Reduced miR-128 in Breast Tumor–Initiating Cells Induces Chemotherapeutic Resistance via Bmi-1 and ABCC5

This article is featured in Highlights of This Issue, p. 6953

Resistance to chemotherapy remains a major obstacle in effective anticancer treatments, resulting in relapse and progression of most malignant tumors. Recently, accumulating evidence has shown that a subpopulation of cancer cells with stem cell–like features, tumor-initiating cells (T-IC), also named cancer stem cells, exist in a variety of human malignancies, including hematologic (1), brain (2–4), breast (5, 6), prostate (7), liver (8), pancreatic (9), and colon (10, 11) cancers. These cells are insensitive to multiple chemotherapeutic regimens and are thought to be the major source of posttreatment tumor recurrence (12–14). Although a number of molecular mechanisms have been associated with drug resistance in T-ICs, such as upregulated ATP-binding cassette transporters, efficient DNA damage response and enhanced antiapoptosis, none of these mechanisms is a stem cell–like property, and thus, their contributions in chemotherapeutic resistance of T-ICs remain controversial (15).

MicroRNAs (miRNA) are short, 20- to 22-nucleotide RNA molecules that repress the translation of mRNAs by base pairing to partially complementary sites in the 3′-untranslated region (UTR) to inhibit the target protein translation (16, 17). miRNAs regulate the biology of cancer cells, including their sensitivity to chemotherapy (18). Aberrant expression of miRNAs is involved in a number of molecular pathways that are related to the mechanisms of chemotherapeutic resistance. Among them, downregulation of miR-15b or miR-16 results in BCL2 overexpression and the MDR seen in gastric and breast cancer cells (19). In addition, miR-21 contributes to chemotherapy resistance in various types of cancers via inhibiting proapoptotic factors, including PTEN, tissue inhibitor of metalloproteinase, and PDCD4. Ectopic expression of miR-451 or miR-326, both of which target the MDR1 gene in chemotherapeutic resistant breast cancer lines, sensitizes them to doxorubicin and VP-16 (20, 21). Others, such as miR-215 and let-7, affect chemotherapy sensitivity by modulating the cell cycles of cancer cells. Our previous study showed a distinct miRNA profile in breast tumor–initiating cells (BT-IC) versus non–T-ICs and showed that miRNAs play crucial roles in maintaining the self-renewal ability, undifferentiated status, and tumorigenic capacity of BT-ICs that are resistant to chemotherapy. Therefore, screening for differentially expressed miRNAs in T-ICs may help to elucidate the stem cell–like machineries underlying chemotherapeutic resistance in these cells.

In this study, we showed that reduction of miR-128 in BT-ICs leads to overexpression of Bmi-1 and ABCC5, two independent targets of miR-128. Ectopic expression of miR-128 decreases cell viability and increases apoptosis and DNA damage in the presence of doxorubicin, hence sensitizes BT-ICs to chemotherapy.

The cell line SK-3rd used in this study was previously established by consecutively passaging the breast cancer cell line SKBR3 in nonobese diabetic severe-combined immunodeficient mice under the pressure of chemotherapy (22). The breast cancer cell lines MCF-7 and SKBR3 were obtained from the American Type Culture Collection.

Mammosphere culture was done as previously reported (22). Cells (1,000 cells/mL) were cultured in suspension in serum-free Dulbecco's modified Eagle's medium (DMEM)-F12 (BioWhittaker), supplemented with B27 (1:50; Invitrogen), 20 ng/mL EGF (BD Biosciences), 0.4% bovine serum albumin (Sigma), and 4 mg/mL insulin (Sigma).

Flow cytometric cell sorting was done on single-cell suspensions using an Epics Altra flow cytometer (Beckman Coulter) as previously reported (22). The antibodies used for cell sorting were anti-CD44 [fluorescein isothiocyanate (FITC) conjugated; BD Biosciences] and anti-CD24 (PE conjugated; BD Biosciences). To deplete the nontumor cells from primary cancer samples, a cocktail of lineage marker antibodies including CD2, CD3, CD10, CD16, CD18, CD31, CD64, and CD140b was used. The purity of the sorted populations was verified by flow cytometry.

Total RNA was extracted using TRIzol (Invitrogen) and treated with RNase-free DNase (Qiagen). Mature miRNA expression analysis was conducted using a TaqMan MicroRNA Assays (Applied Biosystems). Quantitative reverse transcriptase polymerase chain reaction (qRT-PCR) was done using the Roche LightCycler480 Real-Time PCR System, with human U6 as an endogenous control. All reactions were done in a 25-μL reaction volume in triplicate. Primers for mature miR-128 miRNA and U6 snRNA were obtained from Ambion.

Oligonucleotides encoding short hairpin RNA (shRNA) targeting Bmi-1 or ABCC5 were synthesized according to previously published sequences and cloned under the control of the U6 promoter in the lentiviral vector lentilox pLL3.7 as previously reported (23). Lentivirus vectors were generated as described by cotransfecting pLL3.7 carrying a shRNA expression cassette with helper plasmid pCMV-VSV-G and pHR'8.9ΔVPR in 293T cells using FuGENE 6 (Roche). The viral supernatant was collected 48 hours after transfection, and the viral titers were determined by transducing HeLa cells at serial dilutions and analyzing green fluorescent protein (GFP) expression by flow cytometry. The pMSCV-Bmi-1-del constructs were generated as described previously (24), as presented by Dr. Musheng Zeng. Stable cell lines expressing Bmi-1 were generated by infection of the retroviral vectors expressing the particular gene as described (25). The retroviruses were produced by transient transfection of the retroviral vector, together with the pIK packaging plasmid into the 293T cell line. The lentivirus vectors or retrovirus vector were transducted into SK-3rd cells or MCF-7 cells as previously reported (22) at a multiplicity of infection of 1:5. Transduction efficiency, evaluated by GFP expression, was more than 90%. The cells stably expressing miR-128, Bmi-1 shRNA, or ABCC5 shRNA were obtained and established by cell sorting with flow cytometry.

To evaluate the function of miR-128, a pMIR-REPORTTM luciferase reporter vector with a miR-128 target sequence cloned into its 3′-UTR (luc-mir-128; Ambion) was used. The reporter vector plasmid was transfected using Lipofectamine 2000 according to the manufacturer's instructions. To correct for transfection efficiency, a luciferase reporter vector without a miR-128 target was transfected in parallel. Luciferase activity was assayed using a luciferase assay kit (Promega). MiR-128 function was expressed as the percent reduction in the luciferase activity of cells transfected with the reporter vector containing the miR-128 target sequence compared with cells transfected with the vector without the miR-128 target.

Protein extracts were resolved through 8% to 12% SDS-PAGE; transferred to nitrocellulose membranes; and probed with mouse monoclonal antibody against Bmi-1 (Millipore), ABCC5 (Abcam), or β-actin (Proteintech); probed with peroxidase-conjugated secondary antibody (Proteintech); and subsequently visualized by a Gel Doc XR+ System (Bio-Rad).

A total of 10⁵ cells/mL were plated into 96-well microtiter plates, incubated at 37°C, and cultured overnight. Cells were then treated with increasing concentrations of doxorubicin from 0.5 to 10 μg/mL for 24 hours. A total of 5 mg/mL MTT (Sigma) solution was prepared with PBS, and 20 μL of MTT solution was added to each well. Plates were shaken in the dark for 5 minutes, incubated for an additional 4 hours, and administered 150 μL dimethyl sulfoxide (Sigma) in each well before being read at 550 nm in the microplate reader.

A soft agar colony formation assay was done to investigate the ability of cells to grow unattached to a surface. Cells (1,000 cells/mL) were suspended in 2× mammosphere culture medium (DMEM-F12 (Invitrogen) supplemented with B27 (1:25; Invitrogen), 40 ng/mL EGF (BD Biosciences), 0.8% bovine serum albumin (Sigma), and 8 μg/mL insulin (Sigma), and an equal volume of 0.6% agarose. They were subsequently overlaid onto a solidified bottom layer of 2× mammosphere culture medium and an equal volume of 1.2% agarose. Mammosphere culture medium was added on the top of the top agar once a week. After incubation for 20 days at 37°C, the colonies were counted with the naked eye.

Cells were stained with Hoechst 33342 (Sigma) at 0.1 μg/mL in complete medium for 15 minutes at 37°C. Cell morphology was examined using an inverted fluorescence microscope. Nonapoptotic cells showed dim, nonhomogeneous nuclear staining. Apoptotic cells were brightly stained, and the classic progression of chromatin condensation and nuclear fragmentation was visible. Only whole cells were counted. At least 200 cells were counted for each sample, and the percentage of apoptotic cells was calculated.

A comet assay was done by single-cell gel electrophoresis with the Trevigen's Comet Assay Kit as described in the manufacturer's instructions. Briefly, cell suspensions were embedded in agarose and deposited on microscope slides. The slides were incubated for 2 hours at 4°C in lysis solution, followed by 3 washes in neutralizing buffer (0.4 mol/L Tris-HCl, pH 7.5). Electrophoresis was carried out for 20 minutes at 20 V in 0.5% TBE buffer (pH 8). Slides were then stained with SYBR Green I. Tail DNA content and tail length were analyzed with CASP software.

To evaluate apoptosis, cells were collected and washed twice with PBS while spinning at 1,000 rpm for 10 minutes. Cell pellets were resuspended in a FITC-labeled Annexin V and propidium iodide (PI) staining solution (BD Bioscience) and incubated for 15 minutes at room temperature. The samples were then analyzed on a FACSCalibur (BD Bioscience).

Seventy-seven patients with primary breast cancer (2 cm or larger), who consecutively underwent neoadjuvant chemotherapy containing anthracyclines at the Breast Tumor Department, Sun Yat-Sen Memorial Hospital of Sun Yat-Sen University, were enrolled into this study from January 2004 to June 2011 (Supplementary Table S3). Tumor specimens were obtained by core-needle biopsy before starting therapy. Informed consent for the additional core-needle biopsy and experimental use of tumor samples was obtained from all patients, following a protocol approved by the Ethics Committee of the Sun Yat-Sen Memorial Hospital of Sun Yat-Sen University.

The chemotherapy regimen consisted of four 21-day cycles of AC (doxorubicin: 60 mg/m² on day 1; cyclophosphamide: 600 mg/m² on day 1) followed by four 21-day cycles of docetaxel (100 mg/m² on day 1) or six 21-day cycles of TAC (docetaxel: 75 mg/m² on day 1; doxorubicin: 50 mg/m² on day 1; cyclophosphamide: 500 mg/m² on day 1). Tumor size was evaluated by ultrasound or helical computed tomography scan at baseline, 14 to 21 days after the last cycle of AC, and 21 days after the last cycle of docetaxel treatment. The effect of chemotherapy on the tumor was assessed as the 3-dimensional volume reduction rate or tumor response rate. The tumor response was evaluated by the Response Evaluation Criteria in Solid Tumors (RECIST), which is defined as the following: complete response (CR; disappearance of the disease), partial response (PR; reduction of ≥30%), stable disease (SD; reduction <30% or enlargement ≤20%), or progressive disease (PD; enlargement ≥20%).

Tissue slides were prehybridized in a hybridization solution (Boster Company) at 57°C for 2 hours. Ten picomoles of digoxigenin-labeled miRCURY LNA detection probes (Exiqon) complementary to miR-128a/b and/or scrambled microRNA were added and hybridized overnight (16 hours) at a temperature of 21°C below the calculated melting temperature of the LNA probe. After stringent washes, an immunologic reaction was carried out using a mouse monoclonal antibody to digoxigenin (Abcam) and with alkaline phosphatase streptavidin (Zhongshan Golden Bridge Biotechnology Company) to detect biotinylated probes. Slides were mounted with aqueous mounting medium (Maixin Biotechnology Company).

Immunohistochemical analysis was done as reported (22). The paraffin sections were incubated with primary antibody against Bmi-1 (1:100; Millipore), ABCC5 (1:20; Abcam), or Ki67 (1:200; Abcam), respectively. For negative control, isotype-matched antibodies were applied. Immunohistochemical scoring was done without prior knowledge of the clinical response. Ki67 cell proliferation index was determined by counting the number of Ki67-positive cells per 1,000 tumor cells in 10 randomly selected fields (*400) for each section containing the representative tumor sections. Tissue sections were observed under a ZEISS AX10-Imager A1, and all pictures were captured using AxioVision 4.7 microscopy software.

DNA fragmentation in tumor tissues from breast cancer patients was analyzed by following the instructions of the manufacturer of the In Situ Cell Death Detection Kit, POD (Roche, Switzerland). Under the light microscope, apoptotic cells showed brownish staining in the nuclei. The TdT-mediated dUTP-X nick end labeling analysis (TUNEL) apoptosis index was determined by counting the number of TUNEL-positive cells per 1,000 tumor cells in 10 fields (*400) for each section. Areas with extensive necrosis were avoided.

All in vitro experiments were done either in triplicate or in pentuplicate. The results are described as mean ± SD. Statistical analysis was done by 1-way ANOVA, and comparisons among groups were done by the independent sample t test or Bonferroni multiple comparison t test. Survival curves were analyzed by the Kaplan–Meier method and compared by the log-rank test, using GraphPad Prism.

In our previous study (22), which used microarrays to compare the miRNA expression profile in mammospheric SK-3rd with their in vitro–differentiated progeny, we showed a significant reduction in miR-128 expression in BT-ICs. Using qRT-PCR, here we verified that miR-128, which was highly expressed in the parental SKBR3 breast cancer cells, was significantly reduced in mammospheric SK-3rd cells (Fig. 1A, P < 0.01). However, after differentiation, by attaching to the culture plates precoated with collagen IV for 48 hours, miR-128 expression dramatically increased more than 10-fold in the adherent SK-3rd cells and was not significantly different from that of SKBR3 (Fig. 2A, P > 0.05). Similarly, we found that miR-128 expression in mammospheric MCF-7 was reduced to about one-third of that in the adherent MCF-7 cells (Fig. 1A, P < 0.001). To confirm the physiologic relevance of the above results, we examined miR-128 expression using qRT-PCR in the BT-ICs isolated from primary breast cancers. Primary BT-ICs enriched by mammosphere cultures (6) or by sorting for CD44⁺CD24⁻ cells (5) were reduced in miR-128 when compared with either adherently cultured primary cells or tumor cells that were not CD44⁺CD24⁻ (Fig. 1A, P < 0.001).

Transduction of the mammospheric SK-3rd cells with a lentivirus vector carrying miR-128 expression cassette restored the miRNA expression to a level comparable with adherent differentiated SK-3rd cells (Fig. 1B, P < 0.01). Similarly, lenti-mir-128 elevated the level of miR-128 about 8-fold in mammospheric MCF-7 cells, compared with untreated mammospheric MCF-7 cells or those transduced with empty vector (Fig. 1B, P < 0.01).

To further investigate the targeting function of miR-128, we transfected a luciferase reporter vector containing a miR-128 target sequence cloned into its 3′-UTR (luc-mir-128) into mammospheric SK-3rd and differentiated SK-3rd cells and evaluated the percentage of luciferase suppression relative to the cells transfected with a control luciferase reporter vector without a miR-128 target sequence. Luciferase activity was suppressed by 70% in the differentiated SK-3rd cells, whereas there was almost no suppression in the mammospheric SK-3rd cells (Fig. 1C, P < 0.001). Infection of mammospheric SK-3rd cells with lenti-mir-128 restored the suppression of luciferase activity by 63.7% (Fig. 1C, P < 0.001). Cotransfecting the cells with both the reporter vector and the miR-128 ASO significantly weakened the suppression of luciferase activity in adherent SK-3^rd, with abundant endogenous miR-128 and in mammospheric SK-3rd with ectopic miR-128 expression (Fig. 1C, P < 0.001). Analogously, mammospheric MCF-7 cells had almost no change in luciferase activity, whereas in the differentiated MCF-7 cells, luciferase activity was suppressed by 75% (Fig. 1C, P < 0.001). Overexpression of miR-128 in mammospheric MCF-7 by lentivirus restored the suppression of luciferase activity by approximately 68% (Fig. 1C, P < 0.001). Taken together, these data indicate that not only miR-128 expression but also its targeting function was reduced in BT-ICs.

Bmi-1, which is a polycomb ring finger oncogene, has been recently reported as a target gene for miR-128 in glioma cells (26). In addition, with the computer-based sequence analysis software TargetScan, we found that the seed sequence of miR-128 had a perfect match with the 3′-UTR of the mRNA of ATP-binding cassette sub-family C member 5 (ABCC5), also named MDR–associated protein 5 (MRP5; Supplementary Table S2). The targeting function of miR-128 to the 3′-UTRs of Bmi-1 and ABCC5 mRNA was examined using luciferase reporter constructs containing the miR-128 recognition site or a mutated sequence at this site cloned from the 3′-UTR of Bmi-1 mRNA or ABCC5 mRNA immediately downstream of the luciferase gene. As shown in Fig. 2A, transfection of 293T cells with a miR-128 mimic suppressed the luciferase activity of the reporter vectors containing the 3′-UTR of wild-type Bmi-1 or wild-type ABCC5. In addition, miR-128 inhibitors increased the luciferase activity of the reporter vectors containing the 3′-UTR of wild-type Bmi-1 or wild-type ABCC5 in SK-3rd and MCF-7 cells (Supplementary Fig. S1A and B). Therefore, miR-128 regulates Bmi-1 and ABCC5 expression by directly targeting their 3′-UTRs. Immunoblotting with anti–Bmi-1 and anti-ABCC5 antibodies further showed that both the Bmi-1 and ABCC5 proteins were significantly elevated in mammospheric cells obtained from the breast cancer cell lines and primary cells isolated from biopsies of breast cancer patients and established for growth in vitro, compared with the corresponding differentiated cells (Fig. 2B; Supplementary Fig. S1C). However, Bmi-1 and ABCC5 mRNAs measured by qRT-PCR did not differ significantly among the 2 cell lines and the primary breast cancer cells (not shown). Therefore, miR-128 silences Bmi-1 and ABCC5 expression by inhibiting translation.

To further examine whether miR-128 affects Bmi-1 and ABCC5 expression, we infected the mammospheric SK-3rd cells and mammospheric MCF-7 cells with lenti-mir-128 and monitored the protein levels of Bmi-1 and ABCC5 using Western blotting. Bmi-1 and ABCC5 proteins in the mammospheric cells were significantly knocked down. The effect was specific, as Bmi-1 and ABCC5 proteins in the untransduced cells (un) or the cells transduced with an empty vector (nc) remained unchanged (Fig. 2C and D; Supplementary Fig. S1D). Collectively, our data indicate that miR-128 reduction in BT-ICs may result in Bmi-1 and ABCC5 protein elevation by relieving translational repression at the 3′-UTRs of the target genes.

T-ICs with stem cell–like features are more resistant to chemotherapy than are differentiated tumor cells (27, 28). Thus, we further investigated whether miR-128 downregulation in BT-ICs may contribute to doxorubicin resistance. Without doxorubicin, the cell viability of SK-3rd BT-ICs or MCF-7 BT-ICs, as determined by MTT assays, was not significantly changed upon transduction with lenti-miR-128, lenti-Bmi-1-shRNA, lenti-ABCC5-shRNA, or empty lentivirus (Fig. 3A), suggesting that the dysregulation of miR-128, Bmi-1, and ABCC5 is not required for BT-IC survival. However, when SK-3rd BT-ICs or MCF-7 BT-ICs were treated with increasing concentrations of doxorubicin from 0.5 to 10 μg/mL (Fig. 3A), the proportion of viable cells upon transduction with lenti-miR-128 (IC₅₀ = 0.32 μg/mL, P < 0.01), lenti-Bmi-1-shRNA (IC₅₀ = 0.36 μg/mL, P < 0.01), or lenti-ABCC5-shRNA (IC₅₀ = 0.43 μg/mL, P < 0.01) or cotransduction with lenti-Bmi-1-shRNA and lenti-ABCC5-shRNA (IC₅₀ = 0.31 μg/mL, P < 0.01), but not with lentivirus vector alone (IC₅₀ = 1.68 μg/mL, P > 0.05), decreased more rapidly than was observed in the untransduced cells (IC₅₀ = 1.72 μg/mL). At each concentration, doxorubicin induced the highest growth inhibition in BT-ICs with the transduction of lenti-miR-128 or lenti-Bmi-1/ABCC5-shRNAs. Furthermore, infection with lenti-Bmi-1-shRNA and lenti-ABCC5-shRNA synergistically reduced the percentage of viable cells in mammospheric SK-3rd or mammospheric MCF-7 cells under treatment with doxorubicin. On the other hand, restoring Bmi-1 expression by transfection with pMSCV-Bmi-1-del in SK-3rd BT-ICs significantly increased cell viability under the pressure of doxorubicin at the concentrations of 4 μg/mL, 8 μg/mL, and 10 μg/mL (P < 0.01 vs. untransfected control; Supplementary Fig. S2). However, cotransfection of miR-128 mimics and pMSCV-Bmi-1-del in SK-3rd BT-ICs did not elevate cell viability under treatment with doxorubicin (P > 0.05; Supplementary Fig. S2).

Moreover, we investigated the contribution of miR-128 to doxorubicin resistance in BT-ICs isolated from primary breast tumors using mammosphere cultures. When cells were treated with increasing concentrations of doxorubicin from 0.5 to 10 μg/mL, the viability of the adherent differentiated breast cancer cells (IC₅₀ = 0.37 μg/mL) was much lower than that of the untransduced mammospheric breast cancer cells (IC₅₀ = 2.366 μg/mL, P < 0.01) at each concentration. Transduction with lenti-miR-128 (IC₅₀ = 0.45 μg/mL, P < 0.01), but not with the lentivirus vector alone (IC₅₀ = 1.95 μg/mL, P > 0.05), decreased the viability of the primary BT-ICs more than in the untransduced cells (IC₅₀ = 2.366 μg/mL) at each concentration (Supplementary Fig. S3).

To further evaluate the anchorage-independent growth of transduced BT-ICs after treatment with doxorubicin, we employed a soft agar colony formation assay. In the presence of doxorubicin at 1 μg/mL, colony formation in soft agar was reduced approximately 3.4-fold in the SK-3rd BT-ICs with lenti-miR-128 transduction (P < 0.001, Fig. 3B) and 2.3-fold in lenti-miR-128–transduced MCF-7 BT-ICs, but not in the untransduced cells or those transduced with an empty lentivirus vector (P > 0.05). Simultaneous or alternative transduction with lenti-Bmi-1-shRNA and lenti-ABCC5-shRNA significantly inhibited soft agar colony formation in SK-3rd BT-ICs or MCF-7 BT-ICs (P < 0.001, Fig. 3B).

We next examined doxorubicin-induced apoptosis using Hoechst 33342 and Annexin V/PI staining in all transduced BT-ICs. Without chemotherapy, ectopic expression of miR-128 increased the proportion of Hoechst-positive cells approximately 3-fold. In addition, knockdown of Bmi-1 or ABCC5 enhanced the proportion of Hoechst-positive cells 2.3-fold. Simultaneous silencing of Bmi-1 and ABCC5 increased the percentage of Hoechst-positive cells to a level that was not significantly different from that under ectopic expression of miR-128 (Fig. 4A, P < 0.05, vs. untransfected cells). However, treatment with 1 μg/mL doxorubicin dramatically increased the proportion of Hoechst-positive BT-ICs transduced with lenti-miR-128 approximately 10-fold (Fig. 4A, P < 0.001 vs. untreated cells), whereas it did not influence that of the untransduced cells or the cells transduced with an empty lentivirus vector (Fig. 4A, P > 0.05 vs. untreated cells). In agreement with these data, simultaneous or alternative transduction with lenti-Bmi-1-shRNA and lenti-ABCC5-shRNA, but not with an empty lentivirus vector, increased the proportion of Hoechst-positive BT-ICs that were treated with 1 μg/mL doxorubicin approximately 9-fold (Fig. 4A, P < 0.001). In parallel, in the presence of doxorubicin at 1 μg/mL, transduction of the SK-3rd BT-ICs with lenti-mir-128, lenti-Bmi-1-shRNA, lenti-ABCC5-shRNA, or lenti-Bmi-1/ABCC5-shRNAs increased the percentage of Annexin V⁺PI⁺/Annexin V⁺ PI⁻ cells (50.4%, 33.5%, 31.2%, or 54.8%, respectively) about 6- to 8-fold as compared with the untransfected cells or the cells transfected with an empty vector (P < 0.001, Fig. 4D). However, in the absence of chemotherapy, transduction of the SK-3rd BT-ICs with lenti-miR-128, lenti-Bmi-1-shRNA, lenti-ABCC5-shRNA, or lenti-Bmi-1/ABCC5-shRNAs increased the percentage of apoptotic and necrotic cells by only 6.5%, 5.6% 8.4%, or 12.1%, respectively (P < 0.05, Fig. 4D). Similarly, in the presence of doxorubicin at 1 μg/mL, transduction of the MCF-7 BT-ICs with lenti-mir-128, lenti-Bmi-1 shRNA, lenti-ABCC5-shRNA, or lenti-Bmi-1/ABCC5-shRNAs increased the percentage of Annexin V⁺ cells about 7- to 9-fold as compared with the untransfected cells or the cells transfected with an empty vector (P < 0.001, Fig. 4D). However, in the absence of chemotherapy, the percentage of apoptotic and necrotic cells in MCF-7 BT-ICs transduced by lenti-miR-128, lenti-Bmi-1-shRNA, lenti-ABCC5-shRNA, or lenti-Bmi-1/ABCC5-shRNAs was increased by 4.3%, 5.2%, 8.1%, or 11.9%, respectively (P < 0.05, Fig. 4D). Therefore, miR-128 reduction conferred resistance to chemotherapy-induced apoptosis in T-ICs.

Because doxorubicin is a cytotoxic agent causing DNA damage in cancer cells, we further evaluated whether miR-128 and its targets Bmi-1 and ABCC5 are involved in doxorubicin-induced DNA double-strand breaks in SK-3rd BT-ICs and MCF-7 BT-ICs, as measured by comet assay, a method that quantifies DNA damage by the increasing lengths of comet tails and DNA content of the tails. Without chemotherapy, transduction with lenti-mir-128, lenti-Bmi-1-shRNA, or lenti-ABCC5-shRNA or simultaneous transduction with lenti-Bmi-1-shRNA and lenti-ABCC5-shRNA slightly increased the length of the comet tails and the DNA content of the tails, as compared with the untransduced cells or cells transduced with empty vector alone (P < 0.05, Fig. 4B and C). However, 24 hours after doxorubicin treatment, cells transduced with lenti-mir-128, lenti-Bmi-1-shRNA, lenti-ABCC5-shRNA, or lenti-Bmi-1shRNA/ABCC5-shRNA exhibited much longer comet tails and tails with much higher DNA content (P < 0.001, Fig. 4B and C). These results show that miR-128 deficiency and Bmi-1 or ABCC5 overexpression retarded the process of ADM-induced DNA-double strand breaks. Taken together, these data suggest that miR-128 reduction confers chemotherapeutic resistance in BT-ICs via the elevation of Bmi-1 and ABCC5 expression.

To investigate whether miR-128 expression relates to chemotherapy resistance in breast cancer patients, we enrolled 57 breast cancer patients who underwent 6 or 8 cycles of neoadjuvant chemotherapy, including an anthracycline anticancer regimen, and who exhibited CR or PR, defined as chemosensitive, or PD, defined as chemoresistant according to RECIST criteria. We employed in situ hybridization with a digoxigenin-labeled miR-128 probe to examine miR-128 expression in the breast cancer tissue biopsies prior to neoadjuvant chemotherapy and correlated it with the clinical response of the breast tumors to chemotherapy. As shown in Fig. 5A, miR-128 expression was reduced in the chemoresistant breast cancer tissues but upregulated in the chemosensitive tissues. Because CR indicates chemotherapeutic sensitivity of both bulk and stem cell populations, whereas PR does not necessarily involve the elimination of the stem cell population, we further investigated whether miR-128 expression relates to CR in breast cancer patients using real-time RT-PCR with a miR-128–specific primer amplifying both miR-128a and miR-128b. We found that miR-128 expression relative to U6 in CR breast cancer tissues was the highest among the different chemoresponsive breast cancer tissues (Fig. 5B, P < 0.001; Supplementary Table S4A). Because Bmi-1 and ABCC5 are targets of miR-128, we further investigated the expression of Bmi-1 and ABCC5 in breast cancer tissues using immunohistochemistry. In contrast to miR-128, Bmi-1 and ABCC5 expression was higher in the chemoresistant than in the chemosensitive breast tumor tissues (Fig. 5A).

The relationship between miR-128 expression and chemotherapeutic response was further confirmed by examining Ki67 cell proliferation index and TUNEL apoptosis index in paired specimens of another 20 patients obtained by core-needle biopsy prior to neoadjuvant chemotherapy and at surgery following chemotherapy. In agreement with the results of the clinical response, the Ki67 expression and Ki67 index in breast tumors with high miR-128 expression were markedly decreased after chemotherapy, whereas breast tumors with low miR-128 expression showed stable Ki67 expression and Ki67 index (Fig 5C; Supplementary Fig. S4 and Table S5). However, breast tumor tissues with high miR-128 showed an elevated TUNEL index after chemotherapy, whereas breast tumors with low miR-128 expression showed a stable TUNEL index (Fig. 5C; Supplementary Table S5). With the increase in the miR-128 level, changes in the Ki67 index or TUNEL index gradually increased after chemotherapy (Supplementary Fig. S4).

Furthermore, Kaplan–Meier survival curves showed that patients with high miR-128 expression survived significantly longer than those with low miR-128 expression. The median survival for patients with low miR-128 expression was 54 months, whereas the survival rate at 59 months for patients with high miR-128 expression was 100% (Fig. 5D; P < 0.05; Supplementary Table S4B).

In this study, we showed that miR-128 was reduced in BT-ICs, as isolated by mammosphere culturing from 2 breast cancer cell lines and primary BT-ICs from breast cancer patients, which further elevated the protein levels of its 2 targets Bmi-1 and ABCC5. The ectopic expression of miR-128 in BT-ICs reduced both Bmi-1 and ABCC5 expression, while sensitizing BT-ICs to doxorubicin. We also showed that the reduction of miR-128 in invasive breast tumor tissues correlated with poor clinical response to chemotherapy as well as poor patient survival. Therefore, miR-128 reduction in BT-ICs contributes to their chemotherapy resistance by alleviating its repression of Bmi-1 and ABCC5 translation.

Mir-128 has been described as a tumor suppressor, and a reduced level of miR-128 was first identified in glioblastoma (29). Aberrant expression of miR-128 contributes to the malignant phenotypes of cancer cells, such as proliferation (30), cell motility, invasion (31, 32), apoptosis (33), and self-renewal (26). Here, we extended the current knowledge by highlighting the role of miR-128 in the chemotherapy resistance of BT-ICs. Although chemotherapy is the backbone of systemic treatment for most malignancies, its efficacy is hindered by the development of drug resistance, especially in the subpopulation of cancer cells with stem cell–like properties. Therefore, targeting the mechanisms involved in the chemotherapy resistance of T-ICs will help to improve the efficacy of the treatments, with miR-128 being a candidate molecular target for such a purpose.

A crucial characteristic of T-ICs associated with clinical outcome is their resistance to the cytotoxic agents in the microenvironment. We observed that the proapoptotic effect of ectopic miR-128 expression was higher than that from silencing Bmi-1 or ABCC5, as seen in Fig. 4. However, its proapoptotic effect was not definitely higher than when Bmi-1 and ABCC5 were cosilenced. That may be because miR-128 acts as a master regulator of chemoresistance of BT-ICs, presumably by silencing multiple targets, some of which remain to be identified. Moreover, Bmi-1 or ABCC5 may be modulated by a regulator(s) other than miR-128. A number of molecular mechanisms about chemoresistance have been proposed, including the upregulation of multidrug transporters from the ABC superfamily, active DNA damage repair mechanisms, and resistance to apoptosis (34). For example, CD133⁺ lung T-ICs, which display stem cell–like features, were resistant to cisplatin treatment by increased ABCG2 and CXCR4 expression (35). ABCB5 expression in liver cancer stem cells is associated with chemoresistance and reduced survival times of patients with hepatocellular carcinoma (36). These are all similar to our finding that ABCC5 contributes to the chemoresistance of BT-ICs. Glioma stem cells are resistant to radiation via the preferential activation of the DNA damage checkpoint response (37). Other pathways, such as gene mutation (38), aberrant DNA methylation, and histone modification (39), which all play important roles in the resistance of cancer cells to chemotherapeutic agents, may also be involved.

In contrast to other molecular mechanisms involved in the chemotherapy resistance of T-ICs, the downregulation of miR-128 is related to the stem cell–like properties of these cells. Our results confirm that miR-128 directly targets the 3′-UTR of Bmi-1, which is in agreement with a recent study (26). Moreover, the Bmi-1⁺ subpopulation isolated from malignant pleural mesothelioma displays prominent stem cell–like properties and is resistant to cisplatin and pemetrexed (27). Taken together, these previous and current findings indicate that Bmi-1 overexpression is a stem cell–like feature underlying chemotherapy resistance in these cells, which provides a direct link between stem cell–like mechanisms and chemotherapy resistance in T-ICs.

In conclusion, our data show that reduced miR-128 expression plays important roles in the resistance to doxorubicin in BT-ICs via upregulating Bmi-1 and ABCC5. The miR-128/Bmi-1/ABCC5 axis provides a new avenue toward understanding the mechanism of chemoresistance and may help in the development of potential therapeutics against breast cancer.

No potential conflicts of interest were disclosed.

This work was supported by 973 Projects from the Ministry of Science and Technology of China (2010CB912800, 2011CB504203, and 2009CB521706), Natural Science Foundation of China grants (30921140312, 30831160515, 30830110, 30772550, 30671930, 30972785, 30973396, 30973505, and 30801376), a grant for Development of Important New Drugs from the Ministry of Health of China (2011ZX09102-010-02), the Natural Science Foundation of Guangdong Province (8251008901000011 and 9451008901002467), a Young Teacher grant from the HuoYingdong Educational Foundation (121042), the Science and Technology Foundation of the Guangdong Province (2009B030801005), and the Foundation of Guangzhou Science and Technology Bureau (2009Y-C011-1).

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.