Transforming Growth Factor-Β Promotes Prostate Bone

ORIGINAL ARTICLE Transforming growth factor-β promotes prostate bone metastasis through induction of microRNA-96 and activation of the mTOR pathway

MK Siu1,2,6, Y-C Tsai1,3,6, Y-S Chang3, JJ Yin4, F Suau4, W-Y Chen5 and Y-N Liu1,3

Transforming growth factor-β (TGFβ) is enriched in the bone matrix and serves as a key factor in promoting bone metastasis in cancer. In addition, TGFβ signaling activates mammalian target of rapamycin (mTOR) functions, which is important for the malignant progression. Here, we demonstrate that TGFβ regulates the level of microRNA-96 (miR-96) through Smad-dependent transcription and that miR-96 promotes the bone metastasis in prostate cancer. The enhanced effects in cellular growth and invasiveness suggest that miR-96 functions as an oncomir/and metastamir. Supporting this idea, we identiﬁed a downstream target of the TGFβ-miR-96 signaling pathway to be AKT1S1 mRNA, whose translated protein is a negative regulator of mTOR kinase. Our ﬁndings provide a novel mechanism accounting for the TGFβ signaling and bone metastasis.

Oncogene (2015) 34, 4767–4776; doi:10.1038/onc.2014.414; published online 22 December 2014

INTRODUCTION of the AKT1S1–mTORC1 complex followed by disruption of the 12,13 Transforming growth factor-β (TGFβ) signaling exhibits inhibitory effect of AKT1S1. As mTOR activity is tightly 14 complicated features in cancer biology in that TGFβ is considered associated with many aspects of tumorigenesis, the inhibitory as tumor promoter1 and suppressor.1–3 In prostate and other types role of AKT1S1 has raised a great interest of its regulation by of cancer, a general pattern is that TGFβ signaling suppresses mechanisms other than AKT phosphorylation. Emerging evidence tumor growth in normal/benign tissues while enhancing has shown that microRNAs (miRs) have important roles in malignancy (for example, invasion and metastases) in tumorigenic metastases15 and regulate many components in the TGFβ tumor cells.4–6 Current hypothesis to explain this apparent signaling pathway.16 It is possible that, following TGFβ signaling, paradox is that the effector pathways and their associated factors miRs modulate mTOR functions to commit the malignant in response to TGFβ signaling are dependent on individual progression. cellular-context.4–6 In the current study, we have identified microRNA-96 (miR-96) TGFβ signaling is intrinsically complex for the diverse environ- as a key factor in metastatic prostate cancer in response to TGFβ ments in tumor cells. In the case of Smad-dependent canonical signaling. We demonstrated that the Smad-dependent canonical signaling, there are combinatory interactions of type I and type II pathway is the key mechanism responsible for the induction of receptors and receptor-activated Smads along with their corre- miR-96 transcription. We showed that miR-96 targets at the 3′- sponding transcription complexes.7 In addition, following Smad- untranslated regions (3′UTRs) of AKT1S1 mRNA and leads to its independent noncanonical signaling, TGFβ receptors directly downregulation affecting the protein level. As expected, the activate many receptor pathways that response to factors other phosphorylation levels of S6K, a key mTOR kinase substrate, are than TGFβ superfamily.7 Among many Smad-independent positively regulated by miR-96 and inversely correlated with the pathways, the PI3K-AKT-mTOR axis is important for the malignant levels of AKT1S1. In addition, the inhibitory effects of TGFβ-miR-96 progression4,8,9 and pathological bone metastases.10 The signaling on AKT1S1 expression were confirmed in tissue mammalian mTOR exists in two distinct complexes called samples from prostate cancer patients showing an inverse mTORC1 and mTORC2.11 Activation of mTOR in response to correlation between miR-96 and AKT1S1 levels. Finally, based on growth factors results in the phosphorylation of numerous results derived from multiple approaches including cell line substrates, including the phosphorylation of S6 kinase by mTORC1 proliferation/invasion assays, clinical database analyses and mouse and AKT by mTORC2.11 A positive regulation toward mTORC1 by bone metastasis model, we concluded that miR-96 can promote mTORC2 is established, as one of the key regulators for mTORC1 bone metastasis and contribute to reduced survival rates. This kinase is identified to be an AKT substrate, AKT1S1 study supports that miR-96 could be served as a prognostic (also known as PRAS40).12,13 Phosphorylation of AKT1S1 by AKT marker and a potential target of therapeutics for metastatic results in the activation of mTOR kinase due to the dissociation prostate cancers.

1Program for Cancer Biology and Drug Discovery, College of Medical Science and Technology, Taipei Medical University and Academia Sinica, Taipei, Taiwan; 2Department of Anesthesiology, Wan Fang Hospital, Taipei Medical University, Taipei, Taiwan; 3Graduate Institute of Cancer Biology and Drug Discovery, College of Medical Science and Technology, Taipei Medical University, Taipei, Taiwan; 4Cell and Cancer Biology Branch, National Cancer Institute, National Institutes of Health, Bethesda, MD, USA and 5Department of Pathology, Wan Fang Hospital, Taipei Medical University, Taipei, Taiwan. Correspondence: Assistant Professor Y-N Liu or Assistant Professor Y-C Tsai, Graduate Institute of Cancer Biology and Drug Discovery, College of Medical Science and Technology, Taipei Medical University, No. 250, Wu-Hsing Street, Taipei 11031, Taiwan. E-mail: [email protected] or [email protected] 6These authors contributed equally to this work. Received 28 July 2014; revised 21 October 2014; accepted 11 November 2014; published online 22 December 2014 miR-96 regulates mTOR signaling in prostate cancer MK Siu et al 4768 RESULTS Interestingly, we have reported that only AC3 is capable of The level of miR-96 is regulated by the TGFβ-Smad signaling undergoing epithelial–mesenchymal transition (EMT) following TGFβ treatment as confirmed again by the induction of vimentin We have established several clonal cell lines (AC1, AC3 and SC1) 17 derived from the prostate-specific, Pten/TP53 double knockout (Vim, Figure 1b), consistent with our earlier studies. The lack of mice17 to study the roles of microRNAs (miRs) following TGFβ EMT response in SC1 could be due to its late/or terminal stage that signaling.18 Using this system, some miRs (for example, miR-1) contains much lower Smad2 levels and shuts off TGFβ signaling have been characterized with tumor-suppressive functions and (Figure 1a). with downregulation pattern following TGFβ treatment.18 Com- Using this mouse prostate cancer system, we identified that pared with the cell line derived from late-stage sarcomatoid miR-96 was upregulated by TGFβ in a time-dependent manner in carcinoma (SC1), only cell lines derived from adenocarcinoma cell lines derived from adenocarcinoma but not from sarcomatoid (AC1 and AC3) showed constitutive TGFβ signaling as evidenced carcinoma (AC1 and AC3 versus SC1, Figure 1c). The upregulation by elevated phosphorylation levels of Smad2 (Figure 1a). feature of miR-96 is distinct from our previously characterized

Figure 1. TGFβ signaling regulates Smad-dependent miR-96 transcription. (a) Active TGFβ signaling in adenocarcinoma cell lines (AC1 and AC3) but not sarcomatoid carcinoma (SC1). Immunoblot analyses in the PTEN/TP53-deleted mouse prostate cancer cell lines. (b)TGFβ induces EMT only in AC3. Measurement of induction of vimentin (Vim) in response to TGFβ (10 ng/ml) for 24 h. (c)TGFβ induces miR-96 in adenocarcinoma cell lines (AC1 and AC3) but not SC1. Time course of miR-96 induction levels following TGFβ treatment. (d) Reduction of miR-96 levels in cells treated with TGFβ receptor inhibitor, SB431542 (1 μg/ml) for 24 h. (e) Schematic of predicted SBEs in the mouse pri-mir-96 stem–loop promoter. (f) ChIP analyses of predicted SBEs in the pri-mmu-mir-96 promoter region in cells treated with TGFβ (10 ng/ml) for 24 h. The enrichment of each protein to each site is given as a percentage of total input and then normalized to each IgG. (g) The median ﬂuorescent intensity (MFI) value in cell lines transiently transfected with SBEs wild-type (left panel) or mutated (right panel) pri-mmu-mir-96- RFP reporters with TGFβ treatment for 24 h. E-Cad: E-cadherin, Vim: vimentin. Data represent means ± s.e.m., n = 3. * versus vehicle. *Po0.05, **Po0.01, ***Po0.001.

Oncogene (2015) 4767 – 4776 © 2015 Macmillan Publishers Limited miR-96 regulates mTOR signaling in prostate cancer MK Siu et al 4769 miRs (for example, miR-1 and miR-200), which showed down- In addition to FOXO1, we found that AKT1S1 as a potential target regulaton pattern.18 In addition, the irresponsiveness to TGFβ- with two predicted miR-96-binding sites in the 3′UTR region of the induced EMT in both AC1 and SC1 did not warrant a consistent mRNA transcript. We next set up a luciferase reporter system to outcome in predicting miR-96 response, as we observed induction monitor the regulatory roles of TGFβ-Smad-miR-96 signaling on of miR-96 in a time-dependent manner in AC1 but not SC1 AKT1S1 by measuring the ratio of RL, representing the levels of (Figure 1c). Interestingly, when cells were treated with TGFβ type I AKT1S1 3′UTR, and FL, which served as internal control (Figure 2a). receptor inhibitor, the steady-state levels of endogenous miR-96 Consistent with our predicted role of miR-96, we found that TGFβ were reduced (SB431542, Figure 1d), suggesting a background treatment causes downregulation of AKT1S1 levels in AC1 and activation of TGFβ signaling possibly due to serum-derived TGFβ. AC3 cells; on the other hand, inhibition of TGFβ signaling by Besides, this result supports the idea that the regulation of miR-96 inhibitor shows the opposite effects (Figure 2b), further support- is tightly associated with the activity of TGFβ signaling. To further ing the idea that the endogenous AKT1S1 levels is a downstream confirm the association between miR-96 and TGFβ signaling, we event of TGFβ signaling. To directly assess the effects of miR-96 in examined the gene expression profiles in a public prostate cancer the same reporter system, we showed that exogenously expres- data base maintained at Memorial Sloan–Kettering Cancer sing miR-96 precursor (labeled ‘96 pre’) represses the luciferase Center.19 After analyzing the database (containing 98 primary activities and that knockdown of miR-96 (labeled ‘anti-miR-96’) tumor and 13 distant metastasis samples) by gene set enrichment increases the signals (Figure 2c). When monitoring the endogen- analysis, we observed that miR-96 is correlated with TGFβ gene ous mRNA levels of AKT1S1 under the same experimental setting, sets (Supplementary Figures S1A and B). In summary, our results we also found the same regulation pattern in both AC1 and AC3 substantiate the idea that the miR-96 expression is a downstream cell lines (Figure 2d), consistent with the hypothesis that miR-96 event of the TGFβ signaling pathway. reduces the stability of AKT1S1 mRNA. Furthermore, we asked whether the predicted miR-96-binding sites in the 3′UTR of AKT1S1 do provide specificity by monitoring the luciferase Smad-dependent transcription induces miR-96 following TGFβ signaling activities of the reporter constructs containing individual mutations at miR-96 target sites (+120 and +649, Figure 2a). As shown β To investigate how TGF signaling transcriptionally regulates in Figure 2e, we demonstrated that mutation at +649-binding site miR-96 expression, we looked closely at the putative promoter confers resistance to the inhibitory effects of either TGFβ or region upstream of the mouse primary miR (pri-miR) transcripts miR-96, supporting a physical interaction between miR-96 and encoding miR-96, located in the mouse chromosome 6. Following AKT1S1 3′UTR. fi sequence analysis, we identi ed two Smad2/3-binding elements As mTOR signaling is crucial to the malignant progression of (SBEs) within the promoter region (Figure 1e). To further test prostate cancer8–10 and AKT1S1 protein inhibits the mTOR β whether Smad2/3 directly mediates the TGF -dependent functions,13 it is possible that miR-96 promotes aggressive pri-miR-96 transcription, we performed chromatin immuno- prostate phenotypes through downregulation of AKT1S1 and precipitation (ChIP) assay in AC1 and AC3 cell lines following induction of mTOR activities. Indeed, we observed that TGFβ fi quantitative PCR analyses. As shown in Figure 1f, signi cant treatment reduces AKT1S1 protein levels and increases the mTOR increase in Smad2/3 and Smad4 signals only occurred at both phosphorylation, whereas suppression of the TGFβ signaling using SBEs after TGFβ treatment. In addition, we performed promoter specific inhibitor (SB431542) exhibits opposite effects (left panel, assay to examine whether the individual SBE in the promoter Figure 2f). To directly examine whether miR-96 regulates mTOR region is functional. Using fluorescent protein (RFP) as reporter, we activation, we found that overexpression of miR-96 precursor demonstrated that TGFβ indeed increases the RFP levels by increases mTOR phosphorylation and reduces the protein level of monitoring its median fluorescent intensity using either SBE1 or AKT1S1 (right panel, Figure 2f). Similar to the effect of TGFβ SBE2 (left panel, Figure 1g). Furthermore, we characterized the receptor inhibitor, anti-miR-96 reduces mTOR phosphorylation but specificity of Smads binding in the promoter region of pri-miR-96 increases the protein level of AKT1S1 (right panel, Figure 2f). In by introducing point mutation (Figure 1e). We found that only summary, we identified AKT1S1 mRNA as a novel substrate of mutation at SBE2 disrupts the ability of TGFβ to induce RFP miR-96, and the downregulation of AKT1S1 through miR-96 expression in the promoter assay (right panel, Figure 1g). These enhances the mTOR activity following TGFβ signaling. data suggest a molecular mechanism by which TGFβ regulates pri-miR-96 transcription through a physical interaction between TGFβ-miR-96 signaling downregulates AKT1S1 and enhances the Smad2/3/4 complex and the promoter region of pri-miR-96. mTOR functions The identification of AKT1S1 mRNA as a direct target of miR-96 in fi AKT1S1 is a bona de target of miR-96 and tightly regulated by mouse prostate cell lines provides important clues to the function β TGF signaling of miR-96. As the AKT1S1 protein is a negative regulator of To address the molecular mechanism of miR-96, we focused on mTOR,12,13 this newly identified pathway has important implica- searching its targeting substrates that could be involved in tion in the tumor progression. Therefore, we asked whether the promoting malignancy. Earlier studies have shown that miR-96 TGFβ-miR-96-AKT1S1 axis is a general pathway in prostate cancer. suppresses FOXO1 and enhances proliferation.20 In our mouse We performed experiments using human metastatic prostate prostate cancer system, we also observed that TGFβ and miR-96 cancer cell lines to validate the inhibitory effect of TGFβ-miR-96 behave similarly to downregulate FOXO1 levels (Supplementary signaling on AKT1S1 mRNA. As shown in Figure 3a, we found that Figures S1C and D). Using a bi-cistronic luciferase assay system TGFβ treatment induces miR-96 levels in PC3 and RasB121 cell with firefly luciferase (FL) as internal control, we combined Renilla lines. However, under the same conditions, TGFβ treatment luciferase (RL) and the mouse 3′UTR of FOXO1 in one transcrip- markedly suppressed the levels of AKT1S1 mRNA (Figure 3b). tional unit and monitored the luciferase activities. After normal- Next, increasing miR-96 levels by transient transfection (96 pre) ization (RL/FL), we showed that both TGFβ and miR-96 can reduce decreases the endogenous AKT1S1 levels, whereas knockdown of the activity of RL consistent to the finding that miR-96 targets to miR-96 can increase the AKT1S1 levels (anti-miR-96, Figure 3c). We the 3′UTR of FOXO1 leading to its downregulation (Supplementary further performed luciferase reporter assay using the 3′UTR of Figure S1E). Therefore, we validated our cellular system by human AKT1S1. We found that the expression patterns of human recapitulating the inhibitory effects of miR-96 on the tumor AKT1S1 3′UTR are the same as those of mice AKT1S1 3′UTR in suppressor FOXO1. response to TGFβ signaling (Figure 3d versus Figure 2b) and

Figure 2. Activation of TGFβ signaling results in downregulation of AKT1S1 through miR-96-mediated speciﬁc targeting to the 3′ UTR of AKT1S1 in PTEN/TP53-deleted prostate cancer cells. (a) Schematic of predicted miR-96-binding sites (+120 and +649) in full-length 3′ UTR reporter constructs of mouse AKT1S1. SV40 and HSV-TK: promoters. (b) The normalized reporter activity in AC1 and AC3 cells treated with TGFβ or SB431542 for 24 h. RL/FL activities were measured 48 h after transfection. (c) The normalized reporter activity in cells treated with agents that regulate miR-96 levels. (d) Endogenous levels of mouse AKT1S1 in cells treated with agents that regulate miR-96 levels. (e) The normalized reporter activity of mouse AKT1S1 3′UTR containing wild-type or mutated miR-96-binding sites in AC1 and AC3 cells with TGFβ treatment or transient expression of miR-96 precursor (96 pre) for 24 h. (f) Representative western blot analyses with TGFβ and SB431542 using speciﬁc antibodies (left panel) and agents that change the levels of mouse miR-96 precursor (middle and right panels). β-ACTIN is served as loading control. Veh: vehicle for TGFβ or SB431542; EV: empty vector; 96 pre: vector expressing mouse miR-96 precursor; ctrl: control anti-miR. Data represent means ± s.e.m. of separate transfections, n = 3. *Po0.05, **Po0.01, ***Po0.001.

miR-96 status (Figure 3e versus Figure 2c), supporting that the prostate cancer using immunohistochemistry (IHC). First, we TGFβ-miR-96-AKT1S1 axis is a general pathway in prostate cancer. analyzed the mRNA levels in 24 independent prostate tumors When we monitored the protein levels following TGFβ collected from Wan Fang Hospital, Taipei Medical University, treatment, the AKT1S1 level is reduced and the mTOR kinase Taiwan. The samples were divided into two groups based on activity is increased as evidenced by the induced phosphorylation AKT1S1 expression levels, and an analysis of variance showed that of S6 kinase (P-S6K, Figure 3f). Similar to the effect of TGFβ, the tissues with lower AKT1S1 expression (labeled ‘AKT1S1_L’) have overexpression of miR-96 precursor also reduced the AKT1S1 higher miR-96 levels (Figure 4a), which is consistent with the levels and stimulated the mTOR kinase activity showing increased downregulation effect of miR-96 observed in prostate cancer cell phosphorylation levels of S6 kinase (Figure 3f and Supplementary lines (Figure 2 and Figure 3). In addition, an inverse correlation Figure S1D). Upon introducing miR-96 precursor, the endogenous between the miR-96 and AKT1S1 mRNA levels was confirmed by protein levels of mTOR and S6K are also increased (96 pre, statistical analyses in the clinical samples (Figure 4b). We next Figure 3f) consistent with the increased protein synthesis rate performed IHC analysis in distant metastatic tumors to directly following TGFβ signaling.8 We also observed basal levels of TGFβ monitor the AKT1S1 protein levels. Consistent with our finding β signaling likely because of serum-derived TGFβ as TGFβ receptor that induction of miR-96 is TGF -Smad-dependent (Figure 1), we β inhibitor (SB431542) significantly reduces the levels of miR-96 observed strong TGF signaling only in prostate cancer samples (Figure 1d) and the phosphorylation levels of S6 kinase (Figure 3f). with higher miR-96 expression levels as evidenced by the nuclear However, reducing the endogenous miR-96 level by anti-miR-96 phospho-Smad2 staining (upper panel, Figure 4c). However, when overcomes the activation effect of endogenous TGFβ and still we monitored the AKT1S1 protein levels in the continuous tissue reduces the phosphorylation levels of S6 kinase (P-S6K, Figure 3f section, tumor with lower miR-96 levels show intensive AKT1S1 staining compared with the tissue samples with higher miR-96 and Supplementary Figure S1F). We concluded that Smad- fi dependent transcription of miR-96 has an important role in expression (bottom panel, Figure 4c). Quanti cation of the images TGFβ-regulated mTOR functions partly through miR-96-mediated taken from the tissue samples showed that the levels of phospho- Smad2 and AKT1S1 are inverse-correlated (Figure 4d). When we downregulation of AKT1S1. performed gene set enrichment analysis using a prostate cancer data set,19 a set of genes that is downregulated by mTOR TGFβ-miR-96 signaling and AKT1S1 level are inversely correlated inhibitor22 is enriched in the group of cancer expressing lower in human prostate cancer levels of miR-96 (Figure 4e). These results are consistent with our To further confirm the regulatory role of miR-96 and establish the proposed mechanism that the mTOR function is controlled by the inverse correlation between miR-96 and AKT1S1, we sought to status of miR-96 level partly through the regulation of monitor the expression levels directly from the tissue samples of AKT1S1 mRNA.

Figure 3. The TGFβ-miR-96 signaling regulates AKT1S1 levels and mTOR kinase activities in human metastatic prostate cancer cell lines. (a) Levels of miR-96 in PC3 and RasB1 in response to TGFβ treatment for 24 h. (b) Levels of human AKT1S1 determined in PC3 and RasB1 cells with TGFβ treatment for 24 h. (c) Levels of human AKT1S1 in cells treated with agents that regulate miR-96 levels. (d) The normalized reporter activity of the full-length human AKT1S1 3′UTRs with TGFβ or SB431542 treatments for 24 h. The ratios of RL/FL activities were measured 48 h after transfection. (e) The normalized reporter activity of the full-length human AKT1S1 3′UTRs treated with agents that regulate miR-96 levels. (f) Immunoblot assays for human prostate cancer cell lines treated with agents regulating TGFβ signaling (left panel) and miR-96 levels (right panel). Veh: vehicle for TGFβ or SB431542; EV: empty vector; 96 pre: human miR-96 precursor; ctrl: control anti-miR. Data represent means ± s.e.m. of separate transfections, n = 3. *Po0.05, **Po0.01, ***Po0.001.

Induction of miR-96 promotes aggressive phenotypes of prostate whether miR-96 promotes malignancy in cancer phenotypes, we cancers ﬁrst examined the levels of miR-96 in a panel of human prostate Accumulating evidence has shown that miRs have an important cancer cell lines. As shown in Supplementary Figure S1G, we found role in TGFβ-mediated tumor progression.15 To demonstrate that the levels of miR-96 in metastatic cell lines (PC3 and RasB1) are

Figure 4. Establishment of inverse correlation between the TGFβ-Smad-miR-96 pathway and AKT1S1 levels in metastatic prostate cancer patients. (a) Levels of miR-96 in two groups of tissue samples based on AKT1S1 levels (n = 12). AKT1S1_H: samples with high AKT1S1 levels; AKT1S1_L: samples with low AKT1S1 levels. (b) Pearson anticorrelation coefficient of the mean miR-96 to the mean AKT1S1 mRNA expression in primary and metastasis prostate samples (n = 18). Significance was determined by Gaussian population (Pearson) and two-tailed test. (c) Immunohistochemistry staining with antibodies specific for phospho-Smad2 and AKT1S1 in prostate cancer tissue sections with different miR-96 levels. Scale bars represent 100 μm. (d) Inverse correlation of AKT1S1 and phospho-Smad2 protein levels measured by IHC. (e) Correlation of mTOR signaling-responsive genes with miR-96 levels in a clinical prostate cancer database using gene set enrichment analysis (GSEA). (NES, normalized enrichment score; FDR, false discovery rate).

higher than those either in a benign prostatic hyperplasia epithelial samples. Compared with normal tissues, the intensity mean cell line (DHMLE) or nonmetastatic cell lines (LNCap and DU145), expression analyses showed progressive increase in miR-96 indicating a positive role in tumorigenesis. Next, our results showed expression in primary and in metastatic tumor samples that inhibition of miR-96 in the metastatic cell lines results in (Figure 5e), which is consistent with a correlation between declined invasiveness (Figure 5a), whereas overexpression of miR-96 expression and the stages of clinic tumor miR-96 in the nonmetastatic DU145 cell line increases the invasion (Supplementary Figure S2C). Patients with higher miR-96 expres- ability (Figure 5b). We also found that the miR-96-regulated sion also have a poorer survival rate (Figure 5f). invasiveness is correlated with the proliferation rate, as miR-96 On the basis of the effects observed in metastatic (that is, PC3 knockdown reduces the rates in PC3 and RasB1 (Figure 5c), and RasB1) and nonmetastatic (that is, DU145) cell lines, we whereas overexpression of miR-96 increases that in DU145 reasoned that the levels of miR-96 determine the malignancy of (Figure 5d). To conﬁrm that TGFβ promotes malignant phenotypes prostate cancer. To further test this idea, we evaluated the effects of prostate cancer through the miR-96-mediated pathway, we of miR-96 in the nonmetastatic DU145 cell line using a mice performed experiments in cells with miR-96 knockdown and metastatic model. Compared with control vector (EV), the monitored the TGFβ-induced invasiveness and proliferation. As originally nonmetastatic DU145 cells became highly metastatic expectedly, TGFβ treatment can further enhance both invasion and and enriched in the bone marrow after overexpression of miR-96 proliferation rates; however, reducing the levels of miR-96 (96 pre) (Figure 6a). Compared with mice injected with DU145- signiﬁcantly suppressed the malignant phenotypes in response to harboring vector control, those with DU145-expressing miR-96 TGFβ (ctrl. versus anti-miR-96, Supplementary Figures S2A and B). precursor (96 pre) show that the majority of bone marrows in the We further investigated the miR-96 levels in the clinical data,19 femurs were occupied by osteolytic tumor cells (Figure 6b) and containing 28 normal, 98 primary and 13 metastatic prostatic reduced survival rate (Figure 6c). Taken together, our aggregated

Figure 5. Induction of miR-96 contributes to invasiveness in tumor cell lines and correlates with both tumor stages and poor survival rates in patients with prostate cancer. (a) Suppression of miR-96 reduces invasion abilities in PC3 and RasB1 cell lines. (b) Increasing the level of miR-96 enhances invasion efﬁciency in the DU145 cell line. (c) Decreasing the levels of miR-96 reduces the growth rates in PC3 and RasB1 cell lines. (d) Elevated level of miR-96 accelerates the growth rate in DU145. (e) The mean miR-96 levels in human normal (n = 28), primary (n = 98) and metastatic (n = 13) prostate samples. (f) Survival rate of prostate cancer patients (n = 20) with different miR-96-expressing levels. *Po0.05, **Po0.01, ***Po0.001 (a–d; n = 5). Ctrl: anti-miR control, EV: empty vector, 96 pre: vector-expressing miR-96 precursor. results from cell lines, clinical database analyses and mouse Pten/Trp53 deletion showed aggressive and lethal prostate cancer models consistently support the idea that miR-96 promotes compared with either the Pten or Trp53 gene alone in mice oncogenic and metastatic properties in advanced prostate cancer. prostate epithelium.23,24 In the current study, we identiﬁed Our results support the idea that miR-96 acts as a metastamir/or expression of miR-96 following TGFβ treatment in cell lines oncomir partly through its regulation of mTOR functions. As derived from the epithelial prostate adenocarcinoma in the Pten/ shown in Figure 6d, we proposed a model that TGFβ utilizes the Trp53 deletion mice17 (Figure 1). On the basis of several lines of Smad-dependent canonical pathway to stimulate mTOR functions. evidence, we concluded that miR-96 promotes aggressive cancer AKT1S1 suppresses the functions of mTOR through formation of a phenotypes (Figure 6), which could be considered as a biomarker stable complex. The transcribed pri-miR-96 was processed to pre- for diagnostic or prognostic application. Consistent with this miR-96 and miR-96. The recognition of miR-96 to the 3′UTR of notion, it has been shown that the levels of miR-96 were elevated AKT1S1 mRNA results in downregulation of AKT1S1 and disruption in breast and prostate cancers20,25,26 and correlated with WHO of the AKT1S1-mTOR complex. Therefore, TGFβ signaling ensures tumor grades.20 In addition, we and other investigators persistent mTOR function and promotes malignant cancer have found that miR-96 targets several tumor suppressors phenotypes leading to poor prognostic outcomes. including AKT1S1 (Figures 2–4), FOXO1 (Supplementary Figures S1C and D)20,25 and FOXO3a.26 Therefore, miR-96 can be considered as a metastamir/or oncomir in prostate cancer. DISCUSSION The biological function of miR-96, similar to TGFβ, also exhibits It has been shown that PTEN and p53 tumor suppressor genes are paradoxical property. Although we and others have shown that commonly mutated in cancer.23,24 However, only combined miR-96 stands as a metastamir/or oncomir in prostate cancer,

Figure 6. miR-96 promotes bone metastasis and contributes to poor survival rates in prostate cancer. (a) Selected histological images from the femurs of mice injected with DU145 cells expressing vector (top) or miR-96 precursor (bottom). Tumor cells replacing the bone marrow, destroying neighbor bone are shown in areas labeled with arrows. Scale bar: 100 μm. (b) Quantiﬁcation of bone metastasis in mice injected with DU145 cells expressing vector or miR-96 precursor (n = 10). (c) Survival rate of tumor-bearing mice following intracardiac injections of cells from b. EV: empty vector, 96 pre: vector expressing miR-96 precursor. (d) Model for TGFβ signaling regulates mTOR functions leading to bone metastases: TGFβ stimulates mTOR functions through the Smad-dependent canonical pathway to induce pri-miR-96 transcription followed by processing to pre-miR-96 and miR-96. miR-96 downregulates AKT1S1 through targeting its mRNA (dialogue box) and derepresses the inhibitory effects of AKT1S1 on mTOR kinase.

in pancreatic cancer, miR-96 seems to be involved in tumor- suppressor functions.27 On the contrary, late-stage/or metastatic suppressive function by suppressing KRAS.27 It has been observed tumor cells may have acquired mutations that overcome the that re-activation of TGFβ signaling can contribute to cell survival suppressor effects of miR-96. when the RAS signaling is blocked, for example, using cetuximab The statuses of tumor suppressors could be a key factor in or vemurafenib to inhibit the RAS pathway, TGFβ signaling is determining the direction of the paradoxical functions of miR-96. induced to confer survival and resistance.28,29 Therefore, it is Experiments showing miR-96 as a tumor suppressor27 were using possible that TGFβ signaling is normally suppressed in cancer cells several pancreatic cell lines (for example, MIA PaCa-2 and PANC-1) with sustained RAS signaling for growth and proliferation. that exhibited wild-type p53 function (for example, p53-depedent Alternatively, the presence of both RAS and TGFβ signaling cell cycle arrest31). However, the mouse adenocarcinoma cell lines pathways may produce too strong a survival signal that triggers used in our study are p53 null (that is, AC1 and AC3) and the oncogene-induced senescence.29 Early-stage tumor cells may be human prostate cancer cell lines (for example, PC3 and DU145) initiated primarily through RAS signaling,30 and under that are also p53 mutated with poor p53-dependent damage environment, reactivation of TGFβ signaling or exogenously responses.32,33 Using two other prostate cancer cell lines with introducing miR-96 could suppress RAS levels and exhibit tumor wild-type p53 (LNCap and 22Rv1),34 we tried to monitor the

Oncogene (2015) 4767 – 4776 © 2015 Macmillan Publishers Limited miR-96 regulates mTOR signaling in prostate cancer MK Siu et al 4775 effects of miR-96 on invasiveness but failed to detect the invasion In vitro growth assay property even in the parental cells. Nonetheless, when we In vitro growth curves were performed as described in Supplementary compared with the control cell (EV), overexpression of exogenous information using a density of 2 × 103 cells per well. miR-96 precursor did not promote migration rate (96 pre, Supplementary Figure S2D). It is possible that only under p53- IHC staining β defected conditions TGF -miR-96 signaling exhibits tumor pro- The clinical samples used 24 primary prostate tumors were collected from gression functions. However, this interpretation could be further Wan Fang Hospital, Taipei Medical University, Taiwan. Immunohistochem- complicated by the functional androgen-receptor signaling in the 34 istry was performed using AKT1S1 (Cell Signaling, Denvers, MA, USA) and p53 wild-type cell lines (LNCap and 22Rv1). The androgen- phospho-Smad2 (Millipore, Billerica, MA, USA) antibodies at 1:100 dilution receptor-positive cells may attenuate cellular response to TGFβ as described in Supplementary information. through multiple mechanisms in addition to downregulation of the TGFβ receptor.35 FACS analysis It has been shown that AKT1S1 forms a tight complex with fl mTOR kinase and negatively regulates its activity. Phosphorylation Promoter functional analysis using FACS and relative median uorescent intensity value was measured as described in Supplementary information. of AKT1S1 by AKT kinase can dissociate the complex and activate mTOR functions.12,13 In addition to the PI3K-AKT pathway, our studies provide another mechanism by which the mTOR kinase ChIP assay activity is regulated by Smad-dependent miR-96 (Figures 2 and 3 ChIP assays were performed using the EZ magna ChIP A kit (Millipore) with and Supplementary Figure S1F). As mTOR activation is important a modified protocol as described in Supplementary information. in tumor progression, the combinational strategies (phosphorylation and downregulation) to shut off the inhibitory effects of Animal studies AKT1S1 may lead to either hyperactivation or sustained activation Animal work was performed in accordance with a protocol approved by of mTOR kinase. Consequently, the mTOR activity following TGFβ the Taipei Medical University Animal Care and Use Committee. To analyze signaling is ensured to facilitate metastatic progression. Further metastasis, 5-week-old male nude mice (NLAC, Taipei) were injected studies are required to determine whether there are differential intracardiacly with 1 × 105 tumor cells and histology imaging of bone in the roles in the two pathways to activate mTOR functions. femur of tumor-bearing mice were as described.36 The level of bone metastasis in mouse was quantified based on the following criteria: (0) no metastasis, (1) bone lesion covers less than ¼ of the bone width, (2) bone MATERIALS AND METHODS lesion involves ¼–½ of bone width, (3) bone lesion across ½–¾ of bone Cell culture width and (4) bone lesion is more than ¾ of bone width. The derived The AC1 and AC3 cell lines were isolated from 4-month-old PbCre4+; bone metastasis score for each mouse represents the sum of scores of all Ptenfl/fl;TP53fl/fl;Luc+ prostate tumors. Cloned cell lines using limiting bone lesions from four limbs. For survival studies, mice were killed when dilution were established as described.17 The benign prostate hyperplastic one of the following situations applied: 10% loss of body weight, paralysis epithelial DHMLE and metastatic RasB121 cell lines were provided by Dr Yin or head tilting. All animal studies were repeated three times. (NCI, NIH, Bethesda, MD, USA). AC1, AC3 and DHMLE cell lines were cultured in PrEGM media (Lonza, Walkersville, MD, USA). DU145, PC3, Clinical outcome and correlation analyses using human data sets LNCap, 22Rv1 and RasB1 cell lines were cultured in RPMI 1640 media We used mRNA expression data from a public human prostate cancer data supplemented with 10% fetal calf serum (FCS). Transient transfections 19 – were carried out using Lipofectamine RNAiMAX (Invitrogen, Carlsbad, CA, set. The study was conducted under Memorial Sloan Kettering Cancer USA). Dosage of TGFβ and SB431542 was TGFβ (10 ng/ml) and SB431542 Center Institutional Review Board approval on 28 normal, 151 primary and 19 metastatic samples as described in Supplementary information. gene (1 μM). set enrichment analysis was performed using the software from the Broad Institute.37 Real-time RT–PCR Total RNA was isolated using mirVana PARIS RNA isolation system (Ambion, Austin, TX, USA). RT–PCR was performed as described in Supplementary Statistical analysis information. In vivo animal results and clinical outcome analyses are expressed as plots showing the median and box boundaries extending between 25th and 75th percentiles, with whiskers down to the minimum and up to the Western blot analysis maximum value. All in vitro data were presented as means ± s.e.m. Cells were lysed with RIPA buffer containing complete protease inhibitors Statistical calculations were performed with GraphPad Prism (GraphPad (Roche, San Francisco, CA, USA) plus the phosphatase inhibitors (Roche), Software Inc., La Jolla, CA, USA) analysis tools. Differences between 25 mM β‐glycerophosphate, 10 mM sodium fluoride and 1 mM Sodium 18 individual groups were analyzed by one-way or two-way analysis of Vanadate were performed as described. Primary antibodies were variance test. Bonferroni’s post test was used for comparisons among three incubated overnight at 4 °C using the dilutions listed in Supplementary or more groups. Log-rank test was used for survival curve analysis. P-values Tables (Supplementary Table S3). less than 0.05 were considered statistically significant. mR luciferase assay Cells were transfected with 1 μg of mouse or human AKT1S1 3′UTR reporter CONFLICT OF INTEREST and 1 μg of empty vector or plasmid encoding miR-96 precursor. The The authors declare no conflict of interest. psiCHECK-2 vectors contain both firefly and Renilla luciferase reporters. Cell extracts were prepared 24 h after TGFβ (10 ng/ml) or SB431542 (1 μM) treatment, and the luciferase and Renilla activities were measured using ACKNOWLEDGEMENTS the Dual Luciferase Reporter Assay System (Promega, Madison, WI, USA) as This work was jointly supported by grants from the Taipei Medical University-Wan described in Supplementary information. Fang Hospital (102TMU-WFH-05) to Y-N Liu, Taipei Medical University (TMU102-AE1- B30) to Y-C Tsai, the Ministry of Science and Technology (NSC102-2320-B-038-001) of Invasion assay Taiwan to Y-N Liu, and the Ministry of Science and Technology (NSC103-2311-B-038- Invasion assays were carried out as described in Supplemental information 001) of Taiwan to Y-C Tsai. We also thank Dr Ji-Hshiung Chen (Tzu Chi University) for using 1 × 106 cells and RPMI 1640 media supplemented with 10% FCS in reading the manuscript and for his comments and helpful suggestions. We also thank both chamber. Dr Orla Casey for reviewing our manuscript and discussion.

© 2015 Macmillan Publishers Limited Oncogene (2015) 4767 – 4776 miR-96 regulates mTOR signaling in prostate cancer MK Siu et al 4776 REFERENCES 21 Yin JJ, Zhang L, Munasinghe J, Linnoila RI, Kelly K. Cediranib/AZD2171 inhibits 1 Ikushima H, Miyazono K. TGFbeta signalling: a complex web in cancer progres- bone and brain metastasis in a preclinical model of advanced prostate cancer. sion. Nat Rev Cancer 2010; 10: 415–424. Cancer Res 2010; 70: 8662–8673. 2 Bhowmick NA, Chytil A, Plieth D, Gorska AE, Dumont N, Shappell S et al. TGF-beta 22 Majumder PK, Febbo PG, Bikoff R, Berger R, Xue Q, McMahon LM et al. mTOR signaling in fibroblasts modulates the oncogenic potential of adjacent epithelia. inhibition reverses Akt-dependent prostate intraepithelial neoplasia through Science 2004; 303:848–851. regulation of apoptotic and HIF-1-dependent pathways. Nat Med 2004; 10: 3 Li X, Placencio V, Iturregui JM, Uwamariya C, Sharif-Afshar AR, Koyama T et al. 594–601. Prostate tumor progression is mediated by a paracrine TGF-beta/Wnt3a 23 Chen Z, Trotman LC, Shaffer D, Lin HK, Dotan ZA, Niki M et al. Crucial role of p53- signaling axis. Oncogene 2008; 27: 7118–7130. dependent cellular senescence in suppression of Pten-deficient tumorigenesis. 4 Ao M, Williams K, Bhowmick NA, Hayward SW. Transforming growth factor-beta Nature 2005; 436:725–730. promotes invasion in tumorigenic but not in nontumorigenic human prostatic 24 Martin P, Liu YN, Pierce R, Abou-Kheir W, Casey O, Seng V et al. Prostate epithelial epithelial cells. Cancer Res 2006; 66: 8007–8016. Pten/TP53 loss leads to transformation of multipotential progenitors and epi- 5 Lee C, Sintich SM, Mathews EP, Shah AH, Kundu SD, Perry KT et al. Transforming thelial to mesenchymal transition. Am J Pathol 2011; 179: 422–435. growth factor-beta in benign and malignant prostate. Prostate 1999; 39:285–290. 25 Fendler A, Jung M, Stephan C, Erbersdobler A, Jung K, Yousef GM. The anti- 6 Tian M, Schiemann WP. The TGF-beta paradox in human cancer: an update. Future apoptotic function of miR-96 in prostate cancer by inhibition of FOXO1. PLoS ONE Oncol 2009; 5: 259–271. 2013; 8: e80807. 7 Derynck R, Zhang YE. Smad-dependent and Smad-independent pathways in TGF- 26 Lin H, Dai T, Xiong H, Zhao X, Chen X, Yu C et al. Unregulated miR-96 induces cell beta family signalling. Nature 2003; 425: 577–584. proliferation in human breast cancer by downregulating transcriptional 8 Lamouille S, Derynck R. Cell size and invasion in TGF-beta-induced epithelial to factor FOXO3a. PLoS ONE 2010; 5: e15797. mesenchymal transition is regulated by activation of the mTOR pathway. J Cell 27 Yu S, Lu Z, Liu C, Meng Y, Ma Y, Zhao W et al. miRNA-96 suppresses KRAS and Biol 2007; 178: 437–451. functions as a tumor suppressor gene in pancreatic cancer. Cancer Res 2010; 70: 9 Lamouille S, Connolly E, Smyth JW, Akhurst RJ, Derynck R. TGF-beta-induced 6015–6025. activation of mTOR complex 2 drives epithelial-mesenchymal transition and cell 28 Bedi A, Chang X, Noonan K, Pham V, Bedi R, Fertig EJ et al. Inhibition of TGF-beta invasion. J Cell Sci 2012; 125: 1259–1273. enhances the in vivo antitumor efficacy of EGF receptor-targeted therapy. Mol 10 Bertoldo F, Silvestris F, Ibrahim T, Cognetti F, Generali D, Ripamonti CI et al. Cancer Ther 2012; 11: 2429–2439. Targeting bone metastatic cancer: role of the mTOR pathway. Biochim Biophys 29 Sun C, Wang L, Huang S, Heynen GJ, Prahallad A, Robert C et al. Reversible and Acta 2014; 1845: 248–254. adaptive resistance to BRAF(V600E) inhibition in melanoma. Nature 2014; 508: 11 Guertin DA, Sabatini DM. Defining the role of mTOR in cancer. Cancer Cell 2007; 118–122. 12:9–22. 30 Brown K, Quintanilla M, Ramsden M, Kerr IB, Young S, Balmain A. v-ras genes from 12 Sancak Y, Thoreen CC, Peterson TR, Lindquist RA, Kang SA, Spooner E et al. Harvey and BALB murine sarcoma viruses can act as initiators of two-stage mouse PRAS40 is an insulin-regulated inhibitor of the mTORC1 protein kinase. Mol cell skin carcinogenesis. Cell 1986; 46: 447–456. 2007; 25: 903–915. 31 Yadav V, Sultana S, Yadav J, Saini N. Gatifloxacin induces S and G2-phase cell cycle 13 Vander Haar E, Lee SI, Bandhakavi S, Griffin TJ, Kim DH. Insulin signalling to mTOR arrest in pancreatic cancer cells via p21/p27/p53. PLoS ONE 2012; 7: e47796. mediated by the Akt/PKB substrate PRAS40. Nat Cell Biol 2007; 9:316–323. 32 Gogna R, Madan E, Kuppusamy P, Pati U. Re-oxygenation causes hypoxic tumor 14 Zoncu R, Efeyan A, Sabatini DM. mTOR: from growth signal integration to cancer, regression through restoration of p53 wild-type conformation and post- diabetes and ageing. Nat Rev Mol Cell Biol 2011; 12:21–35. translational modifications. Cell Death Dis 2012; 3: e286. 15 Nicoloso MS, Spizzo R, Shimizu M, Rossi S, Calin GA. MicroRNAs--the micro 33 Liu C, Zhu Y, Lou W, Nadiminty N, Chen X, Zhou Q et al. Functional p53 deter- steering wheel of tumour metastases. Nat Rev Cancer 2009; 9:293–302. mines docetaxel sensitivity in prostate cancer cells. Prostate 2013; 73: 418–427. 16 Butz H, Racz K, Hunyady L, Patocs A. Crosstalk between TGF-beta signaling and 34 Skjoth IH, Issinger OG. Profiling of signaling molecules in four different human the microRNA machinery. Trends Pharmacol Sci 2012; 33:382–393. prostate carcinoma cell lines before and after induction of apoptosis. Int J Oncol 17 Liu YN, Abou-Kheir W, Yin JJ, Fang L, Hynes P, Casey O et al. Critical and reciprocal 2006; 28: 217–229. regulation of KLF4 and SLUG in transforming growth factor beta-initiated prostate 35 Mishra S, Deng JJ, Gowda PS, Rao MK, Lin CL, Chen CL et al. Androgen cancer epithelial-mesenchymal transition. Mol Cell Biol 2012; 32: 941–953. receptor and microRNA-21 axis downregulates transforming growth factor beta 18 Liu YN, Yin JJ, Abou-Kheir W, Hynes PG, Casey OM, Fang L et al. MiR-1 and receptor II (TGFBR2) expression in prostate cancer. Oncogene 2014; 33: miR-200 inhibit EMT via Slug-dependent and tumorigenesis via Slug-independent 4097–4106. mechanisms. Oncogene 2013; 32:296–306. 36 Yin J, Pollock C, Tracy K, Chock M, Martin P, Oberst M et al. Activation of the 19 Taylor BS, Schultz N, Hieronymus H, Gopalan A, Xiao Y, Carver BS et al. Integrative RalGEF/Ral pathway promotes prostate cancer metastasis to bone. Mol Cell Biol genomic profiling of human prostate cancer. Cancer Cell 2010; 18:11–22. 2007; 27: 7538–7550. 20 Haflidadottir BS, Larne O, Martin M, Persson M, Edsjo A, Bjartell A et al. Upregu- 37 Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA et al. lation of miR-96 enhances cellular proliferation of prostate cancer cells Gene set enrichment analysis: a knowledge-based approach for interpreting through FOXO1. PLoS ONE 2013; 8: e72400. genome-wide expression profiles. Proc Natl Acad Sci USA 2005; 102: 15545–15550.

Supplementary Information accompanies this paper on the Oncogene website (http://www.nature.com/onc)

Transforming Growth Factor-&Beta; Promotes Prostate Bone

Transforming Growth Factor-Β Promotes Prostate Bone