breast cancer cells respond differentially to modulation of tgfb2 … · 2015. 10. 20. · tumor...

Tumor and Stem Cell Biology

Breast Cancer Cells Respond Differentially toModulation of TGFb2 Signaling after Exposure toChemotherapy or HypoxiaSiobhan K. O'Brien, Liang Chen,Wenyan Zhong, Douglas Armellino, Jiyang Yu,Christine Loreth, Maximillian Follettie, and Marc Damelin

Abstract

Intratumoral heterogeneity helps drive the selection fordiverse therapy-resistant cell populations. In this study, wedemonstrate the coexistence of two therapy-resistant popula-tions with distinct properties that are reproducibly enrichedunder conditions that characterize tumor pathophysiology.Breast cancer cells that survived chemotherapy or hypoxia wereenriched for cells expressing the major hyaluronic acid receptorCD44. However, only CD44hi cells that survived chemotherapyexhibited cancer stem cell (CSC) phenotypes based on growthpotential and gene expression signatures that represent onco-genic signaling and metastatic prowess. Strikingly, we identifiedTGFb2 as a key growth promoter of CD44hi cells that survived

chemotherapy but also as a growth inhibitor of cells thatsurvived hypoxia. Expression of the TGFb receptor TGFbR1and its effector molecule SMAD4 was required for enrichmentof CD44hi cells exposed to the chemotherapeutic drug epiru-bicin, which suggests a feed-forward loop to enrich for andenhance the function of surviving CSCs. Our results revealcontext-dependent effects of TGFb2 signaling in the sametumor at the same time. The emergence of distinct resistanttumor cell populations as a consequence of prior therapeuticintervention or microenvironmental cues has significant impli-cations for the responsiveness of recurring tumors to therapy.Cancer Res; 75(21); 1–12. �2015 AACR.

IntroductionTumor cells that are resistant to therapy can fuel disease relapse

and thus constitute a major barrier to achieving sustained clinicalresponses. Resistance is generally classified as intrinsic or devel-oped and can be mediated by several mechanisms. Moreover,intratumoral heterogeneity may generate a diversity of resistancemechanisms that coexist within the same tumor. Sources of thisheterogeneity include genetic and epigenetic changes, cellularplasticity, tumor microenvironment factors such as hypoxia,immune cell infiltration, exposure to therapeutic compounds,and stochastic changes in signal transduction and gene expression(1–4).

The cancer stem cell (CSC) model provides a mechanistic basisfor intratumoral heterogeneity based on measurable phenotypesand thus constitutes a useful framework for studying therapeuticresistance. CSCs comprise themalignant subpopulation of cells ina tumor that drive tumor growth, metastasis, and relapse (5, 6).CSCs can be resistant to cytotoxic chemotherapies (1, 7) and havealso been observed adjacent to hypoxic regions of tumors, for

example after antiangiogenic therapy (8, 9). Recent studies haveelucidated the coexistence of distinct CSC phenotypic states (10),though the factors that direct cells towards one state versus theother are not known.

Expression of the hyaluronic acid receptor CD44 is significantlyenriched on CSCs in many tumor types (5, 11, 12). The tumor-forming capacity of cells isolated from breast tumors is defined inpart by CD44 expression (11), and in breast cancer patients, theexpression of CD44 in the tumor correlates with poor prognosis(13, 14). In addition to its utility as a cell surfacemarker for CSCs,CD44 directly affects tumor progression, cell proliferation, andextracellular matrix organization (12, 15). CD44 knockdownwaspreviously shown to inhibit invasion, spheroid formation, andtumor growth of MDA-MB-231 cells (16) and promote differen-tiation of primary breast cancer cells (17).

TGFb has been implicated in CSC survival after chemotherapyand in the epithelial–mesenchymal transition (EMT), a processthat has been associated withmetastasis and CSCs (18, 19). Moregenerally, TGFb signaling exerts pleiotropic and context-depen-dent effects on cancer cell growth. For example, TGFb promotescytostasis in normal mammary epithelial cells (20), and con-versely the same signal promotes cell proliferation in late breastcancer (21). While the mechanisms that distinguish the pleiotro-pic effects of the signaling are not fully understood, the down-stream effects of TGFb can be modulated by many factors,including signals from the microenvironment, other oncogenicsignaling pathways, and the expression of SMAD-binding coacti-vators and corepressors.

Here we identify conditions that reproducibly enrich for CSC-like populations that exhibit opposite responses to TGFb signals.TGFb promotes the regrowth of breast cancer cells after cytotoxicchemotherapy yet inhibits their regrowth after hypoxic stress.

Oncology Research Unit, Pfizer Worldwide Research and Develop-ment, Pearl River, New York.

Note: Supplementary data for this article are available at Cancer ResearchOnline (http://cancerres.aacrjournals.org/).

Current address for S.K. O0Brien: Prolong Pharmaceuticals, South Plainfield, NJ.

Corresponding Author: Marc Damelin, Pfizer Worldwide Research and Devel-opment, 401 North Middletown Road, Building 200-4611, Pearl River, NY 10965.Phone: 845-602-7985; E-mail: [email protected]

doi: 10.1158/0008-5472.CAN-15-0650

�2015 American Association for Cancer Research.

CancerResearch

www.aacrjournals.org OF1

Research. on April 4, 2021. © 2015 American Association for Cancercancerres.aacrjournals.org Downloaded from

Published OnlineFirst September 4, 2015; DOI: 10.1158/0008-5472.CAN-15-0650

http://cancerres.aacrjournals.org/

Moreover, pharmacologic inhibition of TGFb signaling elicits theopposite pattern of responses in the two cell populations. Ashypoxic regions of the tumor may not be accessible to chemo-therapy, the distinct populations described here could coexist intumors, with implications for successful long-term treatment.

Materials and MethodsCell culture and treatments

SKBR3 and MCF7 cells were authenticated by short tandemrepeat analysis. SKBR3 cells were maintained in McCoy 5A Medi-um supplemented with 10% FBS and penicillin/streptomycin.MCF7 cells were maintained in Eagle Minimum Essential Medi-um supplemented with 10% FBS, 0.01 mg/mL insulin, 1% non-essential amino acids, and 1% sodium pyruvate. For FACS experi-ments, SKBR3 were treated with 60 to 90 nmol/L epirubicin andMCF7 cells were treated with 150 to 200 nmol/L, approximatelyIC30–IC50. For hypoxia experiments, cells were grown in 1% O2for 5 to 7 days. CD44 activity was stimulated with hyaluronic acid(HA) polymers of either 50 K or 1000 K size (Sigma) by adding200 mg/mL to three-dimensional (3D) cultures. CD44–HA inter-actionwas blockedusing the5F12 antibody (ThermoScientific) at10–25mg/mL. To examine TGFb activity, cellswere treatedwith 25ng/mLTGFb2or 1 to 2mmol/L SB431542 (R&DSystems). Cells in3D growth assays were treated by adding growth factor, HA, ordrug to both Matrigel and media.

RNAi-mediated knockdownCells were transduced with lentiviral particles containing spe-

cific shRNAs directed against CD44 or TGFbR1 (Sigma). Stablytransduced cells were selected and grown in 2 mg/mL puromycin.SMAD4 siRNAs were transfected with Opti-MEM and Lipofecta-mine RNAiMAX. Sequences are provided in Supplementary Data.

Flow cytometryCells were harvested using CellStripper nonenzymatic cell

dissociation buffer (Mediatech) and resuspended in PBS with3%BSA. Direct staining was performed for 30minutes on ice, andCD44 was detected with a FITC-labeled G44-26 antibody (BDBiosciences). ALDH activity was determined by incubating cellswith Aldefluor reagent (Stem Cell Technologies) according tomanufacturer's protocol for 1 hour at 37�C. Analysis was per-formedusing aBDFACSCalibur, and sortingwas performedusinga BD FACSAria 3. Live, CD44hi, and CD44�/lo or Aldefluorhi andAldefluor�/lo cells were isolated, spun at 1,000� g for 10minutesand resuspended inmedia for subsequent analysis and 3Dgrowthassays.

PCRRNA (triplicate samples per condition) was isolated using an

RNeasy Mini Kit (Qiagen) and cDNA was generated using Super-script VILO RT (Life Technologies). qPCR was performed usingTaqMan Fast Advanced Master Mix (Life Technologies) and spe-cific primer-probe sets for total CD44, CD44v8-10, VEGF,TGFbR1, and TGFb-2 (provided in Supplementary Information).Values reported are mean � SEM for at least three replicates.

Clonal growth assaySKBR3 cellswere embedded inMatrigel (Corning) at 1,000 cells

per well (unsorted) or 3,500 to 5,000 cells per well (sorted), whileMCF7 cells were embedded in reduced growth factor Matrigel

(Corning) at 1,000 cells per well and n � 3 for each condition.Cells in Matrigel were overlaid with Mammocult medium (StemCell Technologies). Colonies were counted at 7 days for untreatedand 7 to 14 days after epirubicin or hypoxia. Representativepictures of complete wells were taken using a GelCount instru-ment (Oxford Optronix) and individual colonies were taken at20� on an Olympus IX51 microscope. Values reported are mean� SEM for at least three replicates within each experiment. Eachresult was reproduced in multiple independent biologic experi-ments. Significance was tested using a two-tailed Student t testassuming unequal variance. Matrigel was used because withoutit, SKBR3mammospheres exhibited a high degree of aggregation,especially after epirubicin, and thus were difficult to quantify.

Western blot analysisCells were lysed with MPER (Pierce) supplemented with pro-

tease andphosphatase inhibitors for 30minutes on ice and lysateswere clarified by centrifugation for 5 minutes at 10,000 � g.Twenty-five to 50 mg of lysates were loaded onto 4%–12%SDS-PAGE gels and transferred onto PVDF membrane. Westernblots were performed using GAPDH (Sigma), CD44 (Millipore),and TGFbR1 (Life Technologies). Proteins were visualized usingDyLight-conjugated secondary antibodies (ThermoFisher) andthe LI-COR Odyssey scanner.

Senescence-associated b-galactosidase stainingSenescence was detected using the b-galactosidase staining kit

from Clontech. SKBR3 cells (250,000) were seeded in 6-wellplates, treated with epirubicin or placed in hypoxia, and stainedusing X-gal reagent overnight according to Clontech protocol.Representative pictures were taken at 20� on an Olympus IX51microscope, and results were quantitated by counting stainedand total cells in each picture. Values reported are mean � SEMfor at least three replicates within each experiment. Each resultwas reproduced in multiple independent biologic experiments.

Microarray analysisRNA from sorted cells was isolated using the RNEasy Kit

(Qiagen). cDNA synthesis was performed using the Ovation PicoWTA System (NuGEN) and Ribo-SPIA technique. Ovation PicoWTA products were fragmented and biotin labeled using theEncore Biotin Module (NuGEN). For each sample, 5 mg ofbiotin-labeled cDNA was hybridized to Human Genome U133þ 2.0 oligonucleotide arrays (Affymetrix) using buffers and con-ditions recommended by manufacturer. GeneChips were thenwashed and stained with Streptavidin R-Phycoerythrin (Molecu-lar Probes) using the GeneChip Fluidics Station 450, and scannedwith a Affymetrix GeneChip Scanner 3000. Gene expression datawere processed byMicro Array Suite 5.0 (MAS5) algorithm. Probesets were filtered to obtain robustly expressed qualifiers [averagesignal value of� 50, 100%present call (based onMAS5) in any ofthe sample group]. Microarray data were deposited in the GeneExpression Omnibus (GEO; accession #GSE72362). Gene SetEnrichment analysis was conducted using the Gene Set Enrich-ment Analysis (GSEA) software (Broad Institute; ref. 22).

Master regulator analysisWe interrogated context-specific regulatory or signaling net-

works and applied a network-based systems biology approach,the master regulator inference algorithm (23, 24) to identify keytranscriptional or signaling master regulators associated with

O'Brien et al.

Cancer Res; 75(21) November 1, 2015 Cancer ResearchOF2




CD44hi cell populations induced by chemotherapy or hypoxia.We used a data-driven approach, the ARACNe algorithm (25), toreconstruct breast cancer–specific interactomes from The CancerGenome Atlas (TCGA; ref. 26) gene expression profiles in bothmicroarray (N ¼ 359) and RNASeq (N¼ 950) platforms, against�800 transcription factors and �2,500 signaling proteins. Theparameters of ARACNe were configured as follows: P valuethreshold P ¼ 1E�7, DPI tolerance e ¼ 0, number of bootstrapsNB ¼ 100, and adaptive partitioning algorithm for mutual infor-mation estimation. For the GSEA method in the master regulatorinference algorithm, we applied "maxmean" statistic to score theenrichment of the gene set and used sample permutation to buildthe null distribution for statistical significance.We used the Fishermethod to integrate master regulators predicted frommicroarray-or RNASeq-based TCGA breast cancer networks.

ResultsChemotherapy and hypoxia enrich for a CD44hi population

Cultured SKBR3 breast cancer cells and tumor xenografts estab-lished from them exhibit low expression of CD44, but whentumor-bearing animals were treated with epirubicin, a cytotoxicchemotherapy, the relapsed tumors exhibited high CD44 expres-sion (27). To determine whether this observation could bereproduced in vitro, which would enable mechanistic studies,SKBR3 cells were treated with epirubicin for 3 days at the con-centration that inhibited growth by 50% (IC50). Both the mRNAand protein levels of CD44 increased after the treatment, asrevealed by immunoblot and qRT-PCR, respectively (Fig. 1A andB). The same effectwas observedwhen the cellswere cultured for 5to 9 days in hypoxia (1%O2), a stress that is common inuntreatedtumors and also can be induced by antiangiogenic therapy (Fig.1A and B). As drug perfusion through hypoxic regions can beminimal (28), it is conceivable that chemotherapy and hypoxicstress could simultaneously and differentially affect separateregions of the same tumor.

CD44mRNA was induced to a much greater extent in hypoxiacompared with chemotherapy (Fig. 1B). The induction of CD44in hypoxia required 3 days, compared with 1 day for induction ofVEGF, and therefore CD44 does not appear to be a direct targetof hypoxia inducible factor (HIF). As alternatively spliced vari-ants of CD44 yield protein isoforms with different bindingspecificities for extracellular matrix substrates, cell surface recep-tors, and other signalingmolecules (29), we analyzed the variantsby RT-PCR after each treatment with exon-specific primers (30).CD44S and CD44v8-10 were similarly induced in epirubicin andhypoxia comparedwith untreated cells (Supplementary Fig. S1A).

Flow cytometry was used to distinguish between two scenarios:induction of CD44 expression in all of the cells or enrichment ofthe CD44hi population. Treatment with epirubicin or incubationin hypoxia enriched for CD44hi cells as defined by CD44 cellsurface expression (Fig. 1C); moreover, treatment with paclitaxelhad a similar effect (data not shown). The enrichment persistedlong after each treatment: after one week in normal conditions,the fraction of CD44hi cells had increased further, and after 2weeks, the fraction remained elevated in the epirubicin-treatedculture and returned to its post-treatment level in the hypoxia-treated culture (Fig. 1D and E). Thus the observed enrichments ofCD44hi do not reflect transient or readily reversible processes.Despite these changes inCD44 expression,wedidnot observe anydecrease in the expression of CD24 (Supplementary Fig. S1B),

another marker used in analysis of breast CSC populations andcommonly used in combination with CD44. The expression ofCD44 alone has been sufficient for isolation of CSC populations(11).

Several experiments were performed to study the mechanismthat underlies the enrichment of CD44hi cells. Annexin V stainingrevealed that the apoptotic population in both treatments con-tainedbothCD44hi andCD44�/lo cells (Supplementary Fig. S1C),which suggested that the enrichment was not merely the result ofcell death that was restricted to CD44�/lo cells. Because CD44expression can be induced in senescent fibroblasts (31), weexamined whether chemotherapy and hypoxia induce senescencein this model. Senescence-associated b-galactosidase stainingrevealed an increased number of senescent cells after epirubicintreatment compared with hypoxia (Fig. 1F and G). This resultindicated that epirubicin could induce a quiescent state in cellsand also suggested unique mechanisms of survival that gave riseto the CD44hi population in each condition.

CD44 is required for clonal growth of SKBR3 breast cancer cellsThe above results suggested that CD44 could be an important

determinant of therapeutic resistance in SKBR3breast cancer cells.To determine whether CD44 was required for growth, expressionof CD44 was reduced by shRNA technology (Fig. 2A). The clonalgrowth of SKBR3 in Matrigel was dramatically inhibited in theabsence of CD44: colony number was reduced by at least 80%with each of two independent CD44 shRNA constructs relativeto nontargeting shRNA (Fig. 2B). On tissue culture plastic, CD44knockdown did not have an observable impact on cell growth atnormal passaging density and had a modest effect on clonogenicgrowth, which might reflect the interaction of CD44 with extra-cellular matrix (32).

To further explore the function of the extracellular matrix in theobserved phenotype, we investigated the effect of hyaluronic acid(HA), the primary ligandof CD44.HA increased the cell growth ofSKBR3 but did not restore colony formation of CD44 knockdowncells (Fig. 2C), confirming the CD44-based mechanism of theeffect. Conversely, the addition of an antibody (clone 5F12) thatblocks the interaction of CD44 with HA inhibited HA-stimulatedcell growth in a dose-dependent manner (Fig. 2D). As the knock-down of CD44 had a greater effect on cell growth than 5F12treatment, it is likely that functions of CD44 in addition to HA-based signaling are required for cell growth. Together, these datademonstrate key functions of CD44 in SKBR3 cells, a findingconsistent with results in other cell lines as well as the prognosticsignificance of CD44 in breast cancer (11, 13, 16).

The CSC phenotype of CD44hi cells depends on the treatmenthistory

CSC populations have been enriched following chemotherapytreatment (1, 7) as well as hypoxia that results from rapid tumorgrowthor antiangiogenic treatment (9, 33). Todeterminewhetherthe SKBR3 CD44hi cells that survived these treatments exhibitedCSC phenotypes, viable CD44hi and CD44�/lo populations wereisolated by FACS and then immediately counted and seeded ingrowth assays (Fig. 3A and Supplementary Fig. S2A). qRT-PCRconfirmed higher CD44 mRNA levels in the sorted CD44hi cells,consistent with the cell surface expression (Supplementary Fig.S2B and S2C).

Of the cells that survived epirubicin, CD44hi cells readilyformed colonies, while CD44�/lo cells rarely formed colonies

Resistant Cancer Cells with Opposite Responses to TGFb

www.aacrjournals.org Cancer Res; 75(21) November 1, 2015 OF3




(Fig. 3B and C); comparable results were obtained with twoconcentrations of epirubicin (Fig. 3C). In sharp contrast, of thecells that survived hypoxic stress, the CD44hi and CD44�/lo

populations exhibited comparable growth (Fig. 3B and D); thesame results were obtained at the extreme near-anoxic oxygenlevels of 0.1%O2 (Fig. 3D). Moreover, increasing the stringencyof growth conditions, for example with Cultrex or growthfactor–reduced Matrigel matrices, could not distinguish theclonogenic potentials of CD44hi and CD44�/lo (data notshown). Colonies from sorted cells from untreated, epirubi-

cin-treated and hypoxia-treated cultures exhibited similarcolony morphology (Supplementary Fig. S2D), which enableda direct comparison of colony number. Colonies that grewfrom cells sorted after chemotherapy and hypoxia were com-prised predominantly of CD44�/lo cells regardless of CD44expression at the time of sorting (Supplementary Fig. S2E).Overall, our results demonstrated a striking phenotypic differ-ence in CD44hi populations following two treatments andsuggested that epirubicin, but not hypoxia, was enriching forCD44hi-marked CSCs.

Figure 1.CD44 is expressed in response tocytotoxic chemotherapy or hypoxiain SKBR3 cells. A, CD44 totalprotein expression is increased inchemotherapy and hypoxia. Westernblot analysis was performed using totalprotein extracts of untreated,epirubicin-treated, or hypoxic cells forCD44 and GAPDH. B, CD44 mRNAexpression is induced in epirubicintreatment and hypoxia. cDNA sampleswere generated from SKBR3 cellstreated with 60 nmol/L epirubicin for3 days or incubated in 1% O2 for 1 to 8days. VEGF mRNA was includedas a control for HIF1-controlledtranscription. Results shown areaverage of three or more replicates� SEM for a representative experiment.� , P 0.05. C, flow cytometry for CD44expression. SKBR3 cells wereincubatedwith 90nmol/L epirubicin for3 days or placed in 1% O2 for 7 days.Cells were harvested and stained withCD44-FITC antibody for analysis. D,CD44 is expressed at the cell surfaceafter treatment withdrawal. CD44levels were measured by flowcytometry 0 to 14 days after epirubicinwashout. E, after incubation in hypoxia,cells were placed back in normoxia andCD44 expression was measured byflow cytometry. F, senescence-associated b-galactosidase staining inuntreated, epirubicin-treated, andhypoxic SKBR3. G, quantitationof b-galactosidase staining.�� , P 0.01. N/S, not significant.

O'Brien et al.





Wehypothesized that SKBR3 cells that survive epirubicin versushypoxia would exhibit differential responses to salinomycin, aWnt pathway modulator that selectively inhibited growth ofCD44hi breast CSCs (7). Indeed, salinomycin suppressed theenrichment of CD44hi in epirubicin but not in hypoxia (Fig.3E). Thus, three different phenotypes of CD44hi cells that surviveepirubicin versus hypoxia, mechanism of enrichment, growthpotential, and response to salinomycin, together indicated thatsurprisingly, CD44hi enriches for CSC-like cells in epirubicin butnot hypoxia.

Unique transcriptional profiles of epirubicin-surviving CD44hi

cellsTo identify mechanisms that could explain the treatment-

dependent phenotype of CD44hi cells, transcriptional profileswere generated for sorted CD44hi and CD44�/lo cells aftereach treatment. The comparison of two populations from thesame culture provides a uniquely controlled basis for analysis.Replicate samples from four independent experiments clus-tered together (Fig. 4A). On the basis of the CSC phenotypesspecifically observed for CD44hi cells after epirubicin, itwas initially surprising that hypoxia induced the transcrip-tion of a substantially larger number of genes in CD44hi

versus CD44�/lo compared with epirubicin (Fig. 4B). How-ever, function-based analyses of the transcriptional profilesrevealed that epirubicin CD44hi cells exhibited a responsemore reflective of tumor progression, consistent with theirCSC phenotypes.

GSEA revealed that more transcriptional signatures wereenriched in CD44hi versus CD44�/lo after epirubicin than inhypoxia (Fig. 4C). CD44hi cells that survived epirubicin exhibitedbroader enrichment for signatures related to stem cells, EMT,metastasis, Wnt signaling, cancer progression, and oncogenicsignaling (Supplementary Table S1), consistent with theCD44hiCSCphenotype after epirubicin but not hypoxia.Notably,multiple gene sets related to b-catenin (CTNNB1), a key trans-ducer of Wnt signaling, were observed exclusively after epirubicintreatment (Supplementary Table S1). There was also substantialoverlap in the gene sets enriched by both treatments, and theseshared sets are indicative of aggressive cancer growth and lessdifferentiated tumors, as well as genes associated with cell adhe-sion and motility, which might be connected to the function ofCD44.

CD44hi cells that survived both treatments were enriched forgenes in the TGFb pathway (Supplementary Table S1). The ligandTGFb2 was upregulated in CD44hi cells following both chemo-therapy and hypoxia, as indicated by several microarray probesand confirmed by RT-PCR; themRNA levels of TGFb2, like CD44,were induced to a greater extent by hypoxia than epirubicin,but the CD44hi/CD44�/lo ratios were comparable (Supplemen-tary Fig. S3A and S3B). Interestingly, endoglin, a coreceptor thatregulates TGFbR1/R2 activity, was downregulated after hypoxiabut maintained in epirubicin (Supplementary Fig. S3C).

As a complement to GSEA, we performed an analysis of generegulatory networks to identify master regulators that may haveimplemented the observed transcriptional changes in CD44hi

Figure 2.CD44 expression and interaction withhyaluronic acid is necessary for clonalgrowth of SKBR3 cells. A, flowcytometry for CD44 cell surfaceexpression in nontargeting and shCD44SKBR3. Cells were incubated in 1% O2for 7 days and then stained with anisotype control or CD44-APC. B,Matrigel colony formation assay withSKBR3 cells stably expressingnontargeting shRNA (shNT) or CD44-specific (shCD44) shRNA. Top,quantitation of colony formation;bottom, view of representative wells.Data represent the average of at leastthree samples � SEM. � , P 0.05. C,Matrigel colony formation assay withshNT SKBR3 or shCD44 SKBR3supplemented with hyaluronic acid(HA) of increasing size. D, Matrigelcolony formation of SKBR3 treatedwith a CD44-HA blocking antibody.5F12 antibody or IgG and 200 mg of1000 KHAwere added to bothMatrigeland media.






versus CD44�/lo. Master regulators include transcription factorsas well as upstream signaling factors and are defined by inter-actomes that were reconstructed from datasets in TCGA (25,26). There were more unique master regulators in CD44hi

versus CD44�/lo cells after epirubicin compared with hypoxia(Fig. 4D, Supplementary Table S2), similar to the finding inGSEA. Interestingly, two master regulators related to Wnt sig-

naling were only identified in epirubicin: Wnt5A (FDR ¼ 0.01)and CITED1 (FDR ¼ 0.009). CITED1 interacts with p300/CBPand SMAD4 to promote transcription in response to TGFb andalso plays a role in Wnt-mediated gene expression (34, 35).These epirubicin-specific master regulators are noteworthy inlight of the salinomycin effect and b-catenin signatures thatwere also specific to epirubicin.

Figure 3.CD44hi cells isolated afterchemotherapy and hypoxia arephenotypically distinct. A, schematicfor isolation of CD44hi and CD44�/lo

cells by FACS after epirubicintreatment or hypoxia. Sorted cells wereimplanted in Matrigel and colonieswere counted. B, colony formation ofSKBR3 CD44hi and CD44�/lo cells afterepirubicin or hypoxia. C, CD44hi cellsfrom epirubicin treatment form morecolonies than CD44�/lo. Data representthe average of at least threesamples � SEM. � , P 0.05. D,CD44hi and CD44�/lo cells fromhypoxia form an equal number ofcolonies. E, chemotherapy-treated butnot hypoxicCD44hi cells are sensitive tosalinomycin. SKBR3 cells werecotreated with 3 mmol/L salinomycinand either 90 nmol/L epirubicinor 1% O2.

O'Brien et al.





TGFb exerts opposing effects on cells that survive epirubicinor hypoxia

The enrichment of TGFb signaling factors in CD44hi cells wasprovocative in light of the documented interaction betweenCD44and TGFbR1 to promote cell growth (36) and role of TGFbsignaling inCD44hi breast CSCs (37).Wehypothesized that TGFbsignaling modulated the growth of the cells that survived epir-ubicin and hypoxia. To evaluate the effect of TGFb on the growthof cells that survived the two treatments, sorted cells were sup-plemented with recombinant TGFb2 ligand. Strikingly, TGFb2promoted the growth of CD44hi cells that survived epirubicin, yetinhibited the growth of CD44hi (andCD44�/lo) cells that survivedhypoxia (Fig. 5A and B and Supplementary Fig. S4A and S4B).TGFb2 did not rescue the growth defects of CD44�/lo cells thatsurvived epirubicin (Fig. 5A, Supplementary Fig. S4A). Similareffects were obtained with TGFb1 ligand (data not shown). Thus,despite the upregulation of TGFb2 in both conditions, the ligandexerted opposite effects on the cells based on their treatmenthistory.

In a complementary approach, TGFb signaling was modu-lated pharmacologically with the TGFbR inhibitor SB-431542and the effect on cell growth was determined. SB-431542 is ahighly specific kinase inhibitor of type 1 TGFb receptors (38).Consistent with the above results, SB-431542 inhibited thegrowth of CD44hi cells that survived epirubicin and promoted

the growth of CD44hi and CD44�/lo cells that survived hypoxia(Fig. 5C and D).

Because CD44 did not mark a CSC-specific population afterhypoxia, we examined whether another CSC marker defined apopulation that could regrow after exposure to low oxygen.Alehyde dehydrogenase (ALDH) activity was previously identi-fied as amarker of breast CSCsby staining cellswithAldefluor (39,40), and cells that exhibit this activity have also been identified inhypoxic regions of breast tumors (33). Unlike CD44, there was noenrichment in Aldefluorhi cells in response to epirubicin orhypoxia; moreover, the CD44hi and Aldefluorhi populations hadminimal overlap (Supplementary Fig. S4C). However, of the cellsthat survived hypoxia, Aldefluorhi signal highly enriched forclonal growth (Fig. 5E), which indicated that this activity enrichedfor a CSC-like population in this context. After hypoxia,Aldefluorhi cells responded to TGFb2 treatment in the same wayas CD44hi and CD44�/lo cells: TGFb2 inhibited colony formation(Fig. 5E). The effect of TGFb on cells that survive hypoxia was thusindependent of CD44. Conversely, the effect of TGFb on cells thatsurvive epirubicin was dependent on CD44: it was observed inCD44hi but not CD44lo or Aldefluor-sorted cells (Fig. 5A andSupplementary Fig. S4D).

To determine whether the treatment-dependent effects ofTGFb in SKBR3 were also observed in other breast cancer celllines, we performed similar experiments with MCF7 cells,

Figure 4.Gene expression patterns in SKBR3CD44hi and CD44�/lo cells aftertreatment. A, heat map detailingexpression patterns of genes from fourindependent experiments significantlyupregulated (P 0.05) in CD44hicompared with CD44�/lo fromepirubicin treatment or hypoxia.B, Venn diagram comparing genessignificantly upregulated in CD44hi

versus CD44�/lo in either epirubicinor hypoxia treatment and genesexpressed in both conditions. C, Venndiagram summarizing number of genesets enriched (FDR 0.05) in CD44hiversus CD44�/lo samples after eitherepirubicin or hypoxia treatment andgene sets enriched in both conditions.D, Venn diagram summarizing masterregulators of gene expressionidentified (z score � 2; FDR 0.05) inCD44hi versus CD44�/lo samples afterepirubicin or hypoxia treatment.






Figure 5.TGFb signaling elicits different effects onchemotherapy-treated and hypoxic cells. A,TGFb2 stimulates colony formation of CD44hi

cells from epirubicin treatment. CD44hi andCD44�/lo cells were sorted and embeddedin Matrigel with 25 ng/mL TGFb2. Datarepresent the average of at least threesamples� SEM. �, P0.05. B, TGFb2 inhibitsMatrigel colony formation of hypoxic SKBR3cells. C, inhibition of TGFbR with 1–2 mmol/LSB431542 decreases colony formation inCD44hi cells after epirubicin treatment.SB431542 was added to sorted cells in aMatrigel colony formation assay.D, SB431542treatment enhances colony formation ofCD44hi and CD44�/lo SKBR3 cells fromhypoxia. E, colony formation of Aldefluorhi

and Aldefluor�/lo populations sorted fromSKBR3 incubated in hypoxia. Sorted cellswere embedded in Matrigel with 0 or25 ng/mL TGFb2. CD44hi and CD44�/lo

colony formation is shown for comparison.F, TGFb2 stimulates colony formation ofchemotherapy-treated CD44hi MCF7 cells.Results represent the average of threereplicates � SEM. � , P 0.05. Cells weresorted after treatment with 200 nmol/Lepirubicin and then embedded in Matrigel. G,TGFb2 inhibits colony Matrigel formation ofhypoxic MCF7 cells. Cells were sortedafter exposure to hypoxia and thenembedded in Matrigel.

O'Brien et al.





which are ER-positive, HER2-negative, and represent a differentsubtype of breast cancer than HER2-positive SKBR3. Consistentwith the results in SKBR3, TGFb2 promoted the growth ofMCF7 CD44hi cells that survived epirubicin and inhibited thegrowth of CD44hi cells that survived hypoxia (Fig. 5F and G andSupplementary Fig. S5). These data extended a key finding ofthis study to another breast cancer cell line and suggested thatour conclusions may apply broadly across breast cancer. Inaddition, although the HER2 oncogene is amplified in SKBR3and has been shown to cooperate with TGFb (41, 42), theMCF7 results suggest that the HER2–TGFb interaction is unlike-ly to be relevant in this context.

CSC enrichment in epirubicin is dependent on TGFb signalingOn the basis of the above results and cited literature (36, 37),

we hypothesized a direct mechanistic link between TGFb signal-ing and the CD44hi CSC phenotype in epirubicin. To determinewhether TGFb signaling was required for the enrichment, wereduced TGFbR1 expression in SKBR3 cells by shRNA and then

exposed them to epirubicin or hypoxia. Strikingly, the enrichmentof CD44hi cells in epirubicin was prevented by TGFbR1 knock-down, while the enrichment in hypoxia was not (Fig. 6A and B).

To establishwhether the TGFb-mediated enrichment ofCD44hi

cells in epirubicin occurs at the transcriptional level, we knockeddown the expression of SMAD4, a key transcriptional effectorof TGFb signaling, and exposed the cells to epirubicin. The re-duction of SMAD4 levels prevented the enrichment of CD44hi

CSCs (Fig. 6C). In addition, we measured CD44 mRNA levelsin the TGFbR1 knockdown cells and observed that CD44 levelswere reduced (Fig. 6D). These results demonstrate SMAD-depen-dent transcriptional activation downstream of TGFb engage-ment to enrich the CSC phenotype in epirubicin.

On the basis of these results we hypothesized that inhibition ofCD44 or TGFb signaling would sensitize SKBR3 cells to chemo-therapy. To test this, cells with stable shRNA knockdown of CD44or TGFbR1 were exposed to various concentrations of epirubicinfor 24 hours and then allowed to form colonies. Indeed, SKBR3cells with stable knockdown of CD44 or TGFbR1 exhibited

Figure 6.TGFb dependence of CSC enrichmentin epirubicin. A, Western blot analysisfor TGFbR1 in SKBR3 cells stablyexpressing nontargeting (NT) orTGFbR1-specific shRNAs. B, CD44 cellsurface expression in shNT orshTGFbR1-SKBR3 cells was determinedby FACS after epirubicin treatment orhypoxia. % CD44hi is plotted for arepresentative experiment. C, CD44cell surface expression in SKBR3 cellswith nontargeting (NT) or SMAD4siRNA (SMAD4-1, -2, and -3) wasdetermined by FACS after epirubicintreatment. Cells were transfected withsiRNA 3 days prior to 60 nmol/Lepirubicin treatment for an additional3 days. D, mRNA was isolated fromshNT or shTGFbR1-SKBR3, and qPCRwas then performed using primer-probe sets for TGFbR1 and CD44.Data are the average of three replicates� SEM. � , P 0.05; �� , P 0.01. E,clonogenic growth after epirubicintreatment of parental SKBR3 cells orcells stably transduced withnontargeting (NT), TGFbR1, or CD44shRNA. Cells were treated withepirubicin for 24 hours starting theday after seeding, and colonies werecounted 17 days after seeding.Histogram shows the mean� SD of triplicate samples, normalizedto the untreated sample withappropriate error propagation.






reduced growth after epirubicin treatment relative to the control(Fig. 6E). Similarly, TGFb inhibition increased the radiosensitivityof MCF7 and other breast cancer cells (43), which is likely due tothe same mechanism as both epirubicin and radiation generateDNA damage. Notably, CD44 knockdown in SKBR3 did notmodulate sensitivity to epirubicin in a cell proliferation assay(data not shown), which suggests that the clonogenic effect isdriven by cell survival. The experiments couldnot be performed inMatrigel as CD44 knockdown inhibits growth in this condition(Fig. 2B). Together, these data demonstrate that TGFb signalingis involved in the earliest response to chemotherapy and suggest afeedback loop in which, under certain conditions, TGFb signalinginduces CD44 expression and enriches the CSC phenotype, andthen CD44 modulates the effect of TGFb signals.

DiscussionResistance to anticancer therapies is driven by a vast array of

mechanisms and can vary across tumor types and patientpopulations. Our study has demonstrated the coexistence inthe same tumor of resistant cells with opposite responses toTGFb signals, and thus has revealed a novel layer of complexityof the TGFb pathway in breast cancer. One interpretation of thedata is that different cell populations in a tumor can exhibitearly- or late-stage characteristics based on microenvironment,cellular history, or other factors. For example, cells exposed tochemotherapy may interpret TGFb signals as progrowth, con-sistent with late-stage disease, while cells under hypoxic stressmay interpret TGFb signals as growth-suppressive until theregion becomes vascularized.

This study has identified an unexpected duality of TGFb path-way modulation that builds upon and refines the previouslyidentified links between TGFb and breast CSCs (18, 37). Wedocument opposite effects of TGFb activity that are dependentona cell's previous treatment yet independent ofCD44expression(Fig. 7); the data challenge the existing model of a general CD44–TGFb functional interaction in breast CSCs as well as the assump-tion that all CD44hi breast cancer cells exhibit the same pheno-types. Importantly, the data provide the first evidence that cells inthe same tumor can have opposite signaling activities in responseto TGFb modulation; this dichotomous response is mechanisti-

cally distinct from heterogeneous signaling that reflects variedexpression of a receptor. In addition, we document a role of TGFbsignaling, including SMAD4 dependence, in enriching for theCD44hi CSC phenotype in chemotherapy. TGFb signaling mayenrich for the CSC phenotype by inducing senescence transientlyin epirubicin treatment (Fig. 1F andG), consistentwith previouslycharacterized functions of TGFb (44). Thus, our study providesone mechanistic basis for heterogeneity of CSC phenotypes andmarkers and suggests a feed-forward loop in which, in certaincellular contexts, TGFb promotes the enrichment of CD44hi CSCphenotype and then CD44 functionally interacts with the TGFbpathway to promote cell growth in recovery.

The TGFb pathway interacts with many pathways, and uniquedrivers may influence the proliferative versus antiproliferativeeffects of TGFb. Our results implicate a context-dependent inter-action between CD44 and TGFb receptors, which builds onprevious observations (36). Phosphorylated SMADs could inter-act with distinct transcription factors after chemotherapy orhypoxia; for example, we identified the transcription factor CIT-ED1 as amaster regulator in CD44hi cells that survived epirubicinbut not hypoxia. Signatures of b-catenin, which can interact withSMAD proteins, were enriched in epirubicin but not hypoxia.Together, TGFb and b-catenin could promote an EMT phenotypeand CSC-like growth. Both CITED1 and b-catenin illustrate thepossible interaction between TGFb andWnt pathways in responseto chemotherapy. Signal modulation also could occur at thereceptor complex; for example, the reduced expression of TGFbRcoreceptor endoglin in cells that survived hypoxia but not epir-ubicin (Supplementary Fig. S3C)may favor the growth-inhibitoryeffects of TGFb in hypoxia (45, 46). Interestingly, the cytokine IL8was strongly induced in hypoxia but not epirubicin (Supplemen-tary Fig. S3D), which could be explained by the potent angiogenicproperties of IL8, although IL8 was also implicated in paclitaxel-induced CSCs (18).

Our study builds on previous observations of multiple CSCstates in breast cancer (10) by revealing specific conditions thatenrich for CSC-like populations with distinct phenotypes. CD44expression enriched these cells only after epirubicin, while Alde-fluor activity enriched them after hypoxia; in addition, the CSCsexhibited opposite results to TGFbmodulation. CD44hi cells thatsurvived epirubicin exhibited a gene expression profile that

ChemotherapyTreatment

GrowthGrowth

Recovery

CD44 CD44

IL8;CD44;TGFβ-2

CD44;TGFβ-2;β-Catenin/WntMetastasis

TGFβ-2TGFβ-2

TGFβ-2

TGFβR

SMAD

TGFβR TGFβR

TGFβR

Treatment Recovery

Hypoxia Figure 7.Dichotomy between TGFb signaling inbreast cancer cells that survivechemotherapy versus hypoxia.Among cells that survivechemotherapy, CD44 expressionmarks the population with growthpotential and various characteristicsof CSCs, and TGFb promotes theenrichment of CD44hi cells in a SMAD-dependent manner and promotestheir growth during recovery. Incontrast, among cells that survivehypoxia, TGFb elicits the oppositeresponse during recovery byinhibiting the growth in a CD44-independent manner; CD44 does notenrich for CSC-like cells and is notrequired for growth during recovery.

O'Brien et al.





reflects EMT, consistentwithCD44þbreast cancer patient samples(10). Conversely, ALDH1 expression in patient samples correlat-ed with a mesenchymal-to-epithelial transition and increasedproliferation (10). Our results suggest that CSC states and phe-notypes may continually adapt to changes in the tumor micro-environment and therapeutic insults as well as cumulative geneticchanges.

This study, along with several others (40, 47), demonstratesthat in vitromechanistic studies are a necessary complement to invivo tumor initiation studies in the characterization of intratu-moral heterogeneity and CSC growth mechanisms. CD44hi cellsisolated from dissociated tissues cannot be distinguished by theirunique histories and phenotypes, and the characterization ofthose cells as one population compromises efforts to understandmechanisms of therapeutic resistance. In both breast cancer celllines that we analyzed, our data suggest that CD44hi cells isolatedfrom relapsed tumors would include subpopulations that exhibitopposite responses to TGFb modulation, depending on theirexposure to chemotherapy or hypoxic microenvironment. In thefuture, tumor cells that are genetically engineered to display theirindividual history of DNA damage and exposure to hypoxia, forexample, withfluorescent proteins, could be used to bridge in vitroand in vivo analyses. Other parameters that could also be mon-itored include intracellular levels ofmetabolites and administeredcompounds; the latter is relevant due to challenges of drugdelivery in solid tumors that can result in different exposures indifferent regions of the tumor.

The characterization of the diversity of resistance mechanismsthat may coexist in a tumor will ultimately inform therapeuticstrategies to improve clinical outcome. Novel therapies or mul-tidrug regimens that target resistant tumor cell populations maybe enabled by recent technological advances, such as nanoparti-

cles, that can enable the timed delivery of two different com-pounds (48). Our data have specific implications for the devel-opment of TGFb pathway modulators such as type 1 TGFbreceptor inhibitors (49, 50). In addition, our results suggest thatmicroenvironment-based biomarkers might be valuable comple-ments to tumor cell–basedbiomarkers in optimizing thedesignoftherapeutic regimens.

Disclosure of Potential Conflicts of InterestNo potential conflicts of interest were disclosed.

Authors' ContributionsConception and design: S.K. O'Brien, M. DamelinDevelopment of methodology: S.K. O'Brien, L. Chen, D. ArmellinoAcquisition of data (provided animals, acquired and managed patients,provided facilities, etc.): S.K. O'Brien, L. Chen, D. Armellino, C. LorethAnalysis and interpretation of data (e.g., statistical analysis, biostatistics,computational analysis): S.K. O'Brien, L. Chen, W. Zhong, J. Yu, C. Loreth, M.Follettie, M. DamelinWriting, review, and/or revision of the manuscript: S.K. O'Brien, J. Yu, C.Loreth, M. Follettie, M. DamelinStudy supervision: M. Damelin

AcknowledgmentsThe authors thank Kenneth Geles, Matthew Sung, Kevin Bray, and Shirley

Markant for critical reading of the manuscript and/or helpful discussion, andAndreas Giannakou for technical assistance. S.K. O'Brien acknowledges supportfrom the Pfizer Postdoctoral Training Program.

The costs of publication of this article were defrayed in part by thepayment of page charges. This article must therefore be hereby markedadvertisement in accordance with 18 U.S.C. Section 1734 solely to indicatethis fact.

Received March 6, 2015; revised July 13, 2015; accepted August 4, 2015;published OnlineFirst September 4, 2015.

References1. Bae YH. Drug targeting and tumor heterogeneity. J Control Release 2009;

133:2–3.2. Galon J, Costes A, Sanchez-Cabo F, Kirilovsky A, Mlecnik B, Lagorce-Pages

C, et al. Type, density, and location of immune cells within humancolorectal tumors predict clinical outcome. Science 2006;313:1960–4.

3. Marusyk A, Polyak K. Tumor heterogeneity: causes and consequences.Biochim Biophys Acta 2010;1805:105–17.

4. Vaupel P, Kallinowski F, Okunieff P. Blood flow, oxygen and nutrientsupply, and metabolic microenvironment of human tumors: a review.Cancer Res 1989;49:6449–65.

5. Visvader JE, Lindeman GJ. Cancer stem cells in solid tumours: accumu-lating evidence and unresolved questions. Nat Rev Cancer 2008;8:755–68.

6. Zhou BB, Zhang H, Damelin M, Geles KG, Grindley JC, Dirks PB. Tumour-initiating cells: challenges and opportunities for anticancer drug discovery.Nat Rev Drug Discov 2009;8:806–23.

7. Gupta PB, Onder TT, Jiang G, Tao K, Kuperwasser C, Weinberg RA, et al.Identification of selective inhibitors of cancer stem cells by high-through-put screening. Cell 2009;138:645–59.

8. Ebos JM, Lee CR, Cruz-MunozW, BjarnasonGA, Christensen JG, Kerbel RS.Acceleratedmetastasis after short-term treatment with a potent inhibitor oftumor angiogenesis. Cancer Cell 2009;15:232–9.

9. Paez-Ribes M, Allen E, Hudock J, Takeda T, Okuyama H, Vinals F, et al.Antiangiogenic therapy elicits malignant progression of tumors toincreased local invasion and distant metastasis. Cancer Cell 2009;15:220–31.

10. Liu S, Cong Y, Wang D, Sun Y, Deng L, Liu Y, et al. Breast cancer stem cellstransition between epithelial and mesenchymal states reflective of theirnormal counterparts. Stem Cell Rep 2014;2:78–91.

11. Al-Hajj M, Wicha MS, Benito-Hernandez A, Morrison SJ, Clarke MF.Prospective identification of tumorigenic breast cancer cells. Proc NatlAcad Sci U S A 2003;100:3983–8.

12. Zoller M. CD44: can a cancer-initiating cell profit from an abundantlyexpressed molecule? Nat Rev Cancer 2011;11:254–67.

13. Hurt EM, Kawasaki BT, KlarmannGJ, Thomas SB, FarrarWL. CD44þCD24(�) prostate cells are early cancer progenitor/stem cells that provide amodel for patients with poor prognosis. Br J Cancer 2008;98:756–65.

14. Neumeister V, Agarwal S, Bordeaux J, Camp RL, Rimm DL. In situ iden-tification of putative cancer stem cells by multiplexing ALDH1, CD44, andcytokeratin identifies breast cancer patients with poor prognosis. Am JPathol 2010;176:2131–8.

15. MarhabaR, ZollerM.CD44 in cancer progression: adhesion,migration andgrowth regulation. J Mol Histol 2004;35:211–31.

16. Hiraga T, Ito S, Nakamura H. Cancer stem-like cell marker CD44 promotesbone metastases by enhancing tumorigenicity, cell motility, and hyalur-onan production. Cancer Res 2013;73:4112–22.

17. Pham PV, Phan NL, Nguyen NT, Truong NH, Duong TT, Le DV, et al.Differentiation of breast cancer stem cells by knockdown of CD44: prom-ising differentiation therapy. J Transl Med 2011;9:209.

18. Bhola NE, Balko JM, Dugger TC, Kuba MG, Sanchez V, Sanders M, et al.TGF-beta inhibition enhances chemotherapy action against triple-negativebreast cancer. J Clin Invest 2013;123:1348–58.

19. Singh A, Settleman J. EMT, cancer stem cells and drug resistance: anemerging axis of evil in the war on cancer. Oncogene 2010;29:4741–51.

20. Miettinen PJ, Ebner R, Lopez AR, Derynck R. TGF-beta induced transdiffer-entiation of mammary epithelial cells to mesenchymal cells: involvementof type I receptors. J Cell Biol 1994;127(6 Pt 2):2021–36.






21. Tang B, Vu M, Booker T, Santner SJ, Miller FR, Anver MR, et al. TGF-betaswitches from tumor suppressor to prometastatic factor in a model ofbreast cancer progression. J Clin Invest 2003;112:1116–24.

22. SubramanianA, TamayoP,Mootha VK,Mukherjee S, Ebert BL,GilletteMA,et al. Gene set enrichment analysis: a knowledge-based approach forinterpreting genome-wide expression profiles. Proc Natl Acad Sci U S A2005;102:15545–50.

23. Carro MS, Lim WK, Alvarez MJ, Bollo RJ, Zhao X, Snyder EY, et al. Thetranscriptional network for mesenchymal transformation of braintumours. Nature 2010;463:318–25.

24. Piovan E, Yu J, Tosello V, Herranz D, Ambesi-Impiombato A, Da Silva AC,et al. Direct reversal of glucocorticoid resistance by AKT inhibition in acutelymphoblastic leukemia. Cancer Cell 2013;24:766–76.

25. Margolin AA,Nemenman I, BassoK,Wiggins C, StolovitzkyG,Dalla FaveraR, et al. ARACNE: an algorithm for the reconstruction of gene regulatorynetworks in a mammalian cellular context. BMC Bioinformatics 2006;7Suppl 1:S7.

26. Cancer Genome Atlas N. Comprehensive molecular portraits of humanbreast tumours. Nature 2012;490:61–70.

27. Yu F, Yao H, Zhu P, Zhang X, Pan Q, Gong C, et al. let-7 regulates selfrenewal and tumorigenicity of breast cancer cells. Cell 2007;131:1109–23.

28. Jain RK. Normalization of tumor vasculature: an emerging concept inantiangiogenic therapy. Science 2005;307:58–62.

29. Ponta H, Sherman L, Herrlich PA. CD44: from adhesion molecules tosignalling regulators. Nat Rev Mol Cell Biol 2003;4:33–45.

30. Rajarajan A, Stokes A, Bloor BK, Ceder R, Desai H, Grafstrom RC, et al.CD44 expression in oro-pharyngeal carcinoma tissues and cell lines. PLoSONE 2012;7:e28776.

31. Linskens MH, Feng J, Andrews WH, Enlow BE, Saati SM, Tonkin LA, et al.Cataloging altered gene expression in young and senescent cells usingenhanced differential display. Nucleic Acids Res 1995;23:3244–51.

32. Bourguignon LY. CD44-mediated oncogenic signaling and cytoskeletonactivation during mammary tumor progression. J Mammary Gland BiolNeoplasia 2001;6:287–97.

33. Conley SJ, Gheordunescu E, Kakarala P, Newman B, Korkaya H, HeathAN, et al. Antiangiogenic agents increase breast cancer stem cells viathe generation of tumor hypoxia. Proc Natl Acad Sci U S A 2012;109:2784–9.

34. Shioda T, Lechleider RJ, Dunwoodie SL, Li H, Yahata T, de Caestecker MP,et al. Transcriptional activating activity of Smad4: roles of SMAD hetero-oligomerization and enhancement by an associating transactivator. ProcNatl Acad Sci U S A 1998;95:9785–90.

35. Yahata T, de Caestecker MP, Lechleider RJ, Andriole S, Roberts AB, Issel-bacher KJ, et al. The MSG1 non-DNA-binding transactivator binds to thep300/CBP coactivators, enhancing their functional link to the Smadtranscription factors. J Biol Chem 2000;275:8825–34.

36. Bourguignon LY, Singleton PA, Zhu H, Zhou B. Hyaluronan promotessignaling interaction between CD44 and the transforming growth factorbeta receptor I in metastatic breast tumor cells. J Biol Chem 2002;277:39703–12.

37. ShipitsinM,Campbell LL, Argani P,Weremowicz S, Bloushtain-QimronN,Yao J, et al. Molecular definition of breast tumor heterogeneity. Cancer Cell2007;11:259–73.

38. Inman GJ, Nicolas FJ, Callahan JF, Harling JD, Gaster LM, Reith AD, et al.SB-431542 is a potent and specific inhibitor of transforming growth factor-beta superfamily type I activin receptor-like kinase (ALK) receptors ALK4,ALK5, and ALK7. Mol Pharmacol 2002;62:65–74.

39. Charafe-Jauffret E, Ginestier C, Iovino F, TarpinC,DiebelM, Esterni B, et al.Aldehyde dehydrogenase 1-positive cancer stem cells mediate metastasisand poor clinical outcome in inflammatory breast cancer. Clin Cancer Res2010;16:45–55.

40. Ginestier C, Hur MH, Charafe-Jauffret E, Monville F, Dutcher J, BrownM, et al. ALDH1 is a marker of normal and malignant human mammarystem cells and a predictor of poor clinical outcome. Cell Stem Cell2007;1:555–67.

41. Muraoka RS, Koh Y, Roebuck LR, Sanders ME, Brantley-Sieders D, GorskaAE, et al. Increased malignancy of Neu-induced mammary tumors over-expressing active transforming growth factor beta1. Mol Cell Biol 2003;23:8691–703.

42. Seton-Rogers SE, LuY,Hines LM,KoundinyaM, LaBaer J,Muthuswamy SK,et al. Cooperation of the ErbB2 receptor and transforming growth factorbeta in induction of migration and invasion in mammary epithelial cells.Proc Natl Acad Sci U S A 2004;101:1257–62.

43. Bouquet F, Pal A, Pilones KA, Demaria S, Hann B, Akhurst RJ, et al.TGFbeta1 inhibition increases the radiosensitivity of breast cancer cellsin vitro and promotes tumor control by radiation in vivo. Clin Cancer Res2011;17:6754–65.

44. Katakura Y, Nakata E, Miura T, Shirahata S. Transforming growth factorbeta triggers two independent-senescence programs in cancer cells. Bio-chem Biophys Res Commun 1999;255:110–5.

45. Li C, Hampson IN, Hampson L, Kumar P, Bernabeu C, Kumar S. CD105antagonizes the inhibitory signaling of transforming growth factor beta1on human vascular endothelial cells. FASEB J 2000;14:55–64.

46. OxmannD, Held-Feindt J, Stark AM,Hattermann K, Yoneda T,Mentlein R.Endoglin expression in metastatic breast cancer cells enhances their inva-sive phenotype. Oncogene 2008;27:3567–75.

47. Kondo T, Setoguchi T, Taga T. Persistence of a small subpopulation ofcancer stem-like cells in the C6 glioma cell line. Proc Natl Acad Sci U S A2004;101:781–6.

48. Morton SW, LeeMJ,Deng ZJ, Dreaden EC, Siouve E, Shopsowitz KE, et al. Ananoparticle-based combination chemotherapy delivery system forenhanced tumor killing by dynamic rewiring of signaling pathways. SciSignal 2014;7:ra44.

49. Maier A, Peille AL, Vuaroqueaux V, LahnM. Anti-tumor activity of the TGF-beta receptor kinase inhibitor galunisertib (LY2157299 monohydrate) inpatient-derived tumor xenografts. Cell Oncol 2015;38:131–44.

50. Rodon J, Carducci MA, Sepulveda-Sanchez JM, Azaro A, Calvo E, Seoane J,et al. First-in-human dose study of the novel transforming growth factor-beta receptor I kinase inhibitor LY2157299 monohydrate in patients withadvanced cancer and glioma. Clin Cancer Res 2015;21:553–60.


O'Brien et al.




Published OnlineFirst September 4, 2015.Cancer Res Siobhan K. O'Brien, Liang Chen, Wenyan Zhong, et al. Hypoxia

2 Signaling after Exposure to Chemotherapy orβTGFBreast Cancer Cells Respond Differentially to Modulation of

Updated version

10.1158/0008-5472.CAN-15-0650doi:

Access the most recent version of this article at:

Material

Supplementary

http://cancerres.aacrjournals.org/content/suppl/2015/09/04/0008-5472.CAN-15-0650.DC1

Access the most recent supplemental material at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected] at

To order reprints of this article or to subscribe to the journal, contact the AACR Publications

Permissions

Rightslink site. (CCC)Click on "Request Permissions" which will take you to the Copyright Clearance Center's

.http://cancerres.aacrjournals.org/content/early/2015/10/20/0008-5472.CAN-15-0650To request permission to re-use all or part of this article, use this link



http://cancerres.aacrjournals.org/lookup/doi/10.1158/0008-5472.CAN-15-0650http://cancerres.aacrjournals.org/content/suppl/2015/09/04/0008-5472.CAN-15-0650.DC1http://cancerres.aacrjournals.org/cgi/alertsmailto:[email protected]://cancerres.aacrjournals.org/content/early/2015/10/20/0008-5472.CAN-15-0650http://cancerres.aacrjournals.org/

/ColorImageDict > /JPEG2000ColorACSImageDict > /JPEG2000ColorImageDict > /AntiAliasGrayImages false /CropGrayImages false /GrayImageMinResolution 200 /GrayImageMinResolutionPolicy /Warning /DownsampleGrayImages true /GrayImageDownsampleType /Bicubic /GrayImageResolution 300 /GrayImageDepth -1 /GrayImageMinDownsampleDepth 2 /GrayImageDownsampleThreshold 1.50000 /EncodeGrayImages true /GrayImageFilter /DCTEncode /AutoFilterGrayImages true /GrayImageAutoFilterStrategy /JPEG /GrayACSImageDict > /GrayImageDict > /JPEG2000GrayACSImageDict > /JPEG2000GrayImageDict > /AntiAliasMonoImages false /CropMonoImages false /MonoImageMinResolution 600 /MonoImageMinResolutionPolicy /Warning /DownsampleMonoImages true /MonoImageDownsampleType /Bicubic /MonoImageResolution 900 /MonoImageDepth -1 /MonoImageDownsampleThreshold 1.50000 /EncodeMonoImages true /MonoImageFilter /CCITTFaxEncode /MonoImageDict > /AllowPSXObjects false /CheckCompliance [ /None ] /PDFX1aCheck false /PDFX3Check false /PDFXCompliantPDFOnly false /PDFXNoTrimBoxError true /PDFXTrimBoxToMediaBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXSetBleedBoxToMediaBox true /PDFXBleedBoxToTrimBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXOutputIntentProfile (None) /PDFXOutputConditionIdentifier () /PDFXOutputCondition () /PDFXRegistryName () /PDFXTrapped /False

/CreateJDFFile false /Description > /Namespace [ (Adobe) (Common) (1.0) ] /OtherNamespaces [ > /FormElements false /GenerateStructure false /IncludeBookmarks false /IncludeHyperlinks false /IncludeInteractive false /IncludeLayers false /IncludeProfiles false /MarksOffset 18 /MarksWeight 0.250000 /MultimediaHandling /UseObjectSettings /Namespace [ (Adobe) (CreativeSuite) (2.0) ] /PDFXOutputIntentProfileSelector /NA /PageMarksFile /RomanDefault /PreserveEditing true /UntaggedCMYKHandling /LeaveUntagged /UntaggedRGBHandling /LeaveUntagged /UseDocumentBleed false >> > ]>> setdistillerparams> setpagedevice

breast cancer cells respond differentially to modulation of tgfb2 … · 2015. 10. 20. · tumor...

Documents