Two Different PDGF β-Receptor Cohorts in Human Pericytes Mediate Distinct Biological Endpoints

Abstract

How activation of a specific growth factor receptor selectively results in either cell proliferation or cytoskeletal reorganization is of central importance to the field of pathophysiology. In this study, we report on a novel mechanism that explains how this process is accomplished. Our current investigation demonstrates that soluble platelet derived growth factor- (PDGF)-BB activates a cohort of PDGF-β receptors primarily confined to the lipid raft component of the cell membrane, specifically caveolae. In contrast, cell-bound PDGF-BB delivered via cell–cell contact results in activation and the subsequent up-regulation of a cohort of PDGF β-receptors primarily confined to the non-lipid raft component of the cell membrane. Individual activation of these two receptor cohorts results in distinct biological endpoints, cytoskeletal reorganization or cell proliferation. Mechanistically, our evidence suggests that PDGF-BB-bearing cells preferentially stimulate the non-lipid raft receptor cohort through interleukin 1β-mediated inhibition of the lipid raft cohort of receptors, leaving the non-raft receptor cohort operational and preferentially stimulated. In human skin injected with PDGF-BB and in tissue reparative processes PDGF β-receptors colocalize with the caveolae/lipid raft marker caveolin-1. In contrast, in human skin injected with PDGF-BB-bearing tumor cells and in colorectal adenocarcinoma, activated PDGF β-receptors do not colocalize with caveolin-1. Thus, growth factor receptors are segregated into specific cell membrane compartments that are preferentially activated through different mechanisms of ligand delivery, resulting in distinct biological endpoints.

Lipid rafts are cellular membrane domains that contain high concentrations of cholesterol and sphingolipids. These domains include the flat and related vesicular structures referred to as caveolae. Caveolae, which are formed by the macromolecular oligomerization of the 22-kDa caveolin protein are enriched in a number of vital signal transduction molecules, and contain smaller cohorts of many others.1,2,3,4,5,6 Furthermore, caveolin itself directly binds and/or regulates the activities of a number of these signaling molecules.1 With specific regards to the platelet derived growth factor (PDGF) signaling axis, PDGF-induced signaling occurs in caveolae of many mesenchymal cells,5,6,7,8 and PDGF-receptors are functional in isolated caveolae.8 Based on the apparent signaling events occurring in lipid rafts, and the abundance of molecules involved in multiple signaling pathways, it is inferred that lipid rafts are important loci for signal amplification and cross talk between signaling pathways.1,2,5,6,7,8 Recently emerging evidence shows that lipid rafts also have important specific roles in regulating the activity of cytoskeleton-regulating GTPases, in cytoskeletal organization, in the formation of cell extensions, and in cell motility.9

The PDGF-B chain contains a retention motif that mediates binding to heparan sulfate proteoglycans on cell surfaces.10 This affords PDGF-BB-producing cells alternate modes of ligand delivery to PDGF β-receptor bearing cells, through heterotypic cell-to-cell contacts or as a secreted soluble ligand.11,12 In mesenchymal cell–tumor cell co-cultures, activation of PDGF β-receptors is a consequence of cell–cell contacts, and is not accomplished via soluble PDGF-BB.13 The biological consequences of cell–cell versus secreted ligand remain unknown. Given the central role of PDGF β-receptor activation in pericyte biology during embryogenesis and reactive conditions in the adult organism,14,15,16,17,18,19,20 we chose to investigate the role of caveolae in PDGF β-receptor signaling in primary human pericytes. Activation of PDGF β-receptors in mesenchymal cells leads to several biological endpoints, eg, proliferation and reorganization of the actin cytoskeleton.20 How cells are able to orchestrate signal transduction events leading to different biological endpoints, in response to stimulation by a specific ligand, is not known. Here we demonstrate one mechanism by which context-specific ligand stimulation of a growth factor receptor results in distinct biological endpoints.

Materials and Methods

Antibodies and Other Reagents

The PDGF ß-receptor (PDGFR-B)-specific monoclonal antibody (mAb) PDGFR-B2, which was raised against phosphorylated PDGF ß-receptors, was used at a concentration of 1 μg/ml. At this concentration PDGFR-B2 only detects phosphorylated ie, clusters of activated PDGF β-receptors in vitro.21 The anti-PDGF B-chain mAb clone PGF 007 was from Mochida Co. (Tokyo, Japan). The anti-caveolin-1 (Clone 2234) mAb, anti-caveolin-1 (Clone N-20) mAb, anti-caveolin-2, anti-Fyn mAb, anti-Shc pAb, anti-dynamin mAb, anti-Ras mAb, anti-c-Raf mAb, and the anti-PDGF β-receptor mAb (clone 958) were from Transduction Laboratories BD (San Jose, CA). The anti-pan cytokeratin mAb, anti-epithelial specific antigen mAb used for magnetic bead separation, anti-β-actin mAb, and anti-vinculin mAb were from Sigma-Aldrich (Stockholm, Sweden). The anti-Giα3 polyclonal antibody (pAb), anti-Giα1 pAb, anti-Gsα pAb, and anti-Gβ pAb were from Calbiochem (San Diego, CA). The anti-Src pAb was from Upstate Biotechnologies (Lake Placid, NY). The anti-CD31 mAb, and mouse and rabbit IgG were from DAKO (Glostrup, Denmark). The anti High Molecular Weight-Melanoma Associated Antigen (HMW-MAA) was from Sanbios (Netherlands). The 4G10 anti-phosphotyrosine mAb was a gift of Dr. Brian Drukar (Oregon Health Sciences Center, Portland, OR).

Human Tumor Cell Lines and Isolation of Human Microvascular Pericytes

The MDA-MB-435 human breast carcinoma cell line22 was from Lombardi breast cancer repository (Washington, DC). The SCC-13 human cutaneous squamous carcinoma cell line23,24,25 was kindly provided by Dr. Karen Yee (Harvard Medical School, Boston). Isolation of human microvascular pericytes from neonatal foreskin was performed as described previously.26

Immunolabeling, Confocal Microscopy, and Transmission Electron Microscopy

Double and triple immunofluorescence staining of cells or of 6-μm tissue sections, immuno-enzymatic labeling of 6-μm tissue sections, confocal imaging, quantification of colocalization,14,26,27 and transmission electron microscopy5 were performed as previously described. In labeling studies staining with non-specific IgGs were performed as controls. A BioRad MRC 1024 confocal microscope was used to obtain optical sections, which were analyzed individually and as compiled images.

Co-Culture Experiments and Soluble PDGF-BB Stimulation in Vitro

Cytoskeletal Reorganization Experiments

For analysis of cytoskeletal reorganization, pericytes were seeded on collagen type I-coated coverslips and grown to near confluence in RPMI, containing 20% heat inactivated fetal calf serum (FCS) in 24-well plates. Cells were serum-starved with RPMI medium containing 0.1% FCS for 24 hours before co-culturing experiments. Tumor cell cultures were washed in PBS once and trypsinized for 3 minutes. Tumor cells (2 × 10⁴/well) were seeded onto the near confluent pericyte layers for different time periods (1, 3, 6, 12, and 24 hours) and incubated with or without soluble PDGF-BB in RPMI medium containing 0.1% FCS in the presence or absence of different biological modulators. These modulators were added 2 hours before the addition of PDGF-BB. Alternatively, the individual cell-types were cultured alone and serum-starved as outlined above. These cells were then incubated with or without soluble PDGF-BB for different time periods (15 minutes, and 0.5, 1, 3, 6, 12, and 24 hours) in the presence or absence of different biological modulators. These modulators were added 2 hours before the addition of PDGF-BB. Cells were then washed and fixed in 4% paraformaldehyde for immunofluorescence analysis of cytoskeletal reorganization at the different time points. For this purpose, actin and cell nuclei were visualized by staining with Bodipy Phallacidin from Molecular Probes (Eugene, Oregon) and 4,6-diamidino-2-phenylindole (Sigma), respectively. Semi-quantification was performed as follows: For each condition 5 fields of vision (×200) were randomly chosen. As an indicator of cytoskeletal reorganization the number of cells with circular/membrane ruffles for the 0.5-hour time point, and the number of cells exhibiting cellular extensions, for the 6-hour time point, were manually counted and data presented as percentage of total number of cells counted. Stainings were performed in triplicate for each condition in each experiment. Each experiment was repeated three times.

Proliferation Measurements

For measurements of proliferation, individual cell types or co-cultures were cultured as outlined above, serum starved in RPMI containing 0.1% FCS for 24 hours and then incubated with biological modulators for 24 hours. ³H-thymidine from Amersham (Uppsala, Sweden) was added during the final 4 hours of culture at a concentration of 2 μCi/ml. To assess the individual proliferation rates of pericytes and tumor cells in co-cultures, these two cell types required separation. Co-cultures were incubated for 15 minutes with M450 Dynal magnetic beads from dynal (Oslo, Norway) coated with anti-mouse IgG coupled to a mAb recognizing epithelial specific antigen, whereupon they specifically bound to the tumor epithelial cells. Co-cultures were then briefly trypsinized, beads bound to tumor cells were magnetically separated from the remaining medium containing pericytes and then washed three times in PBS. Cells were then precipitated in 10% trichloroacetic acid, washed in PBS, and solubilized in 0.2 mol/L NaOH containing 2% SDS. Samples were suspended in aqueous scintillation buffer and counts per minute were measured in a scintillation counter. Cross-contamination between the two different cell-types did not exceed 5% and was assessed by preloading one cell type with ³H-thymidine before co-culturing and measuring the activity in the other cell type after magnetic bead separation. A minimum of triplicates was performed for each condition in each experiment. Each experiment was repeated three times.

Metabolic Labeling, Successive Detergent Extraction Method, Immunoprecipitation and Immunoblot Analysis

Pericytes (1 × 10⁶) were seeded in 100-mm tissue culture dishes and grown in RPMI containing 20% FCS to near confluence and then serum starved for 12 hours in RPMI containing 0.1% FCS. Tumor cells were then added on top of the pericyte cultures at a density of 1.5 × 10⁶ cells for different time periods (3, 6, 12, and 24 hours). Alternatively, pericytes were stimulated with 20 ng/ml soluble PDGF-BB for different time periods (15 minutes, and 3, 6, 12, and 24 hours). Cells were subjected to successive detergent extraction, followed by immunoblotting to allow separate analysis of the non-lipid raft versus the lipid raft cell membrane constituents as previously described.6,28,29,30,31,32,33,34,35,36 Briefly, cells were scraped off in PBS + 2 mmol/L EDTA and centrifuged 400 × g for 5 minutes. The cell pellets were resuspended in buffer A [25 mmol/L 2-(N-Morpholino) ethanesulfonic acid (pH 6.5), 150 mmol/L NaCl, 2 mmol/L Na₃VO₄, complete protease inhibitor cocktail (Roche)]. To this an equal volume of the same buffer with 2% Triton X-100 (final concentration of Triton X-100 was 1%) was added, and lysates were incubated on ice for 0.5 hours. Insoluble fractions were pelleted in a microcentrifuge (10,000 × g) for 20 minutes at 4°C. Supernatants were removed (soluble fraction) and the insoluble pellet was resuspended in buffer B [1% Triton X- 100, 10 mmol/L Tris-HCl (pH 7.6), 500 mmol/L NaCl, 2 mmol/L Na₃VO₄, complete protease inhibitor cocktail, 20 mg/ml and octylglucoside (Sigma)] for 0.5 hours on ice. Debris was pelleted in a microcentrifuge (10,000 × g) for 20 minutes at 4°C, and the supernatant was removed (raft fraction). In metabolic labeling experiments, cultures were incubated in methionine/cysteine-free RPMI (Sigma) containing 0.1% FBS for 3 hours. Thereafter 100 μmCi/ml ³⁵S-protein labeling mix from PerkinElmer (NEG-072, Boston, MA) containing labeled methionine and cysteine was added to the culture for 6 to 24 hours. In the metabolic labeling experiments PDGF β-receptors were immunoprecipitated using the PDGF ß-receptor-specific pAb. Lysates and immunoprecipitates from the raft and non-raft membrane fractions were subjected to SDS-polyacrylamide gel electrophoresis (20 μg protein/lane for lysates), transferred, and then immunoblotted as previously described.5,13,32,37 Protein concentrations in membrane fractions were quantified by MicroBCA assay from Pierce (Rockford, IL) using a bovine serum albumin standard. Ponceau S staining of membranes was used to ensure equal loading where appropriate.

Autoradiography

Blotting membranes from metabolic labeling experiments were incubated with Amplify fluorographic reagent (Amersham) for 30 minutes. Membranes were then exposed to Imaging Plates from Fuji Film (Stockholm, Sweden) for 15 to 24 hours, before development in a BAS-2500 Bio-imaging Analyzer (Fuji).

Quantification of Interleukin 1β and PDGF-BB

Mono-cultures and co-cultures were established and incubated for 24 hours in RPMI containing 0.1% FBS as outlined above. Conditioned medium was collected and frozen pending further analysis. The remaining cells were washed twice in PBS and scraped off in PBS containing 2 mmol/L EDTA and centrifuged at 400 × g for 5 minutes. To establish the amount of cell associated interleukin (IL)-1β, pellets were lysed [50 mmol/L Tris-HCl pH 7.5, 0.5 mol/L NaCl, 2 mmol/L EGTA, 1% NP40, 0.25% sodium deoxycholate] for 30 minutes. To establish the amount of PDGF-BB, cultures were either lysed in a lysis buffer (as described above), but with higher NaCl concentration (0.75 mol/L). Samples were centrifuged (10,000 × g) and supernatants were frozen pending further analysis. IL-1β and PDGF-BB concentrations were determined by using an enzyme-linked immunosorbent assay (ELISA) system according to the manufacturer’s instructions. The human IL-1β ELISA (cat. # 88-7010) was purchased from eBioscience (San Diego, CA) and the Human PDGF-BB immunoassay (cat. # DBB00) was purchased from R&D Systems (Abingdon, UK).

Growth Factors and Inhibitors

Recombinant PDGF-AA (20 ng/ml); PDGF-BB (20 ng/ml); IL-1β (20 ng/ml); IL-1 receptor type 1 extracellular domain (1 μg/ml) (IL-1 sRI) were purchased from R&D systems; tyrphostins: 10 μmol/L AG1296 (PDGF receptor inhibitor); 10 μmol/L AG1295 (PDGF receptor inhibitor); 10 μmol/L AG1433 (PDGF β-receptor inhibitor) were purchased from Calbiochem; and 1 μg/ml filipin III was from Sigma. Suramin (100 μmol/L) used to inhibit growth factors with affinity to heparin was from Bayer (Leverkusen, Germany). The blocking pAb rabbit anti-PDGF-BB IgG (30 μg/ml)38 was kindly donated by Dr. CH Heldin (Ludwig Institute, Uppsala, Sweden).

Ex Vivo Experiments in Human Skin

Experiments designed to study tissue responses to tumor cell versus soluble PDGF-BB delivery in intact human skin were performed as follows: 10 × 10 cm sheets of human full thickness skin were obtained immediately after excision following mammary reduction surgery. Experiments were initiated within 1 hour of excision. Skin was incubated in RPMI containing 10% FCS before and during the experiments. 1 × 10⁶ tumor cells were suspended in 100 μl of vehicle consisting of 0.9% saline with 0.2% trypan blue (to mark the injection site) and injected intradermally using a 30-gauge syringe. The injection site and immediate surrounding tissue was excised 2 and 4 hours following tumor cell injection. Alternatively, skin was injected intradermally with 10 ng soluble PDGF-AA, 10 ng soluble PDGF-BB, or vehicle in a total volume of 40 μl using a 30-gauge syringe. The injection site and immediate surrounding tissue was excised 0.5 and 2 hours following injection. Tissues were embedded in OCT, and snap frozen in liquid nitrogen; 6-μm sections were then double- and triple-stained and analyzed using confocal microscopy as described above.

Surgical Specimens

Full thickness biopsies from human tissues were taken from: i) colorectal adenocarcinoma (n = 4); ii) adjacent normal colon (n = 2); iii) pannus formation from synevectomies due to rheumatoid arthritis (n = 3); iv) 7-day-old healing cutaneous wounds (n = 3); and v) adjacent normal skin (n = 2). Biopsy material was snap frozen in liquid nitrogen. The ethics committee, at the Uppsala Academic Hospital, approved the present study.

Statistical Analysis

The non-parametric Mann-Whitney test was used to model the data. The data are unpaired and two-tailed P values were calculated. Data are presented as % ± SEM or % ± SD when applicable. P values below 0.05 were considered significant.

Results

Identification and Characterization of Caveolin and Caveolae in Primary Pericyte Cultures

To characterize lipid rafts and caveolae in isolated primary human pericytes, cultured cells were subjected to biochemical analysis, immunolabeling in situ, or transmission electron microscopy. The caveolae/lipid raft markers caveolin-1 and -2, but not caveolin-3 were detected on the mRNA level (data not shown) and their corresponding proteins were found enriched in the lipid raft fraction of pericytes, denoted as the C-fraction (C). In these cells >90% of the caveolin was located in the rafts, while <10% of the caveolin was found in the non-raft cell fraction, denoted as the S-fraction (S) (Figure 1A). Immunofluorescence analysis demonstrated numerous punctatum of caveolin-1 present in >90% pericytes in vitro (Figure 1B). Immuno-enzyme labeling of tissue sections of human foreskins showed expression of caveolin-1 in vascular-like structures (Figure 1C). Double immunofluorescence labeling experiments of these tissue sections identified caveolin-1 expression in and around PECAM-1 expressing endothelium (Figure 1D). Transmission electron microscopic analysis demonstrated abundant caveolae in freshly isolated human pericytes (Figure 1E). The pericyte raft fractions were also enriched in certain membrane inner-leaflet signaling proteins (Figure 1F). Thus, the lipid rafts/caveolae of pericytes are analogous to the lipid rafts/caveolae described in other cells.1,2,5,6,7,8

Figure 1.

Primary cultured human pericytes contain caveolin-enriched membrane signaling complexes in vitro. A: Pericyte membranes were fractionated and 10 μg of the non-raft (Triton X-100-soluble; S fraction) and lipid raft (caveolar; C) fractions were subjected to SDS- polyacrylamide gel electrophoresis and immunoblot analysis using mAbs specific to caveolin-1 and -2. B: Caveolin-1 is expressed in a punctate pattern in non-stimulated pericytes. C: Caveolin-1 expression in vascular structures (arrows) in human foreskins. D: Double immunofluorescent staining of human foreskins showing caveolin-1 expression in (arrow) and juxtapositioned (block arrow) to PECAM-1 positive endothelium. E: Transmission electron micrograph of pericytes showing characteristic structural caveolae (arrows/arrowhead); and (F) Differential membrane distribution of downstream signaling effectors in pericytes. Asterisks indicate vascular lumens. Scale bar = 20 μm.

Distribution of Activated PDGF β-Receptors and Caveolin-1 in Primary Human Pericyte Cultures in Response to Soluble Ligand Delivery

We further investigated how PDGF β-receptors were distributed between the raft and non-raft fractions. Lysates from the raft and non-raft membrane fractions were quantified with regards to protein content and 20 μg of protein was loaded onto each lane and run on a SDS- polyacrylamide electrophoresis gel, transferred, and then immunoblotted. Immunoblotting with an anti-PDGF β-receptor antibody indicated that the 180-kDa PDGF β-receptor was distributed to both raft (65% to 70%) and non-raft fractions (30% to 35%) of the pericytes (Figure 2A). Treatment of pericytes with 20 ng/ml of soluble PDGF-BB did not substantially alter the relative distribution of the PDGF β-receptors between raft and non-raft fractions compared with unstimulated pericytes. To determine the degree of receptor activation, we immunoblotted with an anti-phosphotyrosine mAb. While soluble PDGF-BB triggers the phosphorylation of receptors in both raft and non-raft membranes, the majority of phosphorylated PDGF β-receptors were concentrated to the raft fraction. In agreement with previous findings PDGF β-receptor levels were down-regulated after 3 hours and remained so for up to 24 hours (data not shown) of PDGF-BB stimulation.21,37

Figure 2.

Activation of PDGF β-receptors in raft and non-raft membrane fractions in primary cultured human pericytes in response to soluble PDGF-BB. A: Levels of expression and tyrosine phosphorylation of PDGF β-receptors in non-raft (Triton X-100-soluble; S fraction) and lipid raft (caveolar; C-fraction) fractions in pericytes treated with 20 ng/ml of soluble PDGF-BB for different time points. A: Top panel, anti-PDGF β-receptor immunoblot. Bottom panel, same blot after stripping and immunoblotting with anti-phosphotyrosine. B–S: Confocal images of double immunofluorescence labeled, fixed human pericytes unstimulated (B–D) and following soluble PDGF-BB stimulation (20 ng/ml) for different time points (E–S), using PDGFR-B2 mAb depicting activated PDGF β-receptors (red; B, E, H, K, N, and Q) in conjunction with caveolin-1 pAb depicting caveolin-1 (green; C, F, I, L, O, and R). Composite images (D, G, J, M, P, and S). Colocalization appears in yellow. B–D: Unstimualted pericytes showing low levels of activated PDGF β-receptors using the PDGFR-B2 antibody. E-G: Pericytes stimulated for 15 minutes with soluble PDGF-BB. Note colocalization of activated PDGF β-receptors and caveolin-1 in focal contacts (arrow). H–J: Pericytes stimulated for 15 minutes with soluble PDGF-BB. Note colocalization between activated PDGF β-receptors and caveolin-1 in circular (arrow) and membrane (block arrow) ruffles. K–M: Pericytes stimulated for 60 minutes with soluble PDGF-BB. Note colocalization between activated PDGF β-receptors and caveolin-1 in the leading edge (arrow) and at the base (block arrow) of lamellipodia. N–P: Pericytes stimulated for 60 minutes with soluble PDGF-BB. Note colocalization between activated PDGF β-receptors and caveolin-1 at the base of and along newly formed cellular extensions (arrows). Q–S: Pericytes stimulated for 6 hours with soluble PDGF-BB. Note colocalization between activated PDGF β-receptors and caveolin-1 in cellular extensions (arrow) forming an elaborate interconnecting homotypic network between several individual cells. Scale bar = 20 μm.

Double immunofluorescence labeling of cells was performed using an antibody (PDGFR-B2) specific for activated PDGF β-receptors in conjunction with an antibody recognizing caveolin-1. Exchanging primary antibodies for non-specific IgGs gave no discernable background staining (data not shown). To assess the colocalization between PDGF β-receptors and caveolin-1 after soluble PDGF-BB stimulation, optical sections were acquired from these cells using confocal microscopy, the degree of colocalization quantified (Table 1) and displayed as compiled images (Figure 2, B–S) or as individual optical sections (see supplemental Figure S1 at http://ajp.amjpathol.org). As previously described21 cells that were not exposed to soluble PDGF-BB, did not show discernable immunoreactivity for PDGF β-receptors using the PDGFR-B2 antibody at a concentration of 1 μg/ml (Figure 2, B–D). On stimulation of pericytes with soluble PDGF-BB 57 ± 7% of PDGF β-receptors colocalized with caveolin-1 (Table 1). This colocalization occurred within several structures known to be involved in cell locomotion and cytoskeletal reorganization that are affected by PDGF-BB20 (Figure 2, E–S). These include: i) focal adhesions at the periphery of the cell (Figure 2, E–G); ii) circular and membrane ruffles (Figure 2, H–J); iii) proximal to, and at the migrating front of cellular lamellipodia, corresponding to foci where active reorganization of actin fibers is known to occur (Figure 2, K–M); iv) the early formation of cellular extensions (Figure 2, N–P); and v) at the base of more well developed cellular extensions that formed elaborate interconnections between pericytes in proximity to one another (Figure 2, Q–S).

Table 1.

Spatial Colocalization between activated PDGFR (a PDGFR) β and Caveolin-1 in the Different Conditions under Investigation

Condition	% of pixels aPDGFRβ-positive and caveolin-1-positive
Human tissues in vivo
Colorectal adenocarcinoma	20 ± 9
Rheumatoid arthritis	60 ± 9*
Wound healing	63 ± 9*
Skin ex vivo injected with
Tumor cells	20 ± 10
sPDGF-BB	80 ± 10*
Pericytes in vitro exposed to
Tumor cells	23 ± 10
sPDGF-BB	57 ± 7*

Quantification of the degree of spatial colocalization of activated PDGF β-receptors and caveolin-1 in pericytes under different experimental conditions and in human tissues. Data are presented as the number of pixels positive for both activated PDGF β-receptors and caveolin-1, as a percentage of the total number of pixels positive for PDGF β-receptors, ± SD. Quantification of colocalization was performed as previously described.13,14 

^*Indicates statistical significance (P < 0.05). Statistical comparisons were made between soluble versus cell-cell mediated ligand delivery in vitro, ex vivo, and between colorectal adenocarcinoma and reparative processes in vivo. Values are mean ± SD. 

Distribution of Activated PDGF β-Receptors and Caveolin-1 in Primary Human Pericytes in Response to Tumor Cell-Mediated Ligand Delivery

To determine whether the lipid raft, non-raft, or both cohorts of PDGF β-receptors responded to cell surface bound PDGF-BB, we co-cultured human pericytes with either the MDA-MB-435 breast carcinoma or the SCC-13 cutaneous squamous carcinoma, two human tumor cell lines of epithelial origin. The PDGF B-chain was not detected in pericytes, nor were PDGF β-receptors detected in either of the tumor cell lines (data not shown) as previously described.13 Since only pericytes expressed PDGF β-receptors, it was possible to use co-culture lysates for analysis of pericyte PDGF β-receptors. Time 0 for these experiments was defined as the earliest time at which tumor cells began to adhere to the pericyte monolayer ie, 1.5 hours after co-cultures had been initiated. Analysis of tyrosine phosphorylation showed that a large proportion of the phosphorylated PDGF β-receptors were confined within the non-raft fraction (Figure 3A). Interestingly and importantly, there was no diminution in the amount of PDGF β-receptors in the raft fraction, arguing against a translocation of receptors from the raft to the non-raft fraction (Figure 3A). Exposure times for these experiments are shorter than those shown in Figure 2 in order not to oversaturate the signal. This analysis of pericyte-tumor cell co-cultures suggested that cell surface-bound PDGF-BB uses a non-raft cohort of PDGF β-receptors for signal transmission.

Figure 3.

Activation of PDGF β-receptors in raft and non-raft membrane fractions in primary cultured human pericytes in response to cell-cell mediated PDGF-BB delivery. A–B: Levels of expression, metabolic labeling and tyrosine phosphorylation of PDGF β-receptors in Triton X-100-soluble (S-fraction) and insoluble (C-fraction) fractions in pericytes co-cultured with MDA-MB-435 and SCC-13 tumor cells. A: Top panel, anti-PDGF β-receptor immunoblot. Bottom panel, same blot after stripping and immunoblotting with anti-phosphotyrosine. B: Metabolic labeling of pericytes unstimulated, stimulated with PDGF-BB or in co-culture with tumor cells. C–J: Confocal images of pericytes co-cultured with tumor cells, using PDGFR-B2 mAb depicting activated PDGF β-receptors on pericytes (red; C and G) in conjunction with anti-caveolin-1 pAb depicting caveolin-1 (green; D and H), or anti-pancytokeratin mAb (blue; E) or anti-PDGF-B chain mAb (blue; I), and their composite images (F and J). Colocalization appears in yellow. C–F: Pericytes and tumor cells co-cultured for 6 hours. Note that activated PDGF β-receptors and caveolin-1 in pericytes juxtapositioned to tumor cells (block arrow) do not colocalize, but are diffusely expressed (arrow); and (G–J) pericytes and tumors cells co-cultured for 6 hours. Note activated PDGF β-receptors on pericytes (arrow) juxtapositioned to PDGF-B chain-expressing tumor cells (block arrow). Scale bar = 20 μm.

Both MDA-MB-435 and SCC-13 co-cultured with pericytes induced a marked and rapid up-regulation of the PDGF β-receptors in the pericyte non-raft fraction (Figure 3A). In addition to the mature PDGF β-receptor band of 180 kDa, both tumor cell lines induced the up-regulation of a 160-kDa band exclusively in the S-fraction that was immuno-reactive with the anti-PDGF β-receptor pAb. This 160-kD band was faintly expressed in the C-fraction in pericyte mono-cultures stimulated with soluble PDGF-BB (Figure 2A). This 160-kDa band corresponded to the PDGF β-receptor precursor10,20 (Figure 3A).

Metabolic labeling of pericytes in mono- and in co-cultures with tumor cells was performed to evaluate the neosynthesis and membrane distribution of PDGF β-receptors (Figure 3B). In the S-fraction of unstimulated pericytes a 160-kD radioactive band emerged, suggesting neosynthesis of the PDGF β-receptor precursor. Following stimulation with soluble PDGF-BB for 6 hours, a 180-kD band, corresponding to the mature PDGF β-receptor, transiently emerged suggesting an increase in neosynthesis of the receptor at 6 hours, followed by a decrease at the 24-hour time point below baseline levels when compared with unstimulated pericytes. In the C-fraction only the 180-kD mature PDGF β-receptor was detected. In the C-fraction, no difference in the intensity of the 180-kDa band was observed between PDGF-BB stimulated pericytes, unstimulated pericytes and co-cultures at either time point (Figure 3B).

In the S-fraction in co-cultures between MDA-MB-435 and pericytes their was an increase in the 160-kD and 180-kD isoforms of the PDGF β-receptor, as compared with unstimulated pericytes and PDGF-BB-stimulated pericytes at the 6-hour time point and to a lesser extent at 24 hours. In co-cultures between SCC-13 and pericytes, while there was a 160-kD band, a corresponding up-regulation, as was seen with MDA-MB-435, was not apparent (Figure 3B).

To assess the colocalization between PDGF β-receptors and caveolin-1 in response to cell-associated PDGF-BB, optical sections were acquired using confocal microscopy, and the degree of colocalization quantified (Table 1) and displayed as compiled images (Figure 3, C–J) or individual optical sections (see supplemental Figure S1 at http://ajp.amjpathol.org). PDGF β-receptors and caveolin-1 were diffusely expressed in the cell membrane and only 23 ± 10% of activated PDGF β-receptors colocalized with caveolin-1 (Table 1). Structures involved in cell locomotion and cytoskeletal reorganization, which were observed in pericytes stimulated with soluble PDGF-BB were not observed in co-cultures (Figure 3, C–J). Furthermore, cytokeratin expressing tumor cells (Figure 3, C–F), which expressed the PDGF-B chain (Figure 3, G–J) were found juxtapositioned to pericytes bearing activated PDGF β-receptors.

Proliferation and Cytoskeletal Reorganization in PDGF β-Receptor-Bearing Pericytes Stimulated with Soluble Versus Tumor Cell-Mediated Ligand Delivery

We further investigated if activation of different PDGF β-receptor cohorts on the cell surface in response to soluble or cell–cell mediated ligand delivery was responsible for different biological endpoints, namely proliferation versus cytoskeletal reorganization (Figure 4 and 5). Proliferation was measured by ³H-thymidine incorporation in cells (Figures 4, B and C, and 5, B and D). As an indicator of cytoskeletal reorganization the number of cells with circular/membrane ruffles at the 0.5-hour time point, and the number of cells exhibiting cellular extensions at the 6-hour time point ± stimulation with soluble PDGF-BB for an additional 30 minutes, were manually counted (Figures 4A and 5C). Proliferation and cytoskeletal reorganization data presented below are based on experiments performed with the MDA-MB-435 human breast carcinoma cell line. Qualitatively similar results were observed using the SCC-13 human cutaneous squamous carcinoma cell line (data not shown).

Figure 4.

Proliferation and cytoskeletal reorganization in response to soluble versus cell–cell-mediated PDGF-BB delivery in the presence of PDGF inhibitors and the caveolae-disrupting agent filipin. Cytoskeletal reorganization (A) and ³H-thymidine incorporation (B and C) in pericyte mono-cultures and co-cultures between pericytes and MDA-MB-435 tumor cells unstimulated or exposed to the caveolae-disrupting agent filipin (1 μg/ml), PDGF-BB (20 ng/ml), the growth factor inhibitor suramin (100 μmol/L), the blocking anti-PDGF-BB pAb (40 μg/ml), and the tyrosine kinase inhibitor AG1296 (10 μmol/L). Pre-incubation of MDA-MB-435 for 1 hour with suramin or PDGF-BB blocking antibodies enhances the effect of these agents. Controls are pericytes in mono-culture or co-cultured cells incubated with 0.1% FCS. *P < 0.05. Statistical comparisons were made for the different conditions against PDGF-BB stimulated pericytes in mono-culture (A, arrows), PDGF-BB stimulated pericytes in mono-culture (B, arrow), and non-stimulated co-cultures between pericytes and MDA-MB-435 tumor cells (C, arrows). Data from three separate experiments performed in triplicate are presented. Values are mean ± SEM or SD where applicable.

Figure 5.

Effect of IL-1β on proliferation and cytoskeletal reorganization in response to soluble versus cell–cell-mediated PDGF-BB delivery. IL-1β protein levels in cell lysates in mono- and co-cultures were measured using an ELISA system and presented as pg IL-1β per mg protein (A). ³H-thymidine incorporation in pericyte mono-cultures (B) and co-cultures (D) and cytoskeletal reorganization (C) between pericytes and MDA-MB-435 tumor cells unstimulated or exposed to PDGF-BB (20 ng/ml), IL-1β (20 ng/ml), and the IL-1β inhibitor IL-1 sRI (1 μg/ml). *P < 0.05. Statistical comparisons were made for the different conditions against PDGF-BB stimulated pericytes in mono-culture (B and C, arrows) and non-stimulated co-cultures between pericytes and MDA-MB-435 tumor cells (D, arrows). Data from three separate experiments performed in triplicate are presented. BDL, below detection limits. Values are mean ± SEM.

Quantification of PDGF-BB was determined using an ELISA assay system. MDA-MB-435 and SCC-13 contained 50 ± 5 and 25 ± 1 pg/mg protein, respectively. PDGF-BB could only be detected when cell membranes were subjected to high concentrations of NaCl thereby dissociating cell surface bound PDGF-BB. PDGF-BB was not detectable using this method in conditioned medium or in the intracellular compartment, thus suggesting that PDGF-BB produced in tumor cells is predominately cell-surface associated.

The results of the proliferation experiments in pericyte mono-cultures are shown in Figure 4B. Proliferation was increased in pericyte mono-cultures (209 ± 70%) in response to soluble PDGF-BB, as compared with unstimulated controls (100 ± 7%). AG1296, used at a concentration that inhibits both PDGF α and β-receptor tyrosine kinases, anti-PDGF-BB IgG and suramin inhibited pericyte proliferation in response to soluble PDGF-BB. Filipin, an agent that disrupts lipid rafts and raft-dependant signal transmission2,8,39 had no significant inhibitory effect on soluble PDGF-BB induced proliferation. As anticipated PDGF-AA, which only binds to the PDGF α-receptor, did not induce a significant increase (129 ± 7%) in pericyte proliferation compared with unstimulated controls, suggesting that PDGF α-receptors play little or no role in the proliferation of these cells.

The results of the cytoskeletal reorganization experiments in pericyte mono-cultures are shown in Figure 4B. In pericyte mono-cultures, soluble PDGF-BB induced cytoskeletal reorganization in a manner consistent with what previously has been described in fibroblasts (Figure 2).21,37 Tyrosine kinase inhibitors targeting the PDGF receptors, blocking anti-PDGF-BB IgG and suramin potently inhibited PDGF-BB induced cytoskeletal reorganization in pericytes (data not shown). Filipin-induced disruption of lipid rafts inhibited soluble PDGF-BB induced cytoskeletal reorganization (Figure 4A). PDGF-AA did not induce cytoskeletal reorganization, suggesting that PDGF α-receptors play little or no role in regulating cytoskeletal dynamics in these cells (data not shown).

When comparing actual counts per minute (cpm) values, mono-cultures of unstimulated pericytes or tumor cells incorporated 19 ± 9% (7300 ± 3900 cpm) and 26 ± 9% (23,000 ± 7200 cpm) ³H-thymidine, respectively, compared with when these cells were grown together in the co-culture system (pericytes = 38,500 ± 17,400 cpm and tumor cells 88,700 ± 37,800 cpm). The results of the proliferation experiments in co-cultures are shown in Figure 4C. In co-cultures both pericyte (29 ± 10%) and tumor cell (25 ± 7%) proliferation were markedly inhibited compared with control co-cultures in the presence of the PDGF receptor tyrosine kinase inhibitor AG1296 approaching levels seen in unstimulated pericyte and tumor cell mono-cultures. Similar results were observed when using other PDGF receptor kinase inhibitors, AG1295 and AG1433 (data not shown). Under co-culture conditions, anti-PDGF-BB IgG or suramin had a marginal inhibitory effect on pericyte proliferation; however, in contrast to experiments performed with tyrosine kinase inhibitors, no effect on tumor cell proliferation was observed. Interestingly, if tumor cells were pre-incubated for 1 hour with anti-PDGF-BB IgG or suramin, extensively washed to remove unbound IgG/suramin, and then co-cultured with pericytes, a marked effect on pericyte and tumor cell proliferation was observed. Pre-incubating tumor cells with anti-epithelial surface antigen IgG or rabbit IgG had no effect on pericyte or tumor cell proliferation. Adding soluble PDGF-BB to co-cultures resulted in a marginal increase in pericyte proliferation. Filipin alone or in combination with soluble PDGF-BB marginally decreased pericyte proliferation and had no effect on tumor cell proliferation under co-culture conditions (Figure 4C).

No cytoskeletal reorganization compared with what was seen in pericyte mono-cultures in response to soluble PDGF-BB was observed in co-cultures even after the addition of soluble PDGF-BB (Figure 4A).

The Effect of IL-1β on Proliferation and Cytoskeletal Reorganization in PDGF β Receptor-Bearing Pericytes Stimulated with Soluble Versus Tumor Cell-Mediated Ligand Delivery

IL-1β blocks downstream PDGF β-receptor signaling initiated in caveolae7,40,41 and is expressed in a number of tumors of epithelial origin.42 IL-1β was not detected in pericyte or MDA-MB-435 mono-cultures but was detected in SCC-13 mono-cultures. In co-cultures between pericytes and both tumor cell lines IL-1β was markedly up-regulated (Figure 5A). IL-1β could not be detected in the conditioned media from any of the culture conditions indicating that IL-1β was not secreted but remained cell associated (data not shown). We tested if IL-1β influences proliferation (Figure 5B and D) and/or cytoskeletal reorganization (Figure 5C) in pericytes stimulated with soluble PDGF-BB or cell–cell-mediated PDGF-BB. IL-1β or IL-1 receptor type 1 extra-cellular domain (IL-1 sRI), which acts as a decoy for IL-1β,43 did not affect soluble PDGF-BB-induced pericyte proliferation (Figure 5B). In pericyte mono-cultures, IL-1β inhibited soluble PDGF BB-induced cytoskeletal reorganization (Figure 5C). Neither IL-1β nor the IL-1 sRI had any effect on proliferation or cytoskeletal reorganization in unstimulated pericytes (data not shown).

In co-cultures, the addition of IL-1β or IL-1 sRI had no effect on baseline pericyte proliferation, only marginal effects on tumor cell proliferation (Figure 5D), and no effect on cytoskeletal reorganization (Figure 5C). However, under co-culture conditions when IL-1β was inhibited using IL-1 sRI, soluble PDGF-BB-induced cytoskeletal reorganization was partially restored in pericytes (Figure 5C). These results suggest that selective inhibition of raft-associated PDGF β-receptor signaling by IL-1β was responsible for the bifurcation of PDGF-BB delivered signals in tissue culture.

Distribution of Activated PDGF β-Receptors and Caveolin-1 in Response to Soluble Versus Tumor Cell-Mediated Ligand Delivery in Intact Human Skin

Injection of 10 ng soluble PDGF-BB into intact human skin led to a clearly discernable punctate staining of activated PDGF β-receptors in structures juxtapositioned to PECAM-1-expressing endothelium (Figure 6, A–D), resembling staining patterns seen in previous studies on human diseased tissues.14,15 In vitro, the PDGFR-B2 antibody is highly specific for activated/phosphorylated PDGF β-receptors.21 To determine whether this specificity is maintained in ex vivo tissue cultures, we tested the PDGFR-B2 antibody on human skin treated with 10 ng soluble PDGF-AA or vehicle (Figure 6, E–G). The lack of discernable punctate staining in these tissues demonstrated that the PDGFR-B2 antibody maintains its specificity toward phosphorylated/activated PDGF β-receptors under these conditions. Using the PDGFR-B2 antibody, activated PDGF β-receptors were predominantly detected on HMW-MAA expressing microvascular pericytes (Figure 6, H–K). In these tissues, 80 ± 10% of activated PDGF β-receptors colocalized with caveolin-1 (Table 1), particularly along pericyte extensions running along the abluminal surface of the endothelium (Figure 6, L–N). As in the in vitro experiments, the addition of soluble PDGF-BB ex vivo led to activation of PDGF β-receptors that colocalized with caveolin-1.

Figure 6.

Activation of PDGF β-receptors in intact human skin in response to soluble PDGF-BB ex vivo. Intact human skin injected with 10 ng of soluble PDGF-BB (A–D, H–N) or 10 ng of soluble PDGF-AA (E–G) and then subjected to immuno-enzyme labeling (A), double (B–G and L–N), and triple (H–K) immunofluorescence staining. PDGF β-receptors (red; B, E, H, and L) were detected in conjunction with PECAM-1 (green; C and F) or caveolin-1 (green; I and M), HMW-MAA mAb (blue; J). Composite images (D, G, K, and N). Colocalization appears in yellow. A: PECAM-1 expressing endothelium in blood vessels (arrows) from skin injected with soluble PDGF-BB. B–D: Intact skin injected with soluble PDGF-BB and harvested after 0.5 hours. Note expression of activated PDGF β-receptors (arrows) in close proximity to, but not in PECAM-1-expressing endothelium. E–G: Intact skin injected with soluble PDGF-AA and harvested after 0.5 hours. Note lack of expression of activated PDGF β-receptors using the PDGFR-B2 mAb in and around PECAM-1 expressing blood vessels (arrows). H–K: Intact skin injected with soluble PDGF-BB and harvested after 0.5 hours. Note colocalization of activated PDGF β-receptors and caveolin-1 in HMW-MAA-expressing microvascular pericytes (arrow). L–N: Intact skin injected with soluble PDGF-BB and harvested after 2 hours. Observe the high degree of colocalization between activated PDGF β-receptors and caveolin-1, notably in pericyte cellular extensions (arrow). Dashed lines (L–N) mark the center of blood the vessel lumen. Scale bar = 20 μm.

Injection of PDGF-BB-bearing tumor cells intradermally into intact human skin resulted in the occurrence of activated PDGF β-receptors in cells that were juxtapostioned to PECAM-1-expressing endothelium (Figure 7, A–F). Tumor cells, identified by their expression of cytokeratin (Figure 7B), were found interspersed among cells expressing activated PDGF β-receptors (Figure 7, C–F). In these experiments, only 20 ± 10% of activated PDGF β-receptors colocalized with caveolin-1 (Table 1). At tissue sites where tumor cells were injected, activated PDGF β-receptors were more abundant, as compared with sites injected with soluble PDGF-BB (Figure 6B compared with 7G). Caveolin-1 immunoreactivity was also reduced in structures positive for activated PDGF β-receptors (Figure 7, G–N). PDGF B-chain positive tumor cells were found interspersed among cells bearing activated PDGF β-receptor (Figure 7, K–N). These data support our in vitro findings that a distinct subpopulation of PDGF β-receptors, not associated with rafts and caveolin-1, are activated in response to PDGF β-receptor ligand delivered via cell–cell contacts in intact human tissues in contrast to soluble ligand, which leads to an activation of PDGF β-receptor associated with rafts and caveolin-1.

Figure 7.

Activation of PDGF β-receptors in intact human skin in response to cell–cell-mediated PDGF-BB delivery ex vivo. Intact human skin injected with 1 × 10⁶ MDA-MB-435 tumor cells (A–N) and then subjected to immuno-enzyme labeling for detection of PECAM-1 (A) and cytokeratin (B) as well as triple immunofluorescence staining (C–N). Activated PDGF β-receptors (red; C, G, and K) were detected in conjunction with PECAM-1 (green; D), caveolin-1 (green; H and L), or cytokeratin (blue; E and I), or PDGF B-chain (blue; M). Composite images (F, J, and N) are shown. Colocalization appears in yellow. (A–B) Intact skin injected with tumor cells and harvested after 2 hours. Note infiltration of cells (block arrows) surrounding PECAM-1 positive structures (A, arrow) that express cytokeratin (B, arrow). C–F: Intact skin injected with tumor cells and harvested after 4 hours. Note activated PDGF β-receptors (arrow) in close proximity to PECAM-1-expressing endothelium and interspersed with cytokeratin-positive tumor cells (block arrow). G–J: Intact skin injected with tumor cells and harvested after 4 hours. Note that activated PDGF β-receptors (arrow) are juxtapositioned to cytokeratin-positive tumor cells (block arrow) and do not colocalize with caveolin-1. K–N: Intact skin injected with tumor cells and harvested after 4 hours. Note activated PDGF β-receptors (arrows) juxtapositioned to PDGF B-chain expressing cells (block arrows) do not colocalize with caveolin-1. Scale bar = 20 μm.

Distribution of Activated PDGF β-Receptors and Caveolin-1 in Vascular Structures in Malignancy Versus Tissue Reparative Processes in Humans

To investigate if the expression patterns observed in vitro and ex vivo could be delineated in vivo, expression patterns and colocalization between activated PDGF β-receptors and caveolin-1 were examined in biopsied malignancies (Figure 8, and Tables 1 and S2 at http://ajp.amjpathol.org) and tissues undergoing reparative processes (Figure 9, Table 1, and supplemental Figures S2–S4 at http://ajp.amjpathol.org). These are examples of reactive tissue conditions in which PDGF-BB and PDGF β-receptors are expressed.14,15,17,18 In tissue sections from colorectal adenocarcinoma, PECAM-1-expressing blood vessels were often observed juxtapositioned to tumor acinar structures (Figure 8A). Caveolin-1 was expressed in PECAM-1-expressing endothelial cells (Figure 8, B–D) and in HMW-MAA-expressing pericytes (Figure 8, E–G), as well in cells not directly associated to vessels. PDGF β-receptors were predominately expressed in HMW-MAA-expressing pericytes as previously described (Figure 8, H–J).14 In the tumor stroma of colorectal adenocarcinoma, only 20 ± 9% of activated PDGF β-receptors colocalized with caveolin-1 (Table 1). Several activated PDGF β-receptor-positive structures expressed little or no caveolin-1, in agreement with the ex vivo experiments performed on human skin. Conversely, there were caveolin-1-positive structures that expressed little or no activated PDGF β-receptors (Figure 8, K–M). At high magnification, the subcellular distribution of caveolin-1 and activated PDGF β-receptors was discernable in the microvessels of colorectal adenocarcinoma (Figure 8, N–P). Here, caveolin-1 was predominately confined to the cell surface, where some degree of colocalization with activated PDGF β-receptors was observed. However, the main fraction of activated PDGF β-receptors had been internalized and was observed in the cytoplasm, where they did not colocalize with caveolin-1 (Figure 8, N–P).

Figure 8.

Activation of PDGF β-receptors in human colorectal adenocarcinoma in vivo. Immuno-enzyme labeling for detection of PECAM-1 (A) and double immunofluorescence labeling (B–P) were performed on sections from human colorectal adenocarcinoma (A–P) analyzed using light (A) and confocal microscopy (B–P). Activated PDGF β-receptors (red; H, K, and N) or caveolin-1 (red, B and E; and green, L and O) were detected in conjunction with PECAM-1 (green; C) or HMW-MAA (green; F and I). Composite images (D, G, J, M, and P). Colocalization appears in yellow. A: Note PECAM-1 positive vessels (arrow) juxtapositioned to tumor glandular structures (block arrow). B–D: Note caveolin-1 expression in PECAM-1 positive endothelium (arrow), as well as in cells in close proximity to the endothelium (block arrow). E–G: Note caveolin-1 expression in HMW-MAA positive pericytes (arrow), as well as in cells in close proximity to pericytes (block arrow). H–J: Note expression of activated PDGF β-receptors in HMW-MAA-expressing pericytes (arrows). K–M: Note lack of colocalization between activated PDGF β-receptors and caveolin-1 (arrows). N–P: High magnification of a microvessel in colorectal adenocarcinoma. Note the lack of colocalization of activated PDGF β-receptors and caveolin-1 in individual cells (arrows). Dashed line marks the center of the blood vessel lumen (N–P). Asterisks denote intracellular compartment of individual cell. Scale bar = 20 μm (A–M); 5 μm (N–P).

Figure 9.

Activation of PDGF β-receptors in pannus formation in rheumatoid arthritis in humans in vivo. Immuno-enzyme labeling for detection of PECAM-1 (A) and double immunofluorescence labeling (B–M) were performed on sections from human pannus formation (A–M) and analyzed using light (A) and confocal microscopy (B–M). Activated PDGF β-receptors (red; H and K) or caveolin-1 (red; B and E; and green, L) were detected in conjunction with PECAM-1 (green; C) or HMW-MAA (green; F and I). Composite images (D, G, J, and M). Colocalization appears in yellow. A: Note PECAM-1 positive vessels (arrow) surrounded by cellular infiltrates (block arrow). B–D: Note caveolin-1 expression in PECAM-1-positive endothelium (arrow), as well as in cells juxtapositioned to the endothelium (block arrow). E–G: Note caveolin-1 expression in HMW-MAA-positive pericytes (arrow) as well as in cells juxtapositioned to pericytes (block arrow). H–J: Note expression of activated PDGF β-receptors in HMW-MAA-expressing pericytes (arrows). K–M: Note colocalization between activated PDGF β-receptors and caveolin-1 (arrows). Asterisks indicate vessel lumen. Continuous line (J) outlines the blood vessel lumen. Dashed line (K–M) marks the center of the blood vessel lumen. Scale bar = 20 μm (A–J), and 10 μm (K–M).

In 7-day-old healing wounds (see supplemental Figures S3–S4 at http://ajp.amjpathol.org) and pannus formation in rheumatoid arthritis (Figure 9) caveolin-1 was expressed in and in close proximity to PECAM-1-expressing endothelial cells (Figure 9, A–D) and in HMW-MAA-expressing pericytes (Figure 9, E–G), as well in cells not directly associated to vessels similar to the distribution pattern observed in colorectal adenocarcinoma. PDGF β-receptors were predominately expressed in HMW-MAA-expressing pericytes in both healing wounds (data not shown) and during pannus formation (Figure 9, H–J), as previously described.14,15 In contrast to what was observed in colorectal adenocarcinoma, in 7-day-old healing wounds and in the pannus of rheumatoid arthritis, 63 ± 9% and 60 ± 9% of activated PDGF β-receptors colocalized with caveolin-1, respectively (Table 1). Thus, the findings in rheumatoid arthritis and wound healing resemble the distribution and degree of colocalization seen in the ex vivo skin experiments when soluble PDGF-BB was injected. In and at the base of vascular sprouts activated PDGF β-receptors and caveolin-1 colocalized to a higher degree than in the remaining portions of the vessel wall (see supplemental Figures S3E–S3G at http://ajp.amjpathol.org). In healing wounds (see supplemental Figures S3H–S3J at http://ajp.amjpathol.org) and in the pannus of rheumatoid arthritis (data not shown) PDGF B-chain expression was observed in a subset of cells in the endothelial lining of larger vessels. Colocalization between activated PDGF β-receptors and caveolin-1 were observed in the abluminal layer of cells located in the cell membrane facing the endothelium, as well as in between adjacent smooth muscle cells (see supplemental Figures S3K–S3M at http://ajp.amjpathol.org).

The investigated biopsies from tumors show that activated PDGF β-receptors in the microvasculature predominantly did not colocalize with caveolin-1, in agreement with our in vitro and ex vivo experiments with tumor cells. Colocalization was observed in biopsies from tissue reparative processes, in agreement with the in vitro and ex vivo experiments with soluble PDGF-BB.

Discussion

In this report, we demonstrate that pericytes express two cohorts of PDGF β-receptors, one in lipid rafts and the other in non-raft membrane compartments. These cohorts are individually and predominantly stimulated by two forms of PDGF-BB, soluble- and cell-membrane bound, and result in two distinct biological endpoints, proliferation and reorganization of the cytoskeleton. These data provide a mechanism by which ligand-producing cells influence the biological outcome of receptor activation in the recipient cell by directing stimulation toward one of the two cohorts of receptors. Our report also describes a novel assay by which the therapeutic potentials of experimental compounds may better be assessed/screened in vitro.

Soluble PDGF-BB predominantly activated a cohort of PDGF β-receptors in human pericytes that resided in caveolin-containing lipid rafts ie, caveolae. Upon soluble PDGF-BB stimulation, no reduction of receptor levels was observed in the non-lipid raft fraction (S-fraction). Furthermore, PDGF β-receptor phosphorylation in the raft cohort (C-fraction) peaked at 15 minutes after soluble PDGF-BB stimulation, preceding the peak phosphorylation of PDGF β-receptors in the S-fraction by 45 minutes, inferring that the two cohorts also differed in the kinetics of PDGF β-receptor phosphorylation after stimulation with soluble PDGF-BB. Our results argue against a translocation of the PDGF β-receptors from the non-raft to the lipid raft compartment upon ligand stimulation. Rather, it is more likely that a population of PDGF β-receptors resides in the lipid rafts of unstimulated cells and becomes activated on stimulation with soluble ligand.

Our results show that the level of PDGF β receptors in the non-raft fraction, but not the lipid raft fraction was rapidly and markedly increased when ligand was delivered via cell–cell contacts during tumor cell–pericyte co-cultures. This could be due to ligand-induced ‘locking’ of the PDGF β-receptor in cell–cell contacts, in effect constraining the receptor to the cell surface, thus physically preventing internalization and degradation. Alternatively, cell–cell contact may influence PDGF β-receptor synthesis and/or degradation processes. The emergence of the 160-kD band corresponding to the PDGF β-receptor precursor led us to believe that the most likely explanation for the observed up-regulation would be an increase in receptor synthesis. Metabolic labeling experiments in part supported this notion. In the MDA-MB-435 co-cultures an increase in neosynthesis was observed. However, a similar up-regulation in the SCC-13 co-cultures was not observed despite the observation that the PDGF β-receptor is up-regulated in co-cultures with both tumor cell lines. Thus, the observed up-regulation of the PDGF β-receptor can in part be explained by increased neosynthesis. However, the relative contribution of “locking” of the receptor at the cell surface or a decrease in degradation in co-cultures cannot be ruled out. Discriminating between these two options is complex and is subject for further study. The increased levels of PDGF β-receptor expression only in the non-raft membrane fraction was a specific result of cell–cell contacts between pericytes and PDGF-BB expressing cells. Thus, ligand-producing cells were not only capable of delivering PDGF-BB to and modifying the biological response in recipient cells, but were also capable of dictating the responsiveness of receptor-bearing cells through receptor up-regulation.

Determining the concentration of PDGF-BB present on tumor cells is of particular importance for understanding the relationship between soluble and membrane-bound PDGF-BB-delivered signals. The actual concentration of ligand available for receptor ligation in cell–cell contacts is difficult to quantify given that the ligand is spatially immobilized.44 However, the level of receptor phosphorylation induced by the ligand in the recipient cell can be used as a measurement of the ligand concentration.13 In a previous study, the amount of PDGF-BB was quantified indirectly by comparing the level of tyrosine phosphorylation of PDGF β-receptors in response to: cell–cell-mediated ligand delivery, ligand liberated from the cell surface via treatment with heparinase, ligand present in tumor cell lysate, and ligand shed into conditioned medium versus stimulation with 20 ng/ml of PDGF-BB for 0.5 hours.13 Variations in tyrosine phosphorylation were relatively constant between the different tumor cell lines studied, regarding the amounts of PDGF-BB from the different PDGF-BB source fractions. Cell–cell-mediated ligand delivery induced tyrosine phosphorylation of PDGF β-receptors equivalent to 10 to 15 ng/ml of soluble PDGF-BB. Medium that had been conditioned by tumor cells over a period of 24 hours resulted in only an equivalent of 1 to 2 ng/ml of soluble PDGF-BB. Thus, with regards to the quantities of PDGF-BB delivered in the soluble versus the cell-associated form, the two conditions are comparable. In the present study, we quantified the amount of cell-associated PDGF-BB and the amount of PDGF-BB secreted into the medium. Importantly, the results showed that PDGF-BB produced in these tumor cells is almost exclusively cell associated, and are secreting undetectable amounts of PDGF-BB (based on our analysis methods). Furthermore, no PDGF-BB could be detected within the intracellular compartment suggesting that PDGF-BB is almost exclusively cell surface associated. Thus, the main mode of ligand delivery in the tumor cells studied is via cell–cell contacts.

Tumor cells required approximately 1.5 hours to adhere to pericyte monolayers when grown in co-cultures. To ensure that there were equivalent levels of adherent cells in each co-culture, we set our first time point at 3 hours post-adhesion, to minimize variation in ‘settling’ time. Measurements of PDGF β-receptor phosphorylation before that time point were inconsistent, likely due to variations in the ability of tumor cells to adhere and establish cell–cell contact with pericytes (data not shown). When soluble PDGF-BB was added at different time points after co-cultures had been established no cytoskeletal reorganization was noted unless IL-1β was inhibited, a result in marked contrast to what was observed in pericyte mono-cultures. Exposing pericyte mono-cultures to soluble PDGF-BB for 3 hours resulted in a down-regulation of PDGF β-receptors, which was sustained for at least 24 hours. No quantitative or qualitative differences in cytoskeletal reorganization were seen in cells that had been exposed to PDGF-BB beyond the 6-hour time point. Thus the time periods chosen for soluble versus cell-associated ligand delivery are overlapping and qualitatively comparable and represent the time periods in which ligand-receptor interaction are resulting in measurable changes in receptor activation.

In our in vitro experimental system, we investigated two PDGF-BB-induced biological endpoints: proliferation and cytoskeletal reorganization. We found that both soluble PDGF-BB and cell–cell contact delivered ligand-induced pericyte proliferation. However, cytoskeletal reorganization was only observed in mono-cultures after the addition of soluble PDGF-BB. The lack of cytoskeletal reorganization in co-cultures was bewildering since both Western analysis of PDGF β-receptor tyrosine phosphorylation and immunofluorescence staining revealed an abundance of activated PDGF β-receptors in co-cultures. However, these activated PDGF β-receptors were confined to the non-raft compartment and did not colocalize with caveolin-1. The lack of cytoskeletal reorganization in co-cultures could be explained if raft PDGF β-receptor signaling was required and that cell–cell-mediated ligand delivery was unable to stimulate receptors in the raft compartment. This does not explain why there was no cytoskeletal reorganization after the addition of soluble PDGF-BB to co-cultures. Another possibility is that other factors expressed in co-cultures assisted in directing the signal by inhibiting downstream PDGF β-receptor signaling specifically in the rafts. In support of the latter, our data suggest that IL-1β inhibited PDGF β-receptor-induced cytoskeletal reorganization, but not proliferation, in response to soluble PDGF-BB. Furthermore, inhibiting IL-1β, produced in the co-culture system, permitted cytoskeletal reorganization to proceed, but only in response to exogenous soluble PDGF-BB. This suggests that tumor cell-produced IL-1β specifically inhibited PDGF β-receptor-mediated signal transmission through the raft compartment. The requirement for exogenous soluble PDGF-BB to elicit a cytoskeletal response in IL-1β-inhibited co-cultures suggests two things: Firstly, that tumor cells secreted a small amount of soluble PDGF-BB, which was insufficient or unable to induce cytoskeletal reorganization. Secondly, that cell–cell-mediated ligand delivery to PDGF β-receptors residing in the non-raft compartment did not play a role in PDGF-BB-induced cytoskeletal reorganization. IL-1β inhibition did not affect proliferation of either tumor cells or pericytes in co-cultures, suggesting that PDGF β-receptors in the raft compartment played little role in PDGF-BB-mediated cell proliferation, and that the non-raft cohort of PDGF β-receptors was regulated by other factors besides IL-1β. How IL-1β alters signal transmission through lipid rafts is not known, but it has been shown that IL-1β-stimulated conversion of sphingomyelin to ceramide, which blocks PDGF signaling, occurs in the rafts of human fibroblasts.7,41,45 Why cell–cell-mediated ligand delivery was unable to stimulate raft resident PDGF β-receptors, even in the presence of IL-1β inhibitors, is a question requiring further study.

Our results suggest that cell-to-cell-mediated ligand delivery predominantly stimulates non-raft PDGF β-receptors, which transduce signals that lead to proliferation. Thus, the biological endpoint resulting from signaling mediated via PDGF β-receptors located in the non-raft compartment differ from PDGF β-receptors located in the raft compartment. This would suggest that mode of ligand delivery dictates the biological effect of the ligand on the recipient cell. Although our data point to a clear distinction between the receptor cohorts (raft versus non-raft) and biological effects (cytoskeletal reorganization versus proliferation) induced by PDGF-BB (soluble versus cell-bound), these distinctions were not absolute as soluble PDGF-BB did induce some pericyte proliferation. Of course, some overlap is anticipated and the absence of any would have been quite extraordinary. This is supported by the observation that the cytoskeletal effects of soluble PDGF-BB in pericytes and other mesenchymal cells occur at substantially lower concentrations of PDGF-BB than do the proliferative effects46 (see supplemental Table S1 at http://ajp.amjpathol.org). The latter phenomenon suggests that the commonly observed proliferative effects of soluble PDGF-BB may be due to the use of high concentrations of soluble PDGF-BB, resulting in some activation of PDGF β-receptors in the non-raft compartment. Disruption of lipid raft structure or inhibition of signal transmission through lipid rafts had little or no effect on pericyte proliferation in response to soluble PDGF-BB or cell–cell-mediated ligand delivery. This suggests that PDGF β-receptor signaling in lipid rafts is of minor importance for pericyte proliferation in response to soluble PDGF-BB or under co-culture conditions. The mechanism behind the downstream divergence after raft versus non-raft PDGF β-receptor stimulation is currently unknown. Some insight has been provided in a report of different pools of Src family kinases that are confined to raft and non-raft membrane domains and that these two different pools of Src family kinases mediate different biological endpoints after PDGF stimulation in murine 3T3 cells.47 Our findings that Fyn and Src are differently localized between the raft and non-raft fraction in primary human pericytes may support this finding.

In agreement with earlier results, suramin and anti-PDGF-BB antibodies were able to inhibit tyrosine phosphorylation of PDGF β-receptors stimulated by soluble PDGF-BB, but not cell-bound PDGF-BB delivered in cell–cell contacts.13 Our present results show that while suramin and anti-PDGF-BB pAb had only a minor effect on pericytes in already established co-cultures, the effect increased dramatically when PDGF-BB on the surface of tumor cells was blocked by pre-incubating these cells with suramin or anti-PDGF-BB pAb. This suggests that antibodies, and even low molecular weight compounds like suramin, have limited access to their targets in cell–cell contacts. Thus, therapeutic efforts aimed at inhibiting growth factor function must take into account the mode by which the growth factor is delivered to the target cell (soluble versus ligand-delivered in cell–cell contacts) as intervention with different classes of compounds may have different biological effects, despite the fact that in each case the same ligand-receptor pair is involved. Our data also suggest that results obtained under co-culture conditions differ from traditional single cell systems and may be more useful in drug screening assays designed to identify potential compounds of therapeutic interest.

We extended our in vitro findings to three human conditions in which pericytes express PDGF β-receptors.14,15,17,18 Immunofluorescence staining of tumor biopsies showed a similar pattern of expression and distribution of activated PDGF β-receptors and caveolin-1 as was observed in co-culture and ex vivo experiments with tumor cells. All these conditions demonstrated a low degree of colocalization of activated PDGF β-receptors and caveolin-1. This suggests that cell–cell-mediated ligand delivery predominates in malignant disease. In contrast, comparison of soluble PDGF-BB treated pericyte mono-cultures and intact skin with tissue reparative processes, such as the formation of granulation tissue during wound healing or pannus in rheumatoid synovitis, which all show a high degree of colocalization between activated PDGF β-receptors and caveolin-1, suggests that soluble PDGF-BB is the main mode of ligand delivery in these in vivo conditions. The higher degree of colocalization between activated PDGF β-receptors and caveolin-1 seen in reparative processes versus colorectal adenocarcinoma may have several explanations. Firstly, soluble secreted ligand may have a more dominant role in PDGF-mediated effects in reparative processes. Secondly, the cells that produce PDGF-BB during reparative processes are non-transformed and may not support cell–cell-mediated PDGF β-receptor stimulation. In support of the latter, non-transformed epithelial cells that express PDGF-BB on their cell surface are not capable of stimulating PDGF β receptor-bearing fibroblasts under co-culture conditions (Kristofer Rubin, Uppsala University, Sweden. personal communication) or in the normal human colon in vivo.13 Most likely, both modes of ligand delivery contribute to any given condition, but the relative contribution of each varies depending on the underlying pathological or reparative processes.

This study may be the first demonstrating that receptors of the same type within and outside lipid rafts are preferentially activated by different modes of ligand delivery, have different kinetics, and may regulate different downstream cellular responses. Thus, we conclude that the distribution of receptors within and outside of lipid rafts may represent a membrane level mechanism for the control of receptor activation and may provide a partial explanation for the ability of a specific growth factor receptor to initiate different biological responses, an issue that has been given much attention but has not been resolved.