Numb^−/low Enriches a Castration-Resistant Prostate Cancer Cell Subpopulation Associated with Enhanced Notch and Hedgehog Signaling

The most recent statistics suggest that aging and a Westernized lifestyle are associated with an increasing trend of prostate cancer incidence and mortality rates worldwide (1, 2). Androgen deprivation therapy in conjunction with surgery or radiation is the mainstream treatment strategy for advanced prostate cancers (3). However, the emergence of castration-resistant prostate cancer (CRPC) remains the predominant cause of death in prostate cancer patients. Unfortunately, our understanding of the molecular etiology of castration resistance remains limited, leading to limited intervention approaches for the treatment of this deadly disease. In addition, prostate cancer cells display great intratumor heterogeneity. Their response to androgen deprivation therapy is also radically different from each other (4–6). The targeted eradication of cell subpopulations with greater castration resistance may serve as an optimal strategy for the treatment of CRPC. The identification of those cell subpopulations and the understanding of their molecular characteristics will be a first step achieving this goal.

Numb was originally identified and investigated in Drosophila. The Numb protein has been found to be distributed asymmetrically during divisions of sensory organ precursors (SOP) in Drosophila embryos, resulting in different cell fates of daughter cells (7, 8). Upon the first division of SOP, Numb is selectively distributed into the anterior daughter cell, which differentiates into neurons or glia, whereas the posterior daughter cell without inheritance of Numb differentiates into outer support cells (9–11). The critical importance of Numb during progenitor cell differentiation and cell fate determination in hematopoiesis, neurogenesis, cardiac morphogenesis, and myogenesis in vertebrates was demonstrated subsequently (12–15). Recently, a growing body of evidence has suggested that Numb may act as a tumor suppressor in various tumor types, including hepatocellular carcinoma, malignant pleural mesothelioma, and breast cancer (16–18). The downregulation of Numb is associated with a poor prognosis in hepatocellular carcinoma (19), salivary gland carcinoma (20), and esophageal squamous cell carcinoma (21). However, the expression of Numb is greatly heterogeneous among different cells within a tumor mass (22–24). The molecular mechanism underlying the downregulation of Numb and the functional difference between Numb highly expressed and underexpressed cancer cells is not well elucidated.

In this study, we aim to elucidate the role of Numb in CRPC development and decipher the transcriptional regulation of the Numb gene in prostate cancer. On the basis a promoter analysis of the Numb gene, we designed a fluorescent protein reporter lentiviral system to distinguish between prostate cancer cells with a high or low Numb expression. We find that Numb^−/low enriches a small subpopulation of smaller and quiescent cells that preferentially express Notch and Hedgehog downstream and stem-cell–associated genes and are associated with greater resistance to androgen deprivation therapy.

Twenty cases of fresh human prostate cancer specimens and paired normal tissues were obtained during surgery at the Department of Urology of Renji Hospital (Shanghai, China). Detailed information regarding the human prostate cancer patients is shown in Supplementary Table S1. Signed informed consent was collected from all participating patients.

Human prostate cancer cell lines, including LNCaP, C4-2B, PC3, and DU145, and 293T cells were purchased from the Institute of Cell Research of the Chinese Academy of Sciences or the ATCC. The cells were recently authenticated via short Tandem Repeat (STR) profiling by the Shanghai Biowing Applied Biotechnology Company.

The Penco Numb-DsRed-DR retroviral reporter was generously donated by Dr. Dean G. Tang (25). The doxycycline (dox)-inducible lentiviral plasmid Notch-ICD was a gift from Rudolf Jaenisch (Addgene plasmid # 61540; ref. 26). We used shRNA to knockdown the expression of Numb and PTCH1. The shRNAs were cloned into lentiviral GV298 vector with IRES-Cherry and puromycin-selective markers (Shanghai GENECHEM Co., Ltd). Detailed information regarding the sequences of shNumb and shPTCH1 is provided in Supplementary Table S4. The fluorescent Numb promoter lentiviral reporter was generated by replacing the pCMV gene in the PLVX-AcGFP1-N1 vector (Clontech, Catalog Nos.632154) with the Numb-2K promoter sequence (from-2913 to +84). The truncated Numb promoter (1K-4K) sequences were amplified and cloned into a pGL3 basic luciferase vector (Promega #E1751) using standard recombinant DNA methods. The RBP-Jκ, GLI and TCF/LEF1 luciferase reporter lentiviruses were purchased from Shanghai Genomeditech Co., Ltd. Detailed information regarding the plasmid components and response elements of the RBPJκ, GLI and TCF/LEF1 luciferase reporter is provided in Supplementary Tables S5 and S6.

Numb promoter (1K-4K) luciferase reporters were co-transfected into cells with an internal control plasmid expressing Renilla luciferase respectively. The firefly luciferase and Renilla luciferase activities were measured using a luminometer with the Dual-luciferase Reporter Assay System (Promega, E1910). The firefly luciferase activity was quantified and normalized to the Renilla activity.

The cells were harvested and suspended in 50 μL serum-free medium and mixed with 50 μL Matrigel (BD Biosciences). Then, the mixture was injected subcutaneously into the flanks of 4-week-old BALB/c nude mice (SLAC). The tumor growth was monitored and recorded weekly after the inoculation. The volume of the tumor was calculated using the following formula: 0.5 × tumor length × tumor width². The mice were sacrificed when the tumors had a 1.0-cm diameter and the tumors were weighed and imaged.

The statistical analysis of all data was carried out using the Prism GraphPad software (LaJolla) via a Student's t test or ANOVA test. Statistical significance was determined two-sided with P values less than 0.05.

More information is provided in the Supplementary Methods and Materials.

To investigate the expression level and clinical significance of Numb in prostate cancers, we first analyzed microarray datasets obtained from the Oncomine database to determine the Numb mRNA expression alterations at different stages of prostate tumorigenesis, including benign prostate hyperplasia (BPH), prostatic intraepithelial neoplasia (PIN), and prostate carcinoma. As shown in Fig. 1A–D, the Numb mRNA levels were markedly reduced in PIN and prostate cancer samples compared to those in the BPH tissues. Importantly, Numb was further downregulated in samples from patients with a higher Gleason Score, metastatic or recurrent prostate cancers. To verify the results of the Oncomine database analysis, we conducted immunoblotting and immunofluorescence experiments using human prostate cancer specimens and paired adjacent normal tissues that were collected during radical prostatectomy. As demonstrated in Fig. 1E, the protein level of Numb was downregulated in 12 of the 20 cases of human prostate cancer specimens compared with that in their matched normal tissues. Among remaining samples, 4 cases of patient samples showed Numb protein levels that were equivalent to their corresponding normal tissues, whereas an increased Numb expression was detected in only 4 cases (Fig. 1E). Next, we performed immunofluorescent staining of Numb in primary normal and prostate cancer human tissues. As shown in Supplementary Fig. S1, the cells in the normal prostate glands generally expressed Numb at a high level, whereas the prostate cancer cells displayed a relatively lower Numb antibody staining intensity, although the Numb expression level significantly varied among the different prostate cancer cells. In addition, using quantitative RT-PCR, we detected a significant decrease in the Numb mRNA levels in murine prostate tumor tissues from probascin-cre: Pten^fl/fl mice compared to those in wild-type murine prostate tissues (Fig. 1F). We then examined the difference in Numb expression between androgen-sensitive and androgen-independent prostate cancer cell lines. As shown in Fig. 1G, the androgen-independent prostate cancer cell lines (C4-2B, PC3, and DU145) expressed a significantly lower amount of Numb mRNA than the androgen-sensitive prostate cancer cell line (LNCaP). Altogether, these data indicated a downregulation of Numb in prostate tumors and a negative correlation between Numb expression and prostate cancer progression.

To explore the functional role of Numb in prostate cancer, we performed gain- and loss-of-function studies by infecting the prostate cancer cell line C4-2B with a Numb-overexpressing retrovirus and the LNCaP cell line with Numb shRNA–expressing lentiviruses (shNumb). According to the immunoblotting experiment results, Numb was successfully overexpressed in the C4-2B cells that were transfected with the Numb-DsRed retroviruses and efficiently knocked down in the LNCaP cells that were infected with the two independent shRNA Sequences 1 and 2 (Fig. 2A). Because Numb has been reported to play a negative role in distinct oncogenic signaling pathways in different cancer types (27–29), we next examined the impact of Numb on potential downstream signaling in prostate cancer. Using luciferase-based Notch (RBP-JK), Hedgehog (GLI), and Wnt (TCF/LEF1) signaling reporters, we were able to detect the activities of each signaling pathways in the prostate cancer cells. As shown in Fig. 2B, the Notch and Hedgehog signaling activities were significantly reduced in the Numb-overexpressing C4-2B cells compared with those in the control cells but were notably increased by the Numb knockdown in the LNCaP cells. In contrast, Wnt signaling was not affected by either Numb overexpression or Numb knockdown according to the TCF/LEF1 luciferase reporter assays (Supplementary Fig. S2A). Consistently with the luciferase reporter experimental data, the overexpression of Numb in the C4-2B cells led to a significant downregulation of the mRNA and protein levels of key components in the Notch and Hedgehog signaling pathways (Notch1, Notch2, Hes1, Hey1, GLI1, and GLI2). The expression of PTCH1, which is a membrane receptor and negative regulator of Hedgehog signaling (30), was upregulated in the Numb-overexpressing C4-2B cells compared with that in the control cells. In contrast, opposite effects were observed when Numb was knocked down in the LNCaP cells (Fig. 2A, C and D). In addition, no significant difference in the protein levels of total β-catenin or phosphorylated β-catenin, which is the major downstream effector of Wnt signaling, was detected in the Numb overexpression or knockdown prostate cancer cells (Supplementary Fig. S2B). Altogether, these data suggested that the Notch and Hedgehog but not Wnt pathways were negatively regulated by Numb in the prostate cancer cells. Consistently, cBioPortal database analysis revealed a negative correlation between Numb and the Notch signaling downstream target gene Hes1 or Hedgehog signaling–specific target gene GLI1 (Supplementary Fig. S3).

Castration resistance in prostate cancer is currently the greatest cause of treatment failure and death in prostate cancer patients. To explore the role of Numb in the development of CRPC, we conducted apoptosis, sphere-forming and cell proliferation analyses in prostate cancer cells using gain or loss of Numb expression under androgen deprivation. As shown in Fig. 3A, the addition of 10% charcoal dextran–stripped serum (CDSS) and 100 μmol/L flutamide, which is an androgen receptor antagonist, to the culture medium resulted in marked cell death in the LNCaP cells, whereas the knockdown of Numb significantly reduced the number of apoptotic LNCaP cells in response to the androgen deprivation. Furthermore, the Numb knockdown in the LNCaP cells resulted in a higher sphere forming capacity than that observed in the control group after androgen was deprived from the culture medium (Fig. 3B).

To further determine whether the activation of Notch and Hedgehog signaling contributes to the inhibitory role of Numb in prostate tumorigenesis and CRPC development, we conducted rescue experiments by restoring of Notch and Hedgehog signaling activities via the overexpression of NICD (Notch intracellular domain) and knockdown of PTCH1 in the Numb-overexpressing C4-2B cells (Fig. 3C). Sphere-forming assay, apoptosis analysis and in vivo xenograft tumor-forming experiments were subsequently carried out. As shown in Fig. 3D, the overexpression of Numb suppressed the sphere-forming capacity of C4-2B cells, whereas the activation of the Notch and Hedgehog signaling pathways abrogated the inhibitory effect of Numb. Significant cell apoptosis was observed in response to the androgen deprivation in the Numb-overexpressing prostate cancer cells, which was attenuated by the upregulation of Notch and/or Hedgehog signaling components (Fig. 3E and F). Moreover, we found that the overexpression of Numb impeded the tumor growth and lowered the tumor incidence in the C4-2B cells (Fig. 3G and H). The restoration of Notch and Hedgehog signaling activities compromised the suppressive role of Numb in the in vivo growth of the prostate cancer cells. Collectively, these results suggested that Numb exerted an inhibitory effect on prostate cancer cells growth and CRPC development through the suppression of the Notch and Hedgehog pathways.

Previous studies have focused on the sub-cellular protein localization and partition of Numb as a cell fate determinant during symmetric or asymmetric cell division (31, 32). Here, we attempted to address a much less investigated but important question regarding the mechanism by which the expression of the Numb gene was regulated at the transcriptional level. Therefore, we first constructed truncated forms of the putative cis-regulatory element of Numb by cloning the human genomic DNA sequence from the −888 bp, −1930 bp, −2913 bp, or −3861 bp to the +84 bp of its transcription start site (TSS) to drive luciferase reporter in pGL3 vectors, which were named Numb-1K, Numb-2K, Numb-3K, and Numb-4K, respectively (Supplementary Fig. S4A). We then transfected the above-mentioned plasmids into 293T, LNCaP and C4-2B cells and analyzed the respective luciferase activities. As shown in Supplementary Fig. S4B, the Numb-2K–transfected cells exhibited the highest luciferase activity of all three tested cell lines, whereas the luciferase reporter activity was remarkably diminished in the Numb-4K transfectants. The Numb-2K fragment was therefore used as the shortest proximal promoter region of Numb in the following experiments.

We then tested whether the activities of the Numb promoter were altered in the Numb-overexpressing or knockdown cells. As shown in Fig. 4A, the Numb-promoter–driven luciferase reporter was markedly inhibited by the Numb knockdown but was notably elevated by the forced expression of Numb, suggesting a positive auto-regulatory of Numb transcription by itself. Considering that the repression of the Notch and Hedgehog signaling pathways contributes to the function of Numb in prostate cancers (Fig. 3C–F), we next examined whether either of these pathways might be required for the auto-regulation of Numb. As shown in Fig. 4B, the inhibition of Hedgehog signaling activity with a treatment of 1 μmol/L cyclopamine, which is a Hedgehog pathway inhibitor (33–35), led to an increased transcription of Numb mRNA in both the LNCaP and C4-2B cells, whereas the Hedgehog signaling activation with a treatment of 0.2 μg/mL Shh or via PTCH1 knockdown resulted in a reduced mRNA level of Numb. Consistently, the Hedgehog signaling antagonist treatment triggered the upregulation of Numb promoter driven luciferase activities in both the LNCaP and C4-2B cells. In contrast, activation of Hedgehog signaling reduced the Numb promoter luciferase activities in both tested prostate cancer cell lines (Fig. 4C). However, we observed that the alteration in the Notch signaling activities induced by the overexpression of NICD or the treatment with 1 μmol/L antagonist DAPT exerted no evident impact on either the Numb promoter activities or the Numb mRNA level in the LNCaP cells and 293T cells (Supplementary Fig. S5). These data suggested that the transcription of the Numb gene was autoregulated by a Numb/Hedgehog signaling axis.

We demonstrated that the transcription of the Numb gene is positively auto-regulated by the Numb protein. Therefore, we hypothesized that the endogenous transcriptional level of Numb mRNA might serve as a faithful indicator of its protein amount. As mentioned above, the prostate cancer cells expressed different levels of the Numb protein (Supplementary Fig. S1). To elucidate the phenotypical and functional heterogeneity between prostate cells with a high and low Numb expression, we constructed a lentiviral reporter system in which the AcGFP reporter, which is a green fluorescent protein derived from Aequorea coerulescens, was driven by the Numb promoter and the puromycin resistance was driven by an independent PGK promoter (Fig. 5A). We then transfected the LNCaP and C4-2B cells with the Numb promoter reporter lentiviruses and established stable Numb promoter reporting cell lines with persistent puromycin selection. To determine whether the AcGFP fluorescence intensity reflected the endogenous expression of Numb, we purified the top 10% GFP^high and bottom 10% GFP^−/low cells, which were also depicted as Numb^high and Numb^−/low prostate cancer cells in subsequent experiments, using fluorescence-activated cell sorting (FACS) and conducted qRT-PCR, immunoblotting and immunofluorescence on the respective subpopulation (Fig. 5B). As shown in Fig. 5C–E, the expression level of AcGFP correlated well with the amount of Numb protein. Therefore, the fluorescence intensity of AcGFP faithfully represented the endogenous Numb expression level.

We cultured the FACS sorted Numb^high and Numb^−/low prostate cancer cells in vitro and monitored the GFP status continuously for 2 to 3 months. As shown in Fig. 5F, the Numb^−/low prostate cancer cells gradually generated both Numb^high and Numb^−/low cells, while the GFP intensity in the Numb^high prostate cancer cells remained high even after 3 months of culture. We also tracked the GFP expression status in single Numb^high and Numb^−/low prostate cancer cell–derived colonies. As shown in Fig. 5G, single LNCaP-pNumb-GFP^−/low cells generated the following 3 different types of colonies: (i) all cells were GFP^−/low, (ii) all cells were GFP^high, (iii) both GFP^high and GFP^−/low cells were present. In contrast, the single GFP^high cells only produced colonies comprised of GFP^high offspring cells. The type I colonies were more common than the other two types of colonies (Fig. 5G). These data suggested that the Numb^−/low cells were capable of giving rise to GFP^high cells, but not vice versa.

To test the effect of Notch or Hedgehog signaling inhibition on the GFP status transition of Numb^−/low and Numb^high cells, we cultured the Numb promoter GFP reporter lentiviruses transfected unsorted prostate cancer cells or sorted Numb^−/low and Numb^high cells with medium containing DAPT, cyclopamine, or a combination of DAPT and cyclopamine, respectively, and monitored the GFP status for 2 weeks. As shown in Supplementary Figs. S6 and S7, we observed that the DAPT treatment did not cause an obvious GFP intensity shift in C4-2B cells at the time of 7 or 14 days, whereas the cyclopamine treatment led to a gradual increase of GFP intensity at the time of 7 and 14 days. Combined treatment of DAPT and cyclopamine resulted in a notable transition of Numb-GFP^−/low cells to Numb-GFP^high cells.

Consistently with the above-mentioned observations in which the Notch and Hedgehog pathways were repressed by Numb, we found that the FACS sorted Numb^−/low prostate cancer cells preferentially expressed Notch signaling-specific genes (Notch1, Notch2, Hes1, and Hey1) and Hedgehog signaling-specific gene (GLI1) and three ligands, including Sonic (Shh), Desert (Dhh) and Indian Hh (Ihh). The expression of PTCH1, which is a negative regulator of Hedgehog signaling, was also underexpressed in the Numb^−/low cells compared with that in the Numb^high prostate cancer cells (Fig. 5H).

We also compared the expression of the stemness-related genes Bmi1, SOX2, OCT4, Nkx3.1 and Nanog and the prostatic differentiation markers AR and PSA between Numb^high and Numb^−/low C4-2B or LNCaP cells. As shown in Fig. 5H, the Numb^−/low prostate cancer cells expressed higher levels of SOX2, OCT4 and Nanog than the Numb^high cells, whereas the differentiation marker PSA had an expression pattern that was opposite to that of the stem cell related genes. Consistently, the mRNA levels of SOX2 and OCT4 were significantly down regulated by the Numb knockdown but increased by the forced expression of Numb (Fig. 2A, C, and D). In addition, because PSA^−/lo prostate cancer cells were previously identified as castration-resistant cancer stem cells, we studied the possible relationship between Numb expression and PSA expression and found that there was indeed a positive correlation between their mRNA and protein levels (Fig. 2A, C and D). Consistently with our data, the Numb mRNA expression was also found to be down-regulated in the PSA^−/lo compared with that in the PSA^hi prostate cancer cell population in the previous study (25). Although AR mRNA expression levels were comparable between Numb^high and Numb^−/low cells, the percentage of cells with nuclear localized AR protein was significantly lower in Numb^−/low prostate cancer cells than Numb^high cells by immunofluorescent staining of AR in the Numb^−/low and Numb^high prostate cancer cells (Supplementary Fig. S8).

Thus, these data indicated that Numb^−/low cells expressed higher levels of stemness genes and lower nuclear localized AR protein and were positively correlated with PSA^−/lo prostate cancer cells.

To investigate the biological differences between the Numb^high and Numb^−/low prostate cancer cells, we evaluated the cell proliferation, sphere-forming, anti-androgen and tumor-initiating capacities in these two populations. As shown in Fig. 5B, the flow cytometric analysis revealed that the Numb^−/low cells were smaller than the Numb^high cells. In addition, using propidium iodide staining and cell cycle analysis, we found that significantly more Numb^−/low prostate cancer cells resided in the G₀–G₁ quiescent stage than the more actively cycling Numb^high cells (Fig. 6A). Consistently, when plated in culture dishes at a low density, single Numb^−/low prostate cancer cells generated much smaller colonies than the Numb^high prostate cancer cells, indicating a shorter doubling time of Numb^high compared with that in the Numb^−/low cells. However, the Numb^−/low LNCaP cells regained their growth advantage under bicalutamide-mediated androgen deprivation (Fig. 6B). Furthermore, when cultured in Matrigel for a serial sphere-formation assay, the Numb^−/low prostate cancer cells formed significantly more and larger spheres than the Numb^high cancer cells (Fig. 6C and D). We then performed an in vivo xenograft tumor growth experiment using purified Numb^−/low and Numb^high cancer cells and monitored the tumor growth rate and tumor weight. As shown in Fig. 6E, we detected an evidently accelerated tumor growth in the C4-2B-Numb^−/low cells compared with that in the C4-2B-Numb^high cells.

We next compared the response with androgen deprivation in the Numb^high and Numb^−/low prostate cancer cells. An apoptosis analysis using Annexin V staining suggested that the androgen deprivation treatment, which was achieved by the inclusion of 10% CDSS and 50 μmol/L flutamide into the culture medium for 2 days, led to much more profound cell apoptosis in the LNCaP-Numb^high cells than in the LNCaP-Numb^−/low cells (Fig. 6F). Altogether, these data provide evidence that Numb^−/low cells are a distinct group of small and slow cycling prostate cancer cells and are associated with greater sphere-forming and tumor-initiating capacities and stronger resistance to androgen deprivation.

We showed that Numb exerted an inhibitory effect on prostate cancer growth and androgen independence by suppressing Notch and Hedgehog signaling (Fig. 3D–H), and that the castration-resistant Numb^−/low prostate cancer cells expressed higher levels of Notch and Hedgehog downstream genes. These findings prompted us to postulate that Notch and Hedgehog signaling activities contribute to the androgen-independent Numb^−/low prostate cancer cell phenotypes. We, therefore, tested whether the blockade of Notch and Hedgehog signaling activation by small molecule inhibitors was able to suppress Numb^−/low prostate cancer cells. As shown in Fig. 6G, the Numb^−/low LNCaP cells were more resistant to bicalutamide-mediated androgen deprivation than the Numb^high cells. DAPT or cyclopamine treatment alone did cause a significant increase of bicalutamide induced apoptosis on prostate cancer cells. But blockage of both Notch and Hedgehog signaling activities markedly potentiate the effect of bicalutamide on Numb^−/low LNCaP cells. Similar effects were observed under flutamide-mediated androgen deprivation combined with DAPT and/or cyclopamine (Supplementary Fig. S9). Thus, the inhibition of the Notch and Hedgehog signaling pathways significantly increased apoptosis in the Numb^−/low cells in response to the androgen deprivation.

Therapeutic intervention for CRPC remains a great scientific and clinical challenge due to our limited understanding of the underlying molecular and cellular mechanisms of this dreadful disease. Herein, we uncover that Numb plays a suppressive role in prostate cancer and the development of androgen independence. We report in this study that Numb is downregulated in prostate cancer and is closely related to the disease's aggressiveness. The Numb knockdown upregulates Notch and Hedgehog signaling activities and promotes resistance to androgen-deprivation therapy in prostate cancer cells. The forced expression of Numb exerts the opposite effects. Moreover, the restoration of Notch and Hedgehog signaling activities abrogated the negative impact of Numb on castration resistance and in vivo tumor formation.

One of the major obstacles in treating CRPC is the immense tumor cells heterogeneity. It has been proposed and validated in multiple tumor types that cancer cells are heterogeneous and hierarchically organized. A small population of cancer cells, which are referred to as cancer stem cells (CSC) or tumor-propagating cells, accounts for the progression, recurrence, and therapy resistance (25, 36–38). Using a Numb promoter reporter lentiviral system, we developed a reliable approach for distinguishing prostate cancer cell populations with a high or low endogenous Numb protein amount. We find that Numb^−/low enriches a distinct group of small- and slow-cycling prostate cancer cells that are associated with higher sphere-forming and tumor-initiating capacities and stronger resistance to androgen deprivation. Stemness-related genes and Notch and Hedgehog signaling downstream genes are preferentially expressed in Numb^−/low prostate cancer cells. Therefore, Numb^−/low prostate cancer cells appear to be a subpopulation with certain features of prostate cancer stem cells, which may serve as a specific target population for anti-CRPC treatment.

Extensive studies during the past few decades revealed that Numb plays a conserved role as a cell fate determinant in various tissue types from Drosophila to mammals. The Numb protein is more frequently segregated into the differentiating daughter cell of an asymmetrically dividing tissue stem cell or recently reported cancer stem cell (39–41). However, less is known concerning the transcriptional regulation of Numb. In this study, we cloned four different lengths of the putative proximal Numb promoter. Using a luciferase reporter assay, we are able to identify the shortest Numb proximal promoter. Additionally, we find that the activity of the Numb proximal promoter is positively regulated by the Numb protein. Numb promoter-Numb protein-Hedgehog signaling forms an autoregulatory axis. This observation may provide novel clues regarding the mechanism by which the daughter cell that retains stemness from the asymmetrically division of a stem cell maintains its low Numb-expressing state.

Numb has been recently shown to negatively regulate oncogenic signaling pathways, such as Notch, Hedgehog, Wnt signaling, and p53 signaling in different cancer types (18, 27, 28, 41, 42). Our study demonstrated that Notch and Hedgehog but not Wnt, signaling pathways are suppressed by Numb in prostate cancers and contribute to the development of castration resistance induced by Numb downregulation. The Notch and Hedgehog pathways have been reported to play important roles in prostate development and tumorigenesis. For example, Notch signaling is critical for normal prostatic cell proliferation and differentiation (43). Hedgehog signaling, in combination with androgen signaling, regulates prostate epithelium regeneration and developmental patterning (44, 45). In prostate cancer, the inhibition of Notch and Hedgehog signaling leads to depletion of docetaxel-resistant tumor-initiating cells (46). Further studies are warranted to determine whether Numb^−/low prostate cancer cells with elevated Notch and Hedgehog signaling activities possess greater resistance to chemotherapy, such as docetaxel.

Collectively, we proposed a model in which low levels of Numb or Numb^−/low status enhances Notch and Hedgehog signaling activities, which promotes prostate cancer cell in vivo growth and resistance to androgen deprivation. The higher activity of Hedgehog signaling, in turn, inhibits the Numb promoter activity, which maintains the low expression of Numb. Pharmaceutical inhibition of both the Hedgehog and Notch signaling pathways may represent a useful approach for the targeted depletion of Numb^−/low CRPC cells.

No potential conflicts of interest were disclosed.

Conception and design: H.H. Zhu, W.-Q. Gao

Development of methodology: D.G. Tang, H.H. Zhu

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Guo, C. Cheng, Z. Ji, M. Wang

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Guo, Z. Ji, H.H. Zhu

Writing, review, and/or revision of the manuscript: Y. Guo, D.G. Tang, H.H. Zhu, W.-Q. Gao

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K. Zhang, C. Cheng, X. Wang, M. Wang, M. Chu

Study supervision: H.H. Zhu, W.-Q. Gao

This study is supported by funds to W-Q Gao from the Chinese Ministry of Science and Technology (2017YFA0102900 and 2013CB945600), the National Natural Science Foundation of China (NSFC, 81372189, and 81630073), the Science and Technology Commission of Shanghai Municipality (16JC1405700), Shanghai Eastern Hospital (Pudong) Stem Cell Research Base Fund, and the KC Wong foundation and by funds to H.H. Zhu from the NSFC (81772743), the State Key Laboratory of Oncogenes and Related Genes (90-16-03), Shanghai Institutions of Higher Learning [The Program for Professor of Special Appointment (Young Eastern Scholar) QD2015002], Shanghai Rising-Star Program (17QA1402100), School of Medicine, Shanghai Jiao Tong University (Excellent Youth Scholar Initiation Grant 16XJ11003), and Ren Ji Hospital (Seed Project RJZZ14-010).

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.