steroidogenic enzyme akr1c3 is a novel androgen receptor … · human cancer biology steroidogenic...

Human Cancer Biology

Steroidogenic Enzyme AKR1C3 Is a Novel AndrogenReceptor-Selective Coactivator that Promotes ProstateCancer Growth

Muralimohan Yepuru1, Zhongzhi Wu1, Anand Kulkarni2, Feng Yin1, Christina M. Barrett1, Juhyun Kim1,Mitchell S. Steiner1, Duane D. Miller1, James T. Dalton1, and Ramesh Narayanan1

AbstractPurpose: Castration-resistant prostate cancer (CRPC) may occur by several mechanisms including the

upregulation of androgen receptor (AR), coactivators, and steroidogenic enzymes, including aldo keto

reductase 1C3 (AKR1C3). AKR1C3 converts weaker 17-keto androgenic precursors to more potent 17-

hydroxy androgens and is consistently the major upregulated gene in CRPC. The studies in the manuscript

were undertaken to examine the role of AKR1C3 in AR function and CRPC.

Experimental Design: LNCaP cells stably transfected with AKR1C3 and VCaP cells endogenously

expressing AKR1C3 were used to understand the effect of AKR1C3 on prostate cancer cell and tumor

growth in nude mice. Chromatin immunoprecipitation, confocal microscopy, and co-immunoprecipita-

tion studies were used to understand the recruitment of AKR1C3, intracellular localization of AKR1C3 and

its interaction with AR in cells, tumor xenograft, and in Gleason sum 7 CRPC tissues. Cells were transiently

transfected for AR transactivation. Novel small-molecule AKR1C3-selective inhibitors were synthesized and

characterized in androgen-dependent prostate cancer and CRPC models.

Results: We identified unique AR-selective coactivator- and prostate cancer growth-promoting roles for

AKR1C3. AKR1C3 overexpression promotes the growth of both androgen-dependent prostate cancer and

CRPC xenografts, with concomitant reactivation of androgen signaling. AKR1C3 interacted with AR in

prostate cancer cells, xenografts, and in human CRPC samples and was recruited to the promoter of an

androgen-responsive gene. The coactivator and growth-promoting functions of AKR1C3 were inhibited by

an AKR1C3-selective competitive inhibitor.

Conclusions: AKR1C3 is a novel AR-selective enzymatic coactivator and may represent the first of more

than 200 knownnuclear hormone receptor coactivators that can be pharmacologically targeted.Clin Cancer

Res; 19(20); 5613–25. �2013 AACR.

IntroductionProstate cancer is themost frequently diagnosed cancer in

men with more than 200,000 new prostate cancer casesestimated for 2012 (American Cancer Society, Facts andFigures). The mainstay of therapy for advanced prostatecancer is the reduction of peripheral androgens to castratelevels. Androgen deprivation therapy creates an androgen-

deficient state in which prostate cancer evolves into castra-tion-resistant prostate cancer (CRPC) and tumor progres-sion occurs (1). Newer therapies that target androgenmetabolizing enzymes and/or androgen receptor (AR) haveshown clinical efficacy, indicating the continued impor-tance of the androgen signaling axis in advanced prostatecancer (2).

A variety of mechanisms are thought to contribute to theemergence of the androgen hypersensitivity that is observedin CRPC. Overexpression of AR and coactivators drives thegrowth and metastasis of CRPC even in response to lowlevels of androgens (3–5). Despite having castrate levels ofserum testosterone, tumor samples obtained from menwith CRPC have only a 60% reduction in intratumoraldihydrotestosterone (DHT). This suggests that intratumoralconversion of weak adrenal androgens or androgen pre-cursors produced de novo in the tumor fuel the growth ofCRPC (6).

Androgen biosynthesis enzyme inhibitors have beenused in the treatment of advanced prostate cancer. Drugssuch as ketoconazole and abiraterone acetate to treat

Authors' Affiliations: 1Preclinical Research and Development, GTx Inc.;and 2Department of Pathology, University of Tennessee Health ScienceCenter, Memphis, Tennessee

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

J.T. Dalton and R. Narayanan share senior authorship.

Corresponding Authors: Ramesh Narayanan, Preclinical Research andDevelopment, GTx Inc., 3 North Dunlap, Memphis, TN 38163. Phone: 901-523-9700; Fax: 901-523-9772; E-mail: [email protected]; andJames T. Dalton, E-mail: [email protected]

doi: 10.1158/1078-0432.CCR-13-1151

�2013 American Association for Cancer Research.

ClinicalCancer

Research

www.aacrjournals.org 5613

on October 17, 2020. © 2013 American Association for Cancer Research. clincancerres.aacrjournals.org Downloaded from

Published OnlineFirst August 30, 2013; DOI: 10.1158/1078-0432.CCR-13-1151

http://clincancerres.aacrjournals.org/

advanced prostate cancer have emphasized the importanceof conversion of androgen precursors either de novo or fromadrenal androgens in maintaining CRPC growth (7).Because these inhibitors inhibit early steps in steroidogen-esis, they not only inhibit androgen biosynthesis, but alsoaffect the levels of other physiologically relevant steroidssuch as glucocorticoids and mineralocorticoids.

Aldo keto reductase 1C3 (AKR1C3 aka 17b-HSD5), anenzyme with 17b-hydroxysteroid dehydrogenase (17b-HSD) activity and important for the biosynthesis of testos-terone and estradiol, is speculated to play a pivotal rolein the emergence of CRPC (8–10). AKR1C3, downstream inthe steroidogenic enzyme cascade, facilitates the conversionof weak androgens androstenedione (A’dione) and 5a-androstanedione (5a-dione) to the more active androgens,testosterone, and DHT, respectively (Fig. 1A). Gene expres-sion studies showed a significant increase in AKR1C3expression in CRPC over normal prostate (9, 10). Althoughthe overexpressionof AKR1C3 inCRPC iswell documented,its role in prostate cancer progression remains unclear, andAKR1C3 is thought to be limited to its role as an enzyme inandrogenbiosynthesis. The interactions ofAKR1C3withARand its role in AR function have not been explored. IfAKR1C3 plays a role in CRPC progression, tumor confinedexpression will make it an ideal tissue-selective therapeutictarget for prostate cancer.

The AKR1C family comprises four members (1C1–1C4;ref. 11, 12). Although all the four isoforms are expressed inliver, their expression varies in other tissues. AKR1C3 isexpressed in many endocrine organs such as prostate,adrenals, breast, and uterus. The four isoforms have highsequence homology sharing more than 83% identity, andtheir crystal structures show conservation of key residues.

Considering that all isoforms, other than AKR1C3, convertpotent to weak androgens, it is important to develop inhi-bitors of AKR1C3 that donot cross reactwith other isoforms(11).

The functions of AR are highly dependent on its coacti-vators.More than 200 coactivators have been identified andmany of them are overexpressed in prostate cancer (13).Several groups successfully used peptides and siRNAs toblock coactivator interactionwith AR, thereby restricting ARfunction and subsequently prostate cancer growth (14).However, the lack of known binding pockets nullifies thecoactivators’ potential to be small-molecule drug targets.

In this study, we show that AKR1C3 is an AR-selectivecoactivator that facilitates the growth of both androgen-dependent prostate cancer and CRPC. AKR1C3 interactswith AR in cells and in advanced prostate cancer samplesand gets recruited to the promoter of anAR-responsive gene.We further show that a novel small molecule that compet-itively binds to the AKR1C3 substrate-binding pocket caninhibit its role as a biosynthetic enzyme and AR coactivator,suggesting that AKR1C3 inhibitors may have utility intargeted therapy of both androgen-dependent prostate can-cer and CRPC.

Materials and MethodsReagents

AR antibody, PG-21, were obtained from Millipore, ARantibody, AR N-20, was obtained from Santa Cruz biotech-nology, AKR1C3 mouse monoclonal antibody wasobtained from Sigma, and AKR1C3 rabbit polycloncalantibody was obtained from Life Technologies. Actin anti-body was procured from Chemicon international. HumanPSA ELISA was procured from R&D systems. Accell siRNAwas procured from Dharmacon. All other reagents usedwere analytical grade from Fisher.

Cell cultureAll cells were obtained from American Type Culture Col-

lection and were grown according to the instructions pro-vided. The cell lines were authenticated by the provider andwere cultured for less than 6months after resuscitation in thelaboratory. For the chromatin immunoprecipitation (ChIP)and co-immunoprecipitation(co-IP) assays, cellswereplatedin 10 cm dishes at 5 million cells per dish in mediumsupplemented with 1% charcoal-stripped FBS (csFBS).

Transfection and transactivation assayPlasmids and transfection assays were described earlier

(15).GRE-LUCandSRC-2 coactivator plasmidswere kindlyprovided by Dr. Nancy L. Weigel and Dr. Bert W O’Malley(BaylorCollege ofMedicine,Houston, TX). Transfections ofLNCaP cells with Amaxa electroporator (Amaxa Inc.,) wereconducted according to the manufacturer’s protocol. StableLNCaP and NIH3T3 cell lines were generated by lentiviralinfection of AKR1C3 cloned into pLenti U6 Pgk-puro vectoras described earlier (16, 17). LacZ and AR adenovirus weremade at Seven Hills Bioreagents.

Translational RelevanceCastration-resistant prostate cancer (CRPC) is charac-

terized by the emergence of a hypersensitive androgensignaling axis after orchiectomy or medical castration.The revival of androgen signaling is believed to be due tohigh intratumoral androgen synthesis, fueled by upre-gulation of steroidogenic enzymes, including aldo ketoreductase 1C3 (AKR1C3). Here, for the first time, usingmolecular and in vivo preclinical models, and humanCRPC tissues, we show that the steroidogenic enzymeAKR1C3 also acts as a selective coactivator for androgenreceptor to promote CRPC growth. Moreover, novelsmall-molecule inhibitors inhibit both the enzymaticand coactivator functions of AKR1C3 resulting in andro-gen-dependent prostate cancer and CRPC regression.These observations identify AKR1C3 as the first novelreceptor- and tissue-selective pharmacologically target-able coactivator that promotes prostate cancer growth.This is also the first in vivo evidence showing the impor-tance of AKR1C3 and the use of its inhibitors in prostatecancer.

Yepuru et al.

Clin Cancer Res; 19(20) October 15, 2013 Clinical Cancer Research5614




Chromatin immunoprecipitation assayChIP assays were conducted with AR N-20 and AKR1C3

rabbit polyclonal antibodies as described earlier (18). Pros-

tate-specific antigen (PSA) enhancer and nonspecific regionprimers and fluorescent probe sequences are describedearlier.

nmol/L (A'dione)

0.01 0.1 1 10 10

01,0

00

10,00

0

PS

A/G

AP

DH

0

5

10

15

20

25

30

35

pmol/L (R1881)

1e-4

1e-3

1e-2

1e-1

1e+0

1e+1

1e+2

1e+3

1e+4

1e+5

1e+6

PS

A/G

AP

DH

0

5

10

15

20

25

30

35

01,0002,0003,0004,0005,0006,0007,0008,000

AKR

1C3/

GAP

DH

0

0.5

1

1.5

C

D

B

Vector = 43.51 pmol/L

AKR1C3 = 13.71 pmol/L

Vector AKR1C3

EC50

Vector

AKR1C3

Vector

AKR1C3

Tu

mo

r vo

lum

e (

mm

3)

Weeks

Weeks

AK

R1C

3/G

AP

DH

TM

PR

SS

2/G

AP

DH

FK

BP

51/G

AP

DH

* * *

Cholesterol

Pregnenolone Progesterone

DHEA Androstenedione

Testosterone

Estrone

Estradiol

DHT

400

350

300

250

200

150

100

50

0

–50

Androstenediol

AKR1C3 AKR1C3 AKR1C3

A

20αα(OH)Proges AKR1C1

Androsterone AKR1C3 17b-HSD6

0 1 2 3 4 5 6 7 80.0 2.5 5.0 7.5 10.0 12.5

70

60

50

40

30

20

10

0

Pe

rce

nt

tum

or

inc

ide

nc

e

0

1

2

3

4

PS

A/G

AP

DH

N.S. AKR1C3 siRNA

0

5

10

15

20

25

TM

PR

SS

2/G

AP

DH

N.S. AKR1C3 siRNA N.S. AKR1C3 siRNA

AK

R1C

3/G

AP

DH

* *

*

FKBP51

Actin

AK

R1

C3

Ve

cto

r

AK

R1

C3

AK

R1

C3

Ve

cto

r

AK

R1

C3

Ve

cto

r

AK

R1

C3

AKR1C3

AR 0100200300400500600

00.5

11.5

22.5

3

0

0.5

1

1.5

2

2.5

Figure 1. AKR1C3enhancesandrogen signaling and prostate cancer xenograft growth. A, steroidogenic enzymesynthesis pathways. Bold arrows indicateback-doorpathway.B,AKR1C3 transfection increases androgen-inducedPSAgeneexpression. LNCaPcellswere transfectedwith 10mgvector (solid line) orAKR1C3(broken line) using Amaxa electroporator. Cells were maintained in serum-free medium for 3 days, treated for 16 hours with indicated concentrations of A'dione(left) or R1881 (middle) and expression of PSAwasmeasured and normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH) by real-time PCR. Rightpanel shows AKR1C3 expression in cells transfectedwith AKR1C3 (closed bars) or vector (open bars; not visible in the figure). EC50 of R1881-induced PSA geneexpression is given above the middle panel. C, AKR1C3 increases DHT-induced LNCaP tumor xenograft implantation and growth. Nude mice (n ¼ 11/group)bearing LNCaP-vector (solid line, diamond) or LNCaP-AKR1C3 (broken line, square) were castrated and a 90-day sustained release DHT pellet (12.5 mg)implanted under the skin. Tumor volume was measured biweekly. Tumor uptake of LNCaP-vector (solid line) or LNCaP-AKR1C3 (broken line) cells is shown asKaplan–Meier plot (right). Expression of indicated genes and proteins was measured in tumors using real-time PCR or Western blotting. Open bars are LNCaP-vector tumors (n ¼ 3) and filled bars are LNCaP-AKR1C3 tumors (n ¼ 6). D, AKR1C3 siRNA reduces androgen signaling. VCaP prostate cancer cells weretransfectedwith nonspecific siRNA (N.S.) or AKR1C3 siRNA, and the expression of androgen-responsive genes wasmeasured in response to 0.1 nmol/L R1881(filledbars). � indicates significance atP < 0.05 fromvector tumorsor vehicle-treated cells. In vitro experiment figures are representative ofn¼ 3 experiments eachconducted in triplicates; cyclo-cyclophilin; LNCaP-AKR1C3-LNCaP cells stably transfected with AKR1C3.

AKR1C3 Is an AR Coactivator

www.aacrjournals.org Clin Cancer Res; 19(20) October 15, 2013 5615




Vehicle

100 nmol/L A'dione

10 nmol/L R1881

AR AKR1C3 MergeA

B

D

C

DAPI

Interaction

Negative

(AKR1C3+IgG)

Case 1 2 3 4 5 6

Pix

el in

ten

sit

y

Negative

250,0000

200,0000

150,0000

100,0000

50,0000

0

Positive

AR

/in

pu

t

–

A'd

ion

e

R1

88

1

–

A'd

ion

e

R1

88

1

LN-Vector LN-AKR

–

A'd

ion

e

R1

88

1

LN-AKR

0

1

2

3

4

525

20

15

10

5

0

AK

R/in

pu

t

–

A'd

ion

e

R1

88

1 –

A'd

ion

e

R1

88

1

LN-Vector LN-AKR

–

A'd

ion

e

R1

88

1

LN-AKR

PSA Enhancer N.S. PSA Enhancer N.S.

AKR1C3

AR

IP

20% input

1 2 3 3

AR IgG

1 2 3Animal. No.

0

2

4

6

ARIgG

Pix

el

inte

ns

ity

IP

IP:AKR1C3

IB:AR

AR:

10% input

R1881 – + – +

Vector AKR1C3

AKR1C3

10% input

IP:AKR1C3

IB:AR

AR:

10% input

Vehicle 10 nmol/L R1881

Figure 2. AKR1C3 physically interacts with AR. A, left, coIP of AR and AKR1C3. LNCaP-AKR1C3 cells were serum starved for 2 days and treated with10 nmol/L R1881 for 6 hours. Cells were harvested, protein extracted, and immunoprecipitated with AKR1C3 antibody and Western blotted for AR.A, right, AR-AKR1C3 colocalization showed by laser confocal microscopy. (Continued on the following page.)

Yepuru et al.





Duolink PLADuolink kit (O’link) was purchased and was used to

determine the interaction between AR and AKR1C3 (19).Images were obtained using deconvolution fluorescentmicroscopy.

Prostate cancer specimensSection of prostate cancer specimens (n ¼ 6) with

Gleason score 7 (4 þ 3) were obtained under The Uni-versity of Tennessee (Memphis, TN), Institutional ReviewBoard approval. Pathologist report indicates a minimumof 60% and maximum of 85% of the samples to bepositive for prostate cancer. The specimens were subjectedto Duolink’s proximity ligation assay (PLA) with AR andAKR1C3 antibodies. For negative control, the sampleswere probed with AKR1C3 antibody and immunoglob-ulin G (IgG).

RNA isolation and gene expressionRNA was isolated using Cells-to-CT Kit (Applied Biosys-

tems) and real-time PCR was conducted using TaqManprimers and probes from Applied Biosystems on ABI7900 (Applied Biosystems).

Growth assayLNCaP cells were plated at 10,000 cells per well of

a 96-well plate in RPMI supplemented with 1% csFBS.The cells were treated as indicated in the figures. Cellviability was measured using sulforhodamine (SRB)reagent.

Tumor xenograft experimentsAll animal protocols were approved by The University

of Tennessee Animal Care and Use Research Committee.Xenograft experiments were carried out as previouslypublished (15). Briefly, a mixture of 2 � 106 LNCaP orVCaP cells were suspended in 0.0375 mL RPMI þ 10%FBS and 0.0625 mL Matrigel/animal and were injectedsubcutaneously. Once the tumor size reached 200 to300 mm3, the animals were castrated or sham operated,randomized, DHT pellets were implanted subcutane-ously or not supplemented, and treated as indicatedin the figures. Tumor volume and body weight weremeasured.

ResultsAKR1C3 overexpression increases AR-dependent geneexpression

Although AKR1C3 is increased in CRPC, its role on geneexpression and growth of prostate cancer cells is not known.PSA gene expression was measured in LNCaP cells trans-fectedwithAKR1C3usingAmaxa electroporator and treatedwith increasing concentrations of A’dione (Fig. 1B, left) orR1881 (Fig. 1B, middle). AKR1C3 overexpression (Fig. 1B,right) increased PSA gene expression in response to bothA’dione and R1881.

AKR1C3 overexpression promotes prostate cancerxenograft implantation and growth

LNCaP-vector or LNCaP-AKR1C3 cells (LNCaP cells sta-bly infected with lentivirus carrying AKR1C3) were injectedsubcutaneously in nude mice. The animals were castratedand supplemented with suboptimal concentration of DHT(Fig. 1C) or left intact (Supplementary Fig. S1). AKR1C3-transfected tumors had a higher incidence of tumor implan-tation (Fig. 1C, right) and better rate of tumor growth(Fig. 1C, left) than LNCaP-vector tumors. Consistently, thetumor incidence was two times higher in LNCaP-AKR1C3–bearing animals than LNCaP-vector–bearing animals.Expression of AR-dependent genes such as FKBP51 andTMPRSS2 (Fig. 1C, lower) increased significantly and theirprotein levels were higher in LNCaP-AKR1C3 tumors (Fig.1C, lower), indicating the role for AKR1C3 to enhanceAR signaling in androgen-dependent prostate cancer andCRPC progression. A significant increase in the androgen-dependent FKBP51 protein expression was observed inLNCaP-AKR1C3 tumors without concomitant increase inAR protein levels (Supplementary Fig. S2).

AKR1C3 siRNA reduces AR signaling in VCaP cellsTo determine whether endogenous AKR1C3 is important

for androgen signaling, VCaP cells were transfected withnonspecific or AKR1C3 siRNA, treated with R1881, andexpression of AR target genes was measured. Reduction inAKR1C3 expression by 70% to 80% reduced R1881-induced PSA and TMPRSS2 gene expression by 30% to40% (Fig. 1D), indicating the important role played byAKR1C3 in AR function.

The ability of AKR1C3 to augment R1881-inducedgene expression, DHT-dependent tumor growth and gene

(Continued.) LNCaP-AKR1C3 cells plated on coverslips and serum starved for 2 dayswere treatedwith vehicle, 100 nmol/L A'dione, or 10 nmol/L R1881 for 6hours. Cells were fixed and immunostained with primary antibodies specific to AR and AKR1C3 and fluorescent-tagged secondary antibodies. Theimmunofluorescent signals were captured by laser confocal microscopy. B, AKR1C3 is recruited to PSA enhancer. LNCaP-AKR1C3 or LNCaP-Vectorcells were maintained in serum-free conditions for 3 days and were treated with 100 nmol/L A'dione or 10 nmol/L R1881 for 2 hours and ChIP assay wasconducted with AR (left) or AKR1C3 (right) antibodies. DNA was purified and real-time PCR was conducted for PSA enhancer or a nonspecific region (N.S.)and normalized to input. Representative data from triplicate experiments is given. C, endogenous AR and AKR1C3 interact in VCaP cells (left) andVCaP xenograft (right). VCaP cells were serum starved for 2 days and treated with vehicle or 10 nmol/L R1881 and immunoprecipitation conducted asdescribed in panel A. Right, protein was extracted from VCaP CRPC xenografts (n¼ 3) and was subjected to immunoprecipitation with AR antibody or IgG.The Iped samples and 20% input samples were fractionated on an SDS-PAGE and Western blotted with AKR1C3 antibody and AR antibody. D, AR andAKR1C3 interact in advanced prostate cancer specimens. Tissue sections from Gleason score 7 prostate cancer (n¼ 6) were subjected to Duolink PLA withAR and AKR1C3 antibodies (interaction) or AR antibody replaced with IgG (negative control bottom). Nucleus was counterstained with 40, 6-diamidino-2-phenylindole (DAPI) and images captured using fluorescent microscopy. Representative images from each sample are displayed. Bar graph of pixel intensityof scanned slides is represented. Figures are representative of n ¼ 3. IB, immunoblotting.






expression, and attenuate R1881 response byAKR1C3 siRNAestablish and support a nonenzymatic function for AKR1C3.

AKR1C3 interacts with AR and is recruited to PSAenhancer

The ability of AKR1C3 to potentiate AR function sug-gested that itmight possibly function as a coactivator. Co-IPstudies conducted in LNCaP-AKR1C3or LNCaP-vector cellstreated with R1881 indicate that AR and AKR1C3 interactedin a ligand-dependentmanner in LNCaP-AKR1C3 cells, butnot in LNCaP-vector cells (Fig. 2A, left).

Immunofluorescence studies in LNCaP-AKR1C3 cellswere conducted using laser confocal microscopy (Fig. 2A,right). Both AR and AKR1C3 were cytoplasmic in theabsence of AR ligands, but translocated into the nucleusupon binding of an AR ligand (R1881) or A’dione. Themigratory patterns for AR and AKR1C3 overlapped sub-stantially, supporting the idea that the two proteins interactwith each other.

To ensure that AKR1C3’s translocation is dependent onAR, NIH3T3 cells stably transfected with AKR1C3 wereinfected with adenovirus expressing Lac Z (SupplementaryFig. S3, top) or AR (Supplementary Fig. S3, bottom) andwere treated with 10 nmol/L R1881. Although cells weretreated with R1881, AKR1C3 was cytoplasmic in theabsence of AR and translocated into the nucleus only inthe presence of AR, indicating the requirement for the ARpresence for AKR1C3 to translocate into nucleus.

Our earlier publication showed that an AR antagonist,SNARE-1, inhibited ligand-dependent AR nuclear translo-cation (15). We tested the translocation of AR and AKR1C3in response to R1881 in the presence or absence of SNARE-1. R1881 efficiently translocated AR into the nucleus andAKR1C3 cotranslocated with AR. However, when cells weretreated with 10 mmol/L SNARE-1, AR only partially trans-located into the nucleus and predominantly remained inthe cytoplasm. AKR1C3 followed the same pattern (Sup-plementary Fig. S4).

The Duolink PLA detects protein–protein interaction byfluorescent visualization (20). DNA attached to the second-ary antibodies is ligated and amplified, and the amplifiedDNAfluoresces red only if the twoproteins are in proximity.LNCaP-AKR1C3 cells were treated with R1881 and sub-jected to PLA. Although AR and AKR1C3 interaction (repre-sented by red fluorescence)was detected in LNCaP-AKR1C3cells (Supplementary Fig. S5A), interaction was undetectedin LNCaP-mock cells or when an antibody was replacedwith IgG (Supplementary Fig. S5A). These results werereproduced in VCaP cells, which contain endogenous ARand AKR1C3 (Supplementary Fig. S5B).

If AKR1C3 interacts with AR, it could also be recruited tothe response element (ARE) of an AR target gene. LNCaP-AKR1C3 or LNCaP-vector cells were treated with R1881 orA’dione, and the recruitment of AKR1C3 to PSA enhancerAREwas examined using ChIP assay. AKR1C3was recruitedto the PSA enhancer, but not to a nonspecific region, both inresponse to its substrate, A’dione, as well as in response toR1881 (Fig. 2B). Although AR was recruited to the PSA

enhancer in LNCaP-vector cells (Fig. 2B), recruitment ofAKR1C3 to the PSA enhancer could not be detected inLNCaP-vector cells due to its limited expression.

To show that AR and AKR1C3 interact in a system whereboth proteins are endogenously expressed, co-IP studieswere conducted in VCaP cells and VCaP CRPC xenografttumors. AR andAKR1C3 interact robustly inVCaP cells (Fig.2C, left) in response to R1881 and inCRPC tumor xenografttissues (Fig. 2C, right).

AKR1C3 and AR interact in advanced prostate cancerTo determine whether the interaction between AR and

AKR1C3 observed in prostate cancer cells and xenografts isalso observed in human prostate cancer, prostate cancerspecimens (n¼ 6; Gleason sum4þ 3¼ 7)were subjected toPLA with AR and/or AKR1C3 antibody. Although interac-tion between the two proteins (as visualized by red fluo-rescent spots) was clearly observed in the tumors (Fig. 2D),the interaction was not detected when one antibody wasreplaced with IgG (Fig. 2D, bottom). The intensity and theextent of interaction varied between samples. The fluores-cent spots were quantified using automated software. Thegraphs below Fig. 2D clearly indicates a robust increase inpixel intensity, a representation of interaction between thetwo proteins, in advanced prostate cancer samples.

AR transactivation as AKR1C3 functional assayAKR1C3 converts A’dione to testosterone resulting in a

ligand with stronger AR activity. We captured this principlein an AR transactivation assay. AR activity in response toA’dione in AKR1C3-transfected cells was higher than the ARactivity in vector-transfected cells (Fig. 3A). AKR1C3 facil-itated AR activation as evidenced by the lower half maximaleffective concentration (EC50) andhigher Emax of A’dione inAKR1C3-transfected cells. Overexpression of AKR1C3 sig-nificantly reduced the EC50 of A’dione to transactivate ARfrom 415 to 175 nmol/L (Supplementary Table S1). Theseresults support the hypothesis that AKR1C3 overexpressionin prostate cancer amplifies or hypersensitizes the AR sig-naling pathway.

AKR1C3 activates AR in response to active androgensThe results observed in LNCaP-AKR1C3 cells, VCaP cells,

and CRPC samples indicated the ability of AKR1C3 tocoactivate AR. We conducted AR transactivation studies inAKR1C3-transfected cells using A’dione and two 17-hydroxy AR agonists (testosterone and DHT; ref. 21). Con-sistent with gene expression results, AKR1C3 increased ARtransactivation in response to all ligands that bind to the AR(Fig. 3B). AKR1C3 not only reduced the EC50 of theseligands, but also increased the maximum level of AR trans-activation (Emax; Supplementary Table S1).

Varying the level of AKR1C3 expression showed thatstarting from 0.5 mg AKR1C3 plasmid DNA increasedAR transactivation in a concentration-dependent manner(Fig. 3C) without altering the expression of AR (Fig. 3D).The increase in AR transactivation facilitated by AKR1C3is comparable with that observed with an established

Yepuru et al.





coactivator, SRC-2 (Supplementary Fig. S6), indicative ofthe robustness in the ability of AKR1C3 to augment ARfunction. AKR1C3 overexpression was confirmed by real-time rtPCR (Fig. 3C, lower). Transactivation experiments indifferent cell types (COS-1 and NIH3T3), various transfec-tion conditions (lipofectamine, fugene, and Amaxa electro-porator), and cells stably expressing AKR1C3 showed thatthe AR activation effect of AKR1C3 was not unique to a celltype or transfection condition (Supplementary Fig. S7).Steroid receptors share sequence homology in many of

their functional domains, facilitating their interaction withthe same coactivator. Transactivation experiments carriedout with glucocorticoid receptor, mineralocorticoid recep-

tor, progesterone receptor, estrogen receptor-a, and perox-isome proliferator-activated receptor g (PPARg) establishedthat AKR1C3 is a selective activator of AR function (Sup-plementary Fig. S8).

AR augmentation effect is selective to AKR1C3Because members of AKR1C family share high sequence

homology, the effect of AKR1C isoforms on AR transactiva-tion in response to R1881 was tested. Interestingly, wefound that R1881-induced AR transactivation was aug-mented only by AKR1C3, but not by AKR1C1, 1C2, or1C4, indicating the selective ability of AKR1C3 to functionas an AR activator (Supplementary Fig. S9).

A

B

C D

nmol/L (A'dione)

0.0001 0.001 0.01 0.1 1 10 100 1,000 10,000

RL

U Vector

AKR1C3

nmol/L (A'dione)

1e-5

1e-4

1e-3

1e-2

1e-1

1e+0

1e+1

1e+2

1e+3

1e+4

1e+5

RL

U

0.00

0.02

0.04

0.06

0.08

0.10

0.12

0.14

0.16

0.18

0.20

nmol/L (Testosterone)

1e-5

1e-4

1e-3

1e-2

1e-1

1e+0

1e+1

1e+2

1e+3

1e+4

1e+5

RL

U

0.00

0.05

0.10

0.15

0.20

0.25

0.30

A'dione Testosterone

nmol/L

1e-4

1e-3

1e-2

1e-1

1e+0

1e+1

1e+2

1e+3

1e+4

1e+5

RL

U

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

Vector0.1 mg 0.25 mg

0.5 mg 0.75 mg

1 mg Vector AKR1C3

AR

ng AKR1C3

0 500 1,000

AK

R1C

3/G

AP

DH

0

2,000

4,000

6,000

8,000

nmol/L (DHT)

1e-5

1e-4

1e-3

1e-2

1e-1

1e+0

1e+1

1e+2

1e+3

1e+4

1e+5

RL

U

0.00

0.05

0.10

0.15

0.20

0.25

0.30

DHT

0.00

0.05

0.10

0.15

0.20

0.25

Figure 3. AKR1C3 augments ARtransactivation in response toactive androgens. A, AKR1C3augments A'dione-dependentAR transactivation. HEK-293cells were transfected with 0.25mg GRE-LUC, 50 ng CMV hAR,10ngCMV-Renilla luciferase, and1 mg vector or AKR1C3. Twenty-four hours after transfection, thecells were treated with a titrationof A'dione. Cells were harvested48 hours after transfection, andfirefly luciferase levels weremeasured and normalized toRenilla luciferase. B, AKR1C3increases active androgen-induced AR transactivation.HEK-293 cells were transfectedas mentioned in A, treated witha titration of the indicatedandrogens, and luciferase assaywas conducted. C, AKR1C3concentration dependentlyincreases R1881-induced ARtransactivation. HEK-293 cellswere transfected with indicatedconcentration of AKR1C3. Totalamount of transfected plasmidswas normalized to 1 mg withvector pCR3.1. The cells weretreated with a titration of R1881,and luciferase assay wasconducted. Bottom panelshows the AKR1C3 RNA levels.D, AKR1C3 does not increase ARexpression. Cells transfectedwith vector or AKR1C3 andtreated with 10 nmol/L R1881were fractionated by SDS-PAGEand Western blotted for AR. Allfigures are representative of n¼ 3experiments. RLU, relative lightunits; solid lines are vectortransfected and broken lines areAKR1C3-transfected cells.






AKR1C3 and SRC-2 synergize to increase ARtransactivation

To determine whether AKR1C3 and a bonafide coactiva-tor, SRC-2, that is overexpressed in prostate cancer (22),synergistically increase AR transactivation, plasmids encod-ing both proteins were transfected and AR transactivationstudies were conducted. Although SRC-2 (SupplementaryFig. S6, top left) and AKR1C3 (Supplementary Fig. S6, topright) concentration dependently increased AR transactiva-tion, cotransfection of suboptimal concentrations of thetwo plasmids synergistically increased AR transactivation

(Supplementary Fig. S6, bottom), suggesting different inter-action sites with AR.

Different regions of AKR1C3 mediate the enzymaticand activation function

To determine the regions responsible for the enzymaticand activation functionsofAKR1C3, truncated andmutatedAKR1C3 constructs were generated in pCR3.1 vector (Sup-plementary Fig. S10A), and AR transactivation studies inresponse to A’dione and R1881 were conducted. Aminoacids 1-182 failed to mediate the effects of A’dione or

A B

C

N

HO

OH

F

F

F

GTx-560

– 10 1 .1 .01 10 1 .1 .01

IndomethacinGTx-560

µmol/L

A'dione

T

% T

GTx-560Indomet

.01 .1 1 100

µmol/L

LN

LN

-AK

R

AKR1C3

Actin

A'dione

T

Veh

Veh

56

0

56

0

59

4

Ind

o

Veh

LNCaP LNCaP-AKR

10 mmol/L finasteride

D

% T 6 9 8 28 11 12 27

– 10 1 0.1 0.01 10 1 0.1 0.01

IndomethacinGTx-560

µmol/L

A’dione

T

% testo

ste

rone

0.01 0.1 1 10

µmol/L

GTx-560

Indomethacin

0

– 560 Indo

% t

es

tos

tero

ne

% t

es

tos

tero

ne

LNCaP

– – 560 Indo

LNCaP-AKR

*

0102030405060708090

100

0102030405060708090

0

2

4

6

8

10

12

14

16

0

5

10

15

20

25

30

Figure 4. GTx-560 inhibits AKR1C3 enzyme activity in vitro. A, structure of GTx-560. B, left, GTx-560 and indomethacin inhibit AKR1C3 enzyme activity.Purified AKR1C3 enzyme was incubated with ATP and [14C]- A'dione (12 mmol/L) in the presence of a titration of GTx-560 or indomethacin for 60 minutes.A'dione and testosteronewere fractionated using TLC and quantified using phosphorimager. The amount of testosterone synthesized is represented as a linegraph (lower). B, right, GTx-560, but not indomethacin, inhibits AKR1C3enzymeactivity inHEK-293cells. HEK-293 cellswere transiently transfectedwith 1mgAKR1C3 and treated with a titration of GTx-560 or indomethacin in the presence of [14C]-A'dione (1 mmol/L). After overnight incubation, A'dione andtestosterone in themediumwere fractionatedusingTLCandwerequantifiedusingphosphorimager. The amount of testosterone synthesized is represented inthe line graph (lower). C, GTx-560, but not indomethacin, inhibits AKR1C3 enzyme activity in adrenal (left) and prostate cancer (right) cell lines. H295R adrenalcells and LNCaP-AKR1C3 cells were treated with [14C]-A'dione (12 mmol/L) in the presence or absence of 10 mmol/L GTx-560 or indomethacin. Medium wascollected and the testosterone synthesized was detected by TLC and quantified using phosphorimager. Western blot analysis showing AKR1C3overexpression in LNCaP cells is shown in the bottom panel. � indicates significance at P < 0.05 of n¼ 3. Representative experiments shown in the figure. D,GTx-560 inhibits finasteride-dependent testosterone formation. LNCaP-Vector or LNCaP-AKR1C3 was treated with [14C]-A'dione (1 mmol/L) in the presenceor absence of 10 mmol/L finasteride, GTx-560, GTx-594, or indomethacin (indo). Percent testosterone formed was calculated and represented as numbersbelow the TLC image. Representative of n¼ 3 experiments is depicted here. T, testosterone; LN, LNCaP-mock; LN-AKR, LNCaP-AKR1C3; veh, vehicle; 560,GTx-560; indo, indomethacin.

Yepuru et al.





R1881. Although full length AKR1C3 (construct A) wasrequired for the effects of A’dione, amino acids 1-282 weresufficient to mediate the effects of R1881 (SupplementaryFig. S10B). The region of AKR1C3 spanning amino acids171-237 (construct G) was sufficient to mediate the effectsof R1881. Point mutation F306A that eliminates the bind-ing of A’dione to AKR1C3 also eliminated its effect on ARtransactivation in response to A’dione, but not to R1881(23). These results were confirmed with PSA gene expres-sion in LNCaP cells expressing construct G or H (Supple-mentary Fig. S10D). The enzymatic functions of theseconstructs were confirmed by thin layer chromatography(TLC; Supplementary Fig. S10C). These results suggest thatthe full-length protein is required to mediate AKR1C3’senzymatic functions, but that amino acids 171-237 weresufficient to mediate the AR activation.

Novel small molecules inhibit AKR1C3 activityWe designed and synthesized a novel series of AKR1C3

inhibitors and compared the activity of the most potentinhibitor, GTx-560 (Fig. 4A), to a knownAKR1C3 inhibitor,indomethacin, in in vitro purified and cell based-enzymeassays (24). Though GTx-560 and indomethacin compara-bly inhibited AKR1C3-dependent conversion of A’dione to

testosterone (Fig. 4B, left) in purified enzyme assays, GTx-560 alone inhibited AKR1C3 enzyme activity in cells(Figs. 4B, right and C), without cross-reacting with a highlyhomologous AKR1C1 (ref. 25; Supplementary Fig. S11).The difference in the IC50s between purified enzyme-based system (�2–3 mmol/L) and cell-based system (10–50 nmol/L) is due to the difference in concentration ofsubstrate A’dione used (12 mmol/L vs. 100 nmol/L).

Because 5a-reductase converts testosterone to DHT (26),we speculated that finasteride, a 5a-reductase inhibitor,would increase testosterone levels in LNCaP-AKR1C3 cells(Fig. 4D). As expected, finasteride blocked the conversion oftestosterone to DHT in LNCaP-AKR1C3 cells, thereby sig-nificantly increasing the testosterone levels. By actingupstream of 5a-reductase in the steroidogenic pathway,GTx-560 completely blocked the formation of testosteronefrom A’dione, indicating the ability of AKR1C3 inhibitors,unlike 5a-reductase inhibitors, to reduce testosteronelevels.

GTx-560 selectively inhibits AKR1C3-dependent ARtransactivation

In agreement with the enzyme inhibition results,GTx-560, but not indomethacin, effectively inhibited the

A B

C

nmol/L1e

-41e

-31e

-21e

-11e

+0

1e+1

1e+2

1e+3

1e+4

1e+5

RL

U

RL

U

0.00.20.40.60.81.01.21.41.61.8

D

Vector = 0.1022 nmol/L

AKR1C3 = 0.072 nmol/L

AKR+560 = 0.1262 nmol/L

EC50

nmol/L (R1881)

1e-5

1e-4

1e-3

1e-2

1e-11e

+0

1e+1

1e+2

1e+3

1e+4

1e+5

0.0

0.5

1.0

1.5

2.0

2.5

3.0

VectorVector+560

DD+560

R1881

– – 0.01 0.1 1 10

– – 0.01 0.1 1 10 – – 0.01 0.1 1 10 – – 0.01 0.1 1 10

560

µmol/L

10 nmol/L A'dione

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

A'dione A'dione A'dione

Vector AKR1C3 17bHSD3

nmol/L

1e-4

1e-3

1e-2

1e-1

1e+0

1e+1

1e+2

1e+3

1e+4

1e+5

RL

U

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

Vector = 415 nmol/L

AKR1C3 = 175 nmol/L

AKR+560= 422 nmol/L

EC50

RL

U

GTx-560

(µmol/L)

RL

U

0.1 1 10

Indomethacin

0.01

0

0.05

0.1

0.15

Figure 5. GTx-560 inhibits AKR1C3 enzyme and coactivator activities in cells. A, GTx-560, but not indomethacin, inhibits AKR1C3-mediated A'dione-inducedAR transactivation. HEK-293 cells transfected as indicated above were treated with 10 nmol/L A'dione alone or in combination with a titration of GTx-560 orindomethacin and luciferase assay conducted. A, right, GTx-560 increases the EC50 of A'dione. HEK-293 cells transfected as indicated above weretreated with a titration of A'dione alone or in combination with 10 mmol/L GTx-560 and luciferase assay was conducted. EC50 values are given on top of thefigure. B, GTx-560 inhibition of AR transactivation is selective to AKR1C3. HEK-293 cells transfected with 1 mg pCR3.1, pCR3.1-AKR1C3, or pCR3.1 17b-HSD3 were treated with 10 nmol/L A'dione in the presence or absence of a titration of GTx-560 and transactivation assay was conducted. C, GTx-560inhibits AKR1C3-dependent R1881-induced AR transactivation. HEK-293 cells transfected as indicated above were treated with a titration of R1881alone or in combination with 10 mmol/L GTx-560 and luciferase assay was conducted. Solid line indicates vector-transfected cell, broken line indicatesAKR1C3-transfected cells, and dotted line indicatesAKR1C3-transfected cells treatedwithGTx-560. EC50 of R1881under various conditions is described. D,GTx-560 requires full-lengthAKR1C3 to inhibit its coactivation effect. HEK-293 cells transfected as indicated in the figurewere treatedwith a titrationof R1881alone or in combination with 10 mmol/L GTx-560 and luciferase assay was conducted. Results shown in the figures are representative of three experiments.RLU, relative light units; 560, GTx-560.






A

0

0.5

1

1.5

2

2.5

3

3.5

4

− − + − + − − + − +GTx-560+ + + +

R1881 A'dione R1881 A'dione

pCR3.1 AKR1C3

PSA

PS

A/G

AP

DH

OD

53

5 n

m

10 nmol/L A'dione10 nmol/L A'dione

− − 56010 µmol/L

B

C

PS

A/G

AP

DH

(%

)

− 560 − 560

−

PS

A/G

AP

DH

(F

old

)

– Fin

63.43

15.49 Actin

PSA

*

*

* *

*

*

D

Vehicle

GTx-560

Vehicle GTx-560

Vehicle

GTx-560

Se

rum

GT

x-5

60

(n

g/m

L)

* ** *

**

0 1 2 3 4

% c

ha

ng

e

Weeks

**

Pintercept = 0.0022

00.20.40.60.8

11.21.4

020406080

100120140160180200

% c

ha

ng

e i

n t

um

or

vo

l.P

SA

(n

g/m

L)

0

150

100

50

0

130120110100908070605040302010

0-10

20406080

100

120140

160180200

0 1

Weeks

2

0123456789

0

50

100

150

200

250

300

GT

x-5

60

GT

x-5

60

GT

x-5

60

GT

x-5

60

Ve

hic

le

Ve

hic

le

Ve

hic

le

0 1,000 2,000 3,000

Tumor volume (mm3)

4,000 5,000

AR

AKR1C3

AR AR

AR AR

A'dione

Testosterone

GTx-560

Figure 6. GTx-560 inhibits androgen signaling and prostate cancer cell and tumor growth. A, GTx-560 inhibits AKR1C3-dependent PSA gene expression andLNCaP-AKR1C3cell growth. LNCaP-AKR1C3cellswere treatedwith vehicle or 10mmol/LGTx-560 in thepresenceofR1881orA'dione. PSAgeneexpressionwas measured and normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH, left). Right, LNCaP-vector (open bars) or LNCaP-AKR1C3(filled bars) cells were treated with vehicle or 10 mmol/L GTx-560 for 3 days. Cells were fixed and stained with sulforhodamine Blue (SRB) and the opticaldensity (OD) was measured at OD 535 nm. B, GTx-560 inhibits PSA gene expression in VCaP cells. VCaP prostate cancer cells were maintained inserum-free medium for 3 days and treated with vehicle, 10 mmol/L GTx-560 (top), or 10 mmol/L finasteride (bottom). (Continued on the following page.)

Yepuru et al.





AKR1C3-dependent A’dione-induced AR transactivation(Fig. 5A), but not the 17b-HSD3- (Fig. 5B) and AKR1C1-dependent (27) transactivation (Supplementary Fig. S13).This inhibition was observed at all AKR1C3 levels, indicat-ing the potency of GTx-560 to block AKR1C3 enzymeactivity (Supplementary Fig. S14).To understand whether a competitive inhibitor of

AKR1C3 such as GTx-560 has the potential to inhibitR1881-induced AKR1C3-dependent AR transactivation,HEK-293 cells transfected with vector or AKR1C3 weretreated with increasing concentrations of R1881 alone orin combination with GTx-560. GTx-560 completely inhib-ited the AKR1C3-dependent transactivation induced byR1881 (Fig. 5C). Though R1881 required amino acids 1-282 of AKR1C3 to elicit its coactivation, GTx-560 requiredthe full-length AKR1C3 to bind and inhibit R1881-inducedactivity (Fig. 5D).

GTx-560 inhibits the AKR1C3-induced PSA geneexpression and cell growthIn concordance with AR transactivation, GTx-560, at 10

mmol/L concentration, was very effective in inhibiting theAKR1C3-dependent A’dione- and R1881-induced PSA geneexpression (Fig. 6A, left) and A’dione-induced cell growth(Fig. 6A, right).

Testosterone levels are reduced by AKR1C3 inhibitionand not by 5a-reductase inhibitionAs intratumoral levels of testosterone can drive prostate

cancer progression, we tested the activity of GTx-560 andfinasteride (Fig. 6B) to reduce AR signaling and testosteroneproduction in VCaP, which are CRPC cells that expressendogenously high levels of AKR1C1, AKR1C3, and 5a-reductase type-1 (Supplementary Fig. S15). Although finas-teride increased PSA gene expression under serum-starvedconditions, GTx-560 significantly inhibited PSA gene expres-sion (Fig. 6B). The ability of finasteride to increase PSA geneexpression could be due to the higher expression of AKR1C1(Supplementary Fig. S15) as finasteride significantlyincreased testosterone levels (Supplementary Table S2).

GTx-560 inhibits prostate cancer xenograft growthAnimals bearing LNCaP-AKR1C3 tumors were castrated

when tumors reached 100 mm3 and were allowed to prog-

ress. Once the tumors reached 200 to 300mm3, the animalswere treatedwith vehicle or GTx-560 at 30mg/kg/d s.c. for 2weeks. The growth of AKR1C3-dependent tumors wasslowed by treatment with GTx-560 (Fig. 6C, top left).Comparable inhibition was also observed in another CRPCxenograft model using VCaP cells, where GTx-560 effective-ly reduced the tumor growth by 50% (Fig. 6C, top right),serum PSA to one-third of that observed in vehicle-treatedgroup (Fig. 6C, bottom left), and tumor PSA completely(Fig. 6C, bottommiddle). To ensure that the tumor volumereduction is due to the exposure to GTx-560, drug concen-tration wasmeasured in the serum of animals bearing VCaPxenograft and correlated with final tumor volume. Asshown in Fig. 6C (bottom right), concentration of GTx-560 in serum inversely correlated with tumor volume,indicating that increased exposure to an AKR1C3 inhibitorsuch as GTx-560 will impede the growth of CRPC tumors.Statistical analysis of this inverse correlation slope wasconducted using regression Bivariate fit (JMP 9.0 from SASinstitute). The results indicate a high level of significancewith P value for intercept being 0.0022 and for the drugconcentration being 0.0351.

DiscussionNumerous treatment options for men who develop

CRPC are evolving with many new therapies in develop-ment. Most of these agents target the AR signaling axis,including abiraterone and prednisone, MDV-3100, TOK-001, TAK-700, andCapesaris (28). The primarymechanismof CRPCprogression is the ability of these cancer cells to useandrogenic precursors from blood or to synthesize de novotestosterone by upregulating steroidogenic enzymes includ-ingCyp17A1 (17, 20 lyase, and17a-hydroxylase), AKR1C3,and 5a-reductase. Abiraterone, by inhibiting Cyp17A1, hasbeen shown to treat CRPC in a postchemotherapy setting byreducing serum PSA, increasing progression-free survival,and improving overall survival. Interestingly, animal mod-els have indicated that resistance to abiraterone treatmentmay be due to increase in the expression of its target,Cyp17A1, as well as higher expression of AR and AKR1C3(8). Moreover, the back door pathways (29–31) to synthe-size DHT (Fig. 1A, filled arrows) despite treatment withabiraterone (29) require 17b-HSDs for biosynthesis ofpotent androgens. With AKR1C3 being themajor 17b-HSD

(Continued.) RNA was extracted and PSA gene expression was measured and normalized to GAPDH. C, GTx-560 reduces the growth of LNCaP and VCaPprostate cancer xenograft growth and androgen-dependent gene expression. Top left, LNCaP-AKR1C3 cells (2 million cells/mouse) were subcutaneouslyimplanted in nude mice (n ¼ 7). Once tumors reach 100 mm3, the animals were castrated and A'dione pellets (5.25 mg 21 day release pellets) weresubcutaneously implanted.Once the tumors re-emerged, the animalswere randomizedand treatedwith vehicle or 30mg/kgGTx-560 subcutaneously. Tumorvolumes were measured biweekly. �, P < 0.05. Top right, GTx-560 reduces VCaP CRPC xenograft growth. VCaP cells (2 million cells/mouse) weresubcutaneously implanted in nudemice.Once tumors reached 200 to 300mm3, the animalswere castrated and the tumorswere allowed to develop asCRPC.Once the tumors re-emerged, the animals were randomized and treated with vehicle (n¼ 6) or 40 mg/kg GTx-560 (n¼ 8) subcutaneously for 4 weeks. Tumorvolumesweremeasuredbiweekly. Bottom left, GTx-560 suppresses serumPSAof animals bearingVCaPCRPCxenograft. PSAwasmeasured in the serumofanimals shown in B, top right. The numbers beneath the bars represent the mean of each group (n¼ 5). C, bottom middle, GTx-560 suppresses tumor PSA.Protein was extracted from tumors shown in E, fractionated by SDS-PAGE and Western blotted for PSA and actin. C, bottom right panel, serum GTx-560inversely correlates with tumor volume. GTx-560 concentrations in serum was measured and correlated with tumor volume shown in B, top right. D, modeldepicting AKR1C3's mechanism of action. The in vitro figures are representative of n ¼ 3 and each experiment was carried out with triplicate samples. fin,finasteride; 560, GTx-560; cyclo, cyclophilin; LNCaP-AKR1C3-LNCaP cells stably transfected with AKR1C3. In A–C, � indicates significance at P < 0.05 fromvehicle-treated samples.






expressed in CRPC, blocking any enzyme other thanAKR1C3 might not be beneficial in CRPC.

The studies presentedherein clearly show thedual role forAKR1C3 as an enzyme that converts androstenedione totestosterone and as a coactivator of AR. Both of the actionscould be inhibited by a competitive AKR1C3 inhibitor,GTx-560. We were surprised that a receptor-type selective coac-tivator had a drug-binding pocket that could serve as atherapeutic target. Although p160 coactivators have histoneacetyl transferase enzyme activity, they are challenging drugtargets due to their lack of a ligand-binding pocket. Even if adrug that targets their interaction surface with the steroidreceptor is designed, because of their interaction with mul-tiple receptors, it will likely not be selective to AR. On theother hand, AKR1C3 is an example of a coactivator that isoverexpressed in advanced prostate cancer and specificallycoactivates AR. Human prostate cancer samples used in ourexperiment had upregulation of AKR1C3 (Fig. 2D), vali-dating other publications. All these make AKR1C3 a poten-tial drug target to treat advanced prostate cancer. We alsoshow for the first time an AKR1C3 inhibitor, GTx-560, iseffective in prostate cancer both in vitro and in vivo, whichprovides validation of AKR1C3 as a therapeutic target forCRPC.

In addition, AKR1C3 also possesses prostaglandin F2(PGF2) synthase activity responsible for the synthesis ofproliferative and anti-differentiating PGF2. Human mye-loid leukemia cells overexpressing AKR1C3 were moreproliferative and resistant to the growth inhibitory effectsof all trans-retinoic acid and 1, 25 (OH)2 vitamin D3. Thiswas due to the deprivation of PGJ2, a ligand for pro-differentiating PPARg . These effects were overcome by treat-ing cells with a nonsteroidal anti-inflammatory drug(NSAID) AKR1C3 inhibitor, indomethacin (32). Althoughwe speculate that the proliferative effects of AKR1C3 in ourstudies were due to its 17b-HSD activity, we cannot rule outits 11-ketoreductase or PGF2 synthase activity as one of thecontributors to its activity in prostate cancer cells.

Because this is the first study depicting a steroidogenicenzyme as a coactivator of AR, we confirmed the interactionusing multiple tools and in different systems. The interac-tion was convincing in both cells and in human prostatecancer tissue. Because immunohistochemistry and immu-nofluorescence techniques gave rise to high autofluores-cence, to overcome this issue, we used PLA, which gavefluorescent signals specific to areas of interaction only (Fig.2D). Although more than 200 coactivators for nuclearreceptors have been discovered to date, AKR1C3 is distinctin being the first to be pharmacologically targetable. Severalquestions remain to be answered for this coactivator func-tion of AKR1C3. The most important being whetherAKR1C3 could function as AR coactivator when AR isactivated by noncanonical signaling pathways such as Src,mitogen-activated protein kinase, and others. Identifyingthis function could further enhance the therapeutic use ofAKR1C3 as an AR-selective coactivator. Others such as theAR domain responsible for AKR1C3 interaction and corre-lation between AR:AKR1C3 interaction and patient survival

data in AKR1C3overexpressing prostate cancer remain to beaddressed.

Interestingly, AKR1C3 interaction with AR in cells is verystrong enabling it to migrate with AR from cytoplasm tonucleus and this migration is purely dependent on AR.Furthermore, the FKBP51 protein expression in LNCaP-AKR1C3 xenograft samples compared with LNCaP-mockxenograft samples (Fig. 1C) provided compelling data thatAR signaling is robust in tissues expressing AKR1C3 even inthe presence of DHT, a response likelymagnified because ofcoactivation.

The evidence in this manuscript clearly suggests a newmechanism for AKR1C3 action in CRPC (Fig. 6D). AKR1C3converts intratumorally the adrenal androgens into testos-terone,whichbinds toARor get converted toDHT, resultingin ligand occupancy of AR. AR then interacts with AKR1C3and gets recruited to the ARE on the promoter of androgen-responsive genes. This recruits several other cofactors lead-ing to magnified transcription and translation of targetgenes. Blocking AKR1C3 will not only eliminate the enzy-matic conversion, but also will block its AR coactivationpotential, that could have been activated by noncanonicalpathways.

Future clinical trials with AKR1C3 inhibitors will beneeded to show their potential to be the next generationof tissue-specific therapeutics for CRPC. However, theunique roles of AKR1C3 as coactivator and androgen bio-synthetic enzyme involved in prostate cancer progressionidentify it as ahighpriority target for study, and indicate thatthe interactions between steroid biosynthetic enzymes andsteroid receptors may be exceedingly complex and involvedin a variety of hormone-dependent cancers.

Disclosure of Potential Conflicts of InterestC.M. Barrett has ownership interest (including patents) in GTx, Inc. M.S.

Steiner is employed (other thanprimary affiliation; e.g., consulting) as aCEOand CMO, and has ownership interest (including patents) in GTx, Inc. D.D.Miller has commercial research grant and is a consultant/advisory boardmember of GTx, Inc. No potential conflicts of interest were disclosed by theother authors.

Authors' ContributionsConception and design:M. Yepuru, Z. Wu, C.M. Barrett, M.S. Steiner, D.D.Miller, J.T. Dalton, R. NarayananDevelopment of methodology: M. Yepuru, M.S. Steiner, J.T. Dalton,R. NarayananAcquisitionofdata (provided animals, acquired andmanagedpatients,provided facilities, etc.): A. Kulkarni, F. Yin, M.S. Steiner, R. NarayananAnalysis and interpretation of data (e.g., statistical analysis, biosta-tistics, computational analysis): M. Yepuru, A. Kulkarni, F. Yin, J. Kim,M.S. Steiner, J.T. Dalton, R. NarayananWriting, review, and/or revision of the manuscript: M. Yepuru, F. Yin,C.M. Barrett, J. Kim, M.S. Steiner, D.D. Miller, J.T. Dalton, R. NarayananAdministrative, technical, or material support (i.e., reporting or orga-nizing data, constructing databases): M. Yepuru, F. Yin, C.M. Barrett,R. NarayananStudy supervision: M. Yepuru, J. Kim, J.T. Dalton, R. Narayanan

AcknowledgmentsThe authors thank Mr. Terrence Costello for assistance with the animal

studies and Ms. Ashley Ezekiel for her assistance with pathology slidepreparation and scanning.

The costs of publication of this article were defrayed in part by thepayment of page charges. This article must therefore be hereby marked

Yepuru et al.





advertisement in accordance with 18 U.S.C. Section 1734 solely to indicatethis fact.

Received May 2, 2013; revised August 7, 2013; accepted August 25, 2013;published OnlineFirst August 30, 2013.

References1. Fitzpatrick JM, Anderson J, Sternberg CN, Fleshner N, Fizazi K,

Rebillard X, et al. Optimizing treatment formenwith advanced prostatecancer: expert recommendations and the multidisciplinary approach.Crit Rev Oncol Hematol 2008;68 Suppl 1:S9–22.

2. Higano CS, Crawford ED. New and emerging agents for the treatmentof castration-resistant prostate cancer. Urol Oncol 2011;29:S1–8.

3. Shiota M, Yokomizo A, Naito S. Increased androgen receptor tran-scription: a cause of castration-resistant prostate cancer and a pos-sible therapeutic target. J Mol Endocrinol 2011;47:R25–41.

4. Taylor BS, Schultz N, Hieronymus H, Gopalan A, Xiao Y, Carver BS,et al. Integrative genomic profiling of human prostate cancer. CancerCell 2010;18:11–22.

5. Yuan X, Balk SP. Mechanisms mediating androgen receptor reacti-vation after castration. Urol Oncol 2009;27:36–41.

6. Labrie F. Blockade of testicular and adrenal androgens in prostatecancer treatment. Nat Rev Urol 2011;8:73–85.

7. Reid AH, Attard G, Barrie E, de Bono JS. CYP17 inhibition as ahormonal strategy for prostate cancer. Nat Clin Pract Urol 2008;5:610–20.

8. Cai C, Chen S, Ng P, Bubley GJ, Nelson PS, Mostaghel EA, et al.Intratumoral de novo steroid synthesis activates androgen receptor incastration-resistant prostate cancer and is upregulated by treatmentwith CYP17A1 inhibitors. Cancer Res 2011;71:6503–13.

9. StanbroughM, Bubley GJ, Ross K, Golub TR, Rubin MA, Penning TM,et al. Increased expression of genes converting adrenal androgens totestosterone in androgen-independent prostate cancer. Cancer Res2006;66:2815–25.

10. Hofland J, van Weerden WM, Dits NF, Steenbergen J, van LeendersGJ, Jenster G, et al. Evidence of limited contributions for intratumoralsteroidogenesis in prostate cancer. Cancer Res 2010;70:1256–64.

11. Brozic P, Turk S, Rizner TL, Gobec S. Inhibitors of aldo-keto reduc-tases AKR1C1-AKR1C4. Curr Med Chem 2011;18:2554–65.

12. Bauman DR, Steckelbroeck S, Penning TM. The roles of aldo-ketoreductases in steroid hormone action. Drug News Perspect 2004;17:563–78.

13. Culig Z, Comuzzi B, Steiner H, Bartsch G, Hobisch A. Expression andfunction of androgen receptor coactivators in prostate cancer.J Steroid Biochem Mol Biol 2004;92:265–71.

14. ChangCY,McDonnell DP. Androgen receptor-cofactor interactions astargets for new drug discovery. Trends Pharmacol Sci 2005;26:225–8.

15. Narayanan R, Yepuru M, Szafran AT, Szwarc M, Bohl CE, Young NL,et al. Discovery and mechanistic characterization of a novel selectivenuclear androgen receptor exporter for the treatment of prostatecancer. Cancer Res 2010;70:842–51.

16. Yang CH, Yue J, Fan M, Pfeffer LM. IFN induces miR-21 through asignal transducer and activator of transcription 3-dependent pathwayas a suppressive negative feedback on IFN-induced apoptosis. Can-cer Res 2010;70:8108–16.

17. YangCH,YueJ,Pfeffer SR,HandorfCR,Pfeffer LM.MicroRNAmiR-21regulates the metastatic behavior of B16melanoma cells. J Biol Chem2011;286:39172–8.

18. Narayanan R, Coss CC, Yepuru M, Kearbey JD, Miller DD, Dalton JT.Steroidal androgens and nonsteroidal, tissue-selective androgenreceptor modulator, S-22, regulate androgen receptor function

through distinct genomic and nongenomic signaling pathways. MolEndocrinol 2008;22:2448–65.

19. Saint-Lu N, Oortwijn BD, Pegon JN, Odouard S, Christophe OD, deGroot PG, et al. Identification of galectin-1 and galectin-3 as novelpartners for von Willebrand factor. Arterioscler Thromb Vasc Biol2012;32:894–901.

20. Soderberg O, Gullberg M, Jarvius M, Ridderstrale K, Leuchowius KJ,Jarvius J, et al. Direct observation of individual endogenous proteincomplexes in situ by proximity ligation. Nat Methods 2006;3:995–1000.

21. Narayanan R, Mohler ML, Bohl CE, Miller DD, Dalton JT. Selectiveandrogen receptor modulators in preclinical and clinical development.Nucl Recept Signal 2008;6:e010.

22. Agoulnik IU, Vaid A, Nakka M, Alvarado M, Bingman WE III, Erdem H,et al. Androgens modulate expression of transcription intermediaryfactor 2, an androgen receptor coactivator whose expression levelcorrelates with early biochemical recurrence in prostate cancer. Can-cer Res 2006;66:10594–602.

23. QiuW, ZhouM, Labrie F, Lin SX. Crystal structures of the multispecific17beta-hydroxysteroid dehydrogenase type 5: critical androgen reg-ulation in human peripheral tissues. Mol Endocrinol 2004;18:1798–807.

24. Byrns MC, Penning TM. Type 5 17beta-hydroxysteroid dehydroge-nase/prostaglandin F synthase (AKR1C3): role in breast cancer andinhibition by non-steroidal anti-inflammatory drug analogs. Chem BiolInteract 2009;178:221–7.

25. Penning TM, Jin Y, Steckelbroeck S, Lanisnik Rizner T, Lewis M.Structure-function of human3alpha-hydroxysteroid dehydrogenases:genes and proteins. Mol Cell Endocrinol 2004;215:63–72.

26. TenniswoodM, Bird CE, Clark AF. The role of androgen metabolism inthe control of androgen action in the rat prostate. Mol Cell Endocrinol1982;27:89–96.

27. Saijo K, Collier JG, Li AC, Katzenellenbogen JA, Glass CK. An ADIOL-ERbeta-CtBP transrepression pathwaynegatively regulatesmicroglia-mediated inflammation. Cell 2011;145:584–95.

28. Yap TA, Zivi A, Omlin A, de Bono JS. The changing therapeuticlandscape of castration-resistant prostate cancer. Nat Rev Clin Oncol2011;8:597–610.

29. Chang KH, Li R, Papari-Zareei M, Watumull L, Zhao YD, Auchus RJ,et al. Dihydrotestosterone synthesis bypasses testosterone to drivecastration-resistant prostate cancer. Proc Natl Acad Sci U S A2011;108:13728–33.

30. Mohler JL, Titus MA, Bai S, Kennerley BJ, Lih FB, Tomer KB, et al.Activation of the androgen receptor by intratumoral bioconversion ofandrostanediol to dihydrotestosterone in prostate cancer. Cancer Res2011;71:1486–96.

31. Mohler JL, TitusMA,Wilson EM. Potential prostate cancer drug target:bioactivation of androstanediol by conversion to dihydrotestosterone.Clin Cancer Res 2011;17:5844–9.

32. Desmond JC, Mountford JC, Drayson MT, Walker EA, Hewison M,Ride JP, et al. The aldo-keto reductase AKR1C3 is a novel suppressorof cell differentiation that provides a plausible target for the non-cyclooxygenase-dependent antineoplastic actions of nonsteroidalanti-inflammatory drugs. Cancer Res 2003;63:505–12.






2013;19:5613-5625. Published OnlineFirst August 30, 2013.Clin Cancer Res Muralimohan Yepuru, Zhongzhi Wu, Anand Kulkarni, et al. GrowthReceptor-Selective Coactivator that Promotes Prostate Cancer Steroidogenic Enzyme AKR1C3 Is a Novel Androgen

Updated version

10.1158/1078-0432.CCR-13-1151doi:

Access the most recent version of this article at:

Material

Supplementary

http://clincancerres.aacrjournals.org/content/suppl/2013/09/09/1078-0432.CCR-13-1151.DC1

Access the most recent supplemental material at:

Cited articles

http://clincancerres.aacrjournals.org/content/19/20/5613.full#ref-list-1

This article cites 32 articles, 13 of which you can access for free at:

Citing articles

http://clincancerres.aacrjournals.org/content/19/20/5613.full#related-urls

This article has been cited by 10 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/19/20/5613To 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-13-1151

http://clincancerres.aacrjournals.org/content/suppl/2013/09/09/1078-0432.CCR-13-1151.DC1

http://clincancerres.aacrjournals.org/content/19/20/5613.full#ref-list-1

http://clincancerres.aacrjournals.org/content/19/20/5613.full#related-urls

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

mailto:[email protected]

http://clincancerres.aacrjournals.org/content/19/20/5613


steroidogenic enzyme akr1c3 is a novel androgen receptor … · human cancer biology steroidogenic...

Documents