Although cancer stem-like cells (CSC) are thought to be the most tumorigenic, metastatic, and therapy-resistant cell subpopulation within human tumors, current therapies target bulk tumor cells while tending to spare CSC. In seeking to understand mechanisms needed to acquire and maintain a CSC phenotype in prostate cancer, we investigated connections between the ETS transcription factor ESE3/EHF, the Lin28/let-7 microRNA axis, and the CSC subpopulation in this malignancy. In normal cells, we found that ESE3/EHF bound and repressed promoters for the Lin28A and Lin28B genes while activating transcription and maturation of the let-7 microRNAs. In cancer cells, reduced expression of ESE3/EHF upregulated Lin28A and Lin28B and downregulated the let-7 microRNAs. Notably, we found that deregulation of the Lin28/let-7 axis with reduced production of let-7 microRNAs was critical for cell transformation and expansion of prostate CSC. Moreover, targeting Lin28A/Lin28B in cell lines and tumor xenografts mimicked the effects of ESE3/EHF and restrained tumor-initiating and self-renewal properties of prostate CSC both in vitro and in vivo. These results establish that tight control by ESE3/EHF over the Lin28/let-7 axis is a critical barrier to malignant transformation, and they also suggest new strategies to antagonize CSC in human prostate cancer for therapeutic purposes. Cancer Res; 76(12); 3629–43. ©2016 AACR.
Prostate cancer is the most common malignancy and the second most frequent cause of cancer-related mortality in men in developed countries (1, 2). Several studies have provided evidence of the presence of self-renewing tumor-initiating stem-like cancer cells in human cancers, including prostate cancer (3–5). Cancer stem-like cells (CSC) can derive from transformation of tissue/adult stem cells or from more differentiated progenitor cells that acquire stem-like properties (6, 7). CSCs within the primary tumors are likely a major source of tumor heterogeneity, disease progression, and treatment failure. Our knowledge of the factors governing the behavior of CSCs in prostate cancer is limited (5, 8). Understanding the pathways controlling the expansion and maintenance of prostate CSCs could be an important step toward development of more effective CSC-directed strategies for treatment of prostate cancer.
ETS transcription factors are important elements in differentiation and developmental programs in many tissues. Expression of ETS factors is tightly regulated according to tissue-specific and time-dependent programs (9, 10). Deregulated expression of ETS factors has oncogenic consequences altering tissue developmental programs and is one of the most frequent findings in human tumors. About 50% of prostate cancers exhibit gene rearrangements and ectopic expression of ETS genes, like ERG and ETV1 (11–14). ESE3/EHF is an ETS factor normally expressed in epithelial cells, including prostate epithelial cells (10). We reported previously that ESE3/EHF is downregulated frequently in prostate tumors (15, 16). More recently, we showed that ESE3/EHF has a key role in controlling the differentiation program of prostate epithelial cells (17). Loss of ESE3/EHF altered cell differentiation and conferred to prostate epithelial cells a CSC-like phenotype along with tumor-initiating and metastatic capability. In clinical samples loss of ESE3/EHF expression marked a subset of prostate tumors with enrichment of CSC transcriptional features and characteristics of clinically aggressive prostate tumors (17). Mechanistically, ESE3/EHF acts as transcriptional activator and repressor of a large network of target genes (17). We showed that in normal prostate epithelial cells ESE3/EHF induces genes of the epithelial cell differentiation and represses genes connected with self-renewal and the CSC phenotype, like Nanog, POU5F1 (Oct4), BMI-1, EZH2 (17). Expression of these genes was conversely upregulated in transformed prostate epithelial cells, prostate cancer cell lines and human tumors with loss of ESE3/EHF expression, indicating their contribution to the acquisition of CSC properties in prostate cancer.
Lin28 is a highly conserved RNA-binding protein and one of the key embryonic stem cell factors (18, 19). Lin28A (LIN28) and its paralog Lin28B (LIN28B) repress the processing of pri- and pre-miRNAs of the let-7 family into mature microRNAs (miRNAs), thus preventing differentiation of embryo stem cells and maintaining self-renewal and pluripotency (20–25). Lin28A and Lin28B are frequently overexpressed in human cancers (26–28). Lin28 promotes neoplastic transformation by repressing let-7 miRNAs, which act as tumor suppressors inhibiting expression of key oncogenes, like RAS, MYC, and HMGA2 (20, 21, 29). In addition, the Lin28/let-7 axis has a critical role by regulating tumor-initiating and self-renewal properties of CSCs in human cancers (30–33).
In the attempt to identify relevant mediators of the prostate CSC phenotype and actionable targets for drug discovery and therapeutic intervention, we examined the relationship between ESE3/EHF and the Lin28/let-7 axis in prostate cancer. We show here that ESE3/EHF represses transcription of Lin28A and Lin28B and concomitantly sustains transcription and processing of let-7 pri-miRNAs to mature let-7 miRNAs. The dual transcriptional and posttranscriptional control exerted by ESE3/EHF on the Lin28/let-7 axis ensures tight control of this key developmental program and proper balance between cell differentiation and self-renewal in the normal prostate epithelium. Targeting the Lin28/let-7 axis in prostate cancer cell lines and tumor xenografts with ESE3/EHF deregulation antagonizes tumor-initiating and self-renewal properties of prostate CSCs and leads to reduced tumor growth, suggesting that it could be a valid strategy for treatment of clinically aggressive prostate cancers.
Materials and Methods
Cell lines, transfection, and selection of cell clones
Immortalized human prostate epithelial cells (PrECs; ref. 15) and RPWE-1 with stable knockdown of ESE3/EHF by shRNAs were established as previously described (16). LNCaP, DU145, and PC3 were obtained from the ATCC, which performs cell line characterization based on DNA profiling (short tandem repeat analysis), and maintained in RPMI-1640 (Gibco) supplemented with 10% FBS. DU145 cells expressing ESE3/EHF were generated after transfection with ESE3/EHF expression vector and selection with G418 (15). Cells were used within 6 months of culturing and regularly checked for Mycoplasma contamination using the MycoAlert Mycoplasma Detection Kit (Lonza).
RNA extraction, quantitative real-time RT-PCR, miRNA precursor, mature miRNA, antagomiR, and siRNAs
RNA was extracted by the Direct-zol RNA MiniPrep Kit (Zymo Research). Quantitative real-time RT-PCR (qRT-PCR) was carried out using 20 ng of RNA as template for the SYBR Green Fast One Step Kit (Qiagen). qRT-PCR primers are reported in Supplementary Table S1. For let-7b miRNA overexpression, cells were transiently transfected with 50 nmol/L of the specific miRNA precursor (pre–miR let-7b, Ambion) or negative control (Control #1, Ambion). For let-7b inhibition cells were transiently transfected with 40 nmol/L of the specific LNA antagomiR (Mercury LNA Power Inhibitor; Exiqon) or a scrambled control (Negative Control A; Exiqon). For miRNA expression analysis, 400 ng of purified RNA was retrotranscribed using the TaqMan MicroRNA Reverse Transcription Kit (Applied Biosystem) with specific primers and the cDNA was subjected to TaqMan Probe-based Real Time PCR using TaqMan MicroRNA Assays and TaqMan Universal PCR Master Mix (Applied Biosystem). The expression was normalized to RNU6 (Control miRNA assay; Applied Biosystem) and 18S Eukaryotic (18S rRNA Endogenous Control; Applied Biosystem). For transient gene knockdown cells were transfected with siRNAs directed to Lin28A and Lin28B (siRNA Silencer select; Ambion) or control (siGL3) siRNA (17) using jetPRIME (Polyplus).
Cell lysates were prepared using RIPA buffer with protease inhibitor cocktail (Roche) and phosphatase inhibitor cocktail (PhosStop; Roche). Total cell extracts were separated by SDS-PAGE and transferred to nitrocellulose membranes (PROTRAN). Antibodies directed to the following proteins were used for immunoblot analysis: total Lin28 (11724-1-AP, Proteintech), Lin28A (16177-1-AP Proteintech), Lin28B (16178-1-AP Proteintech), ESE3/EHF (sc-367574, Santa Cruz Biotechnology), glyceraldehyde-3-phosphatedehydrogenase (GAPDH; Millipore) and tubulin (CP06; Calbiochem).
Soft agar and in vitro prostatosphere forming and self-renewal assay
Soft-agar assays were performed as previously described (34). The prostatosphere assay was previously described (17). The sphere-forming efficiency (SFE) was determined as the percentage of prostatosphere relative to the number of cells plated at the start of the experiment. Each experiment was carried out in triplicate and repeated at least three times.
Expression vectors, reporter constructs, and luciferase assays
The pEF5/FRT/Lin28A-V5 and pCMV6-XL4 LIN28B plasmid were kindly provided by Wilbert and colleagues (35) and by Loughlin and colleagues (36), respectively, and were transfected in cells using jetPRIME (Polyplus). To analyze Lin28 and let-7b promoter activity, we used the following luciferase reporters: pGL3-basic hLin28A-NRSE, provided by Gunsalus and colleagues (37), pGL3-basic-Lin28b P1-IRES, provided by Chang and colleagues (38), and pGL3-let7b-1.5K, provided by D. Wang and colleagues (39). Luciferase reporter assays were performed as previously described (17). Results were normalized to Renilla luciferase and expressed as Relative Luciferase Activity (RLA). Each experiment was performed in triplicate and repeated at least three times.
Synthesis of siRNA for in vivo studies
Guide and passenger sequences of phosphodiester oligoribonucleotides (ORN) for siLin28B-2 were AAAUCCUUCCAUGAAUAGUTT (mass calc.: 6599.1; mass obs.: 6598.2) and ACUAUUCAUGGAAGGAUUUTT (mass calc.: 6656.1; mass obs.: 6655.3; ref. 40). Chemicals for ORN synthesis were from Aldrich and TCI (Sigma-Aldrich Chemie GmbH, D-89555 Steinheim). Phosphoramidites were from Thermo Fisher Scientific. The activator 5-benzylthiotetrazole (BTT) was from Biosolve. ORNs were synthesized on an MM12 synthesizer from Bio Automation Inc. on 500 Å UnyLinker CPG from ChemGenes. Coupling time for phosphoramidites was 2 × 90 seconds. The ORNs were purified on an Agilent 1200 series preparative HPLC fitted with a WatersXBridge OST C-18 column, 10 × 50 mm, 2.5 μm at 60°C. The RNA phosphoramidites were prepared as a 0.08 mol/L solutions in dry acetonitrile (ACN); BTT was prepared as a 0.24 mol/L solution in dry ACN. Oxidizer was prepared as a 0.02 mol/L I2 solution in THF/Pyridine/H2O (70:20:10, w/v/v/v). Capping reagent A was: THF/lutidine/acetic anhydride (8:1:1) and capping reagent B was: 16% N-methylimidazole/THF. Deblock solution was a 3% dichloroacetic acid in dichloroethane. The cleavage from the solid support and the deprotection of the nucleotides was affected by incubation of the CPG-support for 2 hours, at 65°C in gaseous methylamine at 1.8 bar. Deprotection of 2′-TBDMS (tert-butyldimethylsilyl-) was carried out at 1.5 hours, at 70°C in a mixture of dry 1-N-methyl-2-pyrrolidone/triethylamine/trimethylamine.3HF. Running buffer for HPLC purification of ORNs: buffer A (0.1 mol/L triethylammonium acetate), buffer B (methanol): gradient for the DMT-on purification: 20% to 60% buffer B over 5 minutes; gradient for the DMT-off purification: 5% to 35% buffer B over 5 minutes. Fractions containing the product were collected and dried in a miVac duo SpeedVac from Genevac. ORNs were analyzed by LC-MS (Agilent 1200/6130 system) on a Waters Acquity OST C-18 column, 2.1 × 50 mm, 1.7 μmol/L, 65°C. Buffer A: 0.4 mol/L HFIP, 15 mmol/L triethylamine; buffer B: MeOH. Gradient: 7% to 35% B in 14 minutes; flow-rate: 0.3 mL/min.
Animals and tumor xenografts
Mice were purchased from the Harlan Laboratories. Mice were maintained under pathogen-free conditions with food and water provided ad libitum and their general health status was monitored daily. All protocols involving animals were conducted in conformity with the institutional guidelines for animal experimentation and in compliance with national and international policies. Study protocol was approved by the Swiss Veterinary Authority. For subcutaneous tumor xenografts and in vivo self-renewal experiments prostatosphere-derived and bulk adherent growing ESE3KD-PrECs cells were inoculated (2 × 105 cells/site) with Matrigel in the flank of NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice (n = 4/group). For in vivo self-renewal experiments primary tumor xenografts derived from ESE3KD-PrECs prostatospheres were dissociated into single cell suspensions and dissociated tumor cells were implanted (2 × 105 cells/site) subcutaneously as above for two generations (n = 4/group). In vivo serial dilutions experiments were performed with adherent and prostatosphere-derived ESE3KD-PrECs cells by inoculating 2 × 105, 1 × 105 and 0.5 × 105 cells/site subcutaneously in NSG mice (n = 4/group) as above. To assay in vivo tumorigenic capacity following in vitro knockdown, ESE3KD-PrECs, DU145 or PC3 cells (2 × 106 cells/site) transfected with control or Lin28-targeting siRNAs were injected in the flank of NSG mice (n = 4/group). Tumor engraftment and growth were monitored twice a week. For systemic treatment with siRNA, DU145 cells with stable expression of luciferase reporter gene were injected (5 × 106 cells/site) with Matrigel in the flank of athymic nude mice (Balb/c nu/nu, 4–6-weeks-old; n = 4/group). Mice with subcutaneous tumor xenografts were then injected intraperitoneally with siLin28B-2 or control siRNA formulated with in vivo jetPEI (Polyplus Transfection) at the dose of 2 mg/kg/body weight/d three times a week. Tumor growth was monitored as above. To assess in vivo self-renewal capability of tumor cells from DU145 xenografts, cells dissociated from control and siLin28B-2–treated tumors were re-injected (5 × 105 cells/site) with Matrigel in the flank of athymic nude mice (n = 2/group).
Immunohistochemistry and immunocytochemistry
IHC on histologic tissues samples was performed using an antibody detecting total Lin28 protein (Proteintech, 1:25 dilution, code n. 11724-1-AP Rb Poly), and an anti-Ki67 antibody (Lab Vision Corporation, Clone SP6, cat. RT-9106-R7). The specificity of the antibodies was previously confirmed by Western blot analysis. Cell nuclei were counterstained with hematoxylin solution. Positive samples for each antibody and negative samples, in which the primary antibody was omitted, were used as controls. Slides were evaluated by three investigators. Average of values scored by the three investigators was calculated. Following a meeting, each sample was defined positive and negative and/or cytoplasmic and nuclear in full concordance between the three readers. For immunocytochemistry (ICC), harvested cells were washed in PBS by centrifugation and then the concentration adjusted to 5 × 106 cells/mL in PBS. Cells were attached to slides using Cytospin Cytocentrifuge (Thermo Scientific) at 800 rpm for 4 minutes. Cells were fixed and permeabilized with Acetone:Methanol, 1:1. After blocking with 5% BSA, cells were incubated with anti-Lin28 antibody (Proteintech, 1:75, code n. 11724-1-AP Rb Poly). Cell nuclei were counterstained with hematoxylin solution and finally, the sections were dehydrated and mounted in a suitable organic mounting medium.
All steps for flow cytometry were performed in PBS supplemented with 0.5% BSA, and 2 mmol/L EDTA. Fluorescein di-beta-D-galactopyranoside (FDG) was purchased from Invitrogen and used for analysis in accordance with the manufacturer's instructions. Cell sorting was performed with a FACSAria III sorter (BD Biosciences).
Computational search for ETS binding sites on selected gene promoters was performed using Motifviz (biowulf.bu.edu/MotifViz). Chromatin immunoprecipitation (ChIP) was performed with anti-ESE3 (Clone 5A5.5; Lab Vision); anti-acetylated H3 (Upstate, Millipore); anti-H3K9 2met (Upstate, Millipore); anti-H3K27 3met (Upstate, Millipore), and IgG control antibody. Samples were analyzed as previously described (16) by qRT-PCR. Primer sets are reported in Supplementary Table S1.
Tissue samples were collected at Department of Pathology (IRCCS Multimedica, Italy) with the approval of the Ethics Committee of Regione Lombardia, Italy, and patient written informed consent. Primary and metastatic tumor samples were obtained from patients with organ-confined disease treated with radical prostatectomy and from patients with advanced metastatic disease undergoing transurethral resection, respectively. Tissue microarrays containing samples from normal and adenocarcinoma were prepared from paraffin-embedded tissues.
ESE3/EHF, Lin28A, and Lin28B expression correlation analysis
A human prostate cancer dataset (41) was retrieved from GEO (GSE21034). The dataset was based on the microarray platform “Affymetrix Human Exon 1.0 ST Array.” Only raw intensity data for prostate cancer samples were considered. Data were processed in R using the Bioconductor package “oligo” for Affymetrix arrays: sets were separately RMA normalized (with background correction) and quantile normalized at the probe level. Log2, normalized expression values for ESE3/EHF and Lin28A/B genes were extracted and their correlation was tested with the test for association between paired samples with Pearson's product moment correlation coefficient.
ESE3/EHF loss is linked to upregulation of Lin28A and Lin28B in prostate tumors
We previously showed that stable knockdown of ESE3/EHF in immortalized normal prostate epithelial PrECs and RWPE-1 cells (Supplementary Fig. S1A) leads to the acquisition of CSC traits and the upregulation of CSC-related genes (17). Parental prostate epithelial cells and their ESE3KD counterpart represent good models to investigate the pathways activated in prostate CSCs and driving CSC-enriched aggressive prostate tumors. In this context, we examined whether, similar to other CSC genes, expression of Lin28A and Lin28B was upregulated in ESE3KD cells. We observed a significant increase of Lin28A and Lin28B mRNAs in both ESE3KD-PrECs and ESE3KD-RWPE-1 cells (Fig. 1A). The total level of Lin28 protein, measured with an antibody that recognized both Lin28A and Lin28B, was also increased in ESE3KD-PrECs and ESE3KD-RWPE-1 cells (Fig. 1B). Consistently, ESE3KD-PrECs and ESE3KD-RWPE-1 cells exhibited lower levels of let-7 family miRNAs compared with the parental cells (Supplementary Fig. S1B). We reported previously that ESE3KD cells acquire CSC properties and form a larger number of prostatospheres (PS) than parental PrECs and RWPE-1 cells (17). Interestingly, we found that CSC-enriched PS derived from ESE3KD cells had substantially higher levels (3–6-fold) of Lin28A and Lin28B mRNAs compared with bulk ESE3KD-PrECs and ESE3KD-RWPE-1 cells growing as adherent monolayers (Fig. 1C). Consistently, total Lin28 protein was higher in PS of ESE3KD-PrECs compared with adherent growing ESE3KD-PrECs (Fig. 1D). This was confirmed by ICC that showed very intense Lin28 staining in PS derived from ESE3KD-PrECs (Fig. 1D).
Consistent with previous results (17), CSC-enriched PS derived from ESE3KD-PrECs (2 × 105 cells/site) formed tumors (G1 xeno) when implanted subcutaneously in NSG mice whereas adherent growing ESE3KD-PrECs in the same conditions did not (Fig. 1E). Serial dilution experiments confirmed the higher efficiency of tumor initiation by CSC-enriched PS cells compared with adherent growing ESE3KD-PrECs (Fig. 1F). Tumors were detected in 75% (3/4) and 25% (1/4) of mice injected with 2 × 105 and 1 × 105 PS-derived cells. In contrast, no tumors developed in mice injected with the same numbers of adherent growing cells. The minimal number of bulk ESE3KD-PrECs capable of forming tumors in 100% of mice was 2 × 106 cells as determined in additional experiments (see Fig. 3G). Thus, PS-derived ESE3KD-PrECs retained in vivo high tumor-initiating and stem-like properties. Furthermore, cells isolated from the PS-derived xenografts (G2 and G3 xeno) were highly tumorigenic when re-engrafted in NSG mice for consecutive generations, an indication that they retained tumor-initiating and self-renewal capability as expected for stem-like cancer cells (Fig. 1E).
Interestingly, we observed high expression of Lin28A and Lin28B mRNAs in the PS-derived xenografts at the first and consecutive passages in vivo in line with the in vitro data with the PS-enriched cell cultures (Fig. 1G). These findings were intriguing and suggested a link between loss of ESE3/EHF, upregulation of Lin28A/B, and expansion and maintenance of the CSC compartment in prostate cancer cell lines both in vitro and in vivo. Furthermore, these data pointed to Lin28A and Lin28B as possible critical factors driving the CSC phenotype and self-renewing properties in aggressive prostate cancers.
To determine whether the link between ESE3/EHF and Lin28 observed in our cell line models was seen in clinical samples, we analyzed gene-expression data from a large (n = 131) human prostate cancer dataset (Fig. 2A; ref. 41). We found that Lin28A and Lin28B were positively correlated with each other in primary prostate tumors (Pearson correlation = 0.33; P = 9.76E−05). Furthermore, there was a significant inverse correlation of ESE3/EHF with Lin28B (Pearson correlation = −0.516; P = 2.89E−10) and Lin28A (Pearson correlation = −0.216; P = 0.013). We evaluated Lin28 protein expression also by IHC using an antibody that recognized both Lin28A and Lin28B in primary prostate tumors (n = 28) and matched normal tissues (n = 17) and in an independent cohort of metastatic (n = 24) tumor samples. Lin28 staining was low or absent in normal prostate. Primary tumors were positive for Lin28 with weak or moderate Lin28 staining and a slight prevalence of cytoplasmic (35%) over nuclear (27%) staining (Fig. 2B and C). We did not find any correlation between Lin28 expression and Gleason score in primary tumors. However, metastatic tumors exhibited a higher percentage of tumors with positivity for Lin28 and a slight prevalence of cases with nuclear (61%) over cytoplasmic (54%) staining (Fig. 2D and E). Although the in vitro data show higher Lin28 expression in CSC-enriched prostatospheres, it is highly likely that CSCs represent only a small fraction of the Lin28-positive cancer cells detected by IHC both in primary and metastatic tumors. We evaluated Lin28 protein expression also in tumor xenografts derived from human prostate cancer cell lines with different levels of ESE3/EHF expression. We observed higher numbers of positive cells and prevalence of nuclear staining in xenografts of DU145 and PC3 cells, whereas low frequency and prevalently cytoplasmic staining was seen in xenografts of LNCaP cells (Supplementary Fig. S2). Therefore, the inverse correlation with ESE3/EHF and different expression patterns of Lin28 were maintained also in the tumor xenografts.
Lin28A and Lin28B promote malignant transformation and CSC-like phenotype in prostate epithelial cells
To determine the role played by Lin28 in prostate CSC, Lin28A and Lin28B were downregulated in ESE3KD-PrECs and ESE3KD-RWPE-1 cells using specific siRNAs. The extent of the knockdown was assessed by qRT-PCR and Western blot and the impact on Lin28 function demonstrated by the increased level of mature let-7b miRNA (Supplementary Fig. S3A–S3C). Knockdown of Lin28A and Lin28B reduced the ability of the cells to form colonies in soft-agar (Fig. 3A). Moreover, the ability to form PS in non-adherent growth conditions was significantly impaired for two consecutive generations, indicating a prolonged effect on the CSC-like PS forming cells upon a single siRNA transfection (Fig. 3B). We showed previously that induction of cell senescence, a form of growth arrest that antagonizes CSC expansion, coincided with loss of self-renewal capability in prostate CSCs upon c-Myc knockdown (42). We found that knockdown of Lin28A and Lin28B increased the percentage of senescent cells detected by FDG staining in ESE3KD-PrECs and ESE3KD-RWPE-1 cells (Fig. 3C). Thus, induction of cell senescence could contribute to the loss of clonogenic and PS forming potential of transformed ESE3KD cells. Importantly, knockdown of either Lin28A or Lin28B produced similar effects, indicating that both had relevant and independent roles in the induction of these cancer cell properties.
To further support the hypothesis that Lin28A and Lin28B are key factors in the induction of the prostate CSC phenotype, we assessed the effects of their overexpression in immortalized normal prostate epithelial cells. Forced overexpression in RWPE-1 cells, as shown by qRT-PCR and Western blotting (Fig. 3D), led to a substantial increase in colony formation in soft agar (Fig. 3E) and PS (Fig. 3F), indicating that Lin28A and Lin28B promoted the acquisition of tumorigenic and CSC-like properties. Similar effects were observed in PrECs (Supplementary Fig. S3D–S3F) and LNCaP cells (Supplementary Fig. S3G–S3I), in which overexpression of Lin28 increased substantially the formation of colonies in soft-agar and PS.
Impairment of CSC functions could result in persistent reduction of in vivo tumor-initiating capability of cancer cells. We evaluated the effects of Lin28 knockdown on the ability of bulk ESE3KD-PrECs to form tumors when implanted subcutaneously in NSG mice. ESE3KD-PrECs cells (2 × 106 cells/site) transfected with control (siGL3) siRNA formed tumors that grew exponentially after the initial lag phase (Fig. 3G). In contrast, cells transfected in vitro with Lin28A or Lin28B targeting siRNAs formed smaller masses that barely expanded over time. The level of Lin28A and Lin28B mRNA was reduced in the xenografts from cells treated with siLin28A and siLin28B compared with control xenografts (Fig. 3H). Interestingly, the level of total Lin28 protein determined by IHC was reduced in the siLin28-treated xenografts (Fig. 3G). Concomitantly, the expression of various CSC markers, like Nanog, Sox2, POU5F1, KLF4, and BMI-1 (17, 42), was significantly reduced in xenografts treated with siLin28A and siLin28B, consistent with a reduction of the CSC subpopulation (Fig. 3H). These findings suggested the activation of feedback loops that extend the effects of the transient knockdown of Lin28, leading to persistent repression of Lin28 and other CSC genes during the in vivo growth. Accordingly, ex vivo PS-forming assays performed with cells isolated from the tumor xenografts showed a significant and persistent reduction of PS-forming and self-renewal cells in xenografts treated with siLin28A and siLin28B compared with control xenografts (Fig. 3I). Thus, targeting Lin28A/B had a profound effect on the CSCs and their self-renewal capability in vivo. Furthermore, these results indicated that Lin28A/B played key roles in the cell transformation upon loss of ESE3/EHF and sustained the CSC-like phenotype and expansion of the CSC compartment in prostate tumors. Targeting Lin28A/B reversed the CSC properties in prostate cancer cell models in vitro and in vivo, suggesting that it could re-activate a latent differentiation/senescence program and that it might be a valid approach for selective CSC ablation in prostate tumors.
ESE3/EHF exerts dual transcriptional and posttranscriptional control on the Lin28/let-7 axis
ESE3/EHF and Lin28A/B expressions were inversely correlated in cell lines and prostate tumors. Furthermore, Lin28A/B had an important role in determining tumorigenic properties of transformed prostate epithelial cells. To better define the relationship with ESE3/EHF, we searched for ETS binding site (EBS) in the promoters of the Lin28A and Lin28B genes. Computational analysis of transcription factor–binding sequences revealed the presence of multiple EBS in both promoters. To test whether ESE3/EHF bound to the promoters and controlled Lin28A/B transcription, we selected the EBSs with the highest scores and nearest the transcription start sites (TSS) of the respective genes and performed ChIP assays (Fig. 4A). ESE3/EHF bound to the Lin28A and Lin28B promoters in RWPE-1 cells (Fig. 4B). However, no binding was detected in ESE3KD-RWPE-1 cells, in which both Lin28A/B proteins were highly expressed. Consistent with a repressive function on the Lin28A/B promoters, we found enrichment of repressive (H3K9me and H3K27me) histone marks in RWPE-1 cells (Fig. 4C). We performed luciferase reporter assays in RWPE-1 and ESE3KD-RWPE-1 cells using Lin28A and Lin28B promoter constructs (Fig. 4D). Activity of both promoter reporters was significantly reduced in RWPE-1 compared with ESE3KD-RWPE-1 cells, consistent with transcriptional repression by ESE3/EHF.
Lin28A/B block the production of mature let-7 miRNAs (20–25). Accordingly, we observed reduced levels of the mature forms of miRNAs of the let-7 family in ESE3KD-PrECs and ESE3KD-RWPE-1 cells compared with the corresponding parental cells (Supplementary Fig. S1B). This was consistent with reduced processing of let-7 miRNAs in the presence of high Lin28A/B level in cells with low ESE3/EHF expression. In line with this notion, we monitored in greater details the level of mature let-7b in different cells and conditions. Mature let-7b was significantly reduced in ESE3KD cells and in RWPE-1 and LNCaP cells upon forced overexpression of Lin28A (Fig. 5A and B). We hypothesized that let-7 miRNAs could be important mediators of ESE3/EHF prodifferentiation functions in prostate epithelial cells. Accordingly, antagonizing let-7b with a locked nucleic acid (LNA) anti-miRNA increased PS-forming ability of PrECs and RWPE-1 cells inducing an effect similar to Lin28A/B overexpression or ESE3/EHF knockdown (Fig. 5C). Conversely, forced overexpression of pre-let-7b reduced PS formation in DU145 cells (Supplementary Fig. S4A and S4B), phenocopying the effects of ESE3/EHF re-expression (17) or Lin28A/B ablation in these cells.
Interestingly, we observed that ESE3KD cells had also a reduced level of pri-let-7b (Fig. 5D). This suggested that ESE3/EHF could control the production of let-7 miRNAs also at the stage of transcription. Consistently, a high scoring EBS was found near the TSS in the let-7b promoter (Fig. 5E). ChIP assays demonstrated the presence of ESE3/EHF bound to the let-7b promoter in RWPE-1 cells and its absence in ESE3KD-RWPE-1 cells (Fig. 5F). The differential occupancy of the let-7b promoter by ESE3/EHF correlated with enrichment of activating (AcH3) histone marks in RWPE-1 cells compared with ESE3KD RWPE-1 cells (Fig. 5G). To confirm transcriptional activation of the let-7b promoter, we performed a luciferase reporter assay. Activity of the reporter was significantly higher in RWPE-1 compared to ESE3KD-RWPE-1 cells consistent with increased transcriptional activity in cells expressing ESE3/EHF (Fig. 5H). Interestingly, we found EBS in the promoters of other let-7 miRNAs (Supplementary Fig. S5), suggesting that other members of this miRNA family could be transcriptionally activated by ESE3/EHF.
To provide further support for this hypothesis, we assessed the effects of modulating ESE3/EHF expression on let-7b and pri–let-7b in DU145 cells. Expression of both pri–let-7b and mature let-7b was higher in ESE3/EHF-expressing DU145 cells compared with control DU145 cells, in agreement with dual transcriptional and posttranscriptional control of let-7b by ESE3/EHF (Fig. 5I). Thus, ESE3/EHF could sustain expression of let-7 miRNAs by promoting directly pri–let-7 miRNA transcription and indirectly the processing of let-7 precursors to the mature forms through repression of Lin28A/B expression. Interestingly, co-expression of ESE3/EHF in RWPE-1 cells partially rescued the effects of forced overexpression of Lin28A in soft-agar and PS-forming assays (Supplementary Fig. S6A and S6B). This could be attributed to the ability of ESE3/EHF to induce pri-let7 miRNA transcription and, independently of posttranscriptional inhibition of pre-miRNA processing, to partially restore the level of mature let-7b (Supplementary Fig. S6C).
Targeting of Lin28 in prostate cancer cells blocks CSC self-renewal and tumor growth in vivo
We showed previously that re-expression of ESE3/EHF in metastatic DU145 cells resulted in suppression of the malignant phenotype, inhibiting anchorage-independent growth, PS formation, and tumor growth in vivo (17). We reasoned that interfering with the Lin28/let7 axis could have similar effects and reverse the malignant phenotype of prostate cancer cells. DU145 and PC3 cells are models of aggressive, androgen-independent castration-resistant prostate cancer. These cells are highly tumorigenic in mice and have a high fraction of tumor-initiating and CSC-like cells (17, 42). DU145 and PC3 cells do not express ESE3/EHF and exhibit high levels of both Lin28A and Lin28B compared with normal PrECs (Fig. 6A). As seen with ESE3KD cell models, the levels of Lin28A and Lin28B further increased in CSC-enriched PS compared with bulk adherent DU145 and PC3 cells (Fig. 6B). This was particularly evident in DU145 cells with >200-fold increase of Lin28B expression in PS compared with adherent cells. High expression of Lin28 in PS derived from DU145 and PC3 cells was also demonstrated by ICC staining (Fig. 6B). We inhibited Lin28A and Lin28B with siRNAs in DU145 and PC3 cells (Supplementary Fig. S7A). Knockdown of Lin28A and Lin28B was associated with reduced formation of colonies in soft-agar (Fig. 6C) and PS (Fig. 6D) compared with control cells. Furthermore, Lin28A and Lin28B downregulation led to increased levels of let-7b, confirming inhibition of Lin28 function and restoration of let-7 miRNA synthesis (Supplementary Fig. S7B). Interestingly, we found a significant increase of senescent cells upon Lin28A and Lin28B knockdown (Supplementary Fig. S7C).
To assess the effect of Lin28 depletion on in vivo tumorigenicity of DU145 and PC3 cells, cells were first transfected in vitro with control siRNA and the siRNA directed against Lin28B (siLin28B), and then implanted subcutaneously in nude mice. Growth of tumors formed by Lin28B-targeted cells was delayed compared with control cells, leading to a significant difference in tumor size and weight (Fig. 6E). As seen in the ESE3KD-PrEC xenografts, the levels of total Lin28 protein and the proliferation marker Ki67 were significantly reduced in DU145 and PC3 xenografts from Lin28B-targeted cells examined at the end of the in vivo experiment (Fig. 6F). Furthermore, expression of Lin28B and other CSC genes was decreased whereas levels of let-7b increased in Lin28B-targeted tumor xenografts (Fig. 6G). These results indicated that targeting Lin28B affected the CSC-like component and antagonized the development of aggressive prostate tumors.
To further evaluate the potential of Lin28-targeting therapy, we assessed the ability of a siRNA directed to Lin28B to inhibit growth of established tumor xenografts upon systemic delivery. This previously characterized siRNA (siLin28B-2) was selected for its high knockdown efficiency (40). Transfection of siLin28B-2 reduced Lin28B mRNA and concomitantly increased let-7b expression in DU145 cells (Supplementary Fig. S8A and S8B). It also inhibited PS-forming ability of DU145 cells in vitro (Supplementary Fig. S8C). For in vivo systemic delivery siLin28B-2 and the control siRNA were complexed with in vivo jetPEI and administered by intraperitoneal injections at the dose of 2 mg/kg/d. Mice with established DU145 tumor xenografts were treated with siLin28B-2 or control siRNA three times per week for 3 weeks. Treatment with siLin28B-2 reduced tumor growth compared with treatment with the control siRNA (Fig. 7A). In addition, tumor size assessed by in vivo imaging and tumor weight at the end of the treatment were significantly reduced compared with control mice (Fig. 7A). Tumors from siLin28B-2–treated mice showed decreased Lin28 protein and Ki67 staining, indicative of repression of Lin28, and reduced fraction of proliferating tumor cells (Fig. 7B). Furthermore, the expression of several CSC marker genes was significantly reduced in siLin28B-2–treated xenografts along with reduced Lin28B mRNA and upregulation of let-7b (Fig. 7C). Consistent with an effect on the CSCs component, cells extracted from siLin28B-2–treated xenografts exhibited reduced ex vivo PS-forming ability compared with cells from xenografts treated with the control siRNA (Fig. 7D). Furthermore, when tumor cells derived from siLin28B-2–treated xenografts were re-implanted in mice they retained lower in vivo tumor-initiating and ex vivo self-renewal capability, in line with the persistent change in the cell phenotype and loss of CSC properties induced by restoring the normal function of the Lin28/let-7 axis (Fig. 7E). Thus, targeting Lin28B resulted in significant contraction of the CSC compartment impairing tumor initiation, tumor growth, and self-renewal capabilities of prostate cancer cells. Furthermore, depletion of Lin28B was sufficient to affect tumor growth in vivo, in line with the notion that Lin28A and Lin28B have non-redundant functions and cooperate to sustain the tumorigenic and CSC phenotype. Collectively, these data indicate that targeting Lin28A/B function is a valid strategy to antagonize tumor growth and target the CSC compartment in aggressive prostate cancers.
This study provides novel insights on the roles and the regulation of the Lin28/let-7 axis in prostate cancer. We report that Lin28A and Lin28B are highly expressed in prostate cancer and that they sustain the expansion and maintenance of prostate CSCs in cell cultures and tumor xenografts. We found that the epithelial-specific ETS factor ESE3/EHF directly controls transcription of Lin28A, Lin28B and let-7 pri-miRNAs, such as pri–let-7b, in normal prostate epithelial cells. Furthermore, targeting Lin28A and Lin28B had a significant impact on the CSC compartment impairing tumor initiation, tumor growth, and self-renewal capability.
CSCs are a critical source of intratumor heterogeneity within the primary tumor mass and likely the main factors responsible for tumor progression, metastasis, and treatment failures (3). Understanding the molecular mechanisms underlying the induction and expansion of CSCs will help to define new approaches to selectively target these cells and thereby improve cancer treatments (43). Lin28A and Lin28B, which are highly expressed during normal embryogenesis, are frequently upregulated in human cancers and likely play important roles in CSC expansion in many tumors (26, 28, 32). These RNA-binding proteins negatively regulate the processing and maturation of let-7 miRNAs (20, 28, 29, 44, 45). Lin28 proteins bind to let-7 precursors and prevent their processing by DICER and DROSHA. Lin28A has been shown to recruit Zcchc11/TUT4 to let-7 precursors and block processing by DICER in the cytoplasm, whereas Lin28B binds pri–let-7 miRNAs and block processing in the nucleus (22, 28, 44–46). More recent data, however, indicate that Lin28A accumulate also in the nucleus upon methylation to block pri–let-7 biogenesis (47). Intranuclear localization of both Lin28A and Lin28B appears to enhance their oncogenic effects, perhaps providing a more complete block of let-7 miRNA processing. This is in line with our finding of increased intranuclear localization of Lin28 proteins in more aggressive prostate cancer xenografts models and human metastatic tumors. The Lin28 and let-7 interaction is also an example of the complex interplay between miRNAs and their regulators. The existence of a negative feedback loop between let-7 miRNA and Lin28 proteins is well established, whereby increased levels of let-7 miRNAs prevent accumulation of Lin28 proteins and promote further miRNA maturation.
Lin28 proteins function as oncogenes in a variety of human cancers (26). However, the factors regulating the expression of Lin28A and Lin28B in normal and cancer cells are largely unknown (19). Few transcription factors have been identified that regulate the expression of Lin28A or Lin28B (19). We reported previously that the transcription factor ESE3/EHF maintains the differentiation state and restrains transformation and self-renewal in normal prostate epithelial cells (17). Here, we show that ESE3/EHF exerts these functions in part by controlling the level and activity of distinct components of the Lin28/let-7 axis. ESE3/EHF represses transcription of both Lin28A and Lin28B in normal prostate epithelial cells by binding to their promoters and inducing a repressive chromatin state. Thus, ESE3/EHF is the first transcription factor known to repress both Lin28A and Lin28B genes. Coordinate repression of Lin28A and Lin28B by ESE3/EHF allows maturation of let-7 miRNAs to proceed in normal cells. In addition to prevent the posttranscriptional block by Lin28, we demonstrate that ESE3/EHF also controls transcription of let-7 miRNAs. ESE3/EHF binds to the promoter of pri–let-7b and activates its transcription, thus increasing production of mature let-7b miRNA. The promoters of other let-7 miRNAs contain similar EBS, suggesting that ESE3/EHF might positively regulate the transcription of many members of this miRNA family. Notably, let-7 miRNAs are known to downregulate posttranscriptionally the level of Lin28 protein, thus reinforcing the repression of Lin28 transcription by ESE3/EHF. Therefore, in cells depleted of ESE3/EHF both transcriptional and posttranscriptional controls of the Lin28/let-7 axis are altered shifting the balance toward upregulation of Lin28A/Lin28B and diminished production of mature let-7 miRNAs. We show that this is associated with the acquisition of the transformed and stem-like state in normal prostate epithelial cells. Consistent with this, we found that silencing Lin28A and Lin28B or addition of let-7 precursors were sufficient to reverse the stem-like and tumor-initiating phenotype induced by loss of ESE3/EHF. Notably, similar effects were also seen in the prostate cancer cell lines DU145 and PC3, which have low levels of ESE3/EHF and mimic the behavior of aggressive, castration-resistance prostate cancer.
Lin28A and Lin28B function on distinct steps of the processing pathway of pre-miRNAs to mature let-7 miRNAs (28). However, it is unclear whether the two proteins exert complementary or redundant functions with respect to let-7 processing and tumorigenesis (26, 28, 29). In some tumor types prevalent upregulation of either Lin28A or Lin28B has been reported (26, 28, 29). In prostate cancer we found a positive correlation between expression of Lin28A and Lin28B across human tumors, while both were inversely correlated to ESE3/EHF. Thus, consistent with the cell line data, ESE3/EHF apparently set the level of expression of both Lin28A and Lin28B in human prostate cancers and loss of ESE3/EHF was associated with upregulation of both. The frequent co-expression of Lin28A and Lin28B indicates that, although their functions might be partially redundant, both are capable and apparently needed for tumorigenic transformation in prostate cancer. Notably, expression of both Lin28A and Lin28B increased in prostate epithelial cells upon knockdown of ESE3/EHF. In addition, both Lin28A and Lin28B were further elevated in the CSC subpopulation derived from ESE3KD epithelial cells and prostate cancer cell lines compared with the bulk tumor cell population. Thus, loss of ESE3/EHF leads to upregulation of Lin28A and Lin28B, which is more pronounced in CSC-enriched subpopulation and has a direct impact on the growth and self-renewal of CSCs. In this context, therefore, Lin28A and Lin28B appear to have complementary, non-redundant functions and to cooperate to induce cell transformation, tumorigenic, and stem-like phenotype. More in depth studies would allow to dissect the functions of Lin28A and Lin28B in different cell context and clarifying these issues. This would be warranted also in light of the novel roles of Lin28 proteins that might be independent of let-7 miRNA regulation (35, 47, 48). On the other hand, our data show that targeting either Lin28A or Lin28B is sufficient to reverse these malignant phenotypes. The activation of negative feed-back loop between Lin28 and let-7 miRNAs might explain the effects of the knockdown of individual Lin28 on the entire pathway as well as the long-term effects of the transient downregulation of Lin28 proteins. As described in other models (25, 27), increased levels of let-7 miRNAs would lead to posttranscriptional repression of Lin28 protein, thus extending indefinitely the effects of the initial transient depletion by siRNAs. This seems the case in the in vivo experiments that showed significant depletion of Lin28 protein in the tumor xenografts after knockdown of either Lin28A or Lin28B.
Our data point to the Lin28/let-7 axis as a relevant target for developing therapeutics directed to the CSC compartment. Current therapies target and kill preferentially proliferating, partially differentiated tumor cells that constitute the bulk of the tumor, while sparing the rare CSC population (43). Considerable evidence demonstrates that surviving CSCs may cause disease recurrence and drug resistance in prostate cancer (5). Targeting Lin28A and Lin28B might directly antagonize the expansion of the CSC compartment, thus preventing survival of CSCs, generation of therapy-resistant clones, and disease recurrence. Consistently with this notion, Lin28A and Lin28B ablation in ESE3/EHF-KD cells and cancer cell lines resulted in decreased tumorigenic and stem-like properties in vitro. To further test this hypothesis, we knocked down Lin28B in vitro in DU145 cells and then engrafted the cells in mice. Alternatively, we performed systemic treatment of tumor bearing mice with a Lin28B-targeting siRNA by repeated intraperitoneal injections. Both approaches resulted in significant reduction of tumor initiation and growth in mice. Moreover, tumor cells isolated from Lin28B-targeted xenografts had impaired ability to generate PS ex vivo and, notably, failed to form tumors when re-engrafted in mice. These results were consistent with a persistent change of the tumor cell phenotype induced by transient (in vitro) or prolonged (in vivo) targeting of Lin28. Furthermore, they were detectable over multiple generations and passages of the cells in vitro and in vivo.
These data have important implications for therapeutic applications. Collectively, our results indicate that Lin28A and Lin28B support tumor-initiating properties and their inhibition successfully reverts the tumorigenic and stem-like state and tumor development in vivo. The selectivity of Lin28-targeting therapeutics may rely on the increased expression of Lin28 proteins in the prostate CSCs compared with normal cells and bulk tumor cells as well as their role in maintaining the stem-like state of CSCs. Furthermore, the induction of prolonged effects on the cancer cell phenotype even after transient knockdown of Lin28 proteins suggests that continuous exposure to the inhibitors and persistent target inhibition might not be required to achieve beneficial therapeutic effects. Notably, transient short-term inhibition of Lin28 was sufficient to induce persistent effects on the CSC function. This persistent impairment of CSCs might be associated with the induction of cell senescence and irreversible growth arrest that might occur upon Lin28 knockdown. Thus, in combinations with drugs directed to the bulk tumor cells, Lin28-targeting therapeutics might deplete the CSC compartment and allow more persistent control of the tumors, prolong disease-free survival, and reduce disease recurrence. Furthermore, in addition to chemically improved RNAi effectors more suitable for in vivo applications (40), the highly structured interaction between Lin28 proteins and let-7 precursors provides an interesting area for drug discovery and development of innovative approaches for targeting this pathway (23, 36, 45, 49).
In conclusion, our study identifies a relevant upstream mechanism controlling the Lin28/let-7 axis in normal prostate epithelial cells and prostate cancer. The dual control exerted by ESE3/EHF on transcription of Lin28A/Lin28B and let-7 precursors is fundamental to sustain production of let-7 miRNAs and maintain the differentiation state in normal epithelial cells. On the other hand, deregulation of this mechanism in tumors by loss of ESE3/EHF promotes tumorigenesis and affects prominently the CSC compartment, thus pointing to specific strategies to target the CSCs in prostate tumors and prevent disease progression and recurrence.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: D. Albino, G. Civenni, J. Hall, C.V. Catapano, G.M. Carbone
Development of methodology: D. Albino, G. Civenni, C. Dallavalle, M. Roos, L. Curti, C.V. Catapano, G.M. Carbone
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D. Albino, H. Jahns, S. Pinton, G. D'Ambrosio, F. Sessa, G.M. Carbone
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D. Albino, H. Jahns, S. Rossi, C.V. Catapano, G.M. Carbone
Writing, review, and/or revision of the manuscript: D. Albino, M. Roos, J. Hall, C.V. Catapano, G.M. Carbone
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Roos, H. Jahns, S. Pinton, G. D'Ambrosio, F. Sessa, G.M. Carbone
Study supervision: J. Hall, C.V. Catapano, G.M. Carbone
The authors thank Enrica Mira Catò for her assistance with in vivo experiments. The authors declare no competing financial interests.
This work was supported by grants from the Swiss Cancer League (KLS-2648-08-2010, KLS-3243-08-2010, and KLS 3243-08-2013), Swiss National Science Foundation (SNF 310030-146560-2013), ETH Zurich (ETH-01 11-2 and ETH-14 09-3), Swiss Bridge Award, Ticino Foundation for Cancer Research, Virginia Boeger Foundation, and Fidinam Foundation.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.