Prostate cancer is the most common non-cutaneous cancer in American men and the second leading cause of death from cancer (Dennis and Resnick, 2000). Although surgical resection and radiotherapy are potentially curative for localized (organ-confined) disease, conventional chemotherapy and radiotherapy still have limited efficacy against advanced disease (Wang and Waxman, 2000). Androgen ablation often leads to symptomatic improvement in patients with advanced disease, but progression to hormone refractory disease is usually inevitable (Lytton, 2001). Therefore, new approaches to therapy of prostate cancer are greatly needed, not only as alternatives to conventional treatment for adenocarcinoma but also for chemoprevention for patients with prostate cancer precursor lesions, such as high grade prostatic intraepithelial neoplasia, or small organ-confined invasive tumors, as well as for patients with advanced and metastatic prostate cancer.
Thiazolidinediones (TZDs), a new class of anti-diabetic drugs (Murphy and Holder, 2000), have been identified as specific ligands for peroxisome proliferator-activated receptor (PPARγ), a ligand-activated transcription factor belonging to the nuclear hormone receptor superfamily (Rosen and Spiegelman, 2001). PPARγ regulates lipid and lipoprotein metabolism and influences cellular proliferation, differentiation, and apoptosis (Rosen and Spiegelman, 2001). Although PPARγ is predominately expressed in adipose tissue (one of the target tissues for insulin), it has been found in macrophages, vascular smooth muscle cells, endothelial cells, and diverse normal as well as malignant epithelial cells. Studies with cell lines suggest that activation of PPARγ by TZDs suppresses cell growth and may induce differentiation (Fujiwara and Horikoshi, 2000; Rosen and Spiegelman, 2001).
Transcriptional activity of PPARγ is increased upon binding of its ligands. Both natural and synthetic ligands have been described. Fatty acids and derivatives bind with low affinity, whereas certain eicosanoids such as 15-Deoxy-Δ12,14-prostaglandin J2 (15d-PGJ2), 13-hydroxyoctadecadienoic acid (13-HODE), and 15-hydroxyeicosatetraenoic acid (15-HETE) bind PPARγ with higher affinity (Rosen and Spiegelman, 2001). 15-HETE, derived from the activity of epithelial 15-lipoxygenase-2 (15-LOX-2), is the major arachidonic acid metabolite synthesized by benign human prostate, with reduced formation in prostate cancer, and thus represents a candidate endogenous ligand for PPARγ in prostatic epithelial cells (Shappell et al., 1999, 2001b). Rosiglitazone (BRL 49653) is a potent and selective synthetic PPARγ agonist of the TZD class of compounds (Murphy and Holder, 2000).
Previous studies using prostate cancer cell lines showed that agonists of PPARγ induced growth inhibition and cell cycle arrest in vitro, and reduced tumor size of xenografts (Kubota et al., 1998; Butler et al., 2000; Shappell et al., 2001a). Based on the previously reported preclinical data, a phase II clinical study of patients with advanced prostate cancer was conducted. Prolonged stabilization of serum prostate-specific antigen (PSA) was seen in a significant minority of patients treated with the TZD troglitazone (Mueller et al., 2000).
Given the promising preclinical and clinical data regarding the activity of TZDs on prostate cancer, it is worthwhile to further explore the mechanisms whereby PPARγ activation may be tumor suppressive. We used a model system of primary cultures of human prostatic epithelial cells derived from adenocarcinomas of Gleason patterns 3 and/or 4 (Gleason score 6 and 7 tumors), the most common grades of cancer in radical prostatectomy specimens. The expression of PPARγ and the effects of BRL 49653 on cell growth, gene expression, and differentiation were determined. We noted striking temporal and cell type-specific changes in response to BRL 49653. In particular, growth inhibition caused by BRL 49653 was reversible after up to at least 96 h of treatment, and morphologic changes induced by BRL 49653 were not fully developed until after about 7 days of treatment. The phenotype elicited by activation of PPARγ was unique in our experience with prostate cancer cells and could not be directly linked to either classic secretory or neuroendocrine differentiation of prostatic epithelial cells. The new information that we generated with primary cultures will be relevant to developing optimal strategies to target PPARγ for prevention or treatment of prostate cancer.
MATERIALS AND METHODS
Tissues dissected from radical prostatectomy specimens were processed for primary culture of prostatic epithelial cells according to previously described methods (Peehl, 2002). None of the patients had received prior chemical, hormonal, or radiation therapy. Histological assessment was performed as described (Schmid and McNeal, 1992) and each tissue from which cultures were derived contained ≥90% malignant epithelia. Adenocarcinomas were classified according to the Gleason grading system (Gleason, 1988). The primary cell strains E-CA-1, E-CA-2, E-CA-4, and E-CA-5 were from cancers of Gleason grade 3/3. Another cell strain, E-CA-3, was from a cancer composed of 70% Gleason grade 4 and 30% intraductal carcinoma. Characteristics that typically distinguish primary cultures derived from prostatic adenocarcinomas from those derived from normal tissues include ploidy (Brothman et al., 1992), expression of the retinol metabolizing enzyme, lecithin:retinol acyltransferase (Guo et al., 2002), and activity of vitamin D 1α-hydroxylase (Hsu et al., 2001).
BRL 49653 was synthesized by Glaxo Wellcome (Research Triangle Park, NC). Stock solutions of 10 mM were prepared in dimethylsulfoxide (DMSO) and stored at −20°C.
Clonal growth assay
Five hundred cells were seeded into each 60-mm, collagen-coated dish containing serum-free growth medium (Peehl, 2002) and vehicle or BRL 49653. After 10 days of incubation, the cells were fixed with 10% formalin and stained with 0.1% crystal violet (Peehl, 2002). Total cell growth was measured with an Artek image analyzer (Dynatech, Chantilly, VA).
High density growth assay
Ten thousand cells were seeded into each 60-mm, collagen-coated dish containing serum-free growth medium (Peehl, 2002). The next day (day 1), cells from triplicate dishes were counted. Vehicle or BRL 49653 was added to the remaining dishes. On days 4 and 7, cells from triplicate dishes per treatment were harvested and counted. Fresh medium containing vehicle or BRL 49653 was replaced on day 3.
Growth reversibility assay
Semi-confluent populations of cells were grown in medium containing vehicle or 10 μM of BRL 49653. After 1, 2, 3, and 4 days, cells were harvested from vehicle- and BRL 49653-treated dishes and tested for growth potential in clonal growth assays.
Cell cycle analysis
Cells treated with or without BRL 49653 were harvested with trypsin/ethylenediaminetetraacetic acid (EDTA) and fixed by dropwise addition of ice cold 70% ethanol. After 1 h of fixation, cells were rinsed with phosphate-buffered saline (PBS) containing 0.5% Tween and 1% bovine serum albumin (BSA), then incubated with RNase (10 mg/ml) and stained with propidium iodide (20 mg/ml). Analysis of DNA content was carried out on a FACScan flow cytometer and cell-cycle phase distribution was analyzed using Modfit software.
Semi-confluent cells growing in standard serum-free growth medium (Peehl, 2002) were suspended with trypsin/EDTA and centrifuged. The cell pellet was washed in ice-cold PBS, suspended in urea–Tris buffer (9 M urea, 75 mM Tris-HCl, pH 7.5, 0.15 M 2-mercaptoethanol), and sonicated briefly. The supernatant was collected and protein concentration was determined by a BioRad assay (BioRad, Hercules, CA). Typically, 50 or 80 μg of protein were separated by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE), transferred to polyvinylidene fluoride membranes (Osmonics, Westborough, MA) and blocked by PBS with 5% non-fat milk. Proteins were detected with mouse monoclonal anti-PPAR antibody (Santa Cruz Biotechnology, Santa Cruz, CA), anti-β-catenin antibody (BD Transduction Laboratories, San Jose, CA), and anti-glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (Research Diagnostics, Inc., Flanders, NJ). Anti-species horseradish peroxidase-conjugated secondary antibodies were obtained from Dako (Carpinteria, CA) and visual detection was performed using the enhanced chemoluminescence method (Amersham, Piscataway, NJ).
Cells were seeded onto 60-mm, collagen-coated dishes containing complete medium and vehicle or 10 μM of BRL 49653. Fresh medium (with or without BRL 49653) was replaced after 3 days of incubation. Cells were examined microscopically and photos were taken on days 3 and 5.
Oil red O staining
Cells were grown as for immunocytochemistry (see below) and treated with 10 μm BRL 49653 or vehicle for 4 days. Oil red O (Sigma-Aldrich, St. Louis, MO) was dissolved in isopropanol at 0.3%. Cells were rinsed briefly with PBS and stained at 37°C in fresh filtered oil red O solution (60% of saturated stock oil red O/isopropanol). After 30 min of staining, the cells were rinsed once with 60% isopropanol followed by distilled water. The stained cells were examined by microscopy.
Senescence-associated-β-galactosidase (SA-β-gal)-positive cells were detected by the method of Dimri et al. (1995). The presence of positive staining was observed microscopically.
Pellets of treated and untreated cells were fixed in 2.5% glutaraldehyde in PBS for 2 h on ice and processed for electron microscopy by standard techniques.
Cells (inoculated at 104/chamber) were grown in 8-chamber slides (Nalge Nunc International, Naperville, IL), fixed with 2% paraformaldehyde, and permeabilized with ethanol. Non-specific binding was blocked with 10% horse serum, and then cells were incubated with primary antibodies against PSA (DAKO; 1:50), cytokeratin 18 (BioGenex, San Ramon, CA; 1:500), chromogranin A (clone DAK-A3; DAKO; 1:200), and adipophilin (clone AP 125; Research Diagnostics, Inc.; 1:400). After rinsing and incubating with biotinylated secondary anti-species antibodies (Vector Laboratories, Burlingame, CA), labeling was detected with the ABC reagent (Vector Laboratories) and the chromagen diaminobenzidine. After counterstaining with hematoxylin, the slides were coverslipped and examined microscopically.
Quantitative real time RT-PCR
mRNA copy numbers in RNA extracted from cells cultured with or without BRL 49653 were determined for PPARγ, adipocyte-type fatty acid binding protein (aFABP), adipophilin, neutrophil gelatinase-associated lipocalin (NGAL), transforming growth factor beta-stimulated clone 22 (TSC-22), glycerol kinase (GyK), β-actin, and GAPDH by real time quantitative reverse transcription-polymerase chain reaction (RT-PCR) using a Lightcycler fluorescence temperature rapid-air cycler (Roche Molecular Biochemicals, Indianapolis, IN) with cDNA standard curves and the double-stranded DNA-binding fluorescent probe SYBR Green (Shappell et al., 2001a). RNA was extracted using Qiagen RNeasy Midi Kit (QIAGEN, Palo Alto, CA) and the concentration was determined by UV absorbance at 260 nm. Two hundred ng of total RNA were used per analysis, with 20 μl reactions in glass capillary tubes using concentrations of reaction ingredients as described (Shappell et al., 2001a). mRNA copy numbers for the genes of interest were normalized to β-actin expression. The templates for the standard curves and assay primers for PPARγ and aFABP (Shappell et al., 2001a) and GyK (Guan et al., 2002) were as described previously. The cDNA template for standard curves for β-actin was a 762 bp fragment amplified from 1 μg genomic DNA by PCR using the primers 5′-GTG-ACG-TGG-ACA-TCC-GCA-AAG-ACC-3′ (forward) and 5′-TGG-GGG-ACA-AAA-AGG-GGG-AAG-G-3′ (reverse). The purified product was inserted into pCRII using the TA Cloning Kit from Invitrogen (Carlsbad, CA). The 1,272 bp cDNA template and forward and reverse primers for the 452 bp amplimer for GAPDH were purchased from Clontech (Palo Alto, CA). The cDNA templates for adipophilin, NGAL, and TSC-22 were 478, 850, and 435 bp in pPCR-Script (Stratagene, La Jolla, CA), pT7T3D-PAC (Research Genetics, Huntsville, AL), and pcDNA3.1-Zeo (Stratagene), respectively, and were generously provided by Dr. Rajnish A. Gupta (Vanderbilt University Medical Center, Nashville, TN). For real time RT-PCR, the primers and the corresponding amplimer lengths were as follows: β-actin, 5′-AGC-GCA-AGT-ACT-CCG-TGT-G-3′ (forward), 5′-GAC-TGG-GCC-ATT-CTC-CTT-AGA-3′ (reverse), 462 bp amplimer; adipophilin, 5′-CGC-TGT-GAC-TGG-GGC-AAA-AGA-3′ (forward), 5′-ATC-CGA-CTC-CCC-AAG-ACT-GTG-TTA-3′ (reverse), 174 bp amplimer; NGAL, 5′-GCC-CAG-CAG-CCA-CCA-CAG-C-3′ (forward), 5′-TGC-CAG-GCC-TAC-CAC-ATA-CCA-CTT-3′ (reverse), 214 bp amplimer; TSC-22, 5′-CCT-TGC-TGG-GGA-CTG-AAA-A-3′ (forward), 5′-AGC-TGG-GCC-TGA-AAC-TGG-3′ (reverse), 259 bp amplimer. The one-step real time RT-PCR reactions consisted of the following steps: reverse transcription at 55°C for 15 min, denaturation at 95°C for 1 min, amplification for either 40 cycles (GAPDH), 45 cycles (aFABP), 50 cycles (TSC-22), 55 cycles (PPARγ), or 65 cycles (β-actin, adipophilin, and NGAL), and melting curve analysis from 65 or 75°C at a rate of 0.1°C per sec under continuous fluorescence monitoring. The amplification programs for the β-actin, GAPDH, adipophilin, NGAL, and TSC-22 assays consisted of heating at 20°C per sec to 95°C, cooling at 20°C per sec to 55°C, annealing at 55°C for 5 sec, heating at 20°C per sec to 72°C, elongation at 72°C for either 11 sec (adipophilin), 12 sec (TSC-22), 15 sec (NGAL), 19 sec (β-actin), or 20 sec (GAPDH), and heating at 2°C per sec to 87°C (β-actin), at 5°C per sec to 85°C (NGAL), or at 20°C per sec to 82°C (TSC-22), 83°C (aFABP), 84°C (PPARγ), 86°C (adipophilin), or 90°C (GAPDH) for fluorescence acquisition. The specificity of the amplimer in each reaction was confirmed by the melting curve analysis, with initial gel confirmation that this large peak corresponded to the expected amplimer (Shappell et al., 2001a). The contribution to fluorescence signal of any non-specific products and/or primer dimers was eliminated by increasing the temperature to 2°C below the melting temperature of the specific product, which eliminated any other minor cDNAs (which have lower melting temperatures) (Kofron et al., 1999; Shappell et al., 2001a). Copy numbers of mRNA were calculated from serially diluted standard curves generated from the purified cDNA template (Hartel et al., 1999). Serial dilutions (1:10) over a range of 5–6 orders of magnitude were generally used to generate the standard curves. For each assay, the serially diluted standards were simultaneously amplified with the unknown samples to generate a linear standard curve using the fit points method of analysis with four points or the software determined second derivative maximum method. Standard curves for β-actin, GAPDH, PPARγ, aFABP, adipophilin, TSC-22, GyK, and NGAL all had correlation coefficients of 0.99 or 1.00 in each assay. Control samples run in triplicate had a variance of approximately 10%. Untreated cells from multiple analyzed cell strains were always run in the same assay for each gene product, as were the control and BRL-treated cells at different time points. All biologic samples fell on the standard curves, and copy numbers of the unknown samples were calculated using the Lightcycler software (version 3 and version 3.5).
Primary cultures of prostate cancer cells express PPARγ mRNA and protein
Real time RT-PCR was performed to evaluate the expression of the PPARγ gene in primary cultures of prostate cancer cells and the relative levels of PPARγ mRNA in the cells to be used in subsequent proliferation assays. PPARγ mRNA was constitutively expressed in all 4 of 4 cell strains tested (Fig. 1a). The single amplimer demonstrated in melting curve analysis of the real time RT-PCR reactions was identical to that of the PPARγ cDNA standards used for the quantitative standard curve (Fig. 1b), with the proper expected amplimer length of both standards and samples confirmed on agarose gels (not shown). Normalized to β-actin mRNA copy numbers within the same RNA aliquots, the levels of expression of PPARγ mRNA in the individual cell strains were all within 50% of the mean values, with 14.2 ± 5.3 × 10−4 mRNA copies/β-actin mRNA copies [mean ± standard deviation (SD), range 8.4–19.8 × 10−4 mRNA copies/β-actin mRNA copies].
To determine whether PPARγ mRNA was transcribed into protein, immunoblot analyses were performed. Immunoreactive PPARγ was detected in all four of the cell strains evaluated (Fig. 1c).
BRL 49653 inhibited the growth of primary cultures of prostate cancer cells
After confirming that PPARγ was present in primary cultures of prostate cancer cells, we evaluated the ability of the PPARγ-specific ligand BRL 49653 to modulate cell growth. Five cell strains, including the four previously examined for protein expression of PPARγ, were tested in clonal growth assays. Cells were inoculated at 500 cells per dish into media with or without BRL 49653 and growth was measured after 10 days of incubation. Dose-dependent growth inhibition by BRL 49653 was observed (Fig. 2). Four of the cell strains responded very similarly, with half-maximal inhibition occurring at about 2 μM of BRL 49653 and almost complete inhibition at 10 μM. The fifth cell strain (E-CA-3) was less responsive. Growth of this cell strain was inhibited by less than 20% with 2 μM of BRL 49653, and with 10 μM, growth was still about 30% of control. Decreased sensitivity of this cell strain to BRL 49653-induced growth inhibition did not appear to correlate with less expression of PPARγ protein or mRNA compared to the more responsive cell strains (Fig. 1).
Time-dependent growth inhibition by BRL 49653
The ability of BRL 49653 to inhibit growth over time was evaluated in high density growth assays. In these experiments, cells were inoculated at 105 per 60-mm2 dish into medium with or without BRL 49653 and growth was measured by counting cells on days 4 and 7 (Fig. 3). Two concentrations of BRL 49653 were tested. On day 4, repression of cell growth in response to BRL 49653 was evident, with total cell number reduced by approximately 30% with 1 μM and 40% with 10 μM of BRL 49653. Growth inhibition was more pronounced at day 7, when 1 μM of BRL 49653 caused approximately 50% reduction and 10 μM caused approximately 70% reduction of cell number. Therefore, the concentrations of BRL 49653 required to inhibit high density cell growth were similar to those required to inhibit clonal growth, suggesting that the autocrine growth factors that accumulate in high density cultures did not alter response to the PPARγ agonist. In addition, the results of the time course showed that the maximal effects of BRL 49653 were not immediate but took several days to reach full potential.
Reversibility of growth inhibition of prostate cancer cells by BRL 49653
To determine whether growth inhibition by the PPARγ agonist was reversible, cells were tested for their ability to resume growth after removal of BRL 49653. Semi-confluent cell cultures were maintained in media with or without 10 μM of BRL 49653 for 24, 48, 72, or 96 h. At each time point, cells were suspended with trypsin/EDTA and inoculated at clonal densities into medium without BRL 49653 (Fig. 4). The proliferative potential of cells treated for 24 or 48 h with BRL 49653 was equivalent to that of untreated cells. Even after 72 or 96 h of treatment, the growth potential of these cells was still approximately 70∼80% of vehicle-treated cells. These results showed that growth inhibition of prostate cancer cells by BRL 49653 was completely reversible up to 48 h after treatment, and partially reversible up to 96 h after treatment. These results concurred with those of the previously described high density growth assays (Fig. 3), which showed that BRL 49653 required several days to reach full growth-inhibitory potential.
Cell cycle arrest by BRL 49653
We next investigated whether cell cycle arrest accompanied PPARγ-mediated growth inhibition. Cancer cell strains were treated with vehicle or 10 μM of BRL 49653 for 24, 48, and 72 h. At each time point, cell cycle distribution was evaluated by flow cytometry. A significant accumulation in S-phase was found after 24 h of treatment with 10 μM of BRL 49653 compared to control (Table 1). However, S-phase arrest was transient and at 48 and 72 h, cell cycle distribution was equivalent between treated and untreated populations.
Table 1. Effect of BRL 49653 on cell cycle distribution*
E-CA-1 cells were fed fresh medium ± 10 μM BRL 49653 and harvested for cell cycle analysis after 24, 48, and 72 h of treatment. Each entry indicates the proportion of cells (%) in the respective cell cycle compartment, as determined by flow cytometric analysis of propidium iodide-stained cells. The experiment was repeated twice (Exp. 1 and 2).
Morphological alterations induced in prostate cancer cells by BRL 49653
Light microscopic examination of cells treated with BRL 49653 revealed distinctive morphological changes. These changes first became visible after approximately 4 days of treatment and became more pronounced with time, reaching a plateau after approximately 1 week (Fig. 5). Initially, cells became elongated with a morphology resembling that of neuroendocrine-differentiated prostate cancer cell lines induced by cyclic adenosine 3′, 5′-monophosphate (cAMP) (Deeble et al., 2001). Later, cells flattened and developed cytoplasmic vacuoles. The percentage of cells displaying this latter morphology approached 100% after 1 week of treatment with BRL 49653. PPARγ agonists induce adipocyte differentiation associated with accumulation of neutral lipids and potentially induce expression of genes typically associated with adipocyte differentiation in epithelial cells (Mueller et al., 1998). The morphological changes induced in prostate cancer cells by BRL 49653, especially after a week of treatment, were reminiscent of those in cells undergoing adipocyte differentiation. Positive staining with oil red O is indicative of accumulation of neutral lipids that are characteristically found in adipocytes (Esquenet et al., 1997). However, following 7 days of treatment with BRL 49653 (10 μM), no positive staining with oil red O was observed, indicating that the vacuolization induced by BRL 49653 in prostate cancer cells did not reflect accumulation of neutral lipids typical of adipocyte differentiation (not shown). Adipocytes were included as a positive control and showed extensive staining with oil red O.
Immunocytochemical assessment of differentiation induced by BRL 49653
Given that BRL 49653 apparently did not induce adipocyte differentiation in prostatic epithelial cells, we considered other possible types of cellular differentiation that might parallel the observed morphological alterations. Since the earliest morphological features seen in prostate cells treated with BRL 49653 were similar to those of neuroendocrine cells, we used chromogranin A as a marker to evaluate neuroendocrine differentiation (Abrahamsson, 1999). Expression of this marker was not seen by immunocytochemistry in either treated or untreated cells (not shown). Next, we considered that BRL 49653 perhaps enhanced secretory differentiation, characteristic of prostatic luminal cells in normal epithelial glands and paralleled to varying degrees in tumors. Classic markers accompanying secretory differentiation of prostatic epithelial cells include cytokeratin 18 and PSA (Bonkhoff et al., 1994). Primary cultures of prostatic epithelial cells express no or low levels of PSA and this was not changed in cancer cells treated with 10 μM of BRL 49653 for 1 week (not shown). Cytokeratin 18 is expressed heterogeneously in primary cultures, and expression of cytokeratin 18 was also comparable in treated vs. untreated cells (not shown). Therefore, enhanced expression of classic luminal epithelial cell markers of differentiation did not occur in response to BRL 49653.
Changes in gene expression induced by BRL 49653
Several genes have been reported to be upregulated by ligand activation of PPARγ in adipocytes (Murphy and Holder, 2000; Guan et al., 2002) and in epithelial tumor cell lines (Gupta et al., 2001). PPARγ-regulated genes possibly related to intracellular lipid transport and secretory differentiation include aFABP, adipophilin, and NGAL (Gupta et al., 2001). GyK, linked to triglyceride breakdown and “futile” metabolic cycles, was recently shown to be markedly induced in TZD-treated adipocytes (Guan et al., 2002). TSC-22 is a transcription repressor linked to cellular proliferation that is specifically modulated by PPARγ-agonists in epithelial tumor cell lines (Gupta et al., 2003). Levels of expression of these genes in control cells and in cells treated with 10 μM of BRL 49653 for variable time periods were measured by quantitative real time RT-PCR. β-actin mRNA expression (mRNA copy numbers/100 ng total RNA) was similar in all cell strains examined and was not significantly altered by BRL 49653 treatment. Levels of PPARγ-regulated gene expression in control and stimulated cells were thus normalized to β-actin levels.
aFABP, adipophilin, NGAL, and TSC-22 mRNAs were constitutively expressed in all four cell strains examined. Interestingly, the relative levels of these PPARγ-regulated genes seemed to parallel each other and the relative baseline level of PPARγ mRNA in each cell strain (Fig. 6a). In one of the cancer cell strains (E-CA-2) most sensitive to inhibition of growth and induction of morphologic changes by stimulation of PPARγ, BRL 49653 caused marked upregulation of adipophilin mRNA (Fig. 6a). Upregulation was present at 6 h, the earliest time point examined, and was maintained at a similarly high level compared to control cells at 24, 48, and 72 h (Fig. 6b). In contrast, over the same time frame, PPARγ mRNA levels were modulated only slightly, with early minor increased expression and later minor downregulation compared to control cell populations, supporting the impression that the altered expression of adipophilin was truly due to ligand-dependent activation of PPARγ-mediated gene transcriptional regulation. TSC-22 showed a less prominent degree of upregulation (approximately twofold) over the time frame tested, but also remained elevated at later times when PPARγ expression was reduced. In contrast, in this particular cell strain, alterations of aFABP and NGAL by BRL 49653 were not conspicuous over the same time period, and not that dissimilar from relative alterations in PPARγ mRNA (Fig. 6b). GyK was not significantly upregulated by BRL 49653 in two cell strains that we analyzed (not shown). Altered expression of the indicated genes was not a general phenomenon reflecting an overall increased transcriptional activity, as not all genes had similar increases in expression and upregulation of the control housekeeping gene GAPDH was also not observed (not shown).
The upregulation of adipophilin by activation of PPARγ was confirmed by immunocytochemistry. Cytoplasmic adipophilin immunostaining was significantly more prominent in cells treated with 10 μM of BRL 49653 for 7 days compared to untreated cells (Fig. 7). β-catenin, shown to be suppressed in colon cancer cells by PPARγ ligands (Girnun et al., 2002), was also examined by immmunocytochemistry and immunoblot analyses of cells treated with or without BRL 49653 for up to 7 days. At no time point did we observe downregulation of β-catenin protein levels or changes in cellular localization (not shown).
Ultrastructural characteristics of BRL 49653-treated prostate cells
In order to gain more information about the morphological changes induced in prostate cancer cells by activation of PPARγ, electron microscopic examination was performed on cells after 7 days of treatment with 10 μM of BRL 49653. Transmission electron microscopy showed that treated cells contained variable numbers of cytoplasmic vacuoles that ranged from 0.5 μm to 2.0 μm in diameter (Fig. 8). The vacuoles contained scattered phospholipid arrays and membrane fragments as well as occasional dense granules. Most cells contained between 25 and 50 vacuoles per cell profile and in most cells the vacuoles occupied greater than 50% of the cytoplasmic volume. Untreated cells contained only a few enlarged vacuoles that numbered less than 5 per cell. In addition, treated cells showed increased amounts of neutral lipid droplets while in untreated cells lipid droplets were an uncommon finding.
In order to rule out the possibility that the ultrastructural changes observed in treated cells were a manifestation of premature senescence, we tested treated and untreated cells for SA-β-gal activity (Dimri et al., 1995). The incidence of SA-β-gal-positive cells was low in untreated cells and was not increased by treatment with 10 μM of BRL 49653 for up to 7 days.
As the mortality rate of prostate cancer continues unabated, the development of new therapeutic agents is urgently needed. The results of pilot clinical studies suggested that agonists of PPARγ may have anti-tumor activity in patients with prostate cancer (Hisatake et al., 2000; Mueller et al., 2000). PPARγ mRNA is expressed in benign and malignant prostatic tissues (Mueller et al., 2000; Shappell et al., 2001a), and the observed effects of PPARγ ligands on established prostate cancer cell lines in vitro and in xenograft models are consistent with potential therapeutic activity (Kubota et al., 1998; Hisatake et al., 2000; Shappell et al., 2001a). Arachidonic acid metabolites, including 15-HETE and 15d-PGJ2, can activate PPARγ in cell-free systems and intact cells. 15-HETE has been implicated as a possible physiological ligand of PPARγ in the prostate, as 15-LOX-2-derived 15-HETE is the major arachidonic acid metabolite synthesized by benign prostate. Both 15-LOX-2 expression and 15-HETE formation are substantially reduced or absent in most prostate cancers compared to benign prostate (Shappell et al., 2001b). 15-HETE can activate PPARγ and inhibit proliferation of PC-3 prostate carcinoma cells (Shappell et al., 2001a). 15d-PGJ2 is another suspected natural ligand of PPARγ that halts prostate cancer cell growth (Butler et al., 2000), but whether or not this metabolite is actually formed in the prostate or altered in prostate carcinoma remains to be determined.
Because of the involvement of PPARγ in so many critical physiologic and pathologic functions including regulation of insulin sensitivity, energy expenditure, the development of atherosclerosis and anti-tumorigenic effects in a variety of different cancers, great effort has been spent in trying to identify endogenous ligands for PPARγ as well as to develop and characterize the effect of high affinity synthetic ligands. In this study, we investigated the effects of a specific PPARγ agonist, BRL 49653 or rosiglitazone, on primary cultures of human prostatic cancer cells.
Our studies showed that prostate cancer cell strains expressed PPARγ mRNA and protein, similar to established prostate cancer cell lines (Kubota et al., 1998; Butler et al., 2000; Mueller et al., 2000; Shappell et al., 2001a). BRL 49653 suppressed proliferation, with half-maximal inhibition of clonal growth at about 2 μM for the majority of cell strains. This is similar to the level of BRL 49653 reported to inhibit PC-3 cells in soft agar colony-forming assays (Shappell et al., 2001a). One of the primary cancer cell strains was less sensitive than the other four to growth inhibition by BRL 49653, but the reason for this remains to be determined. As has been shown in other studies (Kubota et al., 1998; Butler et al., 2000; Mueller et al., 2000), sensitivity of the primary cultures of prostate cancer cells to growth inhibition by agonists of PPARγ did not directly correlate with levels of expression of PPARγ.
Other factors may influence response to PPARγ ligands. For instance, mitogen-activated protein kinase (MAPK) has been reported to phosphorylate and hence down-regulate PPARγ activity (Hsi et al., 2001). Whether reduced responsiveness to BRL 49653 might reflect higher levels of activated MAPK in the currently employed cell strains remains to be established. As MAPK activation has been reported in high grade human prostate cancer tissues (Gioeli et al., 1999), this pathway could possibly modulate responsiveness of prostate cancer cells in actual patients to systemically administered PPARγ agonists. This possibility needs to be considered in future clinical studies as well as in vitro. It is interesting to note that the less responsive cell strain in our study was derived from a cancer of higher Gleason grade (70% Gleason pattern 4/30% intraductal carcinoma) than the other four cell strains, which were from cancers of Gleason pattern 3 (Gleason score 3 + 3 = 6).
In the current study, growth inhibition by PPARγ stimulation was time-dependent, as shown by rather minimal inhibition after 3 days of treatment with BRL 49653 and significantly greater inhibition after 7 days of treatment. Growth inhibition was also reversible. Cells exposed to BRL 49653 for 2 days recovered almost 100% of their growth potential when inoculated into fresh medium without BRL 49653, and even after 4 days of treatment, growth potential was 75% of cells never exposed to BRL 49653. This observation is relevant to development of clinical protocols and suggests that TZD therapy of prostate cancer may have to be continuous over an extended period to be effective.
Cell cycle arrest caused by BRL 49653 was also transitory. S-phase delay was seen in treated compared to control cells at the end of 24 h with no significant difference in cell cycle distribution between the treated and untreated populations at later time points. In PC-3 cells, a slight increase of cells in G1 was noted after 3 days of treatment with BRL 49653 (Shappell et al., 2001a), somewhat similar to that observed in cancer cell lines established from other tissues (Sarraf et al., 1998). In PC-3, DU 145, and LNCaP prostate cancer cell lines, the PPARγ-ligand 15d-PGJ2 caused accumulation in the S-phase of the cell cycle (Butler et al., 2000), similar to our results with BRL 49653 on primary cultures of prostate cancer cells. However, S-phase arrest in the primary cultures was transitory and likely is not responsible for the antiproliferative effects of BRL 49653. S-phase accumulation of the established cell lines in response to 15d-PGJ2 was also followed by cell death by a non-apoptotic mechanism (Butler et al., 2000), whereas no increased death was noted in the primary cultures in our study or in prostate cell lines (Mueller et al., 2000) after treatment with BRL 49653.
Ligand activation of PPARγ induces cell-specific differentiation (Bar-Tana, 2001). Examples include the promotion of adipocytic differentiation in pre-adipocytes and other mesenchymal cells and the creation of foam cells (cholesterol-loaded macrophages) by activation of PPARγ in peripheral blood monocytes (Bar-Tana, 2001). Whether or not PPARγ agonists actually contribute to differentiation of epithelial cells, including those in the benign prostate, or can promote differentiation in prostate cancer cells, has not been thoroughly investigated. Troglitazone reportedly induced vacuolization in established prostate cancer cell lines, including ultrastructural features of surface invaginations with microvilli and of secondary lysosomes (Kubota et al., 1998). We observed similar ultrastructural features in PC-3 cells stimulated with 15-HETE and BRL 49653 (S.B.S., unpublished observations). The significance of such vacuoles remains unclear, and the effect of PPARγ agonists on other pathways paralleling prostate secretory differentiation remains unclear. Troglitazone was reported to reduce expression of PSA in LNCaP cells (Mueller et al., 2000), but as this agent can directly interfere with androgen receptor-mediated transactivation of this androgen-regulated gene (Hisatake et al., 2000), the relationship of this observation to modulation of true secretory differentiation in prostate cells remains to be defined. Hence, the relationship of the inhibition of prostate cancer cell proliferation by PPARγ agonists to possible modulation of differentiation has not been adequately addressed.
Since PPARγ was initially cloned as a master regulator of adipogenesis, this differentiation-inducing activity of PPARγ is the most thoroughly studied. The ability of PPARγ activation to convert cells to adipocytes is not confined to pre-adipocytes, but has been described as well for fibroblasts, bone marrow stromal cells, liposarcoma, and breast cancer cells (Tontonoz et al., 1997; Mueller et al., 1998). In the current study, although the prominent cytoplasmic vacuolization seen in prostate cancer cells treated with BRL 49653 was reminiscent of foam cell formation or adipogenesis, lack of staining with oil red O ruled out significant accumulation of neutral lipids in these cells, despite occasional lipid noted on ultrastructural examination. Similarly, treatment of the prostate cancer cell lines LNCaP, PC-3, and DU 145 with the PPARγ agonist 15d-PGJ2 failed to induce adipocyte differentiation as also defined by accumulation of neutral lipid (Butler et al., 2000). Vacuolization of tumor epithelial cells by PPARγ agonists has been variably associated with upregulation of a limited number of so-called adipocyte differentiation genes, such as aFABP (AP2) (Mueller et al., 1998; Butler et al., 2000; Shappell et al., 2001a), which may function in lipid processing or secretory function in other cell types.
However, other types of differentiation besides conversion to adipocytes occur in cells upon activation of PPARγ. Differentiation of physiological relevance in the prostate includes neuroendocrine and secretory differentiation. The neuroendocrine marker, chromogranin A, was not induced by treatment of primary cultures of prostate cancer cells with BRL 49653. Immunocytochemical labeling with antibodies against proteins typically expressed by differentiated secretory luminal cells of the prostatic epithelium, PSA and cytokeratin 18, also showed no significant changes in expression. We also ruled out the possibility that the ultrastructural changes induced in prostatic epithelial cells by BRL 49653 were associated with premature senescence by showing absence of an increase in SA-β-gal. At this time, we must conclude that the striking phenotype induced in prostate cancer cells by activation of PPARγ is an atypical form of differentiation. However, it must be kept in mind that development of fully differentiated prostatic secretory cells in culture has been an elusive goal, and that cancer cells might demonstrate aberrant differentiation patterns compared to normal cells. The relevance of the observed phenotype to effects that PPARγ agonists might have on prostatic epithelial cells in vivo may be discerned in future studies by examination of surgically removed prostate tissues following therapy with TZDs.
Genes up- or down-regulated by PPARγ transactivation are still being elucidated. aFABP is increased by PPARγ agonists, paralleling differentiation in adipocytes (Kletzien et al., 1992; Hauner, 2002). By RT-PCR, we have not consistently detected aFABP mRNA in snap-frozen benign or malignant human prostate tissues (Iyengar et al., 2002). In PC-3 cells, treatment with 15-HETE or BRL 49653 resulted in upregulation of aFABP (Shappell et al., 2001a). In contrast, aFABP mRNA was constitutively expressed in the cancer cells utilized in our study and was not appreciably increased by BRL 49653 treatment. GyK and β-catenin, found to be targets of PPARγ in adipocytes (Guan et al., 2002) and colon cancer cells (Girnun et al., 2002), respectively, were also not significantly regulated by BRL 49653 in primary cultures of prostate cancer cells, again demonstrating the cell-specific nature of PPARγ activity.
NGAL and adipophilin are two other genes that are upregulated by PPARγ agonists in tumor cell lines, which is prevented by pre-treatment with a specific PPARγ antagonist (Gupta et al., 2001). Adipophilin is a 48–50 kD protein, previously recognized as a marker for adipocyte differentiation, but which has been demonstrated to be present in a wide range of cell types, including epithelial cells, that have in common the normal function of lipid accumulation or steroid hormone synthesis/secretion (Heid et al., 1998). Interestingly, adipophilin was upregulated by oxidized low density lipoprotein (LDL) in macrophages (Wang et al., 1999), a process known to involve PPARγ activation (Nagy et al., 1998). TSC-22 is specifically upregulated in tumor cell lines by PPARγ agonists (Gupta et al., 2003). The TSC-22 gene product appears to be a transcriptional repressing factor (Kester et al., 1999), regulating cell growth, differentiation, and apoptosis, and which may be reduced in a variety of adenocarcinomas compared to corresponding benign epithelial tissues (Nakashiro et al., 1998; Rae et al., 2000). In contrast to our findings for aFABP, we essentially uniformly detect adipophilin and NGAL mRNA expression (Iyengar et al., 2002) in benign or malignant human prostate tissues. Expression of TSC-22 mRNA, on the other hand, is reduced in malignant compared to benign tissues (S.I., S.B.S., unpublished observations). Whether or not these genes are reduced in tumor vs. benign tissues in a manner regulated by PPARγ remains to be more fully established. Adipophilin, NGAL, and TSC-22 mRNA were constitutively expressed in the cancer cells utilized herein, but only TSC-22 and adipophilin expression was increased by BRL 49653, with adipophilin expression markedly increased. Adipophilin immunostaining is variably present in the cytoplasm of prostate secretory cells in actual human prostatic tissues (Iyengar et al., 2002), and in the current study, treatment with BRL 49653 resulted in increased cytoplasmic immunostaining of adipophilin in growth-inhibited prostate cancer cells. These results show that growth inhibition of primary prostate cancer cells by PPARγ agonists is accompanied by modulation of PPARγ-regulated genes that are normally expressed in prostate and which may contribute to normal prostate cell function and growth regulation. However, other proteins found in secretory epithelial cells of the prostate, such as PSA and cytokeratin 18, were not upregulated by BRL 49653 so that an association between activation of PPARγ and secretory differentiation remains tenuous.
In future studies, we will address the particular role of 15-HETE in regulating growth and differentiation of prostatic epithelial cells. 15-HETE is the product of 15-LOX-2 activity in the secretory cells of the normal prostatic epithelium (Shappell et al., 1999; Jack et al., 2000). This enzyme is absent or reduced in a majority of adenocarcinomas of the prostate, and levels are diminished in the premalignant lesion high grade prostatic intraepithelial neoplasia (Shappell et al., 1999, 2001b; Jack et al., 2000). 15-LOX-2 mRNA, protein, and catalytic activity are variably expressed in primary cultures of benign and malignant prostate tissues as utilized herein (S.B.S., D.M.P., manuscript in preparation). If 15-HETE is a predominant physiological ligand of PPARγ, then our studies suggest that loss of activation of PPARγ due to loss of production of 15-HETE by 15-LOX-2 in prostate cancer cells would lead to increased proliferation and diminished differentiation. Recently, Tang et al. (2002) reported that 15-LOX-2 is a negative cell regulator in normal prostatic epithelial cells, but they did not investigate the role of PPARγ in this process. We conclude that primary cell cultures provide a useful model system to investigate the role of PPARγ in prostatic biology and to generate preclinical data supporting application of PPARγ agonists for prevention or cure of prostate cancer.
This work was supported in part by a CaP CURE Award (to D.M.P.) and a Discovery Grant from the Vanderbilt Ingram Cancer Center (to S.B.S.).