Vascular endothelial growth factor reduces tamoxifen efficacy and promotes metastatic colonization and desmoplasia in breast tumors

Abstract

Clinical studies have shown that decreased tamoxifen effectiveness correlates with elevated levels of vascular endothelial growth factor (VEGF)-A(165) in biopsy samples of breast cancers. To investigate the mechanisms underlying tamoxifen resistance and metastasis, we engineered the estrogen receptor (ER)-positive MCF-7 human breast cancer cell line to express VEGF to clinically relevant levels in a doxycycline-regulated manner. Induction of VEGF expression in orthotopically implanted xenografts that were initially tamoxifen responsive and noninvasive resulted in tamoxifen-resistant tumor growth and metastasis to the lungs. Lung metastases were also observed in a VEGF-dependent manner following tail vein injection of tumor cells. At both primary and metastatic sites, VEGF-overexpressing tumors exhibited extensive fibroblastic stromal content, a clinical feature called desmoplasia. VEGF-induced metastatic colonies were surrounded by densely packed stromal cells before detectable angiogenesis, suggesting that VEGF is involved in the initiation of desmoplasia. Because expression of VEGF receptors R1 and R2 was undetectable in these tumor cells, the observed VEGF effects on reduction of tamoxifen efficacy and metastatic colonization are most likely mediated by paracrine signaling that enhances tumor/stromal cell interactions and increases the level of desmoplasia. This study reveals new roles for VEGF in breast cancer progression and suggests that combination of antiestrogens and VEGF inhibitors may prolong tamoxifen sensitivity and prevent metastasis in patients with ER-positive tumors.

Figure 1

Doxycycline-inducible VEGF expression. Conditioned media were collected from cell cultures of parental MCF-7, pBTE-transfected C9, C9V4, and C9V18, treated with or without 0.2 µg/mL doxycycline (Dox) for 48 h and Western blotted with a specific antibody against human VEGF (R&D Systems). The upper band of VEGF with higher molecular weight is most likely due to the posttranslational modifications. E. coli–expressed human recombinant VEGF lacking posttranslational modifications was used as standard in the first two lanes. Each lane contains the equivalent VEGF secreted by ~500 cells over a 48-h period.

Figure 2

VEGF expression increases tumor growth and vascularization. A, tumor growth in athymic nude mice. C9V4 cells (10⁷), suspended in 200 µL of 50% phenol red–free Matrigel (BD Biosciences), were inoculated into the inguinal mammary fat pad of four groups of ovariectomized athymic nude mice on day 0. A slow-release 0.72-mg E₂ pellet (Innovative Research) was implanted into each mouse the day before the cell inoculation. The control group was provided with drinking water containing 2.5% sucrose only; three other groups were provided with 2.5% sucrose and 0.05 or 0.2 mg/mL of doxycycline (dox) starting the day before the cell inoculation. Doxycycline was removed from one of 0.2 mg/mL groups on day 11. There were six mice for each group (n = 6). Tumor sizes were measured twice a week. Symbols, mean tumor volume; bars, SE. B, relationship between tumor growth and VEGF levels. Cytosolic extracts were prepared from xenograft tumors as described in ref. , and the amount of VEGF present was determined by ELISA (R&D Systems) and normalized to total protein concentration. The mean tumor volume at the end of the study is indicated by the red columns and the average tumor VEGF level of each mouse group (n = 6) is indicated by the green columns. C, VEGF overexpression is associated with increased tumor microvessel density. Paraffin-embedded tumor samples prepared from the animal study were sectioned and immunostained with a rat monoclonal antibody against mouse PECAM-1 (BD Biosciences). Representative images of tumor microvessels from two tumors, one with 0.2 mg/mL doxycycline induction of VEGF expression and the other without. Quantitative analysis, as shown in the histogram, was done by counting the number of PECAM-1–positive blood vessels per microscopic field with five tumor samples for each condition and three microscopic fields per sample (×10 magnification).

Figure 3

Induction of VEGF expression results in tamoxifen-resistant tumor growth and spontaneous lung metastases. A, tumor growth in tamoxifen-treated mice. Four groups of E₂-supplemented mice were inoculated with 10⁷ C9V18 cells on day 0. The day before inoculation, doxycycline was supplied at 0.2 mg/mL in the drinking water for groups 3 and 4. On day 11, when the average tumor size in all groups reached ~200 mm³, the animals in groups 1 and 3 had their E₂ pellet replaced with a tamoxifen pellet (5-mg 60-day release; Innovative Research). B, Ki-67 immunostaining. Paraffin-embedded tumor samples collected from tamoxifen (Tam)-treated xenograft mice were sectioned and immunostained with antibody against Ki-67 (Santa Cruz), a cell proliferation marker. The number of Ki-67–positive cells per microscopic view is approximately doubled when VEGF was induced by doxycycline. Quantitative analysis, as shown in the histogram, was done by counting the number of Ki-67–positive cells per microscopic field. Three microscopic fields (×20 magnification) per tumor and six tumors per group were analyzed. C, macroscopic metastases. X-gal staining of the lungs (top) showed metastatic deposits of LacZ-tagged tumor cells in the tamoxifen-treated tumor-bearing mice that also received doxycycline (group 3). Top left, a mouse lung with a number of metastatic deposits. Top right, an image with higher magnification. Bottom, H&E staining of lung samples confirming the presence of metastatic lesions.

Figure 4

Exogenous VEGF does not affect the sensitivity of C9V18 cells to estrogen or tamoxifen in vitro presumably due to the lack of VEGF tyrosine kinase receptor expression. A, in vitro cell proliferation assay. C9V18 cells were plated at a density of 10⁴ per well in 24-well plates in 10% FBS IMEM. After 24 h, the culture media were exchanged with phenol red–free IMEM containing 0.1% charcoal-stripped calf serum to remove the estrogens overnight. Then, three wells were counted on day 0, and the indicated reagents in 1% charcoal-stripped calf serum, phenol red–free IMEM were added to the cultures in triplicate. The concentrations used were VEGF, 20 ng/mL; FGF-2, 20 ng/mL; 4-hydroxytamoxifen, 10⁻⁷ mol/L; and E₂, 10⁻⁸ mol/L. The media were exchanged with the same conditions on day 4 and the cell numbers were counted with a Coulter counter on day 7. B, real-time quantitative RT-PCR analysis. Expression of VEGF receptors R1, R2, and NP-1 was examined with quantitative RT-PCR as described in Materials and Methods. The number of cells in each culture used for mRNA isolation was determined to allow calculation of the specific number of mRNA molecules per cell. mRNAs of the receptors R1 and R2 were determined to be <1 molecule per cell in MCF-7 and C9V cells, whereas about 200 R1 mRNAs and 5,000 R2 mRNAs are expressed per cell in human umbilical vein endothelial cells (HUVEC).

Figure 5

VEGF expression increased the extent of the desmoplastic response observed at both primary and metastatic sites and facilitates metastatic colonization in an experimental metastasis assay. A, cytokeratin immunostaining. Tumor samples collected from tamoxifen-treated mice were immunostained for an epithelial specific cell marker using mixed antibodies against cytokeratin (Keratin Pan Ab-1, NeoMarkers) and counterstained with hematoxylin. Because C9V tumor cells were cytokeratin positive, the cytokeratin-negative areas were considered to be the stromal-enriched compartment within the tumors. The fraction of the cytokeratin-negative stromal compartment within the total area of the microscopic view was analyzed microscopically with the aid of a counting grid and determining the stromal areas over the total areas for 5 primary tumors in each group using three sections per tumor separated by 100 µm. The results are plotted as percentage of stromal area over total counted areas. Increased desmoplasia correlated with doxycycline induction of VEGF expression. B, FGF-2 ELISA assay. Cytosolic tumor FGF-2 was determined by ELISA and normalized by protein concentration. Tumor FGF-2 levels correlated with the induction of tumor VEGF expression. C, experimental metastasis assay. C9V18 cells (0.5 × 10⁶) were directly introduced into the blood circulation by tail vein injection. Doxycycline was provided in the drinking water at 0.2 mg/mL to one of two E₂-supplemented mouse groups starting the day before cell injection. Mouse lung tissues were harvested at 2 h, 4 wk, and 10 wk after the tail vein injection. Paraffin-embedded lung samples were sectioned and immunostained with the cytokeratin antibodies to identify C9V18 cells. Red arrows, disseminated tumor cells; black arrows, fibroblastic stromal cells that surrounded the cytokeratin-positive C9V18 colonies.