Normal human fibroblasts enable melanoma cells to induce angiogenesis in type I collagen

Normal human fibroblasts enable melanoma cells to induce angiogenesis in type I collagen Lee J. Goldstein, MD, Haiying Chen, MD, Richard J. Bauer, PhD, Stephen M. Bauer, MD, and Omaida C. Velazquez, MD, Philadelphia, Pa

Background. We previously reported that fibroblasts induce human microvascular endothelial cells (HMVECs) to differentiate from monolayer to capillarylike morphology. We now test the hypothesis that fibroblasts modulate angiogenesis in melanoma cells. Methods. We tested 12 human melanoma lines (2 radial growth phase (RGP), 3 vertical growth phase (VGP), and 7 metastatic (MM)) for ability to induce HMVECs to invade/migrate into collagen and form capillarylike networks. HMVEC monolayers were overlaid with 3-dimensional collagen gels embedded with melanoma cells alone (M), fibroblasts alone (F), or a 1:1 mixture of the 2 cells (M+F). After 5 days, gels were removed, fixed, and HMVEC networks were quantified by von Willebrand’s factor (vWF) immunofluorescence. The influence of soluble factors on HMVEC invasion/migration into collagen was assessed with the use of acellular 3-D collagen gels overlaid on HMVEC monolayers, cultured with conditioned media (CM) derived from monolayers of M, F, or M+F. Angiogenic growth factors involved in the observed invasion/migration were identified with the use of a RayBio Cytokine Antibody Array (RayBiotech, Norcross, Ga). Results. Cell line--specific variability in melanoma-supported angiogenesis was observed only when in combination with fibroblasts (analysis of variance [ANOVA], P < .01). Melanoma plus fibroblasts uniformly resulted in a significantly higher angiogenic response than melanoma alone (P < .05). One vertical growth phase and one metastatic melanoma line, while weakly angiogenic alone, induced significantly higher angiogenesis than either fibroblast or melanoma alone (P < .05) when combined with fibroblasts. CM from M or M+F induced significantly less HMVEC invasion/migration into collagen than CM from fibroblasts alone. Interleukin 8, monocyte chemotactic protein-1, and tissue inhibitor of metalloproteinase-2 were identified as significantly elevated in the media derived from M+F cultures, compared with either cell type alone. Conclusion. To our knowledge, this is the first report demonstrating that melanoma-supported angiogenesis in collagen is more significantly influenced by normal skin-derived fibroblasts than by the intrinsic biology of the melanoma cell type. Interleukin 8, monocyte chemotactic protein-1, and tissue inhibitor of metalloproteinase-2 are implicated as potential paracrine factors regulating this observed effect. (Surgery 2005;138:439-49.) From the Hospital of the University of Pennsylvania

ANGIOGENESIS, the formation of new blood vessels, occurs by the local extension of preexisting vessels.1 The endothelial cells (ECs) comprising angiogenic vessels proliferate, migrate, and assemble into new capillaries in response to cellular Presented at the 66th Annual Meeting of the Society of University Surgeons, Nashville, Tennessee, February 9-12, 2005. Reprint requests: Omaida C. Velazquez, MD, Hospital of the University of Pennsylvania, Department of Surgery, 4 Silverstein, 3400 Spruce Street, Philadelphia, PA 19104. E-mail: omaida. [email protected]. 0039-6060/$ - see front matter Ó 2005 Mosby, Inc. All rights reserved. doi:10.1016/j.surg.2005.06.031

and humoral signals, which have not yet been completely defined.2 In malignant cell–induced angiogenesis, as the tumor burden begins to exceed its blood supply, the need for oxygen and nutrients requires tumor cells to induce angiogenesis by secreting angiogenic factors themselves or recruiting other cell types to aid them in this process.3,4 Initially, the study of tumor biology centered specifically on the malignant cell, its genetic derangements, and its biologic abilities to escape regulatory control of cell proliferation and apoptosis. Recently, however, the malignant tissue’s complexity is better appreciated, with a renewed focus on the tumor’s vascular and stromal components.5-7 SURGERY 439

440 Goldstein et al

Melanoma cells are 1 malignant tumor type known to interact with the surrounding stroma via growth factors, as well as cell-cell interactions.7,8 Human melanoma cells produce a variety of substances relevant to the induction of angiogenesis including angiogenic cytokines, extracellular matrix components, and metalloproteinases.9-11 In addition, it has been suggested that some malignant cells are not inducing angiogenesis per se, but rather manipulating the surrounding stromal cells (including EC and fibroblasts) and chemically coercing them to build the needed vascular infrastructure.4,7,8,12 This concept of malignant cells inducing the stroma to do their bidding prompted our initial investigation into the inherent angiogenic abilities of stromal cells. As the predominant cell type in tumor stroma, fibroblasts provide the matrix proteins that become the scaffolding of new tissues.6 Fibroblasts are also known to communicate with melanoma cells via growth factors, cytokines, and cell-cell interactions.10,12,13 After undergoing malignant transformation, cells have been shown to acquire newfound abilities with regard to communication with the stroma.14 On transformation, melanoma cells have been shown to shift cadherin expression from E-cadherin to N-cadherin, resulting in the ability to communicate with fibroblasts via N-cadherin–mediated gap junctions, while at the same time losing this same communication with keratinocytes.12,13 Recent work from our laboratory and others demonstrated the fibroblast’s ability to induce nonmalignant EC vessel formation and tumor angiogenesis.15,16 Previously, we observed that normal dermis-derived human fibroblasts induce human microvascular endothelial cells to invade/ migrate into collagen and form capillary networks, while some metastatic melanoma cell lines were unable to support this type of angiogenesis under identical culture conditions.15 In this in vitro human cell angiogenesis model, fibroblasts decreased EC proliferation, increased EC migration, and inhibited EC apoptosis.15 Our findings that fibroblasts stimulate angiogenesis and the reports that melanoma cells can acquire the ability to communicate with stromal cells led us to hypothesize that fibroblasts modulate angiogenesis in melanoma cells. Three-dimensional collagen-based in vitro assays are an effective and reliable method for assessing angiogenesis, and also have several advantages, including the ability to use human cells and combine multiple cell types, and the advantages of controlled manipulations of extracellular matrix, cell-cell interactions, and secreted factors.15,17 We

Surgery September 2005

employed our previously characterized 3-dimensional (3-D) collagen matrix assay (3DCMA)15,18 to study the angiogenic properties of melanoma cells and their potential interactions with fibroblasts. Implicating fibroblasts as a requisite cell type in the biology of melanoma angiogenesis may open new avenues for anti-angiogenic and anti-neoplastic therapies. METHODS Antibodies and fluorescent reagents. Mouse monoclonal anti-human vWF Ab was purchased from NeoMarkers Inc (Fremont, Calif), AlexaFluor 488 green flourescent goat anti-mouse IgG was obtained from Molecular Probes Inc (Eugene, Ore) and used as a seconday antibody. Hoechst nuclear dye (bisBenzamide at 10 mg/mL) was obtained from Sigma-Aldrich (St Louis, Mo). Cells. All cells and co-cultures were incubated at 37°C in 98% humidified air containing 5% CO2. Fibroblasts: Primary cultures of human dermal fibroblasts kindly provided by Dr M. Herlyn, Wistar Institute, Philadelphia, Pa, were initiated as explant cultures from trypsin-treated and epidermis-stripped neonatal foreskin, as previously reported,19 and cultured in DMEM with glutamine (Gibco/BRL, Gaithersburg, Md), 8 mmol/L HEPES (Sigma, St. Louis, Mo), and 10% fetal bovine serum (FBS) (Hyclone, Logan, Utah). Only passage 1 to 8 was utilized for these experiments. Fibroblast cell cultures were stained for endothelial cell (EC)–specific markers (CD31 and VWF) to confirm the absence of contaminating endothelial cells. Human dermal microvascular endothelial cells: Primary HMVEC, kindly provided by Dr D. Fraker, University of Pennsylvania, Philadelphia, P, were isolated and cultured, as previously described,20,21 on plates coated with collagen type I (1 mg/mL; Organogenesis, Canton, Mass) and passaged in Endothelial Growth Medium Bulletkit (EGM, catalogue # CC-3124; Clonetics/Cambrex Biosciences, East RutherFord, NJ). Only passage 1 to 5 was utilized for these experiments. Melanoma cell lines: Twelve melanoma cell lines kindly provided by Dr M. Herlyn were studied, 2 radial growth lines (WM 3211, WM 35), 3 vertical growth phase lines (WM 115, WM 3248, WM 793), and 7 metastatic growth phase lines (451 Lu, WM 1205, WM 1158, WM 9, WM 278, WM 1232, WM 164). The melanoma cells were cultured in MCDB 153/L15 medium (v/v: 4/1) supplemented with CaCl2 (2 mmol/L), insulin (5 lg/mL), and 2% FBS. 3-D collagen matrix assay (3-DCMA). Reconstruction of capillary-like structures in 3-D collagen

Surgery Volume 138, Number 3 gels was performed as previously described.15 Briefly, 24 well plates were plated with HMVEC monolayers, cultured to 80% confluency, and then overlaid with a 1-mm-thick layer of acellular collagen type I (1 mg/mL) prepared in M199 medium supplemented with L-glutamine, sodium bicarbonate, heparin (100 U/mL), vitamin C (50 lg/mL) and FBS (1%). After polymerization, the collagen layer was overlaid with a second (3 mm) collagen layer containing a combination of 2.5 3 105/mL fibroblasts and 2.5 3 105/mL melanoma cells (total cells 5 3 105/mL). Experimental controls included each melanoma line with the use of both 2.5 3 105/mL and 5 3 105 melanoma cells/ mL in the upper collagen layer, with no added fibroblasts. For additional experimental controls, the second collagen layer was constructed with the use of 5 3 105/mL fibroblasts, 2.5 3 105/mL fibroblasts, or acellular collagen (no cells added to the overlying collagen layers). These 3-D collagen gels were fed EGM medium and incubated at 37°C in 98% humidified air containing 5% CO2 for 5 days (a time point at which the gels are still translucent by virtue of less collagen constriction). Fresh medium was added every 48 hours. Evaluation of 3-D collagen gels. At the completion of each experiment, each gel was fixed in Prefer (Anatech Ltd, Battle Creek Mich), lifted from its incubation well, and processed as a wholemount. Whole-mounts were analyzed by immunofluorescence with the use of antihuman vWF as described previously.18 Briefly, gels were stained with monoclonal anti-vWF antibody followed by an AlexaFluor 488–conjugated secondary antibody and counterstained with Hoechst dye (bisBenzamide at 10 mg/mL). vWF (+) endothelial cell networks were examined and photographed by inverted fluorescence microscopy. Endothelial networks were defined as groups of more than 3 HMVEC-forming polarized branching tubes and quantified by counting 10 random high-power fields (HPFs). Experiments were performed at least in duplicate and independently repeated at least 2 times. Conditioned media experiments. Metastatic Melanoma line WM 1158 was selected for further study of potential secreted angiogenic factors. Conditioned media (CM) was prepared by culturing monolayers of cells (fibroblasts and/or WM 1158) in a T75 flask with 6 mL of EGM media with reduced serum (1%) for 12 hours at 37°C and 5% CO2. CM was collected from 2 3 106 fibroblasts, 2 3 106 melanoma cells, and a combination of melanoma cells and fibroblasts (1 3 106 of each). After 12 hours, the supernatant solution

Goldstein et al 441

was collected and cleared of particulate matter by centrifugation. A quantity of 6 3 104 HMVEC were cultured for 12 hours on bovine collagen–coated, 24-well plates with EGM media. After 12 hours, the supernatant for these cultured cells was aspirated, and the cells were overlaid with a 1-mm layer of acellular type I collagen (1 mg/mL) (Organogenesis, Canton, Mass) prepared in M199 medium supplemented with L-glutamine, NaCO2 (7.5%), and 1% FBS. After polymerization of the acellular collagen layer, 1 mL of the previously prepared CM was added and cells were cultured for 24 hours. At 24 hours, the collagen gels were stained with Hoescht nuclear stain, and the HMVEC migrating into the acellular collagen layer were quantified with the use of Image Pro Plus software (Media Cybernetics, Silver Spring, Md) to count the number of HMVEC found per 10 random HPFs. Individual fibroblast and melanoma CM experiments were repeated in quadruplicate; fibroblast plus melanoma experiments were repeated in triplicate. Angiogenesis protein array. RayBio Human Angiogenesis Antibody Arrays (RayBiotech, Norcross, Ga) were employed to assay cell culture supernatant fluid from the CM experiments. Twenty different angiogenic growth factors were evaluated: angiogenin, epidermal growth factor (EGF), epithelial neutrophil--activating peptide-78 (ENA-78), basic fibroblast growth factor (bFGF), GRO, interferon gamma (IFN-c), insulinlike growth factor 1 (IGF-1), interleukin 6 (IL-6), IL-8, leptin, monocyte chemotactic protein-1 (MCP-1), platelet-derived growth factor BB (PDGF-BB), placenta growth factor (PIGF), rantes, transforming growth factor beta 1 (TGF b1), tissue inhibitor of metalloproteinase 1 (TIMP-1), tissue inhibitor of metalloproteinase 2 (TIMP-2), thrombopoietin, vascular endothelial growth factor (VEGF), and vascular endothelial growth factor D (VEGF-D). Specifically, we assayed for the changes in the released soluble factors by comparing the semiquantitative growth factor levels within culture media from plated collagen, fibroblasts, melanoma, or fibroblasts plus melanoma (Kodak 1D Image; Kodak, Rochester, NJ). Protein levels were quantified against internal controls in the array and compared with other samples as fold increases. The experiment was repeated in quadruplicate. Statistical analysis. Data are presented as means ± SEM. Microsoft Excel (Microsoft Corp, Redmond, Wash) was utilized for statistical analysis. Two-tail t-test was used for statistical comparisons of continuous data. ANOVA was utilized to

442 Goldstein et al

Surgery September 2005

Fig 1. In vitro human angiogenesis assay. vWF immunofluorescence of whole mount 3-D gels from representative study groups. Gels removed from underlying HMVEC monolayer after 5 days of incubation, fixed, stained with vWF (green), and counterstained with Hoescht (blue nuclei). The upper collagen layer overlying the HMVEC monolayer contains radial growth melanoma (RGP) cells (A); vertical growth melanoma cells (B); MM cells (C); Type I collagen without cells (D); RGP cell line + fibroblasts (E); VGP cell line + fibroblasts (F); MM cell line + fibroblasts (G); fibroblasts (H); melanoma cells embedded at 500,000 cells/mL (A-C); melanoma cells + fibroblasts in combination (250,000 cells/mL each) (E-G); fibroblasts embedded at 500,000 cells/mL (H).

compare multiple melanoma cell lines. P < .05 was considered statistically significant. RESULTS We employed a 3DCMA to study the ability of various melanoma cell lines to induce angiogenesis with and without the addition of fibroblasts. As detailed in Methods, after culturing monolayers of HMVECs on collagen-plated wells, media was removed, and a layer of liquid acellular collagen was applied, allowed to polymerize, and then covered with another liquid layer of collagen containing either melanoma cells, fibroblasts, or a combination of the two, as well as collagen-only controls. After complete polymerization, the 3-D gels were cultured in EGM medium, and HMVEC networks growing into the collagen from the underlying HMVEC monolayer were counted at day 5. Figure 1 shows representative fluorescent microphotographs for radial growth, vertical growth, and metastatic melanoma cells, alone or in combination with fibroblasts, as well as collagen-only and fibroblast-only controls. Overall, most melanoma cells alone had very limited to no angiogenic response (Fig 1, A-C). Plated acellular collagen did not support angiogenesis (Fig 1, D). The addition

of fibroblasts enabled significant angiogenesis in all of the melanoma cell lines tested (Fig 1, E-G and Figs 2-4). Fibroblasts alone were highly effective in inducing endothelial cell invasion, migration, and networking in the type I collagen gels (Fig 1, H). Fibroblasts cultured at 5 3 105/mL induced significantly more HMVEC capillary-like networking than those cultured at 2.5 3 105/mL (P < .01; Fig 2). Overall, fibroblasts at 500,000 cells/mL demonstrated significantly higher angiogenic response than 12/12 M alone (P < .01) and 9/12 M+F (1:1, each at 250,000 cells/mL; P < .05; Figs 2-4). Significant heterogeneity was demonstrated in the angiogenic response across the 12 melanoma lines studied, but that variability was statistically significant only when cultured in combination with fibroblasts (ANOVA; P < .01). Figure 2 shows the quantification of the angiogenesis response induced by the RGP melanoma lines. The 2 RGP cell lines studied (WM 3211 and WM 35) were unable to support any HMVEC networking when collagen embedded at a cell density of 250,000 cells/mL. Both RGP cell lines had increases in the HMVEC networks when embedded at 500,000/mL cell density, with WM 3211 having significantly more networking at the higher

Surgery Volume 138, Number 3

Fig 2. Angiogenesis induction by RGP melanoma cell lines. Two RGP melanoma cell lines (WM 3211 and WM 35) were embedded in type I collagen at 2.5 3 105/mL, 5 3 105/mL, alone or in combination with fibroblasts (each cell type at 2.5 3 105/mL) and overlaid on HMVEC monolayers. After 5 days of culture, 3-D gels were removed from the wells, fixed, and assessed by vWF immunofluorescence of gel whole mounts. Networks of HMVEC were quantified by counting 10 random HPFs. Data expressed as mean ± SEM. *WM 3211 at 500,000 cells/mL supported significantly more networks than when cultured at 250,000 cells/mL; P < .01. **Both WM 3211 and WM 35, when in combination with fibroblasts (each cell type at 2.5 3 105/mL), supported significantly more networks, compared with melanoma cells alone; P < .05. ***Fibroblasts (5 3 105/mL) supported significantly more networking than fibroblasts at half concentration (2.5 3 105/mL, black bar), WM 35 in combination with fibroblasts (each cell type at 2.5 3 105/ mL), and both WM 3211 and WM 35 cells alone; P < .01. Black bars indicate cells grown at 2.5 3 105/mL, gray bars indicate cells grown at 5 3 105/mL, and white bars indicate melanoma cells in combination with fibroblasts (each cell type at 2.5 3 105/mL). HMVEC, human microvascular endothelial cell; HPF, high-power field.

density (P < .01). The introduction of fibroblasts enabled a robust angiogenesis response in both RGP cell lines, with the combination of cells yielding significantly more HMVEC networks than melanoma cells alone (P < .05). However, the HMVEC networking supported by the RGP melanoma and fibroblast combinations (each at 250,000 cells/mL) was not significantly higher than fibroblast alone controls at 250,000 cells/mL. Additionally, for RGP line WM 35, the number of HMVEC networks induced by the combination of RGP melanoma cells and fibroblasts (each at 250,000 cells/mL) was significantly less than that of fibroblasts alone at 500,000 cells/mL, (P < .01). Angiogenesis induced by RGP line WM 3211 when combined with fibroblasts did not differ statistically from fibroblasts alone at 500,000 cells/mL.

Goldstein et al 443

Fig 3. Angiogenesis induction by VGP melanoma cell lines. Three VGP cell lines (WM 115, WM 3248, and WM 793) were embedded in type I collagen at 2.5 3 105/mL, 5 3 105/mL, alone or in combination with fibroblasts (each cell type at 2.5 3 105/mL) and overlaid on HMVEC monolayers. After 5 days of culture, 3-D gels were removed from the wells, fixed, and assessed by vWF immunofluorescence of gel whole mounts. Networks of HMVEC were quantified by counting 10 random HPFs. Data expressed as mean ± SEM. *All three VGP melanoma lines in combination with fibroblasts (each cell type at 2.5 3 105/mL) induced significantly more networking than those same melanoma lines cultured independently at 2.5 3 105/mL or 5 3 105/mL; P < .01. **WM 793 in combination with fibroblasts (each cell type at 2.5 3 105/mL) supported significantly more networking than fibroblasts alone (2.5 3 105/mL); P < .05. ***Fibroblasts at 5 3 105/mL supported significantly more networking than WM 115 or WM 3248, when in combination with fibroblasts (each cell type at 2.5 3 105/mL); P < .01. Black bars indicate cells grown at 2.5 3 105/mL, gray bars indicate cells grown at 5 3 105/ mL, and white bars indicate melanoma cells in combination with fibroblasts (each cell type at 2.5 3 105/mL). HMVEC, human microvascular endothelial cell; HPF, high-power field.

Figure 3 depicts the quantification of the angiogenesis response induced by the VGP melanoma lines. Similar to the intrinsic lack of angiogenic response observed with the RGP lines, none of the 3 VGP melanoma cell lines studied alone (WM 115, WM 3248, and WM 793) were able to induce HMVEC networks into the collagen. The addition of fibroblasts enabled strong angiogenesis in all 3 VGP cell lines, with the combination of cells yielding significantly more HMVEC networking than the melanoma cells alone (P < .01). In contrast to the RGP cell lines, the VGP cell line WM 793, when combined with fibroblasts (250,000 of each cell type/mL), supported significantly more HMVEC networking than fibroblasts alone at

444 Goldstein et al

Fig 4. Angiogenesis induction by MM cell lines. Seven MM cell lines (451 Lu, WM 1205, WM 1158, WM 9, WM 278, WM 1232, and WM 164) were embedded in type I collagen at 2.5 3 105/mL, 5 3 105/mL, alone or in combination with fibroblasts (each cell type at 2.5 3 105/mL) and overlaid on HMVEC monolayers. After 5 days of culture, 3-D gels were removed from the wells, fixed, and assessed by vWF immunofluorescence of gel whole mounts. Networks of HMVEC were quantified by counting 10 random HPFs. Data expressed as mean ± SEM. *All seven M melanoma lines in combination with fibroblasts (each cell type at 2.5 3 105/mL) induced significantly more networking than those same melanoma lines alone; P < .01. **PWM 1158 in combination with fibroblasts (each cell type at 2.5 3 105/mL) supported significantly more networking than fibroblasts alone (2.5 3 105/mL); P < .01. ***Fibroblasts at 5 3 105/mL supported significantly more networking than 451 Lu, WM 1205, WM 9, WM 278, WM 1232, and WM 164, when in combination with fibroblasts (each cell type at 2.5 3 105/mL); P < .005. Black bars indicate cells grown at 2.5 3 105/mL, gray bars indicate cells grown at 5 3 105/mL, and white bars indicate melanoma cells in combination with fibroblasts (each cell type at 2.5 3 105/mL). HMVEC, human microvascular endothelial cell; HPF, high-power field.

250,000 cells/mL and F+M combinations with the use of the other 2 VGP lines (WM 115 and WM 3248; P < .05). Similar to the observations for RGP lines, the number of HMVEC networks supported by the combination of VGP melanoma cells and fibroblasts (each at 250,000 cells/mL) was significantly less than fibroblasts alone at 500,000 cells/mL for 2 of the 3 VGP cell lines (WM 115 and WM 3248; P < .01), the exception being the strongly angiogenic VGP WM 793. Figure 4 illustrates the quantification of the angiogenesis response induced by the 7 metastatic melanoma (MM) cell lines studied (451 Lu, WM 1205, WM 1158, WM 9, WM 278, WM 1232, and WM 164). Similar to the findings in RGP and VGP lines, none of the MM cells alone was able to induce significant HMVEC networking. Again, the addition

Surgery September 2005

of fibroblasts rescued the angiogenic response of all 7 MM cell lines, with the combination of cells yielding significantly more HMVEC networking than the melanoma cells alone (P < .01). Unlike the RGP lines and 2/3 of the tested VGP lines, but similar to the VGP line WM 793, MM cell line WM 1158, when combined with fibroblasts (250,000 of each cell type/mL), supported significantly more HMVEC networking than fibroblasts alone (250,000 cells/mL; P < .01). However, that synergism was not bserved in the other 6 MM lines studied. Again, as in the RGP line WM 35 and the VGP lines WM 115 and WM 3248, the number of HMVEC networks supported by the combination of MM melanoma cells and fibroblasts (each at 250,000 cells/mL) was significantly less than fibroblasts alone at 500,000 cells/mL, for 6 of the 7 MM cell lines (451Lu, WM 1205, WM 9, WM 278, WM 1232, and WM 164; P < .05). This was not the case in only 1 metastatic cell line (WM 1158), with the angiogenesis supported by 500,000 fibroblasts/ mL being equivalent to that supported by 250,000 MM/mL + 250,000 fibroblasts/mL. It was this cell line that we selected for further study of potential secreted angiogenic factors. To determine whether secreted angiogenic factors influence EC invasion/migration in our model, we tested the effect of CM derived from melanoma cells, fibroblasts, or a combination of the 2 on HMVEC monolayers overlaid with acellular 3-D collagen gels. Figure 5 depicts the quantification of the numbers of HMVEC within the collagen gels after 24 hours of culture with each of the various CM. A significantly greater number of HMVEC invaded/migrated into the type I collagen with exposure to the fibroblast CM, compared with both melanoma CM and (M+F) CM (P < .01). To further characterize the effects seen in the CM experiments, we utilized a RayBio Human Angiogenesis Antibody Array to assay cell culture supernatant fluid. This array identified 3 factors exhibiting consistent and reproducible elevations in our samples. Figure 6, A shows the comparison of fibroblast-derived media versus a collagen only control. In media derived from fibroblasts, IL-8, MCP-1, and TIMP-2 were identified as significantly increased versus collagen (33.2-, 64.8-, and 23.6fold, respectively; P < .05). Figure 6, B shows the concentration of these factors (IL-8, MCP-1, and TIMP-2) when comparing media derived from melanoma cells only versus media derived from fibroblasts only. Melanoma cells were not found to produce a significantly different amount of IL-8 (1.2-fold; P = .41), or MCP1 (1.4-fold; P = .31) than fibroblasts. Melanoma

Surgery Volume 138, Number 3

Goldstein et al 445

cells were, however, found to produce significantly less TIMP-2 than fibroblasts (0.29-fold; P < .05). Figure 6, C depicts the presence of these same 3 factors (IL-8, MCP-1, and TIMP-2) in media collected from melanoma cells grown in combination with fibroblasts. In this case, all 3 factors were found to be significantly increased in the combination media, compared with media from fibroblasts grown alone (6.1-, 2.3-, and 2.2-fold, respectively; P < .05). When comparing the combination media with melanoma cells alone, all 3 factors were increased; however, significance was reached for IL8 (4.2-fold; P < .01) and TIMP-2 (8.3-fold; P < .01), but not for MCP-1 (1.6-fold; P = .12). DISCUSSION In this study we report the novel findings that human melanoma cells require normal skin-derived fibroblasts to effectively induce angiogenesis in type I collagen. We further report evidence that this unexpected advantage that fibroblasts have over melanoma cells in their ability to induce angiogenesis may be due to increased collagen invasion/migration signals of fibroblasts on microvascular endothelial cells. The angiogenic factors differentially detected within CM, including IL-8, MCP-1, and TIMP-2, imply that soluble or paracrine factors are released from fibroblasts grown in combination with melanoma cells. Although it has been known previously that malignant cells can recruit stromal cells to aid in tumor growth,7,8,12 an essential requirement for fibroblast assistance in melanoma cell angiogenesis has never been reported before. These in vitro data suggest that the ability to induce angiogenesis is not necessarily an inherent feature of the melanoma cell, but perhaps instead that these neoplastic cells recruit stromal cells such as fibroblasts, which in turn induce new vessel growth. The consistency of the findings across 12 different human melanoma lines spanning the 3 key stages in melanoma progression lends strength to these conclusions. Moreover, the significant variability in the extent of angiogenesis induced by varying the melanoma line indicates that the intrinsic biology of each melanoma cell type also significantly contributes to angiogenesis, as one would expect on the basis of prior reports. Finally, the ability of a VGP and MM line (and the absence of an RGP line), when combined with fibroblasts to exceed the ability of an equal number of fibroblasts to induce HMVEC networking, points toward the acquisition of this cooperative effect as melanoma cells become more malignant.

Fig 5. Effect of fibroblast versus melanoma conditioned media on HMVEC invasion/migration into collagen. HMVEC invasion/migration into type I collagen 3-D gels was measured in response to exposure to conditioned media (CM) from MM cells (WM1158), or fibroblasts, alone or in combination. The HMVEC monolayer was overlaid with an acellular collagen 3-D gel and cultured for 48 hours with various CM. Gels were then fixed and HMVEC nuclei within the gels were stained with Hoechst and quantified by software-assisted analysis in 10 random HPFs. Data expressed as mean ± SEM. *CM derived from fibroblasts caused significantly increased HMVEC invasion/migration, compared with CM from melanoma cells or the combination of melanoma cells and fibroblasts; P < .01. **CM derived from M+F caused significantly less HMVEC invasion/migration, compared with CM from melanoma or fibroblasts alone; P < .01. HMVEC, human microvascular endothelial cell; HPF, high-power field.

We have also shown that normal human skin-derived HMVEC will migrate more readily in response to media conditioned from cultured normal human skin--derived fibroblasts than from media obtained from malignant MM cells. While both types of CM were capable of eliciting a migration response, the chemotactic response attributable to soluble fibroblast factors is significantly stronger than the factors secreted by cultured melanoma cells. The combination of our findings lead us to conclude that fibroblasts have inherent angiogenic potential, that this ability is stronger than the angiogenic ability of our cultured melanoma cells, and most notably that, when present together, a synergistic phenomenon can be seen with an angiogenic response significantly larger than either cell type could induce independently, particularly with some VGP and MM cells. The CM results implicate soluble fibroblastderived factors as playing a dominant role in the

446 Goldstein et al

Surgery September 2005

Fig 6. Angiogenesis antibody array detection of conditioned media growth factors. Conditioned media derived from fibroblasts, melanoma cells, or a combination of the two, evaluated with RayBio Human Angiogenesis Antibody Arrays. Protein levels were quantified against an internal control on the array and then graphed as a fold-increase over the comparison. Data expressed as mean fold-increase ± SEM. A, Fold increase in soluble factors in media derived from fibroblasts versus collagen control. *Fibroblasts produced significantly more IL-8, MCP-1, and TIMP-2 than collagen controls; P < .05. B, Fold increase in soluble factors in media derived from melanoma cells versus fibroblasts. *Melanoma cells produced significantly less TIMP-2 than fibroblasts; P < .05. C, Fold increase in soluble factors in media derived from a combination of melanoma cells and fibroblasts (M+F) versus fibroblasts alone, or melanoma cells alone. M+F versus fibroblasts alone represented by black bars; M+F versus melanoma cells alone represented by gray bars. *Media from M+F contained significantly more IL-8, MCP-1, and TIMP-2 than media from fibroblasts alone; P < .05. **Media from M+F contained significantly more IL-8, and TIMP-2 than media from melanoma cells alone; P < .01. IL-8, Interleukin 8; MCP-1, monocyte chemotactic protein1; TIMP-2, tissue inhibitor of metalloproteinase-2.

invasion/migration of HMVEC into type I collagen. The 3DCMA findings also indicate that direct cell-cell interactions are probably playing a role in the observed endothelial cell migration and networking in the collagen angiogenesis

experiments. We detected an increase in 3 factors in CM from fibroblasts and melanoma cells grown together: IL-8, MCP-1, and TIMP-2; moreover, all 3 factors were significantly increased, compared with media from fibroblasts alone, and 2 of the 3 were

Surgery Volume 138, Number 3

significantly increased, compared with media from melanoma cells alone. IL-8 and MCP-1 are known proangiogenic growth factors. IL-8 was identified previously as an upregulated gene during capillary morphogenesis in an in vitro endothelial cell angiogenesis system.22 Endothelial progenitor cells, bone marrow stem cells involved in postnatal vasculogenesis, also have been shown to secrete IL-8, exerting a strong paracrine mitogenic effect on mature endothelial cells.23 MCP-1 has been shown to have varying roles in tumor biology as well as a direct angiogenic effect on endothelial cells, which have been shown to express the receptor for MCP-1, CCR2.24,25 The significantly decreased migration observed with M+F CM, despite the higher levels of soluble proangiogenic factors and the robust response seen in the 3DCMA, would be expected, given the observed overproduction of TIMP-2 in M+F co-culture. TIMP-2 is known to inhibit HMVEC invasion/migration in type I collagen.26,27 Our data indicate that TIMP-2 release increases significantly in media from melanoma cells co-cultured with fibroblasts, compared with either melanoma or fibroblast cells alone. TIMP-2 is a known inhibitor of angiogenesis and also has been shown to decrease EC proliferation in vitro.26,27 Therefore, the observed decrease in EC migration and invasion into type I collagen attributable to M+F CM may be related to the increased presence of the identified anti-angiogenic factor TIMP-2 in the absence of the equally necessary cell-cell interactions observed in the 3DCMA. Physiologically, it would be expected that factors that initially attract EC migration (such as IL-8, MCP-1) would need to be aided by factors that then halt this migration (TIMP-2), in favor of differentiation and capillary formation within the melanoma and stromal cell growth. The utilization of stromal cell mechanisms for angiogenesis also has been shown to exist in other malignant cells including mammary carcinoma, and hepatocellular carcinoma.28 Significant amounts of proangiogenic growth factors have been seen specifically with tumor–associated fibroblasts and immune cells in tumor specimens.29,30 Dong et al16 recently showed, in an attempt to explain the mechanisms of tumor escape after VEGF inactivation, that the recruitment of stromal cells was essential for the angiogenic and tumorigenic characteristics of VEGF-null fibrosarcomas. In addition, they began to unravel the mechanisms by which this action occurs, singling out platelet-derived growth factor-AA (PDGF-AA) and its receptor (PDGFR-a) as the mechanism by which VEGF-null

Goldstein et al 447

fibrosarcomas recruit fibroblasts, which then produce pro-angiogenic VEGF in the tumor stroma. Human microvascular endothelial cells are also known to express receptors for PDGF.31 Other tumor types also have been found to recruit stromal cells for angiogenesis via plateletderived growth factor (PDGF). Guo et al32 showed human gliomas to recruit pericytes and enhance angiogenesis via PDGF-B. In melanoma cells known to produce PDGF, the PDGFR-a has been implicated in the melanoma cell–stimulated fibroblast production of tumor stroma.33 Human melanoma cells, known to not produce PDGF, when transfected with a vector to over express PDGF-B gained the ability to induce a stable connective tissue stroma and avoided tumor necrosis.34 Nontumorigenic immortilized human keratinocytes when transfected with PDGF-B complementary DNA induced a significant stromal and angiogenic response in vivo and caused proliferation of co-cultured fibroblasts in vitro.5 In a mouse melanoma tumor model, Furuhashi et al35 showed paracrine PDGF production to increase tumor vessel pericyte recruitment and increased tumor growth, but were unable to document an increase in actual vessel density. We did not observe a consistent, reproducible change in PDGF levels as measured by our angiogenesis protein array. The role PDGF may play in melanoma angiogenesis and our collagen model will require further study. Differences also have been noted in the angiogenic ability of fibroblasts derived from tumors versus those from benign tissues.36 It has been shown that fibroblasts from normal tissues can have an inhibitory, antitumor effect on preneoplastic breast cells, while fibroblasts derived from tumor confer morphogenic and mitogenic induction to the breast cells.37 A limitation of our study is the use of normal human fibroblasts, as fibroblasts isolated from actual melanoma specimens may exhibit different characteristics with regard to angiogenesis. The difference in behavior between normal fibroblasts and those derived from melanoma specimens will require further studies. In addition to paracrine signals, transformed melanocytes are known to establish new gap junctional activity with fibroblasts attributable to a shift in cadherin expression.13 This communication between joined melanoma cells and fibroblasts was found to be bidirectional. The importance of this cell-cell communication with regard to the induction of angiogenesis in melanomas is unknown, but again lends credence to the idea that angiogenesis involves both paracrine factors as well as cell-cell interactions.

448 Goldstein et al

A clear interplay exists between melanoma cells and the surrounding stromal cells. In this report, we show that the contribution from fibroblasts is an essential component of the melanoma angiogenic response. In vivo, melanoma tumor vasculature is leaky, convoluted, and can be composed of a mix of tumor cells as well as EC.38,39 The stimuli of ischemic growing malignant rests causes the induction of vasculature by more than 1 mechanism, even if it results in a less than optimal vascular bed. Our findings combined with prior reports suggest melanoma cells will make use of any mechanism they can acquire, be it a secreted factor, a cell-cell interaction, or the subversion of local stromal cells, to achieve angiogenesis. It is likely that the variable biology of each melanoma line leads to unique methods by which these malignant cells subvert their surroundings. Our identification of certain soluble factors involved in melanoma angiogenesis is limited by both the inherent biology of the cell lines we studied, as well as the focused set of angiogenic factors assayed by the chosen angiogenesis protein array. Further study will allow the identification of other factors involved in both the paracrine signalling and the cell-cell interactions involved in melanoma angiogenesis. Focused experiments will further resolve the role played in melanoma angiogenesis by other known proangiogenic cytokines, such as PDGF, that were not specifically identified in our current study. For the first time, we demonstrate the utility of normal skin--derived fibroblasts to malignant melanoma cells in helping to induce a neovasculature in our in vitro collagen assay. In addition, we have begun to elucidate some of the paracrine factors by which these cells communicate, including TIMP-2, MCP-1, and IL-8. This communication is likely, but it is only 1 mechanism by which malignant melanomas in vivo can achieve angiogenesis; significant work remains to elucidate the complex interactions between these cells and their environment. However, these novel and unexpected findings point to a new regulatory target in melanoma angiogenesis: the fibroblast. Contrary to the notion that melanoma cells alone support angiogenesis, in our model, even MM cells required fibroblasts to support HMVEC invasion, migration, and capillary-like network formation. REFERENCES 1. Folkman J. Angiogenesis in cancer, vascular, rheumatoid and other disease. Nat Med 1995;1:27-31. 2. Carmeliet P, Jain RK. Angiogenesis in cancer and other diseases. Nature 2000;407:249-57.

Surgery September 2005

3. Hanahan D, Folkman J. Patterns and emerging mechanisms of the angiogenic switch during tumorigenesis. Cell 1996; 86:353-64. 4. Hanahan D, Weinberg RA. The hallmarks of cancer. Cell 2000;100:57-70. 5. Skobe M, Fusenig NE. Tumorigenic conversion of immortal human keratinocytes through stromal cell activation. Proc Natl Acad Sci U S A 1998;95:1050-5. 6. Kunz-Schughart LA, Knuechel R. Tumor-associated fibroblasts (part I): Active stromal participants in tumor development and progression? Histol Histopathol 2002;17:599-621. 7. Ruiter D, Bogenrieder T, Elder D, et al. Melanoma-stroma interactions: structural and functional aspects. Lancet Oncol 2002;3:35-43. 8. Labrousse AL, Ntayi C, Hornebeck W, et al. Stromal reaction in cutaneous melanoma. Crit Rev Oncol Hematol 2004;49:269-75. 9. Srivastava A, Ralhan R, Kaur J. Angiogenesis in cutaneous melanoma: pathogenesis and clinical implications. Microsc Res Tech 2003;60:208-24. 10. Lazar-Molnar E, Hegyesi H, Toth S, et al. Autocrine and paracrine regulation by cytokines and growth factors in melanoma. Cytokine 2000;12:547-54. 11. Streit M, Detmar M. Angiogenesis, lymphangiogenesis, and melanoma metastasis. Oncogene 2003;22:3172-9. 12. Li G, Satyamoorthy K, Meier F, et al. Function and regulation of melanoma-stromal fibroblast interactions: when seeds meet soil. Oncogene 2003;22:3162-71. 13. Hsu M, Andl T, Li G, et al. Cadherin repertoire determines partner-specific gap junctional communication during melanoma progression. J Cell Sci 2000;113(Pt 9):1535-42. 14. Park CC, Bissell MJ, Barcellos-Hoff MH. The influence of the microenvironment on the malignant phenotype. Mol Med Today 2000;6:324-9. 15. Velazquez OC, Snyder R, Liu ZJ, et al. Fibroblast-dependent differentiation of human microvascular endothelial cells into capillary-like 3-dimensional networks. Faseb J 2002; 16:1316-8. 16. Dong J, Grunstein J, Tejada M, et al. VEGF-null cells require PDGFR alpha signaling-mediated stromal fibroblast recruitment for tumorigenesis. Embo J 2004;23:2800-10. 17. Montesano R, Orci L, Vassalli P. In vitro rapid organization of endothelial cells into capillary-like networks is promoted by collagen matrices. J Cell Biol 1983;97(5 Pt 1):1648-52. 18. Liu ZJ, Snyder R, Soma A, et al. VEGF-A and alphaVbeta3 integrin synergistically rescue angiogenesis via N-Ras and PI3-K signaling in human microvascular endothelial cells. Faseb J 2003;17:1931-3. 19. Meier F, Nesbit M, Hsu MY, et al. Human melanoma progression in skin reconstructs: biological significance of bFGF. Am J Pathol 2000;156:193-200. 20. Imcke E, Ruszczak Z, Mayer-da Silva A, et al. CE. Cultivation of human dermal microvascular endothelial cells in vitro: immunocytochemical and ultrastructural characterization and effect of treatment with three synthetic retinoids. Arch Dermatol Res 1991;283:149-57. 21. 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. 22. Sun XT, Zhang MY, Shu C, et al. Differential gene expression during capillary morphogenesis in a microcarrierbased three-dimensional in vitro model of angiogenesis with focus on chemokines and chemokine receptors. World J Gastroenterol 2005;11:2283-90.

Goldstein et al 449

Surgery Volume 138, Number 3

23. He T, Peterson TE, Katusic ZS. Paracrine mitogenic effect of human endothelial progenitor cells - Role of interleukin-8. Am J Physiol Heart Circ Physiol 2005. 24. Conti I, Rollins BJ. CCL2 (monocyte chemoattractant protein-1) and cancer. Semin Cancer Biol 2004;14:149-54. 25. Salcedo R, Ponce ML, Young HA, et al. Human endothelial cells express CCR2 and respond to MCP-1: direct role of MCP-1 in angiogenesis and tumor progression. Blood 2000;96:34-40. 26. Seo DW, Li H, Guedez L, et al. TIMP-2 mediated inhibition of angiogenesis: an MMP-independent mechanism. Cell 2003;114:171-80. 27. Stetler-Stevenson WG, Seo DW. TIMP-2: an endogenous inhibitor of angiogenesis. Trends Mol Med 2005;11:97-103. 28. Fukumura D, Xavier R, Sugiura T, et al. Tumor induction of VEGF promoter activity in stromal cells. Cell 1998;94: 715-25. 29. Hlatky L, Tsionou C, Hahnfeldt P, et al. Mammary fibroblasts may influence breast tumor angiogenesis via hypoxiainduced vascular endothelial growth factor up-regulation and protein expression. Cancer Res 1994;54:6083-6. 30. Lewis JS, Landers RJ, Underwood JC, et al. Expression of vascular endothelial growth factor by macrophages is upregulated in poorly vascularized areas of breast carcinomas. J Pathol 2000;192:150-8. 31. Beitz JG, Kim IS, Calabresi P, et al. Human microvascular endothelial cells express receptors for platelet-derived growth factor. Proc Natl Acad Sci U S A 1991;88:2021-5. 32. Guo P, Hu B, Gu W, et al. Platelet-derived growth factor-B enhances glioma angiogenesis by stimulating vascular

33.

34.

35.

36.

37.

38.

39.

endothelial growth factor expression in tumor endothelia and by promoting pericyte recruitment. Am J Pathol 2003; 162:1083-93. Godden JL, Edward M, MacKie RM. Melanoma cell-derived factor stimulation of fibroblast glycosaminoglycan synthesis–the role of platelet-derived growth factor. Eur J Cancer 1999;35:473-80. Forsberg K, Valyi-Nagy I, Heldin CH, et al. Platelet-derived growth factor (PDGF) in oncogenesis: development of a vascular connective tissue stroma in xenotransplanted human melanoma producing PDGF-BB. Proc Natl Acad Sci U S A 1993;90:393-7. Furuhashi M, Sjoblom T, Abramsson A, et al. Platelet-derived growth factor production by B16 melanoma cells leads to increased pericyte abundance in tumors and an associated increase in tumor growth rate. Cancer Res 2004;64: 2725-33. Kunz-Schughart LA, Knuechel R. Tumor-associated fibroblasts (part II): Functional impact on tumor tissue. Histol Histopathol 2002;17:623-37. Shekhar MP, Werdell J, Santner SJ, et al. Breast stroma plays a dominant regulatory role in breast epithelial growth and differentiation: implications for tumor development and progression. Cancer Res 2001;61:1320-6. Papetti M, Herman IM. Mechanisms of normal and tumorderived angiogenesis. Am J Physiol Cell Physiol 2002;282: C947-70. Maniotis AJ, Folberg R, Hess A, et al. Vascular channel formation by human melanoma cells in vivo and in vitro: vasculogenic mimicry. Am J Pathol 1999;155:739-52.

Editors’ corner SURGERY receives many submissions in the category of case reports, very few of which will be accepted. Our principal benchmark for consideration of publication is that the case or its management provide novel and timely information, as well as insight into a surgical disorder. More precisely, this requirement does not include such criteria as the ‘‘biggest,’’ ‘‘most difficult,’’ ‘‘unusual,’’ or even ‘‘the first ever described.’’ Occasionally

the Editors may suggest to the author that the case contains images of possible interest for our Images in Surgery section. Case reports must be short and focused—not just an excuse for another review of the literature. Andrew L. Warshaw, MD Michael G. Sarr, MD Editors-in-Chief