Down-regulation of intrinsic P-glycoprotein expression in multicellular prostate tumor spheroids by reactive oxygen species.

Intrinsic expression of the multidrug resistance (MDR) transporter P-glycoprotein (Pgp) may be regulated by reactive oxygen species (ROS). A transient expression of Pgp was observed during the growth of multicellular tumor spheroids. Maximum Pgp expression occurred in tumor spheroids with a high percentage of quiescent, Ki-67-negative cells, elevated glutathione levels, increased expression of the cyclin-dependent kinase inhibitors p27Kip1 and p21WAF-1 as well as reduced ROS levels and minor activity of the mitogen-activated kinase (MAPK) members c-Jun amino-terminal kinase (JNK), extracellular signal-regulated kinase ERK1,2, and p38 MAPK. Raising intracellular ROS by depletion of glutathione with buthionine sulfoximine (BSO) or glutamine starvation resulted in down-regulation of Pgp and p27Kip1, whereas ERK1,2 and JNK were activated. Down-regulation of Pgp was furthermore observed with low concentrations of hydrogen peroxide and epidermal growth factor, indicating that ROS may regulate Pgp expression. The down-regulation of Pgp following BSO treatment was abolished by agents interfering with receptor tyrosine kinase signaling pathways, i.e. the protein kinase C inhibitors bisindolylmaleimide I (BIM-1) and Ro-31-8220, the p21ras farnesyl protein transferase inhibitor III, the c-Raf inhibitor ZM 336372 and PD98059, which inhibits ERK1,2 activation. ROS involved as second messengers in receptor tyrosine kinase signaling pathways may act as negative regulators of Pgp expression.

The treatment of cancer is limited by either intrinsic or acquired expression of multidrug resistance transporters, including Pgp. 1 The Pgp-mediated MDR phenotype is generally characterized by a decreased cellular drug accumulation owing to an enhanced drug efflux and is caused by the overexpression of the drug transporter (1)(2)(3). Most in vitro studies have been performed using cell cultures selected for drug resistance by repetitive treatment with anti-cancer agents or using recurrent tumors from patients after chemotherapy (3). However, there is increasing evidence that Pgp-mediated MDR is likewise developing in the multicellular context of tumor tissues, without previous treatment with anti-cancer agents (4 -8). This intrinsic Pgp-mediated MDR may represent a so far underestimated common in vivo situation, which may even worsen the efficacy of chemotherapeutic regimen. Especially, renal and prostate cancers are known as intrinsic drug-resistant cancers, which already have a relatively resistant phenotype to anti-cancer agents even without drug selection (9). This may explain the low response of hormone-insensitive prostate tumors toward cytostatic agents (9,10).
In most strategies developed to reverse the MDR phenotype, chemical-reversing agents are applied, which have in common the ability to reverse the MDR phenotype through the binding to the MDR transporters (11). A more efficient strategy to circumvent MDR would be to down-regulate the expression of genes coding for the transporters. This, however, requires knowledge of the molecular mechanisms and signal transduction pathways that are involved in the regulation of MDRrelated genes and elaboration of sophisticated experimental approaches that efficiently down-regulate the drug transporters. Recently, it has been pointed out that the expression of Pgp may be redox-regulated, because down-regulation of Pgp and reversal of the MDR phenotype was achieved by treatment of tumor cells with tumor necrosis factor ␣ (12), which is known to utilize low levels of ROS in its signal transduction pathway (13). On the other hand, it has been reported that high levels of ROS resulting in severe cellular oxidative stress increased the expression of the MRP-1 (14) and MDR-1b (15) genes.
In previous studies we have reported on an intrinsic Pgp MDR, which develops in quiescent cell areas of large, nonnecrotic multicellular prostate tumor spheroids derived from the androgen-independent cell line DU-145 (7,8). We furthermore demonstrated that the MDR phenotype could be reversed and that Pgp could be down-regulated upon incubation of spheroids with chemical agents that raised intracellular ROS and stimulated cell cycle activity (16). The current study presents evidence that the development of intrinsic Pgp expression is a general phenomenon occurring in multicellular tumor spheroids of different origin, i.e. melanoma, hepatoma, and glioma tumor spheroids. It is hypothesized that endogenously generated ROS are regulating the expression of intrinsic Pgp in multicellular tumor spheroids, because small, exponentially growing spheroids, which robustly generate ROS, display reduced Pgp levels, whereas up-regulation of Pgp in large, nonnecrotic tumor spheroids coincides with decreased ROS levels. It is shown that an elevation of intracellular ROS is downregulating Pgp expression via the activation of receptor tyrosine kinase signaling pathways, which results in the phosphorylation of MAPK members and mitogenic stimulation. The modulation of the intracellular redox state by BSO or other ROS-generating agents is apparently suitable in vitro to circumvent the MDR phenotype and may be useful in vivo to promote the efficacy of chemotherapeutic regimen in anti-cancer treatment.

EXPERIMENTAL PROCEDURES
Culture Technique of Multicellular Tumor Spheroids-The human prostate cancer cell line DU-145 and the melanoma cell line IGR were grown in 5% CO 2 /humidified air at 37°C with Ham's F-10 medium (Life Technologies, Inc., Gaithersburg, MD). The glioma cell line Gli36 and the hepatoma cell line Hepa1 were grown in Dulbecco's modified Eagle's medium (Life Technologies, Inc.). The cell culture media were supplemented with 10% fetal calf serum (Roche Molecular Biochemicals, Mannheim, Germany), 2 mM glutamine, 0.1 mM ␤-mercaptoethanol, 2 mM minimal essential amino acids, 100 IU/ml penicillin, and 100 g/ml streptomycin (ICN Flow, Meckenheim, Germany). Spheroids were grown from single cells. Cell monolayers were enzymatically dissociated with 0.2% trypsin, 0.05% EDTA (ICN Flow) and seeded in siliconized 250-ml spinner flasks (Integra Biosciences, Fernwald, Germany) with 250 ml of complete medium and agitated at 20 rpm using a Cell Spin stirrer system (Integra Biosciences). Cell culture medium was partially (125 ml) changed every day.
Incubation with BSO, Glutamine-reduced Medium, and Ebselen-For incubation with BSO and glutamine-reduced medium, small multicellular tumor spheroids (diameter 130 Ϯ 50 m, day 5) were transferred to bacteriological tissue culture plates (diameter 10 cm) (Becton Dickinson, Meylan, France) filled with 10 ml of F10 cell culture medium. They were subsequently treated for 7 days with 50 M BSO (Sigma, Deisenhofen, Germany), which was added to the cell culture medium. For incubation in glutamine-reduced medium, F10 cell culture medium was supplemented with 10% fetal calf serum, 0.1 mM ␤-mercaptoethanol, 2 mM minimal essential amino acids, 100 IU/ml penicillin, and 100 g/ml streptomycin, whereas glutamine was omitted as an additive. Tumor spheroids were cultivated for 7 days in the glutaminereduced medium. The cell culture medium was changed every 24 h. Ebselen (2-phenyl-1,2-benzisoselenazol-3(2H)-one) (Sigma) was dissolved in Me 2 SO and was used at a final concentration of 1 M.
Doxorubicin Fluorescence Recording and Confocal Laser Scanning Microscopy-The fluorescent cytostatic anti-cancer agent doxorubicin (Sigma Chemical Co.) (excitation at 543 nm, emission, long-pass filter LP 570 nm) was used at a concentration of 10 M unless otherwise indicated. Multicellular tumor spheroids were loaded with doxorubicin at 37°C for 90 min. After recording of doxorubicin fluorescence by confocal laser scanning microscopy, spheroids were transferred to doxorubicin-free cell culture medium and incubated at 37°C for 30 min. Following this incubation doxorubicin retention in multicellular spheroids was evaluated.
Fluorescence recordings were performed by means of a confocal laser scanning setup (LSM 410, Carl Zeiss, Jena, Germany), connected to an inverted microscope (Axiovert 135, Carl Zeiss). Fluorophores were excited using a 0.5-milliwatt helium-neon laser (single excitation at 543 nm). A 25ϫ, numerical aperture 0.8, oil immersion-corrected objective (Carl Zeiss) was applied. Doxorubicin fluorescence was measured in 3600-m 2 areas in a depth up to 80 m from the spheroid periphery.
Immunohistochemical Techniques and Quantitative Immunohistochemistry-Antibody staining was performed on whole mount multicellular spheroids. The monoclonal anti-p27 Kip1 and p21 WAF-1 antibodies were obtained from PharMingen (Hamburg, Germany) and were used in a concentration of 2.5 g/ml. The monoclonal antibodies directed against the NADPH oxidase subunits p47 phox and p67 phox (Dianova Hamburg, Germany) were used in a concentration of 5 g/ml. The polyclonal anti-mdr (Ab-1) antibody (Oncogene Research Products, Cambridge, UK) was used in a concentration of 5 g/ml. The anti-active MAPK polyclonal antibody directed against ERK1,2 (dilution 1:20), the anti-active JNK (dilution 1:20), and the anti-active p38 MAPK (dilution 1:20) polyclonal antibodies were obtained from New England BioLabs (Beverly, MA). The monoclonal anti-proliferating cell protein Ki-67 antibody was obtained from Sigma and used in a concentration of 7.5 g/ml. Spheroids were washed in phosphate-buffered saline (PBS), fixed in either ice-cold methanol/acetone (7:3) or 4% paraformaldehyde (4°C) and permeabilized in PBS supplemented with 1% Triton X-100 (PBST) (Sigma). Subsequently, they were incubated for 1 h in PBST 0.01% containing 10% fat-free milk powder to reduce nonspecific binding and for further 2 h with primary antibody. After washing three times in PBST (0.1%), the spheroids were incubated for 60 min in PBST (0.01%) supplemented with 10% milk powder and either a Cy5-conjugated F(abЈ) 2 fragment goat anti-mouse IgG (HϩL) (concentration 3.25 g/ml), a Cy5-conjugated F(abЈ) 2 fragment goat anti-rabbit IgG (HϩL) (concentration 4.6 g/ml), or a Cy5-conjugated F(abЈ) 2 fragment goat anti-rat (concentration 4.6 g/ml) (all from Dianova, Hamburg, Germany). Excitation was performed using a 633-nm helium-neon laser of the confocal setup. Emission was recorded using a long-pass LP 655-nm filter set.
For quantitative immunohistochemistry, confocal images were recorded from whole mount multicellular spheroids stained with only secondary antibodies (background fluorescence image) and spheroids stained with primary and secondary antibodies. The pinhole settings of the confocal setup were adjusted to yield optical slices of 10-m thickness. After subtraction of background fluorescence, the fluorescence signal (counts) was evaluated by the image analysis software of the confocal setup in 3600-m 2 areas of interest and was routinely exported for further analysis to the Sigma Plot graphic software.
Determination of the Intracellular Redox Levels-Intracellular redox levels were measured using the fluorescent dye 2Ј,7Ј-dichlorodihydrofluorescein diacetate (H 2 DCFDA) (Molecular Probes), which is a nonpolar compound that is converted into a non-fluorescent polar derivative (H 2 DCF) by cellular esterases after incorporation into cells. H 2 DCF is membrane-impermeable and is rapidly oxidized to the highly fluorescent 2Ј,7Ј-dichlorofluorescein (DCF) in the presence of intracellular ROS (17). For the experiments, multicellular tumor spheroids were incubated in E1 medium containing (in millimolar) NaCl 135, KCl 5.4, CaCl 2 1.8, MgCl 2 1, glucose 10, HEPES 10 (pH 7.4 at 23°C), and 20 M H 2 DCFDA dissolved in dimethyl sulfoxide (Me 2 SO) was added. After 5, 10, 15, 20, and 30 min intracellular DCF fluorescence (corrected for background fluorescence) was evaluated in 3600-m 2 regions of interest using an overlay mask. For fluorescence excitation, the 488-nm band of the argon ion laser of the confocal setup was used. Emission was recorded using a long-pass LP515-nm filter set.
Determination of Intracellular Glutathione-Glutathione levels were determined using a commercially available glutathione assay kit (Calbiochem). Tumor spheroids were enzymatically dissociated in 0.1% trypsin/0.05% EDTA. An aliquot of 3.5 ϫ 10 6 cells was lysed in 0.5 ml of meta-phosphoric acid, freeze-thawed, and centrifuged (3000 ϫ g) for 10 min. 50 l of the supernatant was used for the assay. The substitution product obtained with reduced glutathione was transformed under alkaline conditions into a chromophoric thione with a maximal absorbance at 400 nm. The absorbance of the samples was determined using a Dynex (Reutlingen, Germany) microplate reader.
Reverse Transcriptase-Polymerase Chain Reaction (RT-PCR)-MDR-1 gene expression was monitored by RT-PCR. After the indicated growth times for tumor spheroids, culture medium was aspirated and spheroids were washed in PBS. Subsequently the cells were lysed with 1 ml of RNA-clean (PqLab, Erlangen, Germany) and processed according to the manufacturer's protocol to obtain total cellular RNA. 1-g aliquots were electrophoresed through 1.2% agarose-0.67% formaldehyde gels and stained with ethidium bromide to verify the quantity and quality of the RNA. 1 g of the isolated total RNA was reverse-transcribed using random primers and Moloney murine leukemia virus reverse transcriptase (Life Technologies, Inc.) for 60 min at 42°C and 10 min at 70°C. The single-stranded cDNA was amplified by polymerase chain reaction using Taq DNA polymerase (Sigma). 35 cycles were performed under the following conditions: 20 s, 94°C; 30 s, 60°C; 60 s, 72°C. The sequence for hMDR-1 sense and antisense primers were 5Ј-CCCATCATTGCAATAGCAGG-3Ј and 5Ј-G-TTCAAACTTCTGCTCCTGA-3Ј (18). PCR amplification gave a single 167-base pairs (bp) fragment originated from hMDR-1 mRNA. For semi-quantification, PCR conditions were chosen so that the reaction was within the linear exponential phase with respect to the amount of cDNA template and number of cycles performed. Equal amounts of RT-PCR products were loaded on 1.5% agarose gels, and optical densities of ethidium-bromide-stained DNA bands were quantified.
Statistical Analysis-Data are given as mean values Ϯ S.D., with n denoting the number of experiments. In each experiment at least 20 multicellular spheroids were analyzed. Student's t test for unpaired data was applied as appropriate. A value of p Ͻ 0.05 was considered significant.  (Fig. 3). It was evidenced that MDR-1 expression paralleled Pgp expression, i.e. it was low in small tumor spheroids and increased with the development of quiescent cell areas in 5-to 10-day-old tumor spheroids. MDR-1 expression was decreased in large tumor spheroids, which have been previously shown to contain central necrosis (27-day-old) (n ϭ 3) (19).

Transient Expression of Pgp in Multicellular Tumor
To correlate the increased expression of Pgp in DU-145 tumor spheroids with the proliferation status of the tumor cells, the tissue levels of the CDK inhibitors p27 Kip1 and p21 WAF-1 as well as the expression of the proliferation-associated antigen Ki-67 were analyzed. It was observed that p27 Kip1  inversely correlated to the expression of Pgp and declined with increasing age and size of tumor spheroids, indicating a decrease in proliferative activity (n ϭ 3).
The apparent correlation of cell quiescence and increased Pgp expression was furthermore corroborated by investigation of the intrinsic activity of the MAPK members ERK1,2, JNK, and p38 MAPK, which are involved in mitogenic and stressactivated signaling pathways (Fig. 5). As expected, the activity of all investigated MAPK members was low in large, nonnecrotic DU-145 tumor spheroids (diameter 300 Ϯ 50 m, days 14 -15 of tumor spheroid culture), which expressed maximum levels of Pgp (see Fig. 1). Increased JNK activity was observed in small tumor spheroids (diameter 100 Ϯ 50 m, days 3-6 of tumor spheroid culture), as well as large tumor spheroids containing central necrosis (diameter 500 Ϯ 100 m, days 18 -40 of tumor spheroid culture) (n ϭ 3), whereas increased ERK1,2 activity was observed solely in exponentially growing small tumor spheroids. A low basal activity of p38 MAPK was observed in all size classes of tumor spheroids under investigation (n ϭ 3).
Endogenous ROS Generation and Glutathione Content in Multicellular Tumor Spheroids-Cancer cells endogenously generate ROS (21), which are involved in signaling pathways maintaining cell proliferation (22)(23)(24). Hence, it is expected that endogenous ROS generation is high in small tumor spheroids consisting predominantly of proliferating cells, whereas it is low in large, non-necrotic, tumor spheroids, which contain extended areas of cell cycle-inactive, quiescent cells. To determine ROS generation in multicellular DU-145 tumor spheroids, small, large non-necrotic as well as tumor spheroids containing extended central necrosis were stained with the redox-sensitive dye H 2 DCFDA, and the time course of ROS generation was monitored. Indeed our data show that the generation of ROS (evaluated after an incubation time of 20 min with H 2 DCFDA) was most pronounced in small spheroids with 280 Ϯ 25% of the fluorescence increase observed in large, non-necrotic tumor spheroids (set to 100%) (n ϭ 3). In tumor spheroids containing central necrosis, the ROS generation was significantly resumed and amounted to 200 Ϯ 25% of the ROS generation observed in large, non-necrotic spheroids (n ϭ 3) (Fig. 6A).
Intracellular ROS in DU-145 are presumably generated via an NADPH-oxidase-like enzyme, which may be differently expressed during the growth of tumor spheroids. Therefore, tumor spheroids were immunostained for the NADPH-oxidase subunits p47 phox and p67 phox . It was demonstrated that small tumor spheroids exerted high levels of NADPH-oxidase expression, whereas in large non-necrotic tumor spheroids a signifi-cant down-regulation of NADPH oxidase expression was observed (Fig. 6B). In large tumor spheroids containing central necrosis, p47 phox and p67 phox expression was partially resumed which, however, did not reach statistical significance (n ϭ 3).
Intracellular ROS levels are counterbalanced by the antioxidative defense system. Therefore, the intracellular content of glutathione which is an important cellular anti-oxidative thiol, was determined in tumor spheroids after different times of tumor spheroid culture (Fig. 6C). Our data demonstrate that the glutathione content was inversely correlated to ROS levels in tumor spheroids. Low levels of glutathione were determined in small tumor spheroids (6.75 Ϯ 0.09 mM on day 3 of cell culture, n ϭ 3), which generated ROS to significant amounts. However, elevated glutathione levels were found in large, nonnecrotic tumor spheroids with a maximum at day 12 of tumor spheroid cell culture (11 Ϯ 0.85 mM, n ϭ 3) corresponding to spheroid diameters of ϳ250 m. This size class of tumor sphe- Pgp expression obtained from the gel was analyzed by densitometric analysis. Shown is a representative of three independent experiments, which yielded comparable results. C, RT-PCR analysis of MDR-1 gene expression in respect to the cultivation time of multicellular tumor spheroids. After different times of cultivation, as indicated, RNA was prepared from tumor spheroids and the MDR-1-specific (167bp) PCR products were amplified and analyzed by gel electrophoresis, as described under "Experimental Procedures." MDR-1 mRNA expression obtained from the gel was analyzed by densitometric analysis. Shown is a representative of three independent experiments which yielded comparable results. roids consequently generated low levels of ROS (see Fig. 6A). Prolonged culture times (Ͼ12 days) resulted in a decline of the glutathione content, which is in line with the observed elevated ROS levels in large tumor spheroids containing extended central necrosis.

Effects of Intracellular ROS Elevation on Pgp Expression in
Multicellular Tumor Spheroids-The working hypothesis of our present study is based on the assumption that the intracellular redox state of tumor spheroids regulates the expression of the MDR transporter Pgp. To address this issue, intracellular glutathione was reduced either by treatment of small tumor spheroids for 7 days, i.e. from day 5 to day 12 of cell culture with 50 M BSO, which is an irreversible inhibitor of ␥-glutamylcysteine synthetase, the rate-limiting enzyme in glutathione biosynthesis (25) or by incubating tumor spheroids in glutamine-reduced cell culture medium (26). Both experimental conditions significantly increased ROS levels as compared with untreated control spheroids (Fig. 7) (n ϭ 3) and significantly stimulated tumor spheroid growth (data not shown). The increased ROS levels could be completely inhibited by addition of ebselen, which is an effective free radical scavenger (27). Elevation of intracellular ROS in large, non-necrotic multicel-  Note that the glutathione content is elevated in large non-necrotic tumor spheroids (days 7-12). The data were fitted by a Gaussian algorithm. lular spheroids by BSO and glutamine-reduced medium resulted in a significant down-regulation of Pgp (Fig. 8A) from 405 Ϯ 92% to 183 Ϯ 31% and 194 Ϯ 42%, respectively (n ϭ 3). The effects of intracellular elevation of ROS on the MDR phenotype were evaluated by doxorubicin retention experiments. It was observed that incubation of tumor spheroids with BSO and in glutamine-reduced medium resulted in a significant increase in doxorubicin retention from 63 Ϯ 1% to 94 Ϯ 1.5% and 92 Ϯ 6%, respectively (n ϭ 3) (Fig. 8B). The modulatory effects of elevated ROS on the levels of Pgp and on doxorubicin retention could be efficiently reversed when 1 M of the free radical scavenger ebselen was coadministered, indicating that the observed down-regulation of Pgp by glutathione depletion was related to an increase of the intracellular redox state.
Redox-regulation of Pgp may be dose-dependent. To evaluate the effects of increasing ROS concentrations on the expression levels of Pgp, 6-day-old tumor spheroids displaying intermediate Pgp levels were incubated in cell culture media containing increasing concentrations of H 2 O 2 ranging from 1 to 750 M. In parallel experiments cell lethality was assessed by staining dead cells with the cell death marker SYTOX green. Concentrations of 1 M (n ϭ 3) and 200 M (n ϭ 3) H 2 O 2 significantly down-regulated Pgp expression to 128 Ϯ 9% and 126 Ϯ 18%, respectively, whereas Pgp expression was significantly increased to 285 Ϯ 22%, when 750 M H 2 O 2 was administered (Fig. 9A). However, under these conditions cell lethality was significantly augmented (Fig. 9B), indicating that severe oxidative stress results in up-regulation of Pgp, whereas nontoxic low levels of ROS induced down-regulation of Pgp and reversal of the MDR phenotype.
Effects of Intracellular ROS Elevation on p27 Kip1 Levels as Well as ERK1,2 and JNK Activity-H 2 O 2 is known as a medi-ator of cell cycle activity in multicellular tumor spheroids of the DU-145 cell line (28,29). The observed changes in the expression of Pgp may therefore be related to cell cycle stimulation. To address this issue, large non-necrotic tumor spheroids were either incubated with BSO or glutamine-reduced culture medium and the expression of either p27 Kip1 (Fig. 10A) or the activity of ERK1,2 (Fig. 10B) and JNK (Fig. 10C) was assessed. Our data clearly indicate that elevation of intracellular ROS by BSO resulted in a significant down-regulation of p27 Kip1 from 383 Ϯ 90% to 70 Ϯ 50% (n ϭ 3). Comparable results were achieved following incubation with glutamine-reduced medium (data not shown). Consequently, BSO treatment resulted in an increase in phosphorylated ERK1,2 and JNK from 100 Ϯ 25% to 200 Ϯ 25% and 184 Ϯ 46%, respectively (n ϭ 3). The observed effects were efficiently inhibited in the presence of 1 M ebselen. Hence, our data support the notion that ROS-mediated down-regulation of Pgp is accompanied by a stimulation of ERK1,2 and JNK pathways.

Involvement of Receptor Tyrosine Kinase-mediated Signaling Pathways in the Regulation of Pgp Expression-
The data of the present study suggest that the expression of Pgp is associated with cell quiescence and can be down-regulated by mitogenic stimulation of tumor spheroids with low levels of ROS. Furthermore, it has been recently shown that hydrogen peroxide can activate signaling pathways utilized by ROSgenerating growth factors (30). We therefore hypothesized that: 1) mitogenic stimulation by the use of EGF, which has previously been shown to generate intracellular ROS (31), should likewise down-regulate Pgp, 2) the down-regulation of Pgp expression by BSO resulted from an activation of EGFmediated signal transduction pathways (Fig. 11). To verify these assumptions large, non-necrotic tumor spheroids were either treated with tyrphostin AG 1478 (10 M), which selectively inhibits EGF receptor kinase, or incubated for 24 h in serum-free medium and in serum-free medium supplemented with EGF in a concentration range of 2.5-50 ng/ml. The serum-free conditions were chosen to avoid a possible influence of EGF present in the fetal calf serum of the cell culture medium on the outcome of the experiments. Tyrphostin AG 1478 treatment as well as incubation in serum-free medium resulted in a significant up-regulation of Pgp as compared with control conditions. Upon treatment with 2.5, 5, 20, and 50 ng/ml EGF a dose-dependent down-regulation of Pgp to 70 Ϯ 15%, 61 Ϯ 19%, 34 Ϯ 7%, and 55 Ϯ 10%, respectively, was observed as compared with serum-free conditions (set to 100%) (n ϭ 3) (Fig. 11A). Coadministration of 20 ng/ml EGF and the free radical scavenger ebselen (1 M) totally abolished the down-regulation of Pgp observed by treatment with EGF alone, indicating that the down-regulation of Pgp was mediated via ROS. In a parallel set of experiments tumor spheroids were treated with BSO in the presence of specific antagonists of the EGF signaling pathway, i.e. the PKC inhibitors BIM-1 (1 M) and Ro-31-8220 (0.5 M), the Ras antagonist FPT inhibitor III (1 M), the Raf inhibitor ZM 336372 (1 M), and the MEK1 inhibitor PD98059 (20 M), respectively (n ϭ 3) (Fig. 11B). The ROS-mediated downregulation of Pgp by BSO treatment was completely abolished upon inhibition of the Ras-mediated tyrosine kinase signaling pathway, which clearly indicates that EGF-mediated signal transduction pathways are negatively regulating Pgp expression. None of the applied antagonists of receptor tyrosine kinase pathways exerted significant effects on intracellular ROS levels (data not shown).

DISCUSSION
The present study reports on the correlation of endogenous intracellular ROS levels and the intrinsic expression of Pgp in multicellular prostate tumor spheroids of different size classes. The transient expression of Pgp during the growth of tumor spheroids, and its correlation with cell quiescence is apparently a general feature of tumor tissues, because tumor spheroids cultivated from glioma, melanoma, and hepatoma tumor cells displayed comparable features of intrinsic expression of Pgp as observed in the DU-145 prostate cancer cell line. The observation of a transient increase of Pgp-mediated drug resistance during the growth of avascular micrometastases may be of clinical significance, because the outcome of chemotherapeutic treatment critically depends on the chemosensitivity of the treated neoplastic tissue, which may be predicted from levels of Pgp expression as well as the expression of cell cycle-related genes in tumor biopsies. In DU-145 tumor spheroids the intrinsic Pgp expression was clearly correlated to increased levels of the CDK inhibitors p27 Kip1 and p21 WAF-1 , which are increased during cell cycle arrest. Consequently, tumor spheroids with elevated levels of Pgp expression displayed minor endogenous activity of the cell growth-associated MAPK-members ERK1,2 and JNK. There is increasing evidence that the growth of neoplastic tissues is controlled by the generation of ROS, which have been previously shown to arise in Ras-mediated signaling pathways (23). The data of the present study are in notion with these previous studies, because it is demonstrated that increased ROS generation was observed in small tumor spheroids that consist predominantly of exponentially growing Ki-67positive cells as well as in large tumor spheroids containing extended central necrosis. The increase in Ki-67 expression was paralleled by an augmented activity of JNK in small spheroids and large spheroids containing central necrosis as well as an increased activity of ERK1,2 in small tumor spheroids.
The correlation of elevated Pgp levels with reduced levels of ROS as well as diminished activity of ERK1,2 and JNK led us to hypothesize that raising intracellular ROS may down-regulate Pgp expression via up-regulation of ERK1,2 and JNK. An increase in intracellular ROS was achieved by: addition of hydrogen peroxide to the cell culture medium, depletion of intracellular glutathione by the use of BSO and glutaminereduced cell culture medium, as well as treatment with EGF, which has been previously demonstrated to raise intracellular ROS (31). All these experimental procedures resulted in significant down-regulation of Pgp levels, indicating that ROS are involved in the signaling pathways that regulate the expression of Pgp. Interestingly, with hydrogen peroxide concentrations exceeding 500 M an up-regulation of Pgp was observed. A rise in Pgp expression following treatment of cells with hydrogen peroxide has been previously demonstrated (15). In this study millimolar concentrations of H 2 O 2 were adminis-tered, resulting in a parallel up-regulation of poly-(ADP-ribose) polymerase, which is a nuclear enzyme induced by DNA strand breaks. Furthermore, stress-induced expression of Pgp has been recently evidenced under various conditions, i.e. UV radiation (32), low external pH and osmotic shock (33) as well as heat stress (34). Most of these stress factors are accompanied by the generation of intracellular ROS. Under stress conditions the physiological role of Pgp may be related to the prevention of apoptosis (35), presumably by regulating caspase-dependent apoptotic pathways (36). This notion is furthermore supported by studies that demonstrated that overexpression of Pgp is correlated with increased levels of the anti-apoptotic Bcl-xL protein (37). ROS acting as signaling agents in the regulation of cell proliferation are in the nanomolar to micromolar range and are therefore significantly lower than the concentrations necessary to induce DNA strand breakage (38). Hence, intracellular ROS may exert differential effects on the regulation of MDR-related genes, which are dependent on their intracellular concentration.
In the present study raising intracellular ROS by depletion of glutathione resulted in a decline of p27 Kip1 expression, whereas activity of ERK1,2 and JNK was significantly elevated, which indicates that in parallel to the down-regulation of Pgp expression a stimulation of the cell cycle was achieved, suggesting that cell cycle stimulation by ROS and the regulation of Pgp expression may share a common signal transduction pathway. This assumption was supported by experiments demonstrating that pharmacological inhibition of members of the receptor tyrosine kinase pathway utilized by EGF, and involving PKC and Ras/Raf as well as ERK1,2, abolished the observed down-regulation of Pgp following elevation of intracellular ROS. Interestingly, incubation of tumor spheroids in serum-free media and treatment with tyrphostin, which inhibits protein tyrosine kinases, including autophosphorylation of EGF receptor kinase, significantly increased the expression of Pgp in multicellular tumor spheroids. These findings support the notion that EGF present in the fetal calf serum used in the cell culture medium and/or EGF secreted by tumor cells may regulate Pgp expression via the activation of receptor tyrosine kinase signaling pathways.
The apparent effects of ROS on Pgp expression levels and cell cycle activity in multicellular tumor spheroids may be clinically utilized in anti-cancer trials. In EMT-6 tumor spheroids downregulation of p27 Kip1 , which has been discussed as a regulator of MDR (39), by antisense nucleotides increased cell proliferation and sensitized tumor cells to 4-hydroperoxycyclophosphamide (39). A comparable chemosensitization in EMT-6 tumor spheroids was observed by treatment with hyaluronidase, which reduced intercellular communication and activated cell proliferation (40). Furthermore, stimulation of cell proliferation in noncycling hematopoietic progenitor cells and leukemic blasts resulted in a down-regulation of Pgp-mediated MDR (41). An elevation of ROS by depletion of intracellular glutathione sounds reasonable, because BSO has successfully been applied in vivo in cancer patients and was shown to increase the sensitivity of cancer cells toward antineoplastic drugs (42). Because most antineoplastic agents are acting against rapidly cycling cancer cells, the concomitant cell cycle stimulation and down-regulation of Pgp by ROS may offer new avenues for conventional chemotherapy. This may hold especially true for hormone-independent metastatic prostate tumors, which are characterized by a high fraction of cells resting in the G 0 phase of the cell cycle and a pronounced intrinsic resistance to a broad range of anti-cancer agents (9). FIG. 11. Regulation of Pgp expression in multicellular DU-145 tumor spheroids by receptor tyrosine kinase pathways. A, effects of tyrphostin AG 1478 (10 M), which selectively inhibits EGF receptor kinase, serum-free conditions, and different concentrations of EGF on the expression of Pgp in large, non-necrotic tumor spheroids (diameter 300 Ϯ 50 m, day 14 of tumor spheroid culture). After 24 h of incubation the tumor spheroids were fixed and Pgp expression was quantified by immunohistochemistry. The data represent the relative Pgp immunofluorescence (%) in relation to the Pgp immunofluorescence obtained under serum-free conditions (set to 100%). Note, that the down-regulation of Pgp by EGF was totally abolished upon coadministration of the free radical scavenger ebselen. *, p Ͻ 0.05, significantly different from serum-free conditions. B, effects of antagonists of receptor tyrosine kinase pathways on the down-regulation of Pgp by BSO. Note, that all applied antagonists of receptor tyrosine kinase pathways totally abolished the BSO-mediated down-regulation of Pgp. *, p Ͻ 0.05, significantly different from the untreated control.