Inducible Hyaluronan Production Reveals Differential Effects on Prostate Tumor Cell Growth and Tumor Angiogenesis*

Prostate cancer progression can be predicted in human tumor biopsies by abundant hyaluronan (HA) and its processing enzyme, the hyaluronidase HYAL1. Accumulation of HA is dictated by the balance between expression levels of HA synthases, the enzymes that produce HA polymers, and hyaluronidases, which process polymers to oligosaccharides. Aggressive prostate tumor cells express 20-fold higher levels of the hyaluronan synthase HAS3, but the mechanistic relevance of this correlation has not been determined. We stably overexpressed HAS3 in prostate tumor cells. Adhesion to extracellular matrix and cellular growth kinetics in vitro were significantly reduced. Slow growth in culture was restored either by exogenous addition of hyaluronidase or by stable HYAL1 coexpression. Coexpression did not improve comparably slow growth in mice, however, suggesting that excess hyaluronan production by HAS3 may alter the balance required for induced tumor growth. To address this, we used a tetracycline-inducible HAS3 expression system in which hyaluronan production could be experimentally controlled. Adjusting temporal parameters of hyaluronan production directly affected growth rate of the cells. Relief from growth suppression in vitro but not in vivo by enzymatic removal of HA effectively uncoupled the respective roles of hyaluronan in growth and angiogenesis, suggesting that growth mediation is less critical to establishment of the tumor than early vascular development. Collectively results also imply that HA processing by elevated HYAL1 expression in invasive prostate cancer is a requirement for progression.

Prostate cancer progression can be predicted in human tumor biopsies by abundant hyaluronan (HA) and its processing enzyme, the hyaluronidase HYAL1. Accumulation of HA is dictated by the balance between expression levels of HA synthases, the enzymes that produce HA polymers, and hyaluronidases, which process polymers to oligosaccharides. Aggressive prostate tumor cells express 20-fold higher levels of the hyaluronan synthase HAS3, but the mechanistic relevance of this correlation has not been determined. We stably overexpressed HAS3 in prostate tumor cells. Adhesion to extracellular matrix and cellular growth kinetics in vitro were significantly reduced. Slow growth in culture was restored either by exogenous addition of hyaluronidase or by stable HYAL1 coexpression. Coexpression did not improve comparably slow growth in mice, however, suggesting that excess hyaluronan production by HAS3 may alter the balance required for induced tumor growth. To address this, we used a tetracycline-inducible HAS3 expression system in which hyaluronan production could be experimentally controlled. Adjusting temporal parameters of hyaluronan production directly affected growth rate of the cells. Relief from growth suppression in vitro but not in vivo by enzymatic removal of HA effectively uncoupled the respective roles of hyaluronan in growth and angiogenesis, suggesting that growth mediation is less critical to establishment of the tumor than early vascular development. Collectively results also imply that HA processing by elevated HYAL1 expression in invasive prostate cancer is a requirement for progression.
Despite improved detection and diagnosis of prostate cancer, this disease remains the second leading cause of malignant mortality in United States males (1,2). Metastasis may occur with no prior indication of an invasive tumor (3), so prostate cancer progression is difficult to predict. Changes in levels of extracellular matrix molecules such as hyaluronan (HA), 2 a high molecular weight polysaccharide, and its processing enzyme, the hyaluronidase HYAL1, within the prostate extra-cellular matrix have been correlated to invasive prostate cancer progression (4 -11). The molecular mechanisms underlying this correlation could provide important insights for therapeutic development or improved diagnosis of prostate cancer.
HA production is a tightly regulated process that impacts cellular transformation and motility during development (12)(13)(14). Dynamic HA turnover within tissues controls many acute processes such as wound healing or immune function. HA accumulation is the outcome of a balance between the activity of HA synthases (HAS), enzymes that synthesize the linear polymers (15), and hyaluronidases, which process the polymers to biologically potent oligosaccharides (16). Excess quantities of large HA polymers have been reported to suppress cellular growth (17,18) and angiogenesis (19,20), whereas processed oligosaccharides dramatically stimulate angiogenesis (21)(22)(23), and fully degraded oligosaccharides induce apoptosis (14,24,25). HAS isozymes (HAS1, HAS2, and HAS3) have been overexpressed in several tumorigenic cell lines and may impact tumor growth kinetics in a dose-dependent fashion (26). For example, relatively low overexpression of HAS2 augments tumorigenesis, whereas high levels of HAS2 expression either have no effect or suppress growth of subcutaneous tumors.
Interestingly a similar effect was recently shown for the hyaluronidase HYAL1 (5), originally identified as a tumor suppressor (27) despite its subsequent direct correlation with cancer (8,28,29). HA turnover is intricately orchestrated by the hyaluronidases HYAL1 and HYAL2 (30 -32) in conjunction with the HA receptor CD44 (33,34). HYAL1 is secreted and deposited to the extracellular space where it may be retained non-covalently by binding to HA (31). HYAL2 is glycosylphosphatidylinositolanchored at the cell surface, localized to microdomains with CD44 (35), stimulation of which signals HA uptake subsequent to its initial extracellular processing by HYAL1/2. HYAL1 exhibits maximal activity at acidic pH, whereas HYAL2 is active in both acidic and neutral conditions, consistent with intracellular lysosomal function of both enzymes upon internalization (36,37). However, the presence of locally acidic microdomains at individual cell surfaces (35) and within rapidly developing tumors (38,39) may promote inappropriate activation, particularly of HYAL1, in the extracellular compartment. Differential hyaluronidase activity may thereby translate to a gradient of angiogenic and apoptotic oligosaccharides. The importance of a balance between expression levels of HA biosynthetic and processing enzymes for tumor cell growth is further suggested by the finding that HAS2 overexpression may promote growth in cell types with significant hyaluronidase activity, whereas it inhibits growth of cells lacking hyaluronidase (40).
We have shown previously that aggressive prostate tumor cell lines secrete excess HA and retain it at the cell surface in large matrices, a property that differentially mediates the interaction of the tumor cells with stromal and epithelial cell types (41,42). Production of excess HA by prostate tumor cells is catalyzed by elevated levels of HAS2 and HAS3 specifically in aggressive cells. Inhibition of HAS expression in these cells interferes with tumorigenesis, angiogenesis, and metastasis, so the HA product clearly has a role in these processes in a prostate tumor model (43). When we coexpressed HAS2 and HYAL1 in stable tumor cell lines, resultant subcutaneous tumor sizes were 2-3-fold larger than those arising from either HAS2 or HYAL1 individually transfected cells (44), suggesting that combined production and processing of HA contribute synergistically to tumor growth mechanisms.
Because HAS3 is the most dramatically up-regulated isozyme in metastatic prostate tumor cells, we reasoned that its up-regulated expression would have a demonstrable influence on HA matrix production and sought to test the functional relevance of this correlation. We report here the first overexpression of HAS3 in prostate tumor cells. Stable lines were selected for HAS3 overexpression or for HYAL1/HAS3 coexpression to assess the respective outcomes in mice. Because HAS3 overexpression suppressed tumorigenesis, we additionally developed a tetracycline-inducible system for HAS3 overexpression and directly evaluated the correlation between HA production and prostate tumor cell growth in vitro. Our findings support a model in which excess HA suppresses tumorigenesis by diminishing apparent cellular growth and limiting angiogenesis. HYAL1 is thus implicated in promotion of angiogenesis by dissipation of accumulated HA within tumors, which may facilitate its recognition as an angiogenic signal.

EXPERIMENTAL PROCEDURES
Cell Culture, Materials, and Reagents-22Rv1 human prostate adenocarcinoma cells (45) were purchased from ATCC and maintained in RPMI 1640 medium containing 10% fetal bovine serum. Selection, characterization, and maintenance of 22Rv1 stable transfectants in standard medium containing 1.25 mg/ml G418 has been described previously (44). The pIRES2-EGFP vector and plasmids pTet-On and pTRE-Tight that comprise the tetracycline-inducible mammalian expression system were obtained from Clontech. Doxycycline and Tet Systemapproved fetal bovine serum were also from Clontech. Anti-FLAG antibody was purchased from Stratagene (La Jolla, CA). Collagens I and IV were from BD Biosciences. Fibronectin, laminin, and calcein-AM were from Invitrogen, and antibodies to ␣ and ␤ integrins were from Chemicon (Temecula, CA). Human umbilical cord HA, Alcian blue, and tetramethylbenzidine were from Sigma. Streptomyces hyaluronidase was purchased from Calbiochem (EMD Biosciences). Biotinylated hyaluronan-binding protein (HABP) was from Seikagaku (Associates of Cape Cod, East Falmouth, MA).
Transfection and Stable Selection of Prostate Tumor Cells-The plasmid constructs encoding FLAG epitope-tagged HYAL1 or HAS3 were described previously (42,44). 22Rv1 human prostate tumor cells were transfected with 1) vector alone (pIRES2-EGFP), 2) plasmid encoding HYAL1-FLAG, 3) HAS3-FLAG plasmid, or 4) both the HYAL1 and the HAS3 plasmid constructs. Liposome-mediated transfection with FuGENE 6 reagent (Roche Diagnostics) was performed according to the manufacturer's protocol. Gene expression was verified by Western blot and by enzyme activity assays as described. Stable 22Rv1 transfectants were obtained by clonal isolation following incubation for 14 -20 days with G418 (1.25 mg/ml). Multiple clones were tested for expression, and those with average levels of expression were pooled to eliminate the bias from isolated selection. Once established, stable transfectants were maintained in the selection medium.
Western Analysis of Transgene Expression-Conditioned media from cells cultured in 10-cm plates were prepared for Western analysis by diluting 1:1 with SDS gel loading buffer and boiling for 5 min. To prepare whole cell lysates, cultured cells were washed with phosphate-buffered saline and lysed by addition of 0.5 ml of hot SDS gel loading buffer. Cellular lysates were homogenized by shearing through a 26-gauge needle, incubated at 37°C for 20 min, and centrifuged. All samples were normalized for protein content or cell number as indicated. Equal amounts of each sample were separated by SDS-PAGE and transferred to Immobilon-P membrane. After blocking, membranes were probed with anti-FLAG M2 or anti-VP16 antibody as appropriate (2 g/ml), and blotted proteins were detected by enhanced chemiluminescence. For densitometric analysis, samples loaded in triplicate were blotted. Western blots were scanned and converted to grayscale in Adobe Photoshop. Pixel density in identically sized regions encompassing each band was used to calculate the mean Ϯ S.E. for expression levels in HYAL1/HAS3 cotransfectants relative to HAS3 transfectants.
Hyaluronidase Activity Assays-Overnight conditioned media from stably transfected cell lines were concentrated 10-fold and electrophoresed by SDS-PAGE on a 12% polyacrylamide gel containing 0.2 mg/ml HA. After a 1-h incubation at room temperature in 3% Triton X-100, the gel was placed in hyaluronidase assay buffer (50 mM sodium formate, pH 4.0, 150 mM NaCl, and 0.1% bovine serum albumin) at 37°C overnight. Hyaluronidase activity was detected as a clear band at the expected molecular weight for HYAL1 upon staining 1 h with 0.5% Alcian blue and destaining with 7% acetic acid. Cell lysates were assayed similarly, and although several light bands were visible, no significant differences were observed between control and HYAL1-transfected cells indicating that the majority of overexpressed HYAL1 was secreted.
To quantify hyaluronidase activity, we used a microplate assay essentially as described previously (6). Briefly serial dilutions of concentrated conditioned media normalized for protein content were applied in triplicate to HA-precoated microwell plates. Following a 1-h incubation in hyaluronidase assay buffer, wells were washed and developed with biotinylated HABP followed by avidin-biotin-horseradish peroxidase with tetramethylbenzidine substrate. Absorbance at 650 nm was used to interpolate specific activity from a concurrent Streptomyces hyaluronidase standard curve. In competition assays, exogenous HA (average molecular mass of Ϸ20 kDa; Lifecore, Chaska, MN) was included in the 1-h plate incubation.
HA Quantification-HA content of transfected cell culture supernatants was determined by competitive binding assay (6,43,46). Overnight conditioned media from prostate tumor cell cultures were harvested, and cells were counted. Equal volumes of serially diluted media were combined with biotinylated HABP (0.5 g/ml) and incubated in HA-precoated microtiter plates for 6 -8 h. Plates were developed using avidin-biotinhorseradish peroxidase with tetramethylbenzidine as substrate, and absorbance was read at 650 nm. HA concentration was interpolated from a standard curve and normalized to cell number.
HA Pericellular Detection and Quantification-Pericellular HA retention was visualized and quantified by particle exclusion as described previously (42,43). Briefly cell cultures were incubated with 2 mg/ml aggrecan followed by addition of 1 ϫ 10 8 fixed red blood cells. After 15 min, cells were viewed with phase-contrast microscopy and digitally photographed at 400ϫ magnification. Matrix retention was quantified using Adobe Photoshop to calculate relative areas for matrices and cellular boundaries of individual cells. HA matrix thickness is presented as the ratio of matrix area to cell area for each transfectant or condition.
Growth Assays-Two-dimensional growth in culture was assayed as described previously (43). Equivalent passages of each tumor cell line were plated at 5000 cells/well in 24-well plates. At 24-h intervals, quadruplicate wells of each cell line were released with trypsin, neutralized, and manually counted in a hemacytometer. Duplicate counts for each well were averaged to obtain the total cell count per well. Each point is the mean Ϯ S.E. of the total cell counts. Although the graph presents data from a single assay in quadruplicate, the growth trends were reproduced in three additional identical assays.
Mouse Subcutaneous Injection-All mice were cared for and maintained under the supervision and guidelines of the University of Nebraska-Lincoln Institutional Animal Care and Use Committee. Male NOD/SCID mice (The Jackson Laboratory) (eight animals per condition) were injected subcutaneously in each flank with 1 ϫ 10 6 tumor cells suspended in 100 l of serum-free RPMI 1640 medium. After 28 days, mice were sacrificed, and tumors were dissected and weighed. Mean tumor wet weight ϮS.E. was plotted for each cell line. Statistical significance was assigned by Student's two-tailed t test. The experiment was repeated with an additional eight animals per cell line, and similar results were obtained.
Immunohistochemistry and Immunofluorescence-HA content and vascularization of tumors were detected as described previously (43). Briefly after weighing, tumors were divided in halves. One-half was formalin-fixed and embedded in paraffin. The other half was snap frozen in OCT compound. For HA detection, paraffin-embedded tumors were sectioned, dewaxed, probed with biotinylated HABP, and developed using the Vectastain ABC kit (Vector Laboratories, Irvine, CA). Sections were counterstained with Meyer's hematoxylin to visualize cellular boundaries. White light images were collected at 200ϫ magnification. Vascularization of the tumors was assessed in ethanol-fixed frozen sections (8-m thickness) by antibody staining for CD31 as described previously (43,47).
Angiogenesis Quantification-CD31-phycoerythrin-conjugated antibody staining of frozen sections was visualized by fluorescence microscopy. Five random sections from each of three tumors per cell line were digitally photographed with 5-s exposure time. Images saved as TIF files were converted from 16 to 8 bit in Adobe Photoshop, red channel fluorescence was isolated, images were converted to grayscale and inverted, and a black-and-white threshold was arbitrarily set based on levels. The histogram function was used to determine vessel density as represented by density of black pixels at 0 on the black-to-white scale. Average pixel density for each transfectant tumor section was normalized to the average pixel density for untransfected tumor sections. Statistical significance was assigned by Student's two-tailed t test.
Extracellular Matrix Adhesion and Blocking-Microplates were precoated with collagen I, collagen IV, laminin, or fibronectin in serially diluted doses for 1 h at 37°C and blocked with 3% bovine serum albumin. Subconfluent cells were released with 1 mM EDTA, washed, and resuspended in adhesion medium (serum-free RPMI 1640 medium containing 20 mM HEPES, pH 7.4, and 1% bovine serum albumin), and labeled with calcein-AM (25 g/10 6 cells, 20 min at 37°C). Labeled cells (10 4 cells/well) were incubated in quadruplicate wells of the precoated plates for 2 h at 37°C. Non-adherent cells were removed with three gentle washes in adhesion medium. Remaining adherent cells were lysed with 0.2 N NaOH containing 1% SDS. Fluorescence was measured in a microplate spectrophotometer, and the percentage of adherent cells relative to the input number was calculated from a standard curve of each labeled cell culture. For adhesion specificity tests, cells were preincubated with antibodies against integrins ␤1, ␣1, ␣2, or ␣5 (5 g/ml) for 10 min at 37°C prior to placing them in the substrate-coated microplates. Antibodies were present throughout the assay incubation.
Generation and Characterization of Tet-inducible 22Rv1 Cells-22Rv1 cells were transfected with pTet-On using FuGENE 6 reagent and selected with 1.5 mg/ml G418 until isolated colonies were obtained. Individual colonies were amplified in the continuous presence of G418 and screened for expression of the tetracycline transactivator-VP16 fusion protein by Western blots probed with anti-VP16 antibody. VP16positive clones were then tested for inducible gene expression by transient transfection with a GFP or luciferase reporter plasmid. Transfectants were induced by applying media containing 2 g/ml doxycycline (dox) and digitally photographed or harvested to prepare soluble extracts 48 h later. To characterize HAS3 expression, transient transfectants were initially induced with dox for 48 h at which time conditioned media were assayed for HA content as described above, and membrane-enriched fractions were harvested by mechanical disruption and solubilization in 1% Triton X-100 for Western blotting. In subsequent experiments, control and HAS3 transient transfectants were induced with dox 24 h after transfection and plated (20,000 cells/well) for simultaneous assays of growth rate, HA content, and HAS3 expression as indicated in specific figures. In some experiments, exogenous Streptomyces hyaluronidase (0.2 milliunits/ml final concentration per dose) was added daily throughout the time course. To assay the temporal effects of HA induction, transfectants (20,000 cells/well) were plated and induced identically, but dox was removed after day 4. At that time, all media were removed and replaced with fresh media. In additional experiments, cells were identically plated and assayed, but dox was not added until day 4 at which point all standard media were removed and replaced.

RESULTS
Stable Expression of HAS3 and HYAL1-We previously characterized HAS and hyaluronidase isozyme expression and activity in 22Rv1 cells (44). HAS3 was expressed at basal levels comparable to normal prostate and synthesized relatively little HA. Although reverse transcription-PCR indicated expression of HYAL1 message, its enzymatic activity was very low in comparison with that of HYAL1 transfectants. Because our goal was to determine the respective and concerted functions of HYAL1 and HAS3, 22Rv1 cells were deemed an appropriate model in which to overexpress these enzymes.
Stable lines were selected for overexpression of FLAG epitopetagged HYAL1 or HAS3 from a constitutive eukaryotic promoter in the pIRES2-EGFP bicistronic reporter vector. HYAL1/HAS3 cotransfectants were identified that expressed approximately equivalent levels of HYAL1 with respect to the individual HYAL1 transfectants. Conditioned media from each transfectant were compared by Western blot to the vector control ( Fig. 1A) to verify similar expression levels. Corresponding activity of HYAL1 in the control (lane 1), HYAL1 (lane 2), and HYAL1/HAS3 (lane 3) conditioned media was initially measured by HA substrate gel electrophoresis (Fig. 1B). As expected, the control media contained very little hyaluronidase activity, whereas activity in the HYAL1 transfectant and cotransfectant media (lanes 2 and 3) was significantly increased. We then assayed serial dilutions of transfectant conditioned media in a quantitative microplate assay. Interestingly although the specific activity of HYAL1 transfectant media was constant at all dilutions, the specific hyaluronidase activity of HYAL1/ HAS3 media increased with dilution (Fig. 1C). Thus, the HYAL1 activity in these cells was underestimated at the lower dilutions, and its increasing specific activity implied the dilution of an inhibitor or a competitor. We postulated that HA produced in the media by concurrent expression of HAS3 could be competing with HA immobilized on the microplate and artificially reducing the apparent activity of HYAL1. To confirm this, we added two different concentrations of exogenous HA (molecular mass, Ϸ20 kDa) to the HYAL1 microplate assays (Fig. 1D). At the end point, we measured the remaining HA content of the solution and plotted it as a fraction of exogenous HA added. HYAL1 activity was calculated from the amount of HA remaining on the plate and plotted as a fraction of the activity measured in the absence of exogenous HA (white bar). Although the enzyme was clearly active, degrading both solubilized and immobilized HA, its apparent activity in the microplate was significantly competed by inclusion of excess HA.  These results are consistent with HA content of the HYAL1/ HAS3 media reducing the measured HYAL1 specific activity. Nonetheless using the 1:50 dilution data shown in Fig. 1C we estimated that HYAL1 activity of the HYAL1/HAS cells is only Ϸ1.75-fold higher than that of the HYAL1 cells.
Because HA is synthesized at the plasma membrane and concurrently secreted to the extracellular space, HAS3 activity was verified in the stable lines by quantifying HA in conditioned media and at the cell surface. Stable HAS3 transfectants produced Ͼ30-fold more HA than control GFP-or HYAL1-transfected cells ( Fig. 2A). HYAL1/HAS3 cotransfectant media contained Ϸ9-fold more HA than did controls. Expression levels of HAS3 were compared by Western blot of whole cell lysates probed for the FLAG epitope ( Fig. 2A, inset). Triplicate determinations of each of the three lysates showed that the HYAL1/HAS3 cotransfectants (representative sample in lane 3) expressed 69 Ϯ 3.1% of the HAS3-FLAG quantity detected in single HAS3 transfectants (lane 2). Thus, the HA content of HYAL1/HAS3 cultures was only 30% of the HAS3 culture HA content, less than half of what would be expected from the comparable HAS3 protein level. This result suggests that HYAL1, present in the media of the cotransfected cells, either processes HA as it is synthesized or facilitates its removal from the media. Particle exclusion analysis of surface HA showed significant HA retention by the HAS3-transfected cells (Fig.  2C). HYAL1/HAS3 cotransfectants retained additional HA relative to GFP-transfected controls (compare Fig. 2, panels D and B, respectively). Matrix and cell boundaries were traced in digital images to quantify average cellular HA retention (Fig.  2E). Results are consistent with HA accumulation in the media. Collectively these data confirm that both HAS3 and HYAL1 activities are significant in cotransfectants.
HAS3 Decreases Subcutaneous Growth and Vascular Density of Prostate Tumor Xenografts-To assay the effects of HAS3 and HYAL1 overexpression in prostate tumorigenesis, we injected stable 22Rv1 transfectants subcutaneously in flanks of NOD/SCID mice. Tumor growth was monitored over a period of 4 weeks at which time mice were sacrificed and tumors were excised for analysis. Comparison of the mean tumor wet weight showed essentially identical growth of control and HYAL1 transfectants (Fig. 3A). However, both HAS3 and HYAL1/HAS3 transfectants yielded Ϸ40% smaller tumors than GFP controls. Because we previously found that HYAL1 overexpression increased angiogenesis within subcutaneous tumors, we analyzed all tumors for vessel density by staining tumor sections for CD31, a blood vessel endothelial cell surface marker (Fig. 3B). As expected, HYAL1 transfectant tumors were ϳ2-fold more vascular than controls. Vessel density of HAS3 tumors was 62% of controls, consistent with the extent of reduction in tumor size. Although tumor size did not change appreciably between the HAS3 and HYAL1/HAS3 transfectant tumors, vascularization of the cotransfectants was not statistically different from controls. This result suggests that HYAL1 expression may compensate for suppression of angiogenesis by HAS3.
To determine whether there was a correlation between HA levels and suppression of tumor growth, HA content of the tumors was determined histologically in formalin-fixed sec-tions. The virtual absence of HA, visualized as a brown precipitate among the purple hematoxylin-counterstained cells, was evident in control GFP and HYAL1 transfectant tumors (Fig. 3,  C and D). Tumors arising from HAS3 transfectants exhibited extensive accumulation of pericellular HA (Fig. 3E). HA was detected throughout HYAL1/HAS3 cotransfectant tumor sections, although it was not localized entirely in the pericellular space (Fig. 3F). Thus, significant HA deposition in the tumors correlated with suppression of tumor growth. Concurrent expression of HYAL1 was not sufficient to eliminate HA accumulation in these tumors.
HAS3 Overexpression Diminishes Intrinsic Growth Rate and Adhesion to Extracellular Matrix-To address the mechanisms of reduced tumor growth upon HAS3 overexpression, we first assayed growth kinetics of the stable transfectants in culture. Quadruplicate wells plated with equal numbers of cells in suspension were counted manually each day for 7 days (Fig. 4). The growth rate of HAS3 transfectants (filled circles) was Ͼ60% slower than that of controls (filled squares). Interestingly cotransfectants expressing both HYAL1 and HAS3 (open circles) grew at rates identical to controls. Inclusion of exogenous HA in the culture media of control or cotransfected cells throughout the assay had no effects on growth (not shown). These observations are similar to previously reported growth kinetics of HAS2-transfected and cotransfected cells (44). Thus, the presence of hyaluronan in excess reduces the growth rate specifically of cells that are synthesizing it.
Many adherent cell types, including tumorigenic lines derived from epithelial cells (e.g. 22Rv1), are dependent upon engagement of cell adhesion receptors for survival and growth (48). Impairment of adhesion in anchorage-dependent cells leads to anoikis (49) and apparent reduction in growth kinetics. Because hyaluronan accumulation at the cell surface has the potential to influence cell surface interactions physically, we investigated anchorage dependence of our stable transfectants by assaying adhesion to extracellular matrix (ECM) proteins commonly found in basement membranes and connective tissue. Tumor cell suspensions were first incubated in 96-well plates coated with type IV collagen or fibronectin. Although HYAL1 alone did not alter adhesion to these substrates, HAS3 transfectants were 45% less adherent to type IV collagen and 90% less adherent to fibronectin relative to the GFP control cells (Fig. 5A). Importantly HYAL1/HAS3 cotransfectants exhibited control adhesion levels.
Reduced adhesion of HAS3 cells by Ͼ90% was observed when assayed on type I collagen, the major com-  ponent of subcutaneous connective tissue, and laminin (Fig. 5B, gray bars). Reduced adhesion was not due to retention of HA at the cell surface during the assays because adhesion was not restored when exogenous hyaluronidase was included (Fig. 5B, dark bars). Adhesion to ECM proteins and subsequent engagement of survival mechanisms is mediated at the cell surface by heterodimeric integrin receptors (48 -50). Each integrin consists of an ␣ and a ␤ subunit, both of which combine to form an extracellular ligand-binding domain with well characterized specificity for various ECM proteins. A single transmembrane sequence anchors these receptors at the cell surface, and an intracellular signaling domain modulates cellular responses to ligand binding. We used function blocking antibodies to implicate specific integrins in the differential adhesion we observed. Adhesion to all four substrates was abolished by treatment with anti-␤1 antibodies (Fig. 5, C and D, and data not shown). Inhibition of ␣1 and ␣2 integrins accounted for Ϸ70% of control cell adhesion to type IV collagen (Fig. 5C), consistent with cell surface expression of these receptors in 22Rv1 cells and with the previously reported ligand specificity of the ␣1␤1 and ␣2␤1 integrin heterodimers. Dependence of adhesion on ␣2 integrin was no longer observed for the HAS3 transfectants. The limited adhesion of these cells was mediated completely by ␣1␤1 integrin. Similarly despite rather low expression of ␣5 integrin on the cell surface of any of the transfectants (not shown), adhesion of control, HYAL1, and HAS3 cells to the known ␣5␤1 ligand fibronectin was almost completely inhibited with anti-␣5 blocking antibodies (Fig. 5D). HYAL1/HAS3 transfectant adhesion to fibronectin was only inhibited by Ϸ15% with these antibodies, possibly suggesting that integrin use or activation state is altered in these cells. Collectively results are consistent with compromised cell adhesion by HAS overexpression as an underlying cause of reduced tumor cell growth in vitro and in vivo.

Development and Characterization of Inducible 22Rv1 Cells-To
investigate the relationship between excess HA production and cellular growth phenotype, we established an inducible expression system. Three 22Rv1 clones were selected for stable expression of the VP16 tetracycline (tet) transactivator protein that confers tet-dependent transcription (Tet-On 1, 2, and 3). Of these, 22Rv1 Tet-On 1 expressed the highest levels of the transactivator and was characterized for inducible gene expression by transient transfection with constructs encoding GFP or HAS3 (Fig. 6). The fourpanel series shows the results of GFP transient transfection. Bright field images illustrated comparable confluency of cultures in the absence and presence of the tet analog dox (Fig. 6, A and B). Minimal GFP fluorescence was observed in the absence of dox induction (Fig. 6C), and dramatic GFP fluorescence was induced by addition of dox to the culture medium (Fig. 6D).
Similar transfections were performed to evaluate specific induction of HAS3 expression and HA production. Parental 22Rv1 cells and inducible 22Rv1 Tet-On cells were transiently transfected with tet-responsive GFP or HAS3 constructs as indicated (Fig. 6E). Cells were incubated in the presence (ϩ) or absence (Ϫ) of dox prior to collecting conditioned media and cell lysates. In the inset, lysates from 22Rv1 Tet-On transfectants were analyzed by Western blot probed with anti-FLAG antibody. Robust induction of HAS3 was observed with respect to control GFP and uninduced HAS3 transfectants. Some leakiness in expression was observed as the HAS3 band was also faintly visible in the uninduced lysates. HA content of the media correlated well with expression levels of HAS3 protein. As expected, the inducible construct did not confer HA synthesis to the parental 22Rv1 cells that lack the tet transactivator protein. Transfection of 22Rv1 Tet-On with the inducible GFP construct also did not stimulate HA production. A significant amount of HA was produced by the uninduced HAS3 cells, consistent with HAS3 protein expression, but dox incubation induced HA production by an additional 4-fold.
Induced HA Synthesis Impairs Cell Growth but Can Be Compensated with Exogenous Hyaluronidase-We evaluated growth of the 22Rv1 inducible cells as described above for the stable transfectants (Fig. 7A). 22Rv1 Tet-On cells were transiently transfected with inducible constructs for GFP (squares) or HAS3 (circles). Cells were plated in the absence (open symbols) or presence (filled symbols) of dox. Induction of the GFP control did not impact growth. HAS3 transfection reduced growth somewhat even in the absence of dox because the uninduced cells still produced significant amounts of HA. However, induction of HAS3 further impaired growth of the cells to a total of Ϸ50% overall. Considering that the transfection effi- ciency of the cells was also about 50%, these measurements probably underestimated the full suppressive effect of HAS3 overexpression. Importantly adhesion to fibronectin and type IV collagen was not altered by HAS3 induction (data not shown), suggesting that changes in integrin-mediated adhesion are a stable adaptation to long term elevated HA synthesis.
Stable HYAL1/HAS3 cotransfectants exhibited normal growth, so we tested the cells induced for HAS3 expression to determine whether exogenously added hyaluronidase could restore growth kinetics. 22Rv1 Tet-On cells transiently transfected with the inducible HAS3 construct were plated in the absence (Fig. 7B, white bars) or presence (Fig. 7B, gray and black bars) of dox. By day 2, growth of induced cells was reduced by Ϸ30%. Black bars illustrate growth of HAS3-induced cells with daily addition of Streptomyces hyaluronidase (0.2 milliunits/ml) to the culture medium starting at day 3. Impaired cell growth, reduced by 45% as of day 7 in the untreated induced cells (gray bars), was fully relieved by supplementation with hyaluronidase throughout the assay (black bars). To correlate this observation with HA production by the cells, HA content of the media was measured on the indicated days throughout the growth assay (Fig. 7B, inset). HA content of the growth-impaired HAS3-induced cell media (gray bars) was elevated Ϸ2-fold at day 2 and Ϸ10-fold at day 6 relative to the uninduced media (white bars). HA content of media from hyaluronidase-treated cells decreased (following hyaluronidase addition at day 3, black bars) in inverse correlation to the increase in cell growth.
Temporal Control of HA Production Directly Influences Growth Kinetics of Tumor Cells-The inducible system afforded us the advantage of being able to control the temporal expression of HAS3 and correlate resultant HA production with growth curves. 22Rv1 Tet-On cells transfected with the inducible HAS3 construct were plated in the absence (Fig. 8A, white bars) or presence (gray and black bars) of dox. Inhibited cell growth in the dox-induced groups was evident (Ϸ43%) by day 4 (arrowhead). At this time, all cell culture media were replaced: cells represented by white bars were again given standard culture media, and wells represented by gray bars received dox-containing media. Black bars represent cells from which dox-containing media were removed and replaced with standard media at day 4. Removal of induction conditions restored the growth rate of the latter group by day 8. Levels of HA measured in the conditioned media from each group of cells showed an expected drop due to media replacement at day 4, but whereas the induced cells began replacing HA (Fig. 8B, gray bars), induced cells from which dox was removed (black bars) did not resynthesize HA. It should be noted that the magnitude of HA induction was lower when cells were induced at day 4 relative to day 0, probably resulting from expected plasmid copy number decrease following transient transfection. It is also apparent that relatively low amounts of HA synthesis (e.g. Ϸ3-fold induced in the day 4 group) translated to a 31% decrease in cell count by day 6. These results are consistent with a direct correlation between elevated HA production and growth inhibition and demonstrate that this inhibition is rapidly relieved by cessation of HA synthesis.
Finally we plated 22Rv1 Tet-On HAS3 transfectants in the absence (Fig. 8C, white and black bars) or presence (gray bars) of dox. Growth impairment in the induced group was again evident by day 4 (arrowhead) at which time culture media were again replaced. In this experiment, cells represented by the black bars were given dox-containing medium to induce HAS3 expression at day 4. Whereas these cells had exhibited a growth rate identical to the uninduced controls up to day 4, cellular growth was significantly slowed by day 6 (p Ͻ 0.05 for day 6 and day 8). HA levels in the culture media at each time point again reflected a direct relationship between HA production and growth suppression (Fig. 8D). Collectively the results of this work indicate that levels of HA produced by HAS3 overexpression in prostate tumor cells are antiangiogenic and antiproliferative in subcutaneous tumors. Furthermore removal of accumulated HA, either by enzymatic degradation through hyaluronidase S.E. of manual cell counts is plotted. *, p Ͻ 0.01 for HAS3 uninduced transfectants relative to uninduced GFP and for HAS3 induced relative to HAS3 uninduced. B, growth of inducible HAS3 transfectants was similarly assayed in the absence (white bars) or presence (gray and black bars) of dox. The effect of enzymatic HA removal by exogenously added hyaluronidase was compared by assaying growth of the induced transfectants in the absence (gray bars) or presence (black bars) of 0.1 units/ml Streptomyces hyaluronidase added at day 3. *, p Ͻ 0.01 relative to induced, untreated (gray bars versus black bars). Inset, HA content of cell culture conditioned media from the assay was quantified by competitive binding assay at the days indicated.
or reduction of HAS3 expression, correlates directly to increases in tumor cell growth.

DISCUSSION
Accumulation of HA and immunohistochemical detection of HYAL1 in human prostate tumor specimens is predictive of cancer recurrence following therapeutic intervention (4,8). In prostate tumor cell lines, the aggressive potential of the cells correlates with 20-fold excess HAS3, but the relevance of this correlation has not been determined. We stably overexpressed HAS3 in prostate tumor cells and found that the resultant excess HA production by the cells suppressed tumorigenesis in part by retarding intrinsic growth. Excess synthesis of HA led to changes in integrin-mediated cell adhesion that may translate to tumor cell anoikis in the collagen-rich subcutaneous environment but probably also inhibit tumor cell cycle progression. Coexpression of HYAL1 restored adhesion and growth in culture, but HYAL1 activity was insufficient to clear the accumulated HA in vivo. Thus, although HYAL1/HAS3 cotransfectant tumors were vascularized comparably to control tumors, HA polymers persisting within the tumors continued to suppress tumor cell proliferation. Use of an inducible expression system allowed us to determine an inverse correlation between tumor cell HA synthesis and growth kinetics that was independent of effects on cell adhesion, thereby implicating an additional HA receptor-mediated pathway for inhibition of cell cycle progression. Collectively the data support a role for HA in initiating changes in cell surface receptor engagement that may be amplified and exploited subsequently in tumor promotion upon local HA dispersion. Therefore, HA processing by elevated HYAL1 expression in invasive prostate cancer is likely required for progression.
The normal functions of HA are extremely diverse despite its chemical simplicity (for reviews, see Refs. 14 and 25). HA is a linear anionic polymer comprised of a single repeating disaccharide motif critical for regulating cell division, cell motility, and cellular transformation during development (51)(52)(53)(54)(55)(56)(57)(58). HA polymers and HA oligosaccharides have been implicated in distinct aspects of these processes. Responses to HA are mediated by cell surface receptors that cluster when bound to the multivalent HA polymers and disperse when not bound, differentially triggering intracellular signals. For example, HA polymers stabilize focal adhesion contacts by clustering specific cell surface receptors that modulate cytoskeletal linker protein function (18,59). Subsequently engaged signal transduction pathways halt cell cycle progression and inhibit cell motility. Polymers thereby preserve cell-free space in vivo by limiting proliferation and preventing endothelial infiltration for angiogenesis. Disruption of the HA polymer-receptor interaction by clearance of the polymers or by competition of receptor binding with HA oligosaccharides (24, 25, 60 -62) correspondingly may stimulate proliferation, motility, and angiogenesis. Depending upon the size of the oligosaccharides, cells may also undergo apoptosis, anoikis, or autophagy. Thus, it is important for overall tissue homeostasis that HA production and processing are tightly controlled.
Consequences of excess HA production have been investigated previously by manipulation of HAS enzyme expression in several cell types. Mammary carcinoma (63), fibrosarcoma (64), or melanoma cells (65) that overexpress HAS1 or HAS2 are more tumorigenic and/or metastatic in mice. Conversely inhibition of HAS2 or HAS3 decreases HA production and reduces tumorigenic potential of the cells (43,66). Some controversy exists about potential differences in the HA polymers produced by different HAS isozymes. However, there is considerable overlap of the ranges of polymer sizes obtained and no chemical or kinetic evidence for intrinsic differences in HA production among the isozymes. This suggests that differential effects of HA polymers are largely dependent on quantity, which is dictated primarily by HAS expression levels.
The effect of HA on tumorigenesis was recently found by Itano et al. (26) to be concentration-dependent. Stable transformed fibroblast lines selected for different levels of HA production by HAS1 overexpression were proportionally more tumorigenic when synthesizing moderate amounts of HA, but tumor growth was inhibited somewhat in cells that produced excessive quantities. This result helps explain the anomaly in our findings that HAS3 is antiangiogenic and antiproliferative in prostate tumorigenesis. HAS3 was previously overexpressed in TSU cells, formerly thought to be of prostate origin but reclassified as bladder-derived (67), and found to enhance tumorigenesis and metastasis (68). However, levels of HA were not measured with respect to cell culture concentrations so it is not possible to determine whether quantities fell into the tumorigenic range of the curve determined by Itano et al. (26). Hyaluronidase expression was also not characterized in these cells and may have played a role in relief from HA suppression of growth. Of note, HA quantities produced by both HAS3-and HYAL1/HAS3-transfected cells in the current study exceeded the levels shown to suppress tumor growth.
Achieving temporal control of HA production with the HAS3 inducible system afforded an opportunity to directly link the effects of differential HA levels on tumor growth in a constant cellular background. An additional observation was made possible by the comparison of stable lines selected for long term constitutive overexpression of HAS3 with those transiently induced to overexpress HAS3. Suppression of cell growth was observed in both conditions, but adhesion to ECM proteins was only inhibited when HA overproduction was constitutive and stable. Disruption of cell-ECM interactions by altered integrin function has been associated extensively with cell death by anoikis (49,69,70). However, specific integrins (␣1␤1, ␣5␤1, and ␣v␤3) have been shown to interact with the adaptor protein Shc to stimulate an alternative pathway for cell cycle progression (69,71). We used integrin blocking antibodies to define the contribution of specific integrins to the residual adhesion of HAS3 transfectants to type IV collagen and fibronectin. Despite the dependence of control and HYAL1/HAS3-cotransfected cells on multiple integrins for binding to both collagen and fibronectin, HAS3 transfectants were completely dependent on ␣1␤1 and ␣5␤1, respectively. However, significantly reduced adhesion to these substrates implies correspondingly diminished receptor function. Thus, this may translate to a failure of HAS3 transfectants to stimulate cell cycling as well as cell survival in response to ECM proteins in vitro and in vivo.
In contrast, the inducible HAS3 transfectants adhered comparably to all ECM substrates relative to controls. Thus, they are unlikely to be experiencing changes in integrin-mediated survival or cell cycle entry that would account for their decreased growth kinetics. Rather these cells showed growth suppression only with sustained HA synthesis and accumulation in the conditioned media. Thus, it is likely that HA retention at the surface of the tumor cells themselves is transducing the signal for growth suppression. Previous reports of HA-induced cell cycle arrest have implicated the transmembrane HA receptor CD44 (18,59,72,73) as the critical cell surface signaling contact. CD44 promotes cell cycle progression in the absence of HA polymer ligation through differential association with cytoskeletal linker proteins such as ezrin (18,59). The presence of HA polymers elevates a tumor suppressor protein, merlin, which binds the CD44 intracellular signaling domain, precluding direct association of CD44 and ezrin. In fact, ezrin was investigated recently as a cancer progression marker, and although the role in progression was inconclusive, both ezrin and merlin were found to be expressed in prostate tumor cells (74). Thus, reduced cell growth by HAS3 overexpression may result from direct HA suppression of cell cycling through CD44/merlin in addition to cell cycle arrest by altered integrin function.
During normal cellular HA turnover, high molecular mass HA polymers are processed to relatively short polymers and/or oligosaccharides by hyaluronidase-catalyzed hydrolysis (16). Initial extracellular processing of HA to short polymers was demonstrated recently in cultured cells, dependent upon the expression of HYAL2. Cellular uptake of HA was subsequently found to require HYAL1 and was augmented by CD44. The authors proposed that extracellular HYAL1 bound HA and mediated its association with HYAL2 and CD44 at the membrane where HYAL2 initiated hydrolysis. Once inside the cell, HA-HYAL1 complexes were ultimately localized to lysosomes where the low pH activated HYAL1 for complete cleavage of HA to tetrasaccharides. Apparent activity of HYAL2 has also been shown to be enhanced by association in lipid rafts with CD44 and a Na ϩ /H ϩ antiporter that locally acidifies membrane microdomains (35). Thus, although HYAL1 has been shown to have optimal hyaluronidase activity at acidic pH, it is possible through HA-mediated complex formation within these acidic membrane domains that both HYAL1 and HYAL2 may be contributing to extracellular HA processing. Particularly in the context of acidic regions within tumor microenvironments, the excessive presence of either HYAL1 or HYAL2 could be expected to yield excesses of extracellular HA degradation products that would normally be confined within the cell. The outcome could range from cell transformation to angiogenesis.
In fact, HYAL1 and HYAL2 have both been implicated in cancer progression. HYAL2 is found to be elevated in conjunc-tion with HAS2 and CD44 in invasive breast cancer (66,75). HYAL1 is correlated with aggressive cancer progression because its expression in human prostate tumor tissue samples is predictive of poor patient outcome (8). Overexpression of HYAL1 in PC3M, a metastatic prostate tumor cell line, modestly increased lymph node positivity in an orthotopic model (76). We have also determined that constitutive overexpression of HYAL1 in 22Rv1 cells, which are normally non-metastatic, induces metastasis to paraaortic nodes upon orthotopic injection, although there was no effect on growth at the primary site (29). HYAL1 may be required to release tumor cells from cell cycle and motility suppression imposed by HA-rich primary tumor tissues. HAS3 overexpression correlates with metastatic phenotype in multiple lineage-derived tumor cell lines such as prostate and colon (77). HYAL1 expression may rise in tumor tissues subsequently to HAS overexpression as a selected compensatory mechanism for growth suppression. Our finding that exogenous hyaluronidase can relieve growth suppression in vitro further implies that the source of the hyaluronidase in vivo can be exogenous, perhaps secreted by tumor-associated stromal cells or macrophages.
Constitutive HAS3 expression suppressed subcutaneous tumor growth and angiogenesis, which were not compensated by HYAL1 coexpression despite the restored growth of cotransfectants in vitro. One possible explanation for this observation is that HYAL1 activity may be antagonized by endogenous hyaluronidase inhibitors (78). This effect was not observed in our previous studies of HYAL1/HAS2 cotransfectants, however, suggesting rather that HYAL1 levels are insufficient to compensate for the increased HA production by HAS3 (about 5-fold higher than HA production by HAS2 (44)). Hyaluronidases have been shown to be inhibited by excess HA polymer concentrations in vitro (79), so this may also be the case in vivo, particularly depending on local fluctuations in tumor pH. Consistent with this notion, in our initial experiments with exogenous hyaluronidase addition to HAS3-induced tumor cells, we found that daily augmentation of hyaluronidase was essential to observe the effects of restored cell growth. In addition, measurements of HYAL1 activity in our assays of culture conditioned media (Fig. 1C) showed submaximal specific activity until significantly diluted. Furthermore significant HA accumulation was detected in tumor sections, suggesting that although HYAL1 expression ameliorates growth suppression in culture, its ability to do so in vivo requires HA diffusion to relieve inhibition. Diffusion of HA may be more constrained in a three-dimensional tissue context.
Tumor size is a function of apparent tumor cell growth kinetics, whereas vascularization of the tumor is a response to a tumor-provided signal or a tumor-induced environmental cue. The data we have provided in vitro have highlighted two pathways by which HAS3 overexpression could be expected to result in suppression of tumor cell proliferation. Excessive HA could utilize similar signaling mechanisms to suppress the recruitment and proliferation of endothelial cells required for angiogenesis of HAS3-overexpressing tumors. Similar impairment of HYAL1/HAS3 transfectant tumor growth, however, occurred despite unsuppressed angiogenesis. This could be partially explained by Ϸ50% HA clearance of these tumors rel-ative to the HAS3 tumors, but if restored cell cycle entry was solely responsible for vascular development, then tumor cell proliferation in the HA-deficient regions would also be expected. Thus, we propose that angiogenesis is the result of endothelial cell recruitment by a tumor cell product generated within those regions. This recruitment signal could either be HYAL1-bound HA complexes or could be partially hydrolyzed HA polymers/oligosaccharides, response to which would be mediated by endothelial CD44. Further experiments will be required to define aspects of this model.
In summary, we have developed a useful system with which to fine tune levels of HA production and uncouple the temporal effects of HA on growth control versus tumorigenesis. The direct relation of HA production to suppression of tumor growth and angiogenesis resolves the question of whether the product of HAS is intrinsically tumorigenic. Our findings imply that concurrent HA production and adequate processing activity are a requirement for maximal tumorigenic potential. The establishment of a direct relationship between cell growth rate and HA production implies that sudden immediate changes in HA production can control tumor growth. Because HA levels influence tumorigenesis in a dose-dependent fashion, the sudden complete removal of HA production may impair growth of an established tumor in the same way high levels suppressed growth. Thus, the inducible system will permit us to test the relationship between HA and maintenance of established tumors in the same way we have investigated its role in tumor formation. A further advantage of the system is the ability to control HA levels within the same cellular background, thus avoiding some of the pitfalls of clonal selection. It will be important to test this relationship in orthotopic tumorigenesis and metastasis. We are currently developing stable inducible lines to perform these studies.