Cancer stem-like cells (CSC) drive cancer progression and recurrence. Self-renewal expansion of CSC is achieved through symmetric cell division, yet how external stimuli affect intracellular regulatory programs of CSC division modes and stemness remains obscure. Here, we report that the hTERThigh prostate cancer cells exhibit CSC properties, including a stem cell–associated gene expression signature, long-term tumor-propagating capacity and epithelial-to-mesenchymal transition. In promoting the self-renewal symmetric division of hTERThigh prostate cancer cells, WNT3a dramatically decreased the ratio of hTERThigh prostate cancer cells undergoing asymmetric division. Increased WNT/β-catenin signal activation was also detected in hTERThigh prostate cancer cells. hTERT-mediated CSC properties were at least partially dependent on β-catenin. These findings provide novel cellular and molecular mechanisms for the self-renewal of CSC orchestrated by tumor microenvironmental stimuli and intracellular signals. Cancer Res; 77(9); 2534–47. ©2017 AACR.
Although prostate cancer can be surgically excised and effectively treated by androgen deprivation therapy, radiotherapy, or chemotherapy, management of advanced prostate cancer or castration-resistant prostate cancer (CRPC) is still a major obstacle (1). Cancer stem cells (CSC), which are able to propagate new tumor mass and are resistant to conventional anticancer therapies, are considered as the driving force for cancer progression and recurrence (2). Therefore, identification of CSCs and understanding their cellular and molecular regulation mechanisms will be essential for the development of CSC-targeted therapy.
Human telomerase reverse transcriptase (hTERT), an RNA-dependent DNA polymerase, synthesizes the telomeric DNA and adds them to the linear chromosome ends to maintain telomere length (3, 4). In addition to its canonical function in telomere length stabilization, hTERT has also been reported to be actively involved in cell signaling transductions governing cell proliferation, apoptosis, migration, etc (5, 6). hTERT is absent in most human somatic cells due to its transcriptional repression after early stage of embryogenesis (7). In contrast, telomerase activity is relatively high and hTERT expression is detectable in tissue stem or progenitor cells (8). Accumulating evidence suggests that hTERT is upregulated in many human cancers, including prostate cancer, and positively correlated with tumor aggressiveness (9). Moreover, hTERT expression displayed intratumor heterogeneity. Whether hTERT is more preferentially expressed in CSCs and what is the functional contribution of hTERT-expressing tumor cells and hTERT-negative cells to the progression of prostate cancer remain unclear.
Like normal stem cells, CSCs coordinate self-renewal and differentiation to maintain the stem cell pool and, meanwhile, are able to generate new tumor cells (10, 11). Symmetric (SCD) and asymmetric cell divisions (ACD) are key features of stem cells to balance self-renewal and differentiation (12). Although the division modes of stem cells have been extensively studied in development, whether SCD and ACD are present in CSCs and essential in CSC functions remain elusive. If so, how are the external signals and internal molecular events orchestrated to control the division modes of CSCs?
WNT ligands specifically bind to Frizzled and LRP5/6 receptors and activate downstream signaling molecules that lead to stabilization and accumulation of the β-catenin in the nucleus (13). Alterations in expression and localization of β-catenin are common features of advanced prostate cancer (14). Many WNT ligands, including WNT3a, WNT5a, and WNT16b, have been reported to promote epithelial-to-mesenchymal transition (EMT) and CSC activities in various types of cancers (15–17). Interestingly, a recent study demonstrated that localized WNT3a can induce oriented ACD of embryonic stem cells (ESC) and direct distinct cell fates of the daughter cells (18). Whether WNT/β-catenin signaling pathway is involved in the functions of prostate cancer stem cells (PCSC) and whether it operates on the division mode regulation of PCSCs await further investigations.
Here, we employ an hTERT promoter–driven GFP reporter system to separate the hTERT−/low and hTERThigh prostate cancer subpopulation cells and investigate the functional, cellular, and molecular heterogeneities of these two subpopulations. We find that the two subpopulations exhibit different cell division modes and the latter exhibits CSC features, including self-renewal SCDs that are controlled by WNT/β-catenin signaling.
Materials and Methods
Cell lines and patient samples
All cell lines were commercially obtained from the ATCC in 2014. The latest authentication of these cell lines was performed by Shanghai Biowing Applied Biotechnology Co. Ltd. (http://www.biowing.com.cn) using the short tandem repeat genetic analysis in January 2016. LNCaP cells were maintained in RPMI1640 medium, while PC3 cells were cultured in DMEM supplemented with 10% FBS, 100 U/mL penicillin and 100 μg/mL streptomycin in an incubator with 37°C and 5% CO2. Fresh prostate cancer tissues were obtained from prostatectomy surgeries at the Department of Urology, Ren Ji Hospital (Shanghai, China). All samples were collected, stored, and used with informed consent of the patients. The protocol for collection and usage of clinical tissues was approved by the Ren Ji Hospital Ethics Committee.
Plasmids and viruses
The core sequence of hTERT promoter from −1 to −3257 bp was cloned into a lentiviral vector (#17408 Addgene). The cDNA of hTERT with an N-terminal GFP-tag was cloned into the lentiviral vector pLenti-III-HA. shRNA against β-catenin and hTERT and scramble shRNA were constructed into the lentiviral vector pLVTH, respectively. The retroviral vector of Numb-dsRed was a generous gift from Dr. Dean G. Tang from the Department of Pharmacology and Therapeutics of Roswell Park Cancer Institute (Buffalo, NY). Prostate cancer cells were infected with lentiviruses or retroviruses, and 48 hours later, 5 μg/mL puromycin was added into the culture medium for 2 weeks to generate stable expressing cell lines.
Telomerase activity assay
A qPCR-based telomeric repeat amplification protocol assay (qTRAP assay) was used in this study to detect telomerase activity. Detailed procedures of telomerase activity assays are available in Supplementary Materials and Methods.
TCF/LEF reporter activity assay
The T-cell factor/lymphoid enhancer factor (TCF/LEF) luciferase reporter (Shanghai Genomeditech Co., Ltd.) was used to transfect prostate cancer cells together with a Renilla-expressing vector. A dual luciferase reporter assay system (Promega) was used to determine the TCF/LEF–driven luciferase activity at 72 hours posttransfection. The luciferase activity was normalized to the Renilla activity.
Immunoprecipitation, immunoblotting, IHC, and immunofluorescence
The experiments of immunoprecipitation, immunoblotting, IHC, and immunofluorescence in this study were all conventionally performed, and the detailed descriptions are provided in Supplementary Materials and Methods. The detailed information of antibodies used in this study is available in Supplementary Table S1.
FACS and Purification of hTERT−/low and hTERThigh prostate cancer cells
LNCaP and PC3 cells stably expressing hTERT-promoter-GFP were dissociated into single-cell suspension using trypsin/EDTA. For hTERThigh cells, only the top approximately 15% GFP-positive (i.e., hTERThigh) cells were purified for the following experiments. For hTERT−/low cells, the bottom approximately 10% GFP-negative (i.e., hTERT−/low) cells were selected. For PCR and immunoblotting assays, experiments were immediately performed after the GFPhigh and GFP−/low prostate cancer cells were freshly sorted. For immunofluorescence and other functional assays, cells needed 4 to 5 hours to attach to the glass coverslips before we performed the experiments. We have carefully observed that both LNCaP and PC3 cells will not divide within 6 hours before they firmly attach to the glass coverslips using time-lapse recording microscopy.
Time-lapse video microscopy
Purified GFPhigh and GFP−/low LNCaP cells were planted on glass bottom dishes and placed on the stage of Olympus Biostation time lapse system (IX83). The microscope was equipped with an incubator that maintained at 37°C and a humidified 5% CO2 flow. Phase-contrast and GFP images were obtained continuously with a 20× objective lens at a 20-minute interval for 5 days. Data analysis was performed using ImageJ software.
Limiting dilution and clonal formation assay
Single-cell suspension (100 μL) containing 1 to 10 cells was dispensed into each well of a 96-well culture dish. Each well was carefully examined under a light microscope to ensure one single cell per well. The clonal formation assay was performed mainly as described previously (19).
Serial sphere formation assay
GFPhigh or GFP−/low cells were cultured in low-attachment 96-well dish (Corning Costar) in FBS-free RPMI1640 medium supplemented with 20 ng/mL hEGF (Sigma-Aldrich) and 10 ng/mL bFGF (Sigma-Aldrich). After 10 days, the number and volume of the single-cell–derived spheres were counted and measured. The morphologies of the spheres were photographed using an optical microscope. When the diameter of the spheres reached 500 μm, they were digested with trypsin/EDTA and resuspended to generate single-cell suspension. Cells were sorted into GFPhigh and GFP−/low populations and seeded to generate spheres again.
In vivo xenograft and serial tumor transplantation
PC3 cells (1 × 104) in a total volume of 100 μL were mixed with Matrigel at a 1:1 ratio and subcutaneously injected in the right and/or left flanks of 5- or 6-week-old male BALB/C nude mice (Shanghai Laboratory Animal Center, Shanghai, China). Tumors were collected, photographed, and weighed at 50 days after inoculation. Tumor volume was calculated with the formula: volume = 0.5 × length × width2. For the in vivo limiting dilution assay, nude mice were subcutaneously injected with cell suspensions containing 100, 1,000, or 10,000 GFPhigh or GFP−/low PC3 cells. For serial tumor transplantation assay, the first-generation tumors were minced and digested into single-cell suspensions. GFPhigh or GFP−/low cells were sorted again and implanted into nude mice.
Prostate cancer cells were plated as single cells and allowed to progress through one cell division. According to the cell doubling time, we monitored and photographed the phase-contrast and immunofluorescent images of mitotic cells at an 8-hour interval till the two daughter cells were completely separated from each other.
Determining ACD, SCD, coexpression, and inverse expression in cell division modes
Dividing cells in late telophase were validated by enhanced cytoplasmic cleft under a phase-contrast microscope. Then, the GFP images were obtained under a fluorescent microscope. Using ImageJ software, we determined the GFP fluorescent intensities of the two daughter cells with the formula: GFP intensity = IOD/area. If the fluorescent intensity of GFP in the GFPhigh daughter cell was more than 2-fold higher than that in GFP−/low daughter cell, we defined this cell division mode as ACD. If the GFP intensity difference was less than 1.5-fold, we defined this cell division mode as SCD. For immunofluorescence staining assay, dividing prostate cancer cells were fixed and costained with hTERT/Numb or hTERT/β-catenin or TERT/c-Myc antibodies. Taking Numb/hTERT costaining as an example, we calculated and analyzed the fluorescent intensities of Numb and hTERT within one daughter cell. If the fluorescent intensity difference between the two proteins was more than 2-fold, we defined this expression pattern as inverse expression. If the fluorescent intensity difference was less than 0.5-fold, we defined this expression pattern as coexpression.
Prostate cancer samples display heterogeneity in hTERT expression, and elevated hTERT expression is positively correlated with prostate cancer progression
To determine the alteration of hTERT expression during human prostate cancer progression, we first analyzed multiple microarray datasets in the Oncomine database. As shown in Supplementary Fig. S1A–S1C, hTERT mRNA levels were significantly increased in human prostate cancer tissues as compared with the adjacent normal prostatic tissues. Importantly, hTERT mRNA levels were also significantly higher in metastatic prostate cancer samples than that in their matched localized prostate cancer samples. Moreover, upregulated hTERT was positively correlated with high Gleason score in a total number of 89 cases of prostate cancer samples. We also performed IHC staining of hTERT on the benign prostatic hyperplasia tissue (BPH), the prostatic intraepithelial neoplasia tissue (PIN), and prostate cancer tissues with distinct pathologic differentiation status. We found that hTERT was almost absent in BPH tissues. In contrast, hTERT-positive luminal cells began to present in PIN, and their number steadily increased in the well-differentiated and moderately differentiated prostate cancer samples and peaked in the poorly differentiated prostate cancer samples (Supplementary Fig. S1D). Altogether, these data indicate that metastatic and advanced prostate cancers express more hTERT, and the number of hTERT-positive cells inversely correlates with the differentiation state of prostate cancer samples.
Isolation and characterization of hTERT−/low and hTERThigh prostate cancer cell subpopulations
We next asked the question whether there are functional differences between prostate cancer cells highly expressing hTERT and cells with no or low hTERT expression. To separate hTERThigh from hTERT−/low prostate cancer cells, we employed an hTERT-promoter-GFP lentiviral system, in which the expression of GFP was driven by the hTERT promoter. LNCaP or PC3 cells were infected with the lentiviruses and then selected for stable transfection with puromycin (Fig. 1A). We used FACS to purify the top approximately 15% GFP-positive (GFPhigh) and bottom approximately 10% GFP-negative (GFP−/low) prostate cancer cells (Supplementary Fig. S1E and S1F). These GFPhigh cells exhibited a significantly higher mRNA and protein levels of hTERT and stronger telomerase activity compared with their GFP−/low counterparts (Supplementary Fig. S1G–S1I), suggesting that the lentivirus-based reporter system faithfully reflected endogenous hTERT expression and telomerase activity.
hTERThigh prostate cancer cells preferentially express stem cell–associated genes and exhibit CSC properties
Several studies have found that the key regulators in maintaining the stemness of ESCs, including Oct4, Sox2, and Nanog, as well as their activation targets are frequently overexpressed in CSCs in many cancers (20–22). Besides, a set of well-characterized molecules, including CD44, CD133, integrin α2β1, ALDH1A1, and Bmi1, involved in the regulation of CSC self-renewal, metastasis, and drug resistance, has been demonstrated as PCSC markers (23–25). Using qPCR and immunofluorescence staining, we found that hTERThigh prostate cancer cells preferentially expressed these ESC- and PCSC-associated markers (Fig. 1B–E). Among all these markers we examined, Nanog mRNA was found most in hTERThigh LNCaP cells compared with hTERT−/low counterparts, which is consistent with previous studies indicating that Nanog promotes CSC features in both prostatic and hepatocellular carcinoma (20, 26). Quantitative analysis of the expression level of these stem cell–associated genes in hTERT−/low and hTERThigh cells by FACS is shown in Supplementary Fig. S1J. These observations prompted us to investigate whether hTERThigh subpopulation harbors CSCs.
We next assessed the long-term propagation and self-renewal capabilities of hTERT−/low and hTERThigh cells by serial sphere formation assay, serial tumor formation assay, and limiting dilution tumor formation assay. In the serial sphere formation assay, we found that hTERThigh cells were able to form larger spheres with higher efficiencies compared with hTERT−/low cells (Fig. 2A). The larger sphere size and higher sphere formation efficiency of hTERThigh cells could be consistently detected for at least four generations. Interestingly, spheres derived from single hTERThigh cells showed heterogeneity in GFP expression, while spheres produced from single hTERT−/low cells mostly remained GFP negative, suggesting that GFPhigh cells were able to generate GFP−/low cells. In the serial tumor formation assay, we implanted 1 × 104 hTERT−/low or hTERThigh PC3 cells subcutaneously in nude mice and monitored the final tumor incidence, volume, and endpoint weight. Subsequently, the first-generation tumors were isolated, minced, and then digested to generate single-cell suspension. Cells were FACS sorted on the basis of GFP expression again before implantation into nude mice. hTERThigh cells continued to develop significantly larger tumors with higher incidence for a total of four passages in vivo (Fig. 2B). Moreover, limiting dilution assays revealed significant differences in tumorigenic frequency between hTERThigh and hTERT−/low cells. Remarkably, hTERThigh cells were able to initiate tumor formation with as few as 100 cells, whereas hTERT−/low cells required at least 1,000 cells (Fig. 2C). To verify that the high tumor-propagating capacity of hTERThigh prostate cancer cells was mediated by hTERT itself, we used hTERT-overexpressing and shRNA-mediated hTERT knockdown prostate cancer cells in xenograft tumor formation experiments. Indeed, tumor growth was enhanced significantly by forced expression of hTERT and suppressed by hTERT RNAi in PC3 cells (Supplementary Fig. S2A and S2B). Taken together, these data support that hTERThigh prostate cancer cells exhibit CSC features.
CSCs in solid tumors are frequently associated with elevated invasive and migratory capacities (27). In line with that, we found that hTERThigh cells exhibited enhanced cell motility in the scratch assays (Fig. 2D). Rescue experiments by reconstitution of hTERT in hTERT−/low cells or knockdown of hTERT in hTERThigh cells indicated that the increased motility of hTERThigh cells was dependent on hTERT (Supplementary Fig. S2C). EMT, in which cancer cells lose epithelial features including cell polarity and cell–cell adhesion and acquire mesenchymal characteristics, is closely related to cancer metastasis and CSC activities (28). Activation of the EMT regulators and EMT-related signaling pathways has been shown to endow cancer cells with CSC properties (29–33). We then examined the expression of epithelial marker E-cadherin (E-cad) and mesenchymal markers, including N-cadherin (N-cad) and vimentin (Vim), in the two subpopulations. E-cad was preferentially expressed in hTERT−/low cells, whereas N-cad and Vim were more enriched in hTERThigh cells, suggesting that hTERT−/low cells exhibited an epithelial phenotype, whereas hTERThigh cells displayed mesenchymal-like features (Fig. 2E). Moreover, restoration of hTERT in hTERT−/low LNCaP or PC3 cells significantly suppressed E-cad and upregulated N-cad and Vim. Conversely, shRNA-mediated hTERT knockdown significantly abrogated the mesenchymal phenotype of hTERThigh cells (Supplementary Fig. S2D and S2E). These data collectively indicate that hTERT is critical in promoting EMT and cell mobility.
Single hTERThigh prostate cancer cells are prone to form holoclones in clonal cultures
It has been shown that a single prostate cancer cell was capable to form colonies with distinct morphologies: holoclones with regular round shapes and tightly organized 2-demensional structures containing smaller self-renewing cells, paraclones that are loosely packed clones comprised mostly of larger differentiated cells, and meroclones containing a mixture of self-renewing and differentiated cells (19, 34, 35). We found that single hTERThigh prostate cancer cells were able to form all these three types of clones with the frequencies of 15.6% to 31.0% holoclones, 24.1% to 39.2% meroclones, and 44.9% to 45.2% paraclones from every 100 clones. (Supplementary Fig. S3A and S3B). Importantly, overexpression of hTERT significantly increased the holoclone formation frequency, whereas knockdown of hTERT dramatically dampened the holoclone formation efficiency (Supplementary Fig. S3C–S3F).
Using time-lapse video microscopy, we monitored the detailed clone formation processes of single hTERT−/low and hTERThigh cells for 5 days. We found that most single GFP−/low cell–generated paraclones with typically larger progenies and scattered morphologies remained GFP negative within 5 days (Fig. 3A). In contrast, clones derived from single GFPhigh cells displayed distinct morphologies and cellular composition. In single GFPhigh cell–derived holoclones, most of the progenies were GFP positive (Fig. 3B). However, both GFP-positive and GFP-negative cells were present in GFPhigh cell–derived meroclones (Fig. 3C), whereas most of the cells in GFPhigh cell–derived paraclones became GFP negative at day 3 (Fig. 3D). These results suggest that GFPhigh prostate cancer cells are capable of generating GFP−/low daughter cells. Phase-contrast and GFP images of single hTERThigh cell–derived three types of clones were shown in Fig. 3E. We next performed immunofluorescence staining to determine whether the GFP intensity faithfully reflected endogenous hTERT expression in these clone types. Consistently, the vast majority of cells in holoclones expressed high levels of hTERT. A portion of the meroclone cells was positive for hTERT expression, whereas most cells in paraclones stained negative for hTERT (Fig. 3F, left). In addition, the bulk cells from these three types of clones showed the highest hTERT mRNA, protein, and telomerase activity in holoclone cells, and the lowest in paraclones (Fig. 3F, right). In line with our preceding findings, immunofluorescence staining of single hTERThigh-derived clones revealed that cells stained positive for PCSC markers CD44, CD133, and integrin α2β1 were mostly featured with high levels of hTERT expression in holoclones compared with meroclones (Fig. 3G). Collectively, our data indicate that single hTERThigh prostate cancer cells are prone to generate holoclones that are enriched with CSCs, and overexpression of hTERT significantly promotes the holoclone formation efficiency in bulk prostate cancer cells.
hTERThigh prostate cancer cells display distinct cell division modes as compared with hTERT−/low counterparts
Stem cells undergo ACD to generate one stem cell and one differentiated progeny, whereas they undergo SCD to produce two stem cells to expand the stem cell pool (12, 36). Our clone formation results had shown that single GFPhigh prostate cancer cells were capable of generating GFP−/low progenies, leading to our hypothesis that GFPhigh cells, like normal stem cells, probably undergo ACD to produce GFP−/low cells. To test that, we used the time-lapse microscopic imaging to record the division modes of single GFP−/low and GFPhigh prostate cancer cells. In more than 50 videos containing more than 1,000 single-cell mitotic behaviors, we found that GFP−/low LNCaP cells mostly undergo SCD to produce two GFP-negative daughter cells (Fig. 4A). In contrast, three distinct cell division modes were detected in GFPhigh LNCaP cells: SCD type I (SCD I), in which one GFPhigh cell produced two GFP-positive daughter cells (Fig. 4B); SCD type II (SCD II), in which one GFPhigh cell generated two GFP-negative daughter cells (Fig. 4C); and ACD, in which one GFPhigh cell gave rise to one GFP-positive and one GFP-negative daughter cell (Fig. 4D). Quantitative analysis revealed that SCD I was the most common division mode, which accounted for approximately 65% and 80% in LNCaP or PC3, respectively. SCD II ratio was relatively lower, which took up approximately 25% in LNCaP cells and approximately 15% in PC3 cells. ACD constituted approximately 8% and 5% in dividing LNCaP and PC3 cells, respectively. The fact that hTERThigh prostate cancer cells exhibited both ACD and SCD further supported the notion that hTERThigh cells were CSCs. We also analyzed the GFP status of the two subpopulations on days 0 (freshly sorted), 7, and 14 of culture after FACS sorting. Dotted FACS plots indicated that the hTERThigh prostate cancer cells gradually generated hTERThigh and hTERT−/low cells, while the GFP intensity of hTERT−/low prostate cancer cells remained extremely low even after 14 days of culture (Supplementary Fig. S4A).
Using paired-cell assay, we performed immunofluorescence staining of Numb, a well-characterized cell fate determinant that tends to segregate into the more differentiating daughter cell during mitosis (37), in LNCaP and PC3 cells, respectively. We found that the segregation of endogenous hTERT and Numb were mutually exclusive during cell division. In SCD I, both daughter cells expressed high levels of hTERT but exhibited low expression of Numb. In SCDII, both daughter cells exhibited undetectable hTERT but were stained positively for Numb, whereas in ACD, Numb staining was absent in the hTERThigh daughter cells but could be readily detected in the hTERT−/low daughter cells (Fig. 4E and F, left). Quantitative analysis suggested that the frequency of inverse distribution of endogenous hTERT and Numb was significantly higher than the cosegregation frequency (Fig. 4E and F, right). To further substantiate the asymmetric partition of Numb during ACD of hTERThigh cells, we cotransfected LNCaP cells with hTERT-promoter-GFP and Numb-dsRed vectors and monitored the segregation of GFP and dsRed during mitosis (Supplementary Fig. S4B and S4C). Consistently, we found that exogenous Numb was more prone to distribute into the GFP-negative daughter cells during ACD of hTERThigh cells (Supplementary Fig. S4D), and the three typical cell division modes were faithfully reproducible in LNCaP cells relative to GFP status in paired-cell assay (Supplementary Fig. S4E–S4G).
Nuclear β-catenin cosegregates with hTERT in ACD of prostate cancer cells
To investigate the molecular mechanism involved in hTERT-mediated CSC activity, we examined several signaling pathways that are crucial for functions of both prostate epithelial stem cells and PCSCs, including androgen receptor (AR) signaling, sonic hedgehog (SHH) signaling, Notch signaling, and WNT/β-catenin signaling. Among these pathway genes we examined, the central regulator of WNT signaling β-catenin and its downstream targets, including Axin2, c-Myc, and cyclin D1, were most preferentially expressed in hTERThigh rather than hTERT−/low cells (Fig. 5A and B). To test whether WNT/β-catenin signaling was distinctively activated in hTERThigh cells, we performed immunofluorescence staining of β-catenin in both cell subpopulations. hTERThigh cells exhibited enhanced nuclear-localized β-catenin as compared with hTERT−/low cells (Fig. 5C). Moreover, the TCF/LEF luciferase assay suggested a significantly elevated TCF/LEF activity in hTERThigh cells (Fig. 5D). We further conducted paired-cell assays to determine the segregation pattern of endogenous β-catenin, c-Myc, and hTERT in ACD and SCD. In SCD I, when both of the daughter cells expressed high hTERT expression level, β-catenin and c-Myc were also equally enriched in the nucleus of both daughter cells (Fig. 5E, top). In SCD II, when both of the two daughter cells displayed undetectable hTERT, we observed cell membrane–localized β-catenin and the absence of c-Myc expression (Fig. 5E, middle). In contrast, during ACD, the daughter cell with a high level of hTERT also stained enhanced nuclear β-catenin and c-Myc, whereas the other daughter cell with low hTERT displayed clearly membrane β-catenin and undetectable c-Myc (Fig. 5E, bottom). The frequency of coexpression of hTERT with nuclear β-catenin was significantly higher than inverse expression of the two molecules in LNCaP cells. The frequencies of coexpression and inverse expression of hTERT with c-Myc were also showed in Fig. 5E (right). We next asked whether ACD and SCD ratios of hTERThigh prostate cancer cells were affected by WNT ligands. We then compared the cell division modes of hTERThigh prostate cancer cells cultured in WNT3a-containing medium and control medium. We found that the hTERThigh cells treated with WNT3a exhibited an increased SCD I ratio but decreased ACD and SCD II ratios compared with those cells cultured in the regular medium (Fig. 5F). Other WNT ligands, such as WNT16b, which has been reported to promote therapeutic resistance of prostate cancer cells (17), and the noncanonical WNT5a were also analyzed. We found that WNT16b significantly increased SCD I ratio but decreased the ACD and SCD II ratios (Supplementary Fig. S5A). However, WNT5a did not alter the cell division mode ratios of hTERThigh cells (Supplementary Fig. S5B). Moreover, the effects of WNT inhibitors DKK1 and hR-Spondin were also tested. We found that both inhibitors greatly upregulated the ratio of differentiating SCD II and suppressed the self-renewal SCD I (Fig. 5C and D). The effects of these WNT ligands and inhibitors on WNT pathway were confirmed by immunoblotting assays (Supplementary Fig. S5E). Moreover, expression patterns of WNT3a, WNT16b, and WNT receptors, including LRP5 and Frizzled, in hTERT−/low and hTERThigh cells were also examined (Supplementary Fig. S5F–S5H). We found that these ligands and receptors were all present in both subpopulations, and most of them were enriched in hTERThigh cells. These data collectively indicate that canonical WNT ligands, such as WNT3a and WNT16b, promote the expansion of hTERThigh PCSCs by inhibiting ACD and SCD II while enhancing the self-renewal SCD I.
hTERT upregulates WNT signaling target genes and associates with β-catenin
We demonstrated above that hTERThigh prostate cancer cells displayed elevated WNT signaling and hTERT cosegregated with nuclear-localized β-catenin and c-Myc, indicating a positive correlation between hTERT and WNT signaling. We next studied whether WNT signaling can be regulated by hTERT. Overexpression of hTERT in PC3 cells exhibited increased nuclear accumulation of β-catenin (Fig. 6A). We then examined β-catenin and its downstream targets Axin2, c-Myc, and cyclinD1 by qPCR and immunoblotting. Ectopic expression of hTERT upregulated β-catenin, c-Myc, and cyclin D1 expression at both mRNA and protein levels (Fig. 6B and C). It has been reported that hTERT interacts with NF-κB p65 and acts as a transcriptional cofactor (38). We therefore explored whether hTERT physically associates with β-catenin in prostate cancer cells. Coimmunoprecipitation experiments revealed that GFP-tagged hTERT interacted with Flag-tagged β-catenin in 293T cells (Fig. 6D). More importantly, the endogenous hTERT was also found to associate with β-catenin in prostate cancer cells (Fig. 6E). These data suggest hTERT may act as a cofactor and form a complex with β-catenin to activate WNT downstream targets in prostate cancer cells.
We next investigated the correlation between hTERT and WNT/β-catenin activation in clinical prostate cancer samples. Using IHC staining, we found that prostate cancer tissues with increased expression of hTERT also exhibited stronger nuclear-localized β-catenin, whereas prostate cancer samples with low expression of hTERT more frequently showed membrane-bound β-catenin (Fig. 6F). Immunoblotting of hTERT, active β-catenin, c-Myc, and cyclin D1 further confirmed the significant positive correlation between hTERT and WNT/β-catenin signaling (Supplementary Fig. S6).
β-Catenin is crucial for hTERT-mediated CSC traits
The importance of WNT/β-catenin signaling pathway in ESCs and CSCs has been well documented (13, 14). We then explored whether hTERT-mediated CSC properties are dependent on β-catenin. Therefore, we performed a series of rescue experiments by knockdown of β-catenin in prostate cancer cells transfected with hTERT-expressing lentiviruses. The knockdown efficiency of β-catenin by two independent shRNAs in both LNCaP and PC3 cells was confirmed by qPCR and immunoblotting (Supplementary Fig. S7A and S7B). Downregulation of β-catenin markedly abrogated the positive tumor-promoting effects of hTERT (Fig. 7A). We also found that hTERT-induced high holoclone formation frequency was dramatically abolished by β-catenin RNAi (Fig. 7B). Similarly, the scratch assays showed that knockdown of β-catenin significantly attenuated the elevated cell motility mediated by hTERT (Fig. 7C). Moreover, knockdown of β-catenin greatly reversed hTERT-induced EMT in prostate cancer cells (Fig. 7D). Altogether, these data indicate that hTERT-mediated tumor-promoting effects and CSCs features are, at least partially, dependent on WNT/β-catenin signaling.
In this study, we showed that hTERThigh prostate cancer cells highly express stem cell–associated gene signature, display greater serial sphere formation and tumor propagation capacity, and show EMT features. Asymmetric division, a key feature of stem cells, can only be detected in hTERThigh but not hTERT−/low prostate cancer cells (Supplementary Fig. S8). All these data suggest that hTERThigh prostate cancer population is enriched with CSC cells. Moreover, hTERThigh cells display elevated WNT/β-catenin signaling activity. Nuclear-localized β-catenin and c-Myc are significantly cosegregated with hTERT in both ACD and SCD. hTERT-mediated CSC properties are at least partially dependent on β-catenin. Importantly, we further demonstrate that WNT ligands promote the expansion of hTERThigh CSCs by increasing the self-renewing SCD I ratio, meanwhile suppressing ACD and decreasing the differentiating SCD II of hTERThigh prostate cancer cells. Here, we provide novel cellular and molecular mechanisms for the self-renewal of hTERThigh CSCs orchestrated by external WNT ligands and internal WNT/β-catenin signaling.
We also found that hTERT upregulates the holoclone formation, serial tumor formation capacity, and EMT of prostate cancer cells, suggesting that hTERT can not only serve as a PCSC marker but is also required for PCSC functions. The physical interaction between hTERT and β-catenin and hTERT-mediated WNT signaling activation in LNCaP and PC3 cells were also unveiled in our study. Functional rescue experiments demonstrate that β-catenin is required for the tumor-promoting effects of hTERT in prostate cancer cells.
Importantly, CRPC specimens exhibit higher hTERT expression and stronger telomerase activity (39–41). The accumulation and localization of β-catenin in the nucleus are also frequently observed in CRPC (14). In addition, increasing evidence has shown that WNT signaling plays crucial roles in PCSC functions and CRPC progression (42, 43). The connection of hTERT and WNT/β-catenin signaling in prostate cancer reported in this study may shed light on the molecular mechanism of CRPC development. hTERT can promote the nuclear translocation of β-catenin. The interaction between hTERT and β-catenin may elicit specific target genes to facilitate androgen-independent growth and CRPC progression.
Our data from time-lapse microscopy and paired-cell assays vividly show that hTERThigh PCSCs exhibit multiple cell division modes, which mimics the behaviors of stem or progenitor cells to generate cellular diversity during development. Interestingly, a recent study has shown that immobilized WNT3a on beads is capable of directing ACD of ESCs in vitro (18). They discovered that the WNT proximal daughter cell expressed high levels of nuclear β-catenin and pluripotent genes, whereas the distal daughter cell exhibited differentiation hallmarks. In our study, ACD ratio was relatively low and constituted 5% to 8% in hTERThigh prostate cancer cells. It is likely that a lack of cellular polarity and WNT ligand gradient in in vitro cultural conditions might contribute to the low ACD ratio. Future studies utilizing niche cell coculture or immobilized WNT ligands on beads that mimic the in vivo microenvironment may facilitate the cellular polarity establishment of hTERThigh PCSCs. Nevertheless, we found that in ACD, the hTERThigh daughter cell contains nuclear-localized β-catenin, a high level of c-Myc, and undetectable level of the differentiating marker Numb, whereas the hTERT−/low daughter cell exhibits membrane-bound β-catenin, low level of c-Myc, and high level of Numb. WNT3a and WNT16b are found to significantly enhance the self-renewal SCD I and repress ACD as well as the differentiating SCD II of hTERThigh PCSCs. These findings may provide a novel link between the enrichment of hTERThigh PCSCs and elevated expression of WNT ligands in advanced prostate cancer and CRPC specimens. However, the molecular mechanism by which hTERT protein is asymmetrically or symmetrically distributed during PCSC divisions remains unknown.
Collectively, we uncover a WNT/hTERT/β-catenin axis that tightly regulates the self-renewal and differentiation of hTERThigh PCSCs. Our findings suggest that inhibition of WNT/β-catenin signaling combined with hTERT interference may represent a promising strategy to suppress hTERThigh PCSCs and provide new therapeutic approach for localized prostate cancer and CRPC.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: H.H. Zhu, W.-Q. Gao
Development of methodology: K. Zhang, D. Xu
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L. Li, Y. Fang, H.H. Zhu, W.-Q. Gao
Writing, review, and/or revision of the manuscript: D. Xu, H.H. Zhu, W.-Q. Gao
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y. Guo, X. Wang, H. Zhao, Z. Ji, C. Cheng, L. Li, Y. Fang
Study supervision: W.-Q. Gao
The study is supported by funds from the Chinese Ministry of Science and Technology (2012CB966800 to W.Q. Gao, 2013CB967402 to L. Li, and 2013CB945600 to W.Q. Gao), the National Natural Science Foundation of China (NSFC, 81630073 and 81372189 to W.Q. Gao), the State Key Laboratory of Oncogenes and Related Genes (90-16-03 to H.H. Zhu and 90-13-06 to D.W. Xu), Shanghai Institutions of Higher Learning [The Program for Professor of Special Appointment (Young Eastern Scholar, QD2015002) to H.H. Zhu], Shanghai Rising-Star Program (17QA1402100 to H.H. Zhu), School of Medicine, Shanghai Jiao Tong University (Excellent Youth Scholar Initiation Grant 16XJ11003 to H.H. Zhu), Ren Ji Hospital (Seed Project RJZZ14-010 to H.H. Zhu), Science and Technology Commission of Shanghai Municipality (16JC1405700 to W.Q. Gao and 16140902100 to L. Li), KC Wong Foundation (W.Q. Gao), Special Research Foundation of State Key Laboratory of Medical Genomics (W.Q. Gao), and Shanghai Eastern Hospital (Pudong) Stem Cell Research Base Fund (W.Q. Gao).
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.