downregulation of the tgfb pseudoreceptor bambi in non ... · brane-bound inhibitor (bambi), which...

Molecular and Cellular Pathobiology

Downregulation of the TGFb PseudoreceptorBAMBI in Non–Small Cell Lung Cancer EnhancesTGFb Signaling and InvasionSebastian Marwitz1,2, Sofia Depner3,4,5, Dmytro Dvornikov3,5, Ruth Merkle3,4,5,Magdalena Szczygieł3,5, Karin M€uller-Decker6, Philippe Lucarelli3,5,Marvin W€asch3,5, Heimo Mairb€aurl5,7, Klaus F. Rabe2,8,9, Christian Kugler2,8,Ekkehard Vollmer1,2, Martin Reck2,8, Swetlana Scheufele2,10, Maren Kr€oger2,10,Ole Ammerpohl2,10, Reiner Siebert2,10, Torsten Goldmann1,2, and Ursula Klingm€uller3,4,5

Abstract

Non–small cell lung cancer (NSCLC) is characterized by earlymetastasis and has the highest mortality rate among all solidtumors, with the majority of patients diagnosed at an advancedstage where curative therapeutic options are lacking. In this study,we identify a targetable mechanism involving TGFb elevationthat orchestrates tumor progression in this disease. Substantialactivation of this pathway was detected in human lung cancertissues with concomitant downregulation of BAMBI, a negativeregulator of the TGFb signaling pathway. Alterations of epithelial-to-mesenchymal transition (EMT) marker expression wereobserved in lung cancer samples compared with tumor-free

tissues. Distinct alterations in the DNA methylation of the generegions encoding TGFb pathway components were detected inNSCLC samples compared with tumor-free lung tissues. In par-ticular, epigenetic silencing of BAMBI was identified as a hallmarkof NSCLC. Reconstitution of BAMBI expression in NSCLC cellsresulted in a marked reduction of TGFb-induced EMT, migration,and invasion in vitro, along with reduced tumor burden and tu-mor growth in vivo. In conclusion, our results demonstrate howBAMBI downregulation drives the invasiveness of NSCLC, high-lighting TGFb signaling as a candidate therapeutic target in thissetting. Cancer Res; 76(13); 3785–801. �2016 AACR.

IntroductionPrimary lung cancer is one of the most common malignancies

worldwide and has the highest mortality rate among all solidtumors (1), with non–small cell lung cancer (NSCLC) as themostfrequent tumor type comprising approximately 80%of diagnosed

cases. Due to the lack of characteristic early symptoms, 60% oflung carcinomas are detected in a progressed or already metasta-sized state (2) with a 5-year survival of only 16% independent ofthe tumor stage and size (3). Therapeutic interventions dependonthe tumor stage, and only at early stages of tumor progressionpatients are eligible for curative surgery. The majority of patientsare treated with chemotherapy. Still, the available therapies forlung cancer are very limited and rapid progression due to resis-tance to therapy is a major clinical challenge. To develop noveltherapeutic approaches, a detailed understanding of the mechan-isms contributing to lung cancer progression is essential.

The process of epithelial-to-mesenchymal transition (EMT) isknown to contribute to several lung diseases, such as idiopathicpulmonary fibrosis, asthma, and chronic obstructive pulmonarydisease (COPD; ref. 4). With respect to cancer progression,EMT has been observed in cancers of the breast or prostate (5),but this process has not yet been comprehensively studied in thecontext of lung cancer tissues.

TGFb is a cytokine with pleiotropic effects during develop-ment, inflammation, as well as tissue homeostasis. In addition,it exerts tight spatio-temporal control of differentiation, apo-ptosis, and proliferation in both epithelial and immune cells(6). Furthermore, TGFb is known to play a central role in theinduction of the EMT. Under pathologic conditions such aslung cancer, TGFb plays a dual role in the course of carcino-genesis. At primary stages, it acts as a tumor suppressor induc-ing apoptosis and controlling proliferation, whereas at laterstages the tumor-promoting effects take over and promote EMT,invasion, and metastasis. Furthermore, TGFb has immunosup-pressive properties and may enable cancer cells to evade the

1Pathology of the University Hospital of Lübeck and the LeibnizResearch Center Borstel, Borstel, Germany. 2Airway Research CenterNorth (ARCN), Member of the German Center for Lung Research(DZL), Großhansdorf, Germany. 3Systems Biology of Signal Transduc-tion, German Cancer Research Center, Heidelberg, Germany. 4BIO-QUANT, University of Heidelberg, Heidelberg,Germany. 5TranslationalLung Research Center Heidelberg (TLRC), German Center for LungResearch (DZL), Heidelberg, Germany. 6Tumor Models, German Can-cer Research Center, Heidelberg, Germany. 7Medical Clinic VII, SportsMedicine, University Hospital Heidelberg, Heidelberg,Germany. 8Lun-genClinicGroßhansdorf,Grobhansdorf,Germany. 9ChristianAlbrechtsUniversityKiel, Kiel,Germany. 10InstituteofHumanGenetics,Christian-Albrechts-University Kiel and University Hospital Schleswig-Holstein,Campus Kiel, Kiel, Germany.

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

S. Marwitz, S. Depner, and D. Dvornikov contributed equally to this article.

T. Goldmann and U. Klingm€uller share senior authorship.

Current address for P. Lucarelli: Systems Biology Group, Life Sciences ResearchUnit, University of Luxembourg, Belvaux, Luxembourg.

Corresponding Author: Ursula Klingm€uller, German Cancer Reseach Center(DKFZ), Im Neuenheimer Feld 280, 69120 Heidelberg. Phone: 49-6221-424481;Fax: 49-6221-424488; E-mail: [email protected]

doi: 10.1158/0008-5472.CAN-15-1326

�2016 American Association for Cancer Research.

CancerResearch

www.aacrjournals.org 3785

on March 5, 2020. © 2016 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from

Published OnlineFirst May 17, 2016; DOI: 10.1158/0008-5472.CAN-15-1326

http://cancerres.aacrjournals.org/

immune responses (7). The canonical TGFb cascade signalsthrough both TGFb receptors type 1 (TGFBR1) and type 2(TGFBR2). Upon binding of the ligand to TGFBR2, TGFBR1is phosphorylated and recruits the receptor (R)-SMADS—SMAD2 and SMAD3—which form a complex with the commonSMAD, SMAD4, migrate to the nucleus, and stimulate targetgene expression. Negative regulation is exerted by the inhibi-tory SMADs, such as SMAD7 (8), or at the receptor level by thepseudoreceptor bone morphogenetic protein and activin mem-brane-bound inhibitor (BAMBI), which is homologous toTGFBR1 but lacks an active kinase domain (9).

In the context of lung disease, the overexpression of BAMBIhas been observed in the tissues of COPD patients and hypo-thesized to control local inflammation and remodelingprocesses (10). The expression of BAMBI at high levels hasalso been detected in colorectal cancer (11) and ovarian cancer(12) and has been linked to metastasis in colorectal cancer(13). By contrast, BAMBI is epigenetically silenced in high-grade bladder cancer (14) and absent from breast cancer (15).Hence, the specific contribution of BAMBI to tumor progressionremains unresolved.

In this study, we comprehensively examined the TGFb sig-naling pathway in lung cancer and, using the negative regulatorBAMBI as an example, demonstrated that altered TGFb signal-ing contributes to malignant processes, such as EMT, migration,and invasion. The results suggest that TGFb signaling is apromising therapeutic target in lung cancer.

Materials and MethodsPatient material, ethical issue, and tissue processing

Lung tissue samples were obtained from patients who under-went lobectomy or pneumectomy due to therapeutic interven-tions at the LungenClinic Großhansdorf. Residual tumor andtumor-free lung tissues not needed for diagnostic purposes wereHepes-glutamic acid buffer mediated organic solvent protectioneffect (HOPE)-fixed and paraffin-embedded according to themanufacturer's instructions (DCS Innovative Diagnostik-Sys-teme). The use of patient materials was approved by the localethics committee of the University of L€ubeck (AZ 12-220), Ger-many. For the analysis of cancer tissues, 133 archived lung cancerspecimens of either adenocarcinoma or squamous cell carcinomadiagnosis were randomly selected and subjected to IHC. Thegender ratio was 43 female to 90 male patients, and the meanage at time of diagnosis was 66.06 years (� 9.9). Among theinvestigated lung cancer specimens, 74 were squamous cell car-cinomas (55.6%) and 59 adenocarcinomas (44.4%). As controls,23 tumor-free lung specimens were included, four of which werepaired with lung cancer specimens (Supplementary Table S2).A detailed overview of the patient characteristics as well as tumor–node–metastasis staging and other relevant information can befound in Supplementary Table S6.

Isolation of primary alveolar epithelial cells type IITumor-free lung tissues from surgical specimens of patients

who underwent pneumectomy or lobectomy at the Lungen-Clinic Großhansdorf were used for isolation of primary type IIalveolar epithelial cells (AECII). Fresh tissues were rinsed withsterile PBS to remove excess blood, stored in RPMI 1640 (LifeTechnologies) cell culture medium, and supplemented with10% FCS and 1% penicillin/streptomycin at 4�C until further

use. To extract AECII, the tissue was manually dissected withsurgical razors and rinsed with AECII buffer (1.9 mmol/Lcalcium chloride dihydrate, 1.3 mmol/L magnesium sulfateheptahydrate, 136 mmol/L sodium chloride, 6.1 mmol/Lpotassium chloride, 3.2 mmol/L di sodium hydrogen phos-phate, 6.1 mmol/L glucose, and 9.9 mmol/L HEPES) over asieve to remove residual blood. Tissue pieces were dissociatedby incubation for 60 minutes at 37�C and 5% CO2 in AECIIbuffer supplemented with 2 mg/mL Dispase II (Roche AppliedSciences) under constant stirring. The lung tissue pieces werefurther filtered through nylon gaze of decreasing pore size(100 mm, 50 mm, 20 mm) and finally sedimented for 15 minutesat 478 g at room temperature. The resulting cell pellet wasresuspended in 50 mL of AECII buffer supplemented with0.001% Accutase (MilliPore; v/v) and 20 mg/mL DNAseI (Roche Applied Sciences). Each 10 mL of this suspensionwas applied onto 10 mL Biocoll gradient solution (Biochrome)and centrifuged for 25 minutes at 478 g and room temperature.The resulting interphase was washed with AECII buffer at 478 gfor 15 minutes at 4�C and finally resuspended in cell culturemedium. The 5 � 107 cells were seeded per 6-cm petridish and incubated for 20 minutes at 37�C in an incubator toallow adherence of immune cells. A total of 1 � 107 of non-adhered cells were resuspended in 80 mL MACS buffer (PBSpH 7.2 with 0.5% BSA and 2 mmol/L EDTA), mixed with 20 mLanti–CD45-conjugated para-magnetic beads (Miltenyi) andfinally incubated on a rotating device for 15 minutes. Depletionof CD45-positive immune cells with LD columns (Miltenyi)was conducted according to the manufacturer's instructions.

Production of tissue microarray and cellblocksTissue microarrays from HOPE-fixed, paraffin-embedded

tissues were produced as described by Goldmann and collea-gues (16). HOPE-fixed cells were dehydrated and subsequentlyparaffin-embedded according to Marwitz and colleagues (17)to produce cell blocks for immunocytochemistry.

ImmunohistochemistryHOPE-fixed, paraffin-embedded tissues and tissue microar-

rays were cut on amicrotome (Leica SM 2000) to obtain sectionsof 1-mm thickness and mounted on SuperFrostþ (MenzelGl€aser) glass slides. Subsequent deparaffinization was done byincubation in 100% isopropanol for 10 minutes at 65�C. Sec-tions were shortly air dried, rehydrated in 70% acetone for 10minutes at 4�C, and transferred into distilled water for 10minutes at 4�C. Endogenous peroxidases were blocked by incu-bation in 3% H2O2 for 10 minutes. Primary antibodies werediluted with antibody diluent (Zytomed Systems) as following:mouse anti-BAMBI monoclonal (clone 4e8; eBioscience) 1:100;rabbit anti-BAMBI polyclonal (Sigma-Aldrich) 1:100; rabbitanti-SMAD2 (clone D43B4; Cell Signaling Technologies);anti–phospho-SMAD2 (clone 138D4; Cell Signaling Techno-logies); rabbit anti-SMAD3 (clone C67H9; Cell SignalingTechnologies); rabbit anti–phospho-SMAD3 (C25A9; Cell Sig-naling Technologies); rabbit anti-SMAD4 (clone EP618Y; CellSignaling Technologies); rabbit anti-TGFb polyclonal (Abcam);mouse anti-SNON monoclonal (clone 2F6; Abcam); rabbitanti–E-cadherin monoclonal (clone 24E10; Cell Signaling Tech-nologies); mouse anti–N-cadherin monoclonal (5D5; Abcam);and rabbit anti-HMGA1 monoclonal (clone EPR7839; Abcam)all 1:200; mouse anti-SMAD7 (clone 293039; R&D Systems)

Marwitz et al.

Cancer Res; 76(13) July 1, 2016 Cancer Research3786




and rabbit anti-Osteopontin/SPP1 polyclonal (Zytomed Sys-tems) all 1:100. For visualization, a horseradish peroxidase(HRP)–conjugated polymer kit was used according to the man-ufacturer's instructions (ZytoChemPlus Kit; Zytomed Systems).Chromogenic reaction was conducted with permanent 3-ami-no-9-ethylcarbazol (AEC; Permanent AEC Kit; Zytomed Sys-tems) and stopped with distilled water. The slides were subse-quently dehydrated in graded series of ethanol (70%, 90%, 2�95%, 2� 100%, xylene) and mounted with Pertex (Medite).Negative controls were included by omission of primary anti-body in each staining, and photomicrographs were takenwith a CCD camera (Infinity 4; Lumenera). Contrast andbrightness of images were adjusted using FixFoto software(Joachim Koopman Software). For evaluation of target expres-sion, a semiquantitative score was applied based on histologyanalysis of the whole specimen as following: negative (0), focaland weak expression (1), frequent intermediate expression (2),and strong expression and dominating feature of specimen (3).The whole specimen was considered for evaluation of micro-scopic slides.

Immunofluorescence staining of paraffin-embedded tissuesCutting, mounting, and deparaffinization of tissue slides were

conducted as described above. Rehydrated slides were incubatedwith either rabbit anti-pSMAD3 (clone C25A9; Cell SignalingTechnologies) and mouse anti–N-cadherin (clone 5D5; Abcam)or rabbit anti–E-cadherin (clone 24E10; Cell Signaling Tech-nologies) and mouse anti–N-cadherin (clone 5D5; Abcam) alldiluted 1:200 in antibody diluent (Zytomed Systems) for 45minutes at room temperature in a humid chamber. Detection ofprimary antibodies was done with goat anti-rabbit and goatanti-mouse secondary antibodies conjugated with Alexa488(Life Technologies) or Tetramethylrhodamine (Life Technolo-gies), respectively, diluted 1:200 in antibody diluent for 45minutes. Thoroughly washed slides were mounted with Vecta-Shield mounting media with DAPI (Vectorlabs) and kept in thedark at 4�C until read-out.

RNA and DNA isolationThe samples for epigenetic studies were enriched for tumor

cells by microdissection of hematoxylin and eosin (H&E)–stained lung cancer specimens under a microscope. Dissectedcells were transferred into a 1.5-mL reaction tube, and DNAisolation was done with the QIAmp Mini Kit (Qiagen) accord-ing to the manufacturer's instructions. For transcriptomeanalysis and quantitative real-time PCR (qRT-PCR), sectionsfrom paraffin blocks were cut on a microtome and deparaffi-nized prior to RNA isolation. Briefly, each sample was rotatedwith 1 mL xylene for 10 minutes at room temperature. Thetissue was pelleted by centrifugation at 12,000 g for 5 minutes,the supernatant was discarded, and the procedure was repeat-ed once. Deparaffinized tissue was washed twice with 1 mL100% ethanol for 10 minutes at room temperature. Residualethanol was removed by vacuum centrifugation, and driedpellets were homogenized in RLT buffer (Qiagen) with anelectronic pestle (VWR International). RNA was isolated usingthe RNeasy Mini Kit (Qiagen) according to the manufacturer'sinstructions. For qRT-PCR of the cell lines, RNA was extractedusing the RNeasy Mini Plus Kit (Qiagen) according to themanufacturer's instructions.

Microarray analysisRNA integrity for microarray analysis was determined using

the Agilent Bioanalyzer with the RNA Nano 6000 Kit accordingto the manufacturer's instructions (Agilent). Transcriptomeanalysis was conducted as described in the One-Color Micro-array-Based Gene Expression Analysis Low Input Quick AmpLabeling protocol version 6.6. Shortly, 200 ng of total RNA wasmixed with the RNA Spike-In Mix as internal control (AgilentOne Color RNA Spike-In Kit) and reverse transcribed intocDNA by Moloney murine leukemia virus reverse transcriptase.Labeling with Cy3-CTP was conducted with T7 RNA polymer-ase. Cy3-labeled cRNA was purified using the RNeasy Mini Kit,and specific activity was calculated using a NanoDrop 2000spectrophotometer (Thermo Fisher Scientific). The 1,650 ng oflabeled cRNA was hybridized on Agilent Human Gene Expres-sion 4 � 44k V2 arrays and scanned with an Agilent SureScanmicroarray scanner. Raw data were extracted with Agilent Fea-ture Extraction software v11 applying 1-color Green GeneExpression protocol. For hierarchical clustering and heat mapanalysis, Agilent GeneSpring GX software version 12.6 was used.Hierarchical clustering was computed with GeneSpring GX 12.6on entities (genes) and conditions (samples) with squaredEuclidean distance metrics and Ward's linkage rule. Quantile-normalized gene expression data were computed using Direc-tArray software (Oaklabs). Microarray results are deposited tothe GEO database (GEO accession number GSE74706).

cDNA synthesis and qRT-PCRcDNA was generated from 1 mg of total RNA using the Maxima

First Strand cDNASynthesis Kit (ThermoScientific) for tissues andthe High Capacity cDNA Reverse Transcription Kit (AppliedBiosystems) for cell lines according to the manufacturer's instruc-tions. cDNA templates were analyzed by qRT-PCR on a Light-Cycler 480 (Roche Applied Science) thermocycler (cell lines andtissues) using the LightCycler 480 Probes Master with final0.4 mmol/L primer and 0.2 mmol/L fluorescein amidite-labeledhydrolysis probes (Universal Probe Library, Roche Applied Sci-ence). Crossing point values were calculated using the second-derivative-maximum method of the LightCycler 480 Basic Soft-ware (Roche Applied Science). Concentrations were normalizedusing the geometricmeanofb-glucuronidase (GUSB) and esteraseD (ESD) or glyceraldehyde-3-phosphate dehydrogenase(GAPDH) and glucose-6-phosphate 1-dehydrogenase D (G6PD)for cell lines and the geometric mean of ESD, GAPDH, phospho-glycerate kinase 1 (PGK1), hypoxanthine-guanidine phosphori-bosy ltransferase (HPRT), and TATAA-box-binding protein (TBP)for tissues. The BAMBI mRNA quantification for different celltypeswas performed in independent experimentswith at least onecell line overlapping for scaling the results to the other cell types.Primers were designed using the UniversalProbe Library AssayDesign Center (Roche Applied Science). The PCR primersequences used are listed in Supplementary Table S7.

DNA methylation analysisMethylation analysis of HOPE-fixed, paraffin-embedded

tissues was conducted as described elsewhere (18). Briefly,bisulfite conversion of genomic DNA applying the EZ DNAMethylation Kit (ZymoResearch) as well as subsequent DNAmethylation analysis using the Infinium HumanMethyla-tion450k BeadChip (Illumina Inc.) was performed accordingthe manufacturers' instructions. The HumanMethylation450K

Role of BAMBI in Lung Cancer

www.aacrjournals.org Cancer Res; 76(13) July 1, 2016 3787




BeadChip aims at assaying more than 480,000 CpG sitesin parallel. DNA methylation data were analyzed using theGenomeStudio software (v2011.1; methylation module 1.9.0;Illumina, Inc.) applying the default settings. The array's intrin-sic controls were used for normalization of the data, and nofurther background subtraction or normalization steps wereperformed. Identification of differentially methylated CpGloci, principal component analysis (PCA) as well as hierarchi-cal cluster analysis was performed using the OMICS Explorer2.3 (Version 2.3(45), Qlucore) software using DNA methy-lation values (average beta values) obtained from the Geno-meStudio software. HumanMethylation450k BeadChip resultsare deposited to the GEO database (GEO accession numberGSE75008).

To validate the DNA methylation values detected byHumanMethylation450K BeadChip, DNA from 14 lung car-cinoma samples as well as their corresponding normal con-trols was bisulfite-converted using the EpiTect Kit (Qiagen)following the manufacturer's instruction. Subsequently, bisul-fite pyrosequencing of the selected loci was performed as de-scribed by (19) to determine the DNA methylation state atcg05046589 [forward primer: ttttttaggggtgaggttttaggtag; reverseprimer (biotinylated): aaccactcaatcaaaatacatctttaca; sequencingprimer: aggttttaggtaggtg], cg17482424 [forward primer: aggaaa-taggtattagtatgatggtagta; reverse primer (biotinylated): actttac-aacctacccctttt; sequencing primer: gagttgattttttttaaagagtt], andcg07232003 [forward primer: tgtaaagttagatttggaatgtatgg; reverseprimer (biotinylated): actcaaaaaaccttcaaactattca; sequencing pri-mer: tttattggggtatgtatag].

Selection of genes involved in the TGFb signaling pathwayand EMT

The a priori–generated lists of genes for microarray andmethylation analyses for TGFb pathway members wereretrieved from public databases (Wikipathways and KEGG)as well as from literature search. Because the databases con-tained no available list for EMT genes, the set of genes wasadapted from commercially available EMT array sets andliterature search.

Comparison of DNA methylation and gene expression dataTo analyze correlation between differential methylation and

aberrant gene expression in TGFb pathway member genes, themean differential methylation in all tumor-control sample pairswas determined for each locus. Only loci with mean differentialvalues �10% were included in further analysis. Finally, themedian of the means of the CpG loci identified was plottedagainst the log value of the difference in gene expression betweentumor and normal tissues for each gene. R (version 3.1.1) wasused for calculations and preparing graphs.

Cell lines and culture conditionsHuman lung adenocarcinoma cell lines A549, H1975, and

SK-MES1 were purchased from the ATCC. A549 and H1975were authenticated with Multiplex human Cell line Authenti-cation Test (Multiplexion). SK-MES1 cells were used within 6months after purchasing from the cell bank. All cells weregrown in a humidified atmosphere with 5% CO2 at 37�C. AllNSCLC cell lines as well as human primary AECII cells andPhoenix ampho packaging cell line (20) were cultivated inDMEM (Lonza) supplemented with 10% FCS (Gibco) and 1%

penicillin/streptomycin (Gibco). For the GFP-expressing andBAMBI-GFP–expressing cell lines A549 and H1975, 1 mg/mLpuromycin (Sigma) was added. As a starvation medium,DMEM without phenol-red (Lonza) supplemented with1 mg/mL BSA (Sigma), 1% penicillin/streptomycin (Gibco),and 1% L-glutamine (Gibco) was used.

DNA transfectionHuman BAMBI-GFP insert was obtained from the plasmid

pCMV6-AC-GFP (RG200195; OriGene) using BamHI/Pme Irestriction sites. The insert was recloned into the retroviralexpression vector with a Tet-inducible promoter pMOWSIN-TREt (21). As a control, pMOWSIN-TREt-GFP was used. BothpMOWSIN-TREt-BAMBI-GFP and pMOWSIN-TREt-GFP wereused in combination with pMOWS-rtTAM2 encoding the cDNAfor transactivator protein (21). Transfection of the Phoenixampho packaging cell line was performed by calcium phos-phate precipitation. Transducing supernatants were generated24 hours after transfection by passing through a 0.45-mm filter,supplemented with 8 mg/mL polybrene (Sigma), and appliedin the proportion 1:3.5 (pMOWS-rtTAM2:pMOWSIN-TREt-BAMBI-GFP or pMOWSIN-TREt-GFP) for spin infection of1.5 � 105 A549 and H1975 cells. Stably transduced A549 andH1975 cells expressing GFP-tagged human BAMBI or as acontrol only GFP were selected in the presence of 1.0 mg/mLpuromycin (Sigma) 48 hours after transduction.

siRNA transfectionsiRNA against human BAMBI was purchased fromDharmacon

(#LU-019596-00-0002). A549 and SK-MES1 cells were trans-fected with Lipofectamine RNAiMAX (TheroFisher Scientific;#13778150) according to the manufacturer's procedures. Briefly,cells were trypsinized, mixed with required amount of Lipofecta-mine/siRNA mixture, incubated for 5 minutes, and plated inpenicillin/streptomycin-free medium. Twelve hours after siRNAtransfection, medium was replaced with antibiotic-containingDMEM. Cells were used for TGFb stimulation 36 hours aftersiRNA transfection.

Immunoprecipitation and quantitative immunoblottingFor experiments, A549, H1975, and SK-MES1 cells were

cultivated in growth factor (GF)–depleted medium for 24 hoursand stimulated with TGFb (R&D Systems) for indicated timeperiods or left untreated, and then lysed. Before lysis, 1 mL ofcell supernatant was removed for further analysis by Multiplexbead–based assay (BioRad).

Prior to experiments, all cell lines overexpressing GFP orBAMBI-GFP were treated with doxycycline (DOX; 5 mg/mL;Sigma) for 24 hours and then transferred into starvation medi-um supplemented with 1.0 mg/mL puromycin (Sigma) andDOX (5 mg/mL; Sigma) for further 24 hours.

For lysis, cells were scraped in 500 mL lysis buffer (1% NP40,150 mmol/L NaCl, 20 mmol/L Tris–HCl pH 7.4, 10 mmol/LNaF, 1 mmol/L EDTA pH 8.0, 2 mmol/L ZnCl2 pH 4.0,1 mmol/L MgCl2, 2 mmol/L Na3VO4, 20% glycerol, 2 mg/mLaprotinin, and 200 mg/mL AEBSF). Lysates were rotated for30minutes at 4�C and centrifuged for 10minutes at 14,000 rpmand 4�C. Protein concentration was determined by BCAassay, and 50 mg of the lysates were used for total cell lysates(TCL) and 750 mg for immunoprecipitation (IP). For IP,the antibodies against SMAD2/3, SMAD2, SMAD3, or SMAD4

Marwitz et al.





(BD 610843; Cell Signaling Technology; clone D43B4; cloneC67H9; #9515, respectively; dilution 1:100) and Protein Asepharose (GE Healthcare) were added and rotated over nightat 4�C. TCLs and immunoprecipitated proteins were separat-ed by SDS-PAGE, transferred to PVDF (TCLs) and nitrocellu-lose (IP) membranes, and immunoblotting was performedwith anti-SMAD2/3 (BD 610843), anti-SMAD4 (Cell SignalingTechnology; #9515), anti-SMAD7 (R&D; clone 293039), anti-pSmad2 (Cell Signaling; 138D4), anti-pSMAD3 (Cell SignalingTechnology; clone C25A9), anti-BAMBI (Sigma; HPA010866),and anti-GFP (Roche; clones 7.1 and 13.1) antibodies. Blotsprobed with anti-pSMAD2 or anti-pSMAD3 antibodies werestripped by using the stripping buffer (6.25 mL 1 mmol/LTris pH 6.8, 20 mL 10% SDS, 0.7 mL b-Mercaptoethanol)for 20 minutes at 65�C and subsequently reprobed with anti-bodies against total proteins. For all primary antibodies,1:1,000 dilution was used. HRP-conjugated goat anti-rabbit(Dianova; 111-035-144) and goat anti-mouse (Dianova; 115-035-146) were used for enhanced chemiluminescent (ECL)detection (GE Healthcare). IRDye infrared dyes (LI-COR#926-32211 and #926-68070) secondary antibodies were usedfor detection with infrared Odyssey imager (LI-COR). Signalquantification was performed using an ImageQuant system (GEHealthcare). Replicates of dose response data were scaled andaveraged using methods described in (22).

In vitro cell migration assayFor two-dimensional (2D) migration assay, H1975 cells

overexpressing BAMBI-GFP and GFP were seeded in 24-wellplate (Zell Kontakt; # 3231-20) at a density of 5,000 cells/well.Cells were allowed to attach for 18 hours, and after that 5 mg/mLDOX (Sigma) was added into the culture medium. After48 hours, cells were serum-starved for 3 hours and stimulatedwith 0.25 ng/mL TGFb (R&D Systems). After 18 hours ofstimulation, cells were stained with Hoechst (Sigma) andimaged on an environment-controlled microscope (IX81;Olympus). Images were acquired with an UPlanSApo 10�/0.4NA objective lens (Olympus) in 15-minute intervals for22 hours. Nine positions per well (3 � 3 grid) were imagedand stitched with ImageJ plugin. Single-cell tracking was per-formed with ImageJ Mtrack2 plugin. Speed of each trackedcell was calculated by dividing total travelled distance bytotal time, for which cell was tracked. Persistence was calcu-lated by dividing the distance between the first and the lastpoints, where the cell was tracked, by total travelled distance. Theresulting number was multiplied by the square root of time,for which cell was tracked divided by maximal possible trackingtime, in order to penalize cells, which were tracked for a shorterperiod of time.

3D collagen invasion assayThree-dimensional (3D) collagen gels were prepared as

described previously (23). Briefly, ice-cold 1 mol/L HEPESbuffer, 0.7 mol/L NaOH, 10� PBS (pH 8.0), and bovine skincollagen G solution (L1613; Biochrome) were mixed in1:1:2:16 ratio, respectively. Fifty mL of the resulting solutionwas added per well of 96-well plate with a flat bottom (BD#353376). A plate was kept overnight at 4�C and then for atleast 1 hour at 37�C to allow gelation of the collagen. Aftergelation, 10,000 cells per well were seeded on top of the matrix.Expression of GFP or BAMBI-GFP in H1975 cells was induced

with 5 mg/mL DOX. Twenty-four hours after DOX induction,cells were stimulated with 0.25 ng/mL TGFb and left forinvasion for 96 hours. SK-MES1 cells were siRNA transfectedand seeded on a gel. Thirty-six hours after, transfected cellswere stimulated with 1 ng/mL TGFb and left for invasion for96 hours. Afterwards, cells were fixed in 3.7% PFA for 1 hourand subsequently stained with Hoechst (Sigma). Collagen-embedded cells were imaged using LSM710 confocal micro-scope (Carl Zeiss) equipped with EC Plan-Neofluar DIC 10�/0.3 NA objective lens (Carl Zeiss). For each well, 2 � 2 tilez-stack was acquired. Image analysis was performed usingImaris software (Bitplane). Spots detection algorithm wasapplied to assign a spot for fluorescent intensity of each indi-vidual nucleus. Resulting spots were filtered by their z-positionto separate collagen-invaded cells from the cells that remainedon top of the matrix. Percentage of invaded cells was calculatedand used as output.

Lung colonization assay of A549-GFP and A549-BAMBI-GFPcells in nude mice

Animal care and animal experiments were performedaccording to the national guidelines and were reviewed andconfirmed by an Institutional Review Board/ethics committeeheaded by the local animal welfare officer of the GermanCancer Research Center, Heidelberg, Germany. All experimentswere in accordance with the approved guidelines of the respon-sible national authority, the local Governmental Committeefor Animal Experimentation (Regierungspr€asidium Karlsruhe,Germany; license G193/10). Mice were maintained at a12-hour light-dark cycle with unrestricted diet and water. Twomillion A549-GFP and A549-BAMBI-GFP cells in 100 mL ofPBS were i.v. injected into the lateral tail vein of 7- to 8-week-old female NMRI nu/nu mice purchased from Charles Riverand kept on Kliba diet 3307 with 12 randomized mice/group.Three days prior to cell inoculation, treatment of all mice with5 mg/mL DOX (Sigma) in drinking water containing 5%saccharose was started and continued throughout the experi-ment. TGFb was injected i.p. (4 mg/kg bodyweight, dissolvedat 400 ng/mL in 4 mmol/L HCl containing 0.1% mouseserum albumin, 10 mL/g mouse) once daily at days 1, 6, 11,and 16 after cell inoculation. Prior to cervical dislocation onday 70 after cell inoculation, blood was collected under iso-flurane inhalation anesthesia (1–1.5% in O2, 0.5 L/min) fromthe retrobulbar plexus. No mice were lost due to adverse effectsof TGFb.

Mouse lung colonization assay analysisMouse lungs were dissected and separated into left and right

lung and subsequently HOPE-fixed and processed as describedabove. All lung tissue specimens were randomly cut andmounted on SuperFrost slides for H&E staining. For the quan-tification of the lung tumor burden, the right and left lungsof each animal were analyzed by histology-trained investiga-tors (T. Goldmann and S. Marwitz), and images were taken at10� magnification. The area of tumors within the lungs wasmanually tagged within the Infinity Analyze software (Lume-nera), and tumor area per animal was quantified using acalibrated micrometer slide. The total area of lung tumorburden was calculated from the sum of individual tumornodule areas within the alveolar region of both lungs fromeach animal. To calculate the metastatic potential of A549-GFP






and A549-BAMBI-GFP cells, we manually counted tumor nodu-les that consisted of at least two connected tumor cells. Humanorigin of lung tumor cells within the mouse lungs was verifiedvia IHC using rabbit anti human HLA-A antibody (cloneEP1395Y; Abcam). The expression of the human transgenewas confirmed by qRT-PCR (Supplementary Fig. S10). Activa-tion of the TGFb pathway within human lung tumor cellswas investigated by immunohistochemical analysis of SMAD3phosphorylation.

Assessing GFP and BAMBI-GFP expression in mice lungsTo investigate the inducible expression of GFP or BAMBI-

GFP in our animal experiments, total RNA was extracted fromHOPE-fixed, paraffin-embedded tissues as described elsewhere(17). DNA was digested with DNAse I (Thermo Fisher Scien-tific), and 200 to 800 ng of cDNA synthesized with the MaximaFirst Strand cDNA synthesis Kit (Thermo Fisher Scientific)according to the manufacturer's instructions. Oligonucleotides

that target either GFP from A549-GFP control cells (Gene BankAcc. No. U50963.1) or the GFP sequence from donor vectorthat was used for cloning BAMBI-GFP (pCMV RG200195 7.3;OriGene) were generated using the Probe Finder software(Roche Applied Science) with default options and normalizedto expression of human RPL32 gene via qRT-PCR using theRoche Probes Master Mix and Universal Probe library accordingto the manufacturer's instructions with 45 cycles of amplifica-tion on a Roche LightCycler 480 II (5 minutes: 95�C Hot Start;10 seconds 95�C denaturation; and 30 seconds 60�C anneal-ing). Advanced relative quantification was used to displayrelative expression normalized to RPL32 as reference gene.

Statistical analysisResults are shown as mean � SD. Comparisons of datasets

were performed using unpaired t test (test group comparedwith control group). For experiments with human material,two-sided paired t test was used to compare the samples groups

A Adenocarcinoma

TGFβ pSMAD3pSMAD2

SMAD4SMAD3SMAD2

BAMBISNONSMAD7

TGFβ pSMAD3pSMAD2

SMAD4SMAD3SMAD2

BAMBISNONSMAD7

TGFβ pSMAD3pSMAD2

SMAD4SMAD3SMAD2

BAMBISNONSMAD7

B Squamous cell carcinoma

C Tumor-free lung Expression in lung cancer tissues

Freq

uenc

y (%

)

pSMAD3TGFβ

85

pSMAD2 SMAD4 SMAD2 SMAD3 SMAD7 SNON BAMBI

1001.598.599.211.322.671.487.278.2

0

20

40

60

80

3

0

2

1

Expression in aveolar epithelial cells type II of tumor-free tissue34.7 30.48734.8017.460.817.317.3

Freq

uenc

y (%

) 100

0

20

40

60

800

3

2

1

pSMAD3TGFβ pSMAD2 SMAD4 SMAD2 SMAD3 SMAD7 SNON BAMBI

D

Figure 1.BAMBI is absent and the TGFb signaling pathway is abundantly activated in lung cancer tissues. A–C, detection and localization of TGFb pathwayproteins via IHC on HOPE-fixed paraffin-embedded lung adenocarcinomas (n ¼ 59 patients; A), lung squamous cell carcinomas (n ¼ 74 patients; B),and tumor-free lung tissues (n ¼ 23 patients; C). All the images are at �40 magnification. Scale bar, 100 mm. Positive staining is indicated by a redcolor (AEC). For each antibody, a positive signal was observed for at least some of the samples. Representative images are shown for each tumor typeand for tumor-free lung tissue. D, scoring of expression levels in lung cancer tissues (n ¼ 133 patients) and in healthy AECII from tumor-free lungtissue (n ¼ 23 patients). Bar charts indicate semiquantitative scores based on the histologic analysis of the entire specimen as follows: negative (0),focal and weak expression (1), frequent intermediate expression (2), strong expression and dominating feature of specimen (3). Numbers abovethe bars display the total positive cases observed overall in percentage.

Marwitz et al.





(tumor to tumor-free). Comparison of SMAD2/3 activation instably transduced cells was performed using two-way ANOVAwith repeated measurements. Otherwise, the Mann–Whitneyrank sum test was used.

ResultsActivation of the TGFb pathway in human lung cancertissues

To examine the status of the TGFb signaling pathway in lungcancer, we employed IHC to monitor key components ofTGFb signaling in tissues from 133 lung cancer patients withadenocarcinomas (Fig. 1A) and squamous cell carcinomas(Fig. 1B) and in 23 tumor-free lung samples (Fig. 1C). Out ofthose tumor-free specimens, four (17.4%) were matched sam-ples for NSCLC lung cancer tissues. We analyzed the presenceof TGFb signaling components (Fig. 1D) in both lung cancertypes and compared the expression to AECII of the tumor-freetissues. This cell type is considered to be a possible origin ofadenocarcinomas (24) and has been recently discussed toexhibit features of progenitor cells (25).

The TGFb ligand was detected in the majority of the lungcancer tissues (85%) but only in 34.7% of the AECII from thetumor-free samples. Weak staining for SMAD2 and SMAD3was observed in 22.6% and 11.3% of the lung cancer samples,respectively, whereas in the AECII, SMAD3 was not detectableby means of IHC, and SMAD2 was present in 17.4% of thesamples. Immunoblot analysis confirmed 25 times lowerabundance of SMAD3 than SMAD2 in isolated populationsof AECII cells (Supplementary Fig. S9). For the majority of thelung cancer samples, phosphorylation, indicative of pathwayactivation, was observed for SMAD2 (78.2%) and SMAD3(87.2%). In contrast to this, phosphorylation was only presentat low levels in 17.3% of the examined AECII. SMAD4 waswidely expressed in the lung cancer cells (71.4%) and AECII(60.8%), and SNON (SKI novel protein) was detected in thecytoplasm of almost all the lung cancer specimens (98.5%) aswell as in the AECII (87%). Protein expression of the negativeregulator SMAD7, a direct target gene of TGFb signaling, wasdetected in almost all the lung cancer samples (99.2%) and wasprimarily localized in nuclear regions. In the AECII, SMAD7was only moderately expressed (34.8%). Surprisingly, the

A B

Tumor-free lung Tumor Tumor-free lung Tumor

mR

NA

Exp

ress

ion

tum

or (a

.u.) TGFBR1

0.001

0.01

0.1

1

10

mRNA Expression tumor-free lung (a.u.)

0.001 0.01 0.1 1 10

TGFBR2

0.001 0.01 0.1 1 10

0.001

0.01

0.1

1

10

SMAD7

0.001

0.01

0.1

1

10

0.001 0.01 0.1 1 10

BAMBI

0.001 0.01 0.1 1 10 0.001

0.01

0.1

1

10

SMAD2 SMAD3

SMAD4 SNON

BAMBISMAD7C

1,600

1,200

800

400

5,000

4,000

3,000

2,000

1,000

Tumor-free lung Tumor

TGFBR1

mR

NA

Exp

ress

ion

(a.u

.)

40

80

100

0Tumor-free lung Tumor

TGFBR2

14,000

6,000

2,000

0

0 0

0

20

40

60

80

0

200

400

600

800

1,000n.s. n.s.

100

200

300

100

200

300

0 0

400

500

Tumor-free lung Tumor Tumor-free lung TumorTumor-free lung Tumor Tumor-free lung Tumor

mR

NA

Exp

ress

ion

(a.u

.) n.s. n.s.

SQC AC Tumor-free lung SQCAC

10,000

SPP1FOSBFOSL1EGFLEF1SKIRUNX2FOXH1SKILADAM12RHOATGIF1HRASTGFB1LEFTY1SMAD6FOSKLF10ID1TGFB2RUNX3SNAI1JUNBMAPK3TFE3JAK1NFKB1TGFBR1CREBBPSMAD9LTBP1BMP4NOGZNF423ID3SMAD7ZEB1ZEB2ID2ID2BAMBIID4TGFBR3LEFTY2ENGTGFBR2

Figure 2.TGFb-regulated genes are differentially expressed in human lung cancer tissues and tumor-free lung tissues. A, nineteen samples of HOPE-fixedparaffin-embedded tissues from matched lung cancer and tumor-free lung samples were macrodissected and subjected to analysis on AgilentHuman Whole Genome Expression arrays. A list of 80 direct and indirect TGFb pathway members (Supplementary Table S2) as well as up- anddownstream acting mediators were used as inputs for gene expression analysis to analyze the transcriptional regulation of this pathway. Aheatmap of hierarchical clustering computed with a squared Euclidian distance matrix and Ward's linkage rule is shown for the significant resultsfrom paired t tests of tumor-free lung against matched tumor samples with a Benjamini–Hochberg multiple comparison correction (P � 0.05).Relative gene expression levels are displayed for individual cases according to the histologic tumor entity abbreviated as follows: SQC, squamouscell carcinoma; AC, adenocarcinoma. B, quantile-normalized microarray data indicating the expression of TGFb receptors TGFBR1 and TGFBR2,BAMBI, and downstream mediators in matched tumor-free lung tissues and lung cancer samples (n ¼ 18) compared with a two-sided paired t test,with �� , P � 0.01 and �� , P � 0.001 regarded as significant; n.s., not significant. C, validation of the microarray data for individual targets withqPCR assays. Gene expression in tumor-free and tumor samples from the 19 patients. Upregulated genes are displayed above the black line,which indicates a ratio of 1 (equal expression rates in tumor and tumor-free lung tissues), and downregulated genes are displayed below this line.Blue dots, squamous cell carcinoma; red dots, adenocarcinoma.






A

C

D

B

SQC AC Tumor-free lung

2 (13%)

3 (7%)

1 (27%)

–0.2 –0.1 0.0 0.1 0.2Delta DNA-methylation

–4

–2

0

2

4

Log

(del

ta g

ene

expr

essi

on)

DNA-methylation vs. gene expression

E1 (42%)

3 (16%)

2 (23%)

DN

A-m

ethy

latio

n

II I

III IV

Figure 3.Methylation patterns of TGFb pathway members distinguish between lung tumor and tumor-free tissues. A and B, DNA methylation analysis of TGFb pathwaymember genes in tumor and tumor-free lung tissues. Forty-six samples of macrodissected, HOPE-fixed, and paraffin-embedded tissues were subjectedto genome-wide methylome analysis using Illumina's HumanMethylation450K BeadChips. The same a priori list of TGFb pathway members used forthe microarray analysis was used to extract DNA methylation values for the CpG loci located in these genes for an unsupervised hierarchical clusteranalysis (A) and a PCA (B). The data were normalized for visualization (mean ¼ 0). C, correlation of the DNA methylation and gene expression data forTGFb pathway member genes differentially methylated between lung cancer and tumor-free lung tissues. Twenty-six TGFb pathway member genesdifferentially methylated between lung cancer and tumor-free lung tissues (D methylation � 10%) and aberrantly expressed in lung cancer samples(� 1.5-fold) were identified as described in Materials and Methods. The mean differential methylation in all the tumor-control sample pairs wasdetermined for each CpG locus included in the analysis. Subsequently, the median of the means of the CpG loci was calculated for each gene, resultingin one differential DNA methylation value per gene (D DNA methylation). The median was plotted against the log value of the difference in geneexpression between the tumor and normal tissues for each gene. (Continued on the following page.)

Marwitz et al.





protein expression of an additional negative regulator of TGFbsignaling, the TGFb pseudoreceptor BAMBI, was almostcompletely absent from the lung cancer tissues (1.5%) butfrequently detected in the AECII (30.4%). As assessed by theFisher exact test, a statistically significant difference in theprotein expression of TGFb, pSMAD2, pSMAD3, SMAD3,SMAD7, SNON, and BAMBI was observed in the lung cancercells compared with the AECII (Supplementary Table S1).Furthermore, the frequency of pathway activation, as indicatedby the presence of pSMAD2 and pSMAD3 in the same sample,was significantly higher (P � 0.001) in the lung cancer tissuescompared with the tumor-free lung tissues (SupplementaryFig. S1A). No significant differences were observed betweenadenocarcinomas and squamous cell carcinomas (P ¼ 0.0549;Supplementary Fig. S1B); however, a tendency cannot beneglected. IHC scoring of the four matched tumor-free andNSCLC lung cancer tissues showed a similar trend as theanalysis of the entire sample set (Supplementary Table S2).

Differential expression of TGFb-regulated genes in humanlung cancer and tumor-free lung tissues

The observed increased activation of TGFb signaling inlung cancer suggested concomitant alterations at the tran-scriptional level. Therefore, we analyzed the transcriptome of19 matched pairs of tumor-free lung tissues and macrodissectedlung NSCLC tumors with Agilent Human Gene Expressionarrays. An a priori–generated list of genes encoding TGFb corepathway members as well as up- and downstream acting med-iators (Supplementary Table S3) was used as an input file for thestatistical analysis of differentially expressed genes. A paired ttest revealed distinct differences between the lung tumors andmatched tumor-free tissue samples, and the results were clus-tered hierarchically and displayed as a heat map (Fig. 2A;Supplementary Table S5). We observed that some genes wereupregulated in all the lung tumor samples compared with thetumor-free tissues, including regulators of the EMT (SPP1,ADAM12) and regulators of invasion and metastasis (RUNX2,RHOA; Fig. 2A; Supplementary Table S5). Furthermore, certaingenes encoding core components of the TGFb signaling path-way, TGFBR2, SMAD7, and BAMBI (Fig. 2B), were significantlydownregulated in the lung cancer tissues. For BAMBI, the down-regulation of its mRNA corresponded well with the reducedexpression of the BAMBI protein in the lung cancer samples (Fig.1). The divergence between SMAD7 mRNA and protein levelsmight be due to differences in posttranslational regulation intumor and healthy tissue. Possibly, because SMAD7 mRNA istargeted by multiple miRNAs (26, 27), protein translation ofSMAD7 mRNA could be inefficient in healthy tissue. Alterna-tively, proteasomal degradation of SMAD7 mediated e.g. by theE3 ubiquitin ligase Cbl (28) might be impaired in lung cancercells, resulting in elevated SMAD7 protein levels despite reducedmRNA expression. No significant differences between the tumor

and tumor-free samples were observed for the intracellularsignaling mediators SMAD2, SMAD3, SMAD4, or SNON (Fig.2B). The results of the microarray analysis were validated withqRT-PCR (Fig. 2C; Supplementary Fig. S2). Overall, we observedsignificant alterations in the transcription of TGFb pathwaymembers and mediators in lung cancer. In line with the absenceof the BAMBI protein from lung cancer, we identified distinctdifferences in the BAMBI mRNA expression patterns betweentumor-free and tumor lung tissues, indicating a potential role ofBAMBI in lung cancer progression.

Methylation patterns of TGFb pathway members are distinctbetween lung tumor and tumor-free tissues

Changes in epigenetic modifications are observed during car-cinogenesis (29) and have been associated with altered geneexpression in tumors. To determine the methylation patterns ofTGFb pathway members and mediators in lung cancer, we exam-ined lung cancer and matched tumor-free tissue samples usingIllumina's BeadChip technology. The list of TGFb core pathwaymembers that were used for the transcriptome analysis (Sup-plementary Table S3) was again utilized to select CpG loci locatedin the genes of TGFb pathway members. The selected loci wereincluded in unsupervised hierarchical clustering (Fig. 3A) andPCA of DNA methylation values (Fig. 3B). In both analyses, thetumor and tumor-free samples displayed strikingly differentmethylation patterns. Notably, in 21 of the 26 genes with differ-ences of �10% in the median DNA methylation, changes in theDNA methylation between the tumor and tumor-free samplesinversely correlated with gene expression (Fig. 3C). For example,the reduced expression of BAMBI mRNA corresponded toincreasedmethylation. Basedon the reduced expressionof BAMBImRNA and protein in the lung cancer samples, we examinedepigenetic modifications of the CpG loci located in the BAMBIgene region. Analysis of the 11 loci covered by the HumanMethy-lation450k BeadChips revealed differential methylation (Fig. 3Dand E) between the cancer samples and matched tumor-free lungtissues. Bisulfite pyrosequencing and HumanMethylation450KBeadChip analysis showed similar methylation patterns of locilocated in the BAMBI gene region with a Pearson correlationcoefficient of r ¼ 0.958 (Supplementary Fig. S6). Based onthese results, we suggest that major changes in the DNAmethylome occur in lung cancer in the genes encodingTGFb pathway members and mediators. We hypothesized thatthe epigenetic silencing of thenegative regulator of TGFb signalingBAMBI could alter the signaling cascade to a more activated stateand hence influence downstream processes such as TGFb-depen-dent EMT.

Differential expression of EMT markers in human lungcancer and tumor-free lung tissues

Therefore, we re-examined our microarray analysis of pairedlung cancer and tumor-free tissue samples for the differential

(Continued.) Differences in the DNA methylation values between the tumor and tumor-free samples were plotted against the difference in expression(log value) as determined by the array analysis for each gene. For 21 genes, an inverse correlation between DNA methylation and gene expression(quadrants II and IV) was found. By contrast, DNA methylation directly correlated with gene expression in 5 genes (quadrants I and III). D and E, hierarchicalcluster analysis and PCA of 11 CpG loci present on the 450k BeadChip and located in the BAMBI gene demonstrated heterogeneous DNA methylationat the BAMBI locus. In the bar on top of the heatmap, red indicates lung adenocarcinoma, blue indicates lung squamous cell carcinoma, and blackindicates tumor-free lung cancer samples. In the heatmap, blue indicates low DNA methylation values, black indicates intermediate, and yellow indicates highDNA methylation values. The data were normalized for visualization (mean ¼ 0).






expression of EMT markers. With a text-mining search, wegenerated a list of genes involved in EMT that includedgenes regulated by TGFb as well as TGFb-independent EMTgenes (Supplementary Table S4; Fig. 4A). Notably, a hetero-geneous expression pattern of the mRNAs for EMT mediatorswas observed (Fig. 4A). CDH1 (E-cadherin), MMP9, SPP1(Osteopontin), HMGA1, TWIST1, and SOX4, among others(Supplementary Table S5), were significantly upregulated inlung tumor tissues compared with matched control samples.By contrast, classical EMT-inducing transcription factors, suchas ZEB1, ZEB2, and SNAI1, and structural components asso-ciated with the mesenchymal phenotype, such as alpha-smooth muscle actin (ACTA2) and Vimentin (VIM), exhibitedreduced levels of mRNA expression in lung cancer samples(Fig. 4). Validation experiments were performed via qRT-PCRassays (Fig. 4C) and, as revealed in Supplementary Fig. S2,confirmed the results of the transcriptome analysis. The stud-ies further showed that the differential expression of CDH1,SPP1, VIM, and ZEB1 was comparable in adeno- and squa-mous cell carcinomas, whereas for CDH2 and SNAI2, down-regulation was primarily observed in adenocarcinomas(Fig. 4C).

To analyze the EMT-related alterations at the protein level,we examined the presence of E-cadherin (CDH1), N-cadherin

(CDH2), and Osteopontin (SPP1) in the tumor sample cohort(n ¼ 130) via IHC (Fig. 5A; Supplementary Fig. S3 for repre-sentative staining). Again, the tumor samples displayed amixed phenotype with a high frequency of E-cadherin(90.4%), N-cadherin (93.8%), and Osteopontin (76.9%) pro-tein expression. The AECII from the tumor-free samples dis-played a constant expression level of E-cadherin (90%); how-ever, N-cadherin was present in only 5% of the samples, and noOsteopontin expression was observed. Overall, significantdifferences were observed between the lung cancer samplesand AECII regarding the expression frequencies for N-cadherinand Osteopontin (Supplementary Table S1). Furthermore,within the same individual samples (n ¼ 121–126), double-positive staining for pSMAD3/N-cadherin or E-cadherin/N-cadherin was assessed using immunofluorescence (Fig. 5B)and quantified (Fig. 5C). We observed that the majority ofpSMAD3-positive tumors were also positive for N-cadherin(79.4%). In addition, N-cadherin was frequently observed atthe invading front of a tumor (in 76.6% of the analyzedcases; see arrows in Fig. 5B) coinciding with a low E-cadherinsignal. Notably, double positivity for both E-cadherin andN-cadherin was a frequent feature of lung cancer (91.7%),which is consistent with the RNA data and in the single-stainedIHC sample. In summary, the observed alterations in the

B CDH1

mR

NA

Exp

ress

ion

(a.u

.)

20

30

40

10

0Tumor-free

lungTumor

1

2

0

3

CDH2n.s.

Tumor-free lung

Tumor

SPP1

Tumor-free lung

Tumor

4

8

0

12

10

20

0

30

SNAI2

Tumor-free lung

Tumor

n.s.

20

40

0

60

VIM

Tumor-free lung

Tumor

20

100

0

60

SNAI1

Tumor-free lung

Tumor

ZEB1

20

60

0

140

Tumor-free lung

Tumor

CDH1

mR

NA

Exp

ress

ion

tum

or (a

.u.)

0.001

0.01

0.1

1

10

0.001 0.01 0.1 1 10

CDH2

0.01

0.1

10

100

1

0.01 0.1 1 10 100 1,000

SNAI2

0.01

0.1

10

100

1

0.01 0.1 1 10 100 1,000

SPP1

0.01

0.1

10

100

1

0.01 0.1 1 10 100


VIM

0.001

0.01

0.1

1

10

0.001 0.01 0.1 1 10

SNAI1

0.001

0.01

0.1

1

10

0.001 0.01 0.1 1 10

ZEB1

0.1

10

100

1

0.1 1 10 100

FN1

0.001 0.01 0.1 1 10 0.001

0.01

0.1

1

10


mR

NA

Exp

ress

ion

tum

or (a

.u.)

C

A

SQC AC Tumor-free lung

1,000 1,000

SQCAC

4

5

100

SOX2SPP1CLDN1SIX1MMP9TWIST1PLAULOXL2SDC1CDH1GRHL2TIMP1SOX4CLDN7CTNNB1DLG1PRKCIRUNX2LEF1TCF3RUNX2FOXH1N/AHMGA2DSPHMGA1TIMP3CTGFPLGNOTCH1KLF4S100A4TWIST2MMP2ZEB2ACTA2TCF4TIMP2SPARCZEB1CALD1AHNAKVEGFCVIMZYXSMAI1ITGA5MITFPRKCZMSNTJP1CREBBPCEBPBVEGFBRHOA

Figure 4.Differential mRNA expression of EMT markers in human lung cancer and tumor-free lung tissues. A, EMT genes (Supplementary Table S4) significantlydifferentially regulated between macrodissected lung tumors and matched tumor-free lung tissues (n ¼ 19) were identified from microarray analysisby paired t testing with a Benjamini–Hochberg multiple testing correction (P � 0.05). A heatmap of hierarchical clustering was computed with asquared Euclidian distance matrix and Ward's linkage rule. The relative gene expression level is displayed for individual cases according to thehistologic tumor type and is abbreviated as follows: SQC, squamous cell carcinoma; AC, adenocarcinoma. B, quantile-normalized relative expressionlevels for single genes in the EMT data set are compared between tumor-free lung tissues and tumor samples using a two-sided paired t test with�� , P � 0.01 and �� , P � 0.001 regarded as significant; n.s., not significant. C, validation of the microarray data for single genes via qRT-PCR assaysof paraffin-embedded material depicted as relative expression levels. Gene expression in tumor-free lung versus tumor samples from 18 patients.Upregulated genes are above and downregulated genes are below the black line, indicating a ratio of 1 (equal expression rates in tumor and tumor-freelung). Blue dots, squamous cell carcinoma; red dots, adenocarcinoma.

Marwitz et al.





expression of EMT markers in lung cancer included majorchanges in the expression of TGFb target genes.

Abundance and activation of TGFb pathway componentsin NSCLC cell lines

A central alteration of TGFb signaling pathway observed in lungcancer was the loss of the negative regulator BAMBI. To dissect theconsequences of the absence of BAMBI and to assess the contri-bution to lung cancer progression, it was necessary to establish asuitable cellular model system. Therefore, we examined theexpression of BAMBI, TGFb, and several downstream pathwaycomponents in the NSCLC cell lines A549 and H1975 withimmunohistochemical methods (Fig. 6A and B). Similar to thelung cancer tissue samples, the cell lines showed positive signalsfor TGFb, SMAD2, SMAD4, SNON, pSMAD2, and pSMAD3 (Fig.6A and B).However, in contrast to the lung cancer tissues, SMAD7was weakly expressed (Score 1) and was only observed in H1975cells. As in the lung cancer tissues, none of the cell lines expressedthe BAMBI protein (Fig. 6A and B). This was confirmed bythe observation that the amount of BAMBI mRNA was signi-

ficantly lower in the NSCLC cell lines compared with the levelsdetected in healthy AECII (Fig. 6C). In addition, the presence ofTGFb in the supernatants of the NSCLC cell lines was analyzedwith aMultiplex bead-based assay (Supplementary Fig. S4A). Thesecretion of inactive TGFbwas detected in the supernatants of theNSCLC cell lines independent of the cultivation conditions.However, in both NSCLC cell lines, the level of active TGFb waslow. The induction of TGFb signaling in the NSCLC cell lines wasanalyzed by combining enrichment via IP with subsequentimmunoblotting (Supplementary Fig. S4B). Compared with theH1975 cells, higher levels of SMAD2 and SMAD3 proteins wereobserved in the A549 cells. In both cell lines, basal levels ofSMAD2 and SMAD3 phosphorylation were detectable upon GFdepletion or cultivation in regular growth medium. Yet anincrease in the phosphorylation of both SMADs was observedupon TGFb stimulation (Supplementary Fig. S4B). Taken togeth-er, the TGFb signaling in both NSCLC cell lines closely resemblesthe behavior in the lung cancer samples, making the cell linessuitable model systems for gaining insights into underlyingmechanisms.

A

Freq

uenc

y (%

)

0

321

0

20

40

60

93 93.1 66.1

B

0

20

40

60

80

100

Freq

uenc

y (%

)

C

80

Frequency of EMT markers Adenocarcinoma

89 94.5 85.9

Freq

uenc

y (%

)

0

20

40

60

80

Frequency of EMT markers Squamous cell carcinoma

E-cadherin(CDH1)

N-cadherin(CDH2)

Osteopontin(SPP1)

Freq

uenc

y (%

)

0

20

40

60

80

100

Frequency of EMT markers AECII healthy cells

90 5 0

E-cadherin+N-cadherin+

E-cadherin-N-cadherin+

E-cadherin+N-cadherin-

E-cadherin-N-cadherin-

Frequency of E-cadherin/N-cadherin in lung cancer

0

20

40

60

80

100

Freq

uenc

y (%

)

pSMAD3+ N-cadherin+

pSMAD3+ N-cadherin-

pSMAD3- N-cadherin+

pSMAD3- N-cadherin-

Frequency of pSMAD3/N-cadherin in lung cancer

E-cadherin(CDH1)

N-cadherin(CDH2)

Osteopontin(SPP1)

E-cadherin(CDH1)

N-cadherin(CDH2)

Osteopontin(SPP1)

Figure 5.Frequency of EMT markers on protein level differs in human lung cancer and tumor-free lung tissues. A, immunohistochemical scoring of paraffin-embeddedtumor samples (n ¼ 130) and AECII from tumor-free lungs with antibodies recognizing E-cadherin (CDH1), N-cadherin (CDH2), and osteopontin(SPP1) in squamous cell carcinomas, adenocarcinomas, and healthy AECII cells. Bar charts, numbers above the bars display the total positive casesobserved overall in percentage. B, representative images of double immunostaining with either pSMAD3 (green) and N-cadherin (red; top row) orwith E-cadherin (green) and N-cadherin (red; bottom row) in the same paraffin-embedded tumor (n for each marker ¼ 121–126). Arrows, CDH2 expressionat the tumor cell invasion front; scale bar, 50 mm. C, frequency of different combinations of EMT markers as determined by double immunofluorescencestaining (n ¼ 121–127) displayed as percentages.






A

BAMBI

TGFβ

SMAD2

pSMAD2

SMAD3

pSMAD3

SMAD4

SMAD7SNON

H1975

Sco

re

B

0.0

0.5

1.5

2.0

1.0

2.5

3.5

3.0A549

TGFβ pSMAD3pSMAD2SMAD2 SMAD7SMAD4SMAD3 SNONBAMBI

A549

H1975

G

BAMBI siRNA #1Control siRNA BAMBI siRNA #2

H

Sig

nal (

a.u.

)

1 432 50

0

1

2

3

A549pSMAD2

0

1

2

3

4321 50Time (h)

Sig

nal (

a.u.

)

SK-MES1pSMAD2

5432100

1

2

Sig

nal (

a.u.

)

pSMAD3

0

1

2

3

4321 50Time (h)

Sig

nal (

a.u.

)

pSMAD3

BAMBI-GFPGFP

CDH1

0 10 20 30 40 50

CDH2

0 10 20 30 40 50

SNAI2

0 10 20 30 40 50

H1975

1.2

0.8

1.4

0.6

1.2

1.8

2.4

(Fol

d ch

ange

rela

tive

to G

FP-c

ontro

l, t =

0 h

)m

RN

A E

xpre

ssio

n (a

.u.)

Time (h)

Time (h)

0.8

1.0

1.2

1.5

0.5

2.5

0 10 20 30 40 50 0 10 20 30 40 50

0.5

1.5

2.5

A549

0 10 20 30 40 50

0.8

1.2

1.6

A549SNAI1

0

1

2

3

60 12 18 24

SNAI1

SK-MES1

0

1

2

3

PAI1

0

5

10

60 12 18 24

PAI1

0

5

10

15

20

60 12 18 2460 12 18 24

ZEB1

0

5

1

2

3

4

60 12 18 24Time (h)

ZEB1

0

1

2

3

4

60 12 18 24Time (h)

(Fol

d ch

ange

rela

tive

to G

FP-c

ontro

l, t =

0 h

)m

RN

A E

xpre

ssio

n (a

.u.)

TGFβ (ng/mL)

pSMAD2

pSMAD2

0

5

10

0

4

8

12

0.1 0.10.5 0.5

pSMAD3

Sig

nal (

a.u.

)

TGFβ (ng/mL)

H1975

A549

Sig

nal (

a.u.

)

0

5

10

15

0.1 1.00.5

P = 0.0156

1.0

P = 0.0148

1.0

P = 0.9791

0

1

2

3

4

0.1 1.00.5

pSMAD3

P = 0.0104

Sig

nal (

a.u.

)S

igna

l (a.

u.)

BAMBI-GFPGFP

BAMBI siRNA #1Control siRNA

BAMBI siRNA #2

FE

C

mR

NA

Exp

ress

ion

(a.u

.)

AECIIA54

9

H1975

SK-MES1

0.0

0.4

1.2

0.8

BAMBI D

mR

NA

Fol

d ch

ange

rela

tive

to A

EC

II

0

20

40

60

80

100

120

AECII

H1975

-

GFPH19

75-

BAMBI-GFP

A549-

GFP A549-

BAMBI-GFP

BAMBI

CDH1CDH2 SNAI2

n.s.

Figure 6.Altering BAMBI expression affects TGFb signaling in NSCLC cell lines. A, representative immunocytochemical stainings of the indicated proteins inHOPE-fixed paraffin-embedded NSCLC lung cancer cell lines H1975 and A549. The staining was performed as described for the tissue sections.Magnification, �40; scale bar, 50 mm. B, scoring of the expression of the indicated proteins in cell lines A549 and H1975. Bar charts indicate noexpression (0), focal expression (1), frequent expression (2), or expression as the dominant pattern (3). C, relative expression of BAMBI mRNA in theindicated lung cancer cell lines in comparison with healthy AECII. The mean � SD of qRT-PCR determinations from three AECII isolations and fromthree biologic triplicates of the lung cancer cell lines are shown. Statistical analysis was performed using one-way ANOVA test, with �� , P � 0.001regarded as significant; n.s., not significant. D, relative expression of BAMBI mRNA in DOX-treated NSCLC cell lines in comparison with healthy AECII.Data represent the mean from four biologic replicates � SD. Statistical analysis was performed using one-way ANOVA test, with �� , P � 0.001regarded as significant. E, quantification of TGFb dose-dependent phosphorylation of SMAD2 and SMAD3 in H1975-GFP and H1975-BAMBI-GFP aswell as in A549-GFP and A549-BAMBI-GFP cells. (Continued on the following page.)

Marwitz et al.





Altering BAMBI expression affects TGFb signaling in NSCLCcell lines

A striking feature of the examined lung cancer samples was theabsence of the BAMBI protein as well as its reduced mRNAexpression. To obtain insights into mechanisms whereby the lossof BAMBI might contribute to tumor progression, we reconsti-tuted BAMBI expression in the NSCLC cell lines utilizing aretroviral TET-On system that facilitates the inducible expressionof BAMBI-GFP or GFP alone. We achieved 20 and 100 timeshigher BAMBI expression level in A549 and H1975 cells, respec-tively, compared with AECII cells (Fig. 6D). As shown in Fig. 6E,the reconstitution of BAMBI in the H1975 cells resulted in aconsiderable decrease in SMAD2 and SMAD3 phosphorylation atall the TGFb concentrations used for stimulation. In the A549cells, a decrease in SMAD2 and SMAD3 phosphorylation wasobserved at low TGFb concentrations (Fig. 6E; Supplementary Fig.S5A and S5B). To determine whether the reconstitution of BAMBIalters the transcription of genes, we analyzed the TGFb-inducedmRNA expression of selected EMT markers, such as CDH1 (E-cadherin), CDH2 (N-cadherin), and SNAIL2, for 48 hours andobserved that in both A549 and H1975 cells, the mRNA expres-sion levels of each of the examined genes were markedly reducedin the BAMBI-reconstituted cells (Fig. 6F; Supplementary Fig.S5C). These observations correlated with a major reduction atthe protein level detected after 48 hours of TGFb stimulation(Supplementary Fig. S5C). In addition, we performed siRNA-mediated BAMBI knockdown in the lung adenocarcinoma cellline A549 and the lung squamous cell carcinoma cell line SK-MES1 as they showed intermediate reduction of BAMBImRNA incomparison with AECII. We were able to achieve 80% to 90%knockdown efficiency in both cell lines (Supplementary Figs. S7Band S8B). BAMBI knockdown resulted in higher SMAD2 andSMAD3 phosphorylation and enhanced target gene expressionupon TGFb stimulation in both A549 and SK-MES1 cells (Fig. 6Gand H; Supplementary Figs. S7 and S8).

Reconstitution of BAMBI in lung cancer cells leads to reducedinvasion and tumor growth

To determine the effects of BAMBI reconstitution on cellularresponses, we analyzed the TGFb-inducedmigration and invasionof H1975-BAMBI-GFP cells and of H1975-GFP cells as a control.In a 2Dmigration assay, the TGFb treatment resulted in increasedspeed (from 2.5 to 3.3 mm/h) and persistence (from 0.275 to0.425) of cellular movement in control cells, whereas in theBAMBI-GFP–reconstituted cells, the increase was almost entirelyabrogated (Fig. 7A). Furthermore, TGFb treatment resulted in a

3-fold increase of invaded H1975-GFP cells in a 3D collageninvasion assay, whereas in BAMBI-reconstituted cells, suchincrease was not observed, suggesting that the silencing of BAMBIincreased the potential of cells to invade in response to TGFbstimulation (Fig. 7B). Further, we examined the impact of BAMBIloss-of-function on cancer cell invasion. We performed siRNA-mediated knockdown in the squamous lung cell carcinomacell line SK-MES1 that harbors higher basal expression of BAMBImRNA than the adenocarcinoma cell line H1975. The BAMBImRNA knockdown in SK-MES1 cells resulted in increased inva-sion of unstimulated cells and importantly enhanced TGFb-induced invasion for at least one of the siRNA used (Supplemen-tary Fig. S8D).

To investigate the influence of enhanced BAMBI expressionon the metastatic potential and tumor-forming capacity invivo, both A549-GFP control and A549-BAMBI-GFP cells wereinjected into the tail veins of nude mice. The mice were kept inthe presence of DOX to ensure the expression of GFP orBAMBI-GFP, and the TGFb ligand was administrated to facil-itate cancer cell invasion. Per group, 12 animals were injectedand kept for 10 weeks, unless dropout criteria were observedand mice were sacrificed earlier. In the GFP group, 2 mice thatwere sacrificed prematurely and 2 mice that survived until theplanned end-point of the experiment developed histologicallyconfirmed tumors within the lungs. In the BAMBI-GFP cohort,lung lesions were observed in 1 prematurely sacrificed mouseand in 3 mice that survived until the end-point. Overall, 4mice per group developed lung cancer. Macroscopic investiga-tions of the dissected lungs showed that A549-BAMBI-GFPmice developed fewer lung lesions compared with A549-GFPmice (Fig. 7C).

Histologic diagnosis revealed adenocarcinomas in the lungsof both groups (Fig. 7D), and qRT-PCR analysis confirmedthe presence of human A549 cells expressing either GFP orBAMBI-GFP (Supplementary Fig. S10). To determine the influ-ence of DOX-induced GFP or BAMBI-GFP expression on themetastatic potential of A549 cells, the total amount of lungtumor nodules per animal was quantified. Mice injected withGFP-expressing A549 cells showed a median of 68.5 (17 mini-mum, 174 maximum) number of tumor nodules per animal,whereas in mice injected with A549-BAMBI-GFP cells, only 2.5(1 minimum, 25 maximum) nodules per animal wereobserved (Fig. 7E). Furthermore, the analysis of the tumorareas per animal showed a median tumor area of 8.4 � 107

mm2 for A549-GFP–injected mice, whereas the median tumorarea of 2.7 � 106 mm2 was much smaller for mice harboring

(Continued.) For each cell line, the SMAD2/3 phosphorylation was determined by quantitative immunoblotting. The data shown are the mean ofthree biologic replicates (provided in Supplementary Fig. S5). The shaded area indicates the SEM. Statistical analysis was performed using two-wayANOVA with repeated measurements, and P � 0.05 was considered significant. F, expression of the indicated EMT genes in H1975 and A549 cell lines stablytransduced with GFP or BAMBI-GFP was assessed by qRT-PCR. The GFP and BAMBI-GFP expression was induced with DOX for 48 hours, and cellswere stimulated with TGFb (0.1 ng/mL) for the indicated time points. mRNA levels were normalized using the geometric mean of the b-glucuronidase(GUSB) and esterase D (ESD) housekeeping genes. The experiments were performed in six biologic replicates, and the data represent mean values � SD. G,quantification of the dynamics of TGFb induced SMAD2/3 phosphorylation upon BAMBI knockdown in A549 and SK-MES1 cells. A549 and SK-MES1cells were transfected with BAMBI-specific siRNAs or control siRNA for 36 hours (knockdown efficiency is shown in Supplementary Figs. S7 and S8). Afterstimulation with 1 ng/mL TGFb for the indicated time, SMAD2/3 phosphorylation was determined by quantitative immunoblotting. For each cell line,the data shown are the mean of three independent experiments (replicates provided in Supplementary Figs. S7 and S8), and the shaded areaindicates the SEM. H, expression of the indicated EMT genes in A549 and SK-MES1 cell lines transfected with BAMBI-specific siRNAs or control siRNA for36 hours, stimulated with 1 ng/mL TGFb, was assessed by qRT-PCR. RNA levels were normalized using the geometric mean of the GAPDH and G6PDhousekeeping genes. Data shown represent mean � SD from three independent experiments.






A549-BAMBI-GFP. Statistical analysis employing the Mann–Whitney test showed for both analysis significant differences (Pvalues of 0.0295 and 0.0286 for number of nodules and tumorarea, respectively) between the mice harboring A549 cells thatexpress the GFP control or BAMBI-GFP (Fig. 7E). This demon-strates that BAMBI reconstitution in A549 cells impairs notonly TGFb-induced cancer cell invasion in vitro but also in vivo.In addition, immunohistochemical analysis of the human lungtumors within the mouse lungs showed reduced levels ofSMAD3 phosphorylation in A549-BAMBI-GFP cells comparedwith A549-GFP control cells (Fig. 7D), indicating that recon-stitution of BAMBI expression also suppressed the activation ofTGFb signaling in vivo.

DiscussionIn general, the TGFb signaling cascade is viewed as a double-

edged sword in the context of carcinogenesis. At premalignantstages, this pathway favors anti-proliferative and proapoptotic

responses. However in advanced tumors, TGFb-mediatedimmunosuppression and enhancement of invasion promotemetastatic spread (7). In lung cancer, high TGFb serum levelscorrelate with a poor prognosis, lymph node metastasis, andtumor progression (30). Furthermore, the expression of TGFbwas an independent risk factor for the occurrence of pulmonarymetastasis in NSCLC (31).

We presented a comprehensive in-depth analysis of the statusof the TGFb signaling pathway in lung cancer tissues anddiscovered a distinct role of the TGFb pseudoreceptor BAMBI.Strikingly, in almost all the lung cancer tissue samples exam-ined, the negative regulator of TGFb signaling BAMBI wasabsent. In line with this observation, abundant activation ofthe TGFb signaling pathway, indicated by the phosphorylationof both R-SMADs, was observed. On the contrary, in coloncarcinomas, mutations in TGFBR2 (32) and SMAD4 (33) arefrequently observed resulting in the deactivation of signalingpathway. In NSCLC, mutations in the receptor (34) or in theintracellular signal mediators SMAD2 (35) and SMAD4 (36)

Persistence

TGFβ

Per

sist

ence

0.2

0.4

0.6

0.8

BA

Spe

ed (μ

m/h

)

1

2

3

4

5

6

7

Speed

H1975

0

10

20

30

40

50

Inva

ded

cells

(%)

TGFβ

BAMBI-GFPGFP

H1975

H1975-GFP

H1975-BAMBI-GFP

Nuclei of noninvading cellsNuclei of invading cells

DC GFP BAMBI-GFP GFP BAMBI-GFP

H&E

pSMAD3

TGFβ

BAMBI-GFPGFP

0.0

Invasion

Lung

tum

or a

rea

per a

nim

al (μ

m2 )

105

106

107

108

109

GFP BAMBI-GFP

GFP BAMBI-GFP

Lung

tum

or n

odul

es p

er a

nim

al

0.1

1

10

100

1,000

GFP BAMBI-GFPE

Figure 7.Reconstitution of BAMBI expression in lung cancer cells reduces TGFb-mediated invasion capacity in vitro and in vivo. A, persistence and movement speed ofH1975-GFP and H1975-BAMBI-GFP cells, stimulated with 0.25 ng/mL TGFb and imaged for 22 hours. The box is bound by the 25th and 75th quantiles.The whiskers extend to three-halves of the interquartile range. The lines indicate the median, and dots illustrate the 5th and 95th quantiles. Each boxcontains at least 350 cells from two independent experiments, nine biologic replicates in total. Statistical analysis was performed using the Mann–Whitneyrank-sum test; �� , P � 0.001. B, quantification of the number of H1975-GFP and H1975-BAMBI-GFP cells that invaded into the collagen gel. The dataof one representative experiment is shown. Every dot corresponds to one biologic replicate; black line is the median. A second independent replicate isshown in Supplementary Fig. S5F. Statistical analysis was performed using one-way ANOVA, with �� , P � 0.001 regarded as significant. C, representativeimages of mouse lungs 9 weeks after tail vein injection of A549-GFP or A549-BAMBI-GFP cells. Lung macrometastases are indicated with whitearrows. D, histologic analysis of lung tumor burden in mice injected with A549-GFP or A549-BAMBI-GFP. Lung tumor area within the H&E-stainedtissue section of animal lungs was quantified using Infinity Analyze software. Exemplary images of tissue sections from A549-GFP- and A549-BAMBI-GFP–injected animals are shown, and lung tumors are manually marked by black borders (top plots). Activity of the TGFb pathway was determined byIHC detection of pSMAD3 (bottom plots). All images are taken at �10 magnification. Scale bar, 200 mm. E, quantification of human lung tumornodules and lung tumor areas within animal lungs after tail vein injection of A549-GFP and A549-BAMBI-GFP cells. Tumor nodules were counted onH&E-stained sections from each lung of every animal that developed histology-confirmed lung tumors (n ¼ 4 each). The total area of lung tumor peranimal as determined by image analysis on H&E-stained sections (n ¼ 4 each). All data are displayed as box and whiskers plot showing the medianand minimum/maximum. For statistical evaluation of nonparametric data, the Mann–Whitney test (one-tailed) was applied and P � 0.05 (�) was regardedas significant.

Marwitz et al.





are rare. Therefore, NSCLC seems to retain a functional andactive signaling cascade as discussed for breast cancer (7),suggesting a similar role of TGFb in the carcinogenesis of bothtypes of carcinoma. Currently, the role of negative regulatorBAMBI in various cancers is controversially discussed. IncreasedBAMBI expression has been reported for colorectal cancertissues (11, 13) and ovarian cancer cells (12). In contrast,BAMBI was observed to be epigenetically silenced in high-gradebladder cancer (14) and absent in metastasizing melanomacells (37) and breast cancer tissues (15). Our results showedthat in lung cancer tissues, in contrast to matched tumor-freecontrols, both BAMBI mRNA and BAMBI protein were absent.The silencing of BAMBI apparently contributes to malignantdevelopment, tumor progression, and invasion (14, 37).

We hypothesize that the downregulation of BAMBI sensi-tizes lung cancer cells to TGFb and thereby promotes EMT-dependent malignant processes. Central role of TGFb signalingin EMT regulation has been already observed in breast cancer(38) and esophageal cancer (39). For lung cancer, it has beenshown that long-term TGFb treatment induces EMT transitionin A549 and LC31 lung cancer cell lines (40). Furthermore,Prudkin and colleagues observed that the EMT phenotype iscommonly expressed in primary squamous cell carcinoma andadenocarcinoma of the lung (41). Our studies revealed thatBAMBI-negative lung cancer tissues display a mixed EMTphenotype involving the retained expression of epithelialmarker proteins (E-cadherin) along with mesenchymal markerproteins (N-cadherin) and EMT-associated markers such asHMGA1 (42) and Osteopontin (SPP1; ref. 43). This is in linewith the observation that in some metastatic cancers, partialEMT is observed (44). It has been proposed that EMT is ahighly dynamic and reversible multistep process. The mixedexpression of both epithelial and mesenchymal markers withinhuman lung tumor tissues suggests that, in lung cancer, EMT isonly required for a limited time-frame, for example to seedmetastases, or that EMT occurs primarily at the invading frontof tumors and hence is only active at the time of invasivebehavior. This possibility is supported by the findings from theIHC and immunofluorescence data wherein SPP1 and N-cad-herin are observed at the invading front of tumor cells. TheEMT program has been linked to chemotherapy resistance inovarian carcinomas (45), and it has been reported for the A549lung cell line that chemoresistance in this cell line was asso-ciated with TGFb-induced EMT (46). We showed that therestoration of BAMBI expression in NSCLC cell lines reducedTGFb-induced phosphorylation of the R-SMADs. As a conse-quence, TGFb-mediated expression of EMT markers wasmarkedly attenuated, and the induction of cell migration andinvasion was significantly reduced. Furthermore, the in vivoexperiments strongly suggest a pivotal role of BAMBI withregard to tumor growth and metastasizing capability. Theseobservations identify BAMBI as a negative regulator of TGFb-induced EMT responses in NSCLC and suggest a role for BAMBIas a tumor suppressor in lung cancer.

Our epigenetic analysis reveals severe alterations in theDNA methylation patterns of TGFb signaling pathway genesin lung cancer. Accordingly, aberrant DNA methylation ofTGFb signaling pathway genes was observed in gastric adeno-carcinoma (47), showing that the methylation frequency ofTGFBR1 and TGFBR2 in the tissues of high-grade dysplasia andGCA was significantly elevated and associated with mRNA and

protein expression of the two genes. Further, more frequentTGFBR1 and TGFBR2 promoter hypermethylation was detectedin stage III and stage IV esophageal squamous cell carcinomatissue, indicating that promoter methylation of TGFBR1 andTGFBR2 may exist in the early stage of ESCC and play impor-tant roles in TGFBR1 and TGFBR2 gene silencing (48). In ratprostate cancer cell lines, treatment with the demethylatingagent 5-aza-20-deoxycytidine resulted in upregulation of sev-eral genes including TGFBR2, and epigenetic silencing ofTGFBR2 gene was observed in high-grade prostatic intraepithe-lial neoplasia (49). Cheng and colleagues showed that epige-netic silencing of SMAD8 expression by DNA hypermethyla-tion in breast and colon cancers directly correlated with loss ofSMAD8 expression (50). These findings not only underlinethe impact of alterations in the DNA methylation pattern ofTGFb pathway genes in numerous cancer types but also sup-port our conclusion that aberrant DNA methylation mightaffect the activity of the TGFb pathway in lung cancer.

The in-depth analysis of patient samples and mechanisticinsights from in vitro studies presented in this report revealmajor alterations in the TGFb pathway in lung cancer andsuggest that this pathway might be a promising therapeutictarget. To our knowledge, there are currently no completedclinical trials targeting the TGFb pathway in lung cancerpatients. For other cancer types, it has been shown that theradiotherapy-induced elevation of TGFb serum levels in amouse model of breast cancer favors metastasis but can beabrogated by TGFb-targeted antibodies (51). In addition, thetargeting of EGFR signaling by the EGFR antibody Cetuximabin a xenograft model of head and neck cancer resulted inemergence of resistant tumor cells that expressed relativelyhigher levels of TGFb and elevated levels of TGFb in the tumormicroenvironment enable tumor cells to evade antibody-dependent cell-mediated cytotoxicity and resist the antitumoractivity of Cetuximab in vivo. These results show that TGFb is akey molecular determinant of resistance of cancers to EGFR-targeted therapy (52).

A recent study identified MED12, a component of thetranscriptional MEDIATOR complex that is mutated in can-cers, as a determinant of response to ALK and EGFR inhibi-tors. Similar to BAMBI, MED12 negatively regulates TGFbreceptor through physical interaction, and its suppressionresults in activation of TGFb signaling, which is both neces-sary and sufficient for drug resistance. MED12 loss induces anEMT-like phenotype, which is associated with chemotherapyresistance in colon cancer patients and to gefitinib in lungcancer (53).

We suggest that in lung cancer epigenetic silencing ofBAMBI results in elevated TGFb signaling favoring invasion,migration, and EMT. Further, we speculate that an elevatedactivation of TGFb signaling pathway in the absence ofthe negative regulator BAMBI may enhance the immuno-suppressive effects (7) of TGFb in tumor microenviron-ment and promote cancer cell invasion. Therefore, we pro-pose a central role of altered TGFb signaling in lung cancerprogression.

Disclosure of Potential Conflicts of InterestsM. Reck received honoraria from the Speakers Bureau of AstraZeneca,

BMS, Boehringer Ingelheim, Hoffmann-La Roche, Lilly, MSD, and Pfizer. Heis also a consultant/advisory board member for AstraZeneca, BMS,






Boehringer Ingelheim, Hoffmann-La Roche, Lilly, MSD, Novartis, and Pfizer.No potential conflicts of interest were disclosed by the other authors.

Authors' ContributionsConception and design: S. Marwitz, S. Depner, K. M€uller-Decker, M. Reck,R. Siebert, T. Goldmann, U. Klingm€ullerDevelopment of methodology: S. Marwitz, S. Depner, D. Dvornikov,U. Klingm€ullerAcquisition of data (provided animals, acquired and managed patients,provided facilities, etc.): S. Marwitz, S. Depner, D. Dvornikov, R. Merkle,M. Szczygieł, K. M€uller-Decker, P. Lucarelli, M. W€asch, H. Mairb€aurl,K.F. Rabe, M. Reck, S. Scheufele, M. Kr€oger, O. Ammerpohl, T. Goldmann,U. Klingm€ullerAnalysis and interpretation of data (e.g., statistical analysis, biostatistics,computational analysis): S. Marwitz, D. Dvornikov, R. Merkle, M. Szczygieł,P. Lucarelli, M. Reck, S. Scheufele, M. Kr€oger, O. Ammerpohl, R. Siebert,T. GoldmannWriting, review, and/or revision of the manuscript: S. Marwitz, S. Depner,D.Dvornikov, R.Merkle, K.M€uller-Decker, P. Lucarelli, H.Mairb€aurl, K.F. Rabe,C. Kugler, E. Vollmer, M. Reck, O. Ammerpohl, R. Siebert, T. Goldmann,U. Klingm€ullerAdministrative, technical, or material support (i.e., reporting or organizingdata, constructing databases): D. Dvornikov, M. W€asch, C. KuglerStudy supervision: E. Vollmer, R. Siebert, T. Goldmann, U. Klingm€uller

AcknowledgmentsThe authors thank Jasmin Tiebach, Maria Lammers, Kristin Wiczkowski,

Bettina Baron-L€uhr, Anna Schiller, and Lorena Valles for excellent technical

assistance. The support of Damir Krunic from the DKFZ Light Microscopyfacility and of Steffen Schmitt from the DKFZ FACS facility is gratefullyacknowledged. They also thank Bettina Oerhle and Gerald Burgstaller fromOliver Eickelberg's lab at CPC Munich for their help with establishing the 3Dcollagen invasion assay. Finally, they thankDKFZTumorModels core facility forsupport with in vivo mouse experiments.

Grant SupportThis study was funded by the German Center for Lung Research

[Deutsches Zentrum f€ur Lungenforschung (DZL)], the German Ministry ofEducation and Research (BMBF), and in part by the Klara und Werner KreitzStiftung (S. Marwitz). Funding from the BMBF within the CancerSys network"LungSys II" and Virtual Liver Network and funding by an NIH grant1R01DK090347-01 were also acknowledged. Magdalena Szczygieł was sup-ported by the Helmholtz International Graduate School for Cancer Researchat the German Cancer Research Center (DKFZ). Patient tissues were providedby the BioMaterialBank North, which is funded in part by the AirwayResearch Center North (ARCN), member of the German Center for LungResearch (DZL), and is member of popgen 2.0 network (P2N), which issupported by a grant from the German Ministry for Education and Research(01EY1103).

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.

ReceivedMay 20, 2015; revised February 18, 2016; acceptedMarch 11, 2016;published OnlineFirst May 17, 2016.

References1. Ferlay J, Shin HR, Bray F, Forman D, Mathers C, Parkin DM. Estimates of

worldwide burden of cancer in 2008: GLOBOCAN 2008. Int J Cancer2010;127:2893–917.

2. Morgensztern D, Ng SH, Gao F, Govindan R. Trends in stage distributionfor patients with non-small cell lung cancer: A national cancer databasesurvey. J Thorac Oncol 2010;5:29–33.

3. Siegel R, NaishadhamD, Jemal A. Cancer statistics, 2012. CA Cancer J Clin2012;62:10–29.

4. Pain M, Bermudez O, Lacoste P, Royer PJ, Botturi K, Tissot A, et al. Tissueremodelling in chronic bronchial diseases: From the epithelial to mesen-chymal phenotype. Eur Respir Rev 2014;23:118–30.

5. Hugo H, Ackland ML, Blick T, Lawrence MG, Clements JA, Williams ED,et al. Epithelial–mesenchymal and mesenchymal–epithelial transitions incarcinoma progression. J Cell Physiol 2007;213:374–83.

6. Siegel PM, Massague J. Cytostatic and apoptotic actions of TGF-beta inhomeostasis and cancer. Nat Rev Cancer 2003;3:807–21.

7. Massague J. TGFbeta in cancer. Cell 2008;134:215–30.8. Moustakas A, Heldin CH. The regulation of TGFbeta signal transduction.

Development 2009;136:3699–714.9. Onichtchouk D, Chen YG, Dosch R, Gawantka V, Delius H, Massague J,

et al. Silencingof TGF-beta signalling by the pseudoreceptor BAMBI.Nature1999;401:480–5.

10. DromannD, Rupp J, Rohmann K,Osbahr S, Ulmer AJ,Marwitz S, et al. TheTGF-beta-pseudoreceptor BAMBI is strongly expressed in COPD lungs andregulated by nontypeable Haemophilus influenzae. Respir Res 2010;11:67.

11. Sekiya T, Adachi S, Kohu K, Yamada T, Higuchi O, Furukawa Y, et al.Identificationof BMPand activinmembrane-bound inhibitor (BAMBI), aninhibitor of transforming growth factor-beta signaling, as a target of thebeta-catenin pathway in colorectal tumor cells. J Biol Chem 2004;279:6840–6.

12. Pils D, Wittinger M, Petz M, Gugerell A, Gregor W, Alfanz A, et al. BAMBI isoverexpressed in ovarian cancer and co-translocates with Smads into thenucleus upon TGF-beta treatment. Gynecol Oncol 2010;117:189–97.

13. Togo N, Ohwada S, Sakurai S, Toya H, Sakamoto I, Yamada T, et al.Prognostic significance of BMP and activin membrane-bound inhibitor incolorectal cancer. World J Gastroenterol 2008;14:4880–8.

14. Khin SS, Kitazawa R, Win N, Aye TT, Mori K, Kondo T, et al. BAMBI gene isepigenetically silenced in subset of high-grade bladder cancer. Int J Cancer2009;125:328–38.

15. LangDS,Marwitz S,HeilenkotterU, SchummW,BehrensO, SimonR, et al.Transforming growth factor-beta signaling leads to uPA/PAI-1 activationand metastasis: a study on human breast cancer tissues. Pathol Oncol Res2014;20:727–32.

16. Goldmann T, DromannD,Marzouki M, Schimmel U, Debel K, BranscheidD, et al. Tissue microarrays from HOPE-fixed specimens allow forenhanced high throughput molecular analyses in paraffin-embeddedmaterial. Pathol Res Pract 2005;201:599–602.

17. Marwitz S, Abdullah M, Vock C, Fine JS, Visvanathan S, Gaede KI, et al.HOPE-BAL: Improved molecular diagnostics by application of a noveltechnique for fixation and paraffin embedding. J Histochem Cytochem2011;59:601–14.

18. Marwitz S, Kolarova J, Reck M, Reinmuth N, Kugler C, Schadlich I, et al.The tissue is the issue: improved methylome analysis from paraffin-embedded tissues by application of the HOPE technique. Lab Invest2014;94:927–33.

19. Friemel C, AmmerpohlO, Gutwein J, Schmutzler AG, Caliebe A, KautzaM,et al. Array-based DNA methylation profiling in male infertility revealsallele-specific DNA methylation in PIWIL1 and PIWIL2. Fertil Steril2014;101:1097–103 e1.

20. Kinsella TM, Nolan GP. Episomal vectors rapidly and stably producehigh-titer recombinant retrovirus. Hum Gene Ther 1996;7:1405–13.

21. Pfeifer AC, KaschekD, Bachmann J, KlingmullerU, Timmer J.Model-basedextension of high-throughput to high-content data. BMC Syst Biol 2010;4:106.

22. von der Heyde S, Sonntag J, Kaschek D, Bender C, Bues J, Wachter A, et al.RPPanalyzer toolbox: An improved R package for analysis of reverse phaseprotein array data. Biotechniques 2014;57:125–35.

23. Burgstaller G, Oehrle B, Koch I, Lindner M, Eickelberg O. Multiplexprofiling of cellular invasion in 3D cell culture models. PLoS ONE2013;8:e63121.

24. LinC, SongH,HuangC, YaoE,GacayanR, Xu SM, et al. Alveolar type II cellspossess the capability of initiating lung tumor development. PLoS One2012;7:e53817.

Marwitz et al.





25. BorokZ,Crandall ED.More life for a "terminal" cell. Am JPhysiol LungCellMol Physiol 2009;297:L1042–4.

26. ButzH, Racz K,Hunyady L, Patocs A. Crosstalk between TGF-beta signalingand the microRNA machinery. Trends Pharmacol Sci 2012;33:382–93.

27. Gronroos E, Hellman U, Heldin CH, Ericsson J. Control of Smad7 stabilityby competition between acetylation and ubiquitination. Mol Cell 2002;10:483–93.

28. Gruber T, Hinterleitner R, Hermann-Kleiter N, Meisel M, Kleiter I, WangCM, et al. Cbl-bmediates TGFbeta sensitivity by downregulating inhibitorySMAD7 in primary T cells. J Mol Cell Biol 2013;5:358–68.

29. Shen H, Laird PW. Interplay between the cancer genome and epigenome.Cell 2013;153:38–55.

30. Hasegawa Y, Takanashi S, Kanehira Y, Tsushima T, Imai T, Okumura K.Transforming growth factor-beta1 level correlates with angiogenesis,tumor progression, and prognosis in patients with nonsmall cell lungcarcinoma. Cancer 2001;91:964–71.

31. Saji H, Nakamura H, Awut I, Kawasaki N, Hagiwara M, Ogata A, et al.Significance of expression of TGF-beta in pulmonary metastasis innon-small cell lung cancer tissues. Ann Thorac Cardiovasc Surg 2003;9:295–300.

32. Grady WM, Willis JE, Trobridge P, Romero-Gallo J, Munoz N, Olech-nowicz J, et al. Proliferation and Cdk4 expression in microsatelliteunstable colon cancers with TGFBR2 mutations. Int J Cancer 2006;118:600–8.

33. Jones S, Chen WD, Parmigiani G, Diehl F, Beerenwinkel N, Antal T, et al.Comparative lesion sequencing provides insights into tumor evolution.Proc Natl Acad Sci U S A 2008;105:4283–8.

34. Anumanthan G, Halder SK, Osada H, Takahashi T, Massion PP, CarboneDP, et al. Restoration of TGF-beta signalling reduces tumorigenicity inhuman lung cancer cells. Br J Cancer 2005;93:1157–67.

35. Uchida K, Nagatake M, Osada H, Yatabe Y, Kondo M, Mitsudomi T, et al.Somatic in vivo alterations of the JV18-1 gene at 18q21 in human lungcancers. Cancer Res 1996;56:5583–5.

36. Nagatake M, Takagi Y, Osada H, Uchida K, Mitsudomi T, Saji S, et al.Somatic in vivo alterations of the DPC4 gene at 18q21 in human lungcancers. Cancer Res 1996;56:2718–20.

37. DegenWG,WetermanMA, van Groningen JJ, Cornelissen IM, Lemmers JP,Agterbos MA, et al. Expression of nma, a novel gene, inversely correlateswith the metastatic potential of human melanoma cell lines and xeno-grafts. Int J Cancer 1996;65:460–5.

38. Taube JH, Herschkowitz JI, Komurov K, Zhou AY, Gupta S, Yang J, et al.Core epithelial-to-mesenchymal transition interactome gene-expressionsignature is associated with Claudin-low and metaplastic breast cancersubtypes. Proc Natl Acad Sci U S A 2010;107:15449–54.

39. Ohashi S, Natsuizaka M, Wong GS, Michaylira CZ, Grugan KD, Stairs DB,et al. Epidermal growth factor receptor and mutant p53 expand anesophageal cellular subpopulation capable of epithelial-to-mesenchymaltransition throughZEB transcription factors. Cancer Res 2010;70:4174–84.

40. Pirozzi G, Tirino V, Camerlingo R, Franco R, La Rocca A, Liguori E, et al.Epithelial to mesenchymal transition by TGFbeta-1 induction increasesstemness characteristics in primary non small cell lung cancer cell line.PLoS One 2011;6:e21548.

41. Prudkin L, Liu DD, Ozburn NC, SunM, Behrens C, Tang X, et al. Epithelial-to-mesenchymal transition in the development and progression of ade-nocarcinoma and squamous cell carcinoma of the lung. Mod Pathol2009;22:668–78.

42. Pegoraro S, RosG, Piazza S, Sommaggio R, Ciani Y, Rosato A, et al. HMGA1promotes metastatic processes in basal-like breast cancer regulating EMTand stemness. Oncotarget 2013;4:1293–308.

43. Alonso SR, Tracey L, Ortiz P, Perez-Gomez B, Palacios J, Pollan M, et al.A high-throughput study in melanoma identifies epithelial-mesenchy-mal transition as a major determinant of metastasis. Cancer Res 2007;67:3450–60.

44. Debnath J, Brugge JS. Modelling glandular epithelial cancers in three-dimensional cultures. Nat Rev Cancer 2005;5:675–88.

45. Latifi A, Abubaker K, Castrechini N, Ward AC, Liongue C, Dobill F, et al.Cisplatin treatment of primary and metastatic epithelial ovarian carcino-mas generates residual cells with mesenchymal stem cell-like profile. J CellBiochem 2011;112:2850–64.

46. Toge M, Yokoyama S, Kato S, Sakurai H, Senda K, Doki Y, et al. Criticalcontribution of MCL-1 in EMT-associated chemo-resistance in A549 non-small cell lung cancer. Int J Oncol 2015;46:1844–8.

47. Guo W, Dong Z, Guo Y, Kuang G, Yang Z, Shan B. Concordantrepression and aberrant methylation of transforming growth factor-beta signaling pathway genes occurs early in gastric cardia adeno-carcinoma. Mol Biol Rep 2012;39:9453–62.

48. Dong Z, Guo W, Guo Y, Kuang G, Yang Z. Concordant promoter meth-ylation of transforming growth factor-beta receptor types I and II occursearly in esophageal squamous cell carcinoma. Am J Med Sci 2012;343:375–81.

49. Yamashita S, Takahashi S, McDonell N, Watanabe N, Niwa T, Hosoya K,et al. Methylation silencing of transforming growth factor-beta receptortype II in rat prostate cancers. Cancer Res 2008;68:2112–21.

50. Cheng KH, Ponte JF, ThiagalingamS. Elucidationof epigenetic inactivationof SMAD8 in cancer using targeted expressed gene display. Cancer Res2004;64:1639–46.

51. Biswas S, Guix M, Rinehart C, Dugger TC, Chytil A, Moses HL, et al.Inhibition of TGF-beta with neutralizing antibodies prevents radiation-induced acceleration of metastatic cancer progression. J Clin Invest2007;117:1305–13.

52. Bedi A, Chang X, Noonan K, Pham V, Bedi R, Fertig EJ, et al. Inhibition ofTGF-beta enhances the in vivo antitumor efficacy of EGF receptor-targetedtherapy. Mol Cancer Ther 2012;11:2429–39.

53. Huang S, HolzelM, Knijnenburg T, Schlicker A, Roepman P,McDermott U,et al. MED12 controls the response to multiple cancer drugs throughregulation of TGF-beta receptor signaling. Cell 2012;151:937–50.






2016;76:3785-3801. Published OnlineFirst May 17, 2016.Cancer Res Sebastian Marwitz, Sofia Depner, Dmytro Dvornikov, et al.

Signaling and InvasionβCell Lung Cancer Enhances TGFSmall− Pseudoreceptor BAMBI in NonβDownregulation of the TGF

Updated version

10.1158/0008-5472.CAN-15-1326doi:

Access the most recent version of this article at:

Material

Supplementary

http://cancerres.aacrjournals.org/content/suppl/2016/05/17/0008-5472.CAN-15-1326.DC1

Access the most recent supplemental material at:

Cited articles

http://cancerres.aacrjournals.org/content/76/13/3785.full#ref-list-1

This article cites 53 articles, 12 of which you can access for free at:

Citing articles

http://cancerres.aacrjournals.org/content/76/13/3785.full#related-urls

This article has been cited by 2 HighWire-hosted articles. Access the articles at:

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

Subscriptions

Reprints and

[email protected]

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

Permissions

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

.http://cancerres.aacrjournals.org/content/76/13/3785To 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-15-1326

http://cancerres.aacrjournals.org/content/suppl/2016/05/17/0008-5472.CAN-15-1326.DC1

http://cancerres.aacrjournals.org/content/76/13/3785.full#ref-list-1

http://cancerres.aacrjournals.org/content/76/13/3785.full#related-urls

http://cancerres.aacrjournals.org/cgi/alerts

mailto:[email protected]

http://cancerres.aacrjournals.org/content/76/13/3785


downregulation of the tgfb pseudoreceptor bambi in non ... · brane-bound inhibitor (bambi), which...

Documents