Inhibition of I (cid:1) B Kinase Activity by Acetyl-boswellic Acids Promotes Apoptosis in Androgen-independent PC-3 Prostate Cancer Cells in Vitro and in Vivo

Signaling through NF- (cid:1) B has been implicated in the malignant phenotype as well as the chemoresistance of various cancers. Here we show that the natural compounds acetyl- (cid:2) -boswellic acid and acetyl-11-keto- (cid:2) -bo-swellic acid (AK (cid:2) BA) inhibit proliferation and elicit cell death in chemoresistant androgen-independent PC-3 prostate cancer cells in vitro and in vivo . Induction of apoptosis was demonstrated in cultured PC-3 cells by several parameters including mitochondrial cytochrome c release and DNA fragmentation. At the molecular level these compounds inhibit constitutively activated NF- (cid:1) B signaling by intercepting the I (cid:1) B kinase (IKK) activity; signaling through the interferon-stimu-lated response element remained unaffected, suggesting specificity for IKK inhibition. The impaired phosphorylation of p65 and the reduced nuclear translocation of NF- (cid:1) B proteins were associated with down-regulation of the constitutively overexpressed and NF- (cid:1) B-dependent antiapoptotic proteins Bcl-2 and Bcl-x L . In addition, ex- pression probed with specific antibodies, and detected as described previously (20, 21). Mouse monoclonal antibodies were used for actin (Chemicon), p65 phosphorylated at Ser-536 (Cell Signaling Technology), and Bcl-2 (Biomedicals, Switzerland), whereas rabbit polyclonal antibodies were used for I (cid:1) B (cid:3) (Santa Cruz), Bcl-x L , and cyclin D1 (both from Dianova, Germany). Reverse Transcriptase-PCR— Total RNA (1.5 (cid:5) g) extracted using Trizol (Invitrogen) was analyzed by reverse transcriptase-PCR with Moloney murine leukemia virus reverse transcriptase and Taq DNA polymerase (Invitrogen). Primers for Bcl-2, Bcl-x L , and cyclin D1 were as described (22, 23). Conditions were such that the PCR reactions did not reach the saturation phase. Control experiments showed no DNA contaminations. Normalization was carried out using glyceraldehyde 3-phosphate dehydrogenase as the internal standard (24). The PCR products had the appropriate size, and their identity was confirmed by direct sequencing (Abi Prism 310, Applied Biosystems). Cytotoxicity Apoptosis Cytotoxicity was quantified by the release of lactate dehydrogenase from apoptotic cells after the indicated times using a kit (Roche Diagnostics) according to the manufacturer’s instructions. The total amount of lactate dehydrogenase in appropriately treated cell lysates served as a positive control. Additionally, cell proliferation was analyzed by monitoring the mitochondrial reduction of XTT (Roche Diagnostics). P]ATP. blots for IKKs. M GST-IKK (cid:2) fusion M or (cid:2) precipitated cell with 1 (cid:5) g/ml lipopolysaccharide for Extracellular signal-regulated kinase (ERK2) in to the IKK assays using the immu- noprecipitated ERK2 and recombinant GST-Elk-1 fusion protein (Cell as substrate.

Prostate cancer is the most frequently diagnosed malignancy in males in the United States and most other industrialized Western countries (1). It proceeds from a localized, androgendependent disease to an advanced, invasive, and metastatic disease associated with the loss of androgen dependence. A variety of antitumor agents has been tested in patients with androgen-independent prostate cancer, yet without any significant prognostic benefit (1,2). Once the androgen-independent phase develops, prostate cancer is always lethal. Therefore, unconventional strategies targeting novel molecular mechanisms are urgently required.
The development of incurable androgen-independent prostate cancer is apparently mainly due to a lack of an apoptotic response allowing survival of the cells within primary or metastatic tumors (3). The transcriptional activator, nuclear factor B (NF-B), 1 has recently gained much attention as an inhibitor of apoptosis because it was identified as an activator of several target genes that block induction of apoptosis by various pro-apoptotic stimuli including chemotherapeutic agents (3)(4)(5). In chemoresistant androgen-independent prostate cancer cells, NF-B is constitutively activated due to an enhanced turnover of the inhibitor proteins of NF-B, termed IBs (3). Therefore, intercepting NF-B signaling might be an attractive antitumor approach (4 -6).
Secondary metabolites from plants are a rich source of antitumor compounds, and a variety of plant-derived drugs including taxanes, vinca alkaloids, epipodophyllotoxins, and camptothecin analogues have been approved as antitumor drugs (7). We have recently identified the acetyl-boswellic acids (ABAs), acetyl-␤-boswellic acid (A␤BA) and acetyl-11-keto-␤-boswellic acid (AK␤BA), as cytotoxic compounds (8 -10) that can be isolated to chemical homogeneity from the gum resins of various Boswellia species (11). These gum resins are known as frankincense, and extracts thereof have been used in traditional medicine and recently in small clinical trials for the treatment of chronic inflammatory diseases such as rheumatoid arthritis and inflammatory bowel disease (11,12). Previous studies from our laboratory and others have suggested that AK␤BA possesses anti-tumor activity in several malignant glioma (8,13), leukemia (9), and colon cancer (14) cell lines (15,16), which could at least be partially due to the inhibition of human topoisomerases (17).
Here we show that A␤BA and AK␤BA induce apoptosis in the chemoresistant androgen-independent prostate cancer cell line PC-3. Furthermore, we demonstrate that a water-soluble AK␤BA-␥-cyclodextrin complex inhibits tumor growth and induces apoptosis of pre-established PC-3 xenografts in the chick chorioallantoic membrane (CAM) and in nude mice. Finally, we delineate inhibition of IB kinase (IKK) activity by A␤BA and AK␤BA as the molecular mechanism by which these compounds block the constitutive activation of NF-B signaling in PC-3 cells leading to subsequent inhibition of the antiapoptotic proteins Bcl-2 and Bcl-x L and also cyclin D1, a crucial regulator of cell proliferation. These mechanisms finally promote apoptosis of the androgen-independent prostate cancer cells in vitro and in vivo.

EXPERIMENTAL PROCEDURES
Acetyl-boswellic Acids-A␤BA and AK␤BA were isolated from African frankincense and purified to chemical homogeneity (Ͼ99.9% purity) by reversed phase high performance liquid chromatography as described previously (11,12). The compounds were further characterized by mass spectrometry and one-and two-dimensional nuclear magnetic resonance spectroscopy (11).
For the application of A␤BA or AK␤BA in the xenograft models we developed hydrophilic derivatives by generating ␥-cyclodextrin (CD) complexes that allow administration of the lipophilic compounds in aqueous solutions in vivo. Analysis of the compounds in plasma samples was performed after serial extractions on diatomaceous earth and graphitized carbon black followed by reversed phase high performance liquid chromatography and photodiode array detection as described (12,18).
Cell Culture and Xenografts-PC-3 cells and human normal lung fibroblasts (MRC-5 cells) obtained from the American Type Culture Collection were cultured in Ham's F-12-K and in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum under standard conditions, respectively.
For xenotransplantation of PC-3 cells onto the CAM, fertilized chicken eggs were incubated at 37°C at constant humidity (19). On day 6, 0.5 ϫ 10 6 PC-3 cells in the log growth phase were seeded in 20 l of medium/Matrigel (1:1, v/v) in a 6-mm silicone ring placed on the CAM. Starting from day 2 after seeding, cells were topically treated once daily either with A␤BA-CD, AK␤BA-CD, or the CD vehicle. On day 12 the xenografts were histologically analyzed. For the determination of A␤BA and AK␤BA plasma kinetics, blood anticoagulated with 5 mM EDTA was collected from chorioallantoic veins.
For xenotransplantation male NMRI/nu-nu mice (28 -33 g) (Taconic M&B, Ry, Denmark) were inoculated subcutaneously at both dorsal sides in the subscapular region with 1 ϫ 10 6 PC-3 cells in 0.2 ml of media/injection. After 1 week, treatment commenced by single daily intraperitoneal injections of AK␤BA-CD at 100 mol/kg or equivalent amounts of CD vehicle for additional 3 weeks. The tumor size was measured weekly with a caliper; the tumor volume was calculated according to the formula 0.5 ϫ L ϫ W ϫ T (L ϭ length, W ϭ width, T ϭ thickness). At the end of the experiments the tumor and other tissues were recovered for further histological and pathological analysis.
Invasiveness of PC-3 tumors in mice was rated after macropathological inspection of the tumors as non-invasive, slightly invasive when tumors were in close contact with neighboring tissues, or invasive when the tumors had infiltrated neighboring tissues such as the intercostal muscles or the skin. Blood anticoagulated with 5 mM EDTA was drawn from NMRI mice (30 -34 g) by cardiac puncture; plasma was used for the analysis of AK␤BA plasma kinetics (18). All animal experiments were performed in compliance with the institutional guidelines and relevant laws for animal care and protection of Germany; the research protocols were approved by the relevant authorities.
Immunostaining and Histomorphometric Analysis-The tumor tissues from the CAM experiments were fixed, embedded, and serially sectioned (5 m). Every 100 m the slides were processed for staining and immunohistochemistry for human cytokeratin and Ki-67 proliferation antigen as described (19) (both antibodies were from Dako Corp.). The images were digitally recorded at a magnification of ϫ50 with an Axiophot microscope (Carl Zeiss) and a Sony MC-3249 CCD camera using Visupac version 22.1 software (Carl Zeiss). Areas from the digitalized color photomicrographs were analyzed with SimplePCI 5a (Compix) digital imaging software as the number of pixels corresponding to cells expressing Ki-67 antigen compared with all tumor cells. We statistically analyzed 5-m sections throughout the tumor at a distance 100 m apart from each other.
For confocal laser-scanning microscopy (Leica DM IRBE, Leica Microsystems) of p65 cells were treated with 10 M A␤BA or AK␤BA for 6 h. As a marker for the cytosol, ␣-tubulin was immunostained (Sigma-Aldrich), whereas nuclei were stained with Hoe 33342 (Sigma-Aldrich); for p65 immunofluorescence purified rabbit antibodies were used (Santa Cruz).
Cell Extracts and Western Blot Analysis-Aliquots of mitochondria, cytosol, whole cell lysates, nuclear extracts, or immunoprecipitated IKKs containing equal amounts of protein were separated by SDS-PAGE, transferred, probed with specific antibodies, and detected as described previously (20,21). Mouse monoclonal antibodies were used for actin (Chemicon), p65 phosphorylated at Ser-536 (Cell Signaling Technology), and Bcl-2 (Biomedicals, Switzerland), whereas rabbit polyclonal antibodies were used for IB␣ (Santa Cruz), Bcl-x L , and cyclin D1 (both from Dianova, Germany).
Reverse Transcriptase-PCR-Total RNA (1.5 g) extracted using Trizol (Invitrogen) was analyzed by reverse transcriptase-PCR with Moloney murine leukemia virus reverse transcriptase and Taq DNA polymerase (Invitrogen). Primers for Bcl-2, Bcl-x L , and cyclin D1 were as described (22,23). Conditions were such that the PCR reactions did not reach the saturation phase. Control experiments showed no DNA contaminations. Normalization was carried out using glyceraldehyde 3-phosphate dehydrogenase as the internal standard (24). The PCR products had the appropriate size, and their identity was confirmed by direct sequencing (Abi Prism 310, Applied Biosystems).
Cytotoxicity and Apoptosis Parameters-Cytotoxicity was quantified by the release of lactate dehydrogenase from apoptotic cells after the indicated times using a kit (Roche Diagnostics) according to the manufacturer's instructions. The total amount of lactate dehydrogenase in appropriately treated cell lysates served as a positive control. Additionally, cell proliferation was analyzed by monitoring the mitochondrial reduction of XTT (Roche Diagnostics).
Phosphatidylserine expression (25) on the outer membrane leaflet was detected by confocal microcopy of fluorescein isothiocyanate-labeled annexin V (Roche Diagnostics) bound to the membrane of Hoe 33342-counterstained PC-3 cells pretreated with 10 M A␤BA or AK␤BA for 6 h. There was no nuclear staining with propidium iodide, indicating at this time point no increase in membrane permeability. Alternatively, PC-3 cells and human normal lung fibroblasts (MRC-5) were treated with 10 M AK␤BA for 24 h.
Reduced cellular DNA contents were determined by flow cytometric analysis of the sub-G 0 /G 1 formation in propidium iodide-stained PC-3 cells (26) after treatment with A␤BA, AK␤BA or the topoisomerase I inhibitor camptothecin (Calbiochem) for 48 h; in addition, DNA fragmentation was analyzed by agarose gel electrophoresis of DNA from 3 ϫ 10 6 cells treated with 10 M A␤BA or AK␤BA for the indicated times. Isolation of mitochondria and cytosol by density gradient centrifugation as well as analysis of the mitochondrial release of cytochrome c by Western blot using mouse monoclonal antibodies (BD Pharmingen) were performed as described (27).
For the in situ detection of apoptotic cells in paraffin-embedded tissue sections, the terminal deoxynucleotidyltransferase-mediated dUTP nick end labeling (TUNEL) method was used (Roche Diagnostics). The sections were counterstained with hematoxylin.
Electrophoretic Mobility Shift Assay and Nuclear Translocation of Transcription Factors-Nuclear extracts (5 g/sample) from PC-3 cells with or without pretreatment with A␤BA or AK␤BA for 6 h were subjected to electrophoretic mobility shift assays as described previously (20,28). For competition experiments, nuclear extracts were incubated for 30 min with a 50-fold excess of unlabeled specific NF-B or AP-2 oligonucleotides. Nuclear translocation of transcription factors was quantified in nuclear extracts (5 g) from cells treated either with solvent or ABAs for 6 h using the NF-B TransAM enzyme-linked immunosorbent assay (Active Motif) as described (21). Results are expressed as a fold decrease of the three different Rel proteins in nuclear extracts from the treated samples compared with the control samples and were calculated according to nuclear content in treated cells/nuclear content in control cells.
Luciferase Gene Reporter Assay-HEK293 cells grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum were transiently transfected with the vector pNFB-Luc containing four tandem copies of the B enhancer element upstream of the firefly luciferase reporter gene (Clontech) using Superfect (Qiagen) according to the manufacturer's instruction. After 24 h the cells were pretreated with A␤BA or AK␤BA for 1 h before stimulation with 100 ng/ml TNF-␣ (R&D) for 4 h. Luciferase activities normalized to protein contents were determined in the cell lysates using a PlateLumino (Stratec, Germany). Negative controls transfected with the pTal-Luc vector (Clontech) did not show any luciferase activity after TNF-␣ stimulation. Alternatively, the cells were transfected with an interferon-stimulated response element (ISRE) luciferase reporter gene (Stratagene) alone or together with constitutively active form of IRF-3 (IRF-3 5D) (29). Twenty-four hours post-transfection the medium was replaced, and cells were treated with ABAs or solvent for 6 h.
Surface Plasmon Resonance-Binding of p50/c-Rel and p50/p65 heterodimers to NF-B was measured by surface plasmon resonance (SPR). Double-stranded biotinylated DNA containing the NF-B binding site (AGT TGA GGG GAC TTT CCC AGG C) (Thermo Hybaid) was immobilized on a CMD20 B2 SPR sensor chip (Xantec Analytical Systems) (17,30). Mixtures of human recombinant p50 and c-Rel (both from Promega) or p50 and p65 (Active Motif) (200 nM each) in 60 l of NF-B binding buffer (10 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1 mM MgCl 2 , 0.5 mM EDTA, 0.5 mM dithiothreitol, 0.1% bovine serum albumin, 0.1% Nonidet P40, and 3 g of poly[dI-dC]) were preincubated for 30 min at 20°C with 100 M concentrations of either A␤BA or AK␤BA or equivalent amounts of Me 2 SO. The mixture was applied on the sensor chip, and binding to DNA carrying the NF-B binding site was analyzed with a dual channel IBIS II sensor device (Xantec Analytical Systems). The binding of the recombinant proteins to DNA containing the AP-1 consensus binding sequence (CGC TTG ATG AGT CAG CCG GAA) (Thermo Hybaid) served as a control.
Immunoprecipitation and Kinase Assays-Assays were basically performed as described previously (28). IKKs were immunoprecipitated from lysates of 10 7 PC-3 cells using 4 g of antibodies specific for IKK␣ (Santa Cruz) or IKK␤ (BD Pharmingen). The lysis under stringent conditions allowed disruption of the IKK complex and precipitation of the individual kinases (28). Recombinant p65 and a GST-tagged fusion protein corresponding to full-length IB␣ (Santa Cruz) served as substrates. The IKKs were preincubated with the respective ABAs for 15 min before the addition of substrate and [␥-32 P]ATP. As a loading control, the blots were immunostained for the appropriate IKKs. Alternatively, 30 nM concentrations of active human recombinant GST-IKK␣ or His-IKK␤ fusion proteins (Upstate) were treated with 0.1-10 M concentrations of either A␤BA or AK␤BA or solvent and analyzed as indicated above. Extracellular signal-regulated kinase 2 was immunoprecipitated from cell lysates of human monocytes stimulated with 1 g/ml lipopolysaccharide (Escherichia coli serotype 055:B5, Sigma-Aldrich) for 30 min. Extracellular signal-regulated kinase (ERK2) kinase assays were performed in analogy to the IKK assays using the immunoprecipitated ERK2 and recombinant GST-Elk-1 fusion protein (Cell Signaling Technology) as substrate.
Antisense Oligodeoxynucleotide Experiments-For in vitro knockdown of IKK␣ or I⌲⌲␤, phosphorothioate oligodeoxynucleotides (ODN) (Thermo Hybaid) were used. The ODN were selected on the basis of the major predicted secondary structures, i.e. loops (31). The antisense ODN used against IKK␣ and IKK␤ were 5Ј-CAATTATTTTATG-TATT-3Ј and 5Ј-GTCGACGGTCACTGTGT-3Ј, respectively; the sense ODNs were used as controls. Based on a Blast search, the selected sequences did not show any similarity to any other mRNA sequence. PC-3 cells were transfected with 0.1 M IKK-specific sense or antisense phosphorothioate ODNs using Lipofectin (Invitrogen) according to the manufacturer's instructions for 5 h (32). After transfection, media were replaced by Ham's F-12-K supplemented with 10% fetal calf serum, and the cells were allowed to recover for 20 h. Subsequently, the cells were incubated in Ham's F-12-K supplemented with 1% fetal calf serum, and cell proliferation was determined by XTT assay after 4 h.
Statistical Analysis-The results are expressed as the mean Ϯ S.E. Significance was analyzed using the Newman-Keuls test for multigroup comparison or the Wilcoxon rank sum test when appropriate.

A␤BA and AK␤BA Are Cytotoxic for PC-3 Cells in Vitro-
To determine the cytotoxic activity of A␤BA and AK␤BA in the PC-3 cells, we measured release of lactate dehydrogenase from cells damaged by exposure to the ABAs for 12-48 h (Fig. 1, A-C). Viability in terms of lactate dehydrogenase release decreased in A␤BA-and AK␤BA-treated cells with increasing exposure time. These data show that A␤BA and AK␤BA exert a concentration-and time-dependent cytotoxicity on androgenindependent prostate cancer cells. Analysis of the PC-3 cell proliferation in the presence and absence of A␤BA and AK␤BA by the XTT test yielded basically the same results, although the inhibition of proliferation preceded the cytotoxic effects (data not shown).
A␤BA and AK␤BA Induce Apoptosis in PC-3 Cells in Vitro-To define the cytotoxicity of A␤BA and AK␤BA as proapoptotic activity, we analyzed several apoptosis parameters. During apoptosis phosphatidylserine physiologically located at the inner membrane leaflet is flipped to the cell surface, where it serves as an "engulfing" signal for neighboring cells (25). Treatment of PC-3 cells with 10 M A␤BA or AK␤BA for 6 h triggered expression of phosphatidylserine on the outer membrane leaflet, as determined by fluorescein isothiocyanate-labeled annexin V binding ( Fig. 2A). At the concentration used, AK␤BA was more potent than A␤BA.
Apoptotic cells have reduced or hypoploid DNA contents (26) that we monitored as a sub-G 0 /G 1 peak of propidium iodidestained PC-3 cells previously exposed to 10 M A␤BA or AK␤BA for 48 h. Again, AK␤BA was more potent than A␤BA, but it was also more potent than the topoisomerase I poison camptothecin (10 M) that served as positive control (Fig. 2B).
Both pro-and antiapoptotic members of the Bcl-2 family of proteins are localized on the outer mitochondrial membrane, where they compete to regulate the release of cytochrome c. Functional predominance of proapoptotic molecules triggers cytosolic release of cytochrome c, which associates with Apaf-1 and procaspase-9 to form the apoptosome (5,33). Immunoblot analysis of cytochrome c in mitochondria isolated from PC-3 cells treated with ABAs showed a time-dependent loss of cytochrome c that appeared quicker and was more profound in cells pretreated with AK␤BA instead of A␤BA (Fig. 2C). As expected, the mitochondrial loss of cytochrome c was accompanied by a time-dependent increase of cytosolic cytochrome c levels (Fig. 2C). A definite sign of apoptosis is the occurrence of DNA fragmentation by endonuclease activation triggered by caspase 3 (3, 5, 33). Consistent with DNA fragmentation, both A␤BA and AK␤BA induced DNA laddering in PC-3 cells (Fig. 2D). Thus, A␤BA and AK␤BA induce apoptosis in androgen-independent PC-3 prostate cancer cells. In contrast to PC-3 cells, human normal lung fibroblasts (MRC-5 cells) did not undergo apoptosis upon treatment with 10 M AK␤BA, the more potent of both ABAs. At 24 h after AK␤BA treatment the MRC-5 fibroblasts showed neither annexin V binding nor an increased membrane permeability in terms of propidium iodide staining, whereas the AK␤BAtreated PC-3 cells were clearly positive for those signs of apoptosis (Fig. 2E). These data provide evidence for a relatively selective cytotoxicity of AK␤BA on PC-3 cells versus human normal fibroblasts.
A␤BA and AK␤BA Inhibit NF-B Signaling-The resistance of prostate cancer cells including PC-3 cells to apoptosis is believed to be due to constitutive activation of NF-B and subsequent expression of antiapoptotic proteins (3)(4)(5). Electrophoretic mobility shift assay with nuclear extracts from cells exposed to A␤BA or AK␤BA revealed that treatment with those compounds strongly inhibited binding of the extracts to the labeled NF-B binding sites (Fig. 3A). The specificity of NF-B binding was confirmed in competition experiments; a 50-fold molar excess of unlabeled NF-B, but not of AP-2 consensus oligonucleotides, profoundly inhibited binding of the nuclear extracts to the labeled NF-B binding site sequence (Fig. 3A).
Similarly, both A␤BA and AK␤BA concentrationdependently inhibited the TNF-␣-induced NF-B activation in HEK293 cells transiently transfected with a NF-B-driven luciferase gene (Fig. 3B). The viability of the HEK293 cells was not compromised within the course of the experiment. In order to demonstrate the specificity of the inhibitory effects observed, we also analyzed the effects of ABAs on the ISRE-mediated luciferase expression. ABAs did not affect the basal luciferase expression by pISRE-luciferase transfected HEK293 cells. Cotransfection of pISRE-luciferase with the constitutively active form of IRF-3, IRF-3 5D, resulted in a greater than 100 times increase in luciferase expression, which remained unaffected by pretreatment with 10 M concentrations of either A␤BA or AK␤BA (Fig. 3C). This indicates that the ABAs inhibit specifically NF-B activity.
To define the molecular target of A␤BA and AK␤BA in the NF-B pathway, we investigated the effects of the compounds on the binding of recombinant NF-B proteins to double- stranded NF-B consensus oligonucleotides immobilized on a SPR sensor chip (Fig. 3D). The addition of human recombinant p50/c-Rel or p50/p65 heterodimers resulted in an increase of the SPR signal, reflecting the binding kinetics of the proteins; this binding was unaffected by both A␤BA or AK␤BA. The recombinant NF-B proteins did not bind to the AP-1 consensus sequence used as negative control. Together these findings indicate that A␤BA and AK␤BA intercept NF-B signaling, yet they do not directly affect the NF-B binding to DNA.

A␤BA and AK␤BA Inhibit IKK and NF-B-dependent Antiapoptotic Gene Products in PC-3 Cells-To gain further insight
into the mechanism of action of A␤BA and AK␤BA, we next investigated their effects on IB degradation (Fig. 4A). Both compounds time-dependently inhibited IB degradation, which became detectable at 6 h after treatment with AK␤BA yet not earlier than 12 h after treatment with A␤BA.
Generally, p50/p65 dimers are considered to be the most abundant NF-B proteins, and phosphorylation of p65 is required for their full transactivation potential (4,34,35). Immunoblot analysis of p65 phosphorylated on Ser-536 in nuclear extracts from cells pretreated with A␤BA and AK␤BA showed a concentration-dependent inhibition of the p65 phosphorylation that was more pronounced with AK␤BA than with A␤BA (Fig. 4B).
The nuclear localization of p65 heterodimers is determined by the degradation of IB proteins and by post-translational modifications of p65 (34 -36). Therefore, we additionally deter-mined the localization of p65 in PC-3 cells by confocal microscopy (Fig. 4C). Control cells clearly show nuclear localization of p65, which was significantly inhibited, when the cells were treated with A␤BA and AK␤BA. This finding was confirmed by enzyme-linked immunosorbent assay of the Rel proteins contained in the nuclear extracts, demonstrating that A␤BA and, more potently, AK␤BA treatment reduced the levels of p65, c-Rel, and RelB in the nuclei of PC-3 cells (Fig. 4D).
Because of its essential role in the initiation of IB degradation, IKK activity is a crucial step for NF-B signaling (4,34). To clarify a potential effect of A␤BA and AK␤BA on IKKs, we performed kinase assays with IKK␣ and IKK␤ immunoprecipitated from PC-3 cells using recombinant IB␣ and p65 as substrates (Fig. 5A). The kinase assays demonstrate that nontreated PC-3 cells use mainly IKK␣ for the phosphorylation of both IB␣ and p65, whereas IKK␤ contributed only moderately to the phosphorylation of IB␣. The addition of A␤BA and AK␤BA reduced the IKK␣-mediated phosphorylation of both recombinant IB␣ and p65; again AK␤BA proved to be somewhat more effective than A␤BA. IKK␤ immunoprecipitated from PC-3 cells exhibited only very low activity in terms of IB␣ phosphorylation. Consistently, we were unable to detect any significant effect of ABAs on IKK␤ immunoprecipitated from PC-3 cells. The ABA-mediated inhibition of IKKs is apparently specific, because ABAs did not impair the extracellular signal-regulated kinase-2-mediated phosphorylation of Elk-1 (Fig. 5A). The inhibitory effects of ABAs on IKK were further confirmed in in vitro kinase assays using active human recombinant GST-IKK␣ and His-IKK␤. A␤BA and, more potently, AK␤BA inhibited the activity of mainly IKK␣ and to a lesser extent that of IKK␤. As little as 0.1 M A␤BA and AK␤BA affected the phosphorylation of IB␣ by either kinase (Fig. 5B).
NF-B is involved in the expression of antiapoptotic genes such as bcl-2 and bcl-x but also in the expression of the cell cycle regulating cyclin d1 (37)(38)(39). Both A␤BA and AK␤BA induced a concentration-dependent inhibition of the expression of Bcl-2, Bcl-x L , and cyclin D1 at the mRNA level, suggesting transcriptional regulation (Fig. 5C). Consistent with these findings, A␤BA and AK␤BA time-dependently reduced the expression of Bcl-2, Bcl-x L , and cyclin D1 proteins (Fig. 5D). Together these data demonstrate that A␤BA and AK␤BA inhibit NF-B signaling by intercepting IKK activity, which finally leads to a diminished expression of the antiapoptotic proteins Bcl-2 and Bcl-x L as well as to inhibition of cyclin D1, a key activator of cell cycle progression.
Suppression of IKK Expression Reduces Cell Proliferation-To prove that IKK is indeed essential for the PC-3 cancer cell proliferation, we used an in vitro knockdown approach employing antisense ODN. The IKK␣ sense ODN did not significantly affect the cell proliferation, whereas the antisense ODN down-regulated the IKK␣ expression by ϳ55% (Fig. 5E) and significantly reduced the cell proliferation by 45.0 Ϯ 6.0% (n ϭ 3, p Ͻ 0.01) (Fig. 5F). Downregulation of the IKK␤ expression also reduced cell growth; however, to a much lesser extend. Thus, inhibition of IKK is a potential target for antitumor intervention.

A␤BA and AK␤BA Cyclodextrin Complexes Inhibit Growth and Proliferation of PC-3 Xenografts on the Chick Chorioallantoic Membrane and Trigger
Apoptosis-To verify the proapoptotic and antitumor activity of A␤BA and AK␤BA in vivo, we next xenotransplanted PC-3 cells on the CAM of fertilized chicken eggs (19). Immunohistochemical analysis of the tumor sections for human cytokeratin and the proliferation antigen Ki-67 revealed that AK␤BA-CD had a concentration-dependent inhibitory effect on both tumor size and proliferation (Fig. 6A). Using the TUNEL technique for in situ detection of apoptosis, we confirmed induction of apoptosis of PC-3 tumor cells, especially for AK␤BA-CD (Fig. 6A).
Histomorphometric quantification of the Ki-67 antigen expression in the tumor sections (Fig. 6B) confirmed that AK␤BA-CD concentration-dependently and significantly inhibited the proliferation of PC-3 cells, whereas A␤BA-CD was at least 30 times less potent.
The topical application of 100 M AK␤BA-CD yielded only low systemic plasma levels in the chicken embryo, reaching a maximum of ϳ5 M (Fig. 6C); no gross signs of toxicity were detectable during autopsy. These data show that AK␤BA and to a much lesser extent A␤BA exhibit antitumor activity against androgen-independent PC-3 prostate cancer cells xenotransplanted onto the CAM.

AK␤BA-CD Inhibits Growth and Invasiveness of PC-3 Xenografts in Nude Mice and Triggers
Apoptosis-To expand the previous findings to a mammalian system, we xenotransplanted PC-3 cells subcutaneously into nude mice. Treatment by daily intraperitoneal injections with 100 mol/kg of AK␤BA-CD started 1 week after inoculation and continued for another 3 weeks (Fig. 7A). Compared with the CD control group, this treatment significantly reduced the tumor volume in the AK␤BA-CD group by 49.5 Ϯ 11.3% (n ϭ 6, p Ͻ 0.001) and by 52.2 Ϯ 8.5% (n ϭ 6, p Ͻ 0.001) after 2 and 3 weeks, respectively. Furthermore, compared with the CD control group, AK␤BA-CD treatment also clearly reduced the invasiveness of the tumors into the surrounding tissues (Fig. 7B). Consistent with the data from the chorioallantoic membrane model, AK␤BA-CD triggered apoptosis in PC-3 xenotrans-plants in nude mice, as detected by TUNEL staining (Fig. 7C,  lower panel), whereas tumors from CD-treated animals showed only a very few apoptotic cells (Fig. 7C, upper panel).
We also analyzed the plasma kinetics of AK␤BA-CD in mice by reversed phase high performance liquid chromatography (12,18). Twenty minutes after intraperitoneal injection of 100 mol/kg of AK␤BA-CD the plasma levels reached ϳ100 M (Fig. 7D). An initial phase of rapid clearance (t1 ⁄2 ϳ 26 min) was followed by a slower, longer-lasting elimination phase (t1 ⁄2 ϳ 68 Pathological and histological examination of liver, kidney, blood, and bone marrow from AK␤BA-CD-treated animals did not show any overt signs of toxicity (data not shown). Together, these data demonstrate that AK␤BA-CD exerts also antitumor activity against androgen-independent prostate cancer xenotransplanted in nude mice. DISCUSSION With the progression of prostate cancer to androgen independence, the tumor becomes resistant to conventional therapeutic measures including chemotherapy (2). This situation is mirrored in androgen-independent prostate cancer cell lines, such as PC-3 and DU145, which are resistant to the induction of apoptosis by highly potent agents such as staurosporine and camptothecin (6,40,41).
Our data indicate that A␤BA and AK␤BA trigger apoptosis in the chemoresistant and androgen-independent human PC-3 prostate cancer cells both in vitro and in vivo. Interestingly, the more potent of both compounds, AK␤BA, in a concentration able to trigger apoptosis in PC-3 cells did not do so in human normal lung fibroblasts (MRC-5), suggesting specificity for the tumor cells. At the molecular level, we identified down-regulation of the expression of the cell cycle regulator cyclin D1 and of antiapoptotic proteins from the Bcl-2 family by virtue of the inhibitory activity of ABAs on IKK and thereby on the NF-Bdependent gene expression. The reduced viability observed after ABA treatment was indeed due to inhibition of IKK as confirmed by IKK␣and IKK␤-specific antisense ODN. These findings foster hope for a novel unconventional approach to treatment-resistant prostate cancers.
Several plant-derived compounds have been shown to interfere with the NF-B pathway (42)(43)(44)(45)(46). Despite the fact that ursolic and betulinic acids are structurally similar to the ABAs, they do not directly interfere with the IKK activity (44,45). Similarly, the synthetic flavone flavopiridol exerts its activity probably via inhibition of Akt and has no direct effect on IKK (46). Curcumin has been shown to inhibit the activity of immunoprecipitated IKKs; however, these data were not confirmed with purified enzymes nor were effects on other kinases excluded (42).
The transcriptional activator NF-B has potent antiapoptotic functions implicated in oncogenesis and resistance to various therapeutic agents (3)(4)(5)(6). Constitutive activation of NF-B is now held responsible for both the chemoresistance and the highly malignant phenotype of prostate cancer (34,47,48). In contrast to androgen-dependent LNCaP, the androgen-independent prostate cancer cell lines PC-3 and DU145 exhibit a high constitutive activation of IKK and subsequent NF-B signaling (47). The NF-B-mediated transcriptional up-regulation of the expression of the antiapoptotic proteins Bcl-2 and Bcl-x L has been suggested to play a key role for both the resistance to antitumor drugs and the malignant phenotype of androgen-independent versus androgen-dependent prostate cancer (47)(48)(49).
Mature NF-B dimers are trapped in the cytoplasm of unstimulated cells by interaction with the inhibitory proteins termed IBs. It is the multisubunit IKK complex consisting of two catalytic subunits, IKK␣ and IKK␤, and a regulatory subunit IKK␥ that phosphorylates IB proteins on two distinct serine residues, thereby targeting them to rapid ubiquitin-dependent proteolysis that initiates the activation of NF-B (4,34). Although the detailed oligomeric composition and the regulation of the whole IKK complex are still not fully understood, its antiapoptotic role identifies NF-B signaling as an attractive target for antitumor therapy (4 -6, 34).
Our data not only demonstrate that A␤BA and AK␤BA inhibit the IKK complex but also that in PC-3 cells IKK␣ is constitutively activated. The predominant constitutive activation of IKK␣ over IKK␤ is unusual and may be a particular feature of the tumor cells (47,48). The inhibitory effect on IKK activity by ABAs is reflected by the inhibition of phosphorylation of IB␣ and of p65 on Ser-536, resulting in diminished nuclear translocation of p65 and, consequently, reduced expression of NF-B-dependent genes such as bcl-2, bcl-x, and cyclin d1. Overexpression as well as knockdown experiments indicate that p65 phosphorylation on Ser-536 can be catalyzed by both IKK␣ and IKK␤ (36,50). Although phosphorylation of p65 in general and on Ser-536, located within the TA1 transactivation domain, has been implicated in enhanced NF-B transcriptional activity (35), the precise physiological role of the Ser-536 phosphorylation remains to be elucidated (36).
The inhibition of the constitutive IKK activity in PC-3 cells by A␤BA and AK␤BA, as reflected by the predominantly cytosolic localization of NK-B proteins, leads to transcriptional down-regulation of NF-B-dependent genes encoding antiapoptotic proteins of the Bcl-2 family, such as Bcl-2 and Bcl-x L (33,49). Both proteins are prototypic cell death regulators whose function is modulated by complex homo-and heterodimerizations with their proapoptotic homologues such as Bax, thereby modulating the apoptosis-inducing release of cytochrome c from mitochondria (5,33). This aspect may be crucial for the antitumor activity because in tumor cells, Bcl-2 delays or prevents apoptosis to virtually all of the chemotherapeutic agents currently in use (33,51,52). Deregulated overexpression of Bcl-2 has been described in numerous malignant tissues including prostate cancer (5,33). In normal prostate tissue, Bcl-2 expression is restricted to the basal cells of the glandular epithelium, and these cells exhibit resistance to the effects of androgen deprivation (49). In the androgen-dependent LNCaP cells the resistance to apoptosis is increased as a result of Bcl-2 transfection (53). Conversely, antisense Bcl-2 oligonucleotide constructs or down-regulation of Bcl-2 by forced expression of PTEN can be utilized to sensitize otherwise apoptosis-resistant prostate cancer cells to chemotherapeutic agents (51,54,55). For these reasons Bcl-2 has been considered as a major new strategic target for the development of gene knockdown therapies (33, 49, 52, 56). However, with such an approach one still encounters the unsolved difficulties of delivering DNA into solid tumors.
Antiapoptotic Bcl-x L is produced from the long splice variant of the bcl-x gene, whereas proapoptotic Bcl-x S is derived from the short alternative splice form of this gene (33,52,56). Bcl-x L appears to provide at least equivalent or even greater protection against a large number of cytotoxic agents in numerous tumor cell lines, including PC-3 and DU145, than its close homologue Bcl-2 (57). Similar to Bcl-2, forced overexpression of Bcl-x L markedly desensitizes prostate cancer cells to the cytotoxic effects of a variety of chemotherapeutic agents (58). In contrast to benign prostate tissue, androgen-independent prostate cancer cells, such as PC-3 and DU145, are known to express high amounts of Bcl-x L (59). Treatment of those cells with antisense oligonucleotides that shift the splicing pattern of Bcl-x pre-mRNA from the antiapoptotic variant, Bcl-x L , to the proapoptotic variant, Bcl-x S , promotes apoptosis that correlated with the amount of Bcl-x L originally expressed by these cells. In addition, those PC-3 cells exhibited a 10-fold increased sensitivity to cisplatin and 5-fluoro-2Ј-deoxyuridine (59). However, even though Bcl-x L might be an attractive therapeutic target (52,56,57), a therapeutic approach with antisense oligonucleotides might not be feasible because DU145 and LNCaP stably transfected with an antisense bcl-x L expression vector reacted with an increased expression of highly potent NF-Bdependent antiapoptotic proteins, such as XIAP and cIAP-1, rendering those cells resistant to chemotherapy (60).
Cyclin D1 is rate-limiting for the progression of the cell cycle from the G 1 phase to the S phase during which DNA replication occurs (61). IKK␣ is essential for cyclin D1 expression (62) and regulates mitogenic signaling through transcriptional induction of cyclin D1 (63). In line with these findings, the inhibition of IKK activity by ABAs induced transcriptional down-regulation of cyclin D1 in PC-3 cells, which is also reflected by the down-regulation of the Ki-67 proliferation antigen expression in the CAM model. As can be expected, cyclin D1 overexpression in LNCaP cells increased cell growth as well as tumorigenicity when xenotransplanted into mice (64). In line with these data, AK␤BA, which through inhibition of IKK inhibits cyclin D1 but also antiapoptotic Bcl protein expression, was found to reduce tumor growth and invasiveness and to induce apoptosis in our experiments with PC-3-bearing nude mice. These data further demonstrated in vivo efficacy of AK␤BA in a mammalian system despite its rather short half-life. Interestingly enough, these effects occurred in the absence of any major systemic toxicity.
Whether NF-B acts only as a "major roadblock on the path to apoptosis" (4) requiring additional inducers of apoptosis or whether its inhibition is sufficient to trigger apoptosis is currently a subject of intense investigations. Although controversial, it is often assumed that for apoptosis of solid tumors another pro-apoptotic stimulus besides inhibition of NF-B might be required. However, similar to colorectal cancer (4), inhibition of NF-B alone may trigger apoptosis in hormonerefractory prostate cancer (48,65). Also, Palayoor et al. (47,66) reported that inhibition of IKK␣ activation by ibuprofen at a concentration believed to cause a significant reduction of NF-B activity may lead to apoptosis of PC-3 cells. In addition, down-regulation of cyclin D1 in melanoma cells by antisense treatment led to apoptosis in vitro and to tumor shrinkage xenotransplants in nude mice (67). Together, these results suggest that down-regulation of Bcl-2, Bcl-x L , and cyclin D1 could possibly suffice to trigger apoptosis in PC-3 cells.
Of course one might argue that A␤BA and AK␤BA could modulate other processes relevant for cell survival and chemoresistance. Thus, we have previously observed that ABAs might catalytically inhibit topoisomerase I and II␣ in vitro (9,17). However, PC-3 cells are largely resistant against the highly potent topoisomerase poison camptothecin (41). A strong argument for the essential role of IKK for PC-3 tumor growth provides the antisense ODN approach, demonstrating that intercepting IKK␣ and IKK␤ expression indeed reduces cell proliferation. Finally, the inhibitory effects of ABAs on IKK should also contribute to the antiinflammatory efficacy ascribed to those compounds, because IKK is critical for the NF-B-mediated inflammatory response program (68).
In conclusion our data demonstrate that inhibition of IKK activity in androgen-independent prostate cancer by ABAs triggers down-regulation of several target genes crucial for cell proliferation, chemoresistance, and tumor survival. It is appealing that AK␤BA, which is readily available and not afflicted with the inherent problems of delivering oligonucleotides to solid tumors, targets several molecular targets considered important for the development of new target-specific anticancer drugs (4,6,33,41,49,52,56,61). Therefore, ABAs intercepting IKK activity may provide a rewarding novel antitumor approach for so-far incurable androgen-insensitive prostate cancers.