Skip to main content
NCBI home page
Annals of Surgery logoLink to Annals of Surgery
. 2006 Nov;244(5):781–791. doi: 10.1097/01.sla.0000231723.81218.72

Tumor Necrosis Factor-alpha Damages Tumor Blood Vessel Integrity by Targeting VE-Cadherin

Chandrakala Menon *, Antoinette Ghartey *, Robert Canter *, Michael Feldman †, Douglas L Fraker *
PMCID: PMC1856603  PMID: 17060772

Abstract

Background:

Isolated limb perfusion using high-dose human tumor necrosis factor-alpha with melphalan is effective therapy for bulky extremity in-transit melanoma and sarcoma.

Objective:

While it is widely accepted that melphalan is a DNA alkylating agent, the mechanism of selective antitumor effect of tumor necrosis factor-alpha is unclear.

Methods and Results:

Electron microscopic analyses of human melanoma biopsies, pre- and post-melphalan perfusion, showed that the addition of tumor necrosis factor-alpha caused gapping between endothelial cells by 3 to 6 hours post-treatment followed by vascular erythrostasis in treated tumors. In human melanoma xenografts raised in mice, tumor necrosis factor-alpha selectively increased tumor vascular permeability by 3 hours and decreased tumor blood flow by 6 hours post-treatment relative to treated normal tissue. In an in vitro tumor endothelial cell model, tumor necrosis factor-alpha caused vascular endothelial adherens junction protein, VE-cadherin, to relocalize within the cell membrane away from cell-cell junctions leading to gapping between endothelial cells by 3 to 6 hours post-treatment. Phosphotyrosinylation was a prerequisite for movement of VE-cadherin away from endothelial cell junctions and for gapping between endothelial cells. Clinical isolated limb perfusion tumor specimens, at 3 hours postperfusion, showed a discontinuous and irregular pattern of VE-cadherin expression at endothelial cell junctions when compared with normal (skin) or pretreatment tumor tissue.

Conclusions:

Together, the data suggest that tumor necrosis factor-alpha selectively damages the integrity of tumor vasculature by disrupting VE-cadherin complexes at vascular endothelial cell junctions leading to gapping between endothelial cells, causing increased vascular leak and erythrostasis in tumors. VE-cadherin appears to be a potentially good target for selective antitumor therapy.

Isolated limb perfusion with high-dose tumor necrosis factor-alpha (TNF) plus melphalan is effective therapy for bulky extremity in-transit melanoma and sarcoma. TNF selectively damages the integrity of tumor vasculature by phosphorylating vascular endothelial-cadherin (VE-cadherin), resulting in gapping between endothelial cells leading to increased vascular leak and erythrostasis.

The ability of current antineoplastic agents to cause sustained regression of established human tumors is limited.1,2 One strategy to improve response rates is to use isolation perfusion procedures to improve the therapeutic index of antineoplastic drugs.3 Isolated limb perfusion (ILP) is a surgical procedure in which high-dose chemotherapeutics are recirculated within an extremity primarily to treat in-transit melanoma as well as unresectable soft tissue sarcoma.4 ILP with melphalan alone results in a complete response rate of 54% and an overall response rate of 79% in established melanoma.5 The addition ofhigh-dose human recombinant tumor necrosis factor alpha (TNF) to melphalan ILP augments complete response rates to 75% to 90%.6–8

The ability to cause complete and sustained regressions of established bulky solid human tumors with a single 60- to 90-minute regional treatment with TNF plus melphalan is unique and highlights the importance of elucidating the mechanism of this response.2 Isolation perfusion exposes all of the normal tissue of the extremity to high-dose antineoplastic agents, yet there is a selective destruction only within the tumor. It is well established that melphalan is a DNA alkylating agent and is effective because it causes DNA damage in the dividing cells of the tumor.1 There is evidence that TNF targets the tumor vasculature.10–13 This antivascular effect explains the wide range of histologies that respond to TNF plus melphalan ILP. It is also the primary reason that TNF has the most therapeutic benefit against larger tumors14 as it is not dependent on direct cytotoxicity to the malignant cells.15 The disadvantage of this therapy is that it treats only the extremity and delivers no treatment to the rest of the body in this potentially systemic disease. However, if one could understand the mechanism of this response at a cellular and molecular level, one could hypothesize that alternative strategies that reproduce this mechanism of response with systemic therapy may be developed.

Earlier studies have used purely in vitro endothelial cell systems to demonstrate the role of TNF in causing endothelial cell monolayer disruption leading to permeability mediated by the endothelial cell adherence junction protein, VE-cadherin.16–18 The current study is the first of its kind that has used clinical specimens from before and after TNF plus melphalan ILP treatments besides utilizing preclinical in vivo models, and in vitro tumor endothelial cell models to demonstrate that TNF causes tumor selective damage to the vascular integrity mediated by VE-cadherin, which appears to be a molecular target in this effective antitumor therapy.

MATERIALS AND METHODS

Cell Lines and Culture Conditions

Human dermal microvascular endothelial cells (HDMVEC) were isolated from human skin as described previously19 and were grown in EGM-MV medium (BioWhittaker, MD) in sterile 60-mm Permanox culture dishes (Nalgene Nunc International, Rochester, NY). NIH1286 human melanoma cells (gift from Dr. Steven Rosenberg, NCI, Bethesda, MD) were cultured in RPMI medium supplemented with 10% FCS, 2 mmol/L l-glutamine, 100 U/mL penicillin, and 100 μg/mL streptomycin (LTI/GIBCO BRL, Gaithersburg, MD).

Mice, Tumors, and TNF

Six- to 8-week-old athymic nude mice (Harlan, Indianapolis, IN) were injected s.c. in their right flank with 5 × 106 NIH1286 human melanoma cells and tumors were allowed to grow to 8 to 10 mm. Human recombinant TNF-alpha (gift from Knoll Pharmaceuticals, Whippany, NJ) was determined to be endotoxin-free by the standard Limulus assay. Activity was determined by L929 cytotoxicity assay.

Electron Microscopy

Preoperative and postoperative human tumor biopsy tissue was prepared for electron microscopy by fixing in 2% gluteraldehyde in 0.1 mol/L sodium cacodylate buffer (pH 7.4). Fixed tissues were embedded, sectioned, and stained by standard techniques at the Electron Microscopy Core Facility, National Institutes of Health, Bethesda, MD.

PECAM/VE-Cadherin Immunostaining of Human Tumor Cryosections

Cryopreserved tumor sections were fixed in cold acetone on ice for 5 minutes, blocked with phosphate buffered saline (PBS) containing 4% BSA and 5% horse serum and incubated overnight with either a mouse antihuman PECAM antibody (gift from Dr. Steven Albelda, University of Pennsylvania, Philadelphia, PA) or mouse antihuman monoclonal VE-cadherin antibody (Becton Dickinson, Franklin Lakes, NJ). The sections were then incubated sequentially with biotinylated horse antimouse secondary antibody (Vector Labs, Burlingame, CA) and avidin-biotin-peroxidase complex (ABC Elite kit, Vector Labs). The color reaction was developed with the Vector VIP Peroxidase Substrate kit (Vector Labs) and the sections were mounted in glycerol. Slides were photographed at 200× magnification.

Vascular Permeability

Four mice were injected via tail vein with 15 μg of TNF. Four control animals received a comparable volume of 0.5% BSA. Fifteen minutes prior to death, 10 mg of Evans Blue-albumin was injected via tail vein in all mice. Four mice (2 TNF, 2 BSA) were killed at 30 minutes and 4 mice (2 TNF, 2 BSA) at 3 hours and tumor and muscle were harvested, washed in PBS, blotted-dry, and weighed. Each specimen was then placed separately in 3 mL formamide (Sigma, St. Louis, MO), covered, and incubated for 48 hours at 60°C. The resulting solutions were syringe-filtered, and optical density was measured at 620 nm. The results were statistically analyzed using the Student t test.

Blood Flow Measurements

Ten nude mice carrying subcutaneous human melanoma tumors were placed with the help of a mouse holder in front of an experimental high resolution gamma camera linked to a data analyzer; 10 μL of Xe solution (25 μCi) (DuPont Radiopharmaceuticals, North Billerica, MA) was injected percutaneously into the tumor (5 mice) or native subcutaneous tissue (5 mice). Clearance time of Xe was measured at various time-points before and after treatment with i.v. TNF. Gamma counts at 10 seconds intervals over 10 minutes periods were analyzed by a computer program (Nuclear Mac, Scientific Imaging, Littleton, CO) and recorded as t1/2 values. Statistical analysis for differences in mean blood flow at specific time-points in tumor and skin pretreatment and post-treatment was carried out using the Student t test.

Endothelial Cell Exposure to Tumor-Specific ILP (T) and Nontumor ILP Conditions (NT)

To simulate T and NT in an in vitro model, HDMVEC were exposed to either NIH1286 human melanoma conditioned medium (MCM), 2 μg/mL TNF, and 1% O2 [T] or to normal medium (NM), 1 μg/mL TNF and normoxia [NT] in culture. The required level of hypoxia for endothelial cells exposed to T was achieved by enclosing the Petri dishes in specialized calibrated air-tight hypoxia chambers (Nupro Co.) connected to a N2 tank and vacuum pump through a manifold. After equilibration to achieve 1% oxygen, the chambers were incubated at 37°C. HDMVEC were exposed to the above conditions, individually and in combination to determine the relative contribution of each to changes in VE-cadherin expression at endothelial junctions.

Immunofluorescent Staining for VE-Cadherin in HDMVEC and in Clinical Tissue Specimens

HDMVEC were fixed in 3.7% formaldehyde-PBS at RT for 10 minutes, permeabilized with 0.1% Triton X-100-PBS for 5 minutes, rinsed in PBS to remove any remaining detergent, and incubated for 30 minutes in blocking buffer consisting of 1% BSA in PBS. This was followed by incubation with an antihuman VE-cadherin antibody (Becton Dickinson) in blocking buffer for 1 to 2 hours and with FITC conjugated secondary antibody in blocking buffer for 1 to 2 hours after washing with PBS to remove unbound VE-cadherin antibody. The cells were simultaneously fluorescently labeled with phalloidin rhodamine (Molecular Probes, Eugene, OR) and mounted using PermaFluor aqueous mountant (Shandon). The cells were viewed under a laser scanning confocal microscope with a 40× oil immersion lens and digital images were analyzed using Confocal Assistant software. Clinical tissue specimens were obtained from Dr. Fraker’s isolated limb perfusion patients that were treated with TNF plus melphalan after informed consent as per the National Cancer Institute’s IRB approved protocols. Such specimens were first immunofluorescently stained for VE-cadherin using the antihuman VE-cadherin antibody and secondarily stained with an antimouse IgG Texas red antibody. The sections were then stained with an anti-alpha smooth muscle actin antibody that was FITC conjugated (Sigma). The tissue sections were mounted using Permaflour and viewed under a 100× oil immersion lens to give a total magnification of 1000×.

Phosphotyrosinylation Assay

HDMVEC exposed to the various experimental conditions for 30 minutes, or 3 hours, were rinsed in PBS containing 1 mmol/L vanadate fixed in 4% paraformaldehyde for 20 minutes followed by absolute methanol (10 minutes, −20°C), washed, and incubated for 1 hour with FITC-conjugated antiphosphotyrosine antibody (5 μg/mL) (Santa Cruz Biotechnology, Santa Cruz, CA). Cells were mounted and viewed under a 40× oil immersion lens using a confocal laser scanning microscope.

Inhibition of Phosphotyrosinylation

HDMVEC were exposed to T or T plus the specific tyrosine phosphorylation inhibitor, Genistein (40 μg/mL). Cells were incubated at 37°C for 15 to 30 minutes or 6 hours and then fixed and immunofluorescently stained for actin cytoskeleton (Rhodamine) and tyrosine phosphorylation (FITC) (15–30 minutes time-point) or were immunofluorescently stained for actin cytoskeleton (rhodamine) and VE-cadherin expression (FITC) (6 hours time-point) as described above. The cells were analyzed by confocal microscopy for phosphotyrosinylation or reduced junctional VE-cadherin expression and gapping between cells.

Analysis of HDMVEC VE-Cadherin Expression by Flow Cytometry

HDMVEC exposed to T or NT for 30 minutes, 3, 6, or 24 hours were rinsed with 1× PBS, versenized and spun down at 1000g for 5 minutes. The supernatant was discarded and the cells counted and split into 100-μL aliquots of 106 cells each. One aliquot was fixed in 100 μL of 1% paraformaldehyde for 30 minutes and used as unstained normal control. The remaining aliquots were exposed to 1 μg of mouse antihuman VE-cadherin antibody for 30 minutes, washed with PBS and exposed to antimouse FITC IgG1 for 30 minutes followed by washes with PBS. The cells were fixed in 100 μL of 1% paraformaldehyde. An aliquot exposed to mouse antihuman IgG1 and antimouse FITC IgG1 and an aliquot exposed to secondary antibody alone were used as controls. The cells were analyzed by flow cytometry at the University of Pennsylvania’s Flow Cytometry Core Facility.

Analysis of VE-Cadherin Expression by Western Blotting

Protein from T and NT exposed HDMVEC was extracted as described below. HDMVEC were grown in a monolayer on 15-mm polystyrene Petri dishes, rinsed with PBS, and scraped off on ice using a cell scraper in 250 μL/Petri dish of cold lysis buffer containing 20 mmol/L Tris-HCl (pH 7.2), 1 mmol/L EDTA, 1 mmol/L EGTA, 0.1 mmol/L NaCl, 1 mmol/L PMSF, 2 μg/mL aprotinin, 2 μg/mL leupeptin, and 2 μg/mL pepstatin. Cell lysates were homogenized, incubated on ice for 1 hour, and microcentrifuged at 1000 rpm. Supernatant was microcentrifuged at 16,400 rpm (25,000 rcf) for 30 minutes. The supernatant from this spin containing the cytosolic proteins was pipetted out into an Eppendorf tube, labeled, and stored on ice. The pellet was resuspended in ice-cold buffer to remove contaminating cytosolic proteins, microcentrifuged at 16,400 rpm for 30 minutes, and the supernatant was discarded. The pellet was resuspended in 500 μL of ice-cold solubilizing buffer containing 50 mmol/L Tris-HCl (pH 7.4), 5 mmol/L EDTA, 150 mmol/L NaCl, 0.5% NP-40, 0.1% SDS, 1 mmol/L PMSF, 2 μg/mL aprotinin, 2 μg/mL leupeptin, and 2 μg/mL pepstatin. This solution was kept on ice for 1 hour after frequent vortexing and then microcentrifuged at 16,400 rpm for 30 minutes. The supernatant was pipetted out into a separate Eppendorf tube and labeled as membrane proteins. Quantitation of protein in the cytosolic and membrane fractions was carried out using a BCA protein assay kit (Pierce, Rockford, IL). Approximately 5 μg of protein from cytosolic and membrane fractions of T and NT treated HDMVEC was electrophoresed on a 10% SDS-PAGE gel and transferred to polyvinylidene diflouride membranes. Membranes were then blocked for 1 hour at room temperature with 5% nonfat milk in Tris buffered saline (TBS). A polyclonal goat antihuman VE-cadherin antibody (Santa Cruz Biotechnology, Santa Cruz, CA) was added at 1:1000 dilution in 5% nonfat milk in TBS for 1 hour at room temperature. The blots were rinsed with TBS containing 0.1% Tween 20 (TTBS). The blots were incubated with a horseradish peroxidase-conjugated mouse (human adsorbed) antigoat antibody (Santa Cruz Biotechnology) at 1:2000 dilution in 5% nonfat milk in TBS and again rinsed with TTBS. An enhanced chemiluminescence kit (Amersham, Piscataway, NJ) was used to detect immunoreactive bands.

Statistical Analysis

Means and standard errors from means were calculated using Microsoft Excel. Graphpad Instat software was used to carry out a one-way analysis of variance test and the Tukey-Kramer multiple comparisons tests on each set of data. The Student t test was used to compare differences in mean vascular permeability as well as differences in mean blood flow before and after treatment in tumor and control tissue.

RESULTS

TNF Targets Tumor Vascular Endothelial Cell Junctions in Human ILP Patients

ILP using TNF in combination with melphalan is uniquely effective therapy for bulky in transit melanoma of the extremity (Fig. 1a). This regional perfusion delivers high-dose chemotherapy in a single treatment that targets tumor tissue with relatively lesser damage to surrounding normal tissue. Given the fact that these patients have limited treatment possibilities and can potentially face limb amputation, such toxicity to normal tissues appears acceptable.20 Although ILP using melphalan alone has shown significant response rates against melanoma, the addition of TNF to the treatment regimen markedly augments tumor response especially against bulky disease. However, the mechanism underlying the role of TNF in this antitumor therapy has not been elucidated. Electron microscopic analyses of tumor tissue sections from patients treated with melphalan plus TNF ILP show clear evidence of erythrostasis in the tumor vasculature by 24 hours post-treatment (Fig. 1b), in contrast to that of patients treated with melphalan alone (Fig. 1c). Analysis of the tumor cryosections of TNF plus melphalan ILP patients by PECAM immunohistochemistry at a magnification of 200× showed no apparent gross damage to tumor vasculature at 3 hours post-treatment (Fig. 1d). However, at the electron microscopic level, these tumors showed distinct edema in the vasculature as well as widespread gapping between endothelial cells (Fig. 1e). It was also observed that tumor necrosis (data not shown) followed but did not precede such vascular damage and erythrostasis further supporting the hypothesis that TNF exerts its antitumor effect by an antivascular mechanism rather than by direct tumor cell kill.21,22

graphic file with name 24FF1.jpg

Open in a new tab

FIGURE 1. Antivascular effect of TNF in clinical ILP samples. A, Preoperative (left panel) and 13-month postoperative photographs after a 90-minute ILP with TNF and melphalan (right panel) showing effectiveness of melphalan plus TNF ILP in a patient. In the intervening months, this patient received no other therapy. B, Electron micrographs showing a blood vessel with typical erythrostasis in a tumor treated with TNF plus melphalan. C, A normal blood vessel in a tumor treated with melphalan alone ILP. D, Photomicrograph of a PECAM-immunostained human tumor tissue section from a biopsy taken before ILP (left panel) and from the same patient after melphalan plus TNF ILP (right panel) showing no microscopic evidence of vascular damage at 3 hours post-treatment. E, Electron micrograph of a human tumor tissue section from the patient in D showing gapping between endothelial cells surrounding the blood vessel at 3 hours post-treatment.

TNF Increases Tumor Vascular Leakiness

A preclinical human melanoma xenograft model in nude mice was developed to test the hypothesis that TNF selectively increased permeability in tumor vasculature compared with that in TNF-exposed nontumor tissue. Using the Evan’s blue-albumin permeability assay, it was determined that at 30 minutes post-TNF treatment, tumor vascular permeability, on average, was 2 times greater than in muscle from the same animal and about 5 to 8 times greater than that seen in BSA treated tumor or muscle (data not shown). At 3 hours post-TNF treatment, vascular permeability was about 4 times greater in TNF treated tumors than in similarly treated muscle tissue and on average about 7 times greater in BSA treated tumor or muscle. While statistical analysis using the Student t test showed that there was no significant difference at 30 minutes post-treatment, the difference in permeability was significant (P < 0.01) at 3 hours post-treatment (data not shown).

TNF Decreases Tumor Blood Flow

To test whether an increase in vascular permeability was accompanied by a decrease in blood flow in tumors compared with nontumor tissue post-TNF treatment, the xenon clearance method was used in the in vivo human melanoma xenograft model to measure blood flow at baseline and after TNF treatment. While blood flow dropped 18-fold in tumors by 6 hours after TNF treatment, it dropped only 4- to 5-fold in nontumor tissue (skin) (Fig. 2). Statistical analyses using the Student t test showed no significant difference (P > 0.05) between baseline blood flow and that at maximum response at 6 hours after TNF in skin, but there was a significant difference (P < 0.05) in tumor.

graphic file with name 24FF2.jpg

Open in a new tab

FIGURE 2. Effect of TNF on blood flow in tumor versus skin in a preclinical model. Graph showing maximum response in terms of decreased blood flow in TNF treated tumors compared with TNF treated skin from nude mice carrying human melanoma xenografts at 6 hours post-treatment.

TNF Selectively Disrupts Tumor Endothelial Junctions by Changing Localization of VE-Cadherin

To determine the molecular basis of the selective increase in tumor vascular permeability and decrease in tumor blood flow, it was hypothesized that a selective physical disruption of the tumor vascular endothelial junctions occurs after TNF treatment. To test this hypothesis, an in vitro model was developed consisting of confluent HDMVEC that were exposed to either tumor specific ILP conditions (T) or nontumor ILP conditions (NT) for varying periods of time (30 minutes, 3 hours, 6 hours, and 24 hours). T included melanoma conditioned medium, hypoxia (1% oxygen) and TNF (2 μg/mL) while NT included normal cell culture medium, normoxia, and TNF. The data demonstrated that tumor-specific ILP conditions caused a disappearance of VE-cadherin from vascular endothelial junctions resulting in gapping between endothelial cells by 3 hours post-TNF treatment (Fig. 3). Nontumor ILP conditions and other control conditions failed to show a similar effect. It was also observed that endothelial cells exposed to T showed increased tyrosine phosphorylation at endothelial cell junctions as early as 15 to 30 minutes after treatment compared with cells exposed to NT (Fig. 4). In general, melanoma conditioned medium and TNF appeared to have a greater influence on endothelial cell-cell junctional disruption and tyrosine phosphorylation of junctional proteins than hypoxia.

graphic file with name 24FF3.jpg

Open in a new tab

FIGURE 3. Effect of TNF on VE-cadherin expression in HDMVEC. Confocal laser scanning photomicrographs of HDMVEC that were exposed to either control conditions, or nontumor ILP (left panel, middle) or tumor-specific ILP conditions (right panel, bottom) for 3 hours and subsequently immunofluorescently stained with phalloidin rhodamine and an antihuman VE-cadherin antibody that was secondarily labeled with FITC. NM, normal medium; MCM, melanoma conditioned medium. While nontumor ILP conditions (NM, normoxia, TNF) had minimal effect on VE-cadherin expression at endothelial junctions, tumor-specific ILP conditions (MCM, hypoxia, TNF) caused a decreased expression of VE-cadherin and gapping at endothelial junctions.

graphic file with name 24FF4.jpg

Open in a new tab

FIGURE 4. Effect of TNF on phosphotyrosinylation of HDMVEC junctional proteins. Confocal laser scanning photomicrographs showing HDMVEC that were immunofluorescently stained with an antiphosphotyrosine antibody at 15 to 30 minutes post-treatment. A, Cells exposed to normal medium (NM), normoxia, and no TNF showed no phosphorylation. Those exposed to (B) nontumor ILP conditions (NT) and to (C) melanoma conditioned medium (MCM), normoxia, and no TNF showed minimal phosphorylation. D, Cells exposed to tumor-specific ILP conditions (T) showed maximum phosphorylation at endothelial junctions.

To determine whether phosphorylation of junctional proteins was a prerequisite for decreased VE-cadherin expression at endothelial junctions leading to gapping between endothelial cells, HDMVEC were exposed to T or T plus the specific tyrosine phosphorylation inhibitor, Genistein for 15 to 30 minutes. Results showed that while T caused phosphorylation of junctional proteins (Fig. 5a), phosphorylation was absent in the Genistein-treated group (Fig. 5b). In both instances, the cells were still in contact with each other at endothelial junctions (data not shown) showing that 15 to 30 minutes was sufficient time for phosphorylation but was not sufficient for gapping. In the second set of experiments carried out under similar conditions as above, the cells were allowed to incubate for as long as 6 hours and the cells were immunofluorescently stained for actin cytoskeleton and VE-cadherin. Results from this experiment showed that there was not only reduced expression of VE-cadherin at cell-cell junctions (Fig. 5c) but also extensive gapping between HDMVEC in cells that were exposed to tumor-specific ILP conditions (Fig. 5 d). Cells that were exposed to tumor-specific ILP conditions plus Genistein, however, did not show this decreased expression of VE-cadherin at cell-cell junctions (Fig. 5e) and also did not show gapping between cells (Fig. 5f).

graphic file with name 24FF5.jpg

Open in a new tab

FIGURE 5. Effect of inhibition of phosphotyrosinylation on junctional VE-cadherin expression and gapping of endothelial cells. Confocal laser scanning micrographs of HDMVEC that were exposed to tumor-specific ILP conditions (T) or T + Genistein for 15 to 30 minutes (A, B) or 6 hours (C–F). a, HDMVEC exposed to T or (B) T + Genestein for 15 to 30 minutes and immunofluorescently stained for phosphotyrosinylation at cell junctions; c, HDMVEC exposed to T for 6 hours and immunofluorescently stained for VE-cadherin or for (D) actin cytoskeleton. E, HDMVEC exposed to T + Genestein for 6 hours and immunofluorescently stained for VE-cadherin or for (F) actin cytoskeleton. Results show that exposure to T for 15 to 30 minutes caused phosphorylation and that addition of Genistein inhibits this phosphorylation. Exposure to T for 6 hours caused reduction in VE-cadherin expression at cell junctions and also gapping between cells. Genistein inhibits both these phenomena.

To determine whether tumor specific ILP conditions degraded VE-cadherin at endothelial cell junctions, flow cytometric analyses of T or NT treated endothelial cells and VE-cadherin immunohistochemical analyses on clinical specimens from ILP patients was conducted. No significant change (P > 0.05) was seen in overall expression of VE-cadherin in endothelial cells exposed to T with time post-treatment compared with controls in the flow cytometric analyses (Fig. 6a). Western Blot analysis showed that the overall amount of VE-cadherin in the membrane fraction of HDMVEC did not change even after a 15- to 18-hour exposure to T (Fig. 6b). Data from VE-cadherin immunostained tissue sections from human ILP patients that were all exposed to both TNF and melphalan similarly showed that there was no decrease in overall levels of VE-cadherin after exposure to T (Fig. 6c). Collectively, these results when combined with those presented in Figure 3 indicate that after TNF treatment, VE-cadherin moves within the endothelial cell membrane from the region of endothelial cell-cell junctions to nonjunctional areas of the cell membrane, and the above data collectively argue in favor of relocalization of VE-cadherin rather than degradation of the protein subsequent to TNF treatment.

graphic file with name 24FF6.jpg

Open in a new tab

FIGURE 6. Effect of TNF on total VE-cadherin expression in HDMVEC and in clinical samples. A, FACS analysis of HDMVEC exposed to control conditions (▪) or tumor-specific ILP conditions (T) (□) or nontumor ILP conditions (NT) (▨) that were immunofluorescently labeled for VE-cadherin. No significant difference in VE-cadherin expression was observed over time-points studied. B, Western blot analysis showing that VE-cadherin remains in the membrane fraction of the cell even after exposure to T for 12 to 16 hours. Lane 1, cytosolic protein fraction from HDMVEC exposed NT; lane 2, cytosolic protein fraction from HDMVEC exposed to T; lane 3, membrane protein fraction from HDMVEC exposed to NT; lane 4, membrane protein fraction from HDMVEC exposed to T. C, Data from VE-cadherin immunostained clinical specimens from patients before (upper panel) and 3 hours after (lower panel) TNF plus melphalan ILP showing no overall decrease in VE-cadherin expression.

To determine whether the in vitro TNF effects on VE-cadherin expression were relevant in the clinical setting, human melanoma and skin cryosections were analyzed from pre and 3 hours post-TNF plus melphalan ILP that were double immunofluorescently labeled with an antihuman cadherin antibody (red) and an antihuman alpha smooth muscle actin antibody (green) by confocal microscopy at a magnification of 1000×. TNF treated and untreated skin cryosections showed normal continuous distribution of VE-cadherin at vascular endothelial junctions (Fig. 7a, b). While untreated tumor showed a slightly abnormal distribution of VE-cadherin (Fig. 7c, e), TNF treated tumor cryosections showed a highly abnormal, discontinuous, and nonuniform distribution of VE-cadherin at endothelial cell junctions (Fig. 7d, f) in agreement with the in vitro observations.

graphic file with name 24FF7.jpg

Open in a new tab

FIGURE 7. Effect of TNF on junctional VE-cadherin expression in clinical samples from ILP patients. Representative confocal laser scanning photomicrographs of clinical tumor and skin cryosections that were double immunofluorescently labeled with an antihuman VE-cadherin antibody (red) and an anti-alpha smooth actin antibody (green). Skin before TNF plus melphalan ILP (A) and skin 3 hours after TNF plus melphalan ILP (B) showing normal, linear distribution of VE-cadherin distribution at endothelial cell junctions. C and E, Tumor from patients before TNF plus melphalan ILP showing a slightly abnormal distribution of VE-cadherin. D and F, Tumor from patients 3 hours after TNF plus melphalan ILP showing highly abnormal, discontinuous distribution of VE-cadherin.

Taken together, the data indicate that TNF selectively damages tumor vascular integrity by affecting VE-cadherin distribution at endothelial junctions, thereby causing disruption of cell-cell contacts in the endothelial lining of tumor vessels that appear to be more vulnerable to TNF than normal tissue blood vessels. Such a disruption could conceivably cause the gapping seen in TNF-treated human dermal microvascular endothelial cells leading to increased vascular permeability, decreased blood flow, and clumping/adhesion of cells within the vasculature resulting in erythrostasis.

DISCUSSION

Although TNF has direct cytotoxic effects against many tumor cell lines in vitro, clinical observations as well as the known actions of TNF suggest that it targets the tumor vasculature and not the tumor cells directly. Direct TNF-mediated cytotoxicity against tumor cells in vitro occurs within 48 to 72 hours,23 and this is much longer than the time it takes to see tumor necrosis following ILP.6 The mechanism underlying the difference in response to this treatment within the tumor vasculature causing rapid tumor necrosis compared with the normal vasculature in which there is no effect, has not been established.

While tumors are known to be more hypoxic and acidic than normal tissue,24 there are very many important differences between the 2 vasculatures as well. Normal tissues have a regular predictable pattern of blood vessels while tumors are known to have highly abnormal, distended capillaries with leaky endothelial linings, and sluggish blood flow.25 There is evidence that TNF causes increased permeability specifically within tumors compared with normal tissues. Studies using radiolabeled antibodies suggested that the delivery of these large molecules to tumors was augmented significantly by pretreatment with TNF, suggesting a further increase in vascular permeability.26 The precise cellular and molecular changes that occur in the tumor vasculature to cause this increase in vessel permeability following TNF is not known.

At a structural level, vascular endothelial adherens junctions promote intercellular adhesion and contribute to the control of vascular integrity and leakiness in any given tissue, including tumors.27 At a molecular level, the vascular adherens junctions are formed by a transmembrane and cell specific adhesive protein, VE-cadherin, which is linked by its cytoplasmic tail to intracellular proteins.28 VE-cadherin is a 130-kDa cell surface glycoprotein that is constitutively expressed in the vascular endothelium and mediates Ca2+-dependent cell-cell adhesion. It is composed of an N-terminal extracellular domain and a relatively small cytoplasmic domain at the C-terminal side. A single membrane-spanning region connects the 2 domains. The extracellular domains of VE-cadherin molecules from neighboring cells establish a homophilic type of binding resulting in cell-cell adhesion. On the cytoplasmic end, VE-cadherin interacts with beta-catenin, plakoglobin, and p120. Beta-catenin and plakoglobin bind alpha-catenin, which is homologous to vinculin. The cadherin-catenin complex is in contact with the actin cytoskeleton intracellularly via vinculin. This complex organization involving the cadherin-catenin complex and the actin cytoskeleton is known to be important for angiogenesis and neovascularization and thus, for tumor growth and metastases.29 There is also evidence in the literature to show that VE-cadherin expression is important for transferring signals between neighboring cells,30 and for cell-cell adhesion and to regulate vascular permeability.31 It is conceivable, therefore, that any disruption of tumor VE-cadherin would alter the integrity of the vasculature and have an impact on tumor viability.

Evidence in the literature indicates that an increase in vascular permeability is almost always associated with a disruption of the adherens junction complexes that are glued together via cadherin homodimers.32–34 In a pig model, it has been demonstrated that cardiopulmonary bypass is associated with signs of degradation of endothelial and cardio-myocytes adherens junction VE-cadherin leading to increased vascular permeability.33 Another study analyzed the effect of histamine on in vitro permeability in cultured endothelial cells and showed that an induction of tyrosine phosphorylation of VE-cadherin contributes to histamine’s effect on endothelial permeability.34 Also, it has been shown that VEGF increases albumin permeability across endothelial monolayers in vitro and that the structural basis of increased vascular permeability is the rearrangement of endothelial junctional proteins involving the mitogen-activated protein kinase signal transduction pathway.35 Polymorphonuclear leukocytes infiltration into tissues is frequently accompanied by increase in vascular permeability and the structural basis of such a change has been shown to be a disappearance from endothelial cell-cell contacts of adherens junction components including VE-cadherin.36

Our in vitro and in vivo data indicate that the tumor microenvironment is less conducive to strong vascular adherens junctions than normal tissue environments. Using cultured microvascular endothelial cells, tumor supernatant from human melanoma plus hypoxia led to a moderate disruption of cell-cell junctions. The combination of conditioned medium from melanoma cell cultures and hypoxia was used to reproduce the tumor microenvironment. The addition of TNF at the levels used in ILP to this simulated tumor microenvironment caused a complete breakdown of cell-cell contact. This TNF effect was greatly decreased in normoxia without tumor supernatant, thus offering some explanation to the selectivity of the effect on tumor endothelium versus normal tissues. Further evidence for the specificity of TNF to tumor microvasculature comes from the confocal studies of ILP specimens. TNF appears to disrupt the integrity of the tumor vasculature represented by a complete loss of or abnormal/discontinuous VE-cadherin expression at tumor endothelial junctions in relation to skin perfused under the same conditions. Our in vitro data suggest that VE-cadherin molecules from endothelial cell junctions do not undergo degradation but move within the cell membrane away from junctional areas in response to TNF treatment. It is hypothesized that such a change in the endothelial junctions could result in the gapping seen between tumor endothelial cells in vitro and in vivo. The damaged endothelial lining could in turn lead to increased vascular leakiness, decreased blood flow, and resulting hemostasis or erythrostasis. Evidence from angiograms, primarily of human soft tissue sarcomas done before and after perfusion with TNF plus melphalan, show specific obliteration of blood flow in tumor vessels compared with surrounding normal vessels.37 The specific factor in the tumor supernatant in the in vitro studies or in the tumor microenvironment in vivo that is responsible for such an effect is not known. Clearly, the TNF effect is augmented by hypoxia and the tumor supernatant, and the impact these conditions have on endothelial cells is under investigation. Evidence in the literature and our in vitro data also indicate that movement of VE-cadherin away from endothelial cell junctions and gapping between endothelial cells is preceded by increased phosphorylation of junctional proteins.34 Similar results have been obtained in studies using inflammatory mediators such as histamine and TNF on human placental vasculature38 and also by other vascular hyperpermeability inducing factors.39

Taken together, the data obtained from this study suggest that one of the ways in which high-dose TNF selectively damages the vascular lining of tumors is by disrupting VE-cadherin complexes at the vascular endothelial adherens junctions. Such damage to the endothelial lining of the tumor vasculature could conceivably lead to increased vascular leakiness, decreased blood flow, and finally to erythrostasis, which could be lethal to the tumor. TNF is a pleiotropic molecule, and it is possible that its selective antitumor activity has more than one molecular basis that originates from the selective distribution of TNF binding sites on tumor blood vessels.40

Unfortunately, the systemic administration of TNF is severely limited by hypotension reflecting the role TNF plays in sepsis. Pharmacokinetic studies have shown that the intravascular concentration of TNF achieved in ILP is 2000 to 3000 ng/mL.8 However, the normal levels of TNF that can be achieved with acceptable toxicity with systemic administration is 3 log orders less (1–3 ng/mL) due to dose limiting toxicity.41 Clearly, it is impossible to reproduce the conditions achieved in ILP with an identical systemic regimen. Nevertheless, given the distinct physiology of tumors, it appears that the VE-cadherin based molecular mechanism of tumor destruction may be effective and could potentially form the basis of new therapeutic efforts that can be used systemically to selectively target tumors.

ACKNOWLEDGMENTS

The authors thank Dominique Harris for assistance with laboratory techniques related to this project.

Footnotes

Supported in part by the Georgene S. Harmelin Endowment Fund (to D.L.F.).

Reprints: Douglas Fraker, MD, Division of Endocrine and Oncologic Surgery, University of Pennsylvania, 4th Floor, Silverstein Building, 3400 Spruce Street, Philadelphia, PA 19104. E-mail: Frakerd@uphs.upenn.edu.

REFERENCES

  • 1.Devita VT, Hellman S, Rosenberg SA. Cancer: Principles & Practice of Oncology, 6th ed. Philadelphia: Lippincott-Raven, 2001. [Google Scholar]
  • 2.Schuster LM, Haluska F, Fraker DL Skin: malignant melanoma. In: Clinical Oncology, 2nd ed. New York: Churchill Livingstone, 2000:1317–1350. [Google Scholar]
  • 3.Fraker DL. Regional therapies of cancer. In: Norton JA, Bollinger RR, Chang AE, et al, eds. Surgery: Scientific Basis and Current Practice. 2001:1863–1880. [Google Scholar]
  • 4.Fraker DL. Hyperthermic regional perfusion for melanoma and sarcoma of the limbs. Curr Probl Surg. 2001;36:841–908. [DOI] [PubMed] [Google Scholar]
  • 5.Klaase JM, Kroon BB, van Geel AN, et al. Prognostic factors for tumor response and limb recurrence-free interval in patients with advanced melanoma of the limbs treated with regional isolated perfusion with melphalan. Surgery. 1994;115:39–45. [PubMed] [Google Scholar]
  • 6.Leinard J, Ewalenko P, Delmotte JJ, et al. High-dose recombinant tumor necrosis factor alpha in combination with interferon gamma and melphalan in isolation perfusion of the limbs for melanoma and sarcoma. J Clin Oncol. 1992;10:52–60. [DOI] [PubMed] [Google Scholar]
  • 7.Lienard D, Lejeune FJ, Ewalneko P. In transit metastases of malignant melanoma treated by high dose rTNF alpha in combination with interferon-gamma and melphalan in isolation perfusion. World J Surg. 1992;16:234–240. [DOI] [PubMed] [Google Scholar]
  • 8.Fraker DL, Alexander HR, Andrich M, et al. Treatment of patients with melanoma of the extremity using hyperthermic isolated limb perfusion with melphalan, tumor necrosis factor, and interferon gamma: results of a tumor necrosis factor dose escalation study. J Clin Oncol. 1996;14:479–489. [DOI] [PubMed] [Google Scholar]
  • 9.Deleted in proof.
  • 10.Mule JJ, Asher A, McIntosh J, et al. Antitumor effect of recombinant tumor necrosis factor-α against murine sarcomas at visceral sites: tumor size influences the response to therapy. Cancer Immunol Immunother. 1988;26:202–208. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Gutman M, Inbar M, Lev-Shlush D, et al. High dose tumor necrosis factor-alpha and melphalan administered via isolated limb perfusion for advanced limb soft tissue sarcoma results in a >90% response rate and limb preservation. Cancer. 1997;79:1129–1137. [DOI] [PubMed] [Google Scholar]
  • 12.Lejeune FJ. High dose recombinant tumor necrosis factor (rTNFa) administered by isolation perfusion for advanced tumors of the limbs: a model for biochemotherapy of cancer. Eur J Cancer. 1995;31A:1009–1016. [DOI] [PubMed] [Google Scholar]
  • 13.Reugg C, Yilmaz A, Bieler G, et al. Evidence for the involvement of endothelial cell integrin alphaVbeta3 in the disruption of the tumor vasculature induced by TNF and IFN-gamma. Nat Med. 1998;4:408–414. [DOI] [PubMed] [Google Scholar]
  • 14.Fraker D. A Phase III trial of isolated limb perfusion for extremity melanoma comparing melphalan alone versus melphalan plus tumor necrosis factor and interferon gamma. Ann Surg Oncol. 2002;95:8. [Google Scholar]
  • 15.Asher A, Mule JJ, Reichert CM, et al. Studies on the antitumor efficacy of systemically administered recombinant TNF against several murine tumors in vivo. J Immunol. 1987;138:963–974. [PubMed] [Google Scholar]
  • 16.Nwariaku FE, Liu Z, Turnage RH, et al. Tyrosine phosphorylation of vascular endothelial cadherin and the regulation of microvascular permeability. Surgery. 2002;132:180–182. [DOI] [PubMed] [Google Scholar]
  • 17.Friedl J, Puhlmann M, Bartlett DL, et al. Induction of permeability across endothelial cell monolayers by tumor necrosis factor (TNF) occurs via a tissue factor-dependent mechanism: relationship between the procoagulant and permeability effects of TNF. Blood. 2002;100:1334–1339. [PubMed] [Google Scholar]
  • 18.Friedl J, Turner E, Alexander RH. Augmentation of endothelial cell monolayer permeability by hyperthermia but not tumor necrosis factor: evidence for disruption of vascular integrity via VE-cadherin down regulation. Int J Oncol. 2003;23:611–613. [PubMed] [Google Scholar]
  • 19.Richard L, Velasco P, Detmar M. A simple immunomagnetic protocol for the selective isolation and long-term culture of human dermal microvascular endothelial cells. Exp Cell Res. 1998;240:1–6. [DOI] [PubMed] [Google Scholar]
  • 20.Grunhagen DJ, de Wilt JHW, Graveland WJ, et al. The palliative value of tumor necrosis factor alpha based isolated limb perfusion in patients with metastatic sarcoma and melanoma. Cancer. 2006;106:156–162. [DOI] [PubMed] [Google Scholar]
  • 21.Bauer TW, Gutierrez M, Dudrick DJ, et al. A human melanoma xenograft in a nude rat responds to isolated limb perfusion with TNF plus melphalan. Surgery. 2003;133:420–428. [DOI] [PubMed] [Google Scholar]
  • 22.Hagen TL, Eggermont AM. Tumor vascular therapy with TNF: critical review on animal models. Methods Mol Med. 2002;98:227–246. [DOI] [PubMed] [Google Scholar]
  • 23.Sugarman BJ, Aggarwal BB, Hass EP, et al. Recombinant human tumor necrosis factor-alpha: effects on proliferation of normal and transformed cells in vitro. Science. 1985;230:943–945. [DOI] [PubMed] [Google Scholar]
  • 24.Ivanov S. Expression of hypoxia-inducible cell-surface transmembrane carbonic anhydrases in human cancer. Am J Pathol. 2001;158:905–919. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Brown JM, Giaccia AJ. The unique physiology of solid tumors: opportunities and problems for cancer therapy. Cancer Res. 1998;58:1408–1416. [PubMed] [Google Scholar]
  • 26.Folli S, Pelegrin A, Chalandon Y, et al. TNF can enhance radio-antibody uptake in human colon carcinoma xenografts by increasing vascular permeability. Int J Cancer. 1993;53:829–836. [DOI] [PubMed] [Google Scholar]
  • 27.Dejana E, Bazzoni G, Lampugnani MG. VE-cadherin: only an intercellular glue? Exp Cell Res. 1999;252:13–19. [DOI] [PubMed] [Google Scholar]
  • 28.Petzelbauer P, Halama T, Groger M. Endothelial adherens junctions. J Invest Dermatol Symp Proc. 2000;5:10–13. [DOI] [PubMed] [Google Scholar]
  • 29.Dejana E, Lampugnani MG, Martinez-Estrada O, et al. The molecular organization of endothelial junctions and their functional role in vascular morphogenesis and permeability. Int J Dev Biol. 2000;44:743–748. [PubMed] [Google Scholar]
  • 30.Corada M, Liao F, Lindgren M, et al. Monoclonal antibodies directed to different regions of VE-cadherin extracellular domain affect adhesion and clustering of the protein and modulate endothelial permeability. Blood. 2001;97:1679–1684. [DOI] [PubMed] [Google Scholar]
  • 31.Renard N, Nooijen PT, Schakwijk L, et al. VWF release and platelet aggregation in human melanoma after perfusion with TNF alpha. J Pathol. 1995;176:279–287. [DOI] [PubMed] [Google Scholar]
  • 32.Yuan SY. Signal transduction pathways in enhanced microvascular permeability. Microcirculation. 2000;6:395–403. [PubMed] [Google Scholar]
  • 33.Bianchi C, Araujo EG, Sato K, et al. Biochemical and structural evidence for pig myocardium adherens junction disruption by cardiopulmonary bypass. Circulation. 2001;104:1319–1324. [DOI] [PubMed] [Google Scholar]
  • 34.Andriopoulou P, Navarro P, Zanetti A, et al. Histamine induces tyrosine phosphorylation of endothelial cell-to-cell adherens junctions. Arterioscler Thromb Vasc. 1999;19:2286–2297. [DOI] [PubMed] [Google Scholar]
  • 35.Kevil CG, Payne DK, Mire E, et al. Vascular permeability factor/vascular endothelial cell growth factor-mediated permeability occurs through disorganization of endothelial junctional proteins. J Biol Chem. 1998;273:15099–15103. [DOI] [PubMed] [Google Scholar]
  • 36.Del Maschio A, Zanetti A, Corada M, et al. Polymorphonuclear leucocyte adhesion triggers the disorganization of endothelial cell-to-cell adherens junctions. J Cell Biol. 1996;135:497–510. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Olieman AF, van Ginkel RJ, Hoekstra HJ, et al. Angiographic response of locally advanced soft-tissue sarcoma following hyperthermic isolated limb perfusion with tumor necrosis factor. Ann Surg Oncol. 1997;4:64–69. [DOI] [PubMed] [Google Scholar]
  • 38.Leach L, Firth JA. Structure and permeability of human placental microvasculature. Microsc Res Tech. 1997;38:137–144. [DOI] [PubMed] [Google Scholar]
  • 39.Corada M, Mariotti M, Thurston G, et al. Vascular endothelial-cadherin is an important determinant of microvascular integrity in vivo. Proc Natl Acad Sci USA. 1999;96:9815–9820. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Isaka T, Yoshimine T, Maruno M, et al. Morphological effects of TNF alpha on the blood vessels in rat experimental brain tumors. Neurol Med Chir. 1996;36:423–427. [DOI] [PubMed] [Google Scholar]
  • 41.Fraker DL, Alexander HR, Pass HI. Biologic therapy with TNF: clinical applications by systemic and regional administration. In: DeVita V, Hellman S, Rosenberg SA, eds. Biologic Therapy of Cancer, 2nd ed. 1995:329–346. [Google Scholar]

Articles from Annals of Surgery are provided here courtesy of Lippincott, Williams, and Wilkins

RESOURCES