Elevated LIM Kinase 1 in Nonmetastatic Prostate Cancer

Published OnlineFirst October 24, 2014; DOI: 10.1158/1535-7163.MCT-14-0447

Cancer Biology and Signal Transduction Molecular Cancer Therapeutics Elevated LIM Kinase 1 in Nonmetastatic Prostate Cancer Reﬂects Its Role in Facilitating Androgen Receptor Nuclear Translocation Katerina Mardilovich1, Mads Gabrielsen1, Lynn McGarry1, Clare Orange2, Rachana Patel1, Emma Shanks1, Joanne Edwards3, and Michael F. Olson1

Abstract

Prostate cancer affects a large proportion of the male popula- inhibiting proliferation and increasing apoptosis in androgen- tion, and is primarily driven by androgen receptor (AR) activity. dependent prostate cancer cells more effectively than in androgen- First-line treatment typically consists of reducing AR signaling by independent prostate cancer cells. LIMK inhibition blocked hormone depletion, but resistance inevitably develops over time. ligand-induced AR nuclear translocation, reduced AR protein One way to overcome this issue is to block AR function via stability and transcriptional activity, consistent with its effects on alternative means, preferably by inhibiting protein targets that proliferation and survival acting via inhibition of AR activity. are more active in tumors than in normal tissue. By staining Furthermore, inhibition of LIMK activity increased aTubulin prostate cancer tumor sections, elevated LIM kinase 1 (LIMK1) acetylation and decreased AR interactions with aTubulin, indi- expression and increased phosphorylation of its substrate Coﬁlin cating that the role of LIMK in regulating microtubule dynamics were found to be associated with poor outcome and reduced contributes to AR function. These results indicate that LIMK survival in patients with nonmetastatic prostate cancer. A LIMK- inhibitors could be beneﬁcial for the treatment of prostate cancer selective small molecule inhibitor (LIMKi) was used to determine both by reducing nuclear AR translocation, leading to reduced whether targeted LIMK inhibition was a potential prostate cancer proliferation and survival, and by inhibiting prostate cancer cell therapy. LIMKi reduced prostate cancer cell motility, as well as dissemination. Mol Cancer Ther; 14(1); 246–58. 2014 AACR.

Introduction (CRPC) that may quickly progress to metastatic disease with high mortality rates (4). Therefore, a major goal is to identify potential Prostate cancer is the most commonly diagnosed malignancy targets for the development of prostate cancer therapies that target and second leading cause of cancer-related deaths in American AR function in a hormone-independent manner. Such treatments men (1). At the molecular level, prostate cancer development and might not only delay the progression of prostate cancer to CRPC, progression are driven primarily by activity of the androgen but could possibly also be used for the treatment of CRPC, which receptor (AR), a steroid hormone receptor typically localized in maintains and relies upon active AR (4). the cytoplasm in the absence of hormone stimulation (2). In the Targeting the microtubule cytoskeletal network is one approach presence of ligand, ARs translocate to the nucleus to activate the that has been used to achieve the goal of reducing AR signaling. transcription of target genes that control cell-cycle progression, Docetaxel, a microtubule-stabilizing drug commonly used for the cell growth, and survival. As a result, the first line of therapy in treatment of prostate cancer, has been shown to exert its cytotoxic prostate cancer has been to decrease AR activity by hormone effect on prostate cancer cells by inhibiting microtubule-mediated depletion (3). Unfortunately, hormone-ablation therapy often AR nuclear translocation in addition to its direct antimitotic leads to the development of castration-resistant prostate cancer activity (5, 6). Two major issues with docetaxel treatment are that resistance develops over time, and its general antimitotic and microtubule-stabilizing actions result in strong side effects, 1Beatson Institute for Cancer Research, Glasgow, United Kingdom. including alopecia, neutropenia, and anemia. Therefore, an 2Pathology Department, Division of Cancer Sciences and Molecular appealing objective for future prostate cancer drug development Pathology, Western Infirmary, University of Glasgow, Glasgow, United is to identify alternative microtubule regulators, which if inhibited 3 Kingdom. Institute of Cancer Sciences, College of Medical,Veterinary would affect AR function with low nonspecific cytotoxicity. In and Life Sciences, University of Glasgow, Glasgow, United Kingdom. particular, if this target was more active in prostate cancer, its Note: Supplementary data for this article are available at Molecular Cancer inhibition would improve drug selectivity for prostate cancer Therapeutics Online (http://mct.aacrjournals.org/). tumors over normal tissue and contribute to a greater therapeutic Current address for Clare Orange: Department of Pathology, Laboratory Med- window. icine Building, Southern General Hospital, Glasgow, UK. Although best known as regulators of actin–myosin cytoskel- Corresponding Author: M.F. Olson, The Beatson Institute for Cancer Research, etal dynamics (7), LIM kinases 1 and 2 (LIMK1 and LIMK2) also Garscube Estate, Switchback Road, Glasgow, Scotland G61 1BD, UK. Phone: contribute to the regulation of the microtubule cytoskeleton (8– 44-141-330-3654; Fax: 44-141-942-6521; E-mail: [email protected] 10). LIM kinases are highly homologous serine/threonine kinases doi: 10.1158/1535-7163.MCT-14-0447 that are activated by RhoA/ROCK, Rac/PAK, and Cdc42/MRCK 2014 American Association for Cancer Research. signaling pathways (7). The most well-characterized LIMK

246 Mol Cancer Ther; 14(1) January 2015

LIM Kinases in Prostate Cancer

substrates are Cofilin proteins, which are inhibited for their actin- Cell lines and antibodies severing activities when phosphorylated on serine 3 (11). There Cell lines were cultured in RPMI media, supplemented with have been previous reports of elevated LIMK1 expression in 10% FBS and 2 mmol/L L-glutamine (GIBCO). LNCaP-AI cells prostate cancer (12–14), in which it was postulated to have a were cultured in phenol red-free RPMI with 10% charcoal-stripped role in promoting metastasis (15). However, there have been no serum (CSS) and 2 mmol/L L-glutamine (GIBCO). The LNCaP, previous studies that systematically evaluated the expression DU145, PC3, RWPE-1 cell lines were from the ATCC, LNCaP-AI levels of LIMK1, LIMK2, or phosphorylation of their common was made in the laboratory of Hing Leung (Beatson Institute for substrate Cofilin as an indicator of kinase activity in primary Cancer Research, Glasgow, UK) and gifted to us, CWR22 cells were prostate cancer tumor samples accompanied by analysis of pros- from Thomas Pretlow (Case Western Reserve University, Cleve- tate cancer clinical outcomes. land OH). All cell lines were obtained at the beginning of the study We undertook immunohistochemical analysis of a prostate in April 2012 and authenticated using the GenePrint 10 system cancer tissue microarray (TMA) comprised of 164 primary pros- STR multiplex assay (Promega) that amplifies 9 tetranucleotide tate cancer and 23 benign hyperplasia samples from 94 individual repeat loci and Amelogenin gender determining marker. Antibo- patients (16), and identified significant associations of elevated dies: Santa Cruz Biotechnology, AR (N-20 and sc-816), LIMK2 (H- LIMK1 expression and phosphorylation of nuclear Cofilin with 78 and sc-5577); Cell Signaling Technology, p-Cofilin (Ser3, reduced survival of patients with nonmetastatic prostate cancer. 3311), LIMK1 (3842); Abcam, Cofilin (ab54532), LIMK1 Moreover, elevated levels of LIMK1 and cytoplasmic phospho- (ab55414); Sigma, aTubulin (clone DM1A, T9026); Millipore, Cofilin were both associated with significantly higher lympho- MMP-1 (04–1112); Leica, MMP-10 (NCL-MMP10); Novus Bio- vascular invasion. To pharmacologically evaluate whether LIMK logicals, acetylated-a Tubulin (NB600-567). could be a potential prostate cancer drug target, we tested a potent and selective LIMK inhibitor (LIMKi; ref. 17). LIMK inhibition Cytotoxicity assays reduced prostate cancer cell motility, suggesting that blocking Cells were plated in 96-well plates at 2,000 cells per well in LIMK activity could be beneficial as an antimetastatic therapeutic triplicate and treated the next day without changing the media. target in prostate cancer. Interestingly, we observed a cytotoxic Drugs were serial diluted in DMSO, before diluting equal amounts effect of LIMK inhibition that was significantly greater in andro- in media at 2 of the final concentration. Then 100 mLofdrug- gen-dependent prostate cancer cells than in androgen-indepen- containing media was added to each well that already contained dent cells. Treatment of androgen-dependent prostate cancer cells 100 mL of cells. Cells were treated for 72 hours, fixed with 4% with LIMKi reduced AR nuclear translocation and transcriptional paraformaldehyde, and stained with 250 ng/mL DAPI. Plates were activity by altering microtubule dynamics that facilitate AR inter- imaged on a High Content Imaging Operetta system (PerkinElmer) actions with aTubulin, thus inhibiting cell proliferation. There- and nuclei in each well were quantified using Harmony High fore, in addition to a potential role in promoting metastasis, Content Imaging and Analysis Software (PerkinElmer). For cyto- changes in LIMK1 and LIMK2 expression and/or activity might toxicity of RWPE-1 cells, CellTiter-Glo (Promega) Luminescent Cell contribute to AR function in prostate cancer via regulation of Viability assay was used according to the manufacturer's instruc- microtubule cytoskeletal dynamics. These results justify further tions. The effect of each treatment was calculated as the percentage investigation of LIM kinases as potential targets for prostate cancer of change in cell number relative to DMSO-treated control. EC50 therapy. values were calculated from dose–response curves, constrained between 0% and 100%, using Prism 5 (GraphPad) software. Materials and Methods Caspase 3/7 activity TMA and immunohistochemistry Cells were plated and treated as for the cytotoxicity assay, or The prostate cancer TMA, previously described in (16), was transfected with siRNA as in colony formation assay, and plated in comprised of primary prostate cancer tumor samples. Within the 96-well plates. Caspase activity was measured 72 hours after inhib- TMA cohort, 49 samples had available metastasis data, which itor treatment or siRNA transfection using the Caspase-Glo3/7 Assay fi fi were typically identi ed by a bone scan. Survival was de ned as system (Promega) following the manufacturer's instructions. disease-specific, patients who died from intercurrent disease were censored in the analysis. Immunohistochemical staining of TMA Sub-G1 quantification slides was performed as described previously (18), using anti- Cells were plated in 6-well plates at 5 105 cells per well and bodies against LIMK1, LIMK2, and p-Cofilin (18, 19). TMA slides treated the next day with DMSO or 10 mmol/L LIMKi for 24 hours. were scanned and staining intensities were scored using the The percentage of cells with sub-G1 DNA content were measured SlidePath application (Leica Biosystems). The staining was scored and analyzed as described previously (20). low if the sample histoscore was below or equal to the median histoscore for the entire cohort, or high, if above. siRNA knockdown and colony formation assay LNCaP or CWR22 cells were transiently transfected with ON- Statistical analysis TARGETplus SMART pool siRNAs (Thermo Scientific) against Survival differences were determined using the log-rank (Man- LIMK1 or LIMK2, or a nontargeting control (NTC) siRNA, using tel–Cox) test. The Mann–Whitney test was used to compare DreamFect Gold transfection reagent (OZ Biosciences) and plated correlation between protein expression and lymphovascular inva- in 24-well plates at 104 cells per well in duplicate. Twenty-four sion, in Statistical Package for Social Sciences software (SPSS, hours after transfection, media were changed to 1 mL per well of Version19). All other indicated statistical analyses were per- growing media and cells were incubated for 24 hours for lysis and formed in Prism5 (GraphPad) software. The F test was used to knockdown analysis by Western blotting, or incubated for 7 days compare LIMKi EC50 values. for colony formation. To quantify colony formation, cells were

www.aacrjournals.org Mol Cancer Ther; 14(1) January 2015 247

Mardilovich et al.

A LIMK1 LIMK2 Nuclear p-Cofilin Cytoplasmic p-Cofilin

Low

100 100 100 100 µm µm µm µm

100 µm 100 µm 100 µm 100 µm

High

100 100 100 100 µm µm µm µm

100 µm 100 µm 100 µm 100 µm B Nonmetastatic LIMK1 LIMK2 Nuclear p-Cofilin Cytoplasmic p-Cofilin 1.0 1.0 1.0 1.0 Low Low Low Low 0.8 0.8 0.8 0.8 High High High High 0.6 0.6 0.6 0.6

0.4 0.4 0.4 0.4

0.2 0.2 0.2 0.2

Survival rate

Survival rate 0.0 P = 0.035 Log-rank 0.0 0.0 P = 0.034 Log-rank 0.0 0.0 5.0 10.0 15.0 20.0 0.0 5.0 10.0 15.0 20.0 0.0 5.0 10.0 15.0 20.0 0.0 5.0 10.0 15.0 20.0 Years Years Years C Metastatic Years LIMK1 LIMK2 Nuclear p-Cofilin Cytoplasmic p-Cofilin 1.0 1.0 1.0 1.0 Low Low Low Low 0.8 0.8 0.8 High High High 0.8 High 0.6 0.6 0.6 0.6

0.4 0.4 0.4 0.4

0.2 0.2 0.2 0.2

Survival rate

Survival rate 0.0 0.0 P = 0.048 Log-rank 0.0 0.0 0.0 2.5 5.0 7.5 10.0 12.5 0.0 2.5 5.0 7.5 10.0 12.5 0.0 2.5 5.0 7.5 10.0 12.5 0.0 2.5 5.0 7.5 10.0 12.5 D Years Years Years Years LIMK1 LIMK2 Nuclear p-Cofilin Cytoplasmic p-Cofilin 200 100 80 140 80 150 60 120 60 100 100 40 80 40 60 50 20 20 40 P = 0.042 20 0 0 0 0 P = 0.027 no yes no yes no yes no yes Lymphovascular Lymphovascular Lymphovascular Lymphovascular invasion invasion invasion invasion E 2.0 P = 0.03 F P < 0.0001 1.52 1.5 1.0 1.32 0.5 1.15

intensity 0.0 LIMK1 copy number units –0.5 1.00 log2 median-centeredlog2 Prostate Prostate Adeno –1.0 gland carcinoma Prostate carcinoma

248 Mol Cancer Ther; 14(1) January 2015 Molecular Cancer Therapeutics

LIM Kinases in Prostate Cancer

fixed in methanol and stained with 0.2% (v/v) Crystal Violet Real-time PCR solution. Staining was quantified with an ODYSSEY infrared CWR22 cells were pretreated with 3 mmol/L LIMKi, or DMSO imaging system (LI-COR). vehicle control in CSS RPMI without phenol-red for 24 hours, then treated with 1 nmol/L DHT in the presence of the drugs for Coimmunoprecipitation and Western blotting additional 24 hours. RNA was extracted using the RNeasy Mini Kit LNCaP cells were incubated in phenol red-free RPMI with CSS (Qiagen) and reverse-transcribed using the QuantiTect Reverse for 24 hours, and then treated with LIMKi, docetaxel, or DMSO in Transcription Kit (Qiagen) according to the manufacturer's the presence of 1 nmol/L dihydrotestosterone (DHT) for 24 hours. instructions. PSA gene expression was quantified using the DyNA- Cells were lysed with 500 mL of lysis buffer [1 TBS, 1% Triton-X, 1 mo HS SYBR Green qPCR Kit (Thermo Scientific) and the Applied nmol/L EDTA, 0.2 mmol/L Na3VO4, 20 mmol/L NaF, 1 mmol/L Biosystems 7500 Fast Real-Time System (Life Technologies) using 0 0 PMSF, and cOmplete Mini protease inhibitor cocktail (Roche)] primers against PSA (F: 5 -GCAGCATTGAACCAGAGGAG-3 ,R: 0 0 0 per 10-cm plate. Lysates were cleared by 10-minute centrifugation 5 -AGAACTGGGGAGGCTTGAGT-3 ) and 18S (F: 5 -GTAACCC- 0 0 0 and precleared with Protein-A agarose beads (Life Technologies). GTTGAACCCCATT-3 ,R:5-CCATCCAATCGGTAGTAGCG-3 )as After preclearing, beads were removed, and then lysates were an internal control. incubated with anti-AR antibody at 1:25 dilution for 2 hours at 4C, then 30 mL Protein-A agarose beads were added for 1 Immunofluorescence additional hour. Beads were washed three times with TBS and LNCaP cells grown on coverslips were incubated in phenol red- boiled with prewarmed 1% SDS. To quantify the amount of free RPMI with CSS for 24 hours, then treated with indicated drugs immunoprecipitated protein, immunoprecipitated samples and for 24 hours before 2-hour stimulation with 1 nmol/L DHT. Cells total cell lysates (1% of immunoprecipitated) were boiled with were fixed and stained with specific antibodies as described pre- loading dye and Western blotted (20) with anti-AR and aTubulin viously (22). Images were taken using a Zeiss710 laser-scanning antibodies. Quantification of Western blots was performed direct- confocal microscope and analyzed with ZEN2010 (Zeiss) software. ly without signal amplification or X-ray film using infrared emit- ting secondary antibodies and detection with an ODYSSEY infra- Nuclear AR quantification red imaging system (LI-COR). Nuclear AR (volume per nucleus and the percentage of cells with nuclear AR) were quantified using Volocity 3D Image Anal- LIMK1 antibody validation ysis software (PerkinElmer). LIMK1 peptide (ab158818 (Abcam) was spotted on PVDF membranes at 50, 25, or 5 ng in duplicate, and membranes were Acetylated tubulin quantification blocked and blotted as for Western Blotting with LIMK1 primary Immunofluorescence staining intensity of acetylated aTubulin antibody ab55414 (Abcam) or LIMK1 antibody that has been was quantified in ImageJ software using fixed intensity threshold, preincubated with 4 mg/mL LIMK1 peptide or nonspecific control and normalized to total aTubulin immunofluorescence intensity Cofilin peptide (amino acids 1–20). After incubation, membrane levels. was developed as for a Western blot analysis. Immunohistochem- istry was performed as described above using LIMK1 antibody Cell fractionation with or without preincubation with LIMK1 competitor peptide or LNCaP cells were incubated in phenol-red–free RPMI with CSS Cofilin peptide as a nonspecific control. for 24 hours, then treated with 1 nmol/L DHT in the presence of indicated drugs for 24 hours. After treatment, cells were lysed and Luciferase assay fractionated using the NE-PER Nuclear and Cytoplasmic Extrac- fi CWR22 cells were plated in 96-well plates at 2 104 cells per tion Kit (Thermo Fisher Scienti c), according to the manufacturer's instructions. well in triplicates and transfected the next day with p(ARE)3-Luc (21) and CMV (cytomegalovirus)-driven Renilla luciferase plasmids at the 10:1 ratio using X-tremeGENE HP transfection reagent Migration assays (Roche) following the manufacturer's instructions. Next day, cells PC3 and DU145 cells were plated in 96-well plates in quad- were incubated in phenol red-free RPMI with CSS with DHT or ruplicates at 4 104 cells per well and treated the next day as in vehicle control (ethanol), in the presence of indicated drugs for 24 cytotoxicity assay, for 24 hours. After treatment, cell monolayers hours. Luciferase activity was measured using the Dual-Luciferase were scratched, washed, and monitored on the INCUCYTE Kinetic Reporter Assay System (Promega), following the manufacturer's Imaging System (ESSEN BioScience) by continuous imaging every protocol. Firefly Luciferase measurements were normalized to 2 hours for 24 hours. The percentage of wound confluence was Renilla Luciferase values for transfection control and plotted as quantified using INCUCYTE software analysis (ESSEN BioSci- fold change of DMSO-treated control. ence). LNCaP cells were plated in glass-bottom 6-well dishes,

Figure 1. Association of LIMK1, LIMK2, and p-Cofilin with prostate cancer patient outcome. A, human prostate cancer samples were stained with antibodies against LIMK1, LIMK2, or phosphorylated serine 3 Cofilin (p-Cofilin). Representative images of low and high staining for each protein are shown at 10 magnification, inset at 20 magnification; scale bar, 100 mm. B, Kaplan–Meier survival curves showing association of LIMK1, LIMK2, nuclear, and cytoplasmic p-Cofilin levels with patient survival with nonmetastatic; and C, metastatic prostate cancer (blue line, low protein levels; green line, high protein levels). D, expression levels of LIMK1, LIMK2, nuclear, and cytoplasmic p-Cofilin in prostate cancer tumors scored positive or negative for lymphovascular invasion. E, LIMK1 mRNA expression in human prostate cancer tumors and normal gland tissue. Statistical significance was tested by the unpaired Student t test with Welch correction. F, LIMK1 copy- number variations in normal prostate and adenocarcinoma samples determined by TCGA (http://tcga-data.nci.nih.gov/tcga/). Statistical significance was tested by the unpaired Student t test.

www.aacrjournals.org Mol Cancer Ther; 14(1) January 2015 249

Mardilovich et al.

allowed to attach overnight, then medium was changed to CSS patient age, Gleason score, PSA elevation or relapse (Supple- supplemented phenol-red–free RPMI for 24 hours, followed mentary Table S6). by treatment with 1 nmol/L DHT with 5 mmol/L LIMKi or Analysis of publicly available gene-expression data using Onco- DMSO and imaged at three images per well on the Nikon time- mine (29) revealed significantly elevated LIMK1 mRNA expres- lapse imagining system for 20 hours. Cell velocities were sion levels in prostate carcinoma relative to normal gland tissue calculated using ImageJ software for 30 cells per well within (Fig. 1E; ref. 30). Furthermore, analysis using the cBio Cancer each experiment. Genomics Portal (31) of 85 prostate adenocarcinoma tumor samples revealed significantly increased LIMK1 expression or Results amplification in 14 cases (16%), with 5 of 12 patients (42%) having elevated LIMK1 expression undergoing biochemical recur- LIMK expression and activity in prostate cancer rence in contrast with 16 of 68 patients (24%) without alterations LIMK1 has previously been reported to be elevated in limited in LIMK1 expression (32). In addition, disease-free survival samples of human prostate cancer (12–14); however, its associ- trended toward being worse for patients with altered LIMK1 ation with patient outcome had not been determined. In addition, expression (Supplementary Fig. S1B). Analysis of LIMK1 gene neither LIMK2 expression nor phosphorylation of the common copy number by The Cancer Genome Atlas (TCGA; http://tcga- substrate Cofilin had been previously characterized in prostate data.nci.nih.gov/tcga/) indicated that there was a significant cancer. Therefore, we investigated how LIMK1 and LIMK2 expres- increase in prostate adenocarcinoma samples relative to normal sion and activity varied in prostate cancer, and determined how prostate tissue (Fig. 1F). These results are consistent with increased observed variations were associated with patient outcomes using a mRNA contributing to elevated LIMK1 protein in prostate cancer. TMA comprised of 164 primary prostate cancer samples and 23 Together, our findings indicate that elevated LIMK expression and benign hyperplasia samples from 92 individual patients (demo- activity are associated with early-stage prostate cancer growth and graphic, clinicopathologic, and outcome characteristics of the progression. patients detailed in Supplementary Table S1; ref. 16). Using antibodies validated by peptide competition for LIMK1 (Sup- plementary Fig. S1A) or by staining tissues from knockout mice LIMK inhibitor decreases prostate cancer cell motility for LIMK2 (18) or Serine 3 phosphorylated Cofilin (p-Cofilin; Given the previous reports of elevated LIMK1 expression in ref. 19), TMA samples were stained, scanned, and scored for metastatic prostate cancer (12–15) and our findings that blocking staining intensities using an automated algorithm (23). Stain- LIMK activity reduces breast cancer cell invasiveness (33), we ing intensities of each protein were correlated with the histo- sought to determine whether pharmacologic inhibition of LIMK pathologic data of the tumor samples. Examples of low and activity reduces prostate cancer cell motility. We initially com- high staining for LIMK1, LIMK2, nuclear, and cytoplasmic p- pared LIMK1, LIMK2, and Cofilin Serine3 phosphorylation in Cofilin are shown in Fig. 1A. There was a significant association several prostate cancer cells lines, and found that Cofilin phos- between high LIMK1 expression and poor survival in patients phorylation was highest in parental androgen-dependent LNCaP diagnosed with nonmetastatic prostate cancer (P ¼ 0.035; Fig. cells, with lower levels in an androgen-independent LNCaP 1B; Supplementary Table S2), but not in the metastatic disease variant (LNCaP-AI), derived by continuous culture in hor- group (Fig. 1C; Supplementary Table S2). Analysis of LIMK2 mone-depleted media (16), as well as AR-negative PC3 and levels showed a similar trend of high LIMK2 expression asso- DU145 cells (Fig. 2A). Combined LIMK1 and LIMK2 levels ciated with poor patient survival in the nonmetastatic group (P were highest in LNCaP cells, with lower levels in the androgen- ¼ 0.151; Fig. 1B; Supplementary Table S3). However, high independent LNCaP-AI, PC3, and DU145 prostate cancer cells. LIMK2 was significantly associated with increased survival in Treatment of LNCaP cells with 1 or 10 nmol/L DHT did not the metastatic disease group (P ¼ 0.048; Fig. 1C; Supplemen- further increase LIMK1 or LIMK2 protein levels (Supplementary tary Table S3). Because Cofilin has both nuclear import and Fig. S1C). export signals to facilitate nuclear-cytoplasmic shuttling (24), To investigate the role of LIMK activity in prostate cancer we analyzed nuclear or cytoplasmic p-Cofilin levels in the TMA cell motility, we used a potent selective small molecule inhib- samples. Similar to LIMK1, high nuclear p-Cofilin levels itor (N-{5-[2-(2,6-Dichloro-phenyl)-5-difluoromethyl-2H-pyr- showed a significant association with poor patient survival in azol-3-yl]-thiazol-2-yl}-isobutyramide [compound 3 in (ref. 17); the nonmetastatic group (P ¼ 0.034; Fig. 1B; Supplementary hereafter termed LIMKi] that equipotently inhibits LIMK1 and Table S4). Although the role of cytoplasmic Cofilin has largely LIMK2. We confirmed that LIMKi effectively reduced p-Cofilin in been associated with regulation of cytoskeleton dynamics (25), LNCaP cells after 24-hour treatment (Fig. 2B), as we previously nuclear Cofilin may contribute to nuclear actin regulation, gene observed for other cell lines, including MDAMB231 breast cancer transcription and mitosis (26, 27). Together, these results cells (22), NMuMG mouse mammary epithelial cells (34), and indicate that elevated LIMK expression and activity are associ- U2OS human osteosarcoma cells (20). ated with increased mortality in nonmetastatic prostate cancer. As a major regulator of cytoskeleton dynamics, LIMK1 has Although LIMK1 and Cofilin phosphorylation was not associ- been implicated in prostate cancer cell migration and invasion ated with patient mortality in metastatic prostate cancer (Sup- (12, 14, 15). Given that maximal inhibition of Cofilin phos- plementary Tables S2, S4, and S5), elevated LIMK1 and cyto- phorylation was achieved at 10 mmol/L LIMKi (Fig. 2B), we plasmic p-Cofilin were significantly associated with increased tested this concentration on DU145 and PC3 cell migration. lymphovascular invasion (Fig. 1D), a clinical characteristic of Cells were plated in a dense monolayer in 96-well plates, then more aggressive tumors and potential marker of progression to treated the next day with 10 mmol/L LIMKi for 24 hours. After metastatic disease (28). There were no significant associations treatment, PC3 cell migration was analyzed in an automated between LIMK1, LIMK1, or Cofilin phosphorylation with scratch wound-healing assay, with images acquired every 2

250 Mol Cancer Ther; 14(1) January 2015 Molecular Cancer Therapeutics

LIM Kinases in Prostate Cancer

A B C 0 hours 24 hours LIMKi (µmol/L) LNCaP LNCaP-AIPC3 DU145 0 0.3 1 3 10 LIMK1

DMSO LIMK2 p-Cofilin 300 µm p-Cofilin Cofilin Cofilin

AR αTubulin

Figure 2. αTubulin LIMKi inhibits migration of prostate 10 µmol/L LIMKi cancer cell lines. A, LNCaP, LNCaP-AI, PC3, and DU145 cell lines were D P = 0.0002 100 DMSO PC3 100 analyzed for protein levels of LIMK1, 10 μmol/L ﬁ LIMK2, and p-Co lin by Western blot LIMKi 80 analysis. B, LNCaP cells were treated 80 with increasing LIMKi concentrations 60 for 24 hours and analyzed by Western 60 blot analysis for changes in p-Coﬁlin 40 levels. C, cells were plated at high 40 density in 96-well plates and treated 20 with DMSO vehicle control or 20

10 mmol/L LIMKi for 24 hours before Wound confluence (%) Wound confluence (%) 0 0 wounding of the cell monolayer. DMSO 10 mmol/L LIMKi Images show confluence mask (black 0 4 8 12 16 20 24 region) for PC3 cells at 0 and 24-hour Time (h) time points. Cell migration was E fi 100 DMSO DU145 100 P = 0.03 quanti ed by measuring the μ percentage of wound confluence (y- 10 mol/L LIMKi axis) over time (x-axis) relative to an 80 80 initial confluence mask for PC3 (D) and DU145 (E) cells. Results show average 60 60 of three independent experiments performed in quadruplicate, SEM. 40 40 Statistical significance was tested by the unpaired Student t test. F, random 20 20 migration of LNCaP cells was Wound confluence (%) Wound confluence (%) determined by single-cell tracking 0 0 over 20 hours using ImageJ. Five 0 4 8 12 16 20 24 DMSO 10 mmol/L LIMKi independent determinations were Time (h) made of the motility of 30 cells in each condition. Boxes are top and bottom quartiles, whiskers indicate minimum F P = 0.03 P = 0.02 and maximum values. Statistical 0.05 analysis was by one-way ANOVA followed by the post hoc Tukey 0.04 multiple comparison test. 0.03

0.02

0.01 Velocity ( m m/min)

0.00 + + − DMSO − + + DHT(1 nmol/L) −− + LIMKi (5 µmol/L) hours for 24 hours (Fig. 2C). After creating an initial mask, Sensitivity of androgen-dependent prostate cancer cells to LIMK wound confluence was measured by determining the percent- inhibition age of wound area that is occupied by cells over time. There was In addition to testing how LIMKi affected DU145 and PC3 a marked effect of LIMKi on 2D migration of PC3 cells over scratch wound closure, we initially tested LNCaP cells but found time (Fig. 2D, left graph) and at the 24-hour experimental that they did not migrate in this assay format. As an alternative, we endpoint (Fig. 2D, right graph). LIMKi had a more moderate used single-cell tracking to measure random migration velocity. inhibitory effect on the 2D migration of DU145 cells over time Treatment with 1 nmol/L DHT significantly increased migration (Fig. 2E, left graph), which was significant at the 24-hour velocity, which was significantly reversed by 5 mmol/L LIMKi (Fig. experimental endpoint (Fig. 2E, right graph). These results 2F), indicating a role for LIMK activity in androgen-induced suggest that LIMK inhibition could have the therapeutic benefit motility. Expression of the matrix metalloproteinases MMP-10 of reducing prostate cancer dissemination. (35) and MMP-1 (36), which had previously been implicated in

www.aacrjournals.org Mol Cancer Ther; 14(1) January 2015 251

Mardilovich et al.

A 125 125 125 125 LNCaP 100 100 100 100 LNCaP-AI 75 75 75 75 50 50 50 50

% Survival 25 25 25 25 0 0 0 0 –9 –8 –7 –6 –5 –4 –10 –9 –8 –7 –6 –5 –4 –14 –13 –12 –11–10 –9 –8 –9 –8 –7 –6 –5 –4 Log (LIMKi, mol/L) Log (Adriamycin, mol/L) Log (Actinomycin D, mol/L) Log (Camptothecin, mol/L)

B LNCaP LNCaP-AI C P < 0.05 DMSO P = 0.04 1,500 10 µmol/L G1 400 G1 40 LIMKi 1,500

LNCaP LNCaP-AI 300 S G cells 30 1,000 2 1 1,000 G2 S 20 200

Counts 500 sub-G1 sub-G1 500 activity (%) 10 100 Relative caspase3/7 0 0 % of Sub-G 0 0 0 200 400 600 800 1,000 0 200 400 600 800 1,000 010 2 FL2-A LIMKi (μmol/L) PC3 FL2-A LNCaPCWR2 DU145 LNCaP-AI LNCaP CWR22 D E P < 0.001 P < 0.001 250 LNCaP CWR22 250 200 200 150 NTCLIMK1LIMK2LIMK1+2 NTCLIMK1LIMK2LIMK1+2 150 LIMK1 100 100 LIMK2 activity (%) 50 activity (%) 50 Relative caspase3/7 α 0 Relative caspase3/7 Tubulin 0 C 1 2 2 1 2 2 NT C LIMK LIMK NT 1.5 1.5 LIMK1+ LIMK LIMK siRNA LIMK1+ siRNA 1.0 1.0 F siRNA: NTC LIMK1 LIMK2 LIMK1+2 0.5 0.5

0.0 0.0 Relative LIMK1level Relative LIMK1level

C 2 1 2 LNCaP T K1 K2 + TC K K 1+2 N IM IM K1 N IM IM K L L L L IM IM L L siRNA siRNA

1.2 1.2

1.0 1.0 CWR22 0.8 0.8 0.6 0.6 LNCaP CWR22 0.4 0.4 P < 0.05 0.2 0.2 0.6 0.0 0.0 0.8 Relative LIMK2level Relative LIMK2level 1 2 2 + 0.4 TC K1 K2 TC K K 1 0.6 N IM IM K1+2 N IM IM K L L L L IM IM L L 0.4 0.2 siRNA siRNA 0.2

0.0

Relative cell densityRelative cell 0.0 densityRelative cell

K1 K2 +2 TC K1 K2 TC 1 N M M N IM IM K I I K1+2 L L L L M IM I L L siRNA siRNA

Figure 3. The LIMK inhibitor blocks survival of androgen-dependent prostate cancer cell lines. A, LNCaP (red) and LNCaP-AI (green) cell lines were treated with half-log serial dilutions of LIMKi, adriamycin, actinomycin D, and camptothecin (left to right) for 72 hours. Drug cytotoxicity was measured as the percentage of survival relative to DMSO-treated control (100%). Graphs, combined results of three independent experiments performed in triplicate SEM. B, DNA content of LNCaP or LNCaP-AI cells treated with 10 mmol/L LIMKi (red line) or DMSO vehicle (black line) for 72 hours as analyzed by propidium iodide staining followed by flow cytometry. Graph, the results from four independent experiments SEM. Statistical significance was tested by two-way ANOVA; the P value, significant effect of cell type on drug response. C, caspase3/7 activity for each treatment was measured and normalized to DMSO alone–treated control. (Continued on the following page.)

252 Mol Cancer Ther; 14(1) January 2015 Molecular Cancer Therapeutics

LIM Kinases in Prostate Cancer

prostate tumor growth, was not affected by DHT or LIMKi treat- Table 1. Antiproliferative effect of LIMKi on prostate cancer cell lines. All cell ment (Supplementary Fig. S1D). Interestingly, when LNCaP cells lines were treated with half-log serial dilutions of LIMKi for 72 hours. EC50 values were measured from dose–response curves from three independent were treated with 10 mmol/L LIMKi, we readily observed reduced experiments performed in triplicates LNCaP cell numbers. To quantify the effect of LIMKi on prolif- LIMKi EC50 P value, F test eration, we treated LNCaP or LNCaP-AI cells in 96-well plates Cell line (mmol/L) (compared with LNCaP) with half-log serial dilutions of LIMKi. After 72 hours, surviving LNCaP 1.33 — cells were fixed, stained with DAPI, and nuclei numbers quan- CWR22 6.97 <0.0001 tified with an Operetta High Content Imaging System. LNCaP LNCaP-AI 11.11 <0.0001 cells were significantly more sensitive to LIMKi than LNCaP-AI DU145 11.46 <0.0001 < cells (Fig. 3A, left). However, the two cell lines did not differ in PC3 12.68 0.0001 their sensitivities to adriamycin, actinomycin D, or camptothecin (Fig. 3A), all of which exert their cytotoxicity independent of LIMK inhibition. In addition, 10 mmol/L LIMKi treatment tion, and survival. Given the observed effects of LIMKi on andro- induced a substantial increase in the percentage of LNCaP cells gen-dependent prostate cancer cell number and apoptosis, we with sub-G1 DNA content (i.e., <2N), indicative of the induction sought to determine whether LIMKi treatment affected AR sub- of apoptosis, but not in LNCaP-AI cells (Fig. 3B). Moreover, cellular localization. Following 24-hour incubation in hormone- evaluation of the LIMKi rank order of potency on cell number for depleted media (37) to reduce background AR signaling, LNCaP various prostate cancer cell lines revealed androgen-dependent cells were stimulated with 1 nmol/L DHT for 2 hours. Represen- LNCaP and CWR22 cells to be >2 to 8 times more sensitive tative confocal immunofluorescence images show increased than androgen-independent LNCaP-AI, DU145 and PC3 cells nuclear AR staining following DHT stimulation, quantified either (Table 1). Similarly, using a cell viability assay that measures by determining the percentage of cells with detectable nuclear AR ATP levels, immortalized RWPE-1 normal prostate epithelial (Fig. 4A, top) or by measuring the mean volume of AR staining per cells had an EC50 value for LIMKi of 6.67 mmol/L, similar to the nucleus (Fig. 4A, bottom), both of which were reduced by values determined for androgen-dependent CWR22 cells (Table treatment with 3 mmol/L LIMKi or 4 nmol/L docetaxel, a standard 1). Consistent with these results, activity of the apoptosis exe- prostate cancer chemotherapy drug previously shown to inhibit cutioner caspases 3 and 7 was induced by 10 mmol/L LIMKi in AR nuclear translocation (38). Furthermore, the effect of LIMKi LNCaP and CWR22 cells over 3-fold relative to untreated control and docetaxel on DHT-induced AR nuclear translocation was cells, but not in androgen-independent LNCaP-AI, DU145, and supported by cellular fractionation following the same treatment PC3cells(Fig.3C).Theseresults indicate that androgen-depen- used for immunofluorescence experiments. A representative dent prostate cancer cells, which had the highest relative levels of Western blot analysis shows that DHT-induced nuclear AR accu- LIMK expression and Cofilin phosphorylation, were more sen- mulation was reduced by 1 and 3 mmol/L LIMKi or 4 nmol/L sitive to LIMK inhibition than androgen-independent prostate docetaxel (Fig. 4B). Combined results from three independent cancer cells, suggesting that LIMK activity may contribute to AR experiments revealed that 3 mmol/L LIMKi had a comparable function and activity. inhibitory effect on DHT-induced nuclear accumulation of AR as To validate the on-target effects of LIMKi (17), we analyzed the 4 nmol/L docetaxel (Fig. 4B). Interestingly, inhibition of p-Cofilin effect of individual or combined LIMK1 and LIMK2 siRNA-medi- by 4 nmol/L docetaxel was comparable with that observed for 3 ated knockdown (Fig. 3D) on the induction of apoptosis by mmol/L LIMKi, suggesting that docetaxel inhibits LIMK activity as measuring caspase 3/7 activity. In both LNCaP and CWR22 cells, a direct or indirect off-target, which may additionally contribute LIMK2 or combined LIMK1 þ LIMK2 knockdown induced sig- to its antiproliferative mechanism of action in prostate cancer. nificant caspase activity (Fig. 3E). Furthermore, proliferation of Because of the inhibitory effect of LIMKi on AR nuclear accu- androgen-dependent LNCaP and CWR22 cell lines was strongly mulation, we tested the possibility that LIMK inhibition might inhibited by the simultaneous knockdown of both LIMK1 and affect AR protein stability. LNCaP cells were treated with 3 mmol/L LIMK2 in a colony formation assay (Fig. 3F), similar to the effect LIMKi or vehicle (DMSO) control for 24 hours, followed by of LIMKi. These results indicate that LIM kinases have roles in inhibition of de novo protein translation with 20 ng/mL cyclo- supporting proliferation of androgen-dependent prostate cancer heximide (CHX) for the indicated times. By measuring AR protein cells. levels at each time point by Western blotting, we observed less remaining AR protein in the presence of LIMKi compared with LIMK inhibition targets AR activity in androgen-dependent DMSO-treated controls, consistent with LIMKi treatment reduc- prostate cancer cells ing AR protein stability (Fig. 4C). AR is a steroid hormone receptor that translocates into the To test the effect of LIMKi on AR transcriptional activity, we nucleus upon ligand binding to regulate transcription of down- used a Luciferase reporter construct comprised of three consensus stream target genes that promote prostate cell growth, prolifera- AR-response element (ARE) repeats (21) in a transcription–

(Continued.) Graph, combined results of three independent experiments performed in triplicate SEM. Statistical significance was tested by one-way ANOVA and the post hoc Newman–Keuls multiple comparison test, groups of conditions differing significantly from each other as indicated. D, LNCaP or CWR22 cells were transfected with NTC, LIMK1, LIMK2, or LIMK1 þ LIMK2 siRNA oligos as indicated, and Western blotted after 48 hours. Graphs, combined results of three independent experiments performed in duplicate SEM. E, caspase3/7 activity for each treatment was measured and normalized to untransfected control cells. Graph shows combined results of three independent experiments performed in triplicate SEM. Statistical significance was tested by one-way ANOVA and the post hoc Dunnett multiple comparison test, conditions differing significantly from NTC as indicated. F, colony formation of LNCaP or CWR22 cells after 7 days following transfection with indicated siRNA oligos. Graphs, combined results of three independent experiments performed in duplicate SEM. Statistical significance was tested by one-way ANOVA and the post hoc Dunnett multiple comparison test, groups of conditions differing significantly from each other as indicated.

www.aacrjournals.org Mol Cancer Ther; 14(1) January 2015 253

Mardilovich et al.

A DMSO DHT + DMSO DHT + 3 µmol/L LIMKi DHT + 4 nmol/L Dcxl Tubulin 25 P = 0.04 AR 20 15 10 Cells with 5 Nuclear AR (%) 50 µm 50 µm 50 µm 50 µm 0 P = 0.05 60

40 /nucleus) 3 20 ( μ m Mean AR volume Square side = 21.3 µm 0 DMSO + + −− DHT (1 nmol/L) − + ++ LIMKi (3 µmol/L) −−− + Dcxl (4 nmol/L) −−− + B C DMSO 3 µmol/L LIMKi 0 1 3 6 0 1 3 6 CHX (h ) 200 AR AR α Nuclear 150 Tubulin p-Cofilin 100 100 Cofilin

50 P = 0.02

% of Nuclear AR 75 Cell lysate αTubulin 0 −−− + −−− DMSO 50 + + DMSO − + − + +++DHT (1 nmol/L) + +++DHT (1 nmol/L) DMSO −− − −− − LIMKi (µmol/L) 1 3 LIMKi (µmol/L) − −−−1 3 25 3 µmol/L LIMKi − −−− + Dcxl (4 nmol/L) + Dcxl (4 nmol/L)

Remaining AR protein (%) 0 0 2 4 6 Time of CHX treatment (h) D DMSO E 3 mol/L LIMKi 10 mol/L LIMKi 3.5 10 mol/L Bicalutamide 3.0 2.0

2.5 1.5 2.0

1.5 1.0 (fold change) 1.0 0.5 Relative luciferase units 0.5 Relative PSA expression 0.0 0.0 + + − DMSO DHT DHT − + + DHT (1 nmol/L) 0 −− + LIMKi (3 µmol/L)

1 nmol/L DHT 10 nmol/L

Figure 4. Treatment with LIMKi reduces AR function. A, LNCaP cells were stimulated with 1 nmol/L DHT for 2 hours after being hormone starved and treated with DMSO vehicle, 3 mmol/L LIMKi, or 4 nmol/L docetaxel for 24 hours. Cells were fixed and stained with fluorescent antibodies against AR (green), aTubulin (red), or stained with DAPI (blue). Representative confocal immunofluorescence images (top) and 3-dimensional reconstructions from z-stacks (bottom) are shown. The percentage of cells with nuclear AR (top graph) and mean AR volume (mm3) per nucleus (bottom graph) were quantified with Volocity software. Combined results from three independent experiments SEM are shown. Statistical significance was tested by the unpaired Student t test between indicated conditions. B, LNCaP cells were treated as above. Cells were fractionated and nuclear lysates analyzed by Western blot analysis for changes in nuclear AR. Total cell lysates of the same cells were analyzed by Western blot analysis for changes in p-Cofilin. Graph, quantification results of nuclear AR levels normalized to nuclear fraction loading control, plotted as the percentage of change compared with unstimulated cells. Average results of three independent experiments SEM are shown. (Continued on the following page.)

254 Mol Cancer Ther; 14(1) January 2015 Molecular Cancer Therapeutics

LIM Kinases in Prostate Cancer

reporter assay. LNCaP cells could not be adequately transfected quent effects on microtubule dynamics, and by reducing AR for the assay, so androgen-dependent CWR22 cells were used. interaction with aTubulin. Cells were transfected with p(ARE)3–luciferase reporter and Renilla luciferase control plasmids and then treated with 1 or 10 nmol/L DHT for 24 hours in the presence of 3 or 10 mmol/L Discussion LIMKi, DMSO vehicle control, or 10 mmol/L of the antiandrogen In this report, we established a correlation between poor bicalutamide. These experiments revealed that 10 mmol/L LIMKi survival in patients with nonmetastatic prostate cancer and reduced DHT-induced AR transcriptional activity to a similar increased LIMK1 and nuclear phospho-Cofilin levels (Fig. 1B; extent as 10 mmol/L bicalutamide (Fig. 4D). Statistical analysis Supplementary Tables S2 and S4). In addition, elevated LIMK1 by two-way ANOVA revealed a significant effect of LIMKi treat- and phosphorylated cytoplasmic Cofilin were associated with ments on DHT-induced luciferase activity (P ¼ 0.046). Expression increased lymphovascular invasion (Fig. 1D). Associated with of endogenous PSA mRNA was induced by 1 nmol/L DHT these expression patterns, we identified a role for LIM kinases in treatment could be effectively reversed by 3 mmol/L LIMKi (Fig. metastatic prostate cancer cell motility and in promoting AR 4E). Taken together, these experiments indicate that inhibition of function in prostate cancer via their contribution to regulating LIMK activity reduced AR nuclear accumulation, stability, and microtubule dynamics. Intriguingly, inhibition of LIMK activity transcriptional activity in androgen-dependent prostate cancer with a selective small molecule inhibitor had greater cytotoxic cells, which would contribute to the antiproliferative and proa- effects on androgen-dependent than on androgen-independent poptotic effects of LIMKi treatment (Fig. 3). prostate cancer cells (Fig. 3 and Table 1). Treatment of androgen- responsive cells with LIMKi decreased DHT-induced AR nuclear LIMK inhibition affects microtubule stability and AR–aTubulin translocation, protein stability, and transcriptional activity (Fig. interactions 4). Impaired nuclear translocation of AR in the presence of LIMKi AR has been previously shown to interact directly with was likely the product of decreased AR–aTubulin interactions and microtubules, whereas disruption of microtubule dynamics increased aTubulin acetylation, indicative of increased microtu- leads to reduced nuclear AR translocation (38). In addition to bule stability (Fig. 5). Taken together, our results indicate that LIM being a major regulator of actin dynamics, LIMK has also been kinases positively regulate AR function in AR-dependent prostate shown to regulate microtubule dynamics (8–10). Therefore, we cancer, promoting disease development and progression to an tested the effect of LIMKi on the levels and distribution of early locally invasive state. In addition, these studies provide acetylated aTubulin (Ace-aTub), which is associated with pharmacologic evidence, with mechanism-of-action detail, show- increased stabilization and altered dynamics of the microtu- ing that LIMK inhibition has potential as a therapeutic approach bule network (39). To visualize changes in microtubule struc- for the treatment of prostate cancer ture and aTubulin acetylation, we performed confocal immu- The finding of LIMK1 association with poor patient survival in nofluorescence microscopy on LNCaP cells treated with 10 the nonmetastatic prostate cancer group is consistent with a mmol/L LIMKi for 24 hours. Representative Z-plane optical previous observation of correlation between expression of the slices as well as maximum projection images show a noticeable LIMK upstream regulator RhoA with poor patient survival at early change in overall microtubule structure (Fig. 5A). Acetylation stages of prostate cancer and prostate cancer lymph node invasion of aTubulin with increasing concentration of LIMKi was also (40). LIMK1 has been previously reported to be upregulated in detected by immunofluorescence treatment (Fig. 5B), which, prostate cancer tumors and cell lines, in which it was postulated to when combined results from three independent experiments have a role in metastasis largely due to the generally held view of were quantified and normalized to total aTubulin, revealed LIMK1 as a promoter of cell motility and migration (12–14). increased tubulin acetylation with increasing LIMKi concentra- These studies, however, did not examine LIMK1 expression levels tions (Fig. 5B). Western blot analysis showed a similar trend of in nonmetastatic versus metastatic tumor samples, nor was the increased acetylated aTubulin over total aTubulin with increas- link between LIMK1 expression with patient outcome determined ing LIMKi concentrations (Fig. 5C). To determine whether the for either group. LIMKi-induced increase in aTubulin acetylation affected AR– The role of LIMK2 in prostate cancer has largely been over- aTubulin association, LNCaP cells were treated identically as looked. Our results indicate that LIMK2 expression in nonme- for AR nuclear localization analysis in Fig. 4, and AR–aTubulin tastatic prostate cancer is similar to that of LIMK1, with a trend interactions analyzed by coimmunoprecipitation. There was a of elevated LIMK2 associated with poor patient survival and reduction in aTubulin associated with AR immunoprecipitated increased lymphovascular invasion. However, these observa- from cells treated with LIMKi (Fig. 5D); 3 mmol/L LIMKi had a tions did not achieve statistical significance in the number of similar effect to 4 nmol/L docetaxel treatment (Fig. 5D). These samples in the TMA dataset. However, the strong trend of results indicate that LIMK inhibition attenuates AR nuclear association with poor survival in the nonmetastatic patient translocation by increasing aTubulin acetylation with conse- cohort and lymphovascular invasion for all three markers,

(Continued.) C, LNCaP cells were treated with 3 mmol/L LIMKi or DMSO for 24 hours, and then incubated with 20 ng/mL CHX for indicated times. Representative Western blot analysis shows changes in AR protein levels during translational block. AR levels were normalized to aTubulin and plotted as the percentage of change relative to a no-CHX treatment control. Graph, results of three independent experiments SEM. Statistical signiﬁcance was tested by two-way ANOVA, indicated P value reports signiﬁcance of drug treatment. D, CWR22 cells were transiently transfected with p(ARE)3–luciferase reporter and CMV-driven Renilla luciferase constructs. Twenty-four hours after transfection, cells were stimulated with DHT and treated with DMSO vehicle, 3 mmol/L or 10 mmol/L LIMKi, 10 mmol/L bicalutamide for 24 hours. Promoter activity was measured as relative luciferase units and normalized to the activity of unstimulated cells. Results show average values from three independent experiments performed in triplicate SEM. E, PSA expression relative to 18S RNA in CWR22 cells pretreated with 3 mmol/L LIMKi, or DMSO vehicle for 24 hours, then treated with 1 nmol/L DHT in the presence of vehicle or LIMKi for additional 24 hours. Graph, average results for four independent experiments SEM.

www.aacrjournals.org Mol Cancer Ther; 14(1) January 2015 255

Mardilovich et al.

ABDMSO 10 µmol/L LIMKi

1.5

DMSO

Tubulin 20 µm Ace-Tub

3 mol/L LIMKi

m 20 µm

4.5

mol/L LIMKi

20 µm

Maximum projection C 3.5 P < 0.05 P < 0.05 3.0 0.6 2.5 Tubulin 2.0 1.5 Tubulin 0.4

Tub/ 1.0

Tub/ 0.2 (fold change) 0.5 0.0 Ace- α Ace- Tubulin 0.0

Ace- – – αTubulin + DMSO (fluorescence intensity) – 310LIMKi (µmol/L) +–+ ––DMSO –++ ++DHT (1 nmol/L) – – 1310LIMKi (µmol/L)

D IP IP IgG anti-AR + + –––DMSO –++++ DHT (1 nmol/L) 1.5 – – 13– LIMKi (µmol/L) – –––+ Dcxl (4 nmol/L) AR 1.0

IP αTubulin 0.5 AR (fold change) Tub/AR binding αTubulin 0.0 p-Cofilin + +–––DMSO –++++ DHT (1 nmol/L) – – 13 – LIMKi (µmol/L) Cell lysate: 1% IP Cofilin – –––+Dcxl (4 nmol/L)

Figure 5. LIMKi increases aTubulin acetylation and reduces AR–aTubulin interactions. A, LNCaP cells were treated with DMSO vehicle, or 10 mmol/L LIMKi for 24 hours, fixed, and stained with fluorescent antibodies against total aTubulin. Narrow Z-plane images were taken at 0.5-mm intervals, and distance of each image from the starting point is indicated. B, fluorescence staining intensity of acetylated aTubulin (Ace-aTub, green) was quantified and normalized to total aTubulin (red) for each field. Graph, average quantification results for three independent experiments (each using three fields/treatment) SEM. Statistical significance was tested by one-way ANOVA and the post hoc Newman–Keuls multiple comparison test, conditions differing significantly from each other as indicated. C, LNCaP cells were treated with DMSO vehicle or increasing LIMKi concentrations for 24 hours. Cells were lysed and aTubulin acetylation levels determined by Western blot analysis. Acetylated aTubulin (Ace-aTub) was normalized to total aTubulin levels and plotted as fold change over DMSO-treated control. Graph, results of three independent experiments SEM. Statistical significance was tested by one-way ANOVA and the post hoc Dunnett multiple comparison test, condition differing significantly from DMSO only control as indicated. D, LNCaP cells were treated with DMSO vehicle, 3 mmol/L or 10 mmol/L LIMKi, or 4 nmol/L docetaxel (Dcxl) for 24 hours. Cells were lysed and AR was immunoprecipitated (IP). Amount of coimmunoprecipitated aTubulin was determined by Western blot analysis. Total cell lysates (1% of the IP amount) were analyzed by Western blot analysis. Graph, aTubulin levels normalized to AR levels in each IP sample from three independent experiments SEM.

256 Mol Cancer Ther; 14(1) January 2015 Molecular Cancer Therapeutics

LIM Kinases in Prostate Cancer

LIMK1, LIMK2, and p-Cofilin, and the requirement for com- Our results suggest that there may be potential for the use of bined LIMK1 and LIMK2 knockdown by siRNA for full anti- LIMK inhibitors as an AR-dependent prostate cancer–targeted proliferative effects (Fig. 3E) leads us to propose that both therapy. Moreover, the effect of LIMKi on prostate cancer cell LIMK1 and LIMK2 contribute to development and progression migration (Fig. 2) suggests that LIMK inhibitors might have an of early-stage prostate cancer, which is often characterized as AR additional benefit of limiting cancer spread as well as targeting positive and hormone dependent. Indeed, a role of LIM kinase tumor growth. Although AR-negative prostate cancer cell lines signaling in early AR-positive and hormone-dependent prostate showed a relatively weak response to LIMKi in cytotoxicity assays cancer is consistent with the greater sensitivity of androgen- (Fig. 3 and Table 1), LIMK inhibitors may target CRPC cells that still dependent than androgen-independent prostate cancer cell rely on AR for survival. Recent research has shown that CRPC cells lines to LIMKi in cell proliferation and apoptosis detection are reliant on AR signaling, and adapt to hormone depletion assays (Fig. 3 and Table 1). therapy by increasing AR expression and/or upregulating androgen The greater sensitivity of androgen-dependent prostate cancer production (43, 44). This hypothesis is supported by the successful cell lines and RWPE-1 prostate epithelial cells to LIMKi supports treatment of CRPC with taxanes, such as docetaxel that target AR the conclusion that LIMKi targets AR function (Fig. 4 and Table 1). function or abiraterone, which inhibits 17 a-hydroxylase activity to We observed an inhibitory effect of LIMKi on AR nuclear trans- reduce androgen production (45). These findings support the location (Fig. 4A and B) and decreased AR–aTubulin interaction notion that blocking AR activity is the prime target for treatment that were associated with increased tubulin acetylation (Fig. 5). of both drug-na€ve prostate cancer and CRPC. As general microtu- LIM kinases are important regulators of actin cytoskeleton dynam- bule-targeting drugs, taxaneshavehigh overalltoxicity because their ics (7). Given that AR nuclear translocation has been previously target is essential in normal cells, and is not differentially expressed reported to depend on the filamentous actin cross-linking protein in tumor cells. Given the elevated expression of LIM kinases in the filamin (21), an additional possibility is that LIMK inhibition may nonmetastatic poor outcome patient group (Fig. 1), our data affect AR translocation via effects on the actin cytoskeleton. This support the idea that inhibition of LIM kinases could have tumor double effect on both actin and microtubule cytoskeletal net- selective effects while leaving normal tissues relatively unaffected. works that contribute to AR nuclear accumulation may explain the In addition, the finding that LIMK2 knockdown sensitized neuro- apparent selective advantage for elevated LIMK1 expression in the blastoma cells to taxane treatment (46) suggests that the treatment patients with poor-outcome nonmetastatic prostate cancer and of CRPC with docetaxel could be made more effective through the increased sensitivity of androgen-dependent prostate cancer combination therapy with LIMK inhibitors. cell lines to LIMKi. Ligand-bound AR is rapidly degraded by a proteasome-medi- Disclosure of Potential Conflicts of Interest ated degradation pathway if sequestered in the cytoplasm (41, No potential conflicts of interest were disclosed. 42). The inhibition of nuclear AR accumulation in the presence of LIMKi was accompanied by a more rapid decrease in AR protein Authors' Contributions levels (Fig. 4C), possibly due to the accumulation of ligand- Conception and design: K. Mardilovich, M.F. Olson bound AR in the cytoplasm, which would subsequently be tar- Development of methodology: K. Mardilovich, M.F. Olson Acquisition of data (provided animals, acquired and managed patients, geted for degradation. This inhibitory effect on protein stability provided facilities, etc.): K. Mardilovich, M. Gabrielsen, L. McGarry, R. Patel, likely contributes to the LIMKi cytotoxic mechanism of action in J. Edwards androgen-dependent prostate cancer cells. Analysis and interpretation of data (e.g., statistical analysis, biostatistics, We found a significant association between elevated LIMK1 computational analysis): K. Mardilovich, M. Gabrielsen, L. McGarry, E. Shanks, expression and poor survival in nonmetastatic prostate cancer, J. Edwards, M.F. Olson whereas LIMK2 expression had a similar trend in the same patient Writing, review, and/or revision of the manuscript: K. Mardilovich, E. Shanks, M.F. Olson group (Fig. 1). However, we also observed that elevated LIMK2 Administrative, technical, or material support (i.e., reporting or organizing expression in metastatic prostate cancer was associated with better data, constructing databases): C. Orange, R. Patel, J. Edwards, M.F. Olson patient outcome. These differences could indicate a unique role for Study supervision: M.F. Olson LIMK2 in cancer. We previously reported that progressively decreased LIMK2 expression due to promoter methylation was Acknowledgments associated with poor prognosis in patients with colorectal cancer, The authors thank Hing Leung (Beatson Institute for Cancer Research, Glasgow, whereas Limk2 deletion increased tumor burden in a colitis-asso- UK) for advice and gifts of prostate cancer cell lines and test tumor sections. ciated colorectal cancer mouse model (18). We determined that was due to a role of LIMK2 in restraining gastrointestinal stem cell Grant Support proliferation. On the basis of these observations, a possibility is This research was supported a grant from Cancer Research UK (A12966; to M.F. Olson). that there is selection for decreased LIMK2, but not LIMK1, expres- The costs of publication of this article were defrayed in part by the payment of sioninadvancedprostatecancertopromotetheexpansionofstem- page charges. This article must therefore be hereby marked advertisement in like tumor cells. Consistent with our observations, analysis using accordance with 18 U.S.C. Section 1734 solely to indicate this fact. the cBio Cancer Genomics Portal (31) of mRNA levels in 85 prostate adenocarcinoma tumor samples revealed significantly LIMK2 Received May 23, 2014; revised September 24, 2014; accepted October 12, decreased relativetonormaltissuein27cases(31%;ref.32). 2014; published OnlineFirst October 24, 2014.

References 1. American Cancer Society. Cancer facts & ﬁgures 2012. Atlanta: American 2. Matsumoto T, Sakari M, Okada M, Yokoyama A, Takahashi S, Kouzmenko A, et al. Cancer Society; 2012. The androgen receptor in health and disease. Ann Rev Physiol 2013;75:201–24.

www.aacrjournals.org Mol Cancer Ther; 14(1) January 2015 257

Mardilovich et al.

3. Saylor PJ. Prostate cancer: the androgen receptor remains front and centre. 25. Bravo-Cordero JJ, Magalhaes MAO, Eddy RJ, Hodgson L, Condeelis J. Nat Rev Clin Oncol 2013;10:126–8. Functions of cofilin in cell locomotion and invasion. Nat Rev Mol Cell 4. Feldman BJ, Feldman D. The development of androgen-independent Biol 2013;14:405–15. prostate cancer. Nat Rev Cancer 2001;1:34–45. 26. Obrdlik A, Percipalle P. The F-actin severing protein cofilin-1 is required for 5. Gan L, Chen S, Wang Y, Watahiki A, Bohrer L, Sun Z, et al. Inhibition of the RNA polymerase II transcription elongation. Nucleus 2011;2:72–9. androgen receptor as a novel mechanism of taxol chemotherapy in prostate 27. Kalendova A, Kalasova I, Yamazaki S, Ulicna L, Harata M, Hozak P. Nuclear cancer. Cancer Res 2009;69:8386–94. actin filaments recruit cofilin and actin-related protein 3, and their for- 6. Zhu ML, Horbinski CM, Garzotto M, Qian DZ, Beer TM, Kyprianou N. mation is connected with a mitotic block. Histochem Cell Biol 2014;142: Tubulin-targeting chemotherapy impairs androgen receptor activity in 139–52. prostate cancer. Cancer Res 2010;70:7992–8002. 28. Adams J, Cheng L. Lymph node-positive prostate cancer: current issues, 7. Scott RW, Olson MF. LIM kinases: function, regulation and association with emerging technology and impact on clinical outcome. Exp Rev Anticancer human disease. J Mol Med 2007;85:555–68. Ther 2011;11:1457–69. 8. Prudent R, Vassal-Stermann E, Nguyen CH, Pillet C, Martinez A, Prunier C, 29. Rhodes DR, Yu J, Shanker K, Deshpande N, Varambally R, Ghosh D, et al. et al. Pharmacological inhibition of LIM kinase stabilizes microtubules and ONCOMINE: a cancer microarray database and integrated data-mining inhibits neoplastic growth. Cancer Res 2012;72:4429–39. platform. Neoplasia 2004;6:1–6. 9. Acevedo K, Li R, Soo P, Suryadinata R, Sarcevic B, Valova VA, et al. The 30. Magee JA, Araki T, Patil S, Ehrig T, True L, Humphrey PA, et al. Expression phosphorylation of p25/TPPP by LIM kinase 1 inhibits its ability to profiling reveals hepsin overexpression in prostate cancer. Cancer Res assemble microtubules. Exp Cell Res 2007;313:4091–106. 2001;61:5692–6. 10. Gorovoy M, Niu J, Bernard O, Profirovic J, Minshall R, Neamu R, et al. LIM 31. Cerami E, Gao J, Dogrusoz U, Gross BE, Sumer SO, Aksoy BA, et al. The cBio kinase 1 coordinates microtubule stability and actin polymerization in cancer genomics portal: an open platform for exploring multidimensional human endothelial cells. J Biol Chem 2005;280:26533–42. cancer genomics data. Cancer Discov 2012;2:401–4. 11. Klejnot M, Gabrielsen M, Cameron J, Mleczak A, Talapatra SK, Kozielski F, 32. Taylor BS, Schultz N, Hieronymus H, Gopalan A, Xiao Y, Carver BS, et al. et al. Analysis of the human cofilin 1 structure reveals conformational Integrative genomic profiling of human prostate cancer. Cancer Cell changes required for actin binding. Acta Crystallographica Section D 2010;18:11–22. 2013;69:1780–8. 33. Scott RW, Hooper S, Crighton D, Li A, Konig€ I, Munro J, et al. LIM kinases 12. Davila M, Frost AR, Grizzle WE, Chakrabarti R. LIM kinase 1 is essential for are required for invasive path generation by tumor and tumor-associated the invasive growth of prostate epithelial cells: implications in prostate stromal cells. J Cell Biol 2010;191:169–85. cancer. J Biol Chem 2003;278:36868–75. 34. Morin P, Wickman G, Munro J, Inman GJ, Olson MF. Differing contribu- 13. Davila M, Jhala D, Ghosh D, Grizzle W, Chakrabarti R. Expression of LIM tions of LIMK and ROCK to TGFbeta-induced transcription, motility and kinase 1 is associated with reversible G1/S phase arrest, chromosomal invasion. Eur J Cell Biol. 2011;90:13–25. instability and prostate cancer. Mol Cancer 2007;6:40. 35. Maruta S, Miyata Y, Sagara Y, Kanda S, Iwata T, Watanabe SI, et al. 14. Tapia T, Ottman R, Chakrabarti R. LIM kinase1 modulates function of Expression of matrix metalloproteinase-10 in non-metastatic prostate membrane type matrix metalloproteinase 1: implication in invasion of cancer: correlation with an imbalance in cell proliferation and apoptosis. prostate cancer cells. Mol Cancer 2011;10:6. Oncol Lett 2010;1:417–21. 15. Ahmed T, Shea K, Masters JR, Jones GE, Wells CM. A PAK4-LIMK1 pathway 36. Pulukuri SM, Rao JS. Matrix metalloproteinase-1 promotes prostate tumor drives prostate cancer cell migration downstream of HGF. Cell Signal growth and metastasis. Int J Oncol 2008;32:757–65. 2008;20:1320–8. 37. Cao Z WC, Norton-Wenzel CS, Rej R, Davis FB, Davis PJ, Rej R. Effects of 16. TanSS,AhmadI,BennettHL,SinghL,NixonC,SeywrightM,etal.GRP78up- resin or charcoal treatment on fetal bovine serum and bovine calf serum. regulation is associated with androgen receptor status, Hsp70-Hsp90 client Endocr Res 2009;34:101–8. proteins and castrate-resistant prostate cancer. J Pathol 2011;223:81–7. 38. Darshan MS, Loftus MS, Thadani-Mulero M, Levy BP, Escuin D, Zhou XK, 17. Ross-Macdonald P, de Silva H, Guo Q, Xiao H, Hung CY, Penhallow B, et al. et al. Taxane-induced blockade to nuclear accumulation of the androgen Identification of a nonkinase target mediating cytotoxicity of novel kinase receptor predicts clinical responses in metastatic prostate cancer. Cancer inhibitors. Mol Cancer Ther 2008;7:3490–8. Res 2011;71:6019–29. 18. Lourenco FC, Munro J, Brown J, Cordero J, Stefanatos R, Strathdee K, et al. 39. Westermann S, Weber K. Post-translational modifications regulate micro- Reduced LIMK2 expression in colorectal cancer reflects its role in limiting tubule function. Nat Rev Mol Cell Biol 2003;4:938–47. stem cell proliferation. Gut 2014;63:480–93. 40. Schmidt LJ, Duncan K, Yadav N, Regan KM, Verone AR, Lohse CM, et al. 19. Samuel MS, Lopez JI, McGhee EJ, Croft DR, Strachan D, Timpson P, et al. RhoA as a Mediator of Clinically Relevant Androgen Action in Prostate Actomyosin-mediated cellular tension drives increased tissue stiffness and Cancer Cells. Mol Endocrinol. 2012;26:716–35. beta-catenin activation to induce epidermal hyperplasia and tumor 41. Lin HK WL, Hu YC, Altuwaijri S, Chang C. Phosphorylation-dependent growth. Cancer Cell 2011;19:776–91. ubiquitylation and degradation of androgen receptor by Akt require Mdm2 20. Croft DR, Crighton D, Samuel MS, Lourenco FC, Munro J, Wood J, et al. E3 ligase. EMBO J 2002;21:4037–48. p53-mediated transcriptional regulation and activation of the actin cyto- 42. Jaworski T. Degradation and beyond: control of androgen receptor activity skeleton regulatory RhoC to LIMK2 signaling pathway promotes cell by the proteasome system. Cell Mol Biol Lett 2006;11:109–31. survival. Cell Res 2011;21:666–82. 43. Chen CD, Welsbie DS, Tran C, Baek SH, Chen R, Vessella R, et al. Molecular 21. Ozanne DM, Brady ME, Cook S, Gaughan L, Neal DE, Robson CN. determinants of resistance to antiandrogen therapy. Nat Med 2004; Androgen Receptor Nuclear Translocation Is Facilitated by the f-Actin 10:33–9. Cross-Linking Protein Filamin. Mol Endocrinol 2000;14:1618–26. 44. Mohler JL, Gregory CW, Ford OH, Kim D, Weaver CM, Petrusz P, et al. The 22. Scott RW, Hooper S, Crighton D, Li A, Konig I, Munro J, et al. LIM kinases androgen axis in recurrent prostate cancer. Clin Cancer Res 2004;10: are required for invasive path generation by tumor and tumor-associated 440–8. stromal cells. J Cell Biol 2010;191:169–85. 45. Yap TA, Zivi A, Omlin A, de Bono JS. The changing therapeutic landscape of 23. Conway C, Dobson L, O'Grady A, Kay E, Costello S, O'Shea D. Virtual castration-resistant prostate cancer. Nat Rev Clin Oncol 2011;8:597–610. microscopy as an enabler of automated/quantitative assessment of protein 46. Gamell C, Schofield AV, Suryadinata R, Sarcevic B, Bernard O. LIMK2 expression in TMAs. Histochem Cell Biol 2008;130:447–63. mediates resistance to chemotherapeutic drugs in neuroblastoma cells 24. Munsie LN, Desmond CR, Truant R. Cofilin nuclear-cytoplasmic shuttling through regulation of drug-induced cell cycle arrest. PLoS ONE 2013;8: affects cofilin-actin rod formation during stress. J Cell Sci 2012;125:3977–88. e72850.

258 Mol Cancer Ther; 14(1) January 2015 Molecular Cancer Therapeutics

Elevated LIM Kinase 1 in Nonmetastatic Prostate Cancer Reflects Its Role in Facilitating Androgen Receptor Nuclear Translocation

Katerina Mardilovich, Mads Gabrielsen, Lynn McGarry, et al.

Mol Cancer Ther 2015;14:246-258. Published OnlineFirst October 24, 2014.

Updated version Access the most recent version of this article at: doi:10.1158/1535-7163.MCT-14-0447

Supplementary Access the most recent supplemental material at: Material http://mct.aacrjournals.org/content/suppl/2014/10/24/1535-7163.MCT-14-0447.DC1

Cited articles This article cites 45 articles, 15 of which you can access for free at: http://mct.aacrjournals.org/content/14/1/246.full#ref-list-1

Citing articles This article has been cited by 1 HighWire-hosted articles. Access the articles at: http://mct.aacrjournals.org/content/14/1/246.full#related-urls

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

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at Subscriptions [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://mct.aacrjournals.org/content/14/1/246. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.