, and the percentage of wound closure was calculated at each time point. All experiments were done using duplicates of three different prostate LEC cell lines.

In vitro growth. For coculture assays, prostate LECs (200 per well) were mixed with specified prostate cell lines (200 per well) and seeded in 384-well plates in EGM-2 MV medium. For conditioned medium and growth factor experiments, prostate LECs (200 cells well) were seeded in 384-well plates in EGM basal medium/2% FBS with or without VEGFs or with 30% conditioned medium. The concentrations of human recombinant growth factors used were same as those used in the tube formation assays. At days 0 to 4, cells were formalin fixed and immunostained with α-human CD31. The total number of LECs (CD31-positive cells) in each well was counted. All experiments were done in triplicate.

Xenotransplantation and in vivo imaging tumor metastasis. Five-week-old severe combined immunodeficient mice (Animal Resources Centre, Perth, Western Australia, Australia) were anesthetized, and 10⁶ prostate cancer cells (PC-3-DsRed, DU145-DsRed, or LNCaP) were inoculated into the prostate. Tumor growth was monitored weekly using in vivo fluorescence imaging from 24 days after inoculation. Mice were anesthetized, abdominal hair was removed, and fluorescence was emitted from the tumors and/or metastases were detected using the Kodak (Kodak, New Haven, CT) or LAS3000 (Fuji, Tokyo, Japan) Imaging Systems. Images were digitally captured and overlaid onto the X-ray reference image (Kodak Imaging System). Mice were sacrificed 7 to 8 weeks after tumor cell inoculation. At harvest, primary tumors and regional lymph nodes were removed, imaged ex vivo, measured using digital calipers, snap frozen (half), and formalin fixed (half) for further analysis. Primary tumor volume was calculated as length × width² × 0.5 as previously described (18). All animal studies were conducted with the approval of the SVH Animal Ethics Committee and in accordance with NHMRC guidelines.

Quantification of tumor lymphatic vasculature. Lymphatic vessels were identified using LYVE-1 immunostaining, and lymphatic vessel density (LVD) was evaluated using absolute counting (16). Briefly, lymphatic vessels were counted within each tissue section in consecutive intermediate power fields (×200, field diameter 1 mm) to assess LVD in tumor center area (within tumor and >1 mm from tumor margin), tumor border area (within tumor and <1 mm from tumor margin), tumor periphery (normal tissue within 1 mm of tumor margin), and normal tissue away from tumor (tissue 1-10 mm from tumor margin). In four normal mouse prostates, lymphatic vessels were counted throughout the whole tissue section. LVD was expressed as the number of vessels per mm².

Statistical analyses. Data obtained in the tube formation, wound healing, and in vitro growth assays were analyzed using two-way repeated measures ANOVA followed by Bonferroni post-test for multiple comparisons. Analyses of LVD and primary tumor volume of prostate xenografts were done using nonparametric one-way ANOVA followed by Bonferroni post-test for multiple comparisons. The correlations between LVD, cancer lymphatic invasion, and lymph node status were determined using the Pearson correlation test. The results are presented as mean ± SE. P < 0.05 was considered statistically significant. All calculations were done using GraphPad Prism 4.0 (GraphPad Software, San Diego, CA).

Human prostate LECs maintain their lineage-specific phenotype in vitro. In prostate tissue, CD34 was expressed specifically in blood vascular endothelium, whereas the endothelial junction molecule CD31 was expressed in all vascular endothelium (Supplementary Fig. S1). Isolated prostate LECs maintained a typical cobblestone-like endothelial morphology in monolayer culture indistinguishable from BVECs (Fig. 1A). Double-immunofluorescent staining with antibodies to CD31 and the lymphatic-specific transcription factor Prox1 (19) revealed the membrane labeling for CD31 in both cell lineages, whereas only LECs showed nuclear staining for Prox1 (Fig. 1A). Podoplanin, another lymphatic-specific gene (13), was also selectively expressed by prostate LECs in both the membrane and the cytoplasm as determined by D2-40 immunostaining (Fig. 1A). The selective expression of Prox1 and podoplanin in prostate LECs was confirmed at both protein and mRNA levels (Fig. 1B and C). Prostate LECs were further characterized by VEGFR-3 (13) and CEA-CAM (13) expression (Fig. 1B and C). In contrast, prostate BVECs, but not LECs, expressed CD34 and N-cadherin protein (Fig. 1B; ref. 20). Both prostate LECs and BVECs were able to form capillary-like networks in three-dimensional standard Matrigel when resuspended in EGM-2 MV medium (Fig. 1A). Neither prostate LECs nor BVECs expressed fibroblast antigen or cytokeratins (data not shown). Both LECs and BVECs maintained their lineage-specific characteristics for the eight passages examined.

PC-3 and LNCaP, but not DU145 and RWPE-1, induce prostate LEC capillary-like structure formation in vitro. To determine the effect of prostate cells on prostate LEC tube formation, we established a three-dimensional coculture system consisting of DiI-labeled prostate LECs and prostate cancer/normal cell lines. To minimize the effects of exogenous growth factors on prostate LEC tube formation, the mixture of cells was seeded on GFR Matrigel in serum-free DMEM. When cultured alone, PC-3 and LNCaP cells elongated their processes slightly, whereas DU145 and RWPE-1 formed small cell aggregates (data not shown). Prostate LECs formed small tight clusters without sprouting or elongation (Fig. 2A). In coculture, prostate LECs displayed dramatically increased network formation in the presence of PC-3 or LNCaP cells but not DU145 or RWPE-1 cells (Fig. 2A). In the presence of PC-3 or LNCaP cells, prostate LECs started to sprout, elongate, and reorganize to form a network structure within 3 to 4 hours of plating. An extensive LEC network was formed from 6 to 24 hours after plating (Fig. 2A). In contrast, DU145 and RWPE-1 cells had no effect on prostate LECs, which displayed a similar morphology to those in control cultures (Fig. 2A). From 6 hours after plating, there were significant increases in tube length (P < 0.0001), tube area (P < 0.01), and number of tubes (P < 0.01) for prostate LECs cocultured with PC-3 or LNCaP compared with either those cocultured with DU145 or RWPE-1 or those in control cultures (Fig. 2B). Staining of sections through the GFR Matrigel plugs for CD31 and podoplanin confirmed that the LEC network structure was located throughout the gel (Fig. 2E).

To further explore how prostate cancer cells interact with prostate LECs, DiI-labeled LECs were cocultured with either PC-3 or DU145 transfected with enhanced GFP gene (PC-3-EGFP or DU145-EGFP). Within 3 to 4 hours after plating, prostate LECs and PC-3 cells started to elongate, migrate, and form cell-cell contacts. From 6 hours after plating, elongated LECs and PC-3 cells aligned themselves end to end or side by side to form an integrated network structure throughout the Matrigel (Fig. 2C-E). Tumor cells remained negative for endothelial markers. In contrast, DU145 and prostate LECs formed cell aggregates in coculture (Fig. 2C and D).

PC-3 and LNCaP, but not DU145 and RWPE-1, promote prostate LEC migration in vitro. To determine the effects of prostate cells on prostate LEC motility, DiI-labeled LECs were mixed with prostate cell lines before commencement of the wound-healing assay. PC-3 and LNCaP, but not DU145 and RWPE-1, promoted prostate LEC spreading and migration. At 25 hours, prostate LECs cocultured with PC-3 or LNCaP consistently filled the wound, whereas LECs alone or cocultured with DU145 or RWPE-1 failed to cover most of the scratched surface (Fig. 3A,, left). PC-3 and LNCaP significantly increased LEC migration from 9 hours after scraping compared with LECs cultured alone (P < 0.001; Fig. 3B,, left). Interestingly, DU145 significantly reduced LEC migration from 16 hours after scraping (P < 0.05; Fig. 3B,, left). RWPE-1 had no significant effect on LEC migration (Fig. 3B , left).

To further investigate how prostate cells interact with LECs in the coculture wound-healing assay, we plated DiI-labeled LECs mixed with PC-3-EGFP or DU145-EGFP cells. Despite their opposite effects on LEC migration, both PC-3 and DU145 migrated with prostate LECs (Fig. 3A , right).

We then examined the effect of factors secreted by prostate cancer/normal epithelial cells on prostate LEC migration using conditioned medium. Conditioned medium from PC-3 or LNCaP significantly promoted prostate LEC migration (P < 0.01; Fig. 3B,, right), whereas conditioned medium from DU145 or RWPE-1 had no effects on prostate LEC migration (P > 0.05; Fig. 3B , right).

Prostate cancer and normal epithelial cells inhibit growth of prostate LECs in vitro. Prostate cancer/normal epithelial cells were cocultured with prostate LECs to assess their effect on prostate LEC proliferation. All four prostate cell lines significantly inhibited the growth of prostate LECs (P < 0.001; Fig. 3C,, left). From day 3, the inhibitory effect of PC-3, LNCaP, and DU145 cells was stronger than that of RWPE-1 cells (P < 0.05). The inhibitory effects were related to cell number, as higher ratios of prostate cancer/normal epithelial cells to LECs resulted in a further decrease of LEC cell number (data not shown). Conditioned medium collected from the prostate cancer/normal epithelial cells showed similar inhibitory effects on LEC growth (P < 0.05; Fig. 3C , right).

PC-3 and LNCaP, but not DU145, invade lymphatic vessels and metastasize to the lymph nodes in vivo. We next investigated the in vivo behavior of PC-3, LNCaP, and DU145 cells. Microscopic lymphatic invasion, characterized by intravasation of tumor cells into LYVE-1-positive lymphatic vessels, was observed in the tumor periphery and tumor border area in 100% of PC-3 and LNCaP prostate tumors, whereas no tumor-containing lymphatic vessels were observed in any region of DU145 prostate tumors (Fig. 4A). Moreover, 100% of mice in the PC-3 (18 of 18) and LNCaP groups (9 of 9) developed lymph node metastases, whereas no lymph node metastases were observed in DU145 group (18 of 18) as determined by macroscopic examination and histologic evaluation. The presence of microscopic lymphatic invasion was significantly correlated with lymph node status [Pearson correlation coefficient r = 1.000; 95% confidence interval (95% CI), 1.000-1.000; P < 0.0001]. In a subset of mice (PC-3 and DU145 tumor bearing), tumor progression and anatomic localization of tumor metastases were followed temporally using in vivo fluorescence imaging. In PC-3 tumor-bearing mice, regional lymph node metastasis was detected from 24 days after inoculation (Fig. 4B). In contrast, no lymph node metastases were detected in DU145 tumor-bearing mice using in vivo fluorescent imaging (Fig. 4B). Ex vivo fluorescence imaging and immunohistochemistry detecting human mitochondria of the removed lymph nodes confirmed the presence/absence of lymph node metastases (Fig. 4C).

PC-3 cells, but not DU145 prostate tumor cells, enhance lymphatic vessel formation in vivo. As the D2-40 antibody against lymphatic vessel marker podoplanin does not recognize the murine antigen, mouse lymphatic vessels were detected using LYVE-1 immunohistochemistry. The analysis of LNCaP tumors was compromised by the significantly larger tumors, which were greater than that of PC-3 xenografts (P < 0.01) and DU145 xenografts (P < 0.001; Fig. 4D,, top). This was accompanied by numerous necrotic areas in the LNCaP tumors, making it difficult to consistently identify tumor periphery and tumor border areas. Thus, no further analysis of these tumors was done. In PC-3 prostate tumors, lymphatic vessels were preferentially located in tumor periphery and tumor border area (Fig. 4A,, top left). LVD was significantly increased in the tumor periphery (17.79 ± 2.32 mm⁻²) and tumor border area (25.95 ± 3.70 mm⁻²) compared with LVD in normal mouse prostate (3.55 ± 1.14 mm⁻²; P < 0.001) or that in normal tissue away from tumor (2.21 ± 0.62 mm⁻²; P < 0.001), whereas LVD in the tumor center area (3.39 ± 1.36 mm⁻²) was not different from that either in normal mouse prostate or in normal tissue away from tumor (Fig. 4A , bottom right). In contrast, in DU145 tumors there were no LYVE-1-positive lymphatic vessels within the tumor area, although LVD in the tumor periphery (2.92 ± 0.61 mm⁻²) was not different from that in normal tissue away from tumor (2.68 ± 1.52 mm⁻²) or that in normal mouse prostate. Moreover, there was a significant correlation between lymph node status and LVD in the tumor periphery (Pearson correlation coefficient r = 0.9135; 95% CI, 0.6939-0.9777; P < 0.0001) or LVD in the tumor border area (Pearson correlation coefficient r = 0.8877; 95% CI, 0.7100-0.9591; P < 0.0001).

To assess whether the increase of lymphatic vessels in tumor periphery and tumor border area in PC-3 tumors was accompanied by LEC proliferation, Ki-67 immunohistochemistry was done. Ki-67 staining was not detected in any LYVE-1-positive lymphatic vessels despite strong nuclear staining in the positive control tissue (Fig. 4D , bottom).

Differential expression of VEGF family members in prostate cells and xenografts and colocalization of VEGFR-2 and VEGFR-3 in prostate LECs. Because of the established importance of VEGF family members in tumor lymphatic metastasis, we assessed the protein levels of VEGF-A, VEGF-C, and VEGF-D in prostate cell lines and xenografts using Western blotting analysis. For VEGF-A, all four prostate cell lines showed a weak but comparable ∼46 kDa band (Fig. 5A), representing the dimeric glycosylated VEGF₁₆₅. In addition to the well-documented VEGF-A form, both PC-3 and LNCaP cells exhibited a strong ∼57 kDa band (Fig. 5A), which was also present in HUVECs (Fig. 5A). This has previously been shown to be a specific VEGF-A form in PC-3 xenografts (21). In contrast, this VEGF-A form was barely detected in either DU145 or RWPE-1 cells (Fig. 5A). Moreover, PC-3 mouse prostate carcinomas showed a much higher level of ∼57 kDa VEGF-A than DU145 mouse tumors (Fig. 5A). For VEGF-C, detectable level of VEGF-C full-length form (∼58 kDa) was present in PC-3, LNCaP, and DU145 cells (Fig. 5A). This was strongly up-regulated in PC-3 mouse prostate tumors but not in DU145 tumors (Fig. 5A). For VEGF-D, the unprocessed form of VEGF-D (∼53 kDa) was expressed at much higher level in both PC-3 and LNCaP cells than in DU145 and RWPE-1 cells (Fig. 5A). In vivo, VEGF-D expression was moderately increased in PC-3 tumors compared with that in DU145 tumors (Fig. 5A).

Both VEGFR-2 and VEGFR-3 protein and mRNA were readily detectable in prostate LECs (Fig. 1B and C). Double staining for VEGFR-2 and VEGFR-3 revealed the colocalization of these receptors to the same LECs (Fig. 5B). Moreover, VEGFR-2 and VEGFR-3 were colocalized to LYVE-1-positive lymphatic endothelium in PC-3 xenografts (Fig. 5C).

VEGF-A and VEGF-C, but not VEGF-C156S, potently stimulate prostate LEC tube formation, migration, and proliferation in vitro. Due to the pronounced differences in the expression pattern of VEGF family members in prostate cells and their xenografts, we further examined the effects of these proteins on prostate LECs. Both VEGF-A and VEGF-C were potent stimulators of prostate LEC capillary-like network formation in three-dimensional GFR Matrigel (P < 0.01; Fig. 5D,, left). In contrast, VEGF-D and VEGF-C156S, a mutant form of VEGF-C that only binds to and activates VEGFR-3 (22), were much less effective in promoting LEC tube formation (P < 0.01). VEGF-A, VEGF-C, and VEGF-D stimulated significant migration of prostate LECs (P < 0.01), whereas VEGF-C156S, at a 5-fold higher concentration, had a significantly weaker effect than VEGF-C (P < 0.05; Fig. 5D,, middle). Furthermore, VEGF-A, VEGF-C, and VEGF-D potently stimulated the proliferation of prostate LECs (P < 0.01), whereas VEGF-C156S had no effect (Fig. 5D , right).

Compared with VEGF-C, the relative lack of effects of VEGF-C156S on prostate LEC tube formation, migration, and proliferation indicates that VEGFR-3 pathway is a relatively minor contributor to prostate LEC activation. The high amount of VEGF-A and VEGF-C protein in PC-3 and LNCaP cells and xenografts and the coexpression of VEGFR-2 and VEGFR-3 in prostate LECs prompted us to investigate the role of VEGFR-2 on prostate LEC activation induced by prostate cancer cells.

Blockade of VEGFR-2 inhibits tumor-induced activation of prostate LECs. Neutralizing antibody blockade of VEGFR-2 significantly inhibited prostate LEC capillary-like network formation induced by PC-3 (P < 0.01; Fig. 6A and B) and LNCaP cells (P < 0.01; Fig. 6B). In contrast, it had no effect on the elongation and migration of PC-3 cells (Fig. 6A). Furthermore, blockade of VEGFR-2 significantly inhibited prostate LEC migration induced by PC-3 (P < 0.001; Fig. 6C, and D, top) and LNCaP cells (P < 0.001; Fig. 6D,, top) and their conditioned medium (P < 0.01; Fig. 6D,, middle), whereas it had no effect on the motility of the cancer cells (P > 0.05; Fig. 6C  and D, bottom).

Lymph node metastasis is a complex multifactorial process. Our current study has related the propensity of prostate cancer cells to form lymph node metastases in vivo to their ability to modulate prostate LECs in vitro. Activation of prostate LECs was induced by highly lymph node metastatic prostate cancer cells (PC-3 and LNCaP), including a significant increase in LEC migration, differentiation, and tube formation. Poorly lymphatic metastatic prostate cancer cells (DU145) and nontumorigenic prostate epithelial cells (RWPE-1) had no effect on the activation of prostate LECs. In vivo, PC-3 and LNCaP cells, but not DU145 xenografts, invaded prostatic lymphatic vessels and induced a significant increase in the number of lymphatic vessels in tumor periphery and tumor border area, effects were highly correlated with lymph node metastasis. Notably, prostate cancer cells did not promote LEC proliferation in vitro or in vivo. Together, our results provide direct evidence that certain tumor cells can activate LECs in vitro and this activation process is associated with lymph node metastasis in vivo. Consistent with our results, tumor-activated lymphatic capillaries have been shown to promote tumor cell invasion by increasing tumor cell transendothelial migration (23).

We further show that the active interactions between prostate cancer cells and prostate LECs are mediated via VEGFR-2. VEGFR-2 was coexpressed with VEGFR-3 in prostate LECs, and ligand VEGF-A was highly increased in lymph node metastatic prostate cancer cells and xenografts. Blockade of VEGFR-2 potently inhibited prostate LEC migration, differentiation, and tube formation induced by prostate cancer cells. Although prostate cancer cells express VEGFR-2 (24), blockade of VEGFR-2 had no effects on their differentiation or migration. Consistent with our results, VEGFR-2 activation is associated with the lymphangiogenic activity of VEGF-A in vitro (22, 25) and in vivo (26). Additionally, activation of VEGFR-2 by VEGF-A enhances endothelial cell migration via activation of several integrins (27). Among them, integrin α₂β₁ is highly expressed in human LECs (26). VEGF-A has also been shown to bind quiescent endothelial cells (28) and stimulate the differentiation and network formation of nonproliferating endothelial cells (29). Moreover, VEGF-A can promote lymphatic metastasis through a VEGF-C/VEGF-D/VEGFR-3–independent pathway (25, 30); VEGFR-2 blockade significantly inhibits lymph node metastasis in prostate cancer models (31, 32). In human tumors, there is a significant correlation between lymph node status and VEGF-A or VEGFR-2 expression (33, 34). In prostate cancer, increased plasma levels of VEGF-A and activated VEGFR-2 are associated with lymph node metastasis (35). Together, these data suggest that a VEGF-A/VEGFR-2 paracrine loop is important for LEC activation and tumor lymphatic metastasis.

Notably, the lymph node metastatic prostate cancer cells and their xenografts exhibited a large amount of ∼57 kDa protein of VEGF-A rather than its conventional isoforms. Transgenic mice that express only VEGF₁₆₄ or VEGF₁₈₈ show significantly reduced lymphatic vessel formation compared with wild-type controls that express endogenous VEGF-A (36). These results suggest the involvement of novel VEGF-A isoforms in lymphatic endothelial activation. The inhibitory effect of prostate cancer cells on LEC proliferation in vitro implies that cancer cells may also synthesize LEC mitogenesis inhibitors. Similarly, prostate cancer cells (PC-3, LNCaP, and DU145) have been shown to inhibit the in vitro growth of human bone marrow endothelial cells (37).

Targeting VEGFR-2 might also block the heterodimerization of VEGFR-2 with VEGFR-3, which has been proven to be necessary for VEGFR-3 phosphorylation and cellular responses (38, 39), and thus subsequently inhibit the activation of VEGFR-3 induced by prostate cancer cell–derived unprocessed VEGF-C and VEGF-D. However, blocking VEGFR-2 was less effective in halting lymph node metastasis than blocking VEGFR-3 in a VEGF-C-overexpressing breast cancer model (11). This may be explained by the high levels of VEGF-C-inducing activation of lymphatic endothelium directly via the VEGFR-3 pathway. In contrast, in an orthotopic PC-3 prostate cancer model comparable with that used in the current study, which naturally expresses similar levels of VEGF-A and VEGF-C, blocking either VEGF-C or VEGFR-3 had no effect on the incidence of lymph node metastasis (2). Thus, our study, together with this report, indicates that VEGFR-2 signaling pathway plays a critical role in prostate cancer lymphatic metastasis.

In summary, our results provide strong evidence that tumor-induced activation of host LECs via VEGFR-2 signaling underlies prostate cancer lymphatic metastasis. Thus, targeting VEGFR-2 is likely to have beneficial clinical effects not only by inhibiting tumor blood vasculature formation and tumor growth (40) but also through antagonizing the interactions of tumor cells with lymphatic vessels.

Grant support: University of Melbourne International Postgraduate Research Scholarship (Y. Zeng).

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 Prof. Peter Choong and Dr. Crispin Dass (Department of Orthopedics, SVH, Fitzroy, Victoria, Australia) for the use of their inverted fluorescence microscope.