NDY1/KDM2B Functions as a Master Regulator of Polycomb Complexes and Controls Self-Renewal of Breast Cancer Stem Cells

NDY1/KDM2B is a jumonji domain–containing histone lysine demethylase with specificity against histone H3K36me2/me1 and H3K4me3 (1, 2). Its involvement in oncogenesis was originally suggested by studies that identified NDY1/KDM2B and its homolog NDY2/KDM2A as targets of provirus integration in retrovirus-induced rodent lymphomas (1). Subsequently, it was shown that NDY1/KDM2B inhibits replicative senescence and promotes the immortalization and proliferation of mouse embryo fibroblasts (MEF) in culture (2, 3). This is due, at least in part, to the repression of the INK4A-INK4B/Arf locus, via the NDY1-dependent coupling of histone H3K36me2/me1 demethylation, with histone H3K27 trimethylation and histone H2AK119 ubiquitination. Mechanistically, it has been shown that the coupling to histone H3K27 trimethylation is mediated by the recruitment of EZH2, the enzymatic subunit of polycomb repressive complex 2 (PRC2) and that the coupling to histone H2AK119 ubiquitination is mediated by the recruitment of PRC1 (2). Notably, we recently showed that NDY1/KDM2B is induced by FGF-2 via DYRK1A activation and CREB phosphorylation. The induced NDY1 binds the miR101 gene in concert with EZH2, and this results in the repression of miR101 and the upregulation of its target EZH2. The upregulation of EZH2 reinforces this pathway and locks the cells in a state characterized by the overexpression of both NDY1 and EZH2. One type of human cancer where EZH2 is overexpressed, often via the activation of the FGF-2-NDY1-miR101–EZH2 axis is bladder cancer (4). Other recent studies have also shown that NDY1/KDM2B binds CpG islands (CGI) via its CXXC motif and recruits PRC1, which catalyzes histone H2AK119 ubiquitination (1, 2, 5–7). Consistent with these observations, NDY1/KDM2B facilitates stem cell reprogramming of differentiated cells in culture (1, 8, 9) and protects cells from oxidative stress and ROS-induced DNA damage by regulating the expression of antioxidant genes (2, 3, 10). Finally, it is expressed at high levels in various types of stem cells and its expression declines with differentiation (2, 6) The preceding data, combined with observations showing that NDY1 is overexpressed in human T-cell and B-cell acute lymphoblastic leukemia (T-ALL and B-ALL), acute myeloid leukemias (AML; refs. 4 and 11), seminomas (12), and mammary (13, 14) and pancreatic adenocarcinomas (15), strongly suggest that NDY1 functions as an oncogene.

In this study, we knocked down NDY1 in a set of 10 cell lines derived from a broad range of human tumors. In all these cell lines, the NDY1 knockdown consistently induced G₁ delay and senescence and/or apoptosis and interfered with anchorage-dependent and -independent growth. Focusing in a set of two basal-like and two luminal mammary adenocarcinoma cell lines, we showed that the knockdown of NDY1 promotes changes in the expression of cancer stem cell markers, suggesting loss of stemness. Mammosphere assays and orthotopic inoculation of decreasing numbers of the tumor cells in mice confirmed this suggestion by showing that the knockdown of NDY1 results in a decrease in the number of tumor-initiating cells. Consistent with this observation, the knockdown of NDY1 promoted the differentiation of basal-like tumor cells to luminal cells and this change in phenotype remained stable upon orthotopic transplantation of the tumor cells into the mouse mammary gland. Studies addressing the molecular mechanisms of these phenomena revealed that NDY1 functions in concert with EZH2 to repress a set of miRNA, which target the PRC1 and PRC2 subunits, BMI1, RING1B, SUZ12, and EZH2 itself. Although EZH2 and BMI1 also function in this pathway, they cannot rescue the defect induced by the loss of NDY1, which functions as the master regulator of PRC1 and PRC2 by integrating the activities of these complexes. Consistent with these observations, NDY1/KDM2B is expressed at higher levels in basal-like than in luminal mammary adenocarcinomas and its expression is associated with higher rates of relapse after treatment. More important, NDY1-regulated miRNAs are expressed at low levels in normal and neoplastic mammary stem cells, and their expression correlates negatively with the expression of NDY1/KDM2B, as well as the expression of their polycomb targets.

Detailed descriptions of the materials and the methods are presented in the Supplementary Materials.

Cancer cell culture conditions as well as pre-miRs and anti-miRs are listed in Supplemental Materials and Methods. SKOV3, OVCAR3, and OVCAR4 cell lines were kindly provided by Dr. A. Kuliopulos (Molecular Oncology Research Institute, Tufts Medical Center, Boston, MA) and SUM159 cells by Dr. S. Ethier (Hudson-Webber Cancer Karmanos Cancer Institute, Wayne State University, Detroit, MI). All other cell lines were obtained from ATCC. The authors carried out no additional authentication within the last 6 months. Retroviral and lentiviral packaging and infections were carried out using standard methods. Cell proliferation was measured with the MTT assay. Soft agar and sphere formation assays were performed as previously described (4, 16). All studies with cancer cell lines were limited to established cell lines. Patient-derived cell lines were not used.

RNA isolation and real-time RT-PCR were carried using standard procedures. The miRNA levels were normalized based on U6 small nuclear RNA (internal controls). The mRNA levels were normalized based on GAPDH or 18S rRNA (internal control).

Total cell lysates were generated using a Triton X-100 lysis buffer. Western blots of cell lysates resolved by SDS-PAGE were probed with the antibodies listed in the Supplementary Materials.

Chromatin immunoprecipitation (ChIP) was performed using the Chromatin Immunoprecipitation Assay Kit (Millipore, cat no. 17-295).

Cell-cycle distribution was determined as previously described (10). Apoptosis was measured by Annexin V staining and flow cytometry (BD Pharmigen, cat no. 556419) and/or by the Caspase-Glo 3/7 Assay (Promega, cat no. G8091). Cells stained for CD24-FITC/EpCAM-PE/CD49f-APC or ALDH activity (ALDEFLUOR StemCell Technologies, cat no. 01700; CD24-FITC/CD44-APC) were also analyzed by flow cytometry.

Using a protocol approved by the Tufts University IACUC committee, MDAMD-231 cells transduced with shControl or shNDY1 were injected into the mammary gland of NOD/SCID mice at concentrations of 10⁶, 10⁵, 10⁴, or 10³ cells per gland. Significance of differences in cancer-initiating cell numbers between these cells was calculated, using the χ² test, and extreme limiting dilution analysis (ELDA) software (17).

Data on the expression of NDY1/KDM2B in different types of breast cancer and the rates of relapse after treatment were calculated from the study GSE19783 (18). Significance of differences in the rate of recurrence in patients with NDY1/KDM2B high and low tumors was calculated with the Gehan–Breslow–Wilcoxon test. Finally, the correlations between the expression of NDY1, the NDY1-regulated miRNAs and their polycomb targets were made using RNA isolated from a set of 16 primary human breast cancer specimens (Supplementary Table S5). The significance of the correlation between the expression of NDY1 and NDY1-regulated miRNAs, as well as between NDY1 and the genes encoding members of the polycomb complexes in these tumors, was calculated using the Pearson statistical test. Because of the small size of the tumor samples, gene expression was examined only at the RNA level. Tumors were obtained from the Tufts Cancer Center tissue bank, following Institutional Review Board approval (for details, see Supplementary Table S5).

All experiments were performed three times and the data are presented as mean ± SD unless otherwise stated. Two-group comparisons were performed with the Student t-test and P values of 0.05 or less were considered significant (***, P < 0.001, **, 0.001 < P < 0.01; *, 0.01 < P < 0.05).

To elucidate the role of NDY1/KDM2B in the proliferation and survival of cancer cells, we knocked it down in a broad range of established cancer cell lines. Monitoring these cells revealed that the depletion of NDY1 significantly inhibits both live cell accumulation in culture monolayers and colony formation in soft agar (Fig. 1A and B and Supplementary Fig. S1A–S1C), suggesting that NDY1/KDM2B is protumorigenic (19). Four of the cell lines were of mammary epithelial origin and of these two were basal-like (MDAMB-23 and SUM159) and two luminal (T47D and MCF7). Because our focus is on breast cancer, further studies were carried out using these cell lines.

To address the mechanism responsible for the effects of the knockdown on the accumulation of live cells in culture, we first asked whether knocking down NDY1/KDM2B interferes with cell-cycle progression. Flow cytometry of EtBr-stained semiconfluent cell cultures growing under normal tissue culture conditions revealed that the knockdown of NDY1 induces a partial G₁ arrest in all the cell lines (Fig. 1C and Supplementary Fig. S1D), and suggested that NDY1 contributes to progression from G₁ to S.

The knockdown of NDY1 may interfere with the accumulation of live cells in culture also by promoting senescence or apoptosis. In agreement with our earlier observations in MEFs (1), light microscopy of semiconfluent monolayers, stained for β-galactosidase (β-gal), revealed that the knockdown elicits a strong senescence phenotype, which, however, is limited to T47D cells (68% β-gal–positive; Fig. 1D). Flow cytometry of Annexin V–stained MDAMB-231-shNDY1, MCF7-shNDY1, and T47D-shNDY1 cells, and their shRNA controls, revealed that shNDY1 promotes apoptosis, primarily in the first two cell lines (Fig. 1E). We conclude that although the knockdown of NDY1 inhibits G₁ progression in all the tumor cell lines we examined, its ability to induce senescence and apoptosis is selective.

The preceding data addressed the role of NDY1/KDM2B in transformed cells. To determine whether NDY1 is also required for the initiation of transformation, we transduced MCF-10A cells, an immortalized but not transformed mammary epithelial cell line, with shNDY1 or shRNA-control lentiviral constructs and we superinfected them with an H-Ras-V12 retrovirus. Of these cells, only the shControls superinfected with H-Ras-V12-formed colonies in soft agar (Supplementary Fig. S2A and S2B). Cell-cycle analysis of subconfluent monolayer cultures of the same cells showed that the shNDY1 cells accumulate in G₁ (Supplementary Fig. S2C). Finally, shRNA control cells transduced with the H-Ras-V12 retrovirus formed mammospheres when cultured in suspension, whereas the shNDY1 cells did not (Supplementary Fig. S2D). These findings combined show that NDY1 is required not only for the maintenance, but also for the initiation of the cell transformation phenotype.

Tumor cell lines contain populations of cells that possess tumor-initiating properties. These cells tend to form spheres when grown in suspension in defined serum-free media and they are known as tumor initiating, or cancer stem cells (for review, see ref. 20). Tumor-initiating cells in mammary carcinoma cell lines form mammospheres (16, 21). Suspension cultures of MDAMB-231, SUM159, MCF7, and T47D cells, transduced with an shNDY1 lentiviral construct gave rise to fewer and smaller mammospheres than similarly cultured shRNA-control cells (Fig. 2A–C). This suggests that NDY1 is required for the maintenance of tumor-initiating cells, as well as for their proliferation in the mammosphere environment. Reculturing cells harvested from mammospheres of two of the cell lines (MDAMB-231 and MCF7) revealed that the difference of shNDY1 and shRNA-control cells in mammosphere-forming potential increases with passage (Supplementary Fig. S3A), suggesting that NDY1 is also required for the long-term maintenance of the tumor-initiating phenotype.

The preceding results prompted us to analyze the same cells for the breast cancer stem cell markers CD24/CD44 and ALDH (22, 23). Flow cytometry of anti-CD44– and anti-CD24–stained MDAMB231 and MCF7 cells, transduced with shNDY1 or shRNA-control, revealed that shNDY1 causes a decrease in the CD44^high/CD24^low population and a parallel increase in the double-negative and double-positive populations (Fig. 2D and Fig. S3B), indicative of stem cell depletion. Testing the same cells for ALDH activity revealed that shNDY1/KDM2B causes a drop in ALDH^high cells (Fig. 2E). These results combined, support the mammosphere data, suggesting that NDY1/KDM2B is required for the maintenance of the stem cell population.

The stage of differentiation of normal mammary epithelial cells can be determined by the expression of several membrane markers. The expression of the same markers in mammary adenocarcinomas, in combination with the genome wide expression profiles of the tumor cells, have been used to link breast cancer subtypes with the known differentiation stages of normal mammary epithelia. (24–27). Basal, triple-negative tumor cells, represented by the MDAMB-231 and SUM159 lines, have a bipotent myoepithelial/luminal progenitor cell phenotype, whereas luminal tumor cells, represented by the MCF7 and T47D lines, have a differentiated luminal cell phenotype. The results of the mammosphere experiment suggested that NDY1 contributes to the maintenance of mammary tumor cells in an undifferentiated state. To address this hypothesis, MDAMB-231 and SUM159 cells transduced with shNDY1, or shRNA-control lentiviral constructs, were stained with anti-CD49f-APC, anti-EpCAM-PE, and anti-CD24-FITC antibodies. Flow-cytometric 2-color analysis for CD49f/EpCAM or CD49f/CD24 revealed that the knockdown of NDY1 promotes the downregulation of CD49f in concert with the upregulation of EpCAM and CD24 (Fig. 3A). Single-color analysis of the same cells confirmed the preceding results (Supplementary Fig. S4A). These results suggest that NDY1/KDM2B is essential for the maintenance of the basal/myoepithelial/luminal progenitor cell phenotype. This suggestion is supported by the results in Fig. 3B, which show that whereas the basal markers CD10, NOTCH4, and CK14 (25) are dramatically downregulated in shNDY1-transduced cells, the luminal markers CK18, CK19, and GATA3 (25) are upregulated.

In parallel experiments, 2 color and single-color analyses of MCF7 and T47D cells transduced with shNDY1 or shRNA-control constructs also showed that shNDY1 downregulates CD49f and upregulates CD24 (Supplementary Fig. S4B and S4C). The knockdown of NDY1 in these cells promoted the upregulation of GATA3, CK18, and CK19, as in MDAMB-231 and SUM159 cells. However, it also promoted the upregulation rather than the downregulation of NOTCH4 and CD10 (Supplementary Fig. S4D). This suggests that NDY1 may be required for the maintenance of both the basal and the luminal differentiation state.

The hypothesis that NDY1/KDM2B may be expressed at higher levels in basal-like versus luminal breast cancer, which was suggested by the preceding data, was confirmed by comparing its expression in basal-like versus all other types of human mammary adenocarcinomas in the study GSE19783 (Supplementary Fig. S4E; ref. 18). Comparison of the relapse rates of NDY1/KDM2B high and low tumors, following treatment-induced complete remission in the same study, revealed that the high NDY1 tumors tend to relapse earlier. Given the relatively small number of samples, the marginally significant difference of the relapse rates (P = 0.13) is suggestive of a significant relationship.

Infection of shNDY1-transduced T47D cells with a murine NDY1/KDM2B retroviral construct restored NDY1/KDM2B expression (Supplementary Fig. S5A). The biological outcome of this rescue experiment was the restoration of cell-proliferation (Supplementary Fig. S5B), cell-cycle progression (Supplementary Fig. S5C), growth in soft agar (Supplementary Fig. S5D and S5E), mammosphere formation (Supplementary Fig. S5F and 5G), and CD24 expression (Supplementary Fig. S5H). These findings confirmed that the shNDY1 phenotype is because of the knockdown of NDY1/KDM2B and not to off-target effects of shNDY1.

The preceding experiments suggested that NDY1/KDM2B regulates cancer stem cell self-renewal and differentiation. Probing Western blots of shNDY1 or shRNA-control MDAMB-231, MCF7, and T47D cells with antibodies against a panel of PRC1 and PRC2 subunits revealed that the knockdown of NDY1 downregulates the PRC2 subunits EZH2 and SUZ12, but not EED, and the PRC1 subunits BMI1 and RING1B (Fig. 4A). Given the function of PRC1 and PRC2 (for review, see refs. 28–30), these data provide potential mechanistic support to the hypothesis that NDY1 regulates cancer stem cell self-renewal. In agreement with these results, knocking down NDY1 induced the expression of CDKN1A/p21, a known target of PRC1 (31), and RUNX3, WNT1, HOXA4, and PLCβ4, known targets of PRC2 (Fig. 4B; ref. 32). The upregulation of WNT1 was surprising, because WNT signaling has been linked to stem cell self-renewal (for review, see ref. 33). However, additional experiments showed that despite the upregulation of WNT1, the knockdown of NDY1 inhibited the WNT pathway, probably because of changes in other regulators of WNT signaling (Supplementary Fig. S6).

We had previously shown that the upregulation of NDY1 activates an EZH2/miR101 switch, which ultimately locks the cells in a developmental state characterized by high EZH2 and low miR101 (4). We therefore hypothesized that shNDY1 may downregulate EZH2 by reversing this pathway. We also hypothesized that it may downregulate SUZ12, BMI1, and RING1B by similar mechanisms. Measuring the expression of a set of miRNAs that target the subunits of PRC1 and PRC2 listed above (34, 35) in shNDY1 and shControl-transduced MDAMB-231, MCF7, and T47D cells, confirmed that the knockdown of NDY1 de-represses all the miRNAs in this set (Fig. 4C). Chromatin immunoprecipitation experiments revealed that endogenous NDY1/KDM2B binds the genes encoding all these miRNAs, in concert with EZH2. Parallel ChIP experiments showed that whereas EZH2 binds, NDY1 does not bind the promoter region of protein-coding genes that are also repressed by NDY1 via EZH2, such as ADRB2 and WNT1 (Fig. 4D).

Superinfection of shNDY1-transduced MDAMB-231 cells with retroviral constructs of EZH2 or BMI1 failed to rescue the defects in monolayer growth, colony formation in soft agar, and mammosphere formation induced by the NDY1/KDM2B knockdown (Supplementary Figs. S7 and S8). In agreement with these results, EZH2 also failed to increase the binding of NDY1/KDM2B and EZH2 to the PRC1- and PRC2-targeting miRNAs that were de-repressed by the NDY1 knockdown (Supplementary Fig. S7E, left and middle) and as a result, it failed to repress these miRNAs (Supplementary Fig. S7E, right). These results are consistent with the hypothesis that the binding of EZH2 correlates with histone H3K36me2 and H3K36me1 demethylation by NDY1/KDM2B (2, 4) and they support the idea that BMI1 and EZH2 function in concert with NDY1, which is master regulator of PRC1 and PRC2.

The phenotype of the NDY1/KDM2B knockdown in mammary adenocarcinomas depends on the de-repression of NDY1/KDM2B-regulated miRNAs that target subunits of PRC1 and PRC2.

To address the hypothesis that the downregulation of EZH2, SUZ12, BMI1, and RING1B in shNDY1-transduced MDAMB-231 cells is because of the de-repression of miR101, miR181, miR200b, and miR203, shControl and shNDY1-transduced MDAMB-231 cells were transfected with pre-miRs and anti-miRs of these miRNAs, respectively. Western blots of cell lysates probed with the indicated antibodies, confirmed that whereas the pre-miRs upregulate EZH2, SUZ12, BMI1, and RING1B, the anti-miRs downregulate them (Fig. 5A). We conclude that the de-repression of these miRs in shNDY1 MDAMB-231 cells is responsible for the downregulation of the PRC1 and PRC2 subunits.

Pre-miR expression in shControl cells enhanced, whereas anti-miR expression in shNDY1 cells inhibited the activation of caspase-3 and caspase-7 (Fig. 5B). Moreover, pre-miRs inhibited, whereas anti-miRs enhanced cell proliferation in the same cells (Fig. 5C). Finally, the pre-miRs reproduced partially the NDY1/KDM2B knockdown phenotype on the expression of the stem cell markers CD44 and ALDH1 in shControl cells, whereas the anti-miRs partially blocked the NDY1 knockdown phenotype in shND1 cells (Fig. 5D and E). We conclude that NDY1/KDM2B plays an obligatory role in the regulation of miR101, miR181, miR200b, and miR203 and that PRC1 and PRC2, which are under the direct control of these miRNAs, function as critical effectors of tumor cell survival, proliferation, stem cell maintenance, and differentiation.

To determine the effects of shNDY1 in vivo, shNDY1 and shControl-transduced MDAMB-231 cells were inoculated into the mammary gland of NOD/SCID mice. Monitoring tumor growth for up to 70 days after injection revealed that the growth of cells transduced with the shNDY1 construct was significantly delayed (Fig. 6A–C). More important, tumors arising in animals inoculated with shNDY1-transduced cells express lower levels of the stem cell marker ALDH1 and basal marker CD49f but higher levels of CD24 and EPCAM (Fig. 6D), as measured by qRT-PCR in mRNA isolated from these tumors. The lower levels of ALDH1 in the harvested tumors suggest that the reduced expression of ALDH1 in shNDY1-transduced cells is stable, despite microenvironmental signals and the plasticity of the cancer stem cells. However, we should point out that ALDH1 expression in the xenografts was examined only at the RNA and not at the protein level. We conclude that the loss of NDY1/KDM2B inhibits the growth and promotes the differentiation of mammary adenocarcinoma cell lines from the basal-like to the luminal phenotype and interferes with the self-renewal of cancer stem cells, both in culture and in animals.

To determine whether shNDY1 lowers the number of tumor-initiating cells (cancer stem cells), as predicted by the in vitro experiments and by the downregulation of ALDH1 in the shNDY1 tumor xenografts, we repeated the orthotopic inoculation of shControl and shNDY1 cells, using 4 different dilutions of tumor cells (Fig. 6E). Extreme limiting dilution analysis (ELDA; ref. 17) of the results revealed that knocking down NDY1 decreases the cancer stem cell number from 1 of 64,248 in shControl to 1 of 36,5344 in shNDY1 cells, a ∼5.6-fold decrease (P < 0.000693). This confirms that shNDY1 depletes the cancer stem cell compartment of mammary adenocarcinoma cell lines.

To determine whether the NDY1/EZH2/miRNA/PRC axis we observed in tumor cell lines is active in primary human breast cancer, we examined the expression of the NDY1-regulated miRs and their polycomb targets in a set of 16 frozen primary breast cancer specimens obtained from the Tufts Cancer Center tissue bank. Measurement of the Pearson correlation between NDY1/KDM2B, miRNAs, and PRC subunits revealed that the expression of NDY1/KDM2B exhibits a negative correlation with the expression of the NDY1-regulated miRNAs and a positive correlation with the expression of the polycomb targets of these miRNAs (Fig. 7). The P value of some of the comparisons was smaller than 0.1, but higher than 0.05. Given the small number of samples and the consistency in the directionality of the data, we believe that even these not formally significant data are borderline significant. Here we wish to point out that, because of the small size of the tumor samples, the expression of the polycomb targets of the miRNAs regulated by NDY1/KDM2B was examined only at the RNA level. However, given that the same miRNAs regulate their polycomb targets at the RNA level in tumor cell lines in culture, we suggest that the observed miRNA/mRNA correlations are likely to reflect similar miRNA/protein correlations.

Earlier studies had shown that NDY1/KDM2B functions as an oncogene in retrovirus-induced rodent T-cell lymphomas (1). In addition, they provided evidence that NDY1 functions as an oncogene in human bladder and pancreatic carcinomas (4, 15). Experiments in this report focusing on a set of basal-like and luminal mammary adenocarcinoma cell lines provided support to the hypothesis that NDY1/KDM2B functions as an oncogene by showing that NDY1 sustains tumor growth, both in culture and in an orthotopic mammary adenocarcinoma xenograft model. In addition, they revealed that the loss of NDY1/KDM2B depletes the cancer stem cell compartment and promotes the differentiation of cancer cells, both in culture and in animals. Mechanistically, these activities of NDY1 depend on the concerted action of NDY1 and EZH2, which results in the repression of a set of miRNAs that target multiple subunits of PRC1 and PRC2.

The NDY1/EZH2/miRNA/PRC axis was originally observed and experimentally validated in established breast cancer cell lines. Even though patient-derived cancer cell lines were not utilized in this study, the relevance of this axis to human breast cancer was confirmed by observations that in primary tumors, the expression of NDY1/KDM2B exhibits a negative correlation with the expression of the NDY1-regulated miRNAs and a positive correlation with the expression of their polycomb targets. Of the miRNA targets of NDY1/EZH2, which were interrogated in this article, the members of the miR200 family had been shown previously to be expressed at low levels in normal human and mouse mammary stem cells and in breast cancer stem cells isolated from primary human breast cancer specimens (35, 36). These data support the role of NDY1/KDM2B in stem cell self-renewal in vivo. Further support for the essential role of NDY1 in stem cell biology in vivo was provided by the preliminary analysis of an RNA-Seq experiment in shControl and shNDY1-transduced T47D cells. This analysis, which will be presented in full elsewhere, showed that NDY1 also represses miR141, a member of the miR200c cluster, and all the members of the miR182/96/193 cluster (data not shown), which are also expressed at low levels in the normal and neoplastic mammary stem cells (36).

NDY1/KDM2B regulates PRC1 and PRC2 at multiple levels. Data presented in this report showed that NDY1/KDM2B functions in concert with EZH2 to repress a set of miRNAs that target multiple subunits of both PRC1 and PRC2. Equally important, the overexpression of EZH2 or BMI1, both of which were downregulated in shNDY1-transduced cells, failed to restore the phenotype in NDY1 knockdown cells. This finding is consistent with additional observations showing that EZH2 overexpression is not sufficient to repress at least some of its miRNA targets and that catalytically active NDY1 is required to activate the repressive function of EZH2. We conclude that NDY1 functions as a master regulator of both PRC1 and PRC2.

Basal-like breast cancer, a poorly differentiated neoplasm, is characterized by high risk of recurrence after chemotherapy and poor prognosis (37). Experiments presented in this article show that the knockdown of NDY1 inhibits the proliferation and survival of breast cancer cell lines and that it depletes them of cancer stem cells. Consistent with these observations, basal-like breast cancers tend to express higher levels of NDY1 and tumors with high NDY1 tend to recur earlier than tumors with low NDY1, following treatment-induced complete remission (18). Moreover, shNDY1 promotes the differentiation of basal-like tumor cell lines to a luminal cell phenotype. These findings suggest that NDY1/KDM2B promotes the survival and proliferation of tumor cells, as well as the self-renewal of cancer stem cells. Given the relative resistance of cancer stem cells to chemotherapy and their importance in tumor relapse after chemotherapy-induced remission (38), our data highlight the importance of NDY1/KDM2B as a chemotherapeutic target. The fact that luminal type breast cancers have a better prognosis than the basal-like ones is another reason why inhibition of NDY1/KDM2B may alter the response of basal-like tumors to chemotherapy.

Data presented in this report suggest that NDY1/KDM2B may prevent the differentiation of basal-like mammary adenocarcinomas to a differentiated luminal cell phenotype. A partial block at a similar stage of differentiation has been described in the mammary epithelia of BRCA1 mutation carriers (BRCA1^+/mut), who ultimately develop basal mammary adenocarcinomas (39). The apparent phenotypic similarity between BRCA1 mutations and NDY1/KDM2B raises the question whether the two function in the same pathway. This question will be addressed in future studies.

In summary, data presented in this report identify a novel miRNA-dependent mechanism by which NDY1/KDM2B regulates PRC1 and PRC2 and they show that, through this mechanism NDY1/KDM2B regulates the self-renewal and differentiation of tumor-initiating cells in human breast cancer cell lines. These data confirm the importance of NDY1/KDM2B as a therapeutic target for breast and perhaps other human cancers.

No potential conflicts of interest were disclosed.

Conception and design: F. Kottakis, P. Foltopoulou, B.T. Dake, C. Kuperwasser, P.N. Tsichlis

Development of methodology: F. Kottakis, P. Foltopoulou, I. Sanidas, B.T. Dake, P.N. Tsichlis

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): F. Kottakis, P. Foltopoulou, P. Keller, A. Wronski, B.T. Dake, S.A Ezell, Zhu Shend, S.P. Naber, E. McNiel, C. Kuperwasser, P.N. Tsichlis

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): F. Kottakis, P. Foltopoulou, I. Sanidas, P. Keller, B.T. Dake, Zhu Shend, C. Kuperwasser, P.N. Tsichlis

Writing, review, and or revision of the manuscript: F. Kottakis, B.T. Dake, S.P. Naber, P.W. Hinds, E. McNiel, C. Kuperwasser, P.N. Tsichlis

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S.P. Naber, P.N. Tsichlis

Study supervision: P.W. Hinds, P.N. Tsichlis

The authors thank Drs. M. Ewen and G. Sonenshein for critical review of the article and all the members of the P.N.T. laboratory for supporting this work with their intellectual and technical expertise. The authors also thank Dr. A. Kuliopoulos and Dr. S. Ethier for cell lines and the Tufts Cancer Center tissue bank for breast cancer specimens. Finally, the authors thank Dr. A. Touroutoglou for help with the statistical analyses.

P.N. Tsichlis wishes to dedicate this article to the memory of Dr. Patricia Lorenzo, a former member of the NIH Molecular Oncogenesis Study Section and the University of Hawaii Cancer Center. P. Lorenzo was an outstanding scientist and a committed member of the Study Section. She will be remembered for her Science, her courage in the face of terminal illness, and the overall strength of her character.

This work was supported by grants from the NIH (R01 CA109747; P.N. Tsichlis) and the Breast Cancer Research Foundation (NIH/NICHD HD073035, NIH/NCI CA25554, and CA170851; C. Kuperwasser).

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.