Recently, impressive technical advancements have been made in the isolation and validation of mammary stem cells and cancer stem cells (CSC), but the signaling pathways that regulate stem cell self-renewal are largely unknown. Furthermore, CSCs are believed to contribute to chemo- and radioresistance. In this study, we used the MMTV-Neu-Tg mouse mammary tumor model to identify potential new strategies for eliminating CSCs. We found that both luminal progenitor and basal stem cells are susceptible to genetic and epigenetic modifications, which facilitate oncogenic transformation and tumorigenic potential. A combination of the DNMT inhibitor 5-azacytidine and the HDAC inhibitor butyrate markedly reduced CSC abundance and increased the overall survival in this mouse model. RNA-seq analysis of CSCs treated with 5-azacytidine plus butyrate provided evidence that inhibition of chromatin modifiers blocks growth-promoting signaling molecules such as RAD51AP1 and SPC25, which play key roles in DNA damage repair and kinetochore assembly. Moreover, RAD51AP1 and SPC25 were significantly overexpressed in human breast tumor tissues and were associated with reduced overall patient survival. In conclusion, our studies suggest that breast CSCs are intrinsically sensitive to genetic and epigenetic modifications and can therefore be significantly affected by epigenetic-based therapies, warranting further investigation of combined DNMT and HDAC inhibition in refractory or drug-resistant breast cancer. Cancer Res; 76(11); 3224–35. ©2016 AACR.

Cancer stem cells (CSC), a small subpopulation of cells within tumors, have a characteristic feature of self-renewal, a process that drives tumorigenesis and differentiation contributing to cellular heterogeneity in tumors. CSCs are resistant to chemotherapy and radiotherapy and are considered a major obstacle in cancer treatment (1–3). This results in relapse of breast cancer in about 20%–45% of patients within years or decades after treatment. Thus, an effective cancer therapy requires elimination of all tumorigenic cells in the tumor (4). Breast tumors contain a heterogeneous population of cells such as neoplastic epithelial cells, mesenchymal stem cells, infiltrating immune cells, cancer-associated fibroblasts, angiogenic vascular cells, and erythrocytes (5). However, the molecular mechanisms that reprogram normal stem cells into abnormal CSCs are poorly understood.

Stem cells have much longer life span compared with their progeny and therefore, have a greater opportunity to accumulate genetic mutations (6). Hematopoietic stem cells provide the best evidence that normal stem cells could be the target of transforming genetic mutations, which can render them independent of growth signals and undergo uncontrolled proliferation and tumorigenesis. Recent studies have shown that epigenome also plays an important role in cancer initiation and propagation by regulating stem cells (7, 8). For example, ARID1A, a member of SWI/SNF family, is mutated in more than 50% of human cancers; however, this mutation does not directly stimulate tumor formation, rather it determines the epigenetic changes that lead to tumor propagation (9). Thus, the tumorigenic potential of ARID1A resides in its ability to alter the epigenetic profile rather than the DNA sequence. In this context, our recent studies have shown that DNA methyltransferase 1 (DNMT1) plays a critical role in the maintenance of mammary stem/progenitor cells and CSCs (10). Using mammary gland–specific Dnmt1-knockout mice, we have shown that DNMT1 is indispensable for mammary stem cells (MaSC) formation and that Dnmt1 deletion protects mice from mammary tumorigenesis by limiting CSC pool (10). Therefore, targeting the epigenetic modifiers like DNA methylation offers a promising treatment option for human cancers.

Epigenetic modifications represent early events in tumorigenesis (11, 12). Interestingly, unlike genetic mutations, the epigenetic alterations are reversible as proven by the reexpression of tumor suppressor genes by DNMT inhibitors (13). 5-azacytidine (5-AzaC, Vidaza) and 5-aza-2′-deoxycytidine (5-AzaDC, decitabine) are the most successful epigenetic drugs that are most widely used in clinics (14, 15). However, their use is restricted due to their toxicity and poor stability. Interestingly, combinations of 5-AzaC or 5-AzaDC with histone deacetylation inhibitors (HDACi) have been approved by FDA and European Medicines Agency (EMA) for treatment of hematologic malignancies (16). HDACs are upregulated in a wide variety of cancers, and HDACi have long been studied in clinical settings. These inhibitors produce a global effect on the level of acetylation of histone proteins (17). Our recent studies have shown that a combination therapy using 5-AzaC plus butyrate targets CSCs (10). However, the impact of this drug combination on CSCs at genome level has not been investigated. In the current study, we provide evidence that a combination of DNMT and HDAC inhibitors not only reduces the tumor mass but also targets CSCs and differentially regulates genes that are involved in tumor growth. Thus, this combination could be considered as an effective therapeutic strategy for breast cancer treatment. As this combination reduces the pool of drug-resistant CSCs, it can also be used to treat breast cancer patients who have developed resistance to hormone therapies such as trastuzumab.

Cell lines

The human breast cancer cell line, CAL51, was obtained from the DSMZ (Leibniz Institute DSMZ-German Collection of Microorganisms and Cell Cultures) in October 2011. MCF10A4, an invasive breast cancer cell line, was kindly provided by Dr. Fred Miller at the Barbara Ann Karmanos Cancer Institute (Detroit, MI) in April 2009. The mouse mammary tumor cell line 4T1 was obtained from ATCC in July 2011. Cell lines from ATCC and DSMZ have been thoroughly tested and authenticated and morphology, karyotyping, and PCR-based approaches were used to confirm the identity of the cell lines. CAL51 cells were grown in DMEM supplemented with 20% FBS, glutamine (1%), and penicillin (1%). MCF10A4 cells were grown in DMEM/F12 media supplemented with Donor horse serum (5%), glutamine (1%), penicillin (1%), EGF (20 ng/mL), insulin (10 μg/mL), and hydrocortisone (500 ng/mL). 4T1 cells were cultured in RPMI medium supplemented with 10% FBS, glutamine (1%), and penicillin (1%). All these cell lines have been routinely tested for mycoplasma contamination using the Universal Mycoplasma Detection Kit obtained from ATCC and the last mycoplasma test was performed in July 2014. Mycoplasma-free cell lines were used in all of our experiments.

Animals

Balb/cAnNCr (Stock #01B05) mice were obtained from the National Cancer Institute. FVB/NJ (Stock #001800), MMTV-Neu-Tg (Stock #005038), athymic Balb/c (Nu/Nu, Stock #007870), and NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (Stock #00557) mice were obtained from Jackson Laboratories. All these mice were bred and maintained in Augusta University Animal Facility in accordance with the guidelines of the Institutional Animal Care and Use Committee.

Drug treatment

For syngeneic tumor transplant studies, 4T1 cells (0.1 × 106) were injected into the mammary fat pad of Balb/c mice. Twenty-four hours after the tumor cell injection, 5-AzaC (0.5 mg/21-days release) and butyrate (10 mg/21-day release) tablets were implanted using trochar (Innovative Research of America). Similarly, in another group of mice, 24 hours after transplantation, salinomycin (5 mg/kg, intraperitoneal injection) treatment was initiated. For in vitro studies, cells were treated with 5-AzaC (1 μg/mL), 5-AzaDC (1 μg/mL), TSA (100 nmol/L), butryate (1 mmol/L), and salinomycin (1 μg/mL).

RNA-seq analysis

RNA samples extracted from mammospheres and tumorospheres were subjected to cDNA library construction (ScriptSeq RNA-Seq Preparation v2 from Epicenter). The quality of the sample was checked using Agilent Bioanalyzer and then subjected to Illumina sequencing using HiSeq 2000 (paired end 50 cycles V3). CASAVA1.8.2 was used to generate fastq files. To analyze the genome, we used Top Hat 2.0.1 software and detected the differential expression of transcripts (Cuffdiff/Cufflink 2.2.0). We detected raw mapped reads and normalized reads per kilobase per million mapped reads.

Details for single-cell preparation from mammary gland, generation of mammospheres and tumorospheres, flow cytometry, immunofluorescence staining and microscopic imaging, RNA isolation and real-time PCR, cell cycle, and clonogenic assays are given in Supplementary Materials section.

Statistical analysis

Statistical analysis was done using Student t test with two-tail distribution by GraphPad Software. Biologic function and pathway enrichment analysis were carried out using DAVID (http://david.abcc.ncifcrf.gov/), UCSC Cancer Browser (https://genome-cancer.ucsc.edu/proj/site/hgHeatmap/), and Ingenuity Pathway Analysis (http://www.ingenuity.com/products/pathways_ analysis.html) software. GraphPad, Sigma Plot, and Excel were used to draw figures. Kaplan–Meier analyses were used to determine group differences in tumor-free survival (http://kmplot.com/analysis/). The web interface of GOBO was used (http://co.bmc.lu.se/gobo/) for gene set analysis and coexpressed genes.

LinCD49f+CD24+ cells have tumor-propagating and metastatic potential

To distinguish normal MaSC and cancer-propagating stem cells (CSC), we isolated mammary stem/progenitor enriched cell populations from normal mammary glands (16-week-old) and CSC cells from tumor tissues of MMTV-Neu-Tg mice (9-month-old) using CD24/CD49f cell-surface markers. We observed three distinct cell populations in normal mammary glands: LinCD49fhighCD24+ (basal myoepithelial stem cells), LinCD49f+CD24high (luminal progenitor cells), and LinCD49fCD24 (stromal cells). In contrast, there were only two distinct cell populations in tumor tissues: (LinCD49f+CD24+ and LinCD49fCD24; Fig. 1A). As shown in Fig. 1B, the expression of CD49f and CD24 was dramatically increased in tumor tissues compared with normal mammary glands. Previous studies have shown that only a few cells within the tumor, the CSCs, are tumorigenic and possess the metastatic phenotype (18, 19). Thus, to test the tumor-forming potential of LinCD49f+CD24+ and LinCD49fCD24 cells, we injected these cells into mammary fat pad of 12-week-old NOD/SCID mice. As shown in Fig. 1C, LinCD49f+CD24+ cells were able to form tumors while the LinCD49fCD24 cells were not. Furthermore, we investigated the self-renewal capacity of LinCD49f+CD24+ and LinCD49fCD24 cells, using tumorosphere-forming (primary and secondary) and tumor-forming potentials (serial dilution, 102-105 cells). LinCD49f+CD24+ cells were able to form tumorospheres and tumors (as few as 102 cells) but LinCD49fCD24 cells were unable to form either tumorospheres or tumors (Fig. 1D and Supplementary Fig. S1A–S1C). We also tested the mammosphere-forming potential of the three distinct cell populations that were isolated from the normal mammary glands and found that LinCD49fhighCD24+ cells form larger size mammospheres than the LinCD49f+CD24low cells, while LinCD49fCD24 cells were unable to form spheres (Supplementary Fig. S1D).

Figure 1.

LinCD49f+CD24+ cells are tumor-propagating cells. A, representative FACS dot blots showing three distinct cell populations in normal mammary gland: myoepithelial stem cells (LinCD49fhighCD24+), luminal progenitor cells (LinCD49f+CD24high), and stromal cells (LinCD49fCD24). However, MMTV-Neu tumor tissues show only two distinct cell populations: tumor-propagating stem cells (LinCD49f+CD24+) and nontumorigenic cells (LinCD49fCD24; n = 5 mice). B, representative confocal images (×40) showing increased expression of CD49f and CD24 in tumor tissue compared with the normal mammary tissues (n = 3 mice). CD49f, green; CD24, red; DAPI, blue. Scale bars, 100 μm. C, representative tumor growth with LinCD49f+CD24+ cells (102 cells) injected in NOD/SCID mice. No tumor was detected with LinCD49fCD24 cells (105 cells; n = 3 mice). D, representative images (×10) of tumorospheres generated from LinCD49f+CD24+ and LinCD49fCD24 cells. 1,000 cells were used in each group (n = 3 mice). E, representative lung tissue images after tail vein injection of LinCD49f+CD24+ and LinCD49fCD24 cells (0.1 × 106 in each) showing lung metastatic nodules observed in LinCD49f+CD24+ cells but not in LinCD49fCD24 cells (n = 3 mice in each). F, representative H&E images (×40) of nonmetastatic (LinCD49fCD24) and lung metastatic tumor (LinCD49f+CD24+) tissues (n = 3 mice in each). G, representative FACS dot plots for nonmetastatic and metastatic lung tumor tissues showing increased LinCD49f+CD24+ cell population (n = 3 mice).

Figure 1.

LinCD49f+CD24+ cells are tumor-propagating cells. A, representative FACS dot blots showing three distinct cell populations in normal mammary gland: myoepithelial stem cells (LinCD49fhighCD24+), luminal progenitor cells (LinCD49f+CD24high), and stromal cells (LinCD49fCD24). However, MMTV-Neu tumor tissues show only two distinct cell populations: tumor-propagating stem cells (LinCD49f+CD24+) and nontumorigenic cells (LinCD49fCD24; n = 5 mice). B, representative confocal images (×40) showing increased expression of CD49f and CD24 in tumor tissue compared with the normal mammary tissues (n = 3 mice). CD49f, green; CD24, red; DAPI, blue. Scale bars, 100 μm. C, representative tumor growth with LinCD49f+CD24+ cells (102 cells) injected in NOD/SCID mice. No tumor was detected with LinCD49fCD24 cells (105 cells; n = 3 mice). D, representative images (×10) of tumorospheres generated from LinCD49f+CD24+ and LinCD49fCD24 cells. 1,000 cells were used in each group (n = 3 mice). E, representative lung tissue images after tail vein injection of LinCD49f+CD24+ and LinCD49fCD24 cells (0.1 × 106 in each) showing lung metastatic nodules observed in LinCD49f+CD24+ cells but not in LinCD49fCD24 cells (n = 3 mice in each). F, representative H&E images (×40) of nonmetastatic (LinCD49fCD24) and lung metastatic tumor (LinCD49f+CD24+) tissues (n = 3 mice in each). G, representative FACS dot plots for nonmetastatic and metastatic lung tumor tissues showing increased LinCD49f+CD24+ cell population (n = 3 mice).

Close modal

To test the metastatic potential of LinCD49f+CD24+ cells, we serially diluted these cells and injected into 12-week-old NOD/SCID mice. As shown in Fig. 1E, the LinCD49f+CD24+ cells (0.1 × 106) were able to form metastatic lesions in the lung. To confirm that the LinCD49f+CD24+ cells are indeed the drivers of metastasis to the lung, we prepared single-cell suspensions from the lung tissues and analyzed LinCD49f+CD24+ and LinCD49fCD24 cell populations. As shown in Fig. 1G, there were significantly more LinCD49f+CD24+ cells in the lung tissues of mice that were injected with LinCD49f+CD24+ cells than in the lung tissues of mice that were injected with LinCD49fCD24 cells. These findings suggest that LinCD49f+CD24+ cells could play a critical role in driving tumor metastasis to distant organs. Overall, these results provide evidence that LinCD49f+CD24+ cells are tumor-propagating and metastasis-driving CSCs.

Luminal progenitor and basal myoepithelial stem cells equally contribute to mammary tumor development

Despite ample evidence showing that cancer arises from populations of self-renewing stem cells, cancer can also arise from progenitor cells that originate from stem cells. For example, a mutation that enhances β-catenin signaling in granulocyte–macrophage progenitor cells causes a blast crisis in patients with chronic myeloid leukemia (20). This suggests that the mutated progenitor cells acquire the ability of self-renewal, a feature thought to be specific to stem cells, and undergo unlimited growth as cancer cells. To test the cell of origin in mammary tumorigenesis, either from basal myoepithelial stem cells or luminal progenitor cells, we used MMTV-Neu-Tg mouse. About 50%–60% of MMTV-Neu-Tg mice develop spontaneous mammary tumors after acquiring either additional mutations or epigenetic modifications (21). We analyzed different cell populations in mouse mammary glands of MMTV-Neu-Tg premalignant mice (3- to 12-month-old) with no palpable tumors. Surprisingly, some mice as young as 3-month-old showed an abrupt increase in luminal progenitor population (LinCD49f+CD24high), whereas some 12-month-old mice did not show this abnormal phenotype (Fig. 2A), suggesting that these luminal progenitor cells might have undergone genetic/epigenetic modifications.

Figure 2.

Distinct myoepithelial stem cells and luminal progenitor cells represent the cells of origin. A, representative contour plot of FACS gating shows an abnormal increase in luminal cells and shifting of myoepithelial stem cells toward luminal cells in 3-month-old MMTV-Neu premalignant mice. In contrast, 4- and 12-month-old mice show normal level of stem and luminal cells (n = 3 mice in each). B, tumors in Balb/c mice, driven by basal myoepithelial stem cells and luminal progenitor cells, which were derived from premalignant MMTV-Neu mice (n = 3 mice). C, tumor-forming potential of stem cells and luminal cells from premalignant MMTV-Neu mice, and tumor-propagating cells from MMTV-Neu mouse tumor tissue (n = 3 mice). D, frequency of tumor formation by different cell types, derived from MMTV-Neu premalignant mice. E, representative confocal images (×10) of keratin 8 (green), keratin 14 (red), and DAPI (blue) staining for sections of tumor tissues derived from transformed stem cells and luminal cells (n = 3 mice). Scale bars, 100 μm.

Figure 2.

Distinct myoepithelial stem cells and luminal progenitor cells represent the cells of origin. A, representative contour plot of FACS gating shows an abnormal increase in luminal cells and shifting of myoepithelial stem cells toward luminal cells in 3-month-old MMTV-Neu premalignant mice. In contrast, 4- and 12-month-old mice show normal level of stem and luminal cells (n = 3 mice in each). B, tumors in Balb/c mice, driven by basal myoepithelial stem cells and luminal progenitor cells, which were derived from premalignant MMTV-Neu mice (n = 3 mice). C, tumor-forming potential of stem cells and luminal cells from premalignant MMTV-Neu mice, and tumor-propagating cells from MMTV-Neu mouse tumor tissue (n = 3 mice). D, frequency of tumor formation by different cell types, derived from MMTV-Neu premalignant mice. E, representative confocal images (×10) of keratin 8 (green), keratin 14 (red), and DAPI (blue) staining for sections of tumor tissues derived from transformed stem cells and luminal cells (n = 3 mice). Scale bars, 100 μm.

Close modal

To test whether this abnormal cell population leads to formation of mammary tumor, we injected different cell types (LinCD49fhighCD24+, LinCD49f+CD24high and LinCD49fCD24) into mammary fat pad and found that both transformed basal myoepithelial stem cells and luminal progenitor cells were able to form mammary tumors around 120 and 100 days, respectively, whereas the LinCD49fCD24 cells were unable to form mammary tumors (Fig. 2B and D and Supplementary Fig. S2A and S2B). The transformed luminal progenitor cells exhibit earlier onset of mammary tumorigenesis with accelerated tumor growth than the basal myoepithelial stem cells (Fig. 2B and D and Supplementary Fig. S2A and S2B). Interestingly, tumors that developed from MMTV-Neu-Tg basal myoepithelial and luminal progenitor cell types are indistinguishable from the spontaneous tumors that develop in MMTV-Neu-Tg mice, indicating the stem cell–like property of self-renewal (Supplementary Fig. S2C). To confirm this observation further, we stained the tumors derived from basal myoepithelial stem cells and luminal progenitor cells and found that both tumors express keratin 14, basal epithelial markers, and keratin 8, luminal epithelial markers (Fig. 2E). We confirmed this observation further with additional basal epithelial markers (p63 and Nestin) and luminal epithelial marker (Gata3; Supplementary Fig. S2D; ref. 22). These results confirmed that LinCD49f+CD24+ cells are tumorigenic and that both transformed basal myoepithelial stem cells and luminal progenitor cells were able to form mammary tumors.

Transformed luminal progenitor and basal myoepithelial stem cells are susceptible to treatment with DNMT and HDAC inhibitors

As shown in Fig. 2, both luminal progenitor and basal myoepithelial stem cells contribute to mammary tumor development in MMTV-Neu-Tg mice. However, previous studies have shown that the luminal progenitors are the origin of MMTV-Neu-Tg–driven mammary tumor (23) and that basal stem cells are increased during later stages of the mammary tumor development, which contributes to drug resistance and metastasis (24, 25). Furthermore, we have shown that DNMTs, especially DNMT1, play a critical role in regulation of luminal progenitor and basal myoepithelial stem cells and that functional inactivation of DNMT1 significantly reduces both of these populations (10). Therefore, we analyzed the expression of the three major DNMTs (DNMT1, DNMT3A, and DNMT3B) in luminal progenitor and basal stem cells that were isolated from premalignant MMTV-Neu-Tg mice. To confirm the purity and free of cross contamination within the specific cell types, we analyzed keratin 14 and keratin 8 in basal, luminal, and stromal cells and found that there was no cross contamination among these cell types (Fig. 3A). Then, we analyzed DNMT1, DNMT3A, and DNMT3B in these cell populations and found that both luminal progenitor and basal stem cells express similar levels of DNMT1 and DNMT3A but basal stem cells express significantly more DNMT3B than luminal progenitor cells (Fig. 3B). This suggests that DNMT3B could play a critical role in regulation of basal stem cells. To test whether DNMT inhibitors have any effect in these populations, we generated tumorospheres from the tumor tissues derived from MMTV-Neu-Tg mice and treated with pan-DNMTs inhibitors 5-AzaC and 5-aza-2-deoxycytidine (5-AzaDC). We also used HDACi TSA and butyrate (But), either alone or in combination with DNMT inhibitors. We used salinomycin (Salino), the known stem cell inhibitor, as a positive control. We found that DNMT and HDACi significantly reduced CSCs (Fig. 3C and D) but the combination of DNMT and HDAC inhibitors (5-AzaC + But) had much more profound effect in reducing CSC populations.

Figure 3.

Treatment with DNMT and HDAC inhibitors restricts CSCs. A, representative confocal images (×25) of keratin 14 (red), keratin 8 (green), and DAPI (blue) expression in sorted myoepithelial stem cells, luminal progenitor cells, and stromal cells showing free of cross contamination (n = 5 mice). B, real-time PCR showing the relative levels of DNMT1, DNMT3A, and DNMT3B expression in myoepithelial stem and luminal progenitor cells. Data represent mean ± SD for 5 mice. C, representative tumorospheres images (×10) of control and treated (as indicated) for 7 days. D, quantification of primary and secondary tumorospheres derived from control and treated groups. Data represent mean ± SD of three mice with quadruplicate wells (n = 12 samples). E and F, tumorosphere size and number in MCF10A4, CAL51, and 4T1 cells with or without treatment (5-AzaC + But). G, survival curve for mice. 4T1 cells were injected into mammary fat pad of Balb/c mice and then treated with and without 5-AzaC + But and salinomycin (n = 5 mice in each).

Figure 3.

Treatment with DNMT and HDAC inhibitors restricts CSCs. A, representative confocal images (×25) of keratin 14 (red), keratin 8 (green), and DAPI (blue) expression in sorted myoepithelial stem cells, luminal progenitor cells, and stromal cells showing free of cross contamination (n = 5 mice). B, real-time PCR showing the relative levels of DNMT1, DNMT3A, and DNMT3B expression in myoepithelial stem and luminal progenitor cells. Data represent mean ± SD for 5 mice. C, representative tumorospheres images (×10) of control and treated (as indicated) for 7 days. D, quantification of primary and secondary tumorospheres derived from control and treated groups. Data represent mean ± SD of three mice with quadruplicate wells (n = 12 samples). E and F, tumorosphere size and number in MCF10A4, CAL51, and 4T1 cells with or without treatment (5-AzaC + But). G, survival curve for mice. 4T1 cells were injected into mammary fat pad of Balb/c mice and then treated with and without 5-AzaC + But and salinomycin (n = 5 mice in each).

Close modal

We then tested whether the combination of 5-AzaC and butyrate can inhibit CSC signature in human breast cancer cell lines using two metastatic human breast cancer cell lines, MCF10A4 (26) and CAL51 (27), and one mouse metastatic mammary tumor cell line, 4T1 (28). We generated tumorospheres in the presence and absence of 5-AzaC and butyrate. This combination significantly reduced CSC pool (Fig. 3E and F) and colony formation (Fig. Supplementary Fig. S3A and S3B). To confirm the efficacy of this combination in vivo, we implanted 4T1 cells into mammary fat pad of Balb/c mice and then treated with 5-AzaC + butyrate as a combination therapy. Salinomycin was used as a positive control. We found that 5-AzaC + butyrate significantly increased the overall survival compared with salinomycin treatment (Fig. 3G). Overall, these results provide evidence that DNA methylation plays a critical role in the regulation of CSCs in MMTV-Neu-Tg–driven mammary tumor and that DNMT and HDAC inhibitors can be used as a combination therapy to reduce CSC pool.

Treatment with DNMT and HDAC inhibitors differentially regulates genes that are involved in basal stem cell–driven breast cancer

To test the efficacy of 5-AzaC + butyrate on CSCs, we isolated LinCD49f+CD24+ cells from MMTV-Neu-Tg mouse and injected into NOD/SCID mice (Supplementary Fig. S4). Single-cell suspension prepared from the secondary tumors, which were derived from these mice, were then used to generate three-dimensional (3D) cultures. These 3D cultures were treated with 5-AzaC (1 μg/mL) plus butyrate (1 mmol/L) for 72 hours. This provided a unique opportunity to study the therapeutic efficacy of 5-AzaC and butyrate on self-renewing CSCs (Supplementary Fig. S4). We isolated RNA and performed RNA-seq analysis. We found that 5-AzaC + butyrate combination differentially regulated genes that are involved in cell cycle, cell division, kinetochore formation, chromosome segregation, and mitosis (Fig. 4A and B). Using the Ingenuity System Database (IPA) software, we analyzed signaling pathways using the differential gene expression that were altered significantly at or above P < 0.05 level between control and treated groups. This analysis identified cancer and organismal injury and abnormalities as the top two disease and disorders pathways that were altered significantly (Fig. 4C). As IPA showed a close association of genes that are affected in cancer, we performed in-depth molecular and functional analyses that are related to cancer. Interestingly, the top molecular functions altered were cell-to-cell signaling, cellular movement and cell morphology, cell-cycle regulation, programmed cell death, DNA replication, posttranslational modification, and molecular transport (Fig. 4D).

Figure 4.

Treatment with DNMT and HDAC inhibitors differentially regulates genes that are involved in various signaling pathways. A, heatmap, generated from the RNA-seq analysis, showing differential gene expression between control and treated groups (n = 3 mice). B, network map created by IPA software using differentially regulated genes and their function. C, IPA analysis showing the top disease and disorder–based genes that are differentially expressed between control and treated groups. D, molecular function of genes altered by combination of 5-AzaC + But.

Figure 4.

Treatment with DNMT and HDAC inhibitors differentially regulates genes that are involved in various signaling pathways. A, heatmap, generated from the RNA-seq analysis, showing differential gene expression between control and treated groups (n = 3 mice). B, network map created by IPA software using differentially regulated genes and their function. C, IPA analysis showing the top disease and disorder–based genes that are differentially expressed between control and treated groups. D, molecular function of genes altered by combination of 5-AzaC + But.

Close modal

We then investigated gene expression signature in different breast cancer subtypes using UCSC cancer genome browser to create heatmap for genes that were differentially expressed between untreated and treated (5-AzaC + butyrate) groups. Interestingly, genes that were downregulated by 5-AzaC + butyrate were significantly increased in human breast cancer samples, whereas genes that were induced by 5-AzaC + butyrate were decreased in different breast cancer samples (Supplementary Fig. S5). Furthermore, we found that expression of RAD51AP1, NUSAP1, and SPC25 was very high in basal breast cancer subtype compared with other subtypes. This suggests that these genes could play a critical role in basal stem cell–driven breast cancer. To validate these observations, we selected two genes (RAD51AP1 and SPC25) and analyzed the expression of these genes in treated and untreated groups. Combination of 5-AzaC + butyrate inhibited RAD51AP1 gene expression (Fig. 4A; Supplementary Table S1). Previous studies have shown that RAD51AP1 is associated with DNA double strand break repair and homologous recombination (29, 30). Many cancer drugs produce DNA lesions at replication fork; cancer cells might use RAD51, RAD51AP1, and other homologous recombination proteins to repair double strand breaks and maintain the cellular integrity. It has been shown that cell's ability to repair double strand breaks significantly affects the outcome of cancer treatment, whereas cells deficient in homologous recombination and DNA damage repair mechanisms are hypersensitive to drug-induced cell death, resulting in a better chemotherapeutic response and outcome (31).

RAD51AP1 is highly expressed in basal breast cancer

Using the UCSC cancer genome browser and GOBO gene enrichment application, we investigated RAD51AP1 expression in normal and various breast cancer subtypes. We found that RAD51AP1 expression was significantly higher in breast cancer, especially in the basal subtype, than in normal mammary tissue (Fig. 5A). Further detailed analysis of GOBO gene enrichment analysis revealed that RAD51AP1 gene expression is higher in triple-negative basal cell types and in HER-2–positive breast cancer in comparison with hormone-responsive luminal subtypes (Figs. 5B and Supplementary Fig. S6A). Next, we investigated RAD51AP1 gene expression in human tumor subtypes and PAM50 tumor subtypes and found that basal cells express higher levels of this gene (Fig. 5B and Supplementary Fig. S6B). ER-negative human breast tumor subtype expresses high levels of RAD51AP1 (Supplementary Fig. S6C). GSA tumor analysis of RAD51AP1 gene indicates high expression of this gene in grade III histologic tumor types (Fig. 5C). To understand the prognostic value of RAD51AP1 in overall survival, we investigated the Kaplan–Meier plotter integrative bioinformatics interface analysis and found a significant correlation between high RAD51AP1 expression and poor disease-free survival (Fig. 5D). To analyze the expression status of RAD51AP1 in human breast cancer cell lines, we investigated RAD51AP1 mRNA expression levels across different breast cancer cell lines using the GOBO gene enrichment cell line application. We found that triple-negative breast cancer (TNBC) cell lines expressed higher levels of this gene than other cell lines (Fig. 5F). Overall, these findings suggest that the combination of 5-AzaC + butyrate targets RAD51AP1, which in turn affects basal stem cell formation and reduces tumor growth.

Figure 5.

RAD51AP1 is upregulated in basal breast cancer. A, heatmap generated from the The Cancer Genome Atlas database showing relative expression of RAD51AP1 gene in normal and different breast cancer subtypes. B, representative box plot of RAD51AP1 gene expression in different breast cancer subtypes. C, box plot represents RAD51AP1 gene expression for tumor samples stratified according to histologic grade. D, Kaplan–Meier plots represent overall survival of breast cancer patients in whole datasets for breast cancer patients categorized according to RAD51AP1 gene expression. The P value was calculated using a log-rank test. E, GSA tumor box plots showing RAD51AP1 gene expression in human breast cancer cell lines categorized according to the basal A (red), basal B (gray), and luminal (blue) subgroups. F, gene expression analysis of RAD51AP1 in 51 breast cancer cell lines.

Figure 5.

RAD51AP1 is upregulated in basal breast cancer. A, heatmap generated from the The Cancer Genome Atlas database showing relative expression of RAD51AP1 gene in normal and different breast cancer subtypes. B, representative box plot of RAD51AP1 gene expression in different breast cancer subtypes. C, box plot represents RAD51AP1 gene expression for tumor samples stratified according to histologic grade. D, Kaplan–Meier plots represent overall survival of breast cancer patients in whole datasets for breast cancer patients categorized according to RAD51AP1 gene expression. The P value was calculated using a log-rank test. E, GSA tumor box plots showing RAD51AP1 gene expression in human breast cancer cell lines categorized according to the basal A (red), basal B (gray), and luminal (blue) subgroups. F, gene expression analysis of RAD51AP1 in 51 breast cancer cell lines.

Close modal

SPC25 expression is high in basal breast cancer

To establish the rationale for targeting genes that are involved in basal stem cell–driven breast cancer, we also investigated the functional significance of SPC25 in human breast cancer and how the expression of this gene responds to 5-AzaC + butyrate treatment (Supplementary Tables S1 and S2). SPC25 is an essential component of kinetochore-associated NDC80 complex (32). It has been well established that kinetochore function is important for chromosome segregation during mitosis and for maintenance of chromosome stability (33). Furthermore, chromosomal instability is one of the major causes of heterogeneity observed in various human cancers (34). We first tested the expression profile of SPC25 between normal and various breast cancer subtypes using UCSC cancer genome browser and GOBO gene enrichment application. We found that SPC25 expression was significantly increased in HER2-positive and basal tumors compared with normal counterparts (Fig. 6A). SPC25 expression was particularly high in basal tumor types, which represents more stem-like cells (Fig. 6B). Similar to RAD51AP1 gene, SPC25 gene is also highly expressed in basal, ER-negative tumors (Fig. 6B and Supplementary Fig. S7A). GSA tumor analysis shows higher histologic grade and poor prognosis with higher expression of SPC25 (Fig. 6C). To understand the prognostic value of SPC25 in overall survival, we used the Kaplan–Meier plotter analysis and found a significant positive correlation between high SPC25 expression and poor disease-free survival (Fig. 6D). Finally, we checked the expression of SPC25 across the human breast cancer cell lines. We found that TNBC cell lines had high level of SPC25 expression (Fig. 6E and F). Although RAD51AP1 and SPC25 expressions were high in human breast tumor tissues and cell lines and their expression downregulated by 5-AzaC and butyrate, we do not have data to show that RAD51AP1 and SPC25 expressions are directly regulated by DNA methylation or histone acetylation. Further studies are needed to address this issue.

Figure 6.

SPC25 is upregulated in breast cancer. A, heatmap generated from the The Cancer Genome Atlas database showing relative expression of SPC25 gene in normal and different breast cancer subtypes. B, box plot represents SPC25 gene expression in different breast cancer subtypes. C, representative box plot of SPC25 gene expression according to histologic grade. D, Kaplan–Meier plots represent overall survival of breast cancer patients in whole datasets for breast cancer patients stratified according to SPC25 gene expression. The P value was calculated using log-rank test. E, box plots represent SPC25 gene expression in human breast cancer cell lines classified according to the basal A (red), basal B (gray), and luminal (blue) subgroups. F, SPC25 gene expression in 51 breast cancer cell lines.

Figure 6.

SPC25 is upregulated in breast cancer. A, heatmap generated from the The Cancer Genome Atlas database showing relative expression of SPC25 gene in normal and different breast cancer subtypes. B, box plot represents SPC25 gene expression in different breast cancer subtypes. C, representative box plot of SPC25 gene expression according to histologic grade. D, Kaplan–Meier plots represent overall survival of breast cancer patients in whole datasets for breast cancer patients stratified according to SPC25 gene expression. The P value was calculated using log-rank test. E, box plots represent SPC25 gene expression in human breast cancer cell lines classified according to the basal A (red), basal B (gray), and luminal (blue) subgroups. F, SPC25 gene expression in 51 breast cancer cell lines.

Close modal

Breast CSCs have been successfully isolated from human breast tumor tissues and from tumors from various spontaneous mouse mammary tumor models. In our studies, we used CD24 and CD49f cell-surface markers mainly because these two markers are commonly expressed in both human and murine cancers and also in various spontaneous mouse mammary tumor models. Although these two markers do not yield pure luminal progenitor and basal stem cell populations, our studies provide evidence that CD49fhighCD24+ cells exclusively express K14 (a marker for basal myoepithelial stem cells) and that CD49f+CD24high cells express K8 (a marker for luminal progenitor cells) without any cross contamination (Fig. 2E). In our recently published studies, we also used CD61 marker to differentiate mature myoepithelial, mature luminal, and luminal progenitor cells (10). However, all these three cell surface markers (CD49f, CD24, and CD61) are unable to distinguish the normal MaSCs and CSCs. Additional studies are warranted to find new cell-surface markers that are capable of distinguishing between MaSCs and CSCs.

Previous studies have shown that a large majority of mammary tumors, including the mammary tumors that develop in MMTV-Neu-Tg mouse (19), are derived from the luminal progenitor cell type (18, 21) rather than from the basal cell type. However, a recent study has shown that mammary tumors that develop in MMTV-Neu-Tg mouse can originate from both luminal and basal cell types (35). Although our present studies reinforce the previous observations, our findings provide further evidence that tumors originating from the basal cell types are relatively slow growing and are responsible for metastatic progression. Furthermore, our studies provide evidence that both luminal and basal cells require additional mutations to activate the unactivated Neu-Tg to become the tumorigenic phenotype (Figs. 2A-C and Supplementary Fig. S2A and S2B). This suggests that a small portion of basal stem cell population may acquire mutations or epigenetic modifications and serve as the cancer-initiating stem cells in Neu-Tg–induced mammary tumor and that such cells are capable of dedifferentiating into luminal cells. This is supported by the observation that tumors derived from the basal cell type express both K8 and K14, suggesting a transition of basal stem cells into early luminal progenitors (Fig. 2E).

One of the most interesting and unexpected findings of our study is that basal stem and luminal progenitor cells isolated from the premalignant mammary gland of younger mice (3-month-old) were able to propagate mammary tumor in NOD/SCID mice while same cell types isolated from older mice (12-month-old) were unable to form tumors. This suggests that, irrespective of the age, genetic/epigenetic events lead to activation of unactivated Neu-Tg into a transformed mammary tumor–forming phenotype along with reprogramming of the stem/progenitor cells into tumor-propagating cancer stem cells. This is surprising because aging is the single biggest risk factor for tumor development; however, our studies provide evidence that genetic/epigenetic modifications reprogram the basal myoepithelial stem cells and luminal progenitor cells into tumor-propagating CSCs and play a critical role in tumor development. Furthermore, tumors derived from both luminal and basal cell types of premalignant MMTV-Neu-Tg mice showed similar self-renewing tumor-propagating CSC phenotype (Fig. 2B–D and Supplementary Fig. S2C). In addition, tumors developed from these two cell types were indistinguishable from the spontaneous tumors that developed in MMTV-Neu-Tg mice, suggesting that basal cell type has similar tumor-propagating capability as luminal cell type and that this tumor-propagating potential of basal cell type could have a great impact in drug resistance in HER2-positive breast cancer patients treated with trastuzumab. Trastuzumab targets only HER2-positive luminal progenitor cells but not HER2-negative basal myoepithelial stem cells, which are multipotent and quiescent in tumor. This raises the possibility that during chemotherapy, especially in trastuzumab therapy, the tumorigenic basal myoepithelial stem cells escape or become unresponsive to the therapy and then enter into circulation and ultimately lead to tumor metastasis. This is further supported by the previous findings that demonstrated the increased basal cell populations in HER2-resistant tumor after chemotherapy (23, 36). Therefore, it is obligatory to target both transformed luminal progenitor cells and basal myoepithelial stem cells when selecting chemotherapy for breast cancer treatment.

Our findings also provide evidence that LinCD49f+CD24+ cells are metastatic. We used 4T1 cell line, a metastatic cancer cell line that mimics grade IV human breast cancer (36), to test the metastatic potential of LinCD49f+CD24+ cells. 4T1 cell line was established from the myoepithelial cells of a spontaneous mutation-driven mouse mammary tumor (37). Previous studies showed that basal myoepithelial cells are leader cells and that the basal myoepithelial enrichment program helps in the process of metastasis (38, 39). These findings suggest that a few basal myoepithelial stem cells dedifferentiate into tumor-propagating luminal cell type. On the basis of these observations, we speculate that, in addition to epithelial-to-mesenchymal transition (EMT), the basal tumorigenic cell types play a key role in drug resistance, tumor recurrence, and metastasis. Furthermore, our findings provide strong evidence that the basal myoepithelial stem cells play a critical role in metastatic spread, either by themselves or by dedifferentiating into tumor-propagating luminal cell type. Thus, 5-AzaC + butyrate combination could be a novel therapeutic strategy to target basal cell gene signature.

The enhanced expression of RAD51AP1 in a majority of breast cancers compared with normal counterparts suggests the oncogenic role of this gene in human breast cancer. Previous studies have shown that RAD51AP1 expression is upregulated in hepatocellular carcinomas (40), acute myeloid leukemia with complex karyotypic abnormalities (41), and aggressive mantle cell lymphoma (42). Furthermore, downregulation of RAD51AP1 by gene-specific siRNA resulted in growth suppression of cholangiocarcinoma cells. These results indicate that RAD51AP1 expression is essential for the growth of tumor cells and that its inhibition may be a novel therapeutic option for the treatment of breast cancer. On the other hand, CpG methylation of its promoter region leads to the reduced expression of RAD51AP1 in prostate cancer cells (43). Therefore, the tumorigenic potential of RAD51AP1 may be cell type–dependent. Similarly, SPC25 is highly expressed in human breast tumor tissues and in breast cancer cell lines when compared with normal controls. Although the role of SPC25 in tumorigenesis is unknown, studies have shown that it forms a complex with NDC80, NUF2, and SPC24, localized at the kinetochore outer plate after the G2 phase (44–47), and plays a critical role in microtubule–kinetochore attachment and spindle assembly checkpoint in mitosis (48). Furthermore, genetic stability depends primarily on accurate chromosome segregation during the cell cycle and kinetochores are essential for this process. Thus, SPC25 function is required for tumor cells to sustain the genomic stability by maintaining mitotic spindle assembly for their eternal growth. Overall, our studies demonstrate that tumor cell of origin and tumor-propagating cells are regulated by DNA methylation and that inhibition of DNMTs reverses the abnormal self-renewal properties of tumor-propagating cells. Furthermore, our studies unequivocally show that the combination of 5-AzaC and butyrate efficiently blocks mammary tumorigenesis and reduces tumorosphere-forming potential of tumor-propagating cells by reactivating the tumor suppressor genes. Therefore, 5-AzaC/butyrate combination could be an effective adjuvant therapy for the treatment of breast cancer. This combination would not only provide an effective strategy to prevent tumor growth but also prevent relapse of the breast cancer.

No potential conflicts of interest were disclosed.

Conception and design: R. Pathania, P.D. Prasad, M. Thangaraju

Development of methodology: R. Pathania, S. Ramachandran, G. Mariappan, P. Thakur, P.D. Prasad, M. Thangaraju

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R. Pathania, S. Ramachandran, J.-H. Choi, P.D. Prasad, M. Thangaraju

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Ramachandran, G. Mariappan, P. Thakur, H. Shi, J.-H. Choi, S. Manicassamy, R. Kolhe, P.D. Prasad, B.L. Lokeshwar, V. Ganapathy, M. Thangaraju

Writing, review, and/or revision of the manuscript: R. Pathania, S. Ramachandran, S. Manicassamy, P.D. Prasad, B.L. Lokeshwar, V. Ganapathy, M. Thangaraju

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): P.D. Prasad, M. Thangaraju

Study supervision: M. Thangaraju

Other (pathologist and member of research advisory committee for PhD): S. Sharma

The authors thank Dr. Chang and Dr. Kitamura, Integrated Genomic core facility, and Augusta University Cancer Center for their help in RNA-Seq analysis.

This work was supported by grants from the NIH 5R01CA131402 (M. Thangaraju) and 5R01CA15677-07 (B.L. Lokeshwar), Department of Defense (BC074289), and Augusta University Intramural Pilot Study grant, Start-up and Bridge funds. G. Mariappan is a fellow of Department of Biotechnology (DBT), India sponsored overseas fellowship program.

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.

1.
Li
X
,
Lewis
MT
,
Huang
J
,
Gutierrez
C
,
Osborne
CK
,
Wu
MF
, et al
Intrinsic resistance of tumorigenic breast cancer cells to chemotherapy
.
J Natl Cancer Inst
2008
;
100
:
672
9
.
2.
Park
CY
,
Tseng
D
,
Weissman
IL
. 
Cancer stem cell-directed therapies: recent data from the laboratory and clinic
.
Mol Ther
2009
;
17
:
219
30
.
3.
Gupta
PB
,
Onder
TT
,
Jiang
G
,
Tao
K
,
Kuperwasser
C
,
Weinberg
RA
, et al
Identification of selective inhibitors of cancer stem cells by high-throughput screening
.
Cell
2009
;
138
:
645
59
.
4.
Curtis
SJ
,
Sinkevicius
KW
,
Li
D
,
Lau
AN
,
Roach
RR
,
Zamponi
R
, et al
Primary tumor genotype is an important determinant in identification of lung cancer propagating cells
.
Cell Stem Cell
2010
;
7
:
127
33
.
5.
Hanahan
D
,
Coussens
LM
. 
Accessories to the crime: functions of cells recruited to the tumor microenvironment
.
Cancer Cell
2012
;
21
:
309
22
.
6.
Reya
T
,
Morrison
SJ
,
Clarke
MF
,
Weissman
IL
. 
Stem cells, cancer, and cancer stem cells
.
Nature
2001
;
414
:
105
11
.
7.
Ting
AH
,
McGarvey
KM
,
Baylin
SB
. 
The cancer epigenome–components and functional correlates
.
Genes Dev
2006
;
20
:
3215
31
.
8.
You
JS
,
Jones
PA
. 
Cancer genetics and epigenetics: two sides of the same coin?
Cancer Cell
2012
;
22
:
9
20
.
9.
Brower
V
. 
Epigenetics: Unravelling the cancer code
.
Nature
2011
;
471
:
S12
3
.
10.
Pathania
R
,
Ramachandran
S
,
Elangovan
S
,
Padia
R
,
Yang
P
,
Cinghu
S
, et al
DNMT1 is essential for mammary and cancer stem cell maintenance and tumorigenesis
.
Nat Commun
2015
;
6
:
6910
.
11.
Feinberg
AP
,
Ohlsson
R
,
Henikoff
S
. 
The epigenetic progenitor origin of human cancer
.
Nat Rev Genet
2006
;
7
:
21
33
.
12.
Esteller
M
. 
Epigenetics in cancer
.
N Engl J Med
2008
;
358
:
1148
59
.
13.
Ramchandani
S
,
Bhattacharya
SK
,
Cervoni
N
,
Szyf
M
. 
DNA methylation is a reversible biological signal
.
Proc Natl Acad Sci U S A
1999
;
96
:
6107
12
.
14.
Flotho
C
,
Claus
R
,
Batz
C
,
Schneider
M
,
Sandrock
I
,
Ihde
S
, et al
The DNA methyltransferase inhibitors azacitidine, decitabine and zebularine exert differential effects on cancer gene expression in acute myeloid leukemia cells
.
Leukemia
2009
;
23
:
1019
28
.
15.
Daskalakis
M
,
Nguyen
TT
,
Nguyen
C
,
Guldberg
P
,
Köhler
G
,
Wijermans
P
, et al
Demethylation of a hypermethylated P15/INK4B gene in patients with myelodysplastic syndrome by 5-Aza-2′-deoxycytidine (decitabine) treatment
.
Blood
2002
;
100
:
2957
64
.
16.
Gros
C
,
Fahy
J
,
Halby
L
,
Dufau
I
,
Erdmann
A
,
Gregoire
JM
, et al
DNA methylation inhibitors in cancer: recent and future approaches
.
Biochimie
2012
;
94
:
2280
96
.
17.
Kelly
TK
,
De Carvalho
DD
,
Jones
PA
. 
Epigenetic modifications as therapeutic targets
.
Nat Biotechnol
2010
;
28
:
1069
78
.
18.
Shackleton
M
,
Quintana
E
,
Fearon
ER
,
Morrison
SJ
. 
Heterogeneity in cancer: cancer stem cells versus clonal evolution
.
Cell
2009
;
138
:
822
9
.
19.
Visvader
JE
. 
Cells of origin in cancer
.
Nature
2011
;
469
:
314
22
.
20.
Jamieson
CH
,
Ailles
LE
,
Dylla
SJ
,
Muijtjens
M
,
Jones
C
,
Zehnder
JL
, et al
Granulocyte-macrophage progenitors as candidate leukemic stem cells in blast-crisis CML
.
New Engl J Med
2004
;
351
:
657
67
.
21.
Guy
CT
,
Webster
MA
,
Schaller
M
,
Parsons
TJ
,
Cardiff
RD
,
Muller
WJ
. 
Expression of the neu protooncogene in the mammary epithelium of transgenic mice induces metastatic disease
.
Proc Natl Acad Sci U S A
1992
;
89
:
10578
82
.
22.
Lim
E
,
Vaillant
F
,
Wu
D
,
Forrest
NC
,
Pal
B
,
Hart
AH
, et al
Aberrant luminal progenitors as the candidate target population for basal tumor development in BRCA1 mutation carriers
.
Nat Med
2009
;
15
:
907
13
.
23.
Lo
PK
,
Kanojia
D
,
Liu
X
,
Singh
UP
,
Berger
FG
,
Wang
Q
, et al
CD49f and CD61 identify Her2/neu-induced mammary tumor-initiating cells that are potentially derived from luminal progenitors and maintained by the integrin-TGFbeta signaling
.
Oncogene
2012
;
31
:
2614
26
.
24.
Oliveras-Ferraros
C
,
Vazquez-Martin
A
,
Martin-Castillo
B
,
Pérez-Martínez
MC
,
Cufí
S
,
Del Barco
S
, et al
Pathway-focused proteomic signatures in HER2-overexpressing breast cancer with a basal-like phenotype: new insights into denovo resistance to trastuzumab (Herceptin)
.
Int J Oncol
2010
;
37
:
669
78
.
25.
Martin-Castillo
B
,
Oliveras-Ferraros
C
,
Vazquez-Martin
A
,
Cufí
S
,
Moreno
JM
,
Corominas-Faja
B
, et al
Basal/HER2 breast carcinomas: integrating molecular taxonomy with cancer stem cell dynamics to predict primary resistance to trastuzumab (Herceptin)
.
Cell Cycle
2013
;
12
:
225
45
.
26.
Santner
SJ
,
Dawson
PJ
,
Tait
L
,
Soule
HD
,
Eliason
J
,
Mohamed
AN
, et al
Malignant MCF10CA1 cell lines derived from premalignant human breast epithelial MCF10AT cells
.
Breast Cancer Res Treat
2001
;
65
:
101
10
.
27.
Turner
N
,
Lambros
MB
,
Horlings
HM
,
Pearson
A
,
Sharpe
R
,
Natrajan
R
, et al
Integrative molecular profiling of triple negative breast cancers identifies amplicon drivers and potential therapeutic targets
.
Oncogene
2010
;
29
:
2013
23
.
28.
Miller
FR
,
Miller
BE
,
Heppner
GH
. 
Characterization of metastatic heterogeneity among subpopulations of a single mouse mammary tumor: heterogeneity in phenotypic stability
.
Invasion Metastasis
1983
;
3
:
22
31
.
29.
Dunlop
MH
,
Dray
E
,
Zhao
W
,
San Filippo
J
,
Tsai
MS
,
Leung
SG
, et al
Mechanistic insights into RAD51-associated protein 1 (RAD51AP1) action in homologous DNA repair
.
J Biol Chem
2012
;
287
:
12343
7
.
30.
Wiese
C
,
Dray
E
,
Groesser
T
,
San Filippo
J
,
Shi
I
,
Collins
DW
, et al
Promotion of homologous recombination and genomic stability by RAD51AP1 via RAD51 recombinase enhancement
.
Mol Cell
2007
;
28
:
482
90
.
31.
Helleday
T
. 
Homologous recombination in cancer development, treatment and development of drug resistance
.
Carcinogenesis
2010
;
31
:
955
60
.
32.
Janke
C
,
Ortiz
J
,
Lechner
J
,
Shevchenko
A
,
Shevchenko
A
,
Magiera
MM
, et al
The budding yeast proteins Spc24p and Spc25p interact with Ndc80p and Nuf2p at the kinetochore and are important for kinetochore clustering and checkpoint control
.
EMBO J
2001
;
20
:
777
91
.
33.
Yuen
KW
,
Montpetit
B
,
Hieter
P
. 
The kinetochore and cancer: what's the connection?
Curr Opin Cell Biol
2005
;
17
:
576
82
.
34.
Ganem
NJ
,
Storchova
Z
,
Pellman
D
. 
Tetraploidy, aneuploidy and cancer
.
Curr Opin Genet Dev
2007
;
17
:
157
62
.
35.
Zhang
W
,
Tan
W
,
Wu
X
,
Poustovoitov
M
,
Strasner
A
,
Li
W
, et al
A NIK-IKKalpha module expands ErbB2-induced tumor-initiating cells by stimulating nuclear export of p27/Kip1
.
Cancer Cell
2013
;
23
:
647
59
.
36.
van de Rijn
M
,
Perou
CM
,
Tibshirani
R
,
Haas
P
,
Kallioniemi
O
,
Kononen
J
, et al
Expression of cytokeratins 17 and 5 identifies a group of breast carcinomas with poor clinical outcome
.
Am J Pathol
2002
;
161
:
1991
6
.
37.
Aslakson
CJ
,
Miller
FR
. 
Selective events in the metastatic process defined by analysis of the sequential dissemination of subpopulations of a mouse mammary tumor
.
Cancer Res
1992
;
52
:
1399
405
.
38.
Cheung
KJ
,
Gabrielson
E
,
Werb
Z
,
Ewald
AJ
. 
Collective invasion in breast cancer requires a conserved basal epithelial program
.
Cell
2013
;
155
:
1639
51
.
39.
Sheridan
C
,
Kishimoto
H
,
Fuchs
RK
,
Mehrotra
S
,
Bhat-Nakshatri
P
,
Turner
CH
, et al
CD44+/CD24- breast cancer cells exhibit enhanced invasive properties: an early step necessary for metastasis
.
Breast Cancer Res
2006
;
8
:
R59
.
40.
Song
LJ
,
Zhang
WJ
,
Chang
ZW
,
Pan
YF
,
Zong
H
,
Fan
QX
, et al
PU.1 Is Identified as a Novel Metastasis Suppressor in Hepatocellular Carcinoma Regulating the miR-615–5p/IGF2 Axis
.
Asian Pacific J Cancer Prevent
2015
;
16
:
3667
71
.
41.
Schoch
C
,
Kern
W
,
Kohlmann
A
,
Hiddemann
W
,
Schnittger
S
,
Haferlach
T
. 
Acute myeloid leukemia with a complex aberrant karyotype is a distinct biological entity characterized by genomic imbalances and a specific gene expression profile
.
Genes Chromosomes Cancer
2005
;
43
:
227
38
.
42.
Henson
SE
,
Tsai
SC
,
Malone
CS
,
Soghomonian
SV
,
Ouyang
Y
,
Wall
R
, et al
Pir51, a Rad51-interacting protein with high expression in aggressive lymphoma, controls mitomycin C sensitivity and prevents chromosomal breaks
.
Mut Res
2006
;
601
:
113
24
.
43.
Wang
Y
,
Yu
Q
,
Cho
AH
,
Rondeau
G
,
Welsh
J
,
Adamson
E
, et al
Survey of differentially methylated promoters in prostate cancer cell lines
.
Neoplasia
2005
;
7
:
748
60
.
44.
Hori
T
,
Haraguchi
T
,
Hiraoka
Y
,
Kimura
H
,
Fukagawa
T
. 
Dynamic behavior of Nuf2-Hec1 complex that localizes to the centrosome and centromere and is essential for mitotic progression in vertebrate cells
.
J Cell Sci
2003
;
116
:
3347
62
.
45.
MacLeod
AR
,
Rouleau
J
,
Szyf
M
. 
Regulation of DNA methylation by the Ras signaling pathway
.
J Biol Chem
1995
;
270
:
11327
37
.
46.
McCleland
ML
,
Kallio
MJ
,
Barrett-Wilt
GA
,
Kestner
CA
,
Shabanowitz
J
,
Hunt
DF
, et al
The vertebrate Ndc80 complex contains Spc24 and Spc25 homologs, which are required to establish and maintain kinetochore-microtubule attachment
.
Curr Biol
2004
;
14
:
131
7
.
47.
DeLuca
JG
,
Dong
Y
,
Hergert
P
,
Strauss
J
,
Hickey
JM
,
Salmon
ED
, et al
Hec1 and nuf2 are core components of the kinetochore outer plate essential for organizing microtubule attachment sites
.
Mol Bio Cell
2005
;
16
:
519
31
.
48.
Sun
SC
,
Lee
SE
,
Xu
YN
,
Kim
NH
. 
Perturbation of Spc25 expression affects meiotic spindle organization, chromosome alignment and spindle assembly checkpoint in mouse oocytes
.
Cell cycle
2010
;
9
:
4552
9
.

Supplementary data