Metastasis, the dissemination of tumor cells from the primary tumor site, is a major cause of mortality from solid tumors. Metastatic behavior can be facilitated by lymphangiogenesis, the growth of new lymphatic vessels from pre-existing lymphatic vessels.1, 2 Several types of malignancies, including breast,2, 3 melanoma,4 and head and neck,5 predominantly utilize the lymphatic vasculature as the route of dissemination. Tumor lymphatic vessel density (LVD) correlates with increased incidence of intravascular invasion,5, 6 lymph node metastasis,7 cancer recurrence8 and decreased patient survival.3 Thus, understanding the mechanisms of tumor lymphangiogenesis is essential for designing effective anti-metastatic therapies and improving patient survival.
Lymphangiogenesis is thought to be induced primarily by 1 of the 2 well-characterized lymphangiogenic factors: vascular endothelial growth factor C (VEGF-C) or VEGF-D.9, 10 Of these 2 factors, VEGF-C appears to be more clinically significant because its expression is frequently increased in tumors, whereas VEGF-D is highly expressed in normal tissues and often down-regulated in malignancies.11 VEGF-C expression correlates with increased lymphatic vessel invasion,8 metastasis6, 12 and poor survival8, 13 in many human cancers. In experimental models, forced expression of VEGF-C induced intratumoral2, 14, 15 and peritumoral16 lymphangiogenesis associated with increased lymphatic metastasis,2, 16, 17, 18, 19 whereas neutralization of VEGF-C inhibited both processes.15, 20 The lymphangiogenic activity of VEGF-C is mediated by the VEGFR-3 receptor21 which, in adulthood, is primarily expressed on lymphatic endothelial cells (LEC).22, 23 VEGFR-3 activation appears to be critical for tumor dissemination through lymphatics because antibody-mediated blockade of VEGFR-3 has been shown to inhibit both the formation of new tumor lymphatic vessels and lymphatic metastasis.20, 24, 25
Angiogenesis in solid tumors is thought to be primarily driven by vascular endothelial growth factor A (VEGF-A).26, 27 Several recent studies, however, provided extensive evidence demonstrating that VEGF-A can also regulate lymphangiogenesis.25, 28, 29, 30, 31 Nagy et al. were first to show that over-expression of VEGF-A induced new lymphatic vessel formation in the ear and peritoneal lining of nude mice.28 These neo-lymphatics were structurally and functionally abnormal, resembling hyperplasic vessels found in lymphatic malformations, thus suggesting that elevated VEGF-A at malignant and chronically inflamed sites might contribute to pathological lymphangiogenesis. This hypothesis was subsequently supported by studies demonstrating induction of a strong lymphangiogenic response in the rat32 and mouse33 models of corneal injury, chronic inflammation34 and VEGF-A-induced skin tumorigenesis.29 Additionally, VEGF-A mediated lymphangiogenesis has been also shown in T241 fibrosarcoma31 and MDA-MB-435 breast carcinoma cells25 that have been engineered to over-express this factor. In all tumor experimental models studied to date, induction of VEGF-A dependent intratumoral or peritumoral31 lymphatic vessels correlated with lymphatic invasion,29 lymph node and distant metastasis.25, 29, 31 Inhibition of VEGF-A in these models suppressed both new lymphatic vessels and malignant spread, thus supporting a long-suspected causality between tumor lymphangiogenesis and metastasis.
VEGF-A can promote lymphangiogenesis by direct and indirect mechanisms. The indirect mechanism involves VEGF-A mediated recruitment of inflammatory cells that subsequently supply the lymphangiogenic VEGF-C and VEGF-D factors.32 This mechanism appears to play a major role in non-malignant models of cornea injury32, 33 and at sites normally devoid of VEGF-C. In VEGF-C producing tumors, the indirect mechanism appears to be supplemented by a direct action on pre-existing lymphatic vessels through activation of VEGFR-2.25, 29, 35 VEGFR-2, the major signaling receptor for VEGF-A, has been shown to be expressed in LEC36, 37 and to promote LEC proliferation in vitro.29, 35, 38 However, anti-VEGFR-2 antibodies have been less effective than anti-VEGFR-3 antibodies in suppressing tumor lymphangiogenesis,25 suggesting that VEGF-A may regulate the lymphatic endothelium by both direct and indirect mechanisms.
While these reports implicate VEGF-A as a positive regulator of lymphangiogenesis, the molecular mechanisms of its effect on tumor lymphatics and metastatic spread are poorly understood. We, therefore, sought to elucidate the mechanistic role of VEGF-A in tumor lymphangiogenesis and metastasis in an orthotopic model of human breast cancer using the MDA-MB-231 cell line and its luciferase-tagged derivative, 231-Luc+. We herein report that VEGF-A neutralization by anti-VEGF-A 2C3 antibody significantly inhibited both tumor lymphangiogenesis and lymphatic, as well as pulmonary, metastases derived from orthotopically grown breast tumors. This effect was associated with downregulation of the expression of 2 lymphangiogenic mediators: Ang-2 and VEGFR-3. Ang-2 was downregulated in both neoplastic and endothelial cells. The results of these studies implicate VEGF-A in regulating tumor lymphangiogenesis induced by breast tumors, thus suggesting that VEGF-A neutralizing therapies might be used clinically to suppress lymphatic metastasis in breast cancer patients.
Dulbecco's modified Eagle's medium (DMEM), glutamine, sodium pyruvate and nonessential amino acids were obtained from Life Technologies (Grand Island, NY). Purified human VEGF-A165 was purchased from PeproTech (Rocky Hill, NJ). Human Ang-2, mature VEGF-C and VEGF-D proteins were purchased from R & D Systems (Minneapolis, MN). Tetramethylbenzidine (TMB) development kit (Scytek Laboratories, Logan, UT) was used for ELISA. Mouse anesthetic cocktail was a mixture of ketamine (Fort Dodge Animal Health, Fort Dodge, Iowa), xylazine (Phoenix Scientific Inc., St. Joseph, Missouri) and sterile water.
2C3, a mouse IgG2a monoclonal antibody, was raised against recombinant human VEGF-A as described previously.39 C44, an IgG2a mouse anti-colchicine monoclonal antibody from American Type Culture Collection (ATCC, Manassas, VA) served as a negative control for 2C3. MECA32, a pan anti-mouse vascular endothelial cell antibody, was purchased from Developmental Studies Hybridoma Bank (Iowa City, IA). Rat anti-CD31 and rat anti-mouse CD11b antibodies were purchased from PharMingen BD BioScience (San Jose, CA). Rabbit anti-mouse LYVE-1 and hamster anti-mouse podoplanin antibodies were purchased from AngioBio (DelMar, CA). Antibodies against human VEGF-A, anti-mouse VEGFR-2 and VEGFR-3 were purchased from R & D Systems (Minneapolis, MN). Anti-human VEGF-C and anti-human Ang-2 antibodies (cross-reactive with corresponding mouse antigens) were procured from Invitrogen (Carlsbad, CA) and Santa Cruz (Santa Cruz, CA), respectively. HRP-conjugated rabbit anti-rat antibodies were purchased from DacoCytomation (Carpentaria, CA). Rabbit anti-rat and goat anti-rabbit secondary antibodies conjugated with Cy3 and FITC were purchased from Jackson Immunoresearch Labs (West Grove, PA).
Confirmation of VEGF-A specificity of 2C3 antibody by ELISA
Microtiter plates were incubated overnight with 0.25 μg/ml of purified VEGF-A, VEGF-C and VEGF-D diluted in PBS. After washing, plates were blocked with 2% BSA for 2 hr, washed, then incubated for 1 hr with 2C3, anti-VEGF-C or anti-VEGF-D antibodies (1 μg/ml). Bound antibodies were detected by HRP-conjugated secondary antibodies followed by development with TMB substrate (ScyTek, Logan, UT).
Culture of human MDA-MB-231 breast carcinoma cell line and lymphatic endothelial cells
MDA-MB-231 cells were cultured in DMEM containing 10% FBS, 2 mM glutamine, 1 mM sodium pyruvate and 1mM nonessential amino acids at 37°C in 10% CO2. To detach cells, monolayers were incubated with 0.5 mM EDTA dissolved in PBS. Normal rat lymphatic endothelial cells (RLEC) were kindly provided by Dr. Zawieja (Texas A & M University, College Station, TX) and cultured as previously described.38 Lymphatic identity, expression of lymphatic markers (VEGFR-3, podoplanin and LYVE-1) and proliferative responsiveness to human VEGF-A165 and Ang-2 have been established in prior studies.38, 40 All cell lines were sub-cultured twice per week and routinely tested for mycoplasma using an immunodetection kit from Roche Diagnostics GmbH (Penzberg, Germany).
Generation of luciferase-tagged MDA-MB-231 cell line derivative, 231-Luc+
The complete firefly luciferase transcription unit on a Not I restriction fragment consisting of the cDNA, a CMV promoter, and the SV40 splice/polyadenylation signal was sub-cloned from pACCMVluc into pbluescript® vector (Stratagene, La Jolla, CA) to yield pBSCMVluc plasmid. To incorporate the luciferase gene, MDA-MB-231 cells were co-transfected with 1 μg of pBSCMVluc and 200 ng of pKOneo plasmids. Transformants were selected using G-418 (1 mg/ml) and screened for luciferase expression. Positive clones were sub-cloned twice to ensure homogeneity of the population. The luciferase assay of stable clones showed a linear correlation between the number of luciferase-expressing cells (designated as 231-Luc+) and the number of light units. The parental cell line MDA-MB-231 and 231-Luc+ lines displayed identical morphology and grew at identical rates both in culture and upon implantation into the mammary fat pad (MFP) of mice.
Orthotopic model of human breast cancer
MDA-MB-231 or 231-Luc+ cells were implanted into the MFP of 5- to 6-week-old female CB-17 SCID mice (Harlan Sprague-Delaney, Indianapolis, IN), according to published protocol.41 Briefly, mice were anesthetized, then 100 μl cell suspension, containing 4 × 106 cells and 50% Matrigel (Sigma, MO), was injected into the MFP. Every 2–3 days, perpendicular tumor diameters were measured by digital caliper and used to calculate tumor volume according to the formula: volume = Dd2π/6, where D and d equal larger and smaller diameters, respectively. The 231-Luc+ tumors had an identical proliferation rate to their non-transfected counterpart. Animal care was in accordance with institutional guidelines.
Treatment of tumor bearing mice with anti-VEGF-A antibody
Mice were treated according to previously established protocols.42 Briefly, mice bearing 100–150 mm3 MDA-MB-231 tumors were divided into 2 groups of 10–12 per group and treated 3 times weekly by i.p. injection of either 100 μg monoclonal anti-human VEGF-A antibody 2C339 or isotype-matched control antibody C44. Both groups received 15 injections during 5 weeks of treatment. Mice were sacrificed when the average tumor volume in the control group reached 700 mm3. Primary tumors were excised, snap-frozen and used for determination of blood vessel and lymphatic vessel densities. For detection of metastases, mice were implanted with 231-Luc+ tumor cells and treated with 2C3 and C44 antibodies, as described above. To allow maximal development of metastatic foci, mice were sacrificed when the average tumor volume in the control group reached 2000 mm3. Prior studies determined that luciferase activity is primarily associated with tumor proximate axillary lymph nodes and lungs. The anti-metastatic effect of anti-VEGF-A antibody was determined by measuring luciferase activity in tissue extracts derived from 6–10 axillary lymph nodes and lungs.
Detection of metastases by measuring luciferase activity
Lymph nodes and lungs were excised, washed in cold PBS and homogenized in 0.5 ml of cold Cell Culture Lysis Reagent (CCLR) buffer (Promega, Madison, WI) containing protease inhibitor cocktail (Sigma, MO). Cell debris was removed by centrifugation. Protein concentration of cleared lysates was determined by BCA assay (Pierce Biotechnology, Rockford, IL) and adjusted to 10 mg/ml in all samples. About 20 μl of Luciferase Assay Reagent (Promega, Madison, WI) was injected into a tube containing 20 μl of a lysate and luminescence generated in 1 min. The mixture was measured by a single tube luminometer (Berthold, Germany). Values generated by CCLR buffer from tissue homogenates derived from non-tumor bearing mice were considered as background and subtracted from the results. The net results were expressed as relative light units (RLU) normalized per mg of total lysate protein. To assess incidence of metastasis, extracts that had a luminescence signal of 500 light units above background (∼100 light units) were rated positive. This value was chosen because it represented the minimal signal which could be reproducibly detected above background in tissues containing metastases independently confirmed by immunohistochemical detection of human cells.
Immunohistochemical analysis and quantification of blood and lymphatic vessel densities
When the tumors in the control mice reached ∼700 mm3, mice were anesthetized and perfused, and their tumors were excised and snap frozen. Frozen sections were acetone-fixed and incubated for 1 hr at 37°C with 5 μg/ml of antibody against markers of blood endothelium, MECA32,43 or lymphatic endothelium, LYVE-1.44 Primary antibody was detected with appropriate fluorophore-labeled secondary antibody. Double staining was performed using identical reagents applied sequentially to the same section. Quantification of MECA 32-positive structures in 10 0.556 mm2 fields, 2 from each quadrant and 2 from center of the section, determined blood vessel density. Blood vessel density is reported as the mean number of MECA 32-stained structures per mm2 ± SE. Because lymphatic vessels were heterogeneously distributed throughout tumor cross-sections, all vascular structures identified by LYVE-1 antibody were counted. Larger and smaller diameters of each cross-section were measured and used to calculate total section area in mm2. Lymphatic vessel density is reported as the mean number of LYVE-1 positive structures normalized per 100 mm2 ± SE. Sections were quantified from individual mice, then averaged within the experimental group and statistically analyzed.
Immunohistochemical detection and quantification of Ang-2 in tumor and endothelial cells
Goat anti-human/mouse Ang-2 antibody from Santa Cruz (Santa Cruz, CA) was used for immunodetection of Ang-2 in conjunction with MECA32 to identify blood vessels and DAPI to visualize cell nuclei. Anti-Ang-2 and MECA32 antibodies were detected by anti-goat-Cy3 and anti-rat FITC secondary IgG, respectively. An average of Ang-2 positive cells was determined using 5 randomly chosen images within each section taken at magnification of 400×. The percentage of Ang-2 expressing cells was calculated as a fraction of DAPI-stained nuclei within the same field. Data obtained from 8 individual slides of control and 2C3-treated tumors were averaged and expressed as the mean % of Ang-2 positive cells ± SE in each experimental group. The same sections were used to calculate the percentage of Ang-2 expressing blood vessels from total blood vessels identified by MECA32. Ang-2 positive and MECA32-stained vessels were counted on 5–10 randomly chosen images taken at 200× magnification. At least 10 separate fields were chosen to quantify Ang-2 positive vessels in 2C3-treated sections because they had fewer vessels than control tumors. The percentage of Ang-2 positive blood vessels, per 100 blood vessels detected by MECA32, was determined for each section. Data were averaged for all sections from control and treated groups. Results are presented as the mean number of Ang-2 positive vessels ± SE per 100 vessels in each experimental group.
Detection of angiogenic and lymphangiogenic transcripts by endpoint and quantitative RT-PCR
Total RNA was extracted using an RNeasy Mini kit (Qiagen, Valencia, CA), then reverse transcribed using RTG You-Prime Reaction beads (Amersham, Piscataway, NJ) and random hexamer primers (Invitrogen, Carlsbad, CA). Primer sequences, which were designed using GeneRunner software, are available upon request. All primers were tested using human and mouse Universal cDNA (SuperArray, Frederick, MD) as positive controls to determine the sensitivity and species specificity of mRNA detection. Standard endpoint RT-PCR analysis was performed, then visualized and analyzed using a FluroChem5500 (AlphaInnotech, San Leandro, CA) imaging system and software. For qRT-PCR, mRNA expression was quantified using an Applied Biosystem 7500 Real-Time PCR machine. The reaction mix consisted of 12.5 μl of SYBR PCR Master Mix (Applied Biosystems, Foster City, CA), 900 nM of each primer, 1 μl cDNA template, and nuclease free-water in a final volume of 25 μl. A typical reaction consisted of an initial denaturation step at 95°C for 5 min followed by 40 cycles of denaturation at 95°C for 15 s and annealing, extension, and reading at 60°C for 1 min. A final dissociation curve was calculated by linear reaction heating from 60 to 90°C. Data were normalized to beta actin and compared to Universal RNA. Relative mRNA expression was determined using the ΔΔCt method.45
Statistical analysis was performed using GraphPad InStat (GraphPad Inc., San Diego, CA). Results are expressed as mean ± SE. Statistical significance of differences in vascular densities between control and 2C3 antibody treated groups was assessed by unpaired Student's t-test and defined as p < 0.05. The differences in incidence of lymph nodes and lung metastases were analyzed by the Fisher's exact test. The statistical differences of mean tumor volume among experimental groups were analyzed by the nonparametric Kruskal-Wallis test.
Anti-VEGF-A antibody inhibits spontaneous lymph node and lung metastases derived from orthotopic 231-Luc+ tumors
We have previously shown in an orthotopic mouse model that anti-VEGF-A 2C3 antibody inhibited angiogenesis and growth of MDA-MB-231 breast carcinoma tumors.42 The aims of this study were to determine the effect of 2C3 antibody on lymphangiogenesis and metastasis derived from orthotopic MDA-MB-231 tumors. To facilitate detection of metastatic cells, we generated a luciferase-tagged sub-line of MDA-MB-231, designated 231-Luc+. Preliminary study showed high levels of luciferase activity in tumor proximate lymph nodes and lungs, 30 and 50 days, respectively, after tumor implantation (data not shown). Other normal organs including liver, kidney, heart and brain typically did not contain luciferase activity, and therefore, were not included in the metastatic analysis of the present study.
To assess the effect of inhibition of tumor-derived VEGF-A on lymphatic and pulmonary metastasis, mice bearing 100–150 mm3 231-Luc+ tumors were randomized into two groups and treated as described under Methods. Mice were sacrificed when the mean tumor volume in the control group reached 1835 ± 358 mm3. The mean tumor volume in the 2C3-treated group was 642 ± 128 mm3, representing a 65% reduction in primary tumor size compared to mice treated with the control antibody (Table I). Metastatic incidence in the control group (n = 11) was 81% and 72% for lymph node and lungs, respectively. The average metastatic burden in these tissues was 312.7 × 103 ± 63.4 and 78.3 ×103 ± 14.5 RLU/mg protein, respectively (Table I). In contrast, the incidence of lymphatic and pulmonary metastases was decreased by 3.2- and 4.5-fold, respectively (Table I) in the 2C3-treated group (n = 12). The difference in the metastatic incidence between control and 2C3 antibody groups was statistically significant (p = 0.001, by Fisher's exact test). 2C3 therapy also reduced the average metastatic burden in lymph nodes and lungs by 2.5- and 2.4-fold, respectively (Table I). We concluded that inhibition of tumor-derived VEGF-A activity with 2C3 antibody significantly decreased both incidence and total metastatic burden of spontaneously formed lymph node and pulmonary metastases derived from orthotopic 231-Luc+ tumors.
Table I. Incidence and Average Metastatic Burden in Lymph Nodes and Lungs in 231-LUC+ Bearing Mice Treated with Control or Anti-VEGF-A 2C3 Antibodies
CB17 SCID mice bearing orthotopic 231-Luc+ tumors were treated with control or anti-VEGF-A antibody 2C3 (100 μg i.p., 3 times a week). Number of mice in the control and 2C3 treated groups were 11 and 12, respectively. The treatment started when tumor volume was 100–150 mm3 and continued for 5 weeks. Mice were sacrificed when tumors in control group reached 1,800–2,000 mm3 in volume (50–54 days after tumor implantation). Lymph nodes and lungs were harvested, and metastatic burden and incidence were determined by measuring the luciferase activity in tissue extracts.
Lymph nodes (6–10) and 1 lung lobe from each mouse were homogenized in the presence of 0.5 ml of luciferase-extracting buffer and protease inhibitors. The luciferase activity was measured in 20 μl duplicates from each extract. After subtracting background, the results were expressed as relative light units (RLU) normalized per mg of total lysate protein. Mean values ± SE per group are presented.
Expression of blood and lymphatic markers on vessels in MDA-MB-231 orthotopic tumors
Because 2C3 had a significant inhibitory effect on lymphatic metastasis, we hypothesized that VEGF-A may affect lymphangiogenesis induced by breast carcinoma cells. To better understand the mechanism underlying this observation, we wished to examine the effect of this therapy on blood and lymphatic vessel densities as well as on the expression of VEGFR-3. To this end, we first validated the expression of commonly used markers of blood and lymphatic vessels (CD31, MECA32, LYVE-1 and VEGFR-3) specifically in the MDA-MB-231 orthotopic model. This was necessary because some studies reported expression of a blood vascular marker CD31 on lymphatic vessels46 and expression of VEGFR-3 on a subset of blood vessels.47, 48
The results presented in Figure 1 show a mutually exclusive pattern of CD31 and LYVE-1 markers on vessels of MDA-MB-231 tumors (n = 24) (Figs. 1a–1c). An occasional partial overlap was detected in one of ∼50 of LYVE-1 positive structures (Fig. 1c, yellow arrow). As shown on this image, when such overlap occurred, it was typically restricted to a small portion of the vessel. Additionally, we examined another marker specific to mouse blood vasculature, an antibody against MECA32.43 As shown on Figures 1d–1f, none of the LYVE-1 positive vessels was stained with MECA32 antibody. Anti-CD31 and anti-MECA32 antibodies produced identical results with regard to signal intensity, pattern and number of vascular structures detected on all sections of MDA-MB-231 tumors, indicating similar usefulness of both markers for detection of mouse blood vessels in this model.
VEGFR-3 appears to be up-regulated on a subset of human tumor blood vessels, in addition to its main expression on lymphatic endothelium.47, 48 We, therefore, evaluated expression of VEGFR-3 in conjunction with LYVE-1, CD31 and MECA32 markers. Typically, VEGFR-3 and LYVE-1 co-localized, as shown in Figure 1, panel g–i. In contrast, VEGFR-3 co-staining with either of the blood vascular markers CD31 or MECA32 invariably yielded mutually exclusive patterns as shown in Figure 1, panel j–l. We concluded that in our orthotopic xenograft MDA-MB-231 model, CD31 and MECA32 are restricted to blood vessels, whereas LYVE-1 and VEGFR-3 are co-expressed on lymphatic vessels.
Neutralization of VEGF-A inhibits both blood and lymphatic vessels in MDA-MB-231 tumors
VEGF-A has been shown to promote lymphangiogenesis during wound healing,30 inflammation32 and tumorigenesis in models of mouse skin29 and human breast cancer.25 We tested the hypothesis that neutralization of VEGF-A would also inhibit lymphangiogenesis in orthotopic MDA-MB-231 and 231-Luc+ tumor models. Both tumor models produced identical results in support of the premise. To analyze the effect of anti-VEGF-A treatment, frozen tumor sections were double-stained with anti-mouse CD31 and LYVE-1 antibodies detecting blood49 and lymphatic50 markers, respectively. Vessel identity was confirmed on all slides by co-staining with antibodies MECA32 and podoplanin, representing additional markers of blood and lymphatic vessels, respectively.43, 51 Merged tumor images co-stained with blood vessel and lymphatic vessel markers showed non-overlapping patterns of staining (Figs. 1a–1f and 2a, Control), demonstrating that orthotopically grown human MDA-MB-231 tumors induce intratumoral formation of both blood and lymphatic vasculature.
Intratumoral lymphatic vessels were found in all tumors from control animals (n = 33). Lymphatic vessels were mainly located in the tumor margin (100–250 μm from the tumor-MFP boundary), although some vessels were detected as deep as 400 μm away from the tumor boundary. No difference in lymphatic vessel density (LVD) and distribution of intratumoral lymphatic vessels was detected in tumors of different sizes (250–1550 mm3) derived from control mice. LVD among tumors from individual mice varied from 26 to 46 LYVE-1-positive structures, with a mean of 35 ± 9 per 100 mm2. These variations were tumor size-independent. Compared with tumors from control mice, 2C3 treatment significantly reduced the number of lymphatic vessels while, in parallel, reducing blood vessel number (Figs. 2a and 2b). The results were replicated in 4 independent experiments. Of the 33 control tumors, 76% contained more than 10 lymphatic vessels per section (Fig. 2c). In comparison, only 14% of the 2C3-treated tumors (n = 36) had more than 10 lymphatic vessels per section (Fig. 2c). The data strongly support consideration of VEGF-A as an important factor for induction and/or maintenance of tumor lymphatic vasculature.
To exclude cross-reaction with VEGF-C and VEGF-D proteins, the antigen specificity of 2C3 antibody was assessed using ELISA. Human VEGF-A, VEGF-C, or VEGF-D were immobilized in 96-well plates and probed with 2C3, anti-VEGF-C, or anti-VEGF-D antibodies. 2C3 specifically bound human VEGF-A, but not VEGF-C or VEGF-D (Fig. 2d). Control anti-VEGF-C and VEGF-D antibodies detected corresponding lymphangiogenic factors. The results demonstrated specificity of the 2C3 antibody to VEGF-A and excluded cross-reactivity with VEGF-C or VEGF-D as a mechanism for reduction in LVD of the treated tumors. We concluded that suppression of tumor lymphangiogenesis by 2C3 treatment resulted specifically from neutralization of VEGF-A activity.
Anti-VEGF-A treatment reduces macrophage infiltration of MDA-MB-231 tumors
VEGF-A has been shown to induce lymphangiogenesis in the injured cornea by recruiting macrophages that supplied lymphangiogenic factors, VEGF-C and VEGF-D.32 To examine whether reduced LVD following anti-VEGF-A treatment was associated with reduced macrophage infiltration, we performed immunohistochemical staining of control and 2C3-treated tumors using a macrophage-specific marker, CD11b.52 Infiltrating macrophages were detected in both control and 2C3-treated tumors (Figs. 3a and 3b). The density of macrophages was particularly high in necrotic regions in both control and 2C3-treated tumors although significant numbers of CD11b-positive cells (60–100/field at magnification of 400×) were homogenously distributed throughout all tumor regions.
Quantitative analysis of control (n = 12) and 2C3-treated (n = 8) tumors revealed the mean number of infiltrating macrophages was reduced by 32% by 2C3 therapy (Fig. 3c). However, co-staining of tumor sections for CD11b and LYVE-1 revealed no significant spatial association between regions of high LVD located primarily on tumor margins and macrophage infiltrates homogenously detected throughout the tumor mass. Because of a relatively modest decrease of macrophage infiltrates in 2C3-treated tumors and lack of spatial colocalization with LYVE-1 vessels, we concluded that inhibition of macrophage recruitment alone was unlikely to be the predominant mechanism for LVD reduction following anti-VEGF-A treatment.
Anti-VEGF-A treatment does not affect VEGF-C expression in MDA-MB-231 tumors
Because VEGF-A signaling has previously been shown to affect MDA-MB-231 tumor cells in an autocrine manner,53, 54 we explored the hypothesis that VEGF-A may regulate lymphangiogenesis by modulating tumor-derived lymphangiogenic factors such as VEGF-C. RT-PCR detected mRNA for VEGF-C, but not for VEGF-D, in control MDA-MB-231 tumors (Fig. 4a, Ctrl). Both Western blot and immunohistochemical analyses confirmed robust and ubiquitous expression of VEGF-C protein in control tumors (Fig. 4). A comparative analysis of tumors from control and anti-VEGF-A treated mice revealed no significant changes in VEGF-C expression as detected by RT-PCR (Figs. 4a and 4b). To confirm this finding, protein lysates from snap-frozen tumors were subject to Western blot analysis using anti-VEGF-C antibody. The main detectable band by this antibody was the mature, fully processed, 15 kDa form of VEGF-C, followed by a 29/31 kDa doublet. A 29 kDa band was detected with moderate sensitivity whereas the 31 kDa and 43 kDa bands were barely visible. This profile is in agreement with previous publications demonstrating VEGF-C processing,55 and was identical in all tumor samples regardless of anti-VEGF-A antibody treatment (Fig. 4c). Likewise, no significant difference in VEGF-C protein expression was detected immunohistochemically between anti-VEGF-A treated tumors and control tumors (Figs. 4d and 4e). The findings suggest that inhibition of tumor-derived VEGF-A altered neither the total intratumoral concentration of VEGF-C nor VEGF-C proteolytic processing, and also suggests the 2C3 treatment reduces LVD in a VEGF-C independent manner.
Anti-VEGF-A treatment suppresses expression of VEGFR-3 on tumor lymphatic endothelium in vivo
Because VEGF-C expression was not affected, we examined an alternative hypothesis that reduced LVD, observed after anti-VEGF-A treatment, results from decreased expression of VEGFR-3 receptor on lymphatic endothelium. To examine this hypothesis, sections from control and 2C3-treated mice were double stained with rabbit anti-LYVE-1 and goat anti-VEGFR-3 antibodies. As shown on Figure 1, neither LYVE-1 nor VEGFR-3 was observed on MECA32 or CD31-positive vessels. In normal MFP and control tumors, numerous lymphatic vessels were detected, all of which exhibited coincident expression of LYVE-1 and VEGFR-3 (Figs. 5a–5c and 5d–5f). By contrast, the lymphatic vasculature in anti-VEGF-A treated tumors differed both quantitatively and qualitatively from that of normal MFP or control tumors. Quantitatively, the overall density was reduced by 80% (Figs. 2a and 2b). Qualitatively, the lymphatic vessels in 2C3-treated tumors displayed reduced levels of VEGFR-3 on LYVE-1 positive vessels (Figs. 5g–5i). The reduced or absent VEGFR-3 staining on lymphatic endothelium was observed in anti-VEGF-A treated tumors, but not C44 control antibody treated tumors, nor in normal MFP of untreated mice. Thus, the decrease in VEGFR-3 expression on LYVE-1 positive vessels appears to be a direct consequence of VEGF-A neutralization by 2C3 antibody.
These findings show that VEGF-A blockade inhibits expression of VEGFR-3 on newly formed intratumoral lymphatic vessels, but has no deleterious effect on established lymphatic vessels in the nearby normal MFP. This suggests that reduced expression of this receptor on tumor lymphatic endothelium might be partly responsible for suppressing lymphangiogenesis following anti-VEGF-A therapy.
Anti-VEGF-A treatment inhibits expression of angiopoietin-2 (Ang-2) in both tumor cells and tumor vessels
We also considered Ang-2 as a possible mediator of the VEGF-A effect on tumor lymphatic vessels. This hypothesis is based on reported data showing up-regulation of Ang-2 by VEGF-A in cultured endothelial cells56 and a role for Ang-2 and a related protein, Ang-1, in the formation of lymphatic vessels.57, 58, 59 Ang-2 has also been implicated in promoting lymphatic invasion and metastasis in human breast cancer.60 Based on these reports, we hypothesized that anti-VEGF-A treatment could reduce intratumoral Ang-2 expression, subsequently causing a negative effect on generation of new lymphatic vessels.
To test this hypothesis, we compared the expression of Ang-2 in control and anti-VEGF-A antibody treated tumors. Ang-2 transcripts were reduced by 3-fold in 2C3-treated tumors as compared with control samples (p < 0.005) (Figs. 6a and 6b). In contrast, Ang-1 was expressed at comparable levels in both control and 2C3-treated tumors (data not shown). In line with previous reports,61 a two-fold increase in Ang-2 expression was reproducibly observed in RLEC,38 after VEGF-A treatment of 50 ng per ml for 24–48 hr (Fig. 6c). VEGF-A had no effect on Ang-1 or Tie-2 expression (data not shown), suggesting a selective role for increased Ang-2 in VEGF-A mediated lymphangiogenesis.
The sharp decline in Ang-2 mRNA correlated with immunodetection of Ang-2 protein (Figs. 6d and 6e). Tumor sections were double stained with goat anti-mouse/human Ang-2 and rat anti-MECA32 antibodies. In control tumors (n = 8), Ang-2 protein was robustly expressed in the majority of tumor cells (56% ± 6% of DAPI-visualized nuclei) and in some vessels (15.3% ± 3.3% of vessels identified by MECA32 antibody). In contrast to control tumors, all 2C3-treated samples (n = 7) showed drastic reduction of Ang-2 positive structures (Fig. 6d). Both Ang-2 positive cells and vascular structures decreased by 50 and 66%, respectively (p < 0.01). We concluded that VEGF-A can up-regulate Ang-2 expression whereas anti-VEGF-A treatment inhibits the expression of Ang-2 in both tumor and endothelial cells, suggesting that Ang-2 could mediate the anti-lymphangiogenic associated with 2C3 treatment.
VEGF-A and Ang-2 regulate expression of VEGFR-2 and VEGFR-3 in cultured LEC and induce LEC proliferation
Because the reduction in Ang-2 expression correlates with diminished lymphatic vessels in the anti-VEGF-A-treated tumors, we examined the effects of VEGF-A and Ang-2 on lymphatic endothelial cells in vitro. Rat LEC or human H-LLY cells were seeded in DMEM containing 1% FBS and treated with medium alone, VEGF-A165 or Ang-2 (50 ng/ml of either factor) for 24–48 hr. Total RNA was extracted and used for qRT-PCR analysis of VEGFR-2 and VEGFR-3 of rat or human origin. Figure 7a shows that VEGF-A treated RLEC increased VEGFR-2 and VEGFR-3 expression by 73% and 37%, respectively, as compared with medium control (p < 0.02 and 0.03, respectively). The mean values are derived from 6 independent experiments performed in duplicate. Under the same conditions, Ang-2 increased VEGFR-3 expression by 44% (p < 0.03) but had no significant effect on VEGFR-2 (Fig. 7a). A similar pattern of responses to Ang-2 and VEGF-A165 was detected in H-LLY cells (data not shown). These results suggest that one mechanism by which VEGF-A and Ang-2 can enhance lymphangiogenesis is by primarily increasing expression of VEGFR-3 receptor, thereby increasing LEC sensitivity to lymphangiogenic stimuli.
VEGF-A has also been shown to increase proliferation of human LEC directly, via activation of VEGFR-2.37 We report here that Ang-2 and VEGF-A induce proliferation of RLEC in a dose-dependent manner (Fig. 7b). As expected, soluble Tie-2 protein (2 μg/ml) completely neutralized Ang-2 mediated effect. Interestingly, sTie2 also reduced VEGF-A induced proliferation by nearly 50% (Fig. 7b). The data suggest that VEGF-A may regulate lymphangiogenesis, in part, by increasing endogenous Ang-2 expression, which subsequently activates Tie-2 receptor and induces LEC proliferation in an autocrine manner.
The salient findings of this study are that neutralization of tumor-derived VEGF-A in orthotopic MDA-MB-231 breast tumors significantly inhibited lymphatic metastasis and decreased intratumoral LVD. Our findings are in line with recently reported data demonstrating a pro-lymphangiogenic role of VEGF-A during wound healing,30 corneal injury32 and tumorigenesis of skin carcinoma.29 Blocking VEGF-A activation of VEGFR-2 has been shown to suppress tumor lymphangiogenesis and metastasis in breast carcinoma MDA-MB-435 model.25 The novel findings of the present study are that anti-VEGF-A treatment is associated with significant down-regulation of VEGFR-3 protein on residual tumor lymphatic vasculature as well as Ang-2 expression in both tumor cells and tumor endothelial cells. In vitro, Ang-2 activated LEC in a dose-dependent manner and VEGF-A induced LEC proliferation was partly regulated by Tie-2 receptor signaling. These data suggest that reductions in VEGFR-3 and Ang-2 expression represent causative mechanisms underlying the anti-lymphangiogenic effect of anti-VEGF-A therapy.
Treatment with anti-VEGF-A antibody significantly decreased the incidence and the metastatic burden of spontaneous metastases derived from the orthotopically implanted 231-Luc+ tumors (Table I). In the 2C3-treated group, a decrease in the incidence of mice with metastases in lymph nodes and lungs was 70 and 78%, respectively, and the average metastatic burden in lungs was decreased by 91%, as compared with control animals. The results demonstrate a strong correlation between a decrease of intratumoral lymphatic density and a reduction in positive lymph nodes, corroborating previously published data showing positive correlation between over-expression of VEGF-A and lymphatic metastasis in skin tumor and breast carcinoma models.25, 29 Collectively, study data obtained in 3 independent tumor models support the notion that depletion of VEGF-A negatively affects both tumor lymphatics and lymph node metastasis. This study provides additional evidence that VEGF-A plays an important role in the formation of tumor lymphatic vessels and that tumor lymphangiogenesis is a major cause of lymphatic metastasis.
Anti-VEGF-A treatment can reduce metastasis by multiple mechanisms, including inhibition of VEGF-A-induced angiogenesis in primary tumors,62 inhibition of tumor lymphangiogenesis,25 prevention of angiogenic switch in micrometastases63 and blocking macrophage recruitment.64, 65 Our data (Table I and Fig. 2) support the notion that the anti-lymphangiogenic effect of anti-VEGF-A antibody therapy may play an important role in inhibiting metastasis. This conclusion is based on observations in the MDA-MB-231 model demonstrating significant correlation between reduction in tumor lymphatic vessels and positive lymph nodes in the 2C3-treated animals. This is consistent with findings in other tumor models demonstrating that forced expression of VEGF-A correlated with increased lymphangiogenesis and lymphatic and distant metastasis.29, 31 These findings, however, do not exclude the possibility that VEGF-A promotes metastasis by other mechanisms occurring concurrently with enhancement of lymphangiogenesis. Nevertheless, these data suggest that an anti-lymphangiogenic component of anti-VEGF-A therapies might be an important contributor for suppressing both lymphatic and distant metastasis.
The anti-lymphangiogenic effect of anti-VEGF-A 2C3 antibody may occur through several mechanisms. It was previously proposed that VEGF-A induces lymphangiogenesis through macrophage recruitment that supplied VEGF-C and VEGF-D factors.32 Quantification of macrophage infiltrates in MDA-MB-231 tumors showed a 32% decrease in 2C3-treated tumors (Fig. 3). This decrease, however, neither correlated with distribution of LYVE-1 positive vessels nor with VEGF-C expression, which was strong and ubiquitous in both control and 2C3-treated groups (Fig. 4d). Lack of regulation of VEGF-C expression by VEGF-A has also been noted in T241 fibrosarcoma tumor model that over-expressed VEGF-A by 285-fold and displayed higher macrophage infiltration than control tumors.31 We concluded that macrophages may play a more important role in the environments lacking intrinsic VEGF-C (e.g., the injured cornea model32) than in tumors that robustly express VEGF-C protein. These findings suggest that inhibition of lymphangiogenesis by anti-VEGF-A antibodies may occur through modulation of lymphatic sensitivity to VEGF-C rather than by regulating the expression of this factor.
We tested this hypothesis by determining whether anti-VEGF-A treatment affects the expression of VEGFR-3 in lymphatic endothelium. We found that VEGFR-3 receptor was markedly downregulated on lymphatic vessels following 2C3 treatment (Fig. 5). VEGFR-2 receptor was also detected on tumor lymphatic vessels, but the reduction in VEGFR-2 expression following 2C3 treatment was far less remarkable than that of VEGFR-3 (data not shown). The 2C3 antibody treatment has previously been shown to inhibit VEGFR-2 expression on blood vessels in breast42 and pancreatic66 tumor models. A partial suppression of lymphangiogenesis by blocking VEGFR-2 receptor has also been reported, although the treatment was less efficacious than blocking VEGFR-3 receptor in the same model.25 In line with this report,25 we have found that VEGF-A dependent inhibition of lymphangiogenesis was associated with the decrease of predominantly VEGFR-3 receptor on residual lymphatic vessels while the effect on VEGFR-2 was less significant.
Because no molecular connection between VEGF-A signaling and VEGFR-3 expression has been reported, we examined possible indirect mechanisms whereby anti-VEGF-A treatment may affect VEGFR-3 expression. VEGF-A increases expression of Ang-2,56 a ligand for Tie-2 receptor expressed on lymphatic vessels. Ang-2 and related agonist, Ang-1, have been implicated in lymphangiogenesis.59 We, therefore, examined the effect of 2C3 treatment on the presence of Ang-1 and Ang-2 on MDA-MB-231 tumors and found only Ang-2 to be significantly reduced, on both mRNA and protein levels, following 2C3 treatment (Fig. 6). Both tumor cells and tumor endothelial cells demonstrated marked decrease in Ang-2 protein suggesting that depletion of tumor-derived VEGF-A affects both neoplastic and stroma compartments. Consistent with prior observations,61 we measured a 2-fold increase in Ang-2 expression following VEGF-A treatment of cultured LEC (Fig. 6c). Additionally, we report herein that both VEGF-A and Ang-2 increased proliferation of LEC in vitro in a dose-dependent manner and that sTie-2 inhibited the proliferative effects of either factor (Fig. 7). This suggests that the VEGF-A-mediated increase in Ang-2 production in tumor and tumor endothelial cells is responsible for increasing LEC division. While the molecular link between Ang-2 dependent activation and VEGFR-3 expression has yet to be established, the observations of Gale et al.57 that Tie-2 dependent signaling regulates VEGFR-3 expression in vivo suggests the plausibility of such a mechanism. Following this rationale, anti-VEGF-A treatment might reduce LVD indirectly by reducing Ang-2 dependent LEC proliferation and, possibly VEGFR-3 expression in lymphatic endothelium.
VEGF-A expression has been widely correlated with metastasis in breast cancer and other solid tumors that preferentially disseminate through the lymphatic route. This study demonstrates that lymphangiogenesis is promoted by VEGF-A and that the antibody therapy antagonizing this factor inhibits both lymphangiogenesis and lymphatic metastasis. It is, therefore, conceivable that anti-VEGF-A therapies, several of which are currently approved for clinical use, may confer additional benefits to cancer patients by reducing lymphatic metastasis and subsequent mortality.