Cancer Therapeutics, Targets, and Chemical Biology Research

Mithramycin Represses Basal and Cigarette Smoke–Induced Expression of ABCG2 and Inhibits Stem Cell Signaling in Lung and Esophageal Cancer Cells

Mary Zhang1, Aarti Mathur1, Yuwei Zhang1, Sichuan Xi1, Scott Atay1, Julie A. Hong1, Nicole Datrice1, Trevor Upham1, Clinton D. Kemp1, R. Taylor Ripley1, Gordon Wiegand2, Itzak Avital2, Patricia Fetsch3, Haresh Mani6, Daniel Zlott4, Robert Robey5, Susan E. Bates5, Xinmin Li7, Mahadev Rao1, and David S. Schrump1

Abstract Cigarette smoking at diagnosis or during therapy correlates with poor outcome in patients with lung and esophageal cancers, yet the underlying mechanisms remain unknown. In this study, we observed that exposure of esophageal cancer cells to cigarette smoke condensate (CSC) led to upregulation of the xenobiotic pump ABCG2, which is expressed in cancer stem cells and confers treatment resistance in lung and esophageal carcinomas. Furthermore, CSC increased the side population of lung cancer cells containing cancer stem cells. Upregulation of ABCG2 coincided with increased occupancy of aryl hydrocarbon receptor, Sp1, and Nrf2 within the ABCG2 promoter, and deletion of xenobiotic response elements and/or Sp1 sites markedly attenuated ABCG2 induction. Under conditions potentially achievable in clinical settings, mithramycin diminished basal as well as CSC- mediated increases in AhR, Sp1, and Nrf2 levels within the ABCG2 promoter, markedly downregulated ABCG2, and inhibited proliferation and tumorigenicity of lung and esophageal cancer cells. Microarray analyses revealed that mithramycin targeted multiple stem cell–related pathways in vitro and in vivo. Collectively, our findings provide a potential mechanistic link between smoking status and outcome of patients with lung and esophageal cancers, and support clinical use of mithramycin for repressing ABCG2 and inhibiting stem cell signaling in thoracic malignancies. Cancer Res; 72(16); 4178–92. 2012 AACR.

Introduction In addition to being a significant risk factor for major Lung and esophageal cancers are leading causes of cancer- morbidity and mortality in individuals undergoing potentially related deaths worldwide (1). In 2011, these malignancies curative resections (6, 7), cigarette smoking diminishes accounted for an estimated 1.8 million deaths globally; in responses to chemo- and radiation therapy, enhances systemic the United States, nearly 160,000 deaths were attributed to metastases, and decreases survival of patients with locally – lung cancer, whereas 15,000 deaths were due to esophageal advanced or disseminated lung and esophageal cancers (8 carcinoma (2). Presently, 80% of lung cancers and 50% of 11); the mechanisms underlying these phenomena have not esophageal carcinomas are directly attributable to cigarette been fully established. Previously, we reported that under smoke (3, 4). Currently, more than 1.3 billion people smoke; clinically relevant exposure conditions, cigarette smoke hence, the global burden of tobacco-associated thoracic malig- enhances tumorigenicity of lung cancer cells via polycomb- Dickkopf-1 Dkk1 nancies will continue to increase, with particularly devastating mediated repression of ( ), which encodes an consequences in developing countries (5). antagonist of Wnt signaling (12). In unpublished studies, we observed a similar phenomenon in esophageal adenocarcino- ma cells following cigarette smoke exposure. In addition, we Authors' Affiliations: 1Thoracic Oncology Section, 2Gastrointestinal and have observed that cigarette smoke activates miR-31, targeting 3 Hepatobiliary Malignancies Section, Surgery Branch, Laboratory of Dkk1 as well as several other Wnt antagonists in lung cancer Pathology, 4Clinical Pharmacy Department, 5Experimental Therapeutics Section, Medical Oncology Branch, Center for Cancer Research, National cells; constitutive expression of this miRNA significantly Cancer Institute, Bethesda, Maryland; 6Department of Pathology, Penn enhanced proliferation of lung cancer cells in vitro and in vivo State Hershey Medical Center, Hershey, Pennsylvania; and 7Clinical Micro- array Core, University of California, Los Angeles, California (13). In more recent studies, we observed that cigarette smoke mediates epigenetic repression of miR-487b in lung cancer Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/). cells, resulting in overexpression of polycomb group proteins BMI1 and SUZ12, as well as Wnt5a, k-ras, and C-myc, all of Corresponding Author: David S. Schrump, Thoracic Oncology Section, Surgery Branch, National Cancer Institute, Building 10; 4-3942, 10 Center which have been implicated in modulating stem cell pluripo- Drive, Bethesda, MD 20892. Phone: 301-496-2128; Fax: 301-451-6934; tency (14–18); consistent with these observations, knockdown E-mail: [email protected] of miR-487b increases proliferation and tumorigenicity of lung doi: 10.1158/0008-5472.CAN-11-3983 cancer cells (Xi and colleagues; submitted). This study was 2012 American Association for Cancer Research. undertaken to examine if cigarette smoke activates additional

4178 Cancer Res; 72(16) August 15, 2012 Mithramycin and ABCG2

stem cell–associated genes, which enhance the malignant Murine xenograft experiments phenotype of lung and esophageal cancers in an effort to Athymic nude mice were injected in bilateral flanks with develop novel pharmacologic strategies for treatment of these 1 106 parental A549 cells. Approximately 10 days later when neoplasms. palpable tumors were present, mice were randomly assigned to receive mithramycin at 1 or 2 mg/kg body weight or saline administered as intraperitoneal injections on Monday, Wed- Materials and Methods nesday, and Friday for 3 weeks. Tumor dimensions and mouse Cell lines and treatment conditions weights were measured twice weekly. When control tumors Unless otherwise specified, all cancer lines were obtained approached maximum allowable size, all mice were eutha- from American Type Culture Collection. Cells were validated nized, and tumors were excised, weighed, and processed for by periodic HLA typing of laboratory cultures relative to new additional studies. All mouse experiments were approved by cell aliquots from the repository. NCI-SB-EsC1 and NCI-SB- the National Cancer Institute Animal Care and Use Committee EsC2 (EsC1 and EsC2, respectively) were established in our and were in accordance with the NIH Guide for the Care and laboratory from 2 patients with esophageal adenocarcinoma Use of Laboratory Animals. who developed disease recurrence after undergoing induc- tion chemo/XRT and surgery on Institutional Review Board Chromatin immunoprecipitation approval protocols; these cell lines exhibit HLA as well as Quantitative chromatin immunoprecipitation (ChIP) was cytokeratin expression profiles identical to the respective carried out as described (20), with minor modifications. Full primary tumors. All cancer lines were maintained in RPMI methods are described in Supplementary Methods. Antibodies media supplemented with 10% FBS and 1% penicillin/strep- and primers for ChIP are listed in Supplementary Table S1. tomycin (normal media). Primary normal human small airway epithelial cells (SAEC) were obtained from Lonza, Luciferase promoter–reporter transient transfection Inc. and cultured per vendor instructions. Cigarette smoke experiments condensates (CSC) were generated as described (19). For Submitted as Supplementary Methods. smoke exposure experiments, cells were cultured in appro- priate normal media with or without varying concentrations Flow cytometry of CSC. Media and CSC were changed daily. Cells were Flow cytometry for ABCG2 surface expression and side subcultured as necessary and harvested at appropriate times population was carried out as described (21), with minor for further analysis. modifications. Full methods submitted as Supplementary Mithramycin was obtained from either Sigma or the Methods. Developmental Therapeutics Program (National Cancer Institute). For drug exposure treatments, cells were cultured Statistical analysis in normal media with or without CSC. Media was changed SEM is indicated by bars on figures and was calculated using and mithramycin was added at various concentrations for Microsoft Office Excel 2007. All experiments were conducted 24 hours; cells were harvested at indicated time points for with at a minimum of triplicate samples, and all P values were further analysis. calculated with 2-tailed t tests.

RNA isolation, real-time quantitative reverse transcription PCR, and microarray analysis Results Total RNA was isolated and real-time quantitative reverse CSC induces ABCG2 expression in cultured cancer cells transcription PCR (qRT-PCR) was done as described (19), Affymetrix microarrays were used to identify gene expres- using primers and probes listed in Supplementary Table S1. sion profiles in cultured lung and esophageal cancer cells Full details are submitted as Supplementary Methods. mediated by CSC under clinically relevant exposure condi- tions. ABCG2 [also known as breast cancer resistance Immunoblotting and immunofluorescence protein (BCRP)], which encodes a xenobiotic pump protein Submitted as Supplementary Methods. highly expressed in cancer stem cells (22), was one of the most highly upregulated genes in Calu-6, A549, EsC1, and Generation of stable cells expressing shRNA constructs EsC2 cells exposed to CSC (data not shown). Subsequent A549 and EsC2 cells were transfected with validated short qRT-PCR experiments (Fig. 1A, top panel) showed that A549 hairpin RNA (shRNA) constructs targeting ABCG2 or sham and EsC2 had relatively high basal expression of ABCG2, sequences (vector control) according to manufacturer's pro- whichwasincreased2-to4-foldandapproximately8-fold, tocols (Sigma). Stable cells were selected and expanded in respectively, by CSC treatment. In contrast, Calu-6 cells had Puromycin (Sigma) after target gene expression was confirmed relatively low level basal expression of ABCG2, which was by qRT-PCR and immunofluorescence techniques. augmented approximately 25- to 30-fold by CSC. Normal SAEC exhibited very low basal levels and minimal (3- to 4- Proliferation, cell migration, and soft agar clonogenicity fold) induction of ABCG2. CSC-mediated induction of assays ABCG2 was observed across numerous additional non– Submitted as Supplementary Methods. small cell lung cancer lines, but was not observed in a

www.aacrjournals.org Cancer Res; 72(16) August 15, 2012 4179 Zhang et al.

4180 Cancer Res; 72(16) August 15, 2012 Cancer Research Mithramycin and ABCG2

panel of small cell lung cancer lines, all of which had very not be assessed in EsC2 cells, which exhibited poor clonogeni- low levels of endogenous ABCG2 expression (Supplementary city in soft agar. Fig. S1). The lack of validated lines precluded further induction experiments using esophageal cancer cells. CSC increases side population of cultured cancer cells Immunoblot experiments were done to examine ABCG2 Recent studies have indicated that ABCG2 is highly ex- levels in cultured cells before and after CSC exposure. pressed in cancer stem cells, which are enriched within the Representative results are depicted in Fig. 1A, bottom left side population identified by flow cytometry techniques panel. CSC exposure increased ABCG2 levels in lung and (21, 23). As such, additional experiments were done to examine esophageal cancer cells; this phenomenon was less dramatic whether CSC increased side population of cancer cells. Rep- in SAEC. Basal levels of ABCG2 expression in Calu-6 cells and resentative results pertaining to A549 and Calu-6 cells (high SAEC were considerably lower than those observed in A549 and low ABCG2 expressers, respectively), as well as SAEC are and EsC2 cells; detection of ABCG2 in Calu-6 and SAEC depicted in Fig. 2A. Hoechst staining with and without verap- required loading 3-fold more protein lysates than needed to amil revealed a side population fraction of 0.23% in untreated evaluate ABCG2 in A549 and EsC2 cells (200 mgvs.60mg, A549 cells, compared with 0.95% (a 4-fold increase) in A549 respectively). Subsequent flow cytometry experiments cells exposed to CSC for 5 days. The side population fraction in revealed that CSC exposure increased surface expression of untreated Calu-6 cells (0.06%) was much lower than A549 cells ABCG2 in A549, Calu-6, and EsC2 cells (Fig. 1A, bottom right and increased to 0.98% (a 16-fold increase) following CSC panel). This phenomenon was not evident in SAEC, possibly exposure. In contrast, side population fraction of SAEC did attributable to very low basal levels of ABCG2 expression not seem to be increased following CSC exposure. and relatively modest upregulation of ABCG2 following CSC Additional experiments were done to examine whether the exposure. increase in side population coincided with enhanced ABCG2 Additional qRT-PCR experiments were done to assess expression in cancer cells exposed to CSC. Briefly, qRT-PCR timing and duration of ABCG2 induction in cancer cells techniques were used to quantitate ABCG2 expression in mediated by CSC. Results pertaining to Calu-6 and EsC2 sorted side population as well as non–side population fractions cells are depicted in Fig. 1B. In both cell lines, ABCG2 from A549 and Calu-6 cells cultured in normal media with or expression increased significantly 8 to 24 hours following without CSC. Results of this analysis are depicted in Fig. 2B and initiation of CSC exposure. In Calu-6 cells, ABCG2 expres- summarized in Supplementary Table S2. In A549 cells, untreat- sion continued to increase throughout the 10-day smoke ed side population fractions had approximately 2.8-fold higher exposure; in EsC2 cells, ABCG2 expression peaked at 48 ABCG2 expression compared with control non–side popula- hours, plateauing off thereafter. In both cell lines, ABCG2 tion. Following CSC treatment, ABCG2 expression in the A549 expression decreased dramatically and approached pre- side population fraction increased approximately 1.6-fold treatment levels within 1 to 3 days following cessation of (range 1.4–2.8) relative to the control side population fraction; CSC exposure. in contrast, CSC treatment did not seem to increase ABCG2 expression in the non–side population fraction from these ABCG2 modulates proliferation, migration, and cells. The situation was somewhat different for Calu-6 cells. As clonogenicity of cancer cells anticipated, actual ABCG2 levels in side population and non– Additional experiments were carried out to ascertain wheth- side population fractions in these cells were considerably er ABCG2 expression modulated the malignant phenotype of lower than those observed for A549. Furthermore, the relative cultured lung and esophageal cancer cells. Briefly, lentiviral ABCG2 expression level in the side population fraction from shRNA techniques were used to knockdown ABCG2 in A549 untreated Calu-6 cells was quite similar to that observed in the and EsC2 cells exhibiting relatively high-level ABCG2 expres- non–side population fractions, possibly because of the very sion. qRT-PCR and immunoprecipitation experiments con- small side population fraction. Following CSC exposure, firmed significant reduction of ABCG2 expression in ABCG2 ABCG2 expression in the side population fraction increased knockdown cells relative to respective controls (Fig. 1C). approximately 26-fold, whereas ABCG2 expression in the Subsequent experiments showed that knockdown of ABCG2 non–side population fraction increased approximately 3.3-fold significantly decreased proliferation and migration of A549 compared with respective untreated controls. Immunoblot lung cancer cells, and to lesser extent, EsC2 cells (Fig. 1D). In analysis showed that increased ABCG2 expression coincided addition, knockdown of ABCG2 significantly inhibited soft agar with increased ABCG2 levels in side population fractions of clonogenicity of A549 cells (Fig. 1D). This phenomenon could A549 cells following CSC exposure.

Figure 1. Effects of CSC on ABCG2 expression in cultured cells. A, top, qRT-PCR analysis showing dose-dependent induction of ABCG2 expression by CSC in A549, Calu-6, and EsC2 cancer cells, as well as normal SAEC. Bottom left, immunoblot showing increased ABCG2 expression in A549, Calu-6, EsC2, and SAEC cells following CSC exposure. Right, flow cytometry analysis showing increased cell surface ABCG2 expression in A549, Calu-6, and EsC2 cancer cells, but not SAEC following CSC exposure (red, negative control antibody; blue, ABCG2 antibody). B, qRT-PCR analysis of ABCG2 expression in lung and esophageal cancer cells following exposure to and removal of CSC. C, left, qRT-PCR analysis of ABCG2 expression in cells transduced with shRNA-targeting ABCG2 or control shRNA. Right, representative immunoprecipitation experiment depicting decreased ABCG2 expression in A549 cells transduced with shRNA-targeting ABCG2. D, left, MTS assay showing that ABCG2 knockdown inhibits cancer cell proliferation. Middle, scratch assay showing that ABCG2 knockdown diminishes cancer cell migration. Right, ABCG2 knockdown reduces clonogenicity of A549 cells. NM, normal media.

www.aacrjournals.org Cancer Res; 72(16) August 15, 2012 4181 Zhang et al.

4182 Cancer Res; 72(16) August 15, 2012 Cancer Research Mithramycin and ABCG2

Role of Aryl hydrocarbon receptor and Sp1 in ABCG2 elements present in ABCG2-1662. Mutation of the 5 Sp1 sites upregulation by CSC significantly diminished luciferase activity of ABCG2-245 to a PurifiedcarcinogensaswellasHDACinhibitorssuchas level less than that seen for ABCG2 del 198 luc. CSC mediated romidepsin induce ABCG2 expression in cancer cells by aryl luciferase activities of ABCG2-245 following deletion of one of hydrocarbon receptor (AhR) signaling (24, 25). As such, the remaining XRE (ABCG2-XRE del 245 luc) or combined additional experiments were carried out to examine the deletion of this XRE with mutation of the 5 Sp1 sites mechanisms by which CSC induces ABCG2 expression in (ABCG2-Sp1 mut-XRE del 198 luc) were further decreased, lung and esophageal cancer cells. In preliminary experi- approximating those seen for PGL3 basic luc. These data ments, qRT-PCR techniques were used to evaluate ABCG2 suggested that in addition to AhR, Sp1 contributes significantly expression in lung cancer cells exposed to CSC or various to CSC-mediated activation of ABCG2 in cancer cells. AhR agonists for 24 hours. Calu-6 as well as EsC1 cells were chosen for these experiments, as both lines exhibit very low Effects of mithramycin on CSC-mediated induction of levels of endogenous ABCG2 expression and comparable ABCG2 magnitudes of ABCG2 induction following CSC exposure Additional studies were carried out to examine whether (data for EsC1 available on request). The magnitude of mithramycin, a pharmacologic agent that inhibits Sp1-mediated ABCG2 induction mediated by benzopyrene, 3-MC, or TCDD activation of a variety of genes, including DHFR and AhR (27, 28), relative to CSC varied somewhat between the cell lines (Fig. could repress ABCG2 expression in cancer cells. Briefly, cancer 2C; left panel). The AhR antagonist, resveratrol, previously cells were cultured in normal media with or without CSC in the shown to inhibit romidepsin-mediated activation of ABCG2 presence or absence of escalating doses of mithramycin. qRT- (25), only partially abrogated CSC-mediated induction of PCR analysis revealed that 24 hours of treatment with mithra- ABCG2 in these cells (Fig. 2C; right panel). mycin decreased basal levels of ABCG2 and markedly attenuated In additional experiments, shRNA techniques were used to CSC-mediated induction of ABCG2 in lung and esophageal knockdown HDAC6, which is required for activation and cancer cells (Fig. 3B). Downregulation of these genes persisted nuclear transport of AhR in response to tobacco carcinogens for at least 16 hours following removal of mithramycin from (26). qRT-PCR and immunofluorescence experiments con- culture media (data not shown). Additional analysis revealed firmed decreased HDAC6 expression in knockdown cells rel- that mithramycin inhibited basal levels as well as CSC-mediated ative to respective controls transduced with sham sequences upregulation of Sp1 and AhR in these cells. Furthermore, mithra- (Fig. 2D; top panel). Knockdown of HDAC6 only modestly mycin decreased basal as well as CSC-mediated expression of diminished CSC-mediated induction of ABCG2 in A549, EsC1, nuclear factor erythroid related factor 2 (Nrf2), which has been and EsC2 cells (Fig. 2D; bottom panel); the effects of HDAC6 shown recently to modulate ABCG2 expression (29). Immuno- inhibition could not be assessed in Calu-6 cells because of blot experiments showed that mithramycin mediated dose- toxicity of the shRNA construct in these cells. Collectively, dependent decreases in ABCG2, Sp1, AhR, and Nrf2 expression these findings suggested that induction of ABCG2 by CSC was and partially abrogated CSC-mediated increases in levels of not mediated solely by AhR signaling. these transcription factors in A549 and Calu-6 cells (Fig. 3C). In addition to xenobiotic response elements (XRE) that ChIP experiments were done to examine whether mithra- are binding sites for AhR, the ABCG2 promoter contains a mycin modulates basal as well as CSC-mediated levels of number of Specificity protein 1 (Sp1) sites (24) that could Sp1, AhR, and Nrf2 within the ABCG2 promoter. ChIP mediate activation of this gene in response to cigarette smoke. primers encompassed a region containing AhR and Sp1 As such, transient transfection experiments using ABCG2 pro- recognition elements adjacent to the transcription start site; moter–reporter constructs were carried out to examine poten- additional ChIP primers amplified a region containing the tial roles of AhR and Sp1 in mediating ABCG2 activation by CSC. Nrf2 antioxidant response element within the ABCG2 pro- Representative results of these experiments are depicted in Fig. moter (29). Results pertaining to Calu-6 cells (simultaneous- 3A. Following CSC exposure, luciferase activity of ABCG2-1662 ly processed for qRT-PCR experiments in Fig. 3B) are containing 4 XRE sites was 4-fold higher than pGL3 Basic Luc depicted in Fig. 3D. CSC-mediated activation of ABCG2 control in Calu-6 cells. CSC mediated luciferase activity of coincided with increased levels of RNA polymerase II (pol ABCG2-1662 del 198, in which one of the XRE sites had been II) and H3K9Ac, with decreased levels of H3K9Me3 (activa- deleted, was 50% lower than that observed for ABCG2-1662. tion and repression marks, respectively) within the ABCG2 CSC-mediated luciferase activity was also diminished following promoter. CSC induced recruitment of Sp1, AhR, and Nrf2 to transfection with ABCG2-245 containing 2 XRE sites and 5 Sp1 the ABCG2 promoter; these results were most dramatic for

Figure 2. CSC modulates ABCG2 expression and side population fraction in lung and esophageal cancer cells. A, representative flow cytometry analysis showing that CSC increases side population fraction in A549 and Calu-6 lung cancer cells, but not SAEC. B, left, qRT-PCR analysis of ABCG2 expression in side population and non–side population fractions in A549 and Calu-6 cells cultured in normal media CSC. Right, representative immunoblot showing that CSC exposure increases ABCG2 protein levels in side population. C, left, qRT-PCR analysis of ABCG2 expression in Calu-6 and EsC1 cells following exposure to CSC or purified carcinogens, which activate AhR signaling. Right, qRT-PCR analysis showing relatively modest inhibition of CSC-mediated induction of ABCG2 by resveratrol. D, top, qRT-PCR analysis depicting shRNA-mediated knockdown of HDAC6 in lung and esophageal cancer cells. Bottom, qRT-PCR analysis showing effects of HDAC6 knockdown on CSC-mediated induction of ABCG2 in lung and esophageal cancer cells. NM, normal media; SP, side population.

www.aacrjournals.org Cancer Res; 72(16) August 15, 2012 4183 Zhang et al.

Figure 3. Role of AhR, Sp1, and Nrf2 in ABCG2 activation by CSC. A, luciferase activity of ABCG2 promoter reporter constructs following transient transfection into Calu-6 cells. Relative to full length promoter (ABCG2-1662-LUC), luciferase activities of ABCG2 promoter constructs decreased following serial deletions or mutations of XRE and Sp1 elements. B, qRT-PCR analysis of ABCG2, AhR, Sp1, and Nrf2 expression in A549 and Calu-6 cells cultured in normal media with or without mithramycin in the presence or absence of CSC. C, immunoblot analysis of ABCG2, SP1, AHR Nrf2 levels in A549 and Calu-6 cells cultured in the presence or absence of mithramycin with or without CSC. D, quantitative ChIP analysis of the ABCG2 promoter region in Calu-6 cells cultured in normal media with or without CSC in the presence or absence of mithramycin. The dose of mithramycin for ChIP was optimized in preliminary experiments. NM, normal media.

4184 Cancer Res; 72(16) August 15, 2012 Cancer Research Mithramycin and ABCG2

Figure 4. Effects of mithramycin in lung and esophageal cancer cells. A, MTS assays depicting effects of 24-hour mithramycin exposure on proliferation of lung and esophageal cancer cells. B, flow cytometry analysis showing that mithramycin decreases side population in A549 cells. C, effects of intraperitoneal mithramycin (1 or 2 mg/kg Monday–Wednesday–Friday 3) on growth of established subcutaneous A549 xenografts. Left, tumor volumes; middle, tumor masses; right, effects of mithramycin on body mass. D, representative tissue sections from A549 xenografts from control and mithramycin-treated mice. Top, hematoxylin and eosin stains. Bottom, representative immunofluorescence results depicting ABCG2 expression in control tumors and xenografts from mice treated with mithramycin. NM, normal media.

www.aacrjournals.org Cancer Res; 72(16) August 15, 2012 4185 Zhang et al.

Figure 5. Microarray analysis of mithramycin effects on gene expression-cultured A549 and Calu-6 cells and A549 xenografts related to respective controls. A, left top, principal component analysis (PCA) showing highly reproducible results of triplicate samples for each cell line and each treatment condition. Bottom left, Venn diagram showing overlap of genes simultaneously modulated in A549 and Calu-6 cells under 2 in vitro exposure conditions. Right, heat map depicting 1,258 differentially expressed genes modulated by mithramycin. A marked dose-dependent alteration of gene expression profiles was observed in these cells (triplicate samples). B, left, heat map revealing dose-dependent modulation of gene expression by mithramycin in tumor xenografts. Right top, PCA showing highly reproducible results of triplicate samples (derived from 9 tumors for such conditions); bottom, Venn diagram depicting overlap of genes modulated in vivo under both mithramycin doses. C, Venn diagrams showing overlap of differentially expressed genes in cultured A549 and Calu-6 cells exposed to either 50 or 200 nmol/L mithramycin for 24 hours or A549 tumor xenografts from mice treated with 1 or 2 mg/kg mithramycin.

Sp1 and were consistent with aforementioned luciferase Effects of mithramycin on proliferation and experiments. Mithramycin diminished CSC-mediated occu- tumorigenicity of cancer cells pancy of these transcription factors within the ABCG2 Additional experiments were undertaken to examine the promoter; these effects coincided with appropriate altera- effects of mithramycin on proliferation and tumorigenicity tions in pol II, H3K9Ac, and H3K9Me3. of lung and esophageal cancer cells. Results of this analysis

4186 Cancer Res; 72(16) August 15, 2012 Cancer Research Mithramycin and ABCG2

Table 1. Genes differentially expressed in vitro and in vivo by mithramycin

A. Ingenuity canonical pathways–in vitro Differentially expressed genes Molecular mechanisms of cancer MAPK1, SMAD3, ARHGEF7, LRP6, BMPR2, MAP3K5, RBL1, EP300, PTK2, BMPR1A, SOS1, TGFB2, PRKCE, AKT3, GSK3B, HIPK2, PRKD1, PRKCA, PMAIP1, GNA12, CREBBP, SMAD6, GNAQ, CDK6, TCF3, FZD8, BMPR1B, PLCB4, NF1, IRS1, PIK3CB, ARHGEF18, CFLAR TGF-b signalinga MAPK1, SMAD3, SKI, CREBBP, SMAD6, BMPR2, SMURF1, EP300, BMPR1B, BMPR1A, SOS1, TGFB2, SMURF2 TR/RXR activation SLC2A1, NCOA3, EP300, LDLR, SREBF1, AKT3, STRBP, NCOR1, ACACA, PIK3CB, NCOR2, TBL1XR1, RXRA HER-2 signaling in breast cancer SOS1, CDK6, PRKCE, AKT3, PIK3CB, MAP3K5, GSK3B, PARD3, PRKD1, PRKCA, EGFR NSCLC signaling STK4, MAPK1, SOS1, CDK6, AKT3, PDPK1, PIK3CB, RXRA, PRKCA, EGFR mTOR signalinga MAPKAP1, MAPK1, PPP2R5C, PPP2CA, PDPK1, RICTOR, EIF4E, IRS1, PRKAA1, EIF3A, PRKCE, AKT3, PIK3CB, PPP2R5E, PRKD1, PRKCA BMP signaling pathwaya BMPR1B, MAPK1, BMPR1A, CREB1, SOS1, CREBBP, SMAD6, BMPR2, SMURF1, ATF2 Cyclins and cell-cycle regulation HDAC4, PPP2R5C, PPP2CA, CDK6, TGFB2, GSK3B, PPP2R5E Wnt/b-catenin signalinga PPP2R5C, PPP2CA, CSNK1G2, TGFBR3, CREBBP, LRP6, GNAQ, CSNK1A1, TCF3, EP300, FZD8, TGFB2, AKT3, GSK3B, PPP2R5E, TCF7L2 Human embryonic stem cell pluripotencya SMAD3, SMAD6, BMPR2, PDPK1, TCF3, FZD8, BMPR1B, BMPR1A, TGFB2, AKT3, PIK3CB, GSK3B, TCF7L2 DNA methylation and transcriptional MECP2, ARID4B, SAP18 repression signaling Role of Oct4 in mammalian embryonic C3orf63, CCNF, JARID2, IGF2BP1, WWP2 stem cell pluripotencya Role of NANOG in mammalian embryonic FZD8, BMPR1B, MAPK1, BMPR1A, SOS1, BMPR2, AKT3, PIK3CB, GSK3B stem cell pluripotencya Notch signalinga MAML2, NUMB, MAML3, JAG1 NF-kB signaling TRAF3, TNFRSF1A, TGFBR3, CREBBP, BMPR2, MAP4K4, MALT1, EP300, BMPR1B, BMPR1A, IGF1R, AKT3, PIK3CB, GSK3B, EGFR p53 signaling PMAIP1, STAG1, AKT3, PIK3CB, GSK3B, HIPK2, EP300

B. Ingenuity canonical pathways–in vivo Differentially expressed genes Molecular mechanisms of cancer MAP2K6, PMAIP1, TCF4, TGFBR1, BMPR2, CRK, RBL1, TGFBR2, FZD8, BMPR1B, HIPK2, BRCA1, PRKCA TGF-b signalinga MAP2K6, TGFBR2, BMPR1B, TGFBR1, BMPR2 TR/RXR activation LDLR, TBL1XR1, RXRA, NCOA3 HER-2 signaling in breast cancer PARD3, PRKCA NSCLC signaling STK4, RXRA, PRKCA mTOR signalinga PRKAB2, PPP2R5C, PPP2R5E, PDGFC, PRKCA BMP signaling pathwaya BMPR1B, BMPR2 Cyclins and cell-cycle regulation HDAC4, PPP2R5C, CCNB2, PPP2R5E Wnt/b-catenin signalinga TGFBR2, FZD8, TCF4, TGFBR1, PPP2R5C, PPP2R5E Human embryonic stem cell pluripotencya TGFBR2, FZD8, BMPR1B, TCF4, TGFBR1, BMPR2, PDGFC DNA methylation and transcriptional SAP30 repression signaling Role of Oct4 in mammalian JARID2, BRCA1 embryonic stem cell pluripotencya Role of NANOG in mammalian FZD8, BMPR1B, BMPR2 embryonic stem cell pluripotencya Notch signalinga MAML2 NF-kB signaling MAP2K6, TGFBR2, BMPR1B, TGFBR1, BMPR2 p53 signaling PMAIP1, TP53INP1, HIPK2, BRCA1

aStem cell–related signaling pathway.

www.aacrjournals.org Cancer Res; 72(16) August 15, 2012 4187 Zhang et al.

Table 2. Top 100 genes modulated in lung cancer xenografts by both doses of intraperitoneal mithramycin

Downregulated genes Upregulated genes

Probeset ID Gene symbol Fold change Fold change Probeset ID Gene symbol Fold change Fold change (1 mg/kg vs. C) (2 mg/kg vs. C) (1 mg/kg vs. C) (2 mg/kg vs. C) 225846_at ESRP1 –3.7 –8.4 237737_at LOC727770 7.6 17.2 204283_at FARS2 –2.5 –7.7 223218_s_at NFKBIZ 3.8 13.2 214106_s_at GMDS –6.6 –7.1 213348_at CDKN1C 4.8 11.4 203474_at IQGAP2 –2.0 –6.5 229331_at SPATA18 4.2 11.1 213432_at MUC5B –3.4 –6.5 1557765_at LOC643401 2.8 9.9 223278_at GJB2 –2.0 –6.2 232164_s_at EPPK1 3.7 9.8 219077_s_at WWOX –9.8 –5.9 1554195_a_at C5orf46 3.3 8.6 204623_at TFF3 –2.6 –5.9 204803_s_at RRAD 2 8.2 232603_at DCDC5 –2.7 –5.9 212667_at SPARC 3.7 7.2 212942_s_at KIAA1199 –5.4 –5.7 209283_at CRYAB 2.4 7 219410_at TMEM45A –2.7 –5.6 213764_s_at MFAP5 2.6 6.3 227059_at GPC6 –4.4 –5.0 211340_s_at MCAM 3.2 6.2 211596_s_at LRIG1 –2.2 –4.8 221276_s_at SYNC 2.9 5.8 225864_at FAM84B –3.0 –4.7 229275_at IGFN1 2.2 5.5 218788_s_at SMYD3 –2.9 –4.7 227870_at IGDCC4 2.3 5.2 204857_at MAD1L1 –4.6 –4.7 217562_at FAM5C 2.3 4.8 204066_s_at AGAP1 –5.6 –4.6 203889_at SCG5 3.8 4.8 1552673_at RFX6 –3.0 –4.5 203477_at COL15A1 2.7 4.6 218872_at TESC –2.5 –4.4 207173_x_at CDH11 2.3 4.5 204646_at DPYD –4.3 –4.4 1563498_s_at SLC25A45 2.5 4.5 224926_at EXOC4 –2.7 –4.3 201860_s_at PLAT 3.4 4.5 213243_at VPS13B –3.2 –4.3 229711_s_at MDM2 2.3 4.5 203908_at SLC4A4 –3.1 –4.3 223540_at PVRL4 2.7 4.4 229030_at CAPN8 –2.1 –4.2 200974_at ACTA2 3.1 4.4 203699_s_at DIO2 –3.2 –4.2 236656_s_at LOC100130506 2.1 4.3 206010_at HABP2 –2.4 –4.1 218611_at IER5 2.6 4.3 227202_at CNTN1 –2.1 –4.0 215033_at TM4SF1 2.3 4.1 204690_at STX8 –2.8 –4.0 209462_at APLP1 2.2 4 212692_s_at LRBA –2.1 –4.0 203002_at AMOTL2 2.2 4 219732_at RP11-35N6.1 –3.1 –3.9 1552309_a_at NEXN 2.1 3.7 202068_s_at LDLR –2.4 –3.9 201939_at PLK2 2.4 3.6 225081_s_at CDCA7L –2.1 –3.8 225604_s_at GLIPR2 2.9 3.6 212338_at MYO1D –2.4 –3.8 205483_s_at ISG15 2.1 3.5 238440_at CLYBL –2.4 –3.7 205691_at SYNGR3 2.2 3.5 201622_at SND1 –2.9 –3.7 236561_at TGFBR1 2 3.5 201911_s_at FARP1 –2.2 –3.7 223484_at C15orf48 2.2 3.4 211429_s_at SERPINA1 –3.0 –3.7 206857_s_at FKBP1B 2.4 3.4 214303_x_at MUC5AC –2.8 –3.6 225973_at TAP2 2.5 3.4 228570_at BTBD11 –2.2 –3.6 206907_at TNFSF9 2.3 3.4 219307_at PDSS2 –2.0 –3.6 225912_at TP53INP1 2.2 3.4 212093_s_at MTUS1 –2.1 –3.6 202284_s_at CDKN1A 2.3 3.2 228038_at SOX2 –3.5 –3.5 36711_at MAFF 2.1 3.2 243681_at SHANK2 –4.0 –3.5 203153_at IFIT1 2 3.2 212098_at LOC151162 /// –2.1 –3.5 228531_at SAMD9 2.2 3.1 MGAT5 220488_s_at BCAS3 –2.8 –3.5 204430_s_at SLC2A5 2 3.1 203006_at INPP5A –2.8 –3.4 228293_at DEPDC7 2.1 3 (Continued on the following page)

4188 Cancer Res; 72(16) August 15, 2012 Cancer Research Mithramycin and ABCG2

Table 2. Top 100 genes modulated in lung cancer xenografts by both doses of intraperitoneal mithramycin (Cont'd )

Downregulated genes Upregulated genes

Probeset ID Gene symbol Fold change Fold change Probeset ID Gene symbol Fold change Fold change (1 mg/kg vs. C) (2 mg/kg vs. C) (1 mg/kg vs. C) (2 mg/kg vs. C) 223681_s_at INADL –3.2 –3.4 207813_s_at FDXR 2.5 3 205603_s_at DIAPH2 –3.0 –3.3 209277_at TFPI2 2 2.9 228486_at SLC44A1 –2.6 –3.3 209183_s_at C10orf10 2.4 2.8 207414_s_at PCSK6 –2.1 –3.2 204483_at ENO3 2.9 2.8

are shown in Fig. 4. MTS experiments showed that 24 hours in Table 1A; 8 of these canonical pathways are related to stem of mithramycin exposure mediated profound growth-inhib- cell signaling. itory effects in cultured lung and esophageal cancer cells Additional microarray experiments were carried out to (Fig. 4A). Apo-BrdUrd experiments revealed no appreciable examine effects of mithramycin in A549 xenografts (9 each increase in apoptosis in mithramycin-treated cells, suggest- from drug-treated or control mice). Similar to what was ing that the growth-inhibitory effects of this agent were observed following in vitro drug treatment, mithramycin medi- mediated by cell-cycle arrest rather than apoptosis (data not ated highly reproducible, dose-dependent alterations in gene shown). Additional flow cytometry experiments using A549 expression in A549 tumor xenografts (Fig. 5B; left and top right cells suggested that mithramycin decreased side population panel). Using criteria of fold change greater than 2 and P less fraction (Fig. 4B). than 0.05 for drug treatment versus control, 351 and 1,896 Further experiments were carried out to examine if mithra- genes were differentially expressed in xenografts from mice mycin diminished growth of established tumor xenografts. As receiving mithramycin at 1 and 2 mg/kg, respectively, relative shown in Fig. 4C, mithramycin mediated significant dose- to control tumors (Fig. 5B; bottom right panel). A total of 299 dependent growth inhibition of A549 xenografts without genes were modulated by mithramycin under both doses, 100 appreciable systemic toxicities, such as decreased activity, of which are listed in Table 2. ABCG2 was downregulated by skin changes, or significant weight loss. Histopathologic anal- more than 2-fold in xenografts from mice receiving 2 mg/kg but ysis revealed that tumors from mithramycin-treated mice were not 1 mg/kg mithramycin. All 8 of the stem cell–related less glandular in appearance with somewhat less stroma. pathways modulated in vitro by mithramycin were also tar- Furthermore, tumors from mice treated with 2 mg/kg mithra- geted in tumor xenografts by systemic drug treatment, albeit to mycin had 50% fewer mitoses relative to control tumors (data a somewhat lesser degree (Table 1B). A similar phenomenon not shown). Immunofluorescence experiments confirmed that was observed regarding the remaining 8 canonical pathways mithramycin decreased ABCG2 expression in tumor xeno- listed in Table 1A. A variety of networks regulating intracel- grafts (Fig. 4D). leular signaling, DNA damage response, chromatin remodel- Because Sp1 has diverse potential targets (30), Affymetrix ing, and chromosomal replication were inhibited in A549 microarray experiments were conducted to examine global tumor xenografts following mithramycin treatment (Supple- gene expression profiles in Calu-6 and A549 cultured in normal mentary Fig. S3). media, with or without mithramycin for 24 hours. Under Further analysis was undertaken to correlate in vivo effects conditions potentially achievable in clinical settings (50–200 of mithramycin with in vitro drug exposures. A progressive nmol/L 24 hours), mithramycin mediated dramatic dose- dose-dependent increase in genes commonly regulated in vitro dependent alterations in gene expression in cultured lung and in vivo by mithramycin was observed (Fig. 5C). Two to 10% cancer cells. Highly reproducible results were noted among (average 5%) of genes modulated in vitro overlapped with 13% cell lines and within treatment groups (Fig. 5A, top left panel). to 24% (average 18%) of genes altered by in vivo mithramycin Using stringent criteria of fold change greater than 3 and across various treatment comparisons. A total of 337 genes adjusted P less than 0.01 for drug treatment versus control, were simultaneously modulated in cultured A549 and Calu-6 1,582 and 3,771 genes were simultaneously modulated in A549 cells following 200 nmol/L mithramycin and A549 xenografts and Calu-6 cells following 50 and 200 nmol/L mithramycin from mice receiving 2 mg/kg mithramycin i.p.. The top 100 exposures, respectively (Fig. 5A; bottom left panel). A total of repressed and all 43 upregulated genes from these 337 com- 1,258 genes were commonly regulated by mithramycin across 2 monly regulated genes are listed in Supplementary Table S3. cell lines and 2 drug concentrations; the majority of differen- Top molecular and cellular functions mediated by these 337 tially regulated genes were downregulated in both cell lines genes included cell-cycle progression, gene expression, stem (Fig. 5A right panel; Supplementary Fig. S2). Sixteen of the top cell pluripotency, cellular morphology, and death. Represen- cancer-associated pathways, which were downregulated in tative networks mediating chromatin remodeling and TGF- cultured lung cancer cells by mithramycin are listed b/BMP-3 signaling are depicted in Supplementary Fig. S4.

www.aacrjournals.org Cancer Res; 72(16) August 15, 2012 4189 Zhang et al.

Discussion impairing maturation of hematopoietic progenitor cells (36). Of particular relevance regarding our current study are recent The vast majority of lung and esophageal carcinomas are reports showing that increased expression of ABCG2 correlates directly attributable to tobacco abuse (3, 4). Cigarette smoking with chemoresistance and stem-like phenotype of lung and not only facilitates initiation and preclinical progression of esophageal carcinomas (21, 23, 37–39) and decreased survival lung and esophageal cancers but also enhances treatment of patients with these neoplasms (40–42). resistance and dissemination of established malignancies, Despite the fact that knockdown of ABCG2 decreased pro- thereby decreasing overall survival of patients with these liferation, migration, and clonogenicity of lung and esophageal neoplasms (8, 10, 11, 31). Delineation of the mechanisms by cancer cells, our current data do not establish nor imply that which cigarette smoke promotes proliferation of lung and downregulation of ABCG2 is the primary mechanism by which esophageal carcinomas may facilitate development of more mithramycin inhibits proliferation and tumorigenicity of these efficacious treatment regimens for these malignancies. cancer cells. Indeed, our microarray analysis revealed that Our previous studies have shown that cigarette smoke mithramycin significantly downregulated hundreds of genes increases the malignant phenotype of lung and esophageal mediating stem cell signaling, cell-cycle progression, chroma- cancer cells, in part, by upregulating genes, which mediate tin remodeling, and DNA damage response. Although experi- stem cell phenotype (12, 13). Consistent with these observa- ments are underway to examine this issue, it seems unlikely tions, our current experiments revealed that cigarette smoke that constitutive overexpression of ABCG2 would significantly mediates time- and dose-dependent upregulation of ABCG2 in diminish mithramycin-mediated cytotoxicity in cultured can- lung and esophageal cancer cells; this phenomenon seemed to cer cells, unless this agent is a substrate for ABCG2. In all be considerably less pronounced in cultured normal aerodi- likelihood, the antitumor effects of mithramycin are mediated gestive tract epithelia. As expected, CSC-mediated upregula- by direct inhibition of Sp1 binding to promoters of master tion of ABCG2 in cancer cells coincided with increased side genes regulating diverse cellular functions, with subsequent population fraction; these findings suggest, although certainly repression of downstream targets by direct as well as indirect do not prove, that cigarette smoke promotes expansion of mechanisms (30, 43). pluripotent tumor cells (21). Further analysis revealed that Mithramycin, a polyauroleic acid isolated from strepto- CSC-mediated activation of ABCG2 coincided with recruit- myces, was initially evaluated as a chemotherapeutic agent ment of AhR and Nrf2, as well as Sp1 to the ABCG2 promoter. in cancer patients during the 1960s and 70s (44), but was Mithramycin repressed basal as well as CSC-mediated induc- discontinued because of excessive systemic toxicities (45). tion of ABCG2; these effects may be attributable to direct Recently there has been renewed interest in clinical devel- inhibition of binding of Sp1 to DNA (27) and decreased opment of mithramycin and its derivatives because of their expression of Sp1, as well as AhR and Nrf2 (2 potential targets ability to specifically inhibit binding of Sp1 to GC-rich DNA of Sp1; ref. 28), which are known to activate ABCG2. In addition, (27) and downregulate numerous genes mediating prolifer- mithramycin dramatically decreased proliferation and tumor- ation, invasion, and metastasis of cancer cells (46–48). Of igenicity of cancer cells. Growth inhibition coincided with particular interest in this regard are recent studies indicat- downregulation of ABCG2, as well as numerous other genes ing that currently approved agents such as cyclo-oxygenase mediating "stemness," proliferation, and metastatic potential inhibitors markedly enhance mithramycin-mediated inhibi- of cancer cells. To the best of our knowledge, these experiments tion of Sp1 expression/activity in cancer cells (49). Such represent the most comprehensive analysis of mechanisms combinational strategies could enable reduction of mithra- regulating ABCG2 expression in thoracic malignancies follow- mycin doses and, possibly, decrease systemic toxicities in ing exposure to cigarette smoke and are the first to show the clinical settings. potential of mithramycin for targeting ABCG2 as well as other Our current findings have direct translational implications stem cell–related genes in thoracic malignancies. with regard to evaluation of mithramycin in patients with ABCG2 (BCRP) is a member of the ATP-binding cassette thoracic malignancies. Extrapolation of data from previous (ABC) transporters, which functions as a xenobiotic pump in animal studies (50) suggests that mithramycin levels achieved many normal tissues (32). Its substrates include numerous in our xenograft experiments were in the 50 to 200 nmol/L environmental toxins as well as chemotherapeutic agents. Our range over 24 hours; these exposure conditions, which closely experiments showed that knockdown of ABCG2 inhibits pro- approximated those used for our in vitro experiments, are liferation, migration, and clonogenicity of lung and esophageal potentially achievable using previous mithramycin dosing cancer cells; these findings suggest that ABCG2 modulates schedules in humans (44). Collectively, our findings support intracellular processes other than extrusion of xenobiotics. the clinical evaluation of mithramycin as a strategy to repress Several recent studies suggest that ABCG2 is a critical mediator ABCG2 and inhibit signaling pathways mediating "stemness" in of stem cell homeostasis. For example, ABCG2 is an essential thoracic malignancies. upstream mediator of sonic-hedgehog signaling, which has been implicated in maintenance of stemness (33). Further- Disclosure of Potential Conflicts of Interest more, ABCG2 binds to heme, thereby diminishing intracellular No potential conflicts of interest were disclosed. porphyrin levels, rendering stem cells resistant to hypoxia (34). ABCG2 Constitutive expression of protects cardiac stem cells Authors' Contributions from oxidative stress (35) and enhances expansion, although Conception and design: M. Zhang, A. Mathur, M. Rao, D.S. Schrump.

4190 Cancer Res; 72(16) August 15, 2012 Cancer Research Mithramycin and ABCG2

Development of methodology: M. Zhang, Y. Zhang, S. Xi, R.T. Ripley, I. Avital, P. Administrative, technical, or material support (i.e., reporting or orga- Fetsch, H. Mani, X. Li, D.S. Schrump nizing data, constructing databases): M. Zhang, A. Mathur, J.A. Hong, D.S. Acquisition of data (provided animals, acquired and managed Schrump patients, provided facilities, etc.): M. Zhang, A. Mathur, S. Xi, S. Atay, J. Study supervision: M. Rao, D.S. Schrump A. Hong, N. Datrice, C.D. Kemp, R.T. Ripley, G. Wiegand, P. Fetsch, X. Li, D.S. Schrump The costs of publication of this article were defrayed in part by the payment of Analysis and interpretation of data (e.g., statistical analysis, biostatistics, page charges. This article must therefore be hereby marked advertisement in computational analysis): M. Zhang, A. Mathur, S. Xi, J.A. Hong, C.D. Kemp, I. accordance with 18 U.S.C. Section 1734 solely to indicate this fact. Avital, H. Mani, D. Zlott, S.E. Bates, X. Li, M. Rao, D.S. Schrump Writing, review, and/or revision of the manuscript: M. Zhang, A. Mathur, J. A. Hong, N. Datrice, T. Upham, R.T. Ripley, I. Avital, H. Mani, S.E. Bates, X. Li, D.S. Received December 12, 2011; revised May 17, 2012; accepted May 25, 2012; Schrump published OnlineFirst July 2, 2012.

References 1. Jemal A, Bray F, Center MM, Ferlay J, Ward E, Forman D. Global cancer 19. Liu F, Killian JK, Yang M, Walker RL, Hong JA, Zhang M, et al. statistics. CA Cancer J Clin 2011;61:69–90. Epigenomic alterations and gene expression profiles in respiratory 2. Siegel R, Ward E, Brawley O, Jemal A. Cancer statistics, 2011: the epithelia exposed to cigarette smoke condensate. Oncogene 2010;29: impact of eliminating socioeconomic and racial disparities on prema- 3650–64. ture cancer deaths. CA Cancer J Clin 2011;61:212–36. 20. Rao M, Chinnasamy N, Hong JA, Zhang Y, Zhang M, Xi S, et al. 3. Smoking-attributable mortality, years of potential life lost, and pro- Inhibition of histone lysine methylation enhances cancer-testis antigen ductivity losses–United States, 2000–2004. MMWR Morb Mortal Wkly expression in lung cancer cells: implications for adoptive immuno- Rep 2008;57:1226–8. therapy of cancer. Cancer Res 2011;71:4192–204. 4. Zhai R, Chen F, Liu G, Su L, Kulke MH, Asomaning K, et al. Interactions 21. Ho MM, Ng AV, Lam S, Hung JY. Side population in human lung cancer among genetic variants in apoptosis pathway genes, reflux symptoms, cell lines and tumors is enriched with stem-like cancer cells. Cancer body mass index, and smoking indicate two distinct etiologic patterns Res 2007;67:4827–33. of esophageal adenocarcinoma. J Clin Oncol 2010;28:2445–51. 22. Ding XW, Wu JH, Jiang CP. ABCG2: a potential marker of stem cells and 5. Oppeltz RF, Jatoi I. Tobacco and the escalating global cancer burden. novel target in stem cell and cancer therapy. Life Sci 2010;86:631–7. J Oncol 2011;2011:408104. 23. Huang D, Gao Q, Guo L, Zhang C, Jiang W, Li H, et al. Isolation and 6. Kozower BD, Sheng S, O'Brien SM, Liptay MJ, Lau CL, Jones DR, et al. identification of cancer stem-like cells in esophageal carcinoma cell STS database risk models: predictors of mortality and major morbidity lines. Stem Cells Dev 2009;18:465–73. for lung cancer resection. Ann Thorac Surg 2010;90:875–81. 24. Tompkins LM, Li H, Li L, Lynch C, Xie Y, Nakanishi T, et al. A novel 7. Wright CD, Kucharczuk JC, O'Brien SM, Grab JD, Allen MS. Predictors xenobiotic responsive element regulated by aryl hydrocarbon receptor of major morbidity and mortality after esophagectomy for esophageal is involved in the induction of BCRP/ABCG2 in LS174T cells. Biochem cancer: a Society of Thoracic Surgeons General Thoracic Surgery Pharmacol 2010;80:1754–61. Database risk adjustment model. J Thorac Cardiovasc Surg 2009;137: 25. To KK, Robey R, Zhan Z, Bangiolo L, Bates SE. Upregulation of ABCG2 587–95. by romidepsin via the aryl hydrocarbon receptor pathway. Mol Cancer 8. Nguyen SK, Masson-Cote L, Fortin A, Dagnault A. Influence of smok- Res 2011;9:516–27. ing status on treatment outcomes after post-operative radiation ther- 26. Kekatpure VD, Dannenberg AJ, Subbaramaiah K. HDAC6 modulates apy for non-small-cell lung cancer. Radiother Oncol 2010;96:89–93. Hsp90 chaperone activity and regulates activation of aryl hydrocarbon 9. Janjigian YY, McDonnell K, Kris MG, Shen R, Sima CS, Bach PB, et al. receptor signaling. J Biol Chem 2009;284:7436–45. Pack-years of cigarette smoking as a prognostic factor in patients with 27. Blume SW, Snyder RC, Ray R, Thomas S, Koller CA, Miller DM. stage IIIB/IV nonsmall cell lung cancer. Cancer 2010;116:670–5. Mithramycin inhibits SP1 binding and selectively inhibits transcrip- 10. Abrams JA, Lee PC, Port JL, Altorki NK, Neugut AI. Cigarette smoking tional activity of the dihydrofolate reductase gene in vitro and in vivo.J and risk of lung metastasis from esophageal cancer. Cancer Epidemiol Clin Invest 1991;88:1613–21. Biomarkers Prev 2008;17:2707–13. 28. Tang T, Lin X, Yang H, Zhou L, Wang Z, Shan G, et al. Overexpression 11. Shitara K, Matsuo K, Hatooka S, Ura T, Takahari D, Yokota T, et al. of antioxidant enzymes upregulates aryl hydrocarbon receptor expres- Heavy smoking history interacts with chemoradiotherapy for esoph- sion via increased Sp1 DNA-binding activity. Free Radic Biol Med ageal cancer prognosis: a retrospective study. Cancer Sci 2010;101: 2010;49:487–92. 1001–6. 29. Singh A, Wu H, Zhang P, Happel C, Ma J, Biswal S. Expression of 12. Hussain M, Rao M, Humphries AE, Hong JA, Liu F, Yang M, et al. ABCG2 (BCRP) is regulated by Nrf2 in cancer cells that confers side Tobacco smoke induces polycomb-mediated repression of Dickkopf- population and chemoresistance phenotype. Mol Cancer Ther 1 in lung cancer cells. Cancer Res 2009;69:3570–8. 2010;9:2365–76. 13. Xi S, Yang M, Tao Y, Xu H, Shan J, Inchauste S, et al. Cigarette smoke 30. Safe S, Abdelrahim M. Sp transcription factor family and its role in induces C/EBP-beta-mediated activation of miR-31 in normal human cancer. Eur J Cancer 2005;41:2438–48. respiratory epithelia and lung cancer cells. PLoS One 2010;5:e13764. 31. Waller LL, Weaver KE, Petty WJ, Miller AA. Effects of continued 14. Dovey JS, Zacharek SJ, Kim CF, Lees JA. Bmi1 is critical for lung tobacco use during treatment of lung cancer. Expert Rev Anticancer tumorigenesis and bronchioalveolar stem cell expansion. Proc Natl Ther 2010;10:1569–75. Acad Sci U S A 2008;105:11857–62. 32. Ni Z, Bikadi Z, Rosenberg MF, Mao Q. Structure and function of the 15. Pereira CF, Piccolo FM, Tsubouchi T, Sauer S, Ryan NK, Bruno L, et al. human breast cancer resistance protein (BCRP/ABCG2). Curr Drug ESCs require PRC2 to direct the successful reprogramming of differ- Metab 2010;11:603–17. entiated cells toward pluripotency. Cell Stem Cell 2010;6:547–56. 33. Balbuena J, Pachon G, Lopez-Torrents G, Aran JM, Castresana JS, 16. Espada J, Calvo MB, Diaz-Prado S, Medina V. Wnt signalling and Petriz J. ABCG2 is required to control the sonic hedgehog pathway in cancer stem cells. Clin Transl Oncol 2009;11:411–27. side population cells with stem-like properties. Cytometry A 2011;79: 17. Luo F, Hamoudi R, Brooks DG, Patek CE, Arends MJ. Stem cell gene 672–83. expression changes induced specifically by mutated K-ras. Gene Expr 34. Krishnamurthy P, Ross DD, Nakanishi T, Bailey-Dell K, Zhou S, Mercer 2007;14:101–15. KE, et al. The stem cell marker Bcrp/ABCG2 enhances hypoxic cell sur- 18. Kim J, Woo AJ, Chu J, Snow JW, Fujiwara Y, Kim CG, et al. A Myc vival through interactions with heme. J Biol Chem 2004;279:24218–25. network accounts for similarities between embryonic stem and cancer 35. Martin CM, Ferdous A, Gallardo T, Humphries C, Sadek H, Caprioli A, cell transcription programs. Cell 2010;143:313–24. et al. Hypoxia-inducible factor-2alpha transactivates Abcg2 and

www.aacrjournals.org Cancer Res; 72(16) August 15, 2012 4191 Zhang et al.

promotes cytoprotection in cardiac side population cells. Circ Res 43. Reed BD, Charos AE, Szekely AM, Weissman SM, Snyder M. Genome- 2008;102:1075–81. wide occupancy of SREBP1 and its partners NFY and SP1 reveals 36. Ahmed F, Arseni N, Glimm H, Hiddemann W, Buske C, Feuring-Buske novel functional roles and combinatorial regulation of distinct classes M. Constitutive expression of the ATP-binding cassette transporter of genes. PLoS Genet 2008;4:e1000133. ABCG2 enhances the growth potential of early human hematopoietic 44. Kofman S, Eisenstein R. Mithramycin in the treatment of disseminated progenitors. Stem Cells 2008;26:810–8. cancer. Cancer Chemother Rep 1963;32:77–96. 37. Bertolini G, Roz L, Perego P, Tortoreto M, Fontanella E, Gatti L, et al. 45. Kennedy BJ. Metabolic and toxic effects of mithramycin during tumor Highly tumorigenic lung cancer CD133þ cells display stem-like fea- therapy. Am J Med 1970;49:494–503. tures and are spared by cisplatin treatment. Proc Natl Acad Sci U S A 46. Previdi S, Malek A, Albertini V, Riva C, Capella C, Broggini M, et al. 2009;106:16281–6. Inhibition of Sp1-dependent transcription and antitumor activity of the 38. To KK, Yu L, Liu S, Fu J, Cho CH. Constitutive AhR activation leads to new aureolic acid analogues mithramycin SDK and SK in human concomitant ABCG2-mediated multidrug resistance in cisplatin-resis- ovarian cancer xenografts. Gynecol Oncol 2010;118:182–8. tant esophageal carcinoma cells. Mol Carcinog 2012;51:449–64. 47. Venkov C, Plieth D, Ni T, Karmaker A, Bian A, George AL Jr, et al. 39. Li H, Gao Q, Guo L, Lu SH. The PTEN/PI3K/Akt pathway regulates Transcriptional networks in epithelial-mesenchymal transition. PLoS stem-like cells in primary esophageal carcinoma cells. Cancer Biol One 2011;6:e25354. Ther 2011;11:950–8. 48. Grohar PJ, Woldemichael GM, Griffin LB, Mendoza A, Chen QR, Yeung 40. Chen YJ, Huang WC, Wei YL, Hsu SC, Yuan P, Lin HY, et al. Elevated C, et al. Identification of an inhibitor of the EWS-FLI1 oncogenic BCRP/ABCG2 expression confers acquired resistance to gefitinib in transcription factor by high-throughput screening. J Natl Cancer Inst wild-type EGFR-expressing cells. PLoS One 2011;6:e21428. 2011;103:962–78. 41. Li F, Zeng H, Ying K. The combination of stem cell markers CD133 and 49. Jia Z, Gao Y, Wang L, Li Q, Zhang J, Le X, et al. Combined treatment of ABCG2 predicts relapse in stage I non-small cell lung carcinomas. Med pancreatic cancer with mithramycin A and tolfenamic acid promotes Oncol 2011;28:1458–62. Sp1 degradation and synergistic antitumor activity. Cancer Res 42. Tsunoda S, Okumura T, Ito T, Kondo K, Oritz C, Tanaka E, et al. ABCG2 2010;70:1111–9. expression is an independent unfavorable prognostic factor in esoph- 50. Kennedy BJ, Sandberg-Wollheim M, Loken M, Yarbro JW. Studies ageal squamous cell carcinoma. Oncology 2006;71:251–8. with tritiated mithramycin in C3H mice. Cancer Res 1967;27:1534–8.

4192 Cancer Res; 72(16) August 15, 2012 Cancer Research