Prostate Cancer Cell CycleRegulators: Response toAndrogen Withdrawal andDevelopment of AndrogenIndependence

David B. Agus, CarlosCordon-Cardo, William Fox, MarijaDrobnjak, Andrew Koff, David W.Golde, Howard I. Scher

Background:Androgen withdrawal is astandard therapy for prostate cancerthat results in a decrease in tumor vol-ume and a decline in serum prostate-specific antigen in the majority of pa-tients. To understand the factorsassociated with regression of prostatecancers after androgen withdrawal,we studied cell cycle regulator changesin the CWR22 human prostate cancerxenograft model.Methods:Establishedtumors in nude athymic BALB/c micewere sampled at various times afterandrogen withdrawal and after the de-velopment of androgen independence.Changes in the expression of cell cycleregulators were categorized into earlyand mid-to-late events. Results andConclusions: Early events included adecrease in androgen receptor expres-sion, followed by a short-term increasein expression of the p53 and p21/WAF1proteins and a marked decrease in theKi67 proliferative index. Mid-to-lateevents included progressive and sus-tained increases in p27 and p16 pro-tein expression, a decrease in retino-blastoma protein expression, and anincrease in the transcription factorE2F1. Changes in apoptosis (pro-grammed cell death) were not observedat any time after androgen withdrawal.These data suggest that androgen with-drawal results in a cell stress response,in which increased p53 protein pro-duces a cell cycle arrest, without acti-vation of p53-mediated apoptosis. Theproliferative index is further decreasedthrough the action of the cyclin-dependent kinase inhibitors p27 andp16. Androgen-independent sublinesemerged 80–400 days after androgenwithdrawal, and these sublines hadvariable growth phenotypes but wereassociated with mdm2 protein overex-pression and increased expression of

cyclin D1. These results indicate thattumor regression in this human pros-tate cancer model is due to cell cyclearrest rather than to apoptosis and thatthe emergence of androgen indepen-dence is associated with a release fromcell cycle arrest. [J Natl Cancer Inst1999;91:1869–76]

Prostate cancers have a low prolifera-tive index and depend on androgens forgrowth. After castration, these tumors re-gress but later progress to an androgen-independent phenotype that is ultimatelylethal for the patient. Knowledge of themechanisms associated with the inabilityof androgen withdrawal strategies to re-sult in the complete regression of a pros-tate cancer and the mechanisms that con-tribute to androgen independence wouldhave valuable therapeutic implications.Previously, we showed that androgenwithdrawal in mice bearing the CWR22human prostate cancer xenograft(1–3)was associated with a decrease in the pro-liferative index(4). Cellular changes con-sistent with apoptosis (i.e., programmedcell death) were rare at the times evalu-ated. Thus, this study was undertaken todetermine dynamic changes in the expres-sion of critical cell cycle regulators, inthe p53 pathway, and in the retinoblasto-ma protein pathway, associated with theresponse to androgen withdrawal andwith the emergence of the androgen-independent phenotype.

Reproducible changes in expression ofcell cycle regulators were found thatcould be categorized into early andmid-to-late events. Early events wereconsistent with a cell stress responsein which p53 produces a transient cellcycle arrest without activation of p53-dependent cell death programs. Mid-to-late events included an increase in theexpression of the cyclin-dependent ki-nase inhibitors p27 and p16, which areassociated with a further decline and themaintenance of a low proliferative index.Apoptotic changes were not observed af-ter androgen withdrawal. Characteriza-tion of the androgen-independent pheno-type revealed the overexpression ofmdm2, affecting p53 stability, as well asthe increased cyclin D1 expression, af-fecting pRB phosphorylation. These re-sults challenge a commonly held viewthat the regression of prostate cancersafter androgen withdrawal is mediatedexclusively by apoptotic mechanisms.Furthermore, the results suggest that

therapeutic strategies directed at the cellcycle-arrested prostate cancer cells, afterandrogen withdrawal, may be clinicallyimportant.

MATERIALS AND METHODS

Animal StudiesFour- to 6-week-old nude athymic BALB/c male

mice were obtained from the National Cancer Insti-tute-Frederick Cancer Research and DevelopmentCenter, Frederick, MD, and maintained in pressur-ized, ventilated caging. Institutional guidelines forthe proper and humane use of animals in researchwere followed. The CWR22 tumor line was propa-gated in the animals by the injection of a mixturecontaining reconstituted basement membrane (Ma-trigel; Collaborative Research, Inc., Bedford, MA)and minced tumor tissue from an established tumorinto the subcutaneous tissue of the flanks of athymicnude mice(1,2). For the maintenance of serum an-drogen levels, mice were administered 12.5 mg ofsustained-release testosterone pellets (InnovativeResearch of America, Sarasota, FL) subcutaneously.Tumors of approximately 1.5 × 1.0 × 1.0 cm grew3–4 weeks after inoculation. Androgen withdrawalwas accomplished by surgical castration under pen-tobarbital anesthesia and removal of the testosteronepellets. Tumor size was determined by caliper mea-surements of height, width, and depth. Prostate-specific antigen (PSA) assays were performed onthe serum of the mice obtained by tail bleeding andused a Tandem-R PSA immunoradiometric assay(Hybritech, Inc., San Diego, CA). Androgen-independent sublines of the parent CWR22 were ob-tained by following tumors for regrowth and in-creases in serum PSA after androgen withdrawal.The androgen-independent sublines regrew 80–400days after androgen withdrawal. The sublines wereserially passaged in three castrated hosts beforecharacterization as a subline. These were duplicatesperformed at each sampling point in the experi-ments.

Histopathology, MonoclonalAntibodies, andImmunohistochemistry

Tissues were fixed in 10% buffered formalin andembedded in paraffin, and sections (5mm) werestained with hematoxylin–eosin. The followingwell-characterized antibodies were used at the cor-responding final working dilutions: anti-Ki67 mousemonoclonal antibody (MAb) MIB1 (ImmunotechSA, Marseille, France; 1 : 50 dilution), p53 (MAbclone PAb1801; Calbiochem/Oncogene Sciences,Cambridge, MA; 0.2mg/mL), p21/WAF1 (MAbclone 2G12; PharMingen, San Diego, CA; 0.5mg/

Affiliations of authors:D. B. Agus, W. Fox, D.W. Golde, H. I. Scher (Department of Medicine), C.Cordon-Cardo, M. Drobnjak (Department of Pathol-ogy), A. Koff (Sloan-Kettering Institute), MemorialSloan-Kettering Cancer Center, New York, NY.

Correspondence to:David B. Agus, M.D., Me-morial Sloan-Kettering Cancer Center, 1275 YorkAve., New York, NY 10021 (e-mail: d-agus@ski.mskcc.org).

See“Notes” following “References.”

© Oxford University Press

Journal of the National Cancer Institute, Vol. 91, No. 21, November 3, 1999 REPORTS 1869

mL), mdm2 (MAb clone 2A10; gift from Dr. A.Levine, Rockefeller University, New York, NY;1 : 50 dilution), p27/Kip1 (MAb clone DCS72; Cal-biochem/Oncogene Sciences; 0.1mg/mL), p16(MAb clone DCS-50.1/H4; Calbiochem/OncogeneSciences; 2mg/mL), cyclin D1 (MAb clone DCS-6;Calbiochem/Oncogene Sciences; 1mg/mL), E2F1(purified rabbit antiserum, C-20; Santa Cruz Labo-ratories, Santa Cruz, CA; 0.2mg/mL), androgen re-ceptor (MAb clone F39.4.1; BioGenex, San Ramon,CA; 2 mg/mL), bcl-2 (MAb clone 124; Dako Corp.,Carpinteria, CA; 2.5mg/mL), and bax (purified rab-bit antiserum, 1-19; Santa Cruz Laboratories; 0.05mg/mL). An MAb (immunoglobulin G [IgG] sub-class, clone MIgS-KpI; PharMingen; 1 : 50 dilution)and a preimmune rabbit serum were used as negativecontrols. Histologic sections were immersed in boil-ing 0.01% citric acid (pH 6.0) for 15 minutes, al-lowed to cool, and incubated with primary antibod-ies overnight at 4 °C. Biotinylated horse anti-mouseIgG antibodies and biotinylated goat anti-rabbit an-tiserum (both from Vector Laboratories, Inc., Bur-lingame, CA; 1 : 500 dilution) were applied for 1hour, followed by avidin–biotin peroxidase com-plexes for 30 minutes (Vector Laboratories, Inc.;1 : 25 dilution). Diaminobenzidine was used as thefinal chromogen, and hematoxylin was used as thenuclear counterstain. Most markers were evaluatedfor nuclear staining of tumor cells. Exceptions in-cluded the evaluation of bcl-2, which renders apunctated mitochondrial pattern, and bax, whichproduces a diffuse cytoplasmic staining. Positivecells were scored by counting different fields andmore than 500 cells. Data were recorded in a con-tinuum as the percentage of cells stained at the spe-cific time evaluated, and the intensity of stainingwas graded as follows: 0, undetectable; 1+, minimalstaining; and 2+, strong staining. Tumors from atleast two animals were evaluated at each time. Cellcycle regulators were assessed daily after androgenwithdrawal to day 10 and then every 5 days untilday 25.

Flow cytometry assays for assessment of S-phasefraction and cell cycle distribution analyses weredone on tumor blocks from specific times after an-drogen withdrawal. Tumor tissue was microdis-sected to avoid normal tissue contamination and ne-crotic material. Three consecutive 60-mm-thicksections were cut, and the nuclei were disaggre-gated, stained with propidium iodide, and analyzedon a Coulter Excel (Beckman Coulter, Inc., Fuller-ton, CA) with a multicycle soft program for assess-ment of cell cycle profile(5).

To study the microanatomical distribution of ap-optosis, we assayed consecutive sections from theblocks used for immunohistochemistry by a modifi-cation of the method of terminal deoxynucleotidyl-transferase-mediated uridine triphosphate end label-ing (TUNEL) (6), originally described by Gavrieli etal. (7). Briefly, after the exposure of nuclear DNA ofhistologic sections by proteolytic treatment, terminaldeoxynucleotidyltransferase was used to incorporatebiotinylated deoxyuridine at sites of DNA breaks.The signal was amplified by avidin–biotin peroxi-dase complexes and visualized by diaminobenzidinedeposition, enabling conventional histochemicalidentification of reactive nuclei by light microscopy.Nuclear staining was assessed by immunohisto-chemical scoring, counting different fields, andevaluating more than 500 tumor cells(8).

p27 Degradation and TranscriptQuantitative Assays

Extracts were prepared from CWR22 tumors attimes before and after androgen withdrawal. Thetumor extracts were assayed for p27 degradation ac-tivity in vitro as described by Nguyen et al.(9).Digoxigenin-labeled probes were used forin situhybridization, and 1mg of recombinant plasmidpCR™II (Invitrogen Corp., San Diego, CA), con-taining the full-length human p27KIP1 gene, was lin-earized by digestion withBamHI and XbaI to gen-erate antisense and sense transcripts. RNA probeswere generated, and deparaffinized tissue sectionswere stained as previously described(10). Each as-say was done in duplicate.

pRB Phosphorylation Assays

Western blotting was done with protein extractsfrom tumors. Proteins were extracted from OCT (op-timal cutting temperature) compound-embedded tu-mors and resolved on polyacrylamide gels for im-munoblotting with pRB-specific antibodies as

previously described(10). Each assay was done induplicate.

RESULTS

Dynamic Cell Cycle Changes AfterAndrogen Withdrawal

Early events. Androgen withdrawalwas associated with an abrupt decrease inandrogen receptors, detected by nuclearstaining, from 60% (1+) of tumor cellspositive on day 0 (baseline) to 10% bydays 3, 7, and 10 (Fig. 1, A and B). Theproportion of tumor cells with nuclearstaining for p53 increased from a fewscattered nuclei in less than 5% of cells onday 0 to approximately 30% (1–2+) onday 3 (Fig. 1, C and D; maximal staining).The p53 staining decreased to baseline byday 7. After the increase in p53-positivecells, the number of cells expressing p21/

Fig. 1. Sections of tumors from a nude mouse with CWR22 human prostate cancer xenograft immediatelybefore and after androgen withdrawal.Panels AandB: Representative areas of CWR22 tumor sections fromdays 0 and 10, respectively, after androgen withdrawal. Sections are stained with an anti-androgen receptormonoclonal antibody (MAb).Panels CandD: Representative areas of CWR22 tumor sections from days 0and 3, respectively, after androgen withdrawal. Sections are stained with an anti-p53 MAb.Panels EandF:Representative areas of CWR22 tumor sections from days 0 and 5, respectively, after androgen withdrawal.Sections are stained with the anti-p21/WAF1 MAb.Panels GandH: Representative areas of CWR22 tumorsections from days 0 and 10, respectively, after androgen withdrawal. Sections are stained with the anti-Ki67MAb. Bars 4 100 mm.

1870 REPORTS Journal of the National Cancer Institute, Vol. 91, No. 21, November 3, 1999

WAF1 increased from approximately25% (day 0) to a maximum of approxi-mately 50% of the cells (1–2+) by day5 (Fig. 1, E and F) and thereafter gradu-ally decreased. Concurrent with the in-crease in p21/WAF1, a substantial in-crease in p16INK4A was noted. At days0–2 after androgen withdrawal, 5%–10%of cells were p16 positive (1–2+); thispercentage increased to 30% on day 3 andto 40% on days 5–7 (1–2+). The net effectof these early changes was a decrease inthe Ki67 proliferative index, whichranged from 60% to 65% on days 0–3 to20% on day 5 to approximately 5% onday 7 (Fig. 1, G and H). Levels of cyclin

D1 expression also fell from 10% of tu-mor cells (pretreatment and day 0) to ap-proximately 2% on day 7 after androgenwithdrawal.

Expression of mdm2 and bax, otherp53-transactivated proteins, did notchange during this early period. In fact,expression of mdm2 and bax ranged from1% to 5% before and after androgen with-drawal. In addition, we did not detectbcl-2 in prostate cancer cells before orafter androgen withdrawal (data notshown). The apoptotic rate, as assayed byTUNEL, ranged from 0.2% to 0.8% dur-ing the first 7 days (data not shown). Flowcytometry analysis of the CWR22 tumor

cells stained with propidium iodideshowed a decrease in the percentage ofcells in the S-phase fraction from 22%before androgen withdrawal to 5.8% 5days after androgen withdrawal and lessthan 3% by 7–10 days (data not shown).

Thus, early cell cycle events after an-drogen withdrawal are characterized byexit from the cell cycle (summarized inFig. 2, A), which is associated with a cel-lular stress response that is initiated by anincrease in p53, followed by an increasein p21/WAF1 and a resultant early G1/G0-phase arrest, with the lack of an apoptoticresponse. Concurrent with these changesafter androgen withdrawal, the level of

Fig. 2. Panel A: Cell cycle regulators,Ki67, and androgen receptor (AR) ex-pression after androgen withdrawal—dynamic early and mid-late events. Du-plicate tumors were evaluated at eachtime point. The range was consistentlybelow ±10% of the listed value.PanelB: Serum prostate-specific antigen(PSA) (dashed line) and tumor volume(solid line) changes after androgen with-drawal. Values from a representative xe-nografted animal with the CWR22 tu-mor are shown. Androgen independenceis manifested initially by a rise in serumPSA, followed by an increase in tumorvolume.

Journal of the National Cancer Institute, Vol. 91, No. 21, November 3, 1999 REPORTS 1871

PSA fell because PSA transcription isdriven by androgen (Fig. 2, B)(4).

Mid-to-late events.Early events werefollowed by a change in the expression ofpRB. Before androgen withdrawal, the in-tensity of pRB staining was strong (2+),and pRB staining was detected in morethan 80% of the cells (Fig. 3, A). Onabout day 5 after androgen withdrawal,pRB staining was detected in approxi-mately 50% of cells and the intensity hadalso decreased (1–2+). By day 10, weak-to-moderate immunoreactivity (1+) wasdetected in 40%–50% of cells (Fig. 3, B).Western blotting of protein extracts ob-tained revealed that the intense homoge-neous pRB staining was associated withthe hyperphosphorylated pRB products(data not shown). In parallel with the de-crease in pRB expression, there was aslow and progressive increase in E2F1 ex-pression. E2F1-positive staining rangedfrom weak (1+) in 20% of the tumor cells

on day 0 (Fig. 3, C) to moderate (1–2+)in 50% of the tumor cells on day 10 (Fig.3, D) to strong (2+) in more than 70% ofthe tumor cells by day 20. Concomitantwith the rise in E2F1 staining, we ob-served a distinct and sustained increase inp27 levels. Expression of p27 was low toundetectable (0–1+) in the tumor cellsbefore androgen withdrawal (Fig. 3, E)but was detected in 5%–7% (1+) of thecells in the 3 days after castration. p27was detected on day 5 in 40%–50% ofthe cells, on days 7–10 in 60%–70% (Fig.3, F), and on days 15, 20, and 25 in 80%–90% (2+). The level of p27 transcript, asassayed byin situ hybridization, did notchange by the withdrawal of androgen(data not shown). Degradation of p27 inthe tumor was assessed and found to beandrogen dependent. In the presence ofandrogen, there was substantial p27 deg-radation, but this activity declined afterandrogen withdrawal and paralleled the

percentage of cells expressing p27 (Fig. 3,G). Because p27 degradation is a possiblemarker for S-phase cells, the amount ofp27 degradation should be associatedwith the proliferation of cells(9). Expres-sion of p16 continued to increase fromapproximately 40% of tumor cells at day7 to more than 60% at day 10. This ex-pression was sustained to day 30 (Fig. 2,A). The mid-to-late changes in cell cycleregulators after androgen withdrawalwere associated with a low Ki67 prolif-erative index. On day 7, the Ki67 indexwas approximately 5%, decreased to0.8% and 0.2% on days 10 and 15, re-spectively, and was undetectable at days20 and 25. Cyclin D1 was undetect-able from days 7 through 25 but waspresent in the androgen-independent xe-nografts.

Between week 1 and week 3 after an-drogen withdrawal, mdm2, bax, and bcl-2levels did not change relative to those be-fore androgen withdrawal. Also, the rateof apoptosis, determined by the TUNELassay, was low, ranging from 0.7% to1.3% between day 7 and day 25 (data notshown). Flow cytometry assays of propi-dium iodide-stained cells revealed a de-crease in the S-phase fraction from 4% onday 7 to less than 1% 15 days after an-drogen withdrawal (data not shown).

After androgen withdrawal, the mid-to-late cell cycle events are, thus, charac-terized by a progressive and sustained de-crease in cyclin D1 expression, followedby changes in pRB expression (summa-rized in Fig. 2, A). Growth arrest of theprostate cancer cells appears to be main-tained by steady levels of p27 and p16.The prostate cancers did not appear to un-dergo apoptosis after androgen with-drawal but rather underwent a G0/earlyG1-phase cell cycle arrest.

Cell Cycle Protein ChangesAssociated With AndrogenIndependence

Similar to previous observations(2),androgen-independent proliferation wasobserved after 80–400 days, with a serumPSA rise preceding tumor volumechanges (Fig. 2, B). The pattern of expres-sion of cell cycle regulatory proteins inthe five androgen-independent tumors,des ignated CWR22R, CWRSA1,CWRSA3, CWRSA4, and CWRSA6, re-flected a highly proliferative tumor, simi-lar to the CWR22 xenografts. Andro-gen-independent tumors had a Ki67proliferative index of 60%–70%, a homo-

Fig. 3. Sections of tumors from a nude mouse with CWR22 human prostate cancer xenograft immediatelybefore and after androgen withdrawal.Panels AandB: Representative areas of CWR22 tumor sections fromdays 0 and 10, respectively, after androgen withdrawal. Sections are stained with an anti-retinoblastomaprotein monoclonal antibody (MAb).Panels CandD: Representative areas of CWR22 tumor sections fromdays 0 and 10, respectively, after androgen withdrawal. Sections are stained with an E2F1 MAb.Panels Eand F: Representative areas of CWR22 tumor sections from days 0 and 10, respectively, after androgenwithdrawal. Sections are stained with an anti-p27 MAb.Bars 4 100 mm. Panel G: p27 degradation isandrogen dependent. The amount of substrate p27 for the reaction is shown in the lane labeled “Input.”Numbers above each lanerepresent the percent of the input p27 remaining after 2 hours.

1872 REPORTS Journal of the National Cancer Institute, Vol. 91, No. 21, November 3, 1999

geneous and intense staining for androgenreceptor in 60%–80% of CWR22R,CWRSA1, and CWRSA4 cells or 90% ofCWRSA3 and CWRSA6 cells, and, com-pared with the growth-arrested tumors, anincrease in the proportion of pRB-positivecells with strong nuclear staining and adecrease in E2F1 (data summarized inFig. 4, A). There was a decrease in theproportion of cells staining for p27, from80%, when proliferation was low (day25), to approximately 5% in the andro-gen-independent tumors. Consistent withthese findings, there was a decrease in theproportion of cells staining for p16 from60% to less than 3%.

Two important differences between theandrogen-independent sublines and the

parental CWR22 xenograft includedcyclin D1 and mdm2 overexpression. Cy-clin D1 nuclear reactivity was detectedin 20%–25% and mdm2 nuclear over-expression in approximately 30% of theandrogen-independent cells (Fig. 4, B).We did not observe notable changes inp53 (undetectable to 1% positive), bax(<5% positive), bcl-2 (undetectable), orapoptotic rate, as assessed by TUNEL as-say, in the androgen-independent tumorspecimens consistent with the parentalCWR22.

The level of androgen receptor expres-sion was 1+−2+ in 60%–80% of the tumorcells of most androgen-independent sub-lines. Two sublines, CWRSA3 andCWRSA6, expressed higher levels of an-

drogen receptor in a greater percentage oftumor cells (ù90%, 2+ staining). Thegrowth characteristics of these sublineswere similar to those of the other sublines,except that their growth was repressed,rather than stimulated, in the presence ofphysiologic concentrations of androgen.The emergence of androgen indepen-dence in this model system is associatedwith a return of tumor cells to an activecell cycle and increased expression of theoncogenic proteins cyclin D1 and mdm2.In addition, the androgen-repressiblegrowth phenotype is associated with rela-tive androgen receptor overexpression.

DISCUSSION

The current dogma is that androgenwithdrawal in prostate cancer will resultin the apoptotic death of a majority ofcancer cells and that the few remainingcells are resistant and will grow back withan androgen-independent phenotype.Characterization of the androgen-independent phenotype has been limitedby the lack of biopsy material from thesepatients. The CWR22 xenograft model re-capitulates a subset of human prostatecancers in that, after androgen with-drawal, there is a decline in PSA levels.This decline is followed by regression ofthe tumor, a rise in PSA levels, and thenregrowth as an androgen-independentneoplasm(2). In this study, we catego-rized changes in cell cycle regulators afterandrogen withdrawal as early and mid-to-late events. The early cell cycle eventswere consistent with a cellular stress re-sponse associated with an increase in p53,followed by an increase in p21, and atransient early G1/G0-phase arrest, asshown by a decrease in Ki67 expressionand a reduction in the number of cells inS phase. The lack of an apoptotic re-sponse during this period is consistentwith the low-to-undetectable levels of baxand bcl-2 proteins. These data suggestthat the growth arrest observed after an-drogen withdrawal is the result of a fail-ure to activate cell death mechanisms inthis xenograft model and that activatingp53-independent cell death signals maybe an important strategy to effect a com-plete response to androgen withdrawal.

The initial cell stress response in thissystem (an increase in p53 and p21 ex-pression and a block in G1-phase cellcycle progression) is similar to that ob-served after exposure of cells to high-doseUV radiation (11). Ectopic expression ofdecorin(12) and treatment of several hu-

Fig. 4. Panel A:Levels of expression of cell cycle regulators, Ki67, and the androgen receptor in androgen-independent sublines compared with levels in the parental CWR22 xenograft tumor before androgen with-drawal. Duplicate tumors were evaluated at each time point. The range was consistently below ±10% of thelisted value.Panels BandC: Sections of tumors from a nude mouse with CWRSA4 androgen-independentxenograft. Representative areas of CWRSA4 tumor sections stained with monoclonal antibodies to cyclin D1and mdm2, respectively, are shown.Bars 4 100 mm.

Journal of the National Cancer Institute, Vol. 91, No. 21, November 3, 1999 REPORTS 1873

man cancer cell lines with mimosine andaphidicolin (13) have also been reportedto induce increments of p53 and p21 ex-pression, leading to permanent growth ar-rest. This p21-mediated G1-phase arresthas been shown to be dependent on func-tional pRB (14).

After androgen withdrawal, tumor re-gression appears to result from undetect-able cell death or, more likely, fromchanges in cell volume. The increase inp53 staining at day 3, in the absence ofapoptosis, suggests a defect in p53-mediated apoptosis. Androgen with-drawal results in the apoptosis of nonma-lignant prostate epithelial cells and isassociated with a decrease in cell size(15,16).Studies of xenografts and spon-taneous rat prostate cancers have been in-conclusive. Increased apoptotic indiceswere observed in PC-82 and LuCaP hu-man prostate cancer xenograft models(17,18),whereas decreased indices wereobserved in the Dunning R3327PAP ratmodel after androgen withdrawal(19,20).Conflicting results have been observed inhuman prostate cancers after androgenwithdrawal, since increased levels of ap-optosis have been reported by somegroups of investigators(21–23)but not byother groups(24–27). The apparentlycontradicting results may reflect the smallproportion of cells actually undergoingapoptosis at any one time, the inability tosample tumors repeatedly at differenttimes, or the small overall contribution ofapoptotic cell death to human tumor re-gression. An alternative explanation isthat the cell cycle checkpoint status of thevarious tumors is different. The integrityof the checkpoint status can change theresponse to anticancer therapy from cellcycle arrest to cell death in the xenograftmodel system(28).

Although changes in early cell cycleevents were transitory, the mid-to-late re-sponses were progressive and sustained.We found a decrease in cyclin D1 levelsfollowed by changes in pRB expression,changes similar to those described for bu-tyrate-induced G1-phase arrest(29). Bu-tyrate inhibits the mitogen-dependenttranscriptional induction of cyclin D1 andphosphorylation of pRB(30). Withdraw-ing androgen removes a mitogenic signalfor cyclin D1, resulting in pRB hypophos-phorylation and G1-phase arrest. Addi-tional factors associated with the mainte-nance of arrested growth are the observedoverexpression of p21, p27, and p16.

The inhibition of cell cycle progression

and the antiproliferative activity of rapa-mycin have been reported to be dependenton p27 (31). More recently, it has beenreported that the protein kinase inhibitorstaurosporine induces a G1-phase arrestin cells expressing functional pRB(32).In staurosporine-treated cells, the levelsof p27 increase, preventing activationof cyclin E/cyclin-dependent kinase-2 andcontributing to the G1-phase arrest. Inhormone-naive primary prostate cancer,low-to-undetectable p27 levels were asso-ciated with poor clinical outcomes, inde-pendent of tumor stage and grade(10,33–36). However, in one study(10),androgen-independent metastatic bone le-sions had low levels of p27, consistentwith the xenograft findings. Cellular re-sponses to stress, such as hyperthermia,have been shown to induce p16 and toarrest cells in G0/G1 phase in a p53-independent manner(37). Furthermore,the expression of p16 induces decreasedtranscription of the RB gene(38).

Androgen independence was markedby the expression of the oncogenic pro-teins cyclin D1 and mdm2. The increasedmdm2 expression increases p53 transcrip-tional transactivation(39) and also inac-tivates pRB(40). Cyclin D1 is a pivotalmolecule in cell cycle regulation. Its over-expression has been shown to participatein the oncogenesis of many tumors(41,42).

Androgen receptor expression de-creased substantially after androgen with-drawal. The mechanism of this decrease isnot known, but this observation has beenmade in other systems(43).The androgenreceptor in the CWR22 model has a mu-tation in the ligand-binding domain at po-sition 874 (44). Similar to the wild-typeandrogen receptor, the receptor is func-tional in the presence of testosterone anddihydrotestosterone(44). Androgen re-ceptor expression is present again at days10–16 after androgen withdrawal and tovarious amounts in androgen-independentsublines. These results are similar to thehuman condition, where androgen recep-tor is expressed in the majority of un-treated prostate cancers and in most pa-tients whose cancer recurs duringandrogen withdrawal therapy(45). Otherinvestigators(46,47)have shown that am-plification and the increased expression ofa wild-type androgen receptor gene mayplay an important role in androgen-independent prostate cancer. It is of inter-est that the two androgen-independent tu-mors with the highest number of cells

staining for androgen receptor and themost intense staining showed androgen-repressed growth. Androgen repression ofprostate cancer growth has been describedpreviously(48–50).Similar to the currentresults, androgen-repressible lines ex-pressed higher basal levels of androgenreceptor protein and messenger RNA thandid the parental line(50).The overexpres-sion of androgen receptor in androgen-independent prostate cancer may identifya subset of prostate cancers in which an-drogen treatment may be of clinical ben-efit.

This study demonstrates that reproduc-ible cell cycle protein regulator changesoccur with androgen withdrawal in onexenograft model of human prostate can-cer. These changes were associated with acell stress response and the exit of theprostate cancer cells from the active cellcycle with the absence of observed apop-tosis. The cell cycle changes associatedwith the emergence of androgen indepen-dence are more heterogeneous but suggestregulatory pathways for this indepen-dence from hormonal stimulation. Thera-peutic strategies designed to address thecell cycle-arrested prostate cancer cellsafter androgen withdrawal are being in-vestigated in an effort to eliminate theemergence of androgen independence.

REFERENCES

(1) Wainstein MA, He F, Robinson D, Kung HJ,Schwartz S, Giaconia JM, et al. CWR22: an-drogen-dependent xenograft model derivedfrom a primary human prostatic carcinoma.Cancer Res 1994;54:6049–52.

(2) Nagabhushan M, Miller CM, Pretlow TP, Gia-conia JM, Edgehouse NL, Schwartz S, et. al.CWR22: the first human prostate cancer xeno-graft with strongly androgen-dependent and re-lapsed strains bothin vivo and in soft agar.Cancer Res 1996;56:3042–6.

(3) Pretlow TG, Wolman SR, Micale MA, PelleyRJ, Kursh ED, Resnick MI, et al. Xenografts ofprimary human prostatic carcinoma. J NatlCancer Inst 1993;85:394–8.

(4) Agus DB, Golde DW, Sgouros G, BallangrudA, Cordon-Cardo C, Scher HI. Positron emis-sion tomography of a human prostate cancerxenograft: association of changes in deoxyglu-cose accumulation with other measures of out-come following androgen withdrawal. CancerRes 1998;58:3009–14.

(5) Bauer KD, Clevenger CV, Endow RK, MuradT, Epstein AL, Scarpelli DG. Simultaneousnuclear antigen and DNA content quantitationusing paraffin-embedded colonic tissue andmultiparameter flow cytometry. Cancer Res1986;46:2428–34.

(6) Fuks Z, Persaud RS, Alfieri A, McLoughlin M,Ehleiter D, Schwartz JL, et al. Basic fibroblastgrowth factor protects endothelial cells against

1874 REPORTS Journal of the National Cancer Institute, Vol. 91, No. 21, November 3, 1999

radiation-induced programmed cell deathinvitro and in vivo. Cancer Res 1994;54:2582–90.

(7) Gavrieli Y, Sherman Y, Ben-Sasson SA. Iden-tification of programmed cell deathin situ viaspecific labeling of nuclear DNA fragmenta-tion. J Cell Biol 1992;119:493–501.

(8) Osman I, Scher HI, Zhang ZF, Pellicer I,Hamza R, Eissa S, et al. Alterations affectingthe p53 control pathway in bilharzial-relatedbladder cancer. Clin Cancer Res 1997;3:531–6.

(9) Nguyen H, Gitig DM, Koff A. Cell-free deg-radation of p27(kip1), a G1 cyclin-dependentkinase inhibitor, is dependent on CDK2 activ-ity and the proteasome. Mol Cell Biol 1999;19:1190–201.

(10) Cordon-Cardo C, Koff A, Drobnjak M, Capo-dieci P, Osman I, Millard SS, et al. Distinctaltered patterns of p27KIP1 gene expressionin benign prostatic hyperplasia and prostaticcarcinoma. J Natl Cancer Inst 1998;90:1284–91.

(11) Wu L, Levine AJ. Differential regulation of thep21/WAF-1 and mdm2 genes after high-doseUV irradiation: p53-dependent and p53-independent regulation of the mdm2 gene. MolMed 1997;3:441–51.

(12) Santra M, Mann DM, Mercer EW, Skorski T,Calabretta B, Iozzo RV. Ectopic expressionof decorin protein core causes a general-ized growth suppression in neoplastic cellsof various histogenetic origin and requiresendogenous p21, an inhibitor of cyclin-dependent kinases. J Clin Invest 1997;100:149–57.

(13) Ji C, Marnett LJ, Pietenpol JA. Cell cycle re-entry following chemically-induced cell cyclesynchronization leads to elevated p53 and p21protein levels. Oncogene 1997;15:2749–53.

(14) Niculescu AB 3rd, Chen X, Smeets M, HengstL, Prives C, Reed SI. Effects of p21 (Cip1/Waf1) at both the G1/S and the G2/M cell cycletransitions: pRb is a critical determinant inblocking DNA replication and in preventingendoreduplication [published erratum appearsin Mol Cell Biol 1998;18:1763]. Mol Cell Biol1998;18:629–43.

(15) Kyprianou N, Isaacs JT. Activation of pro-grammed cell death in the rat ventral prostateafter castration. Endocrinology 1988;122:552–62.

(16) Kerr JF, Searle J. Deletion of cells by apoptosisduring castration-induced involution of the ratprostate. Virchows Arch B Cell Pathol 1973;13:87–102.

(17) van Weerden WM, van Kreuningen A, ElissenNM, Vermeij M, de Jong FH, van SteenbruggeGJ, et al. Castration-induced changes in mor-phology, androgen levels, and proliferative ac-tivity of human prostate cancer tissue grownin athymic nude mice. Prostate 1993;23:149–64.

(18) Bladou F, Vessella RL, Buhler KR, Ellis WJ,True LD, Lange PH. Cell proliferation and ap-optosis during prostatic tumor xenograft invo-lution and regrowth after castration. Int J Can-cer 1996;67:785–90.

(19) Brandstrom A, Westin P, Bergh A, Cajander S,Damber JE. Castration induces apoptosis in theventral prostate but not in an androgen-

sensitive prostatic adenocarcinoma in the rat.Cancer Res 1994;54:3594–601.

(20) Westin P, Bergh A, Damber JE. Castration rap-idly results in a major reduction in epithelialcell numbers in the rat prostate, but not in thehighly differentiated Dunning R3327 prostaticadenocarcinoma. Prostate 1993;22:65–74.

(21) Montironi R, Pomante R, Diamanti L, Magi-Galluzzi C. Apoptosis in prostatic adenocarci-noma following complete androgen ablation.Urol Int 1998;60 Suppl 1:25–9.

(22) Denmeade SR, Lin XS, Isaacs JT. Role of pro-grammed (apoptotic) cell death during the pro-gression and therapy for prostate cancer [pub-lished erratum appears in Prostate 1996;28:414]. Prostate 1996;28:251–65.

(23) Reuter VE. Pathological changes in benign andmalignant prostatic tissue following androgendeprivation therapy. Urology 1997;49(3ASuppl):16–22.

(24) Westin P, Stattin P, Damber JE, Bergh A. Cas-tration therapy rapidly induces apoptosis in aminority and decreases cell proliferation in amajority of human prostatic tumors. Am JPathol 1995;146:1368–75.

(25) Murphy WM, Soloway MS, Barrows GH.Pathologic changes associated with androgendeprivation therapy for prostate cancer. Cancer1991;68:821–8.

(26) Dhom G, Degro S. Therapy of prostatic cancerand histopathologic follow-up. Prostate 1982;3:531–42.

(27) Tomic R, Bergman B, Hietala SO, AngstromT. Prognostic significance of transrectal fine-needle aspiration biopsy findings after orchiec-tomy for carcinoma of the prostate. Eur Urol1985;11:378–81.

(28) Waldman T, Zhang Y, Dillehay L, Yu J, Kin-zler K, Vogelstein B, et al. Cell-cycle arrestversus death in cancer therapy. Nat Med 1997;3:1034–6.

(29) Xiao H, Hasegawa T, Miyaishi O, Ohkusu K,Isobe K. Sodium butyrate induces NIH3T3cells to senescence-like state and enhances pro-moter activity of p21WAF/CIP1 in p53-independent manner. Biochem Biophys ResCommun 1997;237:457–60.

(30) Vaziri C, Stice L, Faller DV. Butyrate-inducedG1 arrest results from p21-independent disrup-tion of retinoblastoma protein-mediated sig-nals. Cell Growth Differ 1998;9:465–74.

(31) Luo Y, Marx SO, Kiyokawa H, Koff A, Mas-sague J, Marks AR. Rapamycin resistance tiedto defective regulation of p27Kip1. Mol CellBiol 1996;16:6744–51.

(32) Juan G, Gruenwald S, Darzynkiewicz Z. Phos-phorylation of retinoblastoma susceptibilitygene protein assayed in individual lymphocytesduring their mitogenic stimulation. Exp CellRes 1998;239:104–10.

(33) Cote RJ, Shi Y, Groshen S, Feng AC, Cordon-Cardo C, Skinner D, et al. Association ofp27Kip1 levels with recurrence and survival inpatients with stage C prostate carcinoma. J NatlCancer Inst 1998;90:916–20.

(34) Guo Y, Sklar GN, Borkowski A, Kyprianou N.Loss of the cyclin-dependent kinase inhibitorp27(Kip1) protein in human prostate cancercorrelates with tumor grade. Clin Cancer Res1997;3:2269–74.

(35) Cheville JC, Lloyd RV, Sebo TJ, Cheng L,Erickson L, Bostwick DG, et al. Expression ofp27kip1 in prostatic adenocarcinoma. ModPathol 1998;11:324–8.

(36) Yang RM, Naitoh J, Murphy M, Wang HJ,Phillipson J, deKernion JB, et al. Low p27 ex-pression predicts poor disease-free survival inpatients with prostate cancer. J Urol 1998;159:941–5.

(37) Valenzuela MT, Nunez MI, Villalobos M,Siles E, McMillan TJ, Pedraza V, et al. Acomparison of p53 and p16 expression in hu-man tumor cells treated with hyperthermia orionizing radiation. Int J Cancer 1997;72:307–12.

(38) Fang X, Jin X, Xu HJ, Liu L, Peng HQ, HoggD, et al. Expression of p16 induces transcrip-tional downregulation of the RB gene. Onco-gene 1998;16:1–8.

(39) Cordon-Cardo C, Latres E, Drobnjak M, OlivaMR, Pollack D, Woodruff JM, et al. Molecularabnormalities of mdm2 and p53 genes in adultsoft tissue sarcomas. Cancer Res 1994;54:794–9.

(40) Xiao ZX, Chen J, Levine AJ, Modjtahedi N,Xing J, Sellers WR, et al. Interaction betweenthe retinoblastoma protein and the oncoproteinMDM2. Nature 1995;375:694–8.

(41) Lammie GA, Peters G. Chromosome 11q13abnormalities in human cancer. Cancer Cells1991;3:413–20.

(42) Lebwohl DE, Muise-Helmericks R, Sepp-Lo-renzino L, Serve S, Timaul M, Bol R, et al. Atruncated cyclin D1 gene encodes a stablemRNA in a human breast cancer cell line. On-cogene 1994;9:1925–9.

(43) Ruizeveld de Winter JA, van Weerden WM,Faber PW, van Steenbrugge GJ, Trapman J,Brinkmann AO, et al. Regulation of androgenreceptor expression in the human heterotrans-plantable prostate carcinoma PC-82. Endocri-nology 1992;131:3045–50.

(44) Tan J, Sharief Y, Hamil KG, Gregory CW,Zang DY, Sar M, et al. Dehydroepiandros-terone activates mutant androgen receptors ex-pressed in the androgen-dependent humanprostate cancer xenograft CWR22 and LNCaPcells. Mol Endocrinol 1997;11:450–9.

(45) Goldenberg SL, Bruchovsky N. Androgenwithdrawal therapy: new perspectives in thetreatment of prostate cancer. In: Raghavan D,Scher HI, Leibel SA, Lange P, editors. Prin-ciples and practice of genitourinary oncology.Philadelphia (PA): Lippincott-Raven; 1997. p.583–90.

(46) Bentel JM, Tilley WD. Androgen receptorsin prostate cancer. J Endocrinol 1996;151:1–11.

(47) Koivisto P, Kononen J, Palmberg C, TammelaT, Hyytinen E, Isola J, et al. Androgen receptorgene amplification: a possible molecularmechanism for androgen deprivation therapyfailure in prostate cancer. Cancer Res 1997;57:314–9.

(48) Zhau HY, Chang SM, Chen BQ, Wang Y,Zhang H, Kao C, et al. Androgen-repressedphenotype in human prostate cancer. Proc NatlAcad Sci U S A 1996;93:15152–7.

(49) Kokontis JM, Hay N, Liao S. Progression ofLNCaP prostate tumor cells during androgen

Journal of the National Cancer Institute, Vol. 91, No. 21, November 3, 1999 REPORTS 1875

deprivation: hormone-independent growth, re-pression of proliferation by androgen, and rolefor p27Kip1 in androgen-induced cell cycle ar-rest. Mol Endocrinol 1998;12:941–53.

(50) Kokontis J, Takakura K, Hay N, Liao S. In-creased androgen receptor activity and alteredc-myc expression in prostate cancer cells afterlong-term androgen deprivation. Cancer Res1994;54:1566–73.

NOTES

Supported in part by Public Health Servicegrants R01CA30388 (National Cancer Institute)(to D. W. Golde), R01HL42107 (National Heart,Lung, and Blood Institute) (to D. W. Golde), andDK47650 (National Institute of Diabetes andDigestive and Kidney Diseases) (to C. Cordon-Cardo), National Institutes of Health, Departmentof Health and Human Services; and the PepsiCoFoundation and CaP CURE (H. I. Scher and D. B.Agus). D. B. Agus is supported by a Physician’sResearch Training Award from the American Can-cer Society, a CaP CURE Young InvestigatorAward, and the Eleanor and Paul Stephens Founda-tion.

Manuscript received January 29, 1999; revisedAugust 4, 1999; accepted August 30, 1999.

1876 REPORTS Journal of the National Cancer Institute, Vol. 91, No. 21, November 3, 1999