Formation of Hyaluronan- and Versican-rich Pericellular Matrix by Prostate Cancer Cells Promotes Cell Motility*

Previous studies have demonstrated that high levels of hyaluronan (HA) and the chondroitin sulfate proteoglycan, versican in the peritumoral stroma are associated with metastatic spread of clinical prostate cancer. In vitro integration of HA and versican into a pericellular sheath is a prerequisite for proliferation and migration of vascular smooth muscle cells. In this study, a particle exclusion assay was used to determine whether human prostate cancer cell lines are capable of assembling a pericellular sheath following treatment with versican-containing medium and whether formation of a pericellular sheath modulated cell motility. PC3 and DU145, but not LNCaP cells formed prominent polarized pericellular sheaths following treatment with prostate fibroblast-conditioned medium. The capacity to assemble a pericellular sheath correlated with the ability to express membranous HA receptor, CD44. HA and versican histochemical staining were observed surrounding PC3 and DU145 cells following treatment with prostatic fibroblast-conditioned medium. The dependence on HA for integrity of the pericellular sheath was demonstrated by its removal following treatment with hyaluronidase. Purified versican or conditioned medium from Chinese hamster ovary K1 cells overexpressing versican V1, but not conditioned medium from parental cells, promoted pericellular sheath formation and motility of PC3 cells. Using time lapse microscopy, motile PC3 cells treated with versican but not non-motile cells exhibited a polar pericellular sheath. Polar pericellular sheath was particularly evident at the trailing edge but was excluded from the leading edge of PC3 cells. These studies indicate that prostate cancer cells recruit stromal components to remodel their pericellular environment and promote their motility.

Previous studies have demonstrated that high levels of hyaluronan (HA) and the chondroitin sulfate proteoglycan, versican in the peritumoral stroma are associated with metastatic spread of clinical prostate cancer. In vitro integration of HA and versican into a pericellular sheath is a prerequisite for proliferation and migration of vascular smooth muscle cells. In this study, a particle exclusion assay was used to determine whether human prostate cancer cell lines are capable of assembling a pericellular sheath following treatment with versican-containing medium and whether formation of a pericellular sheath modulated cell motility. PC3 and DU145, but not LNCaP cells formed prominent polarized pericellular sheaths following treatment with prostate fibroblast-conditioned medium. The capacity to assemble a pericellular sheath correlated with the ability to express membranous HA receptor, CD44. HA and versican histochemical staining were observed surrounding PC3 and DU145 cells following treatment with prostatic fibroblast-conditioned medium. The dependence on HA for integrity of the pericellular sheath was demonstrated by its removal following treatment with hyaluronidase. Purified versican or conditioned medium from Chinese hamster ovary K1 cells overexpressing versican V1, but not conditioned medium from parental cells, promoted pericellular sheath formation and motility of PC3 cells. Using time lapse microscopy, motile PC3 cells treated with versican but not non-motile cells exhibited a polar pericellular sheath. Polar pericellular sheath was particularly evident at the trailing edge but was excluded from the leading edge of PC3 cells. These studies indicate that prostate cancer cells recruit stromal components to remodel their pericellular environment and promote their motility.
Versican is a large chondroitin sulfate proteoglycan (CSPG) 2 consisting of a core protein (M r Ͼ 400 kDa) with 12-15 chondroitin sulfate (CS) side chains covalently attached. It belongs to a family of extracellular proteoglycans (hyalectins) that bind to hyaluronan (HA) of which aggrecan, the cartilage-specific proteoglycan is the prototype (1). Four versican isoforms have been identified. V0, V1, and V2 result from alternative splicing of two large exons each encoding a glycosaminoglycan (GAG) attachment domain (GAG␣ and GAG␤). V3 lacks these domains and is devoid of GAG side chains. V0 and V1 are the principal isoforms present in the interstitial matrix of most tissues. All versican isoforms contain globular domains at the N terminus (G1) and C terminus (G3). The G1 domain contains two tandem repeat link modules, which bind HA, and the G3 domain consists of a set of lectin-, epidermal growth factor-, and complement-binding protein-like subdomains with structural similarity to the selectin family (1,2).
We have previously demonstrated that elevated levels of versican in peritumoral stroma, is an indicator for disease relapse following surgery for clinically localized prostate (3)(4)(5) and breast cancer (6,7). Our more recent studies demonstrated that purified versican from cultured human prostatic fibroblasts inhibited adhesion of prostate cancer cells to a fibronectin substratum in vitro (8). These findings are consistent with versican modulating tumor cell attachment to the stromal matrix in vivo, a primary facet controlling cancer cell motility and invasion.
CSPGs such as versican interact with other extracellular matrix (ECM) molecules and form macromolecular complexes (9). The expression of CD44 on the cell membrane of chondrocytes is essential to direct assembly of HA pericellular matrix (10). Urinary bladder carcinoma cell lines that express CD44 receptors but minimal HA and versican have also been reported to assemble prominent pericellular matrices following the addition of exogenous HA and aggrecan (10). Evanko et al. (11) recently reported that an HA coat incorporating versican is obligatory for the proliferation and migration of smooth muscle * This work was supported by the Cancer Council of South Australia. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. □ S The on-line version of this article (available at http://www.jbc.org) contains a supplemental movie. 1

cells in vitro.
Participation of versican core protein in matrix assembly through its HA binding property, and its numerous highly negatively charged CS side chains might be integral to modulation of cellular adhesion and enhancement of motility and invasion. In this study we investigated whether prostate cancer cells can utilize the ECM components, HA and versican, secreted by prostatic fibroblasts to assemble a pericellular matrix and promote cancer cell motility.

EXPERIMENTAL PROCEDURES
Cell Culture and Collection of Prostatic Fibroblast-conditioned Medium-The human prostate adenocarcinoma cell lines LNCaP, PC3, and DU145 and Chinese hamster ovary (CHO K1) cell lines were purchased from the American Type Culture Collection. CHO K1 cells overexpressing recombinant V1 (rV1) versican (CHO V1) were kindly provided by Prof. R. LeBaron (Division of Life Science, University of Texas at San Antonio). The prostate cancer cell lines were maintained in 80 cm 2 flasks in complete RPMI 1640 medium (Invitrogen) supplemented with 4 mM L-glutamine, 100 g/ml penicillin, 100 g/ml streptomycin, 2 g/ml amphotericin B, and 5% FBS. The CHO K1 and CHO V1 were maintained as described for the prostate cancer cells but using ␣-MEM nucleoside-free medium (JRH Biosciences, Lenexa, KS) as described previously (12). Fibroblasts were isolated from prostatic tissue obtained with informed consent from patients undergoing transurethral resection of the prostate to resolve the urine voiding symptoms of benign prostatic hyperplasia (8). The prostate fibroblasts were maintained in RPMI containing 5% FBS and were shown to be free of Mycoplasma by PCR analysis. Prostate fibroblasts, CHO K1 or CHO V1 were plated at a density of 1 ϫ 10 4 cells/ cm 2 and incubated in culture medium for 72 h prior to collecting conditioned medium (CM). The CM was used fresh or stored at Ϫ70°C for up to 12 months before use.
Versican Purification and Western Immunoblotting-Versican was isolated from prostate fibroblast or CHO V1 CM (recombinant versican, rV1) using a combination of anion exchange and gel filtration chromatography. Briefly, CM stored at Ϫ70°C was thawed and batch-adsorbed to Q-Sepharose (10 ml/liter of medium, Amersham Biosciences, Uppsala, Sweden) at 4°C for 24 -48 h in the presence of protease inhibitor tablets (Roche Applied Science), batch-eluted with 2 M NaCl in PBS, pH 7.4 and purified by gel chromatography on a Sephacryl S400 (15 ϫ 650-mm column, Amersham Biosciences) as described previously (8). Versican-containing fractions were concentrated 10 -20-fold using Centriprep centrifugal filters (Amicon Bioseparations, Bedford, MA) and Nanosep TM microconcentrators (Pall Gelman Laboratory, Ann Arbor, MI) with M r 50,000 and 300,000 cut-offs, respectively. The molecular integrity of the purified versican samples was determined by immunoblotting with the rabbit antibody to recombinant human versican (Vc) provided by Prof. LeBaron as previously described (13). To determine if other CS proteoglycans were co-purified with versican, a parallel membrane was incubated with antichondroitin sulfate mouse monoclonal antibody (2B6, C-4-S, ICN Biochemicals, Aurora, OH). Visualization was achieved by anti-rabbit IgG or anti-mouse IgG peroxidaseconjugated secondary antibodies (Dako Labs, Glostrup, Denmark) and enhanced chemiluminescence (ECL, Amersham Biosciences). Gels run in parallel were also stained using Coomassie Blue (Difco Laboratories, Surrey, UK) and silver staining (Bio-Rad) to assess the purity of the versican fractions. Results from a typical versican purification are shown in Fig. 1. The highest concentration of versican was found in pooled fractions 9 ϩ 10 (isoforms V0 ϩ V1, ϳ400 kDa, Fig. 1A). A parallel gel immunoblotted with a CS monoclonal antibody demonstrated a broad band at ϳ400 kDa, consistent with the versican, and an additional minor band at ϳ150 kDa (Fig. 1B). This 150-kDa band was present in the versican immunoblot and is likely to be a versican degradation product. These bands but no other contaminating protein bands were observed in the silver stained gel (Fig. 1C). No contaminating bands were observed in fractions 7 ϩ 8 or fractions 9 ϩ 10 in a parallel gel stained with Coomassie Blue (not shown).  4 -6). Only bands corresponding to versican (ϳ400 and 150 kDa) were observed in fractions 7 ϩ 8 and fractions 9 ϩ 10 in a parallel immunoblot with CS-4 (2B6) monoclonal antibody (B) and the silver-stained gel (C). No contaminating proteins bands were detected in fractions 7 ϩ 8 or fractions 9 ϩ 10 in a parallel Coomassie Blue-stained gel (not shown).
Versican Quantitation-As versican levels in purified fractions from prostatic fibroblast CM were too low to be accurately determined by colorimetric protein assays, a versican ELISA was developed. In the absence of a quantified versican standard being available, the versican concentration was determined relative to a highly purified and concentrated fraction of rV1 versican defined as containing 100 arbitrary units/ml. Versican samples (50 l) diluted in PBS containing 0.1% BSA and 0.05% Tween 20 were bound to 96-well tissue culture plates for 2 h at room temperature in PBS and washed with PBS/0.05% Tween 20 prior to the addition of versican monoclonal antibody (12C5, 1/250, developed by Dr. Richard Asher and obtained from by the Developmental Studies Hybridoma Bank, NICHD, University of Iowa). Following an overnight incubation at 4°C, the wells were washed three times with PBS/0.05% Tween 20 and incubated with 100 l of anti-mouse IgG peroxidase-conjugated secondary antibody (1/2000) for 90 min at room temperature. Wells were subsequently washed three times with PBS/0.05% Tween 20. Substrate solution (100 l, 0.4 mg/ml o-phenylenediamine in 0.01% H 2 O 2 , 0.1 M citrate phosphate buffer, pH 5.0) was added to each well and after 15 min at room temperature, the reaction was stopped by addition of 25 l of 2.5 M H 2 SO 4 and absorbance was read at 450 nm. Negative controls included no primary antibody and no secondary antibody and CHO K1 CM and aggrecan instead of versican. A typical standard curve in the versican ELISA had the equation y ϭ 0.054x ϩ 0.039 with an R 2 ϭ 0.986. Relative versican concentrations ranged between 0.1-0.4 units/ml versican in prostatic fibroblast CM and 1.0 -5.0 units/ml versican in the purified versican fractions from prostatic fibroblast CM.
Quantitation of HA Synthesis-The concentration of HA in cell culture supernatants was determined by a competitive binding assay (14). Tissue culture microtiter plates (96 well) were coated with human umbilical cord HA (Sigma-Aldrich) at 50 g/ml in 200 mM sodium carbonate buffer (pH 9.6) for 4 h at 37°C. Unbound HA was removed with four washes of PBS/ 0.05% Tween 20. CM (48 h) was harvested from prostate carcinoma cells (PC3, DU145, and LNCaP) or prostate fibroblasts plated (2 ϫ 10 4 /well) cultured in 5% FBS RPMI in 12-well plates. Final cell numbers were determined after trypsinization by manual counting using a hemacytometer. CM from prostate cells (100 l) was combined with 100 l of biotinylated HAbinding protein (1 g/ml Seikagaku Corp., Tokyo, Japan) and incubated in the HA-precoated wells overnight at room temperature. The plate was washed four times with PBS/Tween 20, developed using a streptavidin HRP-system (Dako Labs) using o-phenylenediamine (Sigma-Aldrich) as a substrate, and the absorbance measured at 490 nm. The mean HA concentration for each sample was interpolated from an HA standard curve performed in parallel and normalized to cell number. Data are presented as g of HA per 10 6 cells.
Visualization of Pericellular Matrix-Pericellular sheath formation was visualized by a particle exclusion assay utilizing the steric exclusion property of the HA gel (11). Prostate cancer cells (PC3, DU145, and LNCaP) were plated (3 ϫ 10 3 ) in 48-well tissue culture plates in 5% FBS RPMI. After 48 h of culture, prostate cancer cells were treated with either prostatic fibroblast CM (0.1-0.4 units/ml versican), purified human versican from prostatic fibroblasts (0.5 unit/ml) rV1 (62.5 milliunits-4 units/ml), bovine aggrecan (0.1 mg/ml, Sigma-Aldrich) or control medium. After 24 h of treatment, 300 l of a suspension of human red blood cells (10 7 /ml, CSL, 0.1% BSA in PBS) was added to the prostate cancer cells and allowed to settle for 10 min. Prostatic fibroblasts (1 ϫ 10 3 ) cultured in 5% FBS RPMI in parallel with the prostate cancer cells were used as positive control. Pericellular sheath formation was quantitated from digital photographs using the Video Pro image analysis system (Leading Edge P/L, Marion, South Australia). The area delimited by red blood cells and the area delimited by the cell membrane was measured to give a sheath to cell ratio. A ratio of 1.0 indicates no matrix. Cells with ratio of Ն2.0 were considered to have a pericellular sheath. To demonstrate the dependence on HA for the formation of the pericellular sheath, cells were subsequently treated with Streptomyces hyaluronidase (Hase, 10 units/ml, Sigma-Aldrich) for 20 min at room temperature. The requirement for CS chains in the formation of pericellular sheath was examined by preincubation of fibroblast CM with chondroitinase ABC (ChABC, 0.1 units/ml, Sigma-Aldrich) for 2 h at 37°C. ChABC activity was subsequently inactivated by treating CM at 95°C for 5 min. Digestion of CS chains and HA by ChABC was determined by dot blot analysis using a mouse monoclonal antibody to native CS (CS-56, Sigma-Aldrich) or biotinylated hyaluronan-binding protein (HABP, 2 g/ml, Seikagaku Corp, Japan) and visualized with anti-mouse IgG peroxidase-conjugated secondary antibodies (Dako Labs) or streptavidin-horseradish peroxidase conjugate (Dako Labs), respectively, followed by enhanced chemiluminescence (ECL, Amersham Biosciences).
CD44 and Bromodeoxyuridine (BrdU) Double Immunofluorescence-The relationship between CD44 expression and PC3 cell proliferation was investigated by double labeling immunofluorescence with CD44 and BrdU antibodies. PC3 cells (3 ϫ 10 3 cells/well) were plated in 8-well tissue culture chamber slides (Nunclon TM Lab-Tek II Chamber slide, RS Glass Slide, Naperville, IL) in 500 l of 5% FBS RPMI for 48 h and treated for 24 h with control medium (5% FBS RPMI, CHO K1 CM, 0.1% BSA RPMI) or versican-containing medium (fibroblast CM, CHO V1 CM, purified versican from prostatic fibroblasts) as described for the particle exclusion assay. The cells were supplemented with 20 M BrdU for the last 4 h of treatment. The proportion of cells forming a pericellular sheath were determined by the particle exclusion assay prior to fixation for 10 min in 4% paraformaldehyde and methanol (Ϫ20°C) for 15 min. Slides were treated with 1 M HCl for 10 min to denature DNA (15) and after neutralizing in PBS, pH 7.4, cells were blocked with 5% goat serum and incubated overnight with anti-BrdU conjugated with Alexa Fluor 594 (2.5 g/ml, Molecular Probes, Eugene, OR) and anti-CD44 (Clone 156-3C11, 2.5 g/ml) antibodies, and then incubated for 1 h in the dark at room temperature with goat anti-mouse immunoglobulins conjugated with Alexa 488 (Molecular Probes). The nuclei were stained with Hoescht dye (5 g/ml, Sigma) for 15 min and mounted with fluorescent mounting medium (Dako Labs). Cells were viewed with an epifluorescence microscope (BX50, Olympus Australia) and imaged using a Spot RT digital camera (Diagnostic Instruments, Sterling Heights, MI).
Boyden Chamber Motility Assays-Chemotactic motility assays were performed in a modified Boyden chamber assay as described previously (16). PC3 cells (5 ϫ 10 5 cells/ ml) were harvested, washed with RPMI medium containing 0.1% BSA (serum-free RPMI), and treated with purified versican from prostatic fibroblast CM (pooled fractions 9 ϩ 10 diluted to 0.4 units/ml) or aggrecan (0.1 mg/ml) Ϯ HA (20 g/ml) or serum-free RPMI Ϯ HA in the absence of versican or aggrecan for 2 h on an oscillating platform (Nutator, Clay Adams, Becton Dickinson, Sparks, MD) at room temperature to enable pericellular sheath formation. An aliquot (20 l) of these cells from each treatment were mixed with 400 l of a suspension of human red blood cells (10 7 /ml, 0.1% BSA in PBS) FIGURE 2. A, pericellular sheath formation by fibroblasts in particle exclusion assay. Prostatic fibroblasts were plated at a density of 1 ϫ 10 3 cells/well in 48-well plates and cultured in 5% FBS RPMI (control RPMI) for 48 h. The same cell was treated with hyaluronidase (Hase, 10 units/ml) for 20 min at room temperature. Red blood cell diameter: 7 m. Magnification: ϫ523. B, pericellular sheath formation by prostate cancer cells following 24 h of treatment with prostate fibroblast CM in the particle exclusion assay. Prostate cancer cells (3 ϫ 10 3 ) were plated in 48-well tissue culture and treated with control RPMI or prostate fibroblast CM (versican 0.1 unit/ml). Red blood cell diameter: 7 m. Magnification: ϫ369. The white arrows in B illustrate PC3 and DU145 cells with a prominent polar pericellular sheath. and allowed to attach for 2 h at 37°C in the presence or absence of Hase (10 units/ml) for assessment of sheath formation. For the motility assay, the lower chambers were filled with serumfree RPMI containing fibronectin (10 g/ml, bovine plasma, Sigma). The PC3 cells (1 ϫ 10 5 , 200 l) as treated above, were added directly to the top of 12-m pore size polyvinylpyrrolidone-free polycarbonate filters (Nucleopore, Pleasanton, CA) and allowed to migrate for 4 h at 37°C in a 5% (v/v) CO 2 incubator. An additional control included treatment of prostate cells with Hase (10 units/ml) during the 4-h migration assay. The filters were stained by Diff-Quik (Baxter, Miami, FL), and migrated cells were counted by light microscopy at ϫ120 magnification in ten random fields. Data are expressed as percentage of control Ϯ S.D. Wound Migration Assays and Time Lapse Photography-Wound migration assays were performed using PC3 cells (2 ϫ 10 4 cells/well) cultured in 8-well chamber slides (Nuclon TM Lab-Tek II Chamber slide, RS Glass Slide) in 500 l of 5% FBS RPMI for 3-4 days. The resulting confluent cell monolayers were wounded with a small spatula coated with parafilm, washed to remove floating cells prior to treating with CHO K1 CM or CHO V1 CM (0.5 units/ml versican). Cell migration was monitored over a 24-h treatment period using an IX 81 microscope (Olympus) equipped with a 37°C incubator (Solent Scientific, Segensworth, UK) and aerated with 5% CO 2 in oxygen. Directional movement and pericellular sheath formation were assessed by time lapse and particle exclusion assays. Cells migrating into the wounded area were imaged over a 24-h period in the presence of red blood cells (10 7 /ml) using Analysis software (Soft Imaging System, Johann-Krane-Weg Munster, Germany). The number of cells migrating, the direction of motion, and proportion of cells forming a pericellular sheath was assessed.
To assess CD44 and HA abundance and localization in motile PC3, the cells were fixed for 10 min in 4% paraformaldehyde and methanol (Ϫ20°C) for 15 min prior to immunofluorescence staining. Slides were blocked with 5% goat serum and incubated overnight with anti-CD44 (2.5 g/ml) and biotinylated HABP (1 g/ml). Cells were incubated for 1 h in the dark at room temperature with goat anti-mouse immunoglobulins conjugated with Alexa 488 and streptavidin conjugated with Alexa 594 and subsequently counterstained with Hoechst dye (5 g/ml). The slides were mounted with fluorescent mounting medium and imaged as described above.
Statistical Analysis-The Student's t test or one-way analysis of variance test and the Dunnet t post hoc test were used to determine statistical significance between control and treatment groups. All analyses were performed using SPSS 11.0 for Windows Software (SPSS Inc., Chicago, IL). Statistical significance was accepted at p Ͻ 0.05.

Pericellular Sheath Formation by Cultured Prostate Cancer
Cells with Prostatic Fibroblast CM-Human prostate fibroblasts produced extensive pericellular sheaths which were entirely removed by digestion with Hase ( Fig. 2A). Greater than 90% of the prostatic fibroblasts displayed HA-dependent matrices, with a mean positive cell to sheath ratio of 2.8 Ϯ 0.9. Prominent pericellular sheaths were also assembled by PC3 and DU145 prostate cancer cells following 24 h of treatment with CM from prostate fibroblasts (Fig. 2B). Pericellular sheaths were not detected around LNCaP prostate cancer cells either in the absence or presence of fibroblast CM (Fig. 2B). Approximately 17% of PC3 (mean positive cell to sheath ratio ϭ 2.9 Ϯ 0.9) and 14% of DU145 (mean positive cell to sheath ratio ϭ 2.6 Ϯ 0.9) cells displayed pericellular sheaths following treatment with fibroblast CM (0.1 units/ml versican, Fig. 3A). In comparison, less than 5% of these cells exhibited pericellular sheaths when cultured in 5% FBS RPMI (Fig. 3A). The proportion of PC3 cells assembling a pericellular sheath was reduced with increasing dilution of fibroblast CM, reaching control levels when diluted at 1:10 (Fig. 3B).
ChABC digestion of the fibroblast CM did not affect pericellular sheath formation by PC3 cells (Fig. 3C). Complete digestion of CS chains in the fibroblast CM was confirmed by dot blot using a CS antibody (Fig. 3D). The requirement of HA for assembly of the pericellular sheath by PC3 cells was demon- strated following treatment with Hase (Fig. 3E). Hase treatment completely removed detectable HA from fibroblast CM; however, the ChABC digestion conditions used did not affect HA binding (Fig. 3D). The polarity of pericellular sheath formation was assessed in photomicrographs of PC3 following versican treatment. Of the 100 cells examined with pericellular sheath, 89 cells exhibited clear asymmetry or polarity of the pericellular sheath as shown in Figs. 2B and 3E.
Pericellular Sheath Formation by PC3 Cells Treated with Purified Versican-Treatment of cultured PC3 cells with purified versican from prostatic fibroblasts (pooled fractions 9 ϩ 10 diluted to 0.4 units/ml) in the absence of exogenous HA resulted in pericellular sheath formation to the same extent as neat fibroblast CM (0.25 units/ml versican) and purified aggrecan (0.1 mg/ml) (Fig. 4A). Pericellular sheath was also formed by PC3 cells following treatment with CM from CHO cells expressing recombinant V1 (CHO V1 CM, 1 unit/ml versican) but not with parental CHO K1 CM. Pericellular sheath formation was again completely removed by treatment with Hase (data not shown). Using an HA competitive binding assay, we found that PC3 cells secreted HA to 40% of the level secreted by prostate fibroblasts. In contrast, minimal HA was found to be produced by DU145 and LNCaP cells (Fig. 4B). The dependence of sheath formation on versican concentration was confirmed following treatment with increasing concentrations of purified rV1 versican (Fig. 4C). Maximal sheath formation by PC3 cells (32 Ϯ 16%) was observed with 0.25 units/ml rV1 versican.
Integration of HA and Versican into Pericellular Sheath-HA production by PC3 was confirmed by histochemical staining with biotinylated HABP. No HA staining was observed in DU145 or LNCaP cells (Fig. 5A). Histochemical staining with biotinylated HABP following treatment with prostate fibroblast CM detected a zone of HA surrounding PC3 and DU145. The HA staining was absent in LNCaP cells and PC3 or DU145 cells pretreated with Hase (10 units/ml, Fig. 5B). Pericellular versican staining corresponding to the region stained for HA was FIGURE 5. CD44 expression and composition of pericellular sheath formation by prostate cancer cells. A, prostate cancer cells treated for 24 h with control 5% FBS RPMI. Cells were fixed and incubated with biotinylated HABP or CD44 monoclonal antibody. Immunostaining detected by immunoperoxidase and counterstained with hematoxylin. Magnification: ϫ338. B, prostate cancer cells treated with prostate fibroblast CM containing 5% FBS and incubated with biotinylated HABP Ϯ Hase (10 units/ml or specific rabbit antibody to versican. Immunostaining detected by immunoperoxidase and counterstained with hematoxylin. Magnification: ϫ338. Insets demonstrate cells with strong polar CD44 immunostaining (Magnification: ϫ677). C, double immunofluorescence labeling with BrdU and CD44 monoclonal antibodies. PC3 prostate cancer cells treated with 5% FBS RPMI or prostate fibroblast CM (versican 0.2 unit/ml). Nuclei are counterstained with Hoescht dye. Nuclei that have incorporated BrdU are shown as pink, and CD44 membrane immunostaining is green. Magnification: ϫ338. Insets demonstrate cells with strong polar CD44 immunostaining (magnification: ϫ677). observed for PC3 and DU145 but LNCaP cells following treatment with prostate fibroblast CM (Fig. 5B). No versican staining was observed in prostate cancer cell lines treated with RPMI containing 5% FBS (data not shown).

Expression of the CD44 HA Receptor by Prostate Cancer Cell Lines and Association with Cell Proliferation-CD44
was immunolocalized to the cell membrane of 37 Ϯ 6% of PC3 cells and 39 Ϯ 8% of DU145 when cultured in RPMI containing 5% FBS under the same conditions as the particle exclusion assay (Fig. 5A). CD44 immunostaining was absent in LNCaP cells (Fig. 5A). Double labeling with BrdU and CD44 monoclonal antibodies was performed to determine if CD44 expression was associated with cell proliferation (Fig. 5C). Neither CD44 expression nor BrdU incorporation was significantly altered following treatment with fibroblast CM (0.2 unit/ml), CHO V1 CM (1 unit/ml) or purified prostatic versican (0.24 unit/ml) compared with control medium (Fig. 6). CD44 immunostaining was present in 25-34% of PC3 cells when treated with control or versican-containing medium and whereas 30 -48% of these PC3 incorporated BrdU (pink nuclei) only 8 -10% of cells both incorporated BrdU and demonstrated immunoreactive CD44 (pink nuclei with green membranous CD44 expression (Figs. 5C and 6). The percentage of PC3 cells with immunoreactive CD44 closely parallelled the capacity to form a pericellular sheath in the different versican-containing media. Interestingly, some of the cells in Fig. 5, A and C exhibited polar CD44 immunostaining.
Pericellular Sheath Formation Enhances Prostate Cancer Cell Motility-A 2-h treatment of PC3 cells with purified versican (0.4 unit/ml) and HA (20 g/ml) at room temperature on an oscillating platform resulted in sheath formation by 31% of cells as determined by a subsequent particle exclusion assay (Fig. 7, A  and B). Pericellular sheath formation by PC3 cells was significantly reduced following treatment with Hase (Fig. 7B). No sheath was formed by PC3 in the presence of HA alone, and in the presence of versican alone, only 17% of PC3 formed a peri-cellular sheath (data not shown). Treatment of PC3 cells with aggrecan (0.1 mg/ml) and HA (20 g/ml) also induced sheath formation, which was reduced following Hase treatment (data not shown). In Boyden chamber assays, the PC3 cells treated with purified versican and HA demonstrated a significantly increased motility rate (203% of control, p ϭ 0.001) compared with cells treated with vehicle (PBS) and HA alone (Fig. 7, C and  D). The increased motility was reversed to control levels in the presence of hyaluronidase (Fig. 7D). Similar results were obtained with an independently purified versican fraction from prostatic fibroblasts. An increased motility rate (187% of control) was also observed following treatment of PC3 cells with aggrecan and HA (data not shown).
Polar Pericellular Sheath Formation and CD44 Expression in Migrating PC3 Cells-A wound migration assay was used to determine whether directional PC3 cell motility was associated with polar pericellular sheath formation following versican treatment. Treatment with CHO V1 CM (0.5 unit/ml versican) significantly increased by 3-fold the number of PC3 cells entering the wounded area after 24 h compared with treatment with CHO K1 CM containing no versican (Fig. 8A). Using the particle exclusion assay up to 40% of PC3 cells (52/130 cells examined in two separate experiments) that migrated into the wound demonstrated pericellular sheath formation following treatment with CHO V1 (Fig. 8B). All cells with pericellular sheath exhibited a polar pattern as shown in Fig. 8B. In addition, we found that up to 30% of PC3 cells (15/50 cells examined from two separate experiments) that migrated in the wound exhibited polar CD44 expression following treatment with CHO V1 CM (Fig. 8C). Polar CD44 expression was not observed in PC3 cells treated with CHO K1 CM. Double labeling with CD44 antibody and biotinylated HABP demonstrated polar CD44 expression and polar HA staining in the same cells following treatment with CHO V1 CM (Fig. 8C).
Time lapse photography combined with the particle exclusion assay enabled us to simultaneously observe directional cell movement and pericellular sheath formation. Polar pericellular sheath formation by motile PC3 cells migrating into the scratch wound over a 2-h treatment period ( Fig. 9 and supplemental materials) was consistently observed. Pericellular sheath was particularly evident at the trailing edge and absent from the leading edge of motile PC3 cells (Fig. 9, white asterisks). No pericellular sheath was observed in non-motile cells (Fig. 9, black asterisks). This finding was confirmed in PC3 cells assessed in three separate experiments.

DISCUSSION
In this study we demonstrate that prostate cancer cells in vitro can utilize the ECM components, HA, and versican, secreted by prostatic fibroblasts to assemble a pericellular matrix and promote their motility. Activation of the host stromal microenvironment is a critical step in tumor metastasis. Stromal cell activation or desmoplasia has been described for many adenocarcinomas including prostate, colon and breast (17) and is characterized by extensive synthesis of ECM components such as collagen, fibronectin, versican, HA, tenascin C, and matrix metalloproteinases (4,7,18,19). A reactive stroma is believed to play an important role in prostate cancer progres- sion and invasion. Prostate stromal cells stimulate the development and modulate the rate of prostate tumorigenesis in mouse xenograft models (20,21). Whereas the specific mechanisms are not known, a strong association between development of prostate cancer metastasis and increased versican or HA in the peritumoral stroma (4,22,23) has implicated these molecules in prostate cancer motility and invasion. The present study supports the role of stromal versican and HA in prostate cancer cell motility.
The structure of versican lends itself to multiple types of interactions through protein-protein and protein-carbohydrate interactions (24). The binding to HA via the G1 domain of versican is well characterized (12). HA is elevated in several cancers, including colon, breast, prostate, bladder, and lung and has been shown to have a critical role in anchorage-independent growth in culture and epithelial mesenchymal transition, both indicators of tumorigenicity in vivo (25,26). HA interacts with several cell-surface receptors, the most common being CD44. HA/CD44 interactions have been shown to induce intracellular signal transduction events, cell aggregation, cell proliferation, migration, and adhesion (27,28). In normal prostate, HA/CD44 interactions are of paramount importance for ductal branching morphogenesis and epithelial growth following androgen treatment (29), and this normal process may become uncontrolled in metastatic disease. An accumulation of HA has been associated with the active invasion of ovarian and mammary carcinoma cells into the mesentery (30). In the study by Yeo et al. (30) mesenteric cells were determined to be the source of the HA as neither the ovarian nor the mammary carcinoma cells exhibited the capacity to synthesize HA in vitro. Whereas versican expression was not examined in these earlier studies, the demonstration that both versican and HA expres- sion are associated with metastatic human prostate cancer (4,23) suggests that when both versican and HA are overexpressed, they act together to promote cancer motility and metastasis. This notion is supported by the findings that versican and HA expression are strongly correlated in tumors (7) (31). An interaction between versican and HA also appears to be critical for the migration of endocardial cushion cells heart development (32) (33). Endocardial cushion cells failed to migrate in both the heart-defect mouse which has a disruption of the versican gene (32) and in the HA synthase 2 knock-out mouse (33). The findings in this study demonstrating that both HA and versican are required for the formation of pericellular sheath and increased motility of prostate cancer cells support their involvement in cancer motility and metastasis.
In this study, PC3 and DU145 but not LNCaP prostate cancer cells were able to form a HA/versican pericellular sheath. The formation of a HA/versican pericellular sheath required cellular expression of CD44, which also correlates with the ability of prostate cancer cell lines (PC3 and DU145) to form metastatic deposits in nude mice (34 -36). The capacity to form a HA/versican pericellular sheath was dependent on the concentration of versican and associated with the level of CD44 expression. CD44 expression was not altered following treatment with CM or purified versican and was not directly associated with cell proliferation. The results in this study indicate that CD44 expression is asynchronous with cell replication and suggest that factors other than the proliferative state of the cell are important in regulating CD44 expression. CD44 immunostaining was polarized on cells in a manner similar to the polarized formation of pericellular sheaths. Polarized CD44 immunostaining has also been described in uropods of migrating neutrophils (37), on the trailing edge of fibroblasts (38) and on the surface of metastatic mouse mammary carcinoma cells (39). Our results suggest that formation of a CD44/HA/versican macromolecular complex aids the motility of prostate cancer cells.
PC3 cells produce endogenous HA and are able to form pericellular sheaths in the presence of exogenous purified versican alone. Similar to our finding, highly metastatic prostate cancer cells PC3 and PC3M-LN4 have been shown to assemble a pericellular sheath upon addition of aggrecan, which resulted in a rapid and specific adhesion to bone marrow endothelial cells (14). However because aggrecan is not produced by prostatic cells, the relevance of these findings to the prostate is not clear, except that it illustrates tissue-specific situations where different HA-aggregating proteoglycans function in a similar manner to induce matrix stabilization. Prostate stromal cells are known to produce versican, and indeed this is up-regulated by tumorderived factors (40) suggesting that induction of versican may be a key aspect of stromal "activation" by tumor cells and tumor progression to the metastatic state. Our results suggest that this is through enhancement of cell motility.
Cell movement requires coordination of biophysical processes including membrane extension, formation of new attachments at the leading edge, generation of contractile force, and detachment at the trailing edge of cells (41). Our recent observation that versican can inhibit prostate cancer cell attachment to fibronectin (8) together with the findings in this study that formation of a HA/versican pericellular sheath promotes prostate cancer cell motility supports a role for versican and HA in tumor cell detachment and motility. Transfection of the HA binding G1 domain of versican can reduce astrocytoma cancer cell adhesion and increase migration of these cells (39). However, the mechanisms whereby HA and versican can mediate their effects on migration are not clear. Studies by Evanko et al. (11) using time-lapse video microscopy have demonstrated that an HA coat incorporating versican is obligatory for the proliferation and migration of smooth muscle cells in vitro.
A pericellular matrix rich in HA/versican may increase cell motility by mediating cell detachment to other matrix compo- nents. In this study, we have demonstrated that formation of an HA/versican pericellular matrix promoted prostate cancer motility in Boyden chamber motility assays using fibronectin as a chemoattractant. In addition we demonstrated polar pericellular sheath formation particularly at the trailing edge of motile PC3 cells in wound migration assays, while there was no evidence of pericellular sheath at the leading edge of cells. We propose that tumor cells form a polarized pericellular sheath through compartmentalized cell surface CD44 expression and subsequent assembly of HA/versican aggregates. The absence of a pericellular sheath at the leading edge of the cell may allow attraction and binding to ECM components such as fibronectin while formation of a pericellular sheath at the trailing edge of cells inhibits cellular binding to ECM components. The combination of binding at the leading edge of cells and inhibition of binding at the trailing edge may enhance the forward motion of the cells. This notion is supported by the demonstration in this study of polarized pericellular sheath formation in motile prostate cancer cells. The description of abundant versican in the trailing tracks of migrating fibroblasts (42) and pericellular sheath at the trailing edge of migrating smooth muscle cells (11) also supports this model.
In summary, we have demonstrated that prostate cancer cells in vitro have the ability to recruit ECM components produced by prostatic stromal cells to promote their motility. These findings are consistent with a scenario whereby the formation of a pericellular sheath in vivo by prostate cancer cells utilizing versican and HA laid down by prostate stromal cells may contribute to the development of metastatic disease.