Selenium Effects on Prostate Cell Growth¹

Selenium is one of the most exciting and promising agents currently under laboratory and translational development for prostate cancer chemoprevention. Secondary clinical findings of Clark et al. (1, 2) are primarily responsible for generating the great interest in this agent. Clark et al. (1, 2) conducted a double-blind, placebo-controlled trial of selenium to prevent basal cell or squamous cell skin carcinoma in 1312 subjects (including 974 men) with a history of nonmelanoma skin cancer. Although negative regarding the primary skin cancer end point, the trial produced the strikingly positive secondary finding that prostate cancer incidence was reduced by 63% among men in the selenium arm(versus in the placebo arm). This finding is supported by epidemiological data (3) and prompted our present laboratory study of the in vitro biological effects of selenium on normal human prostate cells and human prostate cancer cells.

For the past decade, prostate cancer has been second only to nonmelanoma skin cancer as the most commonly diagnosed cancer in American men. Lifetime risks for prostate cancer are 16.6 and 18.1% for Caucasian and African-American men, respectively (4). Therefore, the public health implications of identifying an effective chemopreventive agent against this disease are tremendous.

Although well established as an essential micronutrient of animal and human diets, the activity of selenium in human prostate carcinogenesis is only just beginning to undergo evaluation. Selenium is an essential element in redox pathways and has been identified as one of the many dietary factors with the potential to modify carcinogenesis (5, 6, 7, 8). This element is widely distributed in soil,forages, and grains (9). It is present in plants in both inorganic (e.g., selenite and selenate) and organic(e.g., selenocysteine and selenomethionine) forms (10). Organic forms of selenium such as SeMet³are more bioavailable than are inorganic forms such as Na₂SeO₃ (11). The metabolism and bioavailability are influenced by absorption in the gastrointestinal tract, transport in the blood, metabolism and storage in tissues, and excretion rate in the urine and feces (11).

In the present study, we examined selenium effects on the growth, cell cycle, and apoptosis of prostate cancer cells and normal prostate cells. We evaluated the effects of both an inorganic selenium(Na₂SeO₃) and organic selenium (SeMet) in primary tissue culture. We used both androgen-responsive (LNCaP) and androgen-unresponsive (PC-3 and DU145)cancer cells. The normal cell types were PrECs, PrSM cells, and PrSt cells. To our knowledge, this is the first report of a comparative in vitro analysis of the apoptosis and growth arrest effects of selenium in a comprehensive panel of human prostate cancer and normal cells.

Primary cultures of normal human prostate cells PrEC, PrSt, and PrSM were purchased from Clonetics (San Diego, CA). These primary cell cultures were maintained in defined culture medium (Clonetics) and passaged using trypsin-EDTA (Clonetics) according to the manufacturer’s instructions. The LNCaP cell line was isolated from a lymph node, PC-3 cell line from a bone metastasis, and DU145 cell line from a brain metastasis of prostate cancer patients. These cells were maintained in DMEM and F-12 (mixed 1:1) medium supplemented with 10% fetal bovine serum and passaged using trypsin-EDTA.

Cells were plated into 96-well tissue culture plates and allowed to establish monolayers overnight. Either Na₂SeO₃ or SeMet in various concentrations was added to chosen wells and grown for 24, 48, or 72 h. The assay was terminated by washing cells with PBS and then fixing them with 1% paraformaldehyde in PBS. Fixed cells were washed and stained with 0.2% crystal violet in 50 mm3-(cyclohexylamino)-1-propanesulfonic acid buffer at pH 9.5 for 30 min. Plates were washed, allowed to dry, and then solubilized using 10%glacial acetic acid. Samples were analyzed by using a Biolumin 960 microtiter plate reader (Molecular Dynamics, San Diego, CA) at an absorption wavelength of 590 nm and normalized based on control well absorbency. These data were then reduced to growth inhibition percentages in relationship to untreated control samples. The inhibitory concentration of 50% growth (IC₅₀)was determined using a Statview statistics program (SAS Institute,Inc., Cary, NC) by using a logically weighted least squares regression algorithm called Lowess at a tension parameter setting of 66. Independent experiments were performed using trypan blue exclusion. Cells were evaluated by light microscopy on a hemocytometer to determine viability, and these experiments provided comparable results.

Cells were grown in 24-well plates and treated with various concentrations of Na₂SeO₃or SeMet. Cells were harvested using 0.25% trypsin, 2 mmEDTA in PBS, resuspended in ∼100 μl of Cytofix, and stored in an Eppendorf tube at 4°C. After placing cells in Cytofunnel, they were spun onto coated Cytoslides at 750 rpm in a Cytospin 2 centrifuge(Shandon, Pittsburgh, PA). Cells were postfixed with 4%paraformaldehyde in PBS for 1 h at 25°C and permeabilized for 10 min with 1% Triton X-100 in 0.1 m citrate buffer (pH 6.0). Samples were washed with PBS and stained with either DAPI or Hoechst 33258 stain (Molecular Probes, Eugene, OR) at a final concentration of 1.0 μm overnight at 4°C, washed in distilled H₂O, and mounted in Prolong antifade solution(Molecular Probes). Slides were analyzed by light microscopy, and data were acquired using digital image analysis. Our image analysis system uses a Quantix air-cooled, black and white CCD camera (Photometrics,Tucson, AZ) that is driven by IPlabs software (Scanalytics, Inc.,Fairfax, VA) using a Macintosh G3 computer (Apple Computer, Cupertino,CA). This image acquisition equipment is attached to an IX70 inverted research light microscope equipped with epi-illumination objectives(Olympus America, Inc., Lake Success, NY).

Laminin was adsorbed to 12-mm round coverslips by incubating for 1 h at 25°C in 24-well tissue culture plates at a concentration equivalent to 1 μg/cm². After washing, cells were plated on laminin-coated coverslips 24 h prior to exposure to selenium compounds. After treatment, the coverslips were washed with PBS and fixed with 1% paraformaldehyde in PBS. Coverslips were washed again with PBS, and the cells were permeabilized with 1% Triton X-100 in 0.1 m citrate buffer at pH 6.0. Cells were then incubated with an Apotag (Roche, Indianapolis, IN) enzymatic reaction mixture containing terminal deoxynucleotidyl transferase and FITC-dUTP according to the manufacturer’s instructions, with slight modification. Coverslips were inverted on top of 30 μl of reaction mixture on parafilm that had been placed over #1 Whatman paper moistened with distilled H₂O lying on the bottom of a glass baking tray. After covering the tray with a plastic cover,it was placed in a humidified incubator, and the reaction was performed at 37°C for 90 min. Coverslips were washed by dipping 10 times in a beaker containing 1 liter of distilled H₂O. They were then mounted on glass slides using Prolong antifade solution(Molecular Probes). This method was also used to stain the Cytoslides described above. Slides were analyzed by light microscopy, and data were acquired and analyzed on our digital image acquisition system.

A similar TUNEL method was used to analyze samples by FACS. The APO-DIRECT system (Phoenix Flow Systems, Inc., San Diego, CA) was used for FACS analysis according to the manufacturer’s instructions with slight modification. Floating cells that had been grown on 10-cm dishes and treated with the different selenium compounds were harvested by washing monolayers with CMF-PBS. The cells in wash buffer were centrifuged at 300 × g for 5 min at 4°C and placed on ice. The monolayers were trypsinized for 10 min in PBS containing 2 mm EDTA and 0.25% trypsin. These samples were pooled and resuspended in 1% paraformaldehyde in PBS and fixed for 15 min on ice. Samples were centrifuged at 800 × g for 5 min and resuspended in PBS containing 1% Triton X-100 and incubated on ice for 10 min. Cells were washed with PBS followed by centrifugation,resuspended in 70% ethanol, and stored at −20°C until analyzed with the APO-DIRECT protocol according to the manufacturer’s instructions.

Cells grown on coverslips using methods described above were also stained to detect caspase-3 activation or P-Tyr15-p34/cdc2 (New England Biolabs). Samples were blocked with Superblock (Pierce Chemical Co.,Rockford, IL) solution in Tris-buffered saline (TBS) containing 0.1%Tween 20 (Pierce Chemical Co.). Primary rabbit polyclonal antibodies that recognize either the activated caspase-3 or P-Tyr15-p34/cdc2 were diluted in this Superblock solution and incubated in 24-well tissue culture dishes overnight at 4°C. Samples were rinsed two times with PBS and incubated with Texas Red-X goat antirabbit secondary antibody or Alexa 488 goat antirabbit secondary antibody (Molecular Probes). Samples were rinsed three times in PBS and one time in distilled H₂O and mounted on glass slides using Prolong antifade mounting medium (Molecular Probes). Slides were then analyzed by epi-fluorescence microscopy, and data were acquired using digital image analysis.

After staining cells on coverslips with primary and secondary antibodies, we counterstained cells with DAPI to identify the nuclear DNA and Alexa-594-phalloidin (Molecular Probes) to identify cytoplasmic actin. Cells were counterstained with 500 nm DAPI stain and 1 unit/ml Alexa-594-phalloidin in Superblock solution containing 0.1%Tween 20 overnight at 4°C. Samples were rinsed three times in PBS and one time in distilled H₂O and mounted on glass slides using Prolong antifade solution (Molecular Probes). Slides were then analyzed by light microscopy, and data were acquired using digital image analysis.

Cell monolayers were washed twice with cold CMF-PBS and removed from tissue culture plates by scraping into cold CMF-PBS on ice. Samples were then pelleted at 12,000 × g and solubilized. The packed cells were solubilized 3:1 (v/v) in solubilization buffer to cells in a buffer solution containing 1%NP40, 1 mm NaMoO₄, 30 mm NaF, 1 mm EDTA, 1 mm NaVO₄, 30 mm Tris (pH 8.3), 100 mmNaCl, and Complete protease inhibitor cocktail (Roche). Protein levels were determined using the Bradford protein assay on a Biolumin 960(Molecular Dynamics, San Diego, CA) microtiter plate reader. Proteins were separated on Bis-Tris 4–12% polyacrylamide gradient electrophoresis gels and electrotransferred using the NuPage system(Novex, San Diego, CA) to 0.45-μm nitrocellulose membranes(Schleicher and Schuell, Keene, NH). Membranes were blocked with 3%bovine serum albumin in TBS with 1% Tween 20 (TBST) and incubated with primary antibodies at 4°C in the same blocking solution. After washing three times in TBST, blots were incubated with the suitable horseradish peroxidase-conjugated secondary antibodies (Pierce Chemical Co.) and washed again three times. SuperSignal West (Pierce Chemical Co.) chemiluminescence detection procedures were followed according to the manufacturer’s instructions. Blots were exposed to Hyperfilm (Amersham, Arlington Heights, IL), followed by development in a Kodak automatic film processor (Eastman Kodak, Rochester, NY). Western data were digitized on a laser scanning densitometer (Molecular Dynamics, San Diego, CA) and quantified using NIH image version 1.62. Data were analyzed for statistical significance using Statview (SAS Institute, Inc., Cary, NC).

An agarose feeder layer was prepared by combining 12 ml of 2×DMEM with 6 ml of 1.8% melted agar that was cooled to just above 37°C, and the mixture was placed into six-well plates after the addition of various concentrations of selenium compound. After the feeder layer solidified, a low percentage agar layer was prepared by combining 5 ml of cells in DMEM/F-12 and 10% fetal bovine serum with 1 ml of 1.8% melted agar. The DU145 cells were plated at 1500 cells per well, and the LNCaP or PC-3 cells were plated at 2000 cells per well. The final mix of cells, medium, and agar was cooled to just above 37°C and selenium compound was added in the same concentrations as in the feeder layer. Cells were placed in a humidified incubator and fed periodically with a fresh agar-selenium mixture. Colonies were evaluated at 2 weeks using bright-field illumination through a ×1.25 lens on an IX70 inverted microscope (Olympus). Data were digitized using a Quantix CCD camera (Photometrics) and analyzed using IPlabs software (Scanalytics).

Fig. 1 and Table 1 illustrate the concentration-dependent and time-dependent effects of both Na₂SeO₃ and SeMet. Na₂SeO₃ and SeMet were each given in concentrations ranging from 1–500 μm. The growth of prostate cancer cells LNCaP, PC-3, and DU145 was inhibited at lower IC₅₀ concentrations by both Na₂SeO₃ and SeMet at 72 h than was the growth of normal prostate cells PrEC, PrSM, and PrSt (Table 1). Over the experimental time course, we observed a significant suppression of tumor cell growth compared with normal cells(Fig. 1). Na₂SeO₃ was consistently a more potent growth inhibitor than SeMet (Table 1). Generally, PrECs were least effected by increasing concentrations of Na₂SeO₃ and SeMet (Fig. 1). The strongest growth inhibiting activity of Na₂SeO₃ or SeMet was in androgen-responsive LNCaP cells. Na₂SeO₃ at IC₅₀ concentrations between 0.2 and 3.9μ m was a potent growth inhibitor of all prostate cancer cells at 72 h compared with an effective IC₅₀ concentration range of 22–38μ m in primary prostate cells. SeMet treatment demonstrated the greatest differential IC₅₀between prostate carcinoma cells (1–90 μm) and normal prostate cells (>500 μm) at 72 h, suggesting that SeMet more selectively inhibited growth of prostate cancer cells,compared with normal cells (Table 1).

Normal and malignant prostate cell lines were examined for evidence of apoptosis after exposure to 1–500 μm of Na₂SeO₃ or SeMet. We present apoptosis data based primarily on 10 μm of Na₂SeO3 and 500 μm of SeMet at 48 h, which were the best conditions for presenting side-by-side comparisons between tumor and normal cells. These conditions also were necessary experimentally, because there was a tendency of the LNCaP cells to lift off the tissue culture substrata at longer times or higher concentrations.

Selenocompounds induced less apoptosis in normal prostate cells compared with tumor cells when we initially examined cells grown on laminin-coated coverslips or Cytoslides using fluorescence microscopy. In these initial experiments, apoptosis was examined by triple staining coverslips or double staining Cytoslides containing untreated cells and those treated with either Na₂SeO₃ or SeMet (Fig. 2,A). Coverslips were triple stained (Fig. 2,A) using the FITC-dUTP TUNEL method (inset with green stain,outlined in white), counterstained with DAPI(blue stain) to study DNA condensation, and then immunostained for activated caspase-3 (inset with redstain, outlined in white). Cytoslides (Fig. 2,A) were double stained with FITC-dUTP TUNEL method(insets with green stain, outlined in yellow) then counterstained with DAPI (inset with blue stain, outlined in yellow). Control samples demonstrated little evidence of apoptosis shown in Cytospin samples by FITC-dUTP-TUNEL (Fig. 2,A, inset with green stain,outlined in yellow). Few selenium-treated normal cells stained for FITC-dUTP, condensed DAPI, and activated caspase-3 (Fig. 2,A). In contrast, selenocompounds induced high levels of apoptosis in prostate carcinoma cells, demonstrated by nuclear DNA condensation that colocalized with FITC-dUTP and activated capase-3(Fig. 2,A). These staining patterns were most prevalent in the Na₂SeO₃-treated cells when compared with the SeMet-treated samples, although SeMet caused a significant difference in staining of tumor cells in contrast with normal (Fig. 2 A).

We quantified the DNA condensation by analyzing digital images of DAPI-or Hoechst-stained Cytospin samples (Fig. 2,B). Many cells detached from laminin-coated substrata as they began to die. We collected the nonadherent cells and the remaining adherent cells after EDTA-trypsin harvesting and combined these two cell subpopulations. DNA condensation was observed within 48 h in prostate cells after Na₂SeO₃ as well as after SeMet treatment (Fig. 2,B). The androgen-responsive LNCaP carcinoma cells exhibited the highest level of DNA condensation after selenium treatment, followed by PC-3 and DU145 cells, in order of sensitivity (Fig. 2,B). Selenium-treated PrECs exhibited between 7.5- and 23-fold less DNA condensation compared with tumor cells (Fig. 2,B). Selenium-treated PrSM and PrSt cells exhibited between 4.7- and 5.3-fold more DNA condensation than PrECs but 1.9–3.8-fold less DNA condensation than tumor cells (Fig. 2 B).

We quantified the DNA fragmentation that occurred during apoptosis using TUNEL-FACS analysis. We collected and pooled the nonadherent cells and the remaining monolayer cells as described above. TUNEL-FACS analysis demonstrated FITC-dUTP incorporation into the DNA of Na₂SeO₃ as well as SeMet-treated dying cells (Fig. 2,C). The androgen-responsive LNCaP cells exhibited the highest level of nicked-end DNA labeling after treatment with both selenium compounds. The PC-3 cells were also sensitive to selenium treatments, followed by DU145 cells (Fig. 2 C). SeMet- and Na₂SeO₃-treated PrEC and PrSt cells did not exhibit significant levels of DNA fragmentation,whereas PrSM exhibited ∼25% fragmentation with Na₂SeO₃.

PARP is a substrate of the caspase-3 protease during apoptosis;therefore, we examined cell lysates for PARP cleavage after selenium treatment by Western analysis (Fig. 2,D). There were substantial differences observed in PARP expression between normal cells and tumor cells. The intact PARP molecule migrates at M_r 116,000, and the antibody(clone 4C10-5) we used recognizes the intact M_r 116,000 molecule and two cleavage products, M_r 85,000 and M_r 48,000 subfragments. We observed a profound difference for basal PARP expression of the intact M_r 116,000 molecule in normal prostate cells compared with tumor cells (Fig. 2,D). The untreated PrECs, PrSM cells, and PrSt cells did not express the intact PARP molecule when grown to ∼80% confluence (Fig. 2,D, column 1), whereas tumor cell lines expressed distinct baseline levels(Fig. 2 D, columns 2 and 3). These findings may reflect the ability of normal cells to undergo contact inhibition.

PARP cleavage in Na₂SeO₃-treated tumor cells was complete with no intact M_r 116,000 PARP observed at the earliest 24-h treatment point. A M_r 85,000 cleavage product appeared within 24 h in the LNCaP, PC-3, and DU145 cells (Fig. 2,D,column 2). In comparison, the cleavage of PARP in the SeMet-treated cells gradually occurred over the full time course (Fig. 2 D, column 3). In the SeMet-treated DU145 cells, the predominant cleavage fragment was a M_r48,000 form, whereas in the other cells, the predominant cleavage product was the M_r 85,000 form. This variation in cleavage fragment profiles may reflect heterogeneity of the cleavage sites for the multiple forms of caspases that result in a context-specific or cell-specific proteolytic processing mechanism (12). Whether the various cleavage fragments exert different effects on a given cell population remains to be determined.

PI staining of cellular DNA, followed by FACS analysis, provides a useful method for evaluating the cell cycle (Fig. 3,A). We observed PI staining of apoptotic cells at sub-G₀-G₁ levels of the cell cycle that corresponded to TUNEL-FACS data. The SeMet-treated normal prostate cells did not exhibit the same extent of sub-G₀-G₁ subpopulation cells as did the prostatic carcinoma cells (Fig. 3, A and B). Androgen-responsive LNCaP cells (Fig. 3, A, column 2, and B) exhibited the highest total increase in sub-G₀-G₁ cell fraction of any line examined after treatment with Na₂SeO₃ (23.1%) or SeMet(41.5%). We also observed an increase in sub-G₀-G₁ cells after Na₂SeO₃ of 13.5% in PC-3 cells or 14.7% in DU145 cells. Similarly,sub-G₀-G₁ cells after SeMet treatment of either PC-3 cells or DU145 cells increased 12.1 or 11.2%(Fig. 3, A and B).

A higher percentage of untreated PrECs were in the G₂ phase of the cell cycle compared with untreated tumor cells (Fig. 3, A, column 1, and C). Greater than 20% of the PrECs and PrSM cells were in the G₂ phase, which most likely reflects contact inhibition typical of normal cells. Treatment of both normal and tumor cells with Na₂SeO₃ resulted in a smaller change in the G₂ subfraction in PrEC(3%), PrSM (−3%), and PrSt (2%) cells as opposed to LNCaP (6%),PC-3 (4%), and DU145 (12%) tumor cells. SeMet treatment selectively increased the percentage of the G₂ subfraction cancer cells compared with normal cells (Fig. 3,C). When we examined the percentage increase in G₂ cell cycle phase cells after SeMet treatment, there was a modest 8% increase in PrECs, 4% increase in PrSM cells, and a 1% increase in PrSt cells. In contrast, SeMet caused a significant increases in G₂ cells (Fig. 3 C), particularly in LNCaP (13%), PC-3 (32%), and DU145 (20%) cells.

We also assessed the selenium effects on percentages of cells in S-phase. In PrEC cells,Na₂SeO₃ produced a 3%increase and SeMet a 12% increase in S-phase cells. In PrSM cells,Na₂SeO₃ and SeMet produced increased S-phase cells by 11 and 6%, respectively. In PrSt cells,Na₂SeO₃ and SeMet produced increased S-phase cells by 12 and 6%, respectively. In LNCaP tumor cells, we observed little change in S-phase after Na₂SeO₃ (3% increase) or SeMet (5% decrease). In contrast, treating the PC-3 cells and DU145 cells with Na₂SeO₃increased their respective S-phase components by 28 and 26%, whereas SeMet treatment caused <2% increases.

The phosphorylation of cdc2 on Tyr15 by WEE1 and MIK protein kinases causes cells to arrest in the G₂-M transition phase of the cell cycle (13, 14). To determine whether the accumulation of cells in the G₂-M phase observed by PI-FACS is associated with phosphorylation of cdc2 on Tyr15, we examined cells by immunofluorescence. We used optimized concentrations of Na₂SeO₃ (10μ m) and SeMet (500 μm)for cell cycle studies. Treatment of all cells with Na₂SeO₃ caused an increase in nuclei labeled with anti-phospho-cdc2 (Tyr15) antibody. Tumor cells demonstrated observable increases in P-Tyr15-p34/cdc2 after SeMet treatment that were not apparent in the normal cells (Fig. 4,A). When we examined nuclei using DAPI counterstaining, the P-Tyr15-p34/cdc2 colocalized mainly with the nuclei. In some instances,there also was some perinuclear staining that was observed in some of the SeMet-treated PC-3 samples (Fig. 4,A, white arrows),which also occurred in the DU145 (data not shown). When we examined cellular morphology using Alexa-594-conjugated phalloidin as a counterstain that labels actin microfilaments, we observed differential effects on cytoskeletal structures that depended on which compound was used, Na₂SeO₃ or SeMet. The Na₂SeO₃ treatment caused the disadherence of both normal and tumor cells that was preceded by rounding up of the cells. PC3 cells (Fig. 4 A, yellow bracket) and DU145 cells (data not shown) specifically formed numerous spike-shaped actin filament bundles in response to Na₂SeO₃ treatment. In contrast, SeMet treatment primarily caused tumor cells to either round-up and detach or spread out their cytoskeleton. The most intense P-Tyr15-p34/cdc2 staining appeared to associate with the nuclei of cells that had rounded-up.

The percentage of P-Tyr15-p34/cdc2-containing cells relative to DAPI-stained nuclei was determined by counting using digital capture and image analysis (Fig. 4,B). Normal PrECs exhibited little phosphorylation of P-Tyr15-p34/cdc2, whereas PrSM and PrST exhibited a more extensive presence of phosphoprotein in the Na₂SeO₃-treated samples. All of the selenium-treated LNCaP cells that remained attached to laminin substrata exhibited P-Tyr15-p34/cdc2 label. A high proportion of the selenium-treated PC-3 and DU145 cells that remained attached were positive for P-Tyr15-p34/cdc2 (Fig. 4 B).

To confirm that cdc2 was phosphorylated on Tyr15, we examined detergent-solubilized cell lysates after selenium treatment for 6 and 48 h by Western analysis. After SeMet treatment, Cdc2 phosphorylation ranged from 75 to 127% higher at 6 h in tumor cells compared with 2–15% higher at 6 h in normal cells by densitometry. In epithelial cells, there was only a 0.4% increase in cdc2 phosphorylation after treatment with Na₂SeO₃ (Fig. 4,C,Lane 2) and 2.4% increase with SeMet-(Fig. 4,C, Lane 3)treated samples. There was a low level of P-Tyr15-p34/cdc2 in the untreated tumor cells at the 6 and 48 h (Lane 1). There was a statistically significant increase in P-Tyr15-p34/cdc2 phosphoprotein formation in tumor cells after treatment with Na₂SeO₃ at 6 h(P < .0001) that was reduced at 48 h, most likely reflecting the loss of viable cells because of increased apoptosis(Fig. 4,C, Lane 2). There was a significant increase in P-Tyr15-p34/cdc2 in all of the tumor cells after 6 h of SeMet treatment (P < 0.0001) that remained higher than controls through 48 h (Fig. 4,C, Lane 3). These data are consistent with the increase in G₂-M-phase tumor cells observed using PI-FACS (Fig. 3,C) and the immunostaining of adherent cells (Fig. 4 A). Taken together,these data support the notion that G₂-M cell-cycle arrest of tumor cells after SeMet treatment involves cdc2-Tyr15 phosphorylation.

We also examined the number and size distribution of tumor colonies that formed in soft agar in the presence of selenium compounds. Computerized imaging and segmentation analyses were performed to determine the colony number and average colony size. SeMet inhibits the number of colonies formed from single cell clones (Fig. 5). Selenium compounds profoundly affected the size distribution of colonies that grew in soft agar, reflecting the clonal growth or expansion rate of each colony. The size and number of LNCaP, PC-3, and DU145 prostate cell colonies were decreased in the presence of organic SeMet in all cancer cell lines (Fig. 5). Tumor cell colony size and number were also inhibited by treatment with Na₂SeO₃ (data not shown).

Our present in vitro study provides direct support and rationale for selenium as a prostate cancer prevention agent. Our major findings are: (a) selenium has differential effects in prostate cancer versus normal cells; (b)selenium-induced growth inhibition and apoptosis in cancer cells are dose-dependent; and (c) selenium has greater activity in androgen-sensitive (LNCaP) versus androgen-insensitive (PC-3 and DU145) cancer cells. We present the first report of differential in vitro effects of selenium on growth inhibition and apoptosis in a comprehensive panel of human prostate cancer and normal cells. Our major result of differential selenium-induced growth inhibition, apoptosis, and cell cycle arrest in prostate cancer (but not normal) cells is consistent with the earlier positive selenium epidemiological and clinical findings.

The biology of selenium, including toxic chemical effects, has been studied extensively in other cell systems. Many biological effects of selenium supplementation involve incorporation of selenocysteine into proteins such as SeGPX and thioredoxin reductase via selenocysteine insertion sequences (15, 16, 17). Although selenium primarily functions after insertion into the active sites of enzymes, it also can exhibit certain chemical activity independent of protein incorporation. Chemical activity ranges from direct oxidation of nucleotides,proteins, and cofactors (18, 19) to chemically altering the binding interactions between retinoids and the nuclear retinoic acid receptors (20). Profound chemical effects, however,generally were observed in studies involving inorganic forms of selenium at high concentrations. Shen et al.(18) found that 10 μm of Na₂SeO₃ induced apoptosis in HepG2 cells because of generation of reactive oxygen species.

Consistent with our findings, certain reports show extensive toxicity of Na₂SeO₃ on both normal and tumor cells but little effect of SeMet on normal cells. A study of a series of selenocompounds showed that Na₂SeO₃ and selenocystamine were the most potent agents in reducing glutathione and DNA damage,leading to apoptosis in normal mouse keratinocytes (21). SeMet had no significant chemical effects on normal mouse keratinocytes. Kajander et al. (22) also examined the in vitro cytotoxicity of SeMet and Na₂SeO₃ in a large panel of leukemia, lymphoma, and hepatoma cells, along with CHO, BHK, and normal fibroblasts. In this study (22), the selenium IC₅₀ concentrations observed in the tumor cells(30–135 μm) were similar to ours (Table 1),but the normal fibroblasts IC₅₀ concentrations were lower (160 μm). However, the normal fibroblasts required a much longer time period (10 days) than tumor cells (3 days) to reach 50% cell death. Redman et al.(23) demonstrated recently the in vitroinhibition of growth and induction of apoptosis in DU145 cells after SeMet treatment, which produced lesser such effects in normal diploid fibroblasts. Consistent with our findings (Table 1), they observed very high IC₅₀ levels in treating normal fibroblasts(1 mm), compared with micromolar IC₅₀ levels in treating cancer cells (23). These and our studies demonstrate a differential effect of SeMet at lower concentrations and shorter times to induce death in tumor cells compared with normal cells.

Understanding the biological activity of selenium is complicated by the variety of available dietary forms of selenium and by the dynamics of selenium metabolism. In addition to SeMet and Na₂SeO₃, other selenocompounds, e.g., monomethylated selenium, may have chemopreventive activity (8, 24). Ip et al.(25) have reported that monomethylated selenium compounds can cause growth inhibition and apoptosis in rat mammary gland and epithelial cells in vivo. SeMet and other selenium compounds are dynamically metabolized into a wide array of products, and these derivatives may directly influence the chemopreventive effects of selenium (8, 24). The different metabolic pathways of SeMet and Na₂SeO₃ both produce H₂Se, which, after activation to selenophosphate, is a source of selenium for synthesis into selenoproteins (8, 24). H₂Se also can be further metabolized (for excretion) by a series of methylation reactions that generate additional reactive selenium species with potential chemopreventive properties (8, 24).

Other studies have reported differential rates of DNA fragmentation in the three prostate cancer cell lines that are consistent with our observations describing TUNEL-FACS. Analysis of DNA fragmentation has been a useful method for analyzing apoptosis. Tang et al. (26) performed an extensive apoptosis and DNA fragmentation analysis on human prostate cells after serum removal. The rank order of overall survivability of cells from lowest to highest was benign prostate hyperplasia, normal human prostate cells, LNCaP, PC-3,primary carcinoma, and DU145 cells. This ranking coincided with DNA fragmentation profiles (26). The onset of apoptosis and the DNA fragmentation of the cell began to increase at day 2 in both the LNCaP and PC3 cells but were more gradual and protracted in the DU145 cells (26). These data are consistent with our observations of nicked-DNA formation by TUNEL-FACS (Fig. 2, B and C).

Many studies have examined the proteolysis of PARP by caspase-3 to demonstrate apoptotic activity. Our observations that PARP cleavage profiles are substantially different between prostate tumor and normal cells emphasize the selective nature of selenium effects on inducing tumor cell apoptosis (Fig. 2 D). There are a growing number of biological properties attributed to PARP function that are coming under debate. PARP has zinc-finger DNA binding properties that enable the protein to detect and signal the occurrence of DNA strand breaks (27). PARP becomes enzymatically active when bound to DNA lesions. When its enzymatic activity is high, PARP converts nicotinamide dinucleotide to nicotinamide that can deplete ATP stores in the cell (28). The result of PARP activation is increased poly-ADP ribosylation of numerous nuclear proteins, including PARP itself. PARP activation occurs in response to genotoxic agents,such as alkylating agents and γ-irradiation. Another important PARP activity is to bind directly to retinoid X receptors and repress retinoid X receptor ligand-dependent transcriptional activities.

An important finding of the current report suggests that there may be a profound difference in PARP expression between contact-inhibited cells and cancer cells. The baseline expression of PARP in normal prostate cells was undetectable when we examined 30 μg of total protein from whole-cell lysates (Fig. 2 D). Our findings are consistent with studies of PARP gene expression in vivo during prostate involution (29). PARP expression was virtually undetectable in rat ventral prostates, when examined after castration,which caused the involution of prostatic tissue (29). PARP expression increased over the next 4 days in association with apoptotic cell death (29). Prostate tissue mRNA samples were examined by Northern blot analysis to demonstrate a time-dependent increase in PARP that occurred during prostate involution (29). The pattern of PARP expression was confirmed by in situ hybridization. PARP expression was isolated to just a few epithelial cells in the control animals and increased dramatically in the epithelium and surrounding stromal tissues as prostatic involution progressed (29). The time frame for production of an M_r 85,000 subfragment after Na₂SeO₃ treatment was between 24 and 48 h in PrSM and PrSt cells (data not shown). Even with such dramatic increases in the M_r85,000 proteins during these time periods, there continued to be very little intact M_r 116,000 PARP. These findings suggest that PARP production was induced because of apoptotic stress in contact-inhibited normal cells. Our findings further suggest that when PARP is produced, most of any intact molecules are rapidly degraded to inactive subfragments to minimize the effects of the intact molecule in contact-inhibited cells. These findings may be related to transient burst activity causing polyadenosinediphosphate-ribosylation of nuclear proteins as well after transients in PARP cleavage by caspase-3 (30, 31, 32). Selenium may exert its effects on this pathway through the selenoprotein, thioredoxin reductase, also described independently as a gene associated with retinoid-IFN-induced mortality (33). On the basis of these findings, it seems likely that thioredoxin reductase may induce caspase-3 activity in cancer cells. If selenium supplementation causes an increase in thioredoxin reductase in prostate cells, this may help explain the increase in caspase-3 activity (34).

In contrast with the absence of baseline PARP expression in normal cells, we found substantial baseline PARP levels in all three prostate tumor cell lines (Fig. 2,D, columns 2 and 3). Uncleaved PARP expression was the highest in androgen-responsive LNCaP cells and was associated with the greatest sensitivity to both Na₂SeO₃and SeMet treatment (Fig. 2 D). The PC-3 and DU145 cells also expressed higher baseline levels of PARP and responded to both Na₂SeO₃ and SeMet treatment to a greater degree than did normal prostate cells. The differences in basal PARP expression may help explain why normal cells are more resistant to selenium-induced apoptosis than tumor cells. Because the tumor cells express higher basal levels of M_r 116,000 protein, even the slightest elevation of caspase-3 activity may lead to the cleavage of available PARP. This may lead to more rapid elimination of DNA repair capacity,making the tumor cells more sensitive to selenium treatment (35). These findings suggest the importance of PARP expression in the differential responses of normal prostate cells and prostate cancer cells to selenium.

Cell cycle analysis using DNA binding dyes often helps determine the status of cell populations in either growth fractions, growth arrest,or apoptosis (36). Dying cells accumulate in the sub-G₀-G₁ fraction of the cell cycle as they undergo apoptosis (36). These fractions represent cells that are dead or dying, including fragmented cell nuclei or apoptotic debris (37). We observed a selective increase in the sub-G₀-G₁fraction of tumor cells compared with normal cells after Na₂SeO₃ or SeMet treatment(Fig. 3, A and B). The sub-G₀-G₁ fraction cell cycle profiles correlate with TUNEL-FACS labeling. The sub-G₀-G₁ fraction is underestimated by the computer because of debris from cell death (Fig. 3A, column 2). Cells that form decreased fluorescence subfraction, like those that we observed, have been shown previously to consist of apoptotic cells by ultrastructural analysis (37).

Cell cycle arrest at the G₂-M transition phase is one control mechanism that cells exert on proliferation. Apparently, G₂-M arrest protects cells from the lethality that results from undergoing cell division before repairing DNA damage (14). Eliminating cell growth by halting progression through the G₂-M transition state can also be caused by cell-cell contact inhibition as cells reach confluency (38). We observed many cells that were arrested in the G₂-M phase of the cell cycle (Fig. 3, A and C). There were many PrECs and PrSM normal cells arrested in the G₂-M cell cycle that associated with cell contact-induced growth inhibition (38). There was only a slight increase in normal cells at the G₂-M cell cycle checkpoint when they were treated with Na₂SeO₃ or SeMet (Fig. 3,C). In contrast, there was an increase in tumor cells in the G₂-M phase after treatment with selenium, particularly when using SeMet (Fig. 3, A and C). Cells arrested at the G₂-M cell cycle checkpoint doubled in all three tumor cell lines compared with a<10% increase in all of the normal cells.

Growth arrest in the G₂-M transition phase of the cell cycle has many important potential causes and outcomes. Reactive oxygen species, for example, can produce free radicals that promote cells to arrest in G₂-M (39). Free radicals can induce DNA and protein damage that activate protective mechanisms, resulting in growth arrest. Radiation damage is an important area of therapeutic intervention that relies on the generation of free radicals to induce macromolecular damage in cells. Studies have shown that G₂-M-arrested cells exhibit an increased sensitivity to ionizing radiation (40). When human prostatic carcinoma cell lines were examined using low-dose radiation exposure, there was a certain degree of variability in radiation sensitivity that depended on the cell line examined. The PC-3 cells that exhibited the most substantial increase in G₂-M-arrested cells was also the most sensitive to low-dose ionizing radiation. It is interesting to speculate that SeMet treatment may selectively enhance the radiosensitivity of prostatic carcinoma cells.

Growth control mechanisms rely heavily on phosphorylation or dephosphorylation of specific cell cycle molecules. cdc2 forms a complex with cyclin B1 during the G₂-M phase that controls cell mitosis (41). The phosphorylation status of cdc2 controls the progression of cells through the G₂-M phase of the cell cycle. Growth arrest at the G₂-M cell cycle phase is mediated by phosphorylation of cdc2 on Tyr15 and threonine 14 by WEE1 and MIK protein kinases (13, 14). The progression through G₂ into mitosis is controlled by cdc25B dephosphorylase activity on Tyr15 and threonine 14 (42). We observed a selective increase in tumor cells that were phosphorylated on Tyr15 of cdc2 when we examined these cells by immunofluorescence (Fig. 4,A). There was an increase in cytoplasmic distribution of P-Tyr15-p34-cdc2, particularly in the SeMet-treated PC-3 and DU145 prostatic carcinoma cells (Fig. 4,A). These data are particularly interesting in view of the recent finding that p34-cdc2 is preferentially located in the cytoplasm of human head and neck squamous cell carcinomas (43). In our studies, elevated levels of cdc2 phosphorylation were particularly apparent after SeMet treatment of tumor cell lines. These findings were verified when we examined the phosphorylation status of cdc2 by Western blotting, which revealed a significant increase in P-Tyr15-p34-cdc2(Fig. 4 C). There may be a negative regulation of cdc2 phosphorylation involving WEE1 that is influenced by p53. Leach et al. (14) reported recently that WEE1 protein kinase may be down-regulated because of p53 activation that resulted in the cumulative dephosphorylation of cdc2, leading to growth arrest and apoptosis. The inactivation of p53 using antisense oligonucleotides, followed by UV irradiation in normal human fibroblasts, was also shown to enhance sensitivity to DNA-damaging agents, implicating p53 in the protection of normal cells. The importance of our findings that selenium can selectively induce G₂-M arrest in prostatic carcinoma by the involvement of cdc2 are highly significant if these events lead to enhanced cell death or potential radiosensitization (44).

Sinha et al. (45) reported recently that S-G₂-M phase growth arrest occurred after Na₂SeO₃ treatment and G₁ arrest occurred after MSC treatment in asynchronously dividing mouse mammary epithelial cells (TM6; Ref. 45). These findings are consistent with our results using Na₂SeO₃ in asynchronously dividing prostate cancer cells (PC-3 and DU145). In a second study,these authors established how MSC inhibition affected the timing of the cell cycle by examining synchronized TM6 cells (46). In both studies, they attributed growth inhibition of TM6 cells by MSC to G₁ growth arrest (45, 46). Ip et al. (47) also reported that mammary intraductal proliferations induced by methylnitrosourea showed an increase in p27Kip1 (a G₁ checkpoint protein) expression,when rats were fed MSC compared with Na₂SeO₃. In TM6 mammary tumor cells, Na₂SeO₃preferentially increased cells in S-G₂-M and MSC primarily increased cells arrested in G₁. In human prostate cancer cells, we found that Na₂SeO₃ had greater effects on S-phase than did SeMet but lesser effects on cell arrest in G₂-M. These data demonstrate that different chemical forms of selenium can vary in their effects on the cell cycle and that selenium can act differently in different cell systems.

We observed a decrease in anchorage-independent growth in the presence of SeMet that reflects a potential decrease in tumorigenicity. The high sensitivity of colony formation to SeMet [e.g., SeMet substantially more active in anchorage-independent (Fig. 5) than in monolayer (Fig. 1) PC-3 cells] may involve cell-to-cell interactions related to solid tumor development (48, 49).

The results of our present study are relevant to the molecular mechanisms of selenium actions in prostate carcinogenesis, which are only beginning to be elucidated. Our findings that selenium selectively induces growth inhibition and apoptosis in prostate cancer cells versus in prostate normal cells supports the study of selenocompounds for prostate cancer chemoprevention.

We thank Dr. Reuben Lotan for contributions to the manuscript. We also thank Norma Llansa and Laura Schaefer for technical support and Kendall Morse for editorial assistance.