Abstract

Triple negative breast cancer (TNBC) is known to contain a high percentage of CD44+/CD24−/low cancer stem cells (CSCs), corresponding with a poor prognosis despite systemic chemotherapy. Chloroquine (CQ), an antimalarial drug, is a lysotropic reagent which inhibits autophagy. CQ was identified as a potential CSC inhibitor through in silico gene expression signature analysis of the CD44+/CD24−/low CSC population. Autophagy plays a critical role in adaptation to stress conditions in cancer cells, and is related with drug resistance and CSC maintenance. Thus, the objectives of this study were to examine the potential enhanced efficacy arising from addition of CQ to standard chemotherapy (paclitaxel) in TNBC and to identify the mechanism by which CQ eliminates CSCs in TNBCs. Herein, we report that CQ sensitizes TNBC cells to paclitaxel through inhibition of autophagy and reduces the CD44+/CD24−/low CSC population in both preclinical and clinical settings. Also, we are the first to report a mechanism by which CQ regulates the CSCs in TNBC through inhibition of the Janus-activated kinase 2 (Jak2)—signal transducer and activator of transcription 3 signaling pathway by reducing the expression of Jak2 and DNA methyltransferase 1. Stem Cells  2014;32:2309–2323

Introduction

Triple negative breast cancer (TNBC) is defined by the absence of expression of estrogen receptor, progesterone receptor, and human epidermal growth factor receptor 2 (Her2). It is an aggressive and lethal form of breast cancer with relative lack of targeted therapeutic options and poor prognosis [1, 2]. It constitutes roughly 15% of breast cancer cases and accounts for 25% of breast cancer mortality principally due to early relapse and frequent metastasis [1, 2]. In addition, TNBC has a higher percentage of CD44+/CD24−/low cancer stem cells (CSCs) compared to other types of breast cancer [1, 3]. We have previously reported an enrichment of CD44+/CD24−/low CSCs following chemotherapy in women with locally advanced breast cancer, suggesting resistance of CSCs to conventional therapy [4]. As chemotherapy is the only choice of treatment for TNBC patients, a viable strategy is needed to target CSCs in addition to conventional chemotherapy [5, 6].

Autophagy constitutes a salvage pathway for recycling nutrients that has been implicated in various aspects of cancer, including cell survival under stress conditions, drug resistance, and metastasis [7, 8]. Moreover, autophagy-mediated metabolic coupling between cancer cells and neighboring stromal cells has been identified as a mechanism of cancer survival, growth, and resistance to therapy [9, 10]. Additionally, recent reports emphasize critical roles of autophagy in maintenance of CSC self-renewal in breast cancer [11, 12], leading to the exploration of potential antiautophagy strategies for eliminating CSCs in preclinical and clinical studies.

Chloroquine (CQ) is an antimalarial drug known to inhibit autophagy by disrupting lysosomal stability and function [7]. Previously, we identified a CD44+/CD24−/low mammosphere (MS)-forming treatment-resistant gene expression signature using biopsies obtained from women with primary breast cancer [4]. Through network analysis of the gene signature, CQ was repositioned as a putative anticancer drug against CD44+/CD24−/low CSCs. Interestingly, CQ has been identified as a CSC targeting agent for other aggressive cancers including breast cancer [11, 12], glioblastoma multiforme [13], and chronic myeloid leukemia [14]. However, the mechanism by which CQ affects the CD44+/CD24−/low CSCs remains unclear.

We investigated the therapeutic potential of CQ in combination with paclitaxel (PTX) on the CD44+/CD24−/low CSC population, and determined the value and feasibility of incorporating CQ with chemotherapy for treatment of therapy-resistant TNBC. We hypothesized that CQ affects the CSC self-renewal through the inhibition of autophagy. Our findings suggest that CQ reduces the CD44+/CD24−/low CSCs population in TNBC cells through autophagy and by downregulation of Janus-activated kinase 2 (Jak2) signaling pathway with a concomitant inhibition of DNA methyltransferase 1 (DNMT1) expression.

Materials  and Methods

Materials and Cell Culture

TNBC cell lines (Hs578t, MDA-MB-231, HCC1937, and HCC38) were purchased from American Type Culture Collection (Manassas, VA), with the exception of SUM159PT (Asterand, Detroit, MI). All cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM) (Invitrogen, Grand Island, NY) and 10% fetal bovine serum (FBS) (Thermos Scientific Hyclone, Rockford, IL) in a humidified 5% CO2 incubator at 37°C. SUM159PT cells were first maintained in F-12 (Invitrogen) containing 10% FBS, insulin (Invitrogen) (5 µg/ml), and hydrocortisone (StemCell Technologies Inc., Vancouver, Canada) (1 µg/ml), then adjusted to DMEM (high glucose and glutamine) with 10% FBS. All chemicals were purchased from Sigma unless otherwise specified. CQ was first dissolved in Dulbecco’s phosphate-buffered salines with no calcium and magnesium (DPBS) (Invitrogen) at the concentration of 0.1 M (kept in −80°C) and diluted further in DPBS (CQ 1 mM). All CD marker antibodies and mouse IgG isotype antibodies were purchased from BD Biosciences, San Jose, CA. Rabbit polyclonal anti-p-Jak2, rabbit monoclonal anti-Jak2, rabbit polyclonal anti-pSTAT3–705, rabbit polyclonal anti-pSTAT3–727, mouse monoclonal signal transducer and activator of transcription 3 (STAT3), and mouse monoclonal anti-Actin antibodies were purchased from Cell Signaling Technology, Danvers, MA. Mouse monoclonal anti-DNMT1, rabbit polyclonal anti-SOCS1, and mouse monoclonal anti-SOCS3 were purchased from Santa Cruz Biotechnology, Inc., Dallas, TX. SYTOX blue nucleic acid stain (SYTOX-Blue) was purchased from Invitrogen for nuclear staining of dead cells.

In Silico Drug Repositioning for Breast CSCs

Our previously published gene expression data of breast CSCs (CD44+/CD24−/low and MS-forming treatment-resistant cells) was used for in silico drug repositioning analysis (GSE7513, SE7515, and GSE10281) [4]. The cancer signaling bridges (CSBs)-based drug repositioning computational modeling method was applied to derive specific CSCs signaling pathways [15, 16].

MS Assay

MS assay was performed as previously described with minor modification [4, 17]. Modified methods are described in Supporting Information Materials and Methods.

Fluorescence-Activated Cell Sorting Analysis

Cell lines and clinical samples were stained with antibodies against CD44-APC and CD24-FITC for fluorescence-activated cell sorting (FACS) analysis and cell sorting as previously described [17]. A single-arm, phase two clinical trial (NCT01446016) is currently active and enrolling patients at our institution. Patients with metastasis or locally advanced breast cancer previously treated with anthracyclines underwent treatment with a combination of taxane and CQ. Biopsies were then obtained at baseline and at day 42 after treatment. FACS analysis and sorting were performed at the Houston Methodist Hospital Research Institute flow cytometry core using BD FACS Fortessa (BD Biosciences, San Jose, CA) for FACS analysis of CSCs and BD FACS Aria II (BD Biosciences) for cell sorting.

Western Blot and Immunoprecipitation Assays

Western blotting and immunoprecipitation experiments were performed with the listed primary and matching secondary antibodies as described previously [18]. Detailed procedures are described in Supporting Information Materials and Methods.

In Vivo Experiments

All animal procedures were approved by the Methodist Hospital Research Institute Animal Care and Use Review Office. Athymic nude mice (Hsd:Athymic Nude-Foxn1nu) (5 weeks old; 20–23 g) were purchased from Harlan Laboratories, Inc., Houston, TX. Detailed methods are described in Supporting Information Materials and Methods.

Immunofluorescence Staining for the Colocalization of Jak2 and SOCS3

Cells were fixed and stained using antibodies listed in Supporting Information Materials and Methods as described previously [18].

Real-Time PCR for SOCS1 and SOCS3

Real-time PCR for SOCS1 and SOCS3 was performed as described previously [17] with minor modifications. Detailed methods are described in Supporting Information Materials and Methods.

SOCS3 Promoter PCR for Methylation Analysis

For the PCR primer design, sequences of proximal SOCS3 promoter regions (−5,676 and +2,633) was obtained from the NCBI reference sequence (NC_000017.10 GI:224589808) for Homo sapiens chromosome 17, GRCh37.p13 Primary Assembly. Primers were then designed using primer3 [19] to result in approximately 200–250-bp of PCR products. The sequences and the site of each primer are indicated in Supporting Information Table S1.

Methyl-CpG-Binding Domain Protein-Enriched DNA Sequencing Assay and Data Analysis

Methylated DNA from control and CQ-treated MDA-MB-231 cells was eluted using the MethylMiner Methylated DNA Enrichment Kit (Life Technologies) following the manufacturer’s instructions as described below. Genomic DNA was sonicated to ∼300-bp fragments. Methylated DNA was captured by methyl-CpG-binding domain proteins and subsequently eluted in 1 M salt buffer for precipitation. Libraries were generated from eluted DNA (>10 ng) for single-end 50-bp sequencing following the protocols from Illumina (San Diego, CA). Methyl-CpG-binding domain protein-enriched DNA sequencing (MBDCap-seq) libraries were sequenced using the Illumina HiSeq 2000 system protocols. Image analysis and base calling were performed with the standard Illumina pipeline. Using the ELAND algorithm, unique reads (up to 50 bp reads) were mapped to the human reference genome (hg19) with Bowtie version 0.12.7 [20] with reported parameters [21]. Further analysis of the MBDCap-seq data was performed by the Houston Methodist Research Institute Genomics Core as described in Supporting Information Materials and Methods.

Statistical Analysis

We used two-tailed Student’s t test for comparison of two groups and one-way ANOVA for multiple group comparison. Two-way ANOVA was used for all animal experiments. Each value reported represents the mean of at least three replicate experiments with SD. The values in the animal experiments represent the mean of 10 individual mice per group with SEM. Data were tested for normal distribution, and Student’s t test and ANOVA were used to determine statistical significance. To account for multiple comparisons, Tukey’s multiple comparison tests for one-way ANOVA and Bonferroni post-tests for two-way ANOVA were performed with Graphpad Prism 5.0 (Graphpad Software, Inc., La Jolla, CA). In all cases, p values < .05 were considered statistically significant.

All other materials and methods are described in Supporting Information Materials and Methods.

Results

In Silico Drug Repositioning

Differentially expressed genes (p < .001, Student’s t test) from the CD44+/CD24−/low and MS-forming treatment-resistant cells were used to identify CSC pathways (p < .05, Fisher exact two-tailed test). The enriched pathways included: NOTCH, VEGF, PTEN, sonic Hedgehog, Wnt/β-catenin, JAK/STAT, P53, and PI3K/AKT signaling. The CSB-analysis was then performed to extend the incomprehensive pathways and establish crosstalks within pathways [15, 16]. The signaling networks included 140 gene nodes for the CD44+/CD24−/low cells (Fig. 1A and Supporting Information Fig. S1A) and 153 gene nodes for the MS-forming treatment-resistant cells (Supporting Information Fig. S1B). After mapping all gene nodes to the drug database, a total of 21 drugs, including CQ, auranofin, and arsenic trioxide, were identified as candidate drugs which could target the CSC pathways. We chose to focus on CQ, which has been clinically used for several decades, displaying a safe toxicity profile, alone and in combination with PTX.

Chloroquine targets breast cancer stem cells (CSCs) of triple negative breast cancer (TNBC). (A): Twenty-one FDA-approved drug targets on seven signaling nodes in the network signaling pathway for CD44+/CD24−/low tumor cells. Circles denote gene nodes, yellow links are protein-protein interactions, gray squares are drugs, and gray links are drug-target interactions. The color of every gene node is determined by a statistical score, that is, −log(p-value), where p-value is the Student’s t test statistical value considering all the probes’ value for the gene in the groups. The data scale is between 0 and 8. 0 is for no-color and 8 for dark red circle. A node in a blue rectangle is enlarged for better view. The red oval circle indicates chloroquine. The enlarged whole network is also presented in Supporting Information Figure S1A. (B): CQ reduced primary MSFE of Hs578t, MDA-MB-231, and HCC1937 TNBCs and the secondary MSFE in MDA-MB-231 and SUM159PT cells in a dose-dependent manner. (C): CQ reduced the CD44+/CD24−/low CSC population alone or in combination with PTX. CQ1 and CQ5 indicates CQ at 1 µM or at 5 µM, respectively, while CQ1-PTX or CQ5-PTX indicates treatment with CQ at the corresponding doses in combination with PTX (5 nM). Groups treated with CQ alone were compared to control (B, C), and the combination treatment groups were compared to the PTX alone group (C). Con: control (DMSO), PTX: paclitaxel (in DMSO), and CQ-PTX: combination of chloroquine and paclitaxel. One-way ANOVA (Tukey) was used for statistical analysis. This figure represents the mean difference in suppression of MSFE or CD44+/CD24−/low CSC population with SD. Asterisk indicates p < .05. (D): CQ and PTX combination treatment reduced CD44+/CD24−/low CSCs in patients with drug-resistant recurrent tumors. Pre indicates pretreatment and post indicates post-treatment. The MS assay and fluorescence-activated cell sorting analysis for the CSCs were repeated at least three times with similar results. Representative results are shown. Abbreviations: CQ, chloroquine; MSFE, mammosphere-forming efficiency; PTX, paclitaxel.

CQ Inhibits MS Formation and Reduces CD44+/CD24/low Populations in TNBC Cell Lines

To determine whether CQ would have an effect on decreasing MS-forming efficiency (MSFE), we performed a dose response experiment for CQ in four different TNBC cell lines, Hs578t, MDA-MB-231, SUM159PT, and HCC1937 as shown in Figure 1B. Even though sensitivity to CQ varied according to cell line, we found that CQ at 1 or 5 µM effectively decreased primary MSFE in Hs578t, MDA-MB-231, and HCC1937 TNBC cell lines (Fig. 1B), and also secondary MS formation in SUM159 and MDA-MB-231 cells (Fig. 1B) by specifically targeting the CD44+/CD24−/low populations (Supporting Information Fig. S2A). Hs578t and HCC1937 cells did not form secondary MS under the same culture conditions. Similarly, we observed a significant dose-dependent reduction in CD44+/CD24−/low populations (15%–50%) with CQ treatment alone or in combination with PTX, correlating with the observed decrease in primary and secondary MSFE (Fig. 1C). Additionally, we found that CQ reduced breast CSCs identified by aldehyde dehydrogenase1 activity through ALDEFLUOR assay as described previously [22]. CQ alone showed significant reduction of ALDEFLUOR-positive cells in MDA-MB-231 (50-fold decrease) and SUM159PT (8-fold decrease) (Supporting Information Fig. S2B).

CQ-PTX Treatment Reduced CD44+/CD24/low Population in Patients

A clinical trial is currently underway to evaluate the efficacy of CQ in combination with PTX in women with treatment-refractory advanced or metastatic breast cancer. Consistent with in vitro results, the combination treatment of CQ and PTX reduced the CD44+/CD24−/low population by five- to sixfold in two patients after treatment cycles (Fig. 1D). However, a minimal reduction of the CSC population was observed in one patient. These results support the preclinical findings and confirm the potential for improved patient response resulting from the combination of CQ and taxane therapy.

Inhibition of Autophagy by CQ Sensitizes TNBC Cells to PTX

We next investigated whether the reduction of CSCs by CQ could be correlated with inhibition of autophagy, thus sensitizing TNBCs to chemotherapy. First, inhibition of autophagy was confirmed by observing accumulation of autophagosomes in Hs578t cells treated with CQ (1 µM) alone and in combination with PTX (5 nM) using transmission electron microscopy (TEM). Autophagosomes were not detected in either control or PTX-treated cells (Fig. 2A). Additionally, CQ induced puncta formation (green) and inhibited the formation of PTX-induced autophagolysosomes (yellow) in MDA-MB-468 cells, expressing GFP-tagged LC3B (Supporting Information Fig. S3A). The inhibition of autophagy was further confirmed by detection of LC3B-II and upregulated p62 in all cells treated with CQ alone or in combination with PTX (Fig. 2B). In PTX-treated cells, a marginal increase in LC3B-II along with a partial increase or decrease in p62 was observed (Fig. 2B), indicating autophagy induction. Enhanced antitumor effects of the combination treatment over PTX alone were confirmed by increased cleaved caspase-3 (Fig. 2B) and by enhanced apoptosis measured by Annexin V and/or Sytox-Blue positive cell populations (Supporting Information Fig. S3B). Additionally, CQ alone increased cleaved caspase-3 in Hs578t and MDA-MB-231 cells (Fig. 2B). Thus, these results suggest that CQ may be used in combination with chemotherapy in TNBC cells.

CQ inhibits PTX-induced autophagy and sensitizes triple negative breast cancer cell lines to PTX. (A): TEM images of autophagosomes in Hs578t cells after 48 hours treatment with Con (DMSO), CQ (1 µM), PTX (5 nM), or combination of CQ and PTX. Solid black arrows indicate autophagosomes. The TEM experiment was performed once and the representative images are shown here. The black bars are 2 µm in length. (B): Western blotting results for p62, LC3B, and cleaved caspase 3 from lysates of Hs578t, MDA-MB-231, and SUM159PT cells after 48-hour treatments as indicated. We confirmed similar effects of CQ in independently repeated experiments. CQ attenuated (C) growth of orthotopic MDA-MB-231 G/L tumors and (D) inhibited spontaneous lung metastasis of the tumors; Con: PBS and CQ: 10 mg/kg i.p. daily (n = 10 per group). Total photon counts were normalized by setting the minimum at 3.50 × 105 and the maximum at 1.05 × 108 (p/second per cm2 per sr) and transformed by log10 to analyze the effects of CQ on metastasis and tumor burden in lungs. Two-tailed Student’s t test was used for the statistical comparison. (E): CQ-PTX attenuated tumor recurrence (MDA-MB-231 G/L) compared to PTX line. (F): The CQ-PTX combination therapy reduced the growth of orthotopic SUM159PT tumors (n = 10). All mice received the scheduled dose of each drug as indicated in Materials and Methods. Two-way ANOVA (Boneferroni post-tests) was used to compare tumor volumes in all animal experiments (C, E, F). All control groups received PBS as the vehicle (*, p < .05; **, p < .01; ***, p < .001). Abbreviations: CQ, chloroquine; PTX, paclitaxel.

In Vivo Inhibition of Tumor Growth and Lung Metastasis by CQ

We observed a significant 50% (p < .0001) in vivo growth inhibition in orthotopic MDA-MB-231 G/L tumors by CQ treatment alone compared to controls (Fig. 2C). Additionally, the CQ treatment prevented spontaneous lung metastasis from 90% in controls to 20% in treatment mice, with significant reduction of tumor burden in lungs (p < .003) (Fig. 2D). We next compared the impact of CQ-PTX treatment against PTX alone in MDA-MB-231 G/L orthotopic tumor models. The combination treatment reduced tumor size by 50% compared to PTX alone (p < .001) (Fig. 2E). Furthermore, we observed significantly slower tumor recurrence in CQ-PTX-treated mice compared to PTX alone treatment arm; 20% of the mice in the CQ-PTX group showed complete regression of tumor during the treatment cycle with no recurrence observed. Moreover, an additional 20% of the mice in the CQ-PTX group showed consistent reduction in tumor size even after the last treatment, in contrast to continuous tumor growth observed in all mice in the PTX group (data not shown). The antitumor effects of CQ-PTX were also confirmed in the SUM159PT orthotopic xenograft model involving a 4-week treatment of Control (PBS) CQ (10 mg/kg, daily, i.p.), PTX (15 mg/kg, twice per week, i.p.), or in combination. Consistently, the CQ-PTX combination treatment arm was the only group to show significant inhibition of tumor growth while CQ alone or PTX alone showed no statistical difference in tumor volume compared to controls (Fig. 2F). These results may suggest that CQ enhances the antitumor effects of PTX by decreasing the CSCs.

CQ Reduces Breast CSCs In Vivo

For CSC analysis, additional cohorts of mice bearing either MDA-MB-231 (n = 7) or SUM159PT (n = 5) orthotopic tumors were treated for 2 weeks with vehicle, CQ (10 mg/kg, daily), PTX (15 mg/kg, twice per week), or the combination, CQ-PTX. We confirmed the enhanced anticancer effects of CQ-PTX in both tumor cell lines compared to the control group or PTX alone (Fig. 3A, 3B). Additionally, we found that PTX significantly increased the Aldeflour+ CSCs by threefold in MDA-MB-231 tumors (Fig. 3C) and the CD44+/CD24−/low CSCs by twofold in SUM159PT models (Fig. 3D) compared to controls. We did not observe any significant change in the CSC population by CQ alone, but CQ in combination with PTX reduced the PTX-induced CSC population to control levels in both tumor cell lines (Fig. 3C, 3D). We further investigated the tumorigenic potential of tumors by testing sphere-forming ability. Interestingly, the PTX-induced CSC increase correlated well with the increased MSFE in both the primary and the secondary MS of MDA-MB-231 and SUM159PT tumors compared to the controls (Fig. 3E, 3F). The CQ-PTX combination treatment significantly inhibited the PTX-induced primary MSFEs of the two tumor cell lines comparable to control levels in the primary MS, and further reduced the MSFE more than four times lower than controls in the secondary MS for both MDA-MB-231 (Fig. 3E) and SUM159PT tumors (Fig. 3F). CQ did not alter the sphere forming ability compared to controls in the primary MS, but reduced the secondary MSFE by fourfold in MDA-MB-231 tumors (Fig. 3E) and twofold in SUM159PT tumors (Fig. 3F). Finally, we confirmed the CSC targeting effects of CQ through a limiting dilution assay for MDA-MB-231 tumors using three dilutions; 75,000 (75k), 25,000 (25k), and 5,000 (5k) cells. CQ or CQ combination with PTX completely inhibited tumor formation for 6 weeks in all three dilutions of cells compared to controls or PTX (Fig. 3G). As anticipated, the PTX-mediated CSC increase also correlated well with higher tumor incidence rates at cell each dilution assay compared to controls; 100% versus 38% at 75k, 50% versus 13% at 25k, and 75% versus 38% at 5k dilutions (Fig. 3G). Also, by pairwise comparison, we confirmed that CQ significantly reduced the CSC frequencies in tumors compared to controls or the PTX treatment group (Fig. 3G). Together, these results strongly support the CSC-targeting effects of CQ in vivo.

CQ reduces the cancer stem cell (CSC) population in vivo. (A, B): Effects of short-term treatments on the growth of orthotopic MDA-MB-231 and SUM159PT tumors in athymic nude mice. Two-way ANOVA (Bonferroni post-tests) was used for statistical analysis. CQ prevented PTX induced CSC enrichment as indicated by Aldefluore + population in MDA-MB-231 tumors (C) and the CD44+/CD24−/low population in SUM159PT tumors (D). CQ reduced the MSFE of MDA-MB-231 (E) and SUM159PT tumors (F). MS indicates mammosphere formation assay. One-way ANOVA (Tukey) was used for statistical analysis. (G): Limiting dilution assay for testing the tumor initiation ability in Scid-beige mice bearing MDA-MB-231 tumors received with short-term treatments (n = 8 per group); 75k: 75,000, 25k: 25,000, and 5k: 5,000 tumor cells. (H): A pairwise comparison for CSC frequencies. p value indicates the significance of the difference on the CSC frequency of the compared two groups. For the statistical analysis, extreme limiting dilution analysis [23] was used. Con (PBS), CQ (10 mg/kg, i.p. daily), PTX (15 mg/kg, twice per week, i.p.), CQ-PTX, the combination (*, p < .05; **, p < .01; ***, p < .001). Abbreviations: CQ, chloroquine; MS, mammosphere; PTX, paclitaxel.

CQ Inhibits Jak2-STAT3 Signaling Pathway in CSCs

As the Jak2/STAT3 signaling pathway is critical for maintenance of breast CSCs [5], we investigated the effects of CQ, PTX, and the combination on this signaling pathway. The phosphorylation of STAT3 (Tyr705) was compromised by CQ alone, PTX, or CQ-PTX in Hs578t and SUM159PT cells, while CQ-PTX was most effective at inhibiting phosphorylation (Fig. 4A). Analogously, we observed significant reduction of pSTAT3 by CQ or CQ-PTX compared to controls in MDA-MB-231 cells. However, PTX induced a substantially higher phosphorylation of STAT3 (Fig. 4A). The changes in STAT3 phosphorylation were correlated with the phosphorylation status of Jak2 in all three cell lines. Interestingly, we observed significant reduction of Jak2 expression by CQ-PTX in all three cell lines (Fig. 4A). We next investigated the Jak2-STAT3 signaling pathway in sorted CD44+/CD24−/low CSC and non-CSC populations of SUM159PT cells when treated with either CQ, PTX, or in combination, CQ-PTX. We observed a reduction of Jak2 phosphorylation in CSCs by CQ, PTX, and CQ-PTX, with the most significant inhibition achieved with CQ-PTX compared to controls (Fig. 4B). In non-CSCs, only the combination treatment inhibited Jak2 phosphorylation. However, we found substantial reduction in Jak2 following CQ-PTX treatment only in the CSCs (Fig. 4B). Additionally, CQ inhibited pSTAT3–705, albeit, less significantly than CQ-PTX treatment, only in CSCs of SUM159PT, while PTX alone showed no effects (Fig. 4B). In non-CSCs, pSTAT3–705 was upregulated by CQ, PTX, and CQ-PTX. Consistently, the combination treatment also reduced the phosphorylation of STAT3 at S727 in CSCs (Fig. 4B). Furthermore, CQ alone or in combination with PTX substantially inhibited the PI3K/Akt/mTOR pathway, an alternate pathway that can activate STAT3 in breast CSCs [24], through activation of PTEN (Supporting Information Fig. S4). These results suggest that CQ may affect CSCs by inhibiting activation of STAT3 and by reducing Jak2 expression.

CQ regulates the Jak2-STAT3 pathway in CSCs. (A): Results of Western blot assays for pSTAT3–705, STAT3, pJak2, Jak2, and Actin on lysates of Hs578t, MDA-MB-231, and SUM159PT cells treated for 48 hours with a four-treatmet regimen; DMSO (Con) or CQ (1 µm), PTX (5 nM) in DMSO, or the combination. (B): Western blot analysis for the detection of pSTAT3–705, pSTAT3–727, STAT3, pJak2, Jak2, and Actin on lysates of sorted CD44+/CD24−/low CSCs or non-CSCs of SUM159PT cells. These cells were incubated in MC+ containing 2% FBS and treated with the four-treatment regimen for 48 hours. (C): Western blot analysis of SOCS1 and SOCS3 expression in SUM159PT CSCs after treatment of PTX or CQ-PTX for 24, 36, and 48 hours. (D): Immunoprecipitation of SOCS3 with Jak2 in SUM159PT CSCs treated with PTX (5 nM) or CQ (1 µM)-PTX (5 nM). Western blot for SOCS3 and Jak2 is shown. Normal IgG showed no nonspecific bands at the expected molecular size (data not shown). (E): Confocal microscope images of SOCS3 colocalization with Jak2 in SUM159PT CSCs treated with PTX or CQ-PTX for 48 hours. Jak2 (red) and SOCS3 (green). Hoechst 33342 (100 ng/ml) stains for the nucleus. Colocalization of SOCS3 and Jak2 is depicted in the separate black-white cell images in which the white dots indicate the colocalization. All images were taken using an Olympus FV1000 confocal microscope with a ×100 objective lens and ×2.5 optical zoom. The white bars indicated 10 µm in length. (F): Western blot assay of pSTAT3–705, STAT3, pJak2, Jak2, SOCS3, and Actin on lysates of SUM159PT CSCs in which SOCS3 was silenced prior to the 48 treatment with CQ-PTX. siRNA-Scr: siRNA with scrambled RNA sequences, siRNAi-SOCS3: siRNA with SOCS3 targeting RNA sequences. These are representative results of independent duplicate or triplicate experiments. Abbreviations: CQ, chloroquine; CSC, cancer stem cell; PTX, paclitaxel.

CQ-PTX Induces the Expression of Suppressor of Cytokine Signaling Families in CSCs

Since SOCS1 and SOCS3 are known to induce Jak2 degradation upon its activation [25, 26], we investigated whether the suppressor of cytokine signaling (SOCS) family plays a role in CQ-mediated Jak2/STAT3 deregulation. Gene expression analysis by RT-PCR showed no alteration of Jak2 gene expression under any treatment (data not shown). In SUM159PT CSCs, a time-dependent increase in SOCS1 and SOCS3, and reciprocal decrease in pJak2 and Jak2, was found following CQ-PTX treatment compared to PTX alone at 48 hours (Fig. 4C). However, in an immunoprecipitation assay, SOCS3 was found associated with Jak2 and not SOCS1 in SUM159PT CSCs (Fig. 4D). Using immunofluorescence colocalization imaging, the increased interaction of Jak2 with SOCS3 was confirmed in SUM159PT CSCs treated with CQ-PTX in comparison to PTX alone (Fig. 4E). Finally, we were able to rescue Jak2 expression by silencing SOCS3 using siRNA in SUM159PT CSCs treated with CQ-PTX (Fig. 4F). Moreover, silencing SOCS3 expression increased Jak2 protein level in normal culture conditions, hinting at the Jak2 regulating nature of SOCS3 in SUM159PT CSCs (Supporting Information Fig. S5). Taken together, these results confirm that CQ-PTX treatment resulted in the expression of SOCS1 and SOCS3 and enhanced interaction of SOCS3 with Jak2, causing reduction of Jak2 protein level in CSCs.

CQ Suppressed the Expression of DNMT1 in CSCs

The expression of SOCS1 and SOCS3 can be regulated by DNA methylation [27, 28]. To that end, we found that the CQ-PTX combination treatment significantly reduced DNMT1 in of Hs578t, SUM159PT, and MDA-MB-231 bulk tumors compared to controls or PTX alone treatment (Fig. 5A). Likewise, we also observed significantly reduced DNMT1 by CQ or CQ-PTX compared to controls and PTX alone, respectively, in CSCs and non-CSCs of SUM159PT, while PTX increased DNMT1 expression in both populations of cells (Fig. 5B). The negative effects of CQ-PTX on DNMT1 expression in CSCs of basal-like TNBCs HCC1937 and HCC38 (Fig. 5B) were further confirmed. The changes in DNMT1 protein levels induced by CQ or CQ-PTX significantly correlated with changes in global DNA methylation. In Hs578t and MDA-MB-231 cells, CQ alone induced hypomethylation by 50% (p < .0001) and 8% (p < .05), respectively (Fig. 5C). PTX also induced hypomethylation in Hs578t by 50% (p < .0001), while no changes were observed in MDA-MB-231 cells. CQ-PTX induced the most significant hypomethylation in both cell lines compared to controls or to PTX. In SUM159PT bulk tumor cells, no changes in methylation were observed following CQ treatment, while PTX or CQ-PTX induced substantial hypermethylation (Supporting Information Fig. S6). However, CQ induced global hypomethylation in CSCs of SUM159PT by 50% (p < .001) while PTX induced hypermethylation (p < .0001) compared to controls (Fig. 5C). CQ-PTX reduced global methylation by 10% relative to PTX treatment (p < .05) (Fig. 5C). It is important to note that more than 85% in Hs578t and 97% of MDA-MB-231 cells were CD44+/CD24−/low. Thus, we confirmed that the increase in SOCS1 and SOCS3 expressions was due to the downregulation of DNMT1 in SUM159PT CSCs (Fig. 5D). However, we found a fourfold increase in SOCS3 mRNA alone in CSCs treated with CQ-PTX compared to PTX, while no difference in SOCS1 mRNA was detected (Fig. 5E). This result suggests that SOCS1 upregulation might be an indirect effect of DNA hypomethylation. Consequently, we observed CQ-PTX induced hypomethylation in three different promoter regions of SOCS3 after CQ-PTX treatment in SUM159PT CSCs compared to PTX (Fig. 5F). We also confirmed the effects of CQ-PTX on DNMT1, pSTAT3, and Jak2 in vivo (Supporting Information Fig. S7A, S7B). Taken together, our data suggest that CQ regulates the Jak2-STAT3 pathway to target CSCs through DNA methylation of SOCS3 in the presence of PTX.

CQ induced DNA hypomethylation by reducing DNMT1 expression. (A): Western blot analysis of DNMT1 from extracts of Hs578t, MDA-MB-231, and SUM159PT cells, and (B) extracts of cancer stem cells (CSCs) or non-CSCs of SUM159PT and CSCs of HCC1937 and HCC38 cells. All cells were treated for 48 hours with the four-treatment regimen; DMSO (Con) or CQ (1 µM), PTX (5 nM) in DMSO, or the combination. (C): Measurement of global DNA methylation using genomic DNA extracted from Hs578t and MDA-MB-231 cells and DNA from SUM159PT CSCs. All cells were treated with the four-treatment regimen. One-way ANOVA (Tukey) was used for statistical analysis. (D): Western blot assay for DNMT1, SOCS1, and SOCS3, and Actin after silencing DNMT1 in PTX-treated SUM159PT CSCs for 30 hours. (E): Expression of SOCS1 or SOCS3 transcripts in SUM159PT CSCs treated with either PTX (5 nM) or CQ (1 µM) and PTX (5 nM) for 48 hours. Student’s t test was used for statistical analysis. (F): Images of DNA gel electrophoresis of SOCS3 promoter PCR to confirm CQ-induced DNA hypomethylation. SUM159PT CSCs were treated with either PTX or CQ-PTX for 48 hours before DNA extraction. Primer indicates the corresponding primer sets for PCR. Input indicates PCR reactions with the sonicated genomic DNA. N-IgG is the commercially available mouse normal IgG used for MeIP as a negative control. 5-mC Ab is the monoclonol antibody recognizing 5-methycytosine of DNA for MeDIP. The black arrow indicates the expected PCR products. Images were cropped from the original found in Supporting Information Figure S8. All experiments were repeated at least two or three times with similar results, and the representative results were presented (*, p < .05; **, p < .02; ***, p < .0001). Abbreviations: CQ, chloroquine; DNMT1, DNA methyltransferase 1; PTX, paclitaxel.

Jak2-STAT3 and DNMT1 Synergistically Regulate TNBC CSCs

Using siRNAs, we examined the impact of silencing Jak2, STAT3, and DNMT1, on TNBC CSCs. The silencing efficiency in Hs578t, MDA-MB-231, and SUM159PT cells was confirmed by detection of DNMT1, Jak2, and STAT3 using Western blot assay (Fig. 6A). As shown in Figure 6B, silencing either of the genes resulted in reduction of the CD44+/CD24−/low population by 5%–10% in Hs578t and MDA-MB-231 cells. The reduction of CSCs was more significant when two of the three genes were silenced simultaneously in Hs578t and MDA-MB-231 cells, resulting in an approximate 15%–20% reduction of CSCs. However, the most significant reduction of CSCs was observed when all three genes were silenced simultaneously, resulting in roughly 25%–30% reduction of CSCs (Fig. 6B). Contrary to the aforementioned cell lines, SUM159PT cells showed a significant 50% reduction of CSCs following silencing of a single gene, with effects enhanced through silencing of Jak2 or STAT3 with DNMT1. However, in SUM159PT, the most effective CSC reduction was achieved when all three genes were silenced simultaneously. An MS assay was then performed after silencing each gene using specific siRNA in all three cell lines. Contrary to the FACS analysis of the CD44+/CD24−/low CSCs, the silencing of DNMT1, Jak2, or STAT3 altered MSFE more dramatically, with roughly a 30%–70% reduction of MSFE observed in MDA-MB-231 and SUM159PT cells compared to controls (Fig. 6C). In Hs578t cells, STAT3 silencing alone was effective at inhibiting MSFE by 70% (Fig. 6C). STAT3 silencing was more effective at reducing MSFE than either DNMT1 or Jak2 in all three cell lines. Interestingly, severely compromised MSFE was observed when any two of the three genes were silenced (Fig. 6C). Although there was additional reduction of MSFE by three-gene silencing compared to two-gene silencing, no significance was found except in SUM159PT cells (Fig. 6C). These results confirm that DNA methylation plays a critical role in maintenance of breast CSCs concomitantly with Jak2-STAT3 signaling.

Both DNMT1 and the Jak2-STAT3 pathway are critical for cancer stem cells (CSCs) of triple negative breast cancer. (A): Confirmation by Western blot analysis in Hs578t, MDA-MB-231, and SUM159PT cells for DNMT1, Jak2, and STAT3. GAPDH indicates the loading control. (B): Fluorescence-activated cell sorting analysis of CD44+/CD24−/low CSCs after silencing using the indicated single or combination of corresponding RNAi. (C): Impact of gene silencing on MSFE. D indicates silencing of DNMT1, J for Jak2, S for STAT3, D/J for the silencing of DNMT1 and Jak2, D/S for DNMT1 and STAT3, J/S for Jak2 and STAT3, and D/J/S for the silencing of DNMT1, Jak2, and STAT3. Scr indicates RNAi of scrambled sequences and serves as control. One-way ANOVA (Tukey) was used for statistical analysis. All experiments were repeated three times with similar results. The representative results are presented (*, p < .05; **, p < .01; ***, p < .001). Abbreviations: DNMT1, DNA methyltransferase 1; MSFE, mammosphere-forming efficiency.

CQ Rewrites DNA Methylation in MDA-MB-231 Cells

Changes in DNA methylation by MBD-enriched DNA from MDA-MB-231 cells were analyzed after 48 hours CQ treatment. Substantial differences were observed in the number and make-up of Model-based analysis of ChIP-seq (MACS) defined methyl-CpG-binding domain protein-enriched peaks within the proximal promoter region (−5,000 to +200) of protein coding genes (Fig. 7A). Upon more detailed differentiation analysis of MACS defined MDB-enriched peaks between the CQ and control treatments (MAnorm [29]), the proximal promoter regions of 359 genes uniquely methylated in the control treatment compared to CQ and 136 exclusively methylated in the CQ treatment were identified. To assess any biological significance of these genes with affected proximal regulatory regions, we conducted functional enrichment analysis with GeneCodis3 [30, 31]. Roughly one-third of the genes with hypomethylated proximal promoters following CQ treatment were allocated into four functional groups (p ≤ 9.06 × 10−6); protein, nucleotide, ATP, and RNA binding functions (Fig. 7B). The majority of the genes with hypermethylated proximal promoter regions in the CQ treatment group were predicted to have binding functions to zinc ion, protein, nucleotide, beta-catenin, metal ion, and single-stranded RNA (p ≤ 7.83 × 10−5) (Fig. 7C). Enriched genes are listed in Supporting Information Tables S2 and S3. Additionally, the uniquely methylated genes in controls were enriched only for one KEGG-enriched pathway, protein processing in endoplasmic reticulum (ER) (p < .0002), while genes for CQ were enriched for pathways in cancer (p = 4.43 × 10−6) and the Wnt signaling pathway (p < .0003) (Fig. 7D). Thus, these results suggest that CQ can regulate CSCs by affecting multiple signaling pathways through DNA methylation via downregulation of DNMT1, and through inhibition of the PI3K/Akt/mTOR and Jak2-STAT3 pathways (Fig. 7E).

CQ induces hypomethylation and resets the DNA methylome. (A): A heatmap showing unique methylation status of the proximal promoter region (−5,000 to +200) of protein coding genes in MDA-MB-231 cells treated for 48 hours with PBS (Con) or CQ (1 µM). The numbers on the left of the heatmap indicate the total number of genes with the uniquely methylated promoters. Methylation degree increases from blue to yellow (log10 transformed tag counts). (B, C): Schematic diagrams illustrating results of the functional enrichment analysis with GeneCodis for the identified genes. The p values for the functional enrichment test were (B) p ≤ 9.06 × 10−6 and (C) p ≤ 7.83 × 10−5. (D): KEGG-enriched pathway of genes that are hyper- or hypo-methylated by CQ in MDA-MB-231 cells. (E): A schematic diagram illustrating how CQ can affect triple negative breast cancer CSCs. Red indicates inhibition and blue activation. The red dotted line indicates possible inhibition and the blue dotted line represents activation through DNA methylation. Abbreviations: CQ, chloroquine; CSC, cancer stem cell.

Discussion

CQ, an autophagy inhibitor, was named as a potential repositioned drug candidate for treatment against CSCs through in silico network analysis of gene signatures specific for drug-resistant CD44+/CD24−/low cells derived from patient biopsies. Based on our observation of CSC enrichment following chemotherapy [4, 32], autophagy was hypothesized as an underlying mechanism to maintain viable CSC populations in TNBC. This is further supported by previous studies, suggesting autophagy as a key regulator of breast CSCs [11, 12]. To this end, we demonstrated the anti-CSC activity of CQ through the reduction of MSFE and the CD44+/CD24−/low CSCs. This reduction of CSCs correlates well with the inhibition of PTX-induced autophagy and with increases in apoptosis. As CSCs have been implicated in metastasis and recurrence [22, 33-35], we confirmed the anti-CSC effects of CQ in vivo through inhibition of tumor growth, prevention of spontaneous lung metastasis, and attenuation of tumor recurrence. The enhanced antitumor effects were accompanied with suppression of CSC enrichment following PTX treatment and significantly impaired tumor initiation ability in vivo. More importantly, we found a significant reduction of CD44+/CD24−/low CSC populations in patients who underwent clinical trials involving the combination therapy of CQ with taxanes. Thus, our data strongly support the anti-CSC activity of CQ against CSCs in TNBC through autophagy inhibition.

The Jak2-STAT3 pathway was compromised by CQ alone or in combination with PTX. A significant inhibition of the Jak2 phosphorylation by CQ alone was observed in all cell lines examined. We suspect that CQ may induce ER stress which mediates inhibition of Jak2 phosphorylation through inhibition of autophagy, downregulation of the PI3K/Akt/mTOR pathway, and hypomethylation of ER stress-related genes in MDA-MB-231 cells. Kimura et al. [36], and Um et al. [37] reported similar ER stress-mediated inhibition of Jak2-STAT3 pathway. However, the inhibitory effects of CQ on Jak2-STAT3 were most profound following combination therapy, as demonstrated by a decrease in phosphorylation and expression of Jak2 in all cell lines examined. Moreover, the inhibitory effect on Jak2 expression was CSC specific. These results are in agreement with previous reports on the essential role of the Jak2-STAT3 signaling pathway for growth and maintenance of CD44+/CD24−/low breast CSCs [5, 24]. Additionally, the decrease in Jak2 was accompanied with a reduction of DNMT1 expression that correlated well with the global DNA hypomethylation in CSCs. Similar to Jak2-STAT3, DNMT1 is an important gene expression regulator in normal stem cells as well as CSCs [38, 39]. In leukemia, haploinsufficiency of DNMT1 is known to impair leukemogenesis and self-renewal of leukemia stem cells [40]. Furthermore, the epigenetic role of STAT3 has been described for inhibition of tumor suppressor genes through interaction with DNMT1 [41, 42]. Thus, our findings suggest that CQ regulates CSCs through epigenetic regulation in addition to the inhibition of autophagy.

SOCS1 and SOCS3 have been identified as versatile negative regulators of the Jak2-STAT3 signaling pathway [43-45]. Along with downregulation of Jak2, the combination treatment induced expression of SOCS1 and SOCS3 as well as interaction of SOCS3 with Jak2 in CSCs. Additionally, SOCS1 and SOCS3 expression were inversely proportional to the expression of DNMT1, while the opposite was observed following PTX treatment alone. SOCS1 and SOCS3 are known to interact with Jak2 and induce its degradation [25, 26, 43-45]. Moreover, the expression of SOCS1 and SOCS3 is tightly regulated by DNA methylation [27, 28]. Thus, we believe that CQ regulates the Jak2/STAT3 signaling pathway in CSCs through deregulation of DNA methylation mediated by loss of DNMT1 expression. In order to determine whether Jak2, STAT3, or DNMT1 was critical for CSC maintenance, sequential gene silencing was performed for all the three genes. Our findings indicate that simultaneous silencing of Jak2, STAT3, and DNMT was most effective in reducing CD44+/CD24−/low CSCs and significantly impaired the sphere-forming ability. This study defines a possible mechanism of CQ for inhibition of CSCs through regulation of the Jak2/STAT3 and DNA methylation through DNMT1.

Conclusions

In summary, this is the first study that identifies a CQ-mediated decrease in CD44+/CD24−/low CSC due to inhibition of the Jak2-STAT3 signaling pathway through expression of SOCS1 and SOCS3, which in turn deregulates Jak2 expression. Moreover, this is the first study to demonstrate that inhibition of the Jak2-STAT3 pathway is associated with downregulation of DNMT1 and subsequent global DNA hypomethylation. More importantly, these preclinical findings are reflected in a currently ongoing clinical trial involving CQ-PTX treatment, where significant reduction in CD44+/CD24−/low populations has been observed. Herein, we report that CQ reduces CSCs in TNBC by altering the Jak2-STAT3 pathway and DNMT1 expression in addition to autophagy inhibition. Subsequent analysis of CQ-mediated changes in epigenome and gene expression in combination with other epigenetic inhibitors, such as histone deacetylase inhibitors, may enable refinements in strategies targeting TNBC CSC subpopulations.

Acknowledgments

This work was supported by NIH/NCI Grants R01 CA138197, U54 CA149196, Golfers against Cancer, Breast Cancer Research Foundation, Causes for a Cure, Team Tiara, Emily W. Herrman Cancer Research Laboratory, and Komen for Cure KG 081694. We declare that none of the authors have any financial interest related to this work.

Author Contributions

D.S.C. and E.B.: study conception, design, collection, assembly, analysis, interpretation of data, and manuscript writing; Y.-S.K.: collection and/or assembly, and analysis of data and manuscript writing; A.A.R.: provision of study material and patients; H.Z.: assembly and analysis of data and manuscript writing; T.H.-M.H. and C.-L.C.: collection of sequencing data; G.J.: assembly of data; M.D.L.: data analysis and interpretation of clinical data; L.A.B., W.Q., and H.H.W.: collection and assembly of data; S.M.G. and B.D.: data analysis; M.F. and S.T.C.W.: administrative support and final approval of manuscript; J.C.: data analysis and interpretation of data, manuscript writing, financial support, administrative support, and final approval of manuscript. D.S.C. and E.B. contributed equally to this article.

Disclosure  of Potential Conflicts  of Interest

The authors indicate no potential conflicts of interest.

References

1

Foulkes
 
WD
,
Smith
 
IE
,
Reis-Filho
 
JS
.
Triple-negative breast cancer
.
N Engl J Med
 
2010
;
363
:
1938
1948
.

2

Dent
 
R
,
Trudeau
 
M
,
Pritchard
 
KI
, et al.
Triple-negative breast cancer: Clinical features and patterns of recurrence
.
Clin Cancer Res
 
2007
;
13
:
4429
4434
.

3

Idowu
 
MO
,
Kmieciak
 
M
,
Dumur
 
C
, et al.
CD44(+)/CD24(-/low) cancer stem/progenitor cells are more abundant in triple-negative invasive breast carcinoma phenotype and are associated with poor outcome
.
Hum Pathol
 
2012
;
43
:
364
373
.

4

Creighton
 
CJ
,
Li
 
X
,
Landis
 
M
, et al.
Residual breast cancers after conventional therapy display mesenchymal as well as tumor-initiating features
.
Proc Natl Acad Sci USA
 
2009
;
106
:
13820
13825
.

5

Marotta
 
LL
,
Almendro
 
V
,
Marusyk
 
A
, et al.
The JAK2/STAT3 signaling pathway is required for growth of CD44(+)CD24(-) stem cell-like breast cancer cells in human tumors
.
J Clin Invest
 
2011
;
121
:
2723
2735
.

6

Gupta
 
PB
,
Onder
 
TT
,
Jiang
 
G
, et al.
Identification of selective inhibitors of cancer stem cells by high-throughput screening
.
Cell
 
2009
;
138
:
645
659
.

7

Janku
 
F
,
McConkey
 
DJ
,
Hong
 
DS
, et al.
Autophagy as a target for anticancer therapy
.
Nat Rev
 
2011
;
8
:
528
539
.

8

Kenific
 
CM
,
Thorburn
 
A
,
Debnath
 
J.
 
Autophagy and metastasis: Another double-edged sword
.
Curr Opin Cell Biol
 
2010
;
22
:
241
245
.

9

Martinez-Outschoorn
 
UE
,
Lin
 
Z
,
Ko
 
YH
, et al.
Understanding the metabolic basis of drug resistance: Therapeutic induction of the Warburg effect kills cancer cells
.
Cell Cycle
 
2011
;
10
:
2521
2528
.

10

Sanchez
 
CG
,
Penfornis
 
P
,
Oskowitz
 
AZ
, et al.
Activation of autophagy in mesenchymal stem cells provides tumor stromal support
.
Carcinogenesis
 
2011
;
32
:
964
972
.

11

Cufi
 
S
,
Vazquez-Martin
 
A
,
Oliveras-Ferraros
 
C
, et al.
Autophagy positively regulates the CD44(+) CD24(-/low) breast cancer stem-like phenotype
.
Cell Cycle
 
2011
;
10
:
3871
3885
.

12

Gong
 
C
,
Bauvy
 
C
,
Tonelli
 
G
, et al.
Beclin 1 and autophagy are required for the tumorigenicity of breast cancer stem-like/progenitor cells
.
Oncogene
 
2013
;
32
:
2261
2272
.

13

Sotelo
 
J
,
Briceno
 
E
,
Lopez-Gonzalez
 
MA
.
Adding chloroquine to conventional treatment for glioblastoma multiforme: A randomized, double-blind, placebo-controlled trial
.
Ann Intern Med
 
2006
;
144
:
337
343
.

14

Bellodi
 
C
,
Lidonnici
 
MR
,
Hamilton
 
A
, et al.
Targeting autophagy potentiates tyrosine kinase inhibitor-induced cell death in Philadelphia chromosome-positive cells, including primary CML stem cells
.
J Clin Invest
 
2009
;
119
:
1109
1123
.

15

Jin
 
G
,
Fu
 
C
,
Zhao
 
H
, et al.
A novel method of transcriptional response analysis to facilitate drug repositioning for cancer therapy
.
Cancer Res
 
2012
;
72
:
33
44
.

16

Zhao
 
H
,
Jin
 
G
,
Cui
 
K
, et al.
Novel modeling of cancer cell signaling pathways enables systematic drug repositioning for distinct breast cancer metastases
.
Cancer Res
 
2013
;
73
:
6149
6163
.

17

Schott
 
AF
,
Landis
 
MD
,
Dontu
 
G
, et al.
Preclinical and clinical studies of gamma secretase inhibitors with docetaxel on human breast tumors
.
Clin Cancer Res
 
2013
;
19
:
1512
1524
.

18

Kim
 
JH
,
Choi
 
DS
,
Lee
 
OH
, et al.
Antiangiogenic antitumor activities of IGFBP-3 are mediated by IGF-independent suppression of Erk1/2 activation and Egr-1-mediated transcriptional events
.
Blood
 
2011
;
118
:
2622
2631
.

19

Rozen
 
S
,
Skaletsky
 
H
.
Primer3 on the WWW for general users and for biologist programmers
.
Methods Mol Biol
 
2000
;
132
:
365
386
.

20

Langmead
 
B
,
Trapnell
 
C
,
Pop
 
M
, et al.
Ultrafast and memory-efficient alignment of short DNA sequences to the human genome
.
Genome Biol
 
2009
;
10
:
R25
.

21

Feng
 
J
,
Liu
 
T
,
Qin
 
B
, et al.
Identifying ChIP-seq enrichment using MACS
.
Nat Protocols
 
2012
;
7
:
1728
1740
.

22

Charafe-Jauffret
 
E
,
Ginestier
 
C
,
Iovino
 
F
, et al.
Breast cancer cell lines contain functional cancer stem cells with metastatic capacity and a distinct molecular signature
.
Cancer Res
 
2009
;
69
:
1302
1313
.

23

Hu
 
Y
,
Smyth
 
GK
.
ELDA: Extreme limiting dilution analysis for comparing depleted and enriched populations in stem cell and other assays
.
J Immunol Methods
 
2009
;
347
:
70
78
.

24

Zhou
 
J
,
Wulfkuhle
 
J
,
Zhang
 
H
, et al.
Activation of the PTEN/mTOR/STAT3 pathway in breast cancer stem-like cells is required for viability and maintenance
.
Proc Natl Acad Sci USA
 
2007
;
104
:
16158
16163
..

25

Ungureanu
 
D
,
Saharinen
 
P
,
Junttila
 
I
, et al.
Regulation of Jak2 through the ubiquitin-proteasome pathway involves phosphorylation of Jak2 on Y1007 and interaction with SOCS-1
.
Mol Cell Biol
 
2002
;
22
:
3316
3326
.

26

Kershaw
 
NJ
,
Murphy
 
JM
,
Liau
 
NP
, et al.
SOCS3 binds specific receptor-JAK complexes to control cytokine signaling by direct kinase inhibition
.
Nat Struct Mol Biol
 
2013
;
20
:
469
476
.

27

Yoshikawa
 
H
,
Matsubara
 
K
,
Qian
 
GS
, et al.
SOCS-1, a negative regulator of the JAK/STAT pathway, is silenced by methylation in human hepatocellular carcinoma and shows growth-suppression activity
.
Nat Genet
 
2001
;
28
:
29
35
.

28

He
 
B
,
You
 
L
,
Uematsu
 
K
, et al.
SOCS-3 is frequently silenced by hypermethylation and suppresses cell growth in human lung cancer
.
Proc Natl Acad Sci USA
 
2003
;
100
:
14133
14138
.

29

Shao
 
Z
,
Zhang
 
Y
,
Yuan
 
GC
, et al.
MAnorm: A robust model for quantitative comparison of ChIP-Seq data sets
.
Genome Biol
 
2012
;
13
:
R16
.

30

Tabas-Madrid
 
D
,
Nogales-Cadenas
 
R
,
Pascual-Montano
 
A
.
GeneCodis3: A non-redundant and modular enrichment analysis tool for functional genomics
.
Nucleic Acids Res
 
2012
;
40
:
W478
483
.

31

Nogales-Cadenas
 
R
,
Carmona-Saez
 
P
,
Vazquez
 
M
, et al.
GeneCodis: Interpreting gene lists through enrichment analysis and integration of diverse biological information
.
Nucleic Acids Res
 
2009
;
37
:
W317
322
.

32

Li
 
X
,
Lewis
 
MT
,
Huang
 
J
, et al.
Intrinsic resistance of tumorigenic breast cancer cells to chemotherapy
.
J Natl Cancer Inst
 
2008
;
100
:
672
679
.

33

Sheridan
 
C
,
Kishimoto
 
H
,
Fuchs
 
RK
, et al.
CD44+/CD24- breast cancer cells exhibit enhanced invasive properties: An early step necessary for metastasis
.
Breast Cancer Res
 
2006
;
8
:
R59
.

34

Liu
 
H
,
Patel
 
MR
,
Prescher
 
JA
, et al.
Cancer stem cells from human breast tumors are involved in spontaneous metastases in orthotopic mouse models
.
Proc Natl Acad Sci USA
 
2010
;
107
:
18115
18120
.

35

Merlos-Suarez
 
A
,
Barriga
 
FM
,
Jung
 
P
, et al.
The intestinal stem cell signature identifies colorectal cancer stem cells and predicts disease relapse
.
Cell Stem Cell
 
2011
;
8
:
511
524
.

36

Kimura
 
K
,
Yamada
 
T
,
Matsumoto
 
M
, et al.
Endoplasmic reticulum stress inhibits STAT3-dependent suppression of hepatic gluconeogenesis via dephosphorylation and deacetylation
.
Diabetes
 
2011
;
61
:
61
73
.

37

Um
 
HJ
,
Min
 
KJ
,
Kim
 
DE
, et al.
Withaferin A inhibits JAK/STAT3 signaling and induces apoptosis of human renal carcinoma Caki cells
.
Biochem Biophys Res Commun
 
2012
;
427
:
24
29
.

38

Gereige
 
LM
,
Mikkola
 
HK
.
DNA methylation is a guardian of stem cell self-renewal and multipotency
.
Nat Genet
 
2009
;
41
:
1164
1166
.

39

Mathews
 
LA
,
Crea
 
F
,
Farrar
 
WL
.
Epigenetic gene regulation in stem cells and correlation to cancer
.
Differentiation
 
2009
;
78
:
1
17
.

40

Trowbridge
 
JJ
,
Sinha
 
AU
,
Zhu
 
N
, et al.
Haploinsufficiency of Dnmt1 impairs leukemia stem cell function through derepression of bivalent chromatin domains
.
Genes Dev
 
2012
;
26
:
344
349
.

41

Lee
 
H
,
Zhang
 
P
,
Herrmann
 
A
, et al.
Acetylated STAT3 is crucial for methylation of tumor-suppressor gene promoters and inhibition by resveratrol results in demethylation
.
Proc Natl Acad Sci USA
 
2012
;
109
:
7765
7769
.

42

Zhang
 
Q
,
Wang
 
HY
,
Marzec
 
M
, et al.
STAT3- and DNA methyltransferase 1-mediated epigenetic silencing of SHP-1 tyrosine phosphatase tumor suppressor gene in malignant T lymphocytes
.
Proc Natl Acad Sci USA
 
2005
;
102
:
6948
6953
.

43

Alexander
 
WS
,
Starr
 
R
,
Metcalf
 
D
, et al.
Suppressors of cytokine signaling (SOCS): Negative regulators of signal transduction
.
J Leukoc Biol
 
1999
;
66
:
588
592
.

44

Elliott
 
J
,
Johnston
 
JA
.
SOCS: Role in inflammation, allergy and homeostasis
.
Trends Immunol
 
2004
;
25
:
434
440
.

45

Babon
 
JJ
,
Nicola
 
NA
.
The biology and mechanism of action of suppressor of cytokine signaling 3
.
Growth Factors
 
2012
;
30
:
207
219
.

This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/open_access/funder_policies/chorus/standard_publication_model)