ExpressionofAndrogenandEstrogenSignalingComponents and ...containing 25 mL of ISOGEN reagent (Nippon Gene) and tissues were stored at 80 C before RNA isolation. Quantitative reverse-transcription

Biology of Human Tumors

ExpressionofAndrogenandEstrogenSignalingComponentsand Stem Cell Markers to Predict Cancer Progression andCancer-SpecificSurvival in PatientswithMetastatic ProstateCancer

Tetsuya Fujimura1, Satoru Takahashi2, Tomohiko Urano3,4, Kenichi Takayama3,4, Toru Sugihara1,Daisuke Obinata2, Yuta Yamada1, Jimpei Kumagai1, Haruki Kume1, Yasuyoshi Ouchi3, Satoshi Inoue3,4, andYukio Homma1

AbstractPurpose: Genes of androgen and estrogen signaling cells and stem cell–like cells play crucial roles in

prostate cancer. This study aimed topredict clinical failure by identifying these prostate cancer-related genes.

Experimental Design: We developed models to predict clinical failure using biopsy samples from a

training set of 46 and an independent validation set of 30 patients with treatment-na€�ve prostate cancer with

bonemetastasis. Cancerous and stromal tissues were separately collected by laser-capturedmicrodissection.

We analyzed the association between clinical failure andmRNA expression of the following genes androgen

receptor (AR) and its related genes (APP, FOX family, TRIM 36,Oct1, and ACSL 3), stem cell–like molecules

(Klf4, c-Myc, Oct 3/4, and Sox2), estrogen receptor (ER), Her2, PSA, and CRP.

Results: Logistic analyses to predict prostate-specific antigen (PSA) recurrence showed an area under the

curve (AUC) of 1.0 in both sets for Sox2,Her2, andCRP expression in cancer cells, AR and ERa expression in

stromal cells, and clinical parameters. We identified 10 prognostic factors for cancer-specific survival (CSS):

Oct1, TRIM36, Sox2, and c-Myc expression in cancer cells; AR, Klf4, and ERa expression in stromal cells; and

PSA,Gleason score, and extent of disease.On the basis of these factors, patients were divided into favorable-,

intermediate-, and poor-risk groups according to the number of factors present. Five-year CSS rates for the 3

groupswere 90%,32%, and12%in the training set and75%,48%, and0%in the validation set, respectively.

Conclusions: Expression levels of androgen- and estrogen signaling components and stem cell markers

are powerful prognostic tools. Clin Cancer Res; 20(17); 4625–35. �2014 AACR.

IntroductionThe pioneering work of Huggins and Hodges (1) showed

that prostate cancer is sensitive to androgen deprivationtherapy (ADT); however, at least two problems are associ-ated with ADT. First, durability of ADT varies among pros-tate cancer patients (2). For example, the median survivaltime was 13 months for patients with prostate-specificantigen (PSA) nadirs of >4 ng/mL, 44 months for patients

with PSAnadirs of 0.2–4ng/mL, and75months for patientswith PSA nadirs of�0.2 ng/mL (2). Second, ADT is initiallyeffective as a treatment for advanced prostate cancer, but notwhenprostate cancer acquires a castration-resistant status. Arecent study proposed that stem cell–like prostate cancercells with a pluripotent phenotype are involved in castra-tion-resistant prostate cancer (CRPC; ref. 3). In cases withstem cell–like prostate cancer cells, ADT may actually stim-ulate cancer progression (3). Therefore, evaluation of boththe durability of ADT and presence of stem cell–like cellcomponents in prostate needle biopsy samples before ADTprescription is the first step in personalized medicine forpatients with metastatic prostate cancer.

Determination of the therapeutic strategy for breast can-cer commonly depends on the expression patterns of estro-gen receptor (ERa), progesterone receptor (PR), and Her 2in needle biopsy samples (4). However, pretreatment diag-nosis by estimating gene expression is not yet prevalent forpatients with prostate cancer. The growth-inhibitory effectson prostate cancer cells are associated with the status ofsteroid nuclear receptors and related genes of prostatecancer–related molecules, such as androgen receptor (AR;refs. 5, 6), AR-related genes (5, 6), amyloid precursor

1Department of Urology, Graduate School of Medicine, The University ofTokyo, Tokyo, Japan. 2Department of Urology, Graduate School of Med-icine, The Nihon University, Tokyo, Japan. 3Department of Geriatric Med-icine,Graduate School ofMedicine, TheUniversity of Tokyo, Tokyo, Japan.4Department of Anti-Aging Medicine, Graduate School of Medicine, TheUniversity of Tokyo, Tokyo, Japan.

Note: Supplementary data for this article are available at Clinical CancerResearch Online (http://clincancerres.aacrjournals.org/).

Corresponding Author: Tetsuya Fujimura, Department of Urology, Grad-uate School of Medicine, The University of Tokyo, Bunkyo-ku, Tokyo113-8655, Japan. Phone: 813-5800-8753; Fax: 813-5800-8917; E-mail:[email protected]

doi: 10.1158/1078-0432.CCR-13-1105

�2014 American Association for Cancer Research.

ClinicalCancer

Research

www.aacrjournals.org 4625

on March 10, 2020. © 2014 American Association for Cancer Research. clincancerres.aacrjournals.org Downloaded from

Published OnlineFirst July 1, 2014; DOI: 10.1158/1078-0432.CCR-13-1105

http://clincancerres.aacrjournals.org/

protein (APP; ref. 7), forkhead-box (FOX) family proteins(8–11), octameter transcription factor (Oct1; ref. 12), andERs (5, 13, 14). However, the distribution and functions ofthese steroid nuclear receptors vary depending on the sitesfrom which prostate cancer samples are obtained (13–15).For example, theAR gene is expressed in luminal, basal, andstromal cells. AR functions as a proliferation stimulator instromal cells, as a suppressor in epithelial basal cells, and asa survival factor in epithelial luminal cells (15). Therefore,cancer and stromal cells should be separately collected forthe analysis of prostate cancer samples.

Here, we attempted to predict PSA recurrence and cancer-specific survival (CSS) by measuring expression of genesrelated to prostate cancer in patients with bone metastasisusing laser-captured microdissection (LCM) technique.

Materials and MethodsTissue selection and patient characteristics of thetraining set

Formalin-fixed, paraffin-embedded sections of the pri-mary tumors were obtained from 46 treatment-na€�veconsecutive patients (age, 58–87 years; mean age, 74years) diagnosed with bone metastatic prostate cancerbetween 2001 and 2009 by transrectal ultrasound–guidedbiopsy. This study was approved by our institutionalethics committee. Before treatment, serum PSA levelswere 8.6 to 13,700 ng/mL (mean, 219 ng/mL). Thesections were evaluated by two pathologists, and thetumors were assigned to Gleason scores (GS) of �7(n ¼ 8), 8 (n ¼ 11), 9 (n ¼ 26), and 10 (n ¼ 1). The clini-cal primary tumor (cT) stages were 2 (n ¼ 6), 3a (n ¼ 15),3b (n ¼ 14), and 4 (n ¼ 11). The clinical regional lymphnode (cN) stages were 0 (n ¼ 21) and 1 (n ¼ 25). Todiagnose bone metastasis, we performed bone scintigra-

phy using technetium-99m-methylene diphosphonate inall the patients. Computed tomography was used in 3 of46 patients to distinguish from bone degenerativechanges. On the basis of the number or extent of metas-tases, the scans were divided into five extent of disease(EOD) grades as follows: 0, normal or abnormal due tobenign bone disease; 1, number of bony metastases lessthan 6, each of which is less than 50% the size of avertebral body (one lesion about the size of a vertebralbody would be counted as two lesions): 2, number ofbone metastases between 6 and 20, size of lesions asdescribed above; 3, number of metastases greater than 20but fewer than a "super scan"; and 4, "superscan" or itsequivalent (i.e., more than 75% of the ribs, vertebrae, andpelvic bones; ref. 16). EOD was 1 (n ¼ 25), 2 (n ¼ 10), 3(n ¼ 9), and 4 (n ¼ 2).

All patients received ADT by medical or surgical cas-tration with or without the administration of antiandro-gen agents, bicalutamide (n ¼ 22), chlormadinone (n ¼3), flutamide (n ¼ 3), and estramustine phosphonate(n ¼ 2; Table 1). PSA relapse was defined by consecutiveincrease in serum PSA levels to above the patient’s PSAnadir (17). If PSA relapse occurred during initial ADT,either new antiandrogen agent was added or switched toanother. When patients were switched to other antian-drogen agents, antiandrogen withdrawal syndrome waschecked. From September 2008, systemic chemotherapyby docetaxel was also administered every 3 or 4 weeks. Ifprostate cancer became hormone- and chemotherapy-refractory, patients received best supportive care. A totalof 37 patients (80%) experienced relapse. The mean timeto PSA relapse was 972 � 1,193 days (range, 5–4,616days). The mean follow-up duration was 1,650 � 1,319days (range, 79–5,961 days). At the end of the follow-upperiod, 9 patients (19%) were alive without PSA relapse,whereas 15 (33%) were alive with biochemical or clinicalrecurrence. Twenty-two patients (48%) died of prostatecancer during the follow-up period.

Patient characteristics of the validation setAn independent cohort of 30 patients with prostate

cancer with bone metastasis (age, 59–91 years; mean age,68 years) between 2001 and 2011 were enrolled. Beforetreatment, serum PSA levels were 5.8 to 8,428 ng/mL(mean, 498 ng/mL). The tumors were assigned to Gleasonscores of�7 (n¼ 4), 8 (n¼ 4), 9 (n¼ 16), and 10 (n ¼ 6).The cT stages were 2 (n ¼ 4), 3a (n ¼ 9), 3b (n ¼ 8), and 4(n ¼ 9). The cN stages were 0 (n ¼ 5) and 1 (n ¼ 25). EODwas 1 (n ¼ 11), 2 (n ¼ 9), 3 (n ¼ 4), and 4 (n ¼ 6; ref. 16).

All patients received ADT by medical or surgical castra-tion with or without the administration of antiandrogenagents, bicalutamide (n ¼ 27), and estramustine phospho-nate (n ¼ 3). A total of 25 patients (83%) experiencedrelapse. The mean time to PSA relapse was 571 � 877 days(range, 6–4,616 days). The mean follow-up duration was1,143 � 1,226 days (range, 71–6,395 days). At the end ofthe follow-up period, 5 patients (16%) were alive withoutPSA relapse, whereas 13 (43%) were alive with biochemical

Translational RelevanceAndrogen and estrogen signaling play crucial roles in

prostate cancer. Stem cell–like cells are also known to beinvolved in prostate cancer progression. In the presentstudy, we investigated the expression of androgen andestrogen signaling components and stem cell markers topredict prostate-specific antigen (PSA) recurrence andcancer-specific survival (CSS) in patients with metastaticprostate cancer. Discriminant analysis using the mRNAexpression of AR, ERa, Sox2, Her2, CRP, and clinicalparameters highly predicted PSA recurrence. We identi-fied 10 prognostic factors for CSS: Oct1, TRIM36, Sox2,and c-Myc, AR, Klf4, and ERa expression as well as PSA,Gleason score, and extent of disease. On the basis ofthese factors, we propose a new risk classification forCSS. Taken together, the expression pattern of androgenand estrogen signaling components and stem cell mar-kers predicted PSA recurrence and CSS in patients withprostate cancer characterized by bone metastasis.

Fujimura et al.

Clin Cancer Res; 20(17) September 1, 2014 Clinical Cancer Research4626




or clinical recurrence. Twelve patients (40%) died of pros-tate cancer during the follow-up period.

Laser-captured microdissectionTissue sections (10 mm) were deparaffinized and rehy-

drated using graded ethanol and rinsed in diethylpyrocar-bonate-treated water. After staining with 0.05% toluidineblue solution (WAKO), the tissue sections were separatedinto groups of cancer cells and stromal cells by laser micro-dissection (Leica 6500). To obtain sufficient materials, we

collected 30 to 460 acini of epithelium within an area of3,865 mm3 in average, and all of the surrounding stromawere collected. Tissues were collected into Eppendorf capcontaining 25 mL of ISOGEN reagent (Nippon Gene) andtissues were stored at �80�C before RNA isolation.

Quantitative reverse-transcription PCRTotal RNA was extracted using ISOGEN PB kit (Nippon

Gene). Tissue samples were incubated 15 minutes withproteinase K and extraction buffer at 50�C then added toan ISOGEN-LS (NipponGene).BoundRNAwaspurified inaseries ofwash steps to remove cellular components. ResidualDNAwasdigested by incubating the eluatewithDNase. RNAquality and RNA quantity was assessed using a NanoDropND-1000 spectrophotometer (Japan SCRUM Inc). The ratioof absorbance at 260nmand280nmwas 1.7:2.0, anda totalof 40 ng RNA was used. First-strand cDNA was generatedusing PrimeScript RT Master Mix (Takara). Quantitativereverse-transcription PCR (qRT-PCR) was performed with150 nmol/L primers using 7300 Real-Time PCR system(Applied Biosystems); one cycle at 50�C for 2 minutes, onecycle at 95�C for 10minutes, 40 cycles at 95�C for 15 secondsand at 50�C for 1 minute, one cycle at 95�C for 15 seconds,one cycle at 59�C for 30 seconds, andone cycle at 95�C for 15seconds. mRNA expression was normalized relative toGAPDH and the average relative expression of 4 timesexamination was adopted.

PrimersThe size of PCR amplicons (base pair: bp) and the

sequences of the PCR primers used are shown below:

AR (59 bp) AR forward: 50-GCTGCAAGGTCTTCTTCAAAAGA-30

AR reverse: 50-GCTGGCGCACAGGTACTTCT-30

Oct1 (130 bp) Oct1 forward: 50-CCTGCTTTCTTTTGCGGTAG-30

Oct1 reverse: 50- GTTCTGTTTTCGCCCAACAT-30

FOXO1 (128 bp) FOXO1 forward: 50-CTGCATCCATGGACAACAAC-30

FOXO1 reverse: 50- AGGCCATTTGGAAAACTGTG-30

FOXA1 (78 bp) FOXA1 forward: 50-CATTGCCATCGTGTGCTTGT-30

FOXA1 reverse: 50- CCCGTCTGGCTATACTAACACCAT-30

FOXP1 (65 bp) FOXP1 forward: 50-ACCGCTTCCATGGGAAAT-30

FOXP1 reverse: 50- CCGTTCAGCTCTTCCCGTATT-30

APP (121 bp) APP forward: 50-GAGACACCTGGGGATGAGAA-30

APP reverse: 50- CTTGACGTTCTGCCTCTTCC-30

ACSL3 (110 bp) ACSL3 forward: 50-GACACAAGGGCGCATATCTT-30

ACSL3 reverse: 50- AGGGTGGCAAATGTTACAGC-30

TRIM36 (136 bp) TRIM36 forward: 50-CGTTGTTCTGCCTTTCACAA-30

TRIM36 reverse: 50- GCAACACCAGCTGACAGAAA-30

Oct3/4 (110 bp) Oct3/4 forward: 50-AGTGAGAGGCAACCTGGAGA-30

Table 1. Correlation between men with andwithout PSA recurrence in patients with bonymetastatic prostate cancer (n ¼ 46)

Groups

Clinical findings

PSArecurrence(n ¼ 37)

Without PSArecurrence(n ¼ 9) P

Age, y 70 � 8.0 72 � 4.3 0.20Serum PSA (ng/mL) 1,109 � 2,421 1,170 � 2,777 0.15Gleason score�7 7 1 0.558 7 49 22 410 1 0

Clinical T stage2 4 2 0.273a 11 43b 13 14 9 2

Clinical N stage0 14 5 0.611 23 4

EOD1 18 7 0.282 8 23 9 04 2 0

MABNo 14 3 0.58Yes 21 6

Initial antiandrogen agentBicalutamide 17 5Chlormadinone 3 0Flutamide 2 1Estrogens 2 0

PSA nadir<0.01 1 9 <0.00010.01–0.1 16 00.1� 20 0

Time to PSA nadirafter ADT (days)

424 � 286 419 � 303 0.31

Abbreviations: MAB, maximum androgen blockade withantiandrogen agents.

Androgen and Estrogen Signaling and Stem Cell Markers in Prostate Cancer

www.aacrjournals.org Clin Cancer Res; 20(17) September 1, 2014 4627




Oct3/4 reverse: 50- ACACTCGGACCACATCCTTC-30

Sox2 (95 bp) Sox2 forward: 50-CAAGATGCACAACTCGGAGA-30

Sox2 reverse: 50- GCTTAGCCTCGTCGATGAAC-30

Klf4 (127 bp) Klf4 forward: 50-ACTCGCCTTGCTGATTGTCT-30

Klf4 reverse: 50- AGTTAACTGGCAGGGTGGTG-30

c-Myc (123 bp) c-Myc forward: 50-TCAAGAGGCGAACACACAAC-30

c-Myc reverse: 50- TAACTACCTTGGGGGCCTTT-30

CRP (107 bp) CRP forward: 50-TGGTCTTGACCAGCCTCTCT-30

CRP reverse: 50- CGGTGCTTTGAGGGATACAT-30

Her2 (132 bp) Her2 forward: 50-ACCAAGCTCTGCTCCACACT-30

Her2 reverse: 50- ACTGGCTGCAGTTGACACAC-30

ERb (175 bp) ERb forward: 50-AAGAAGATTCCCGGCTTTGT-30

ERb reverse: 50- CTTCTACGCATTTCCCCTCA-30

Klf5 (81 bp) Klf5 forward: 50-CACCTCCATCCTATGCTGCT-30

Klf5 reverse: 50- AGTTAACTGGCAGGGTGGTG-30

ERa (153 bp) ERa forward: 50-AGCACCCTGAAGTCTCTGGA-30

ERa reverse: 50- GATGTGGGAGAGGATGAGGA-30

GAPDH (80 bp) GAPDH forward: 50-GGTGGTCTCCTCTGACTTCAACA-30

GAPDH reverse: 50- GTGGTCGTTGAGGGCAATG-30

We have previously uncovered the AR transcriptionalnetwork in prostate cancer cells by chromatin immuno-precipitation (ChIP) combined with DNA microarray(ChIP-chip) and cap analysis gene expression (CAGE)(18). In addition, on the basis of previous studies, weselected APP (7), FOX family proteins (FOXO1, FOXA1,and FOXP1; refs. 9–11), tripartite molecule 36 (TRIM;ref. 18), Oct1 (12), and ACSL 3 (18) for this study. Stemcell–like markers [Oct3/4, Sox2, Kru��ppel-like factor (Klf4),and c-Myc; refs. 19, 20) and prostate cancer–relatedgenes (CRP,Her2, ERb, Klf5, and ERa) were also evaluated(21–23).

AntibodiesWe performed immunohistochemistry of AR and Klf4 to

evaluate a correlation with its mRNA expression.Because only a small amount of the biopsy samples was

available, we chose two antibodies; AR (AR441; ref. 24) andKlf4 (ab72543; ref. 25) which had been reported to beadequate for immunohistochemical analysis. Mousemonoclonal antibody for AR (AR441) and rabbit polyclon-al antibody for Klf4 (ab72543) were purchased from Dakoand Abcam, respectively.

Immunohistochemical analysisImmunohistochemical analysis for AR and Klf4 was

performedwith the streptavidin-biotin amplificationmeth-od using an EnVisionþ Visualization Kit (Dako) for AR andKlf4, as previously described (12, 26). Theprimary antibody

against AR and Klf4 (1:50 dilution) was applied and incu-bated at room temperature for 1 hour. The sections werethen rinsed in PBS and incubated at room temperature withEnVisionþ for 1 hour. The antigen–antibody complex wasvisualized with 3, 30-diaminobenzidine (DAB) solution [1mmol/L DAB, 50 mmol/L Tris-HCl buffer (pH 7.6), and0.006% H2O2].

Immunohistochemical assessmentThe sample slides were evaluated for staining intensity

(ref. 26; none, weak, moderate, and strong) based onlabeling index (LI; ref. 27). LIs were determined bycounting the percentage of cells with positive immuno-reactivity in 1,000 cells (27). Two pathologists (T. Fuji-mura and S. Takahashi) independently evaluated thetissue sections, and the average LI was used. We definedpositive immunoreactivity as showing moderate or strongimmunoreactivity.

Statistical analysesCorrelations between age, pretreatment serum PSA

levels, mRNA levels, and PSA relapse were evaluated usingthe Wilcoxon test. Associations between PSA relapse,Gleason scores, clinical stage, and antiandrogen therapywere assessed using c2 tests. Correlations between mRNAexpression, PSA relapse, and clinical parameters werestatistically analyzed using logistic regression analyses.Appropriate variables indicating�2 F values were selectedby stepwise method. CSS curves were plotted using theKaplan–Meier method and verified using the log-ranktest. Cox-hazard proportional analysis was used for esti-mating the relationship between mRNA expression andCSS. A correlation between the mRNA expression and LIwas evaluated using Spearman rank-correlation coeffi-cient. JMP 9.0 software (SAS Institute) was used forstatistical analyses, and P < 0.05 was considered statisti-cally significant.

ResultsRelationship between PSA relapse andclinicopathologic data

We divided the patients into two groups according toPSA recurrence during the follow-up period: the PSArecurrence group (n ¼ 37) and the no-recurrence group(n ¼ 9; Table 1). No significant correlations were foundbetween PSA recurrence and clinicopathologic character-istics such as age, pretreatment serum PSA levels, Gleasonscores, clinical stage, EOD, or therapeutic regimens (Table1). However, all patients without PSA recurrence achieveda PSA nadir of <0.01, which is the measurement limit inour institute, a proportion significantly higher than thatin the PSA recurrence group (P < 0.0001). Time to PSAnadir following ADT among the PSA recurrence and no-recurrence groups was 424 � 286 days and 419 � 303days, respectively (P ¼ 0.31). The CSS rate was signifi-cantly worse in men with PSA recurrence than in the otherpatients (P ¼ 0.0045).

Fujimura et al.





Relative mRNA expression of AR, AR-related genes,stem cell–like markers, and prostate cancer-relatedgenes between the PSA recurrence and no-recurrencegroupsThe relative mRNA expression of AR, AR-related genes,

stem cell–like markers, and prostate cancer–relatedgenes is shown in Table 2. Expression of AR in bothcancer and stromal cells was significantly stronger inmen without PSA recurrence than in the other group(P ¼ 0.0026 and 0.013, respectively). Expression of APP,TRIM36, Klf4, c-Myc, and ERb in stromal cells frommen without PSA recurrence was significantly increased(P ¼ 0.0018, 0.047, 0.032, 0.044, and 0.032,respectively).

Comparison of area under the curve for clinicalparameters and gene expression in prostate needlebiopsy samples for predicting PSA recurrenceLogistic regression analyses for predicting PSA recur-

rence using age, serum PSA levels, Gleason scores, Tstage, N stage, and EOD showed relatively high areaunder the curves (AUCs; 0.83; Fig. 1A); however, 11patients (23%) were misclassified by discriminantanalyses. In contrast, the AUCs for Sox2, Her2, and CRPmRNA expression in cancer cells; AR and ERa mRNAexpression in stromal cells; and clinical parameterswere 1.0 in men with PSA recurrence. Only 2 patients(4%) were misclassified by discriminant analyses(Fig. 1B).

Correlation between CSS and gene expressionWe compared prognostic clinical parameters and gene

expression profiles using Cox proportional hazard anal-yses (Table 3). Cutoff values for age, serum PSA levels, Tstage, EOD, and relative mRNA expression for each genewere determined using receiver operating characteristic(ROC) curves. Decreased expression of Oct1, TRIM36,Sox2, and c-Myc in cancer cells and decreased expressionof AR, Klf4, and ERa in stromal cells were significantprognostic factors in univariate hazard analyses (HR: 2.6,2.9, 3.0, 2.7, 3.8, 4.1, and 2.5, respectively; P ¼ 0.031,0.0015, 0.045, 0.022, 0.0067, 0.0014, and 0.0034,respectively). Increased serum PSA levels (�335 ng/mL),increased Gleason scores (�8), and high EOD (�2) werealso correlated with CSS (HR: 2.7, 3.5, and 2.9; P¼ 0.027,0.046, and 0.016, respectively). Multivariate analyseswere not performed because of the following reasons. Toavoid multicollinearity problem, a correlation matrix wasconstructed among 10 prognostic parameters. The Spear-man rank-correlation coefficient test showed that relativeexpression of Oct1, TRIM36, c-Myc, and SOX2 in cancercells has significant correlation with one another. Relativeexpression of AR, ERa, and KLF4 in stromal cells has alsosignificant correlation with one another. Therefore, weused 10 factors to classify 3 prognostic groups because oflacking independent variables. Furthermore, the numberof cancer-specific death (n ¼ 24) events was small relativeto 10 prognostic factors.

Risk classification according to gene expression andcorrelation with CSS

According to the number of cancer-specific risk factorsdescribed above (Oct1, TRIM36, Sox2, and c-Myc expressionin cancer cells;AR,Klf4, and ERa expression in stromal cells;serum PSA levels � 335 ng/mL, Gleason scores � 8, andEOD � 2), we divided the patients into favorable-, inter-mediate-, andpoor-risk groups,whichhad0–3, 4–7, and8–10 risk factors, respectively. Significant differences wereobserved in CSS rates among the 3 groups (favorable vs.intermediate, P ¼ 0.0013; favorable vs. poor; P < 0.0001;and intermediate vs. poor, P ¼ 0.0059; Fig. 1C). Five-yearCSS rates for the favorable-, intermediate-, and poor-riskgroups were 90%, 32%, and 12%, respectively.

Validation study in an independent cohort of 30patients with bony metastatic prostate cancer

To validate the reproducibility, we performed logisticregression analyses, discriminant analyses for predictingPSA recurrence, and made new risk classification for CSSin an independent cohort of 30 patients with prostatecancer with bone metastasis. Logistic regression analysesfor predicting PSA recurrence using clinical findings showedrelatively high AUCs (0.95; Fig. 2A); however, 9 patients(30%) were misclassified by discriminant analyses. In con-trast, theAUCs for Sox2,Her2, andCRPmRNAexpression incancer cells; AR and ERamRNA expression in stromal cells;and clinical parameters were 1.0 in men with PSA recur-rence. Four patients (13%) were misclassified by discrim-inant analyses (Fig. 2B). According to the above classifica-tion, we also divided the patients into favorable-, interme-diate-, and poor-risk groups. Clinical significance wasshown in CSS rates among the 3 groups (favorable vs.intermediate, P¼ 0.11; favorable vs. poor; P¼ 0.0025; andintermediate vs. poor, P ¼ 0.033; Fig. 2C). Five-year CSSrates for the favorable-, intermediate-, and poor-risk groupswere 75%, 48%, and 0%, respectively.

Immunohistochemistry and correlation betweenimmunoreactivity andmRNA expressionof AR andKlf4

Among sevenprognostic genes,we selectedAR andKlf4 toinvestigate the relationship between immunoreactivity andmRNA expression of these genes. Figure 2D–I shows theresults of immunohistochemical analyses forAR andKlf4 inprostate cancer. Immunostaining ofARwas identified in thenuclei of cancer and stromal cells (Figure 2D–F), whereasthat of Klf4 was localized in both nuclei and cytoplasm incancer and stromal cells (Figure 2G–I). The LIs of AR incancer and stromal cells were 55� 41 (0–100) and 19� 29(0–100), respectively. The LIs of Klf4 in cancer and stromalcells were 50� 39 (0–100) and12� 17 (0–70), respectively(Supplementary Fig. S1A–S1D). Two pathologists indepen-dently evaluated the tissue sections and the average LI wasused. The SEs of LIsAR in cancer and stromal cells were 0.35and 0.06, respectively. The SEs of LIs of Klf4 in cancer andstromal cells were 1.25 and 0.99, respectively. The Spear-man rank-correlation coefficient (r) between immunore-active score andmRNAexpressionwas shownas follows:AR






in cancer and stromal cells (r ¼ 0.73, P < 0.0001 and r ¼0.56, P ¼ 0.0001, respectively), Klf4 in cancer and stromalcells (r ¼ 0.08, P ¼ 0.61 and r ¼ 0.56, P < 0.0001,respectively; Supplementary Fig. S1).

DiscussionThe most common initial systemic therapy for metastatic

prostate cancer isADT;however, thedurabilityofADTvaries.Amongpatientswithmetastaticprostate cancerwho receivedADT, the median survival time was 13 months for patientswith PSA nadirs of >4 ng/mL, 44 months for patients withPSA nadirs of 0.2–4 ng/mL, and 75months for patients with

PSA nadirs of�0.2 ng/mL (2). Several relevant nomogramsare used to predict progression-free survival and CSS inpatients with prostate cancer (28). These nomograms areestimatedaccording to age, serumPSA levels,Gleason scores,and clinical stage. Clinical parameters showed a degree ofpredictive power in the present study. These clinical para-meters do not sufficiently reflect cancer aggressiveness ordurability of treatments such as ADT, radiation, or chemo-therapy because clinical parameters do not accurately corre-late with prostate cancer cell behavior.

The durability of ADT may be influenced by AR and AR-related gene profiles in prostate cancer cells. ADT has an

Table 2. Comparison of relative mRNA expression between men with PSA recurrence and without PSArecurrence (mean � SD; n ¼ 46)

GenePSA recurrence(n ¼ 37)

Without PSArecurrence (n ¼ 9) P

Androgen-related genesAR Cancer 20 � 47 33 � 24 0.0026

Stroma 11 � 14 38 � 31 0.013Oct1 Cancer 4.7 � 8.0 6.6 � 12 0.67

Stroma 3.4 � 5.3 9.9 � 20 0.35FOXO1 Cancer 2.6 � 5.1 8.3 � 22 0.30

Stroma 3.1 � 6.4 0.78 � 1.1 0.29FOXA1 Cancer 0.64 � 1.9 0.30 � 0.35 0.66

Stroma 0.79 � 2.1 0.11 � 0.23 0.15FOXP1 Cancer 0.66 � 1.7 0.33 � 0.57 0.55

Stroma 0.94 � 4.2 0.081 � 0.15 0.37APP Cancer 10 � 21 21 � 32 0.14

Stroma 38 � 99 39 � 38 0.018ACSL3 Cancer 4.8 � 12 1.6 � 2.0 0.71

Stroma 3.4 � 7.2 8.3 � 16 0.09TRIM36 Cancer 0.78 � 1.6 0.92 � 1.6 0.25

Stroma 2.7 � 12 3.4 � 6.9 0.047Stem cell–like markersOct3/4 Cancer 9.8 � 13 18 � 36 0.59

Stroma 9.2 � 12 22 � 32 0.31Sox2 Cancer 0.58 � 1.0 0.61 � 1.1 0.58

Stroma 2.7 � 9.1 1.0 � 1.1 0.12Klf4 Cancer 4.5 � 8.5 5.5 � 8.7 0.57

Stroma 9.2 � 23 9.4 � 13 0.032c-MyC Cancer 8.3 � 12 15 � 15 0.14

Stroma 17 � 37 26 � 40 0.044Prostate cancer–related genesCRP Cancer 4.8 � 9.7 2.6 � 3.0 0.53

Stroma 8.3 � 33 3.9 � 5.9 0.54Her2 Cancer 1.2 � 2.8 1.7 � 4.4 0.83

Stroma 2.8 � 6.1 2.1 � 2.5 0.53ERb Cancer 3.2 � 6.5 8.0 � 15 0.28

Stroma 6.7 � 17 8.0 � 9.3 0.032Klf5 Cancer 7.2 � 11 6.7 � 8.2 0.86

Stroma 25 � 68 12 � 11 0.41ERa Cancer 0.18 � 0.44 0.18 � 0.38 0.98

Stroma 0.18 � 0.27 1.9 � 3.7 0.16

Fujimura et al.





inhibitory effect on prostate cancer; however, ADT nega-tively selects stem cell–like prostate cancer cells, an impor-tant component of CRPC (3, 29, 30). Germann and col-leagues found that castration induced the expression of 4essential transcription factors required for reprogramming,self-renewal, and pluripotency in differentiated somaticcells, namely Oct4, Sox2, Klf4, and NANOG (3). Therefore,before ADT is prescribed, an accurate prediction is criticalfor treatment decisions, which range from ADT alone toinitial combination therapy with other therapeutic agentssuch as docetaxel, cabazitaxel, zoledronic acid, and deno-sumab (31), that are tightly linked to prognoses. In thisrespect, we determined molecular indicators of ADT stabil-ity and cancer progression in patients with metastatic pros-tate cancer by evaluating AR, AR-related genes, stem cell–related genes, and prostate cancer–related genes.

Molecular diagnosis using immunohistochemistry,FISH, and DNA, RNA, or miRNA analyses is discussedin a recent article (32–38). The studies in which gene setsof stem cell–like cells, micro-RNA, or cell-cycle progres-sion markers reflect more aggressive disease are limited tolocalized prostate cancer (33–35). Although severalreports have shown that certain gene expression profilesin primary prostate cancers correlate with poor progno-sis, no consensus has been reached regarding specificprognostic markers (32). In addition, controversial datamay reflect contamination of samples by other tissueelements such as aggressive tumors, normal prostateepithelium, and normal stromal components. Therefore,LCM techniques may influence the accuracy of themolecular diagnosis of prostate cancer. A recent studyusing frozen samples showed that low PSA/HK3 mRNAexpression in prostate cancer was associated withincreased risk of biochemical recurrence in patients withintermediate preoperative serum PSA levels (2–10ng/mL; ref. 36). Fresh frozen samples are preferable forRNA analysis of prostate biopsy samples. However, arefined LCM technique efficiently provided mRNA andmiRNA from formalin-fixed biopsy samples (37, 38). Inthis study, we showed that the durability of ADT is acancer-specific prognostic factor in patients with meta-static prostate cancer using paraffin-embedded needlebiopsy samples.

Gene expression profiles have been successfully used todefine breast cancer subclasses with different biologic beha-viors and responses to therapy (4). Gene expression-baseddiagnostic tests are currently in clinical use for assessing therisk of recurrence and for predicting the benefits of adjuvantchemotherapy in patients with localized ER-positive andlymph node-negative breast cancers (39). In contrast tobreast cancers, in which the status of estrogen receptors inprimary tumors is commonly used tomake therapeutic andprognostic decisions, the status of AR protein expressiondoes not seem to be as useful as in prostate cancer (32). Onepossible explanation for this is that AR expression is het-erogeneous and changes over time (31). Therefore, mea-surement of the expression of selected AR downstreamtargets can provide information on the individual function-al status of ARs in prostate cancer cells. We previouslydefined an AR transcriptional network in prostate cancercells using ChIP–chip and CAGE assays (18) and alsoanalyzed the AR-related genes APP, Oct1, and FOXP1 inprostate cancer (7, 11, 12). In these experiments, APP andOct1 were correlated with prostate cancer aggressivenessand both were considered therapeutic targets (7, 12). SomeFOX proteins that are involved in cell growth and differen-tiation as well as in embryogenesis and longevity are knownto be AR-related genes (8). In the current study, we showedrelationships between biochemical recurrence andAR, ERa,Sox2,CRP, andHer2 expression andbetweenprognoses andexpression of AR,Oct1, TRIM36, Sox2, Klf4, c-Myc, and ERa.Functional analysis of TRIM36 proteins, a subfamily ofRING type E3 ubiquitin ligases, remains unresolved, andfurther investigation of the TRIM family is required in

Figure 1. Logistic regression analyses for predicting PSA recurrence afterandrogen deprivation therapy and CSS classified by the number of riskfactors in patients with bony metastatic prostate cancer (n ¼ 46). A,correlations of age, serum PSA levels, Gleason score (GS), T stage,N stage, and EOD with PSA recurrence after ADT. The AUC for PSArecurrence was 0.83. B, PSA recurrence after ADT can be predicted byanalyzing mRNA expression in biopsy samples using LCM. Combinedexpression ofSox2,CRP, andHer2 in cancer cells,AR andERa in stromalcells, and previously described clinical parameters strongly predictedPSA recurrence (AUC ¼ 1.0). C, risk factors included decreasedexpression ofOct1, TRIM36,Sox2, and c-Myc expression in cancer cells,AR, Klf4, and ERa in stromal cells, increased serum PSA levels (�335ng/mL), high Gleason scores (�8), and high EOD (�2). Patients in thefavorable- (n¼ 15), intermediate- (n¼ 18), and poor- (n¼ 13) risk groupshad 0–3, 4–7, and 8–10 risk factors, respectively. CSS rates differedsignificantly among the 3 groups (favorable vs. intermediate, P¼ 0.0013;favorable vs. poor, P < 0.0001; and intermediate vs.poor, P ¼ 0.0059).






association with prostate cancer. The correlation betweensuppressed Oct1, Sox2, and c-Myc expression in cancer cellsand poor prognosis may reflect the sensitivity of these genesto ADT. Recent studies determined the localization andfunction of Klf4 in prostate cancer cells (40, 41). In thesestudies, Klf4was downregulated in prostate cancer cell linesand metastatic prostate cancer tissue (41), and RNA activa-

tor-mediated overexpression of Klf4 inhibited prostate can-cer cell growth/survival and arrested cell-cycle progression(41). Together with the present data, Klf4 appears to exert apowerful inhibitory effect in prostate cancer.

Recent therapeutic strategies for advanced andmetastaticprostate cancer include ADT combined with antiandrogenagents such as bicalutamide or flutamide, secondary

Table 3. Cox proportional hazard regression analysis of CSS in metastatic prostate cancer (n ¼ 46)

Gene expression(low vs. high) Cutoff Foci HR 95% index P

AR 0.27 Cancer 2.0e-6 1.3–1.3 0.07814 Stroma 3.8 1.4–13 0.0067

Oct1 1.5 Cancer 2.6 1.1–6.4 0.0311.1 Stroma 1.8 0.78–4.3 0.16

FOXO1 0.027 Cancer 2.5 0.95–5.9 0.0638.3 Stroma 1.1 0.32–6.9 0.89

FOXA1 1.3 Cancer 0.27 0.072–1.8 0.151.3 Stroma 0.42 0.14–1.8 0.22

FOXP1 6.4 Cancer 0.094 0.013–1.8 0.0991.6 Stroma 0.35 0.064–6.6 0.39

APP 15 Cancer 1.8 0.59–7.6 0.332.3 Stroma 0.69 0.24–1.7 0.45

ACSL3 0.78 Cancer 1.5 0.65–3.8 0.334 Stroma 1.7 0.48–11 0.45

TRIM36 0.13 Cancer 2.9 1.2–7.3 0.0150.93 Stroma 2.6 0.99–9.3 0.053

Oct3/4 16 Cancer 2.3 0.66–14 0.224.5 Stroma 1.9 0.82–4.9 0.13

Sox2 0.53 Cancer 3.0 1.0–13 0.0450.46 Stroma 1.6 0.65–4.5 0.32

Klf4 0.65 Cancer 1.3 0.56–3.1 0.531.6 Stroma 4.1 1.7–11 0.0014

c-Myc 3.8 Cancer 2.7 1.2–6.8 0.02224 Stroma 0.79 0.31–2.4 0.66

PSA 0.21 Cancer 1.8 0.71–4.2 0.210.40 Stroma 2.0 0.86–5.1 0.10

CRP 0.20 Cancer 1.4 0.46–3.5 0.520.74 Stroma 1.8 0.77–4.3 0.18

Her2 5.9 Cancer 2.1 0.44–38 0.4114 Stroma 1.2 0.26–22 0.82

ERb 0.63 Cancer 2.3 0.97–5.4 0.0590.60 Stroma 1.9 0.72–4.5 0.19

Klf5 2.1 Cancer 2.2 0.87–5.2 0.0911.5 Stroma 1.0 0.34–2.6 0.95

ERa 0.042 Cancer 0.61 0.27–1.4 0.260 Stroma 2.5 1.1–6.1 0.0034

Age �85 vs. <85 0.49 0.17–2.1 0.30Serum PSA �335 vs. <335 2.7 1.1–6.5 0.027Gleason score �8 vs. �7 3.5 1.0–22 0.046T stage �4 vs. �3 1.2 0.47–2.9 0.66N stage 1 vs. 0 1.5 0.62–3.9 0.38EOD �2 vs. �1 2.9 1.2–7.0 0.016

NOTE: The appropriate cutoff value of each gene was decided by ROC curve including age, serum PSA levels, T stage, and EOD.

Fujimura et al.





hormonal manipulation using adrenal testosterone inhibi-tors, low-dose diethylstilbestrol therapy, steroid therapy,somatostatin analog therapy, and chemotherapy (31). Inthe present study, we identified patients in whom thedurability of ADT was poor or moderate on the basis ofmolecular diagnoses. We also validated the models of bothPSA reccurence and CSS in an independent metastaticprostate cancer cohort. Further investigations in relativelarge cohort prove the reproducibility, molecular diagnosisof needle biopsy samples may be generalized for strategicinterventions. Immunohistochemical analysis using biopsysamples may be easier to perform in clinical settings thanestimating mRNA expression using LCM. Therefore,

patients predicted tohaveCRPC in the early phaseof diseasemay be given other therapeutic interventions before theprogression of aggressive phenotypes.

This study found that genes expressed in stromal cells,such as AR, Klf4, and ERa, were correlated with cancerprogression. Expression of these genes may contribute tothe stromal–epithelial interactions in prostate cancer pro-gression. The stromal cells in the prostatic tissue consist ofmyofibroblasts, fibroblasts, and smoothmuscle cells withina connective tissue matrix. Numerous growth factors pro-duced by the stromal cells, including transforming growthfactors, platelet-derived growth factors, fibroblast growthfactors, and EGFs, are crucial for prostate cancer growth (42,

Figure 2. Logistic regressionanalyses for predicting PSArecurrence after ADT and CSSclassified by the number of riskfactors in an independent cohort ofpatients with bony metastaticprostate cancer (n ¼ 30) andimmunohistochemistry using anti-AR and Klf4 antibodies. A,correlations of age, serum PSAlevels, Gleason score (GS), T stage,N stage, and EOD with PSArecurrence after ADT. The AUC forPSA recurrence was 0.95. B,combined expression of Sox2,CRP, and Her2 in cancer cells,AR and ERa in stromal cells, andpreviously described clinicalparameters strongly predictedPSArecurrence (AUC ¼ 1.0). C, riskfactors included decreasedexpression ofOct1, TRIM36, Sox2,and c-Myc expression in cancercells, AR, Klf4, and ERa in stromalcells, increased serum PSA levels(�1228 ng/mL), high Gleasonscores (�8), and high EOD (�2).Patients in the favorable- (n ¼ 7),intermediate- (n ¼ 20), and poor-(n ¼ 3) risk groups had 0–3, 4–7,and 8–10 risk factors, respectively.CSS rates differed between the 3groups (favorable vs. intermediate,P ¼ 0.11; favorable vs. poor,P ¼ 0.0025; and intermediate vs.poor, P ¼ 0.033). D and E,immunohistochemical staining forAR (D–F) and Klf4 (G–I) antibodiesin prostate cancer. Strong (D),moderate (E), and weak (F) stainingof ARwas identified in the nuclei ofcancer and stromal cells. Strong(G), moderate (H), and weak (I)immunoreactivity for Klf4 wasobserved in both nuclei andcytoplasm of cancer cells.Moderate (G and H) and weak (I)immunostaining for Klf4 was seenin nuclei and cytoplasm of stromalcells.






43). These growth factors change microenviroment aroundthe stromal cells (43). Clark and colleagues created abioengineered microenviroment using tissue recombina-tion that involved mixing of stromal and epithelial cellpopulations (43). In coculture with cancer-associated fibro-blasts (CAF), but not nonmalignant prostatic fibroblasts,BPH-1 cells showed a more aggressive phenotype withincreased motility and a more direct way of cell migration(43). In addition, the secretion of growth factors is influ-enced by steroid hormones because stromal cells expressAR, ERa, and ERb that play significant roles in prostaticepithelial cell growth (42). For example, Risbridger andcolleagues investigated prostatic tissue recombinants estab-lished from wild-type ERa, and its knockout (KO) mice(44). Presence of squamousmetaplasia induced by estrogenwas evaluated in each combination of tissue recombinants,such as wt-stroma (S) þ wt-epithelia (E), aERKO-S þaERKO-E, wt-S þ aERKO-E, and aERKO-S þ wt-E. Squa-mousmetaplasia induced by estrogenwas found only in thewt-S þ wt-E group, which suggested that both stromal andepithelial ERa are required to achieve full response toestrogen in terms of developing squamous metaplasia(44). Stromal AR also influences oncogenic epithelium cellgrowth (45). In Lai and colleagues (45), the authors estab-lished an animal model with AR deletion in stromal fibro-muscular cells in PTEN deleted from chromosome 10(Pten)þ/�mouse. Prostatic intraepithelial neoplasia (PIN)was developed via changing tumor microenvironment,such as the alteration of angiogenesis and immune cellsinfiltration in the model (45). Moreover, AR degradationenhancer, ASC-J9, suppresses PIN development via stromalAR degradation (45). These findings suggest that modula-tors of stromal–epithelial interactions may be useful asfuture therapeutic agents.

In conclusion, we demonstrated a predictive model ofPSA recurrence and CSS in patients with prostate cancerwith bone metastasis by measuring the expression of pros-

tate cancer–related genes in needle biopsy samples. Inter-mediate- or poor-risk patients with bonymetastatic prostatecancer would be candidates for further clinical trials inaddition to ADT.

Disclosure of Potential Conflicts of InterestNo potential conflicts of interest were disclosed.

Authors' ContributionsConception and design: T. Fujimura, S. Takahashi, T. Urano, Y. Yamada,S. Inoue, Y. HommaDevelopment of methodology: T. Fujimura, S. Takahashi, T. Sugihara,Y. YamadaAcquisitionofdata (provided animals, acquired andmanagedpatients,provided facilities, etc.): T. Fujimura, T. SugiharaAnalysis and interpretation of data (e.g., statistical analysis, biosta-tistics, computational analysis): T. Fujimura, S. Takahashi, K. Takayama,S. InoueWriting, review, and/or revision of the manuscript: T. Fujimura,S. Takahashi, T. Urano, K. Takayama, J. Kumagai, S. Inoue, Y. HommaAdministrative, technical, or material support (i.e., reporting ororganizing data, constructing databases): T. Fujimura, K. Takayama,D. Obinata, H. Kume, Y. Ouchi, Y. HommaStudy supervision: S. Takahashi, J. Kumagai, S. Inoue, Y. Homma

AcknowledgmentsThe authors thank Tomoko Yamanaka and Yasuho Saito for their tech-

nical assistance and Takuhiro Yamagichi and Tempei Miyaji for their kindadvice regarding statistical analysis.

Grant SupportThis work was supported by Grants of the Cell Innovation Program from

theMinistry of Education, Culture, Sports, Science&Technology, Japan (to S.Inoue), by grants from the Japan Society for the Promotion of Science (to T.Fujimura, S. Takahashi, and S. Inoue), by Grants-in-Aid from the MHLW,Japan (to S. Inoue), and by the Advanced Research for Medical ProductsMining Program in Health Sciences, National Institute of Biomedical Inno-vation, Japan (to S. Inoue).

The costs of publication of this article were defrayed in part by thepayment of page charges. This article must therefore be hereby markedadvertisement in accordance with 18 U.S.C. Section 1734 solely to indicatethis fact.

Received April 22, 2013; revised June 2, 2014; accepted June 2, 2014;published OnlineFirst July 1, 2014.

References1. Huggins C, Hodges CV. Studies on prostatic cancer. The effect of

castration, of estrogen and of androgen injection on serum phospha-tases in metastatic carcinoma of the prostate. Cancer Res 1941;1:293–7.

2. Hussain M, Tangen CM, Higano C, Scelhammer PF, Faulkner J,Crawford ED, et al. Absolute prostate-specific antigen value afterandrogen deprivation is a strong independent predictor of survival innewmetastatic prostate cancer: data from southwest oncology grouptrial 9346 (INT-0162). J Clin Oncol 2006;24:3984–90.

3. Germann M, Wetterwald A, Guzman-Ramirez N, van der Pluijm G,Culig Z, Cecchini MG, et al. Stem-like cells with luminal progenitorphenotype survive castration in human prostate cancer. Stem Cells2012;30:1076–86.

4. Carlson RW, Allred DC, Anderson BO, Burstein HJ, Edge SB, FarrarWB, et al. Metastatic breast cacer, version 1.2012: featured updates tothe NCCN guidelines. J Natl Canc Netw 2012;10:821–9.

5. Tammela T. Endocrine treatment of prostate cancer. J Steroid Bio-chem Mol Biol 2004;92:287–95.

6. Culig Z, Steiner H, Bartsch G, Hobisch A. Mechanism of endocrinetherapy-responsive and –unresponsive prostate tumors. Endocr RelatCancer 2005;12:229–44.

7. TakayamaK, TsutusumiS,Suzuki T,Horie-InoueK, IkedaK,KaneshiroK, et al. Amyloid precursor protein is a primary androgen target genethat promotes prostate cancer growth. Cancer Res 2009;69:137–42.

8. Katoh M, Igarashi M, Fukuda H, Nakagama H, Katoh M. Cancergenetic and genomics of human FOX genes. Cancer Lett 2013;328:198–206.

9. Zhang H, Fang J, Yao D, Wu Y, Ip C, Dong Y. Activation of FOXO1 iscritical for anticancer effect methylseleninic acid in prostate cancercells. Prostate 2010;70:1265–73.

10. Sahu B, Laakso M, Ovaska K, Mirtti T, Lundin J, Ranniko A, et al. Dualrole of FoxA1 in androgen receptor binding to chromatin, androgensignaling and prostate cancer. EMBO J 2011;30:3962–76.

11. Takayama K, Horie-Inoue K, Ikeda K, Urano T, Murakami K, Haya-shizaki Y, et al. FOXP1 is an androgen-responsive transcription factorthat negatively regulates androgen receptor signaling in prostatecancer cells. Biochem Biophys Res Commun 2008;374:388–93.

12. Obinata D, Takayama K, Urano T, Murata T, Kumagai J, Fujimura T,et al. Oct1 regulates cell growth of LNCaP cells and is a prognosticfactor for prostate cancer. Int J Cancer 2012;130:1021–8.

13. Kawashima H, Nakatani T. Involvement of estrogen receptors inprostatic tissues. Int J Urol 2012;19:512–22.

Fujimura et al.





14. Harkonen P, Makela SI. Role of estrogens in development of prostatecancer. J Steroid Biochem Mol Biol 2004;92:297–305.

15. Niu Y, Chang TM, Yeh S, Ma WL, Wang YZ, Chang C. Differentialandrogen receptor signals in different cells explain why androgen-dep-rivation therapy of prostate cancer fails. Oncogene 2010;29:3593–604.

16. Soloway MS, Hardeman SW, Hickey D, Raymond J, Todd B, SolowayS, et al. Stratification of patientswithmetastatic prostate cancer basedon disease on initial bone scan. Cancer 1988;61:195–202.

17. Mottet N, Bellmunt J, Bolla M, Joniau S, Mason M, Matveev V, et al.EAU guidelines on prostate cancer. Part II: treatment of advanced,relapsing, and castration-resistant prostate cancer. Eur Urol 2011;59:572–83.

18. Takayama K, Tsutsumi S, Katayama S, Okayama T, Horie-Inoue K,Ikeda K, et al. Integration of cap analysis of gene expression andchromatin immunoprecipitation analysis on array reveals genome-wide androgen receptor signaling in prostate cancer cells. Oncogene2011;30:619–30.

19. Takahashi K, Yamanaka S. Induction of pluripotent stem cells frommouse embryonic and adult fibroblast cultures by defined factors. Cell2006;126:663–76.

20. Schoenhals M, Kassambara A, De Vos J, Hose D, Moreaux J, Klein B.Embryonic stem cell markers expression in cancers. BiochemBiophysRes Commun 2009;380:157–62.

21. Saito K, Kihara K. Role of C-reactive protein in urological cancers:a usuful biomarker for predicting outcomes. Int J Urol 2013;20:161–71.

22. Tobiume M, Yamada Y, Nakamura K, Aoki S, Zennami K, Kato Y, et al.Significant prognostic factor of immunohistochemical HER-2 expres-sion using initial prostate biopsy specimenswithM1b prostate cancer.Prostate 2011;71:385–93.

23. Nakajima Y, Akaogi K, Suzuki T, Osakabe A, Yamaguchi C, SunaharaN, et al. Estrogen regulates tumor growth through a nonclassicalpathway that includes the transcription factors ERb and KLF5. SciSignal 2011;4:ra 22.

24. de Matteris A, Guidi A, Di Paolo B, Franco G, Revoltella RP.Endothelin-1 in human prostatic carcinoma treated with andro-gen withdrawal: an immunohistochemical study. Cancer 2001;91:1933–99.

25. Verma V, Huang B, Kallingappa PK, Oback B. Dual kinase inhibitionpromotes pluripotency in finite bovine embryonic cell lines. StemCellsDev 2013;1:1728–42.

26. Murata T, Takayama K, Urano T, Fujimura T, Ashikari D, Obinata D,et al. 14-3-3z, a novel androgen-responsive gene, is upregulated inprostate cancer and promotes prostate cancer cell proliferation andsurvival. Clin Cancer Res 2012;18:5617–27.

27. Takayama K, Horie-Inoue K, Katayama K, Suzuki T, Tsutsumi S, IkedaK, et al. Androgen-responsive long noncoding RNA CTB1-AS pro-motes prostate cancer. EMBO J 2013;32:1665–80.

28. Shariat SF, Kattan MW, Vickers AJ, Karakiewicz PI, Scardino PT.Critical review of prostate cancer predictive tools. Future Oncol 2009;5:1555–84.

29. Tang Y, Hamburger AW, Wang L, Khan MA, Hussain A. Androgendeprivation and stem cell markers in prostate cancers. Int J Clin ExpPathol 2010;3:128–38.

30. Pfeiffer MJ, Smit FP, Sedelaar JP, Schalken JA. Steroidogenicenzymes and stem cell markers are upregulated during androgendeprivation in prostate cancer. Mol Med 2011;17:657–64.

31. Beltran H, Beer TM, Carducci MA, de Bono J, GleaveM, HussainM, atal. New therapies for castration-resistant prostate cancer: efficacy andsafety. Eur Urol 2011;60:279–90.

32. Choudhury AD, Eeles R, Freedland SJ, Isaacs WB, Pomerantz MM,Schalken JA, et al. The role of genetic markers in the management ofprostate cancer. Eur Urol 2012;62:577–87.

33. Market EK,MizunoH,VazquezA, LevineAJ.Molecular classification ofprostate cancer using curated expression signatures. Proc Natl AcadSci U S A 2011;108:21276–81.

34. Martnes-Uzunova ES, Jalava SE, Dits NF, van Leenders GJ, TrapmanJ, Bangma CH, et al. Diagnostic and prognostic signatures from thesmall non-coding RNA transcriptome in prostate cancer. Oncogene2012;31:978–91.

35. Cuzick J, Swanson GP, Fisher G, Brothman AR, Berney DM, ReidJE, et al. Prognostic value of an RNA expression sig-nature derived from cell cycle proliferation genes in patients withprostate cancer: a retrospective study. Lancet Oncol 2011;12:245–55.

36. Sterbis JR, Gao C, Furusato B, Chen Y, Shaheduzzaman S,Ravindranath L, et al. Higher expression of the androgen-regulatedgene PSA/HK3 mRNA in prostate cancer tissues predicts bio-chemical recurrence-free survival. Clin Cancer Res 2008;14:758–63.

37. Nonn L, Vaishnav A, Gallagher L, Gann PH. mRNA and micro-RNAexpression analysis in laser-capture microdissected prostate biop-sies: valuable tool for risk assmessment and prevention traials. ExpMol Pathol 2010;88:45–51

38. Rogerson L, Darby S, Jabbar T, Mathers ME, Leung HY, Robson CN,et al. Application of transcript profiling in formalin-fixed paraffin-embedded diagnostic prostate cancer needle biopsies. BJU Int 2008;102:364–70.

39. Oakman C, Santarpia L, Di Leo A. Breast cancer assessmenttools and optimizing adjuvant therapy. Nat Rev Clin Oncol 2010;7:725–32.

40. Magnen CL, Bubendorf L, Ruiz C, Zlobec I, Bachmann A, Heberer M,et al. Klf4 transcription factor is expressed in the cytoplasm of prostatecancer cells. Eur J Cancer 2013;49:955–63.

41. Wang J, Place RF, Huang V, Wang X, Noonan EJ, Maryar CE, et al.Prognostic value and function of KLF4 in prostate cancer: RNAs andvector-mediated overexpression identify KLF4 as an inhibitor of tumorcell growth and migration. Cancer Res 2010;70:10182–91.

42. Berry PA, Maitland NJ, Collins AT. Androgen receptor signaling inprostate: effects of stromal factors on normal and cancer stem cells.Mol Cell Endocrinol 2008;288:30–7.

43. Clark AK, Taubenberger AV, Taylor RA, Niranjan B, Chea ZY,Zotenko E, et al. A bioengineered microenvironment to quantita-tively measure the tumorigenic properties of cancer-associatedfibroblasts in human prostate cancer. Biomaterials 2013;34:4777–85.

44. Risbridger G, Wang H, Young P, Kurita T, Wang YZ, Lubahn D, et al.Evidence that epithelial and mesenchymal estrogen receptor-alphamediates effects of estrogen on prostatic epithelium. Dev Biol 2001;229:432–42.

45. Lai KP, Yamashita S, Huang CK, Yeh S, Chang C. Loss of stromalandrogen receptor leads to suppressed prostate tumorigenesis viamodulation of pro-inflammatory cytokines/chemokines. EMBO MolMed 2012;4:791–807.






2014;20:4625-4635. Published OnlineFirst July 1, 2014.Clin Cancer Res Tetsuya Fujimura, Satoru Takahashi, Tomohiko Urano, et al. CancerCancer-Specific Survival in Patients with Metastatic ProstateStem Cell Markers to Predict Cancer Progression and Expression of Androgen and Estrogen Signaling Components and

Updated version

10.1158/1078-0432.CCR-13-1105doi:

Access the most recent version of this article at:

Material

Supplementary

http://clincancerres.aacrjournals.org/content/suppl/2014/07/14/1078-0432.CCR-13-1105.DC1

Access the most recent supplemental material at:

Cited articles

http://clincancerres.aacrjournals.org/content/20/17/4625.full#ref-list-1

This article cites 45 articles, 12 of which you can access for free at:

Citing articles

http://clincancerres.aacrjournals.org/content/20/17/4625.full#related-urls

This article has been cited by 1 HighWire-hosted articles. Access the articles at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected]

To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at

Permissions

Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://clincancerres.aacrjournals.org/content/20/17/4625To request permission to re-use all or part of this article, use this link



http://clincancerres.aacrjournals.org/lookup/doi/10.1158/1078-0432.CCR-13-1105

http://clincancerres.aacrjournals.org/content/suppl/2014/07/14/1078-0432.CCR-13-1105.DC1

http://clincancerres.aacrjournals.org/content/20/17/4625.full#ref-list-1

http://clincancerres.aacrjournals.org/content/20/17/4625.full#related-urls

http://clincancerres.aacrjournals.org/cgi/alerts

mailto:[email protected]

http://clincancerres.aacrjournals.org/content/20/17/4625


Documents

ExpressionofAndrogenandEstrogenSignalingComponents and ...containing 25 mL of ISOGEN reagent (Nippon Gene) and tissues were stored at 80 C before RNA isolation. Quantitative reverse-transcription