Upload
others
View
7
Download
0
Embed Size (px)
Citation preview
Integrated Systems and Technologies
Molecular Portraits of Epithelial, Mesenchymal,and Hybrid States in Lung Adenocarcinoma andTheir Relevance to SurvivalMark J. Schliekelman1, Ayumu Taguchi2, Jun Zhu3, Xudong Dai3, Jaime Rodriguez2,Muge Celiktas4,Qing Zhang1, Alice Chin1, Chee-HongWong1, HongWang4, Lisa McFerrin1,Suhaida A. Selamat5, Chenchen Yang5, Evan M. Kroh1, Kavita S. Garg1, Carmen Behrens6,Adi F. Gazdar7, Ite A. Laird-Offringa5, Muneesh Tewari1,8,9, Ignacio I.Wistuba2,Jean P. Thiery10,11, and Samir M. Hanash4
Abstract
Epithelial-to-mesenchymal transition (EMT) is a key processassociated with tumor progression and metastasis. To definemolecular features associated with EMT states, we undertook anintegrative approach combining mRNA, miRNA, DNA methyla-tion, and proteomic profiles of 38 cell populations representativeof the genomic heterogeneity in lung adenocarcinoma. The result-ing data were integrated with functional profiles consisting of cellinvasiveness, adhesion, and motility. A subset of cell lines thatwere readily defined as epithelial or mesenchymal based on theirmorphology and E-cadherin and vimentin expression eliciteddistinctivemolecular signatures. Other cell populations displayedintermediate/hybrid states of EMT, with mixed epithelial and
mesenchymal characteristics. A dominant proteomic feature ofaggressive hybrid cell lines was upregulation of cytoskeletal andactin-binding proteins, a signature shared with mesenchymal celllines. Cytoskeletal reorganization preceded loss of E-cadherin inepithelial cells in which EMT was induced by TGFb. A set oftranscripts corresponding to the mesenchymal protein signatureenriched in cytoskeletal proteins was found to be predictive ofsurvival in independent datasets of lung adenocarcinomas. Ourfindings point to an association between cytoskeletal and actin-binding proteins, a mesenchymal or hybrid EMT phenotype andinvasive properties of lung adenocarcinomas. Cancer Res; 75(9);1789–800. �2015 AACR.
IntroductionEpithelial-to-mesenchymal transition (EMT) is a process in
embryonic development that allows polarized epithelial cells toconvert to loosely organized mesenchymal cells (1). The transi-
tion from an epithelial to a mesenchymal phenotype fosters cellmovement during gastrulation and later during morphogeneticevents such as neural crest ontogeny. Mesenchymal cells reachingthe target site engage in a new differentiation program allowingdevelopment of diverse tissue types (2–4). The same features ofEMT, namely loss of cell adhesion, increased migration, andinvasion that aid metazoan development provide a likely mech-anism for tumor progression with loss of an epithelial phenotypein more aggressive tumors (5–8).
Regulation of EMT is complex and multilayered, with diversegrowth factors, miRNAs, genetic mutations, and epigenetic altera-tions all having been shown to play a role. TGFb, hepatocytegrowth factor, Notch, orWnt can serve as initiating factors of EMT(9–13). Several miRNAs regulate EMT through inhibition ofeither effector genes or the signaling axis. The miR-200 familyand miR-34 inhibit EMT, whereas miR-21 has an opposite effect(14–16). Several studies demonstrated a role for DNA methyla-tion in regulating miR-200 and an altered DNA methylationprofile associated with EMT (17–20). Other EMT regulatorymechanisms include zinc-finger transcription factors Snail1,Snail2, Zeb1, Zeb2, and the basic helix–loop–helix family mem-bers Twist1 and Twist2 which control expression of downstreamgenes and cellular features associated with EMT, such as celladhesion and polarity (21–23). The transient nature of EMT hasbeen considered as aiding distant sitemetastasis through a reversemesenchymal-to-epithelial transition after invading cells havecolonized distant sites (8, 24). Recent work by Lu and colleaguesproposed occurrence of a hybrid epithelial-mesenchymal statewith a determination between the three phenotypes being
1Division of Public Health Sciences, Fred Hutchinson Cancer ResearchCenter, Seattle, Washington. 2Department of Translational MolecularPathology, The University of Texas MD Anderson Cancer Center,Houston, Texas. 3Department of Genetics and Genomic Sciences,Institute of Genomics and Multiscale Biology, Mount Sinai School ofMedicine, New York, New York. 4Department of Clinical Cancer Pre-vention,TheUniversityof TexasMDAndersonCancerCenter, Houston,Texas. 5Department of Surgery, Biochemistry and Molecular Biology,Norris Cancer Center, Keck School ofMedicine, University of SouthernCalifornia, LosAngeles,California. 6DepartmentofThoracic/HeadandNeck Medical Oncology,The University of Texas MDAnderson CancerCenter, Houston, Texas. 7Hamon Center for Therapeutic OncologyResearch,DepartmentofPathology,Universityof TexasSouthwesternMedical Center, Dallas, Texas. 8Division of Clinical Research, FredHutchinson Cancer Research Center, Seattle, Washington. 9Divisionof Human Biology, Fred Hutchinson Cancer Research Center, Seattle,Washington. 10Institute of Molecular Cell Biology, Singapore. 11Depart-ment of Biochemistry, Yong Loo Lin School of Medicine, NationalUniversity of Singapore, Singapore.
Note: Supplementary data for this article are available at Cancer ResearchOnline (http://cancerres.aacrjournals.org/).
Corresponding Author: Samir M. Hanash, Department of Clinical Cancer Pre-vention, MDAnderson Cancer Center, 6767Bertner Avenue, Houston, TX 77030.Phone: 713-745-5242; Fax: 713-792-1474; E-mail: [email protected]
doi: 10.1158/0008-5472.CAN-14-2535
�2015 American Association for Cancer Research.
CancerResearch
www.aacrjournals.org 1789
on April 12, 2017. © 2015 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from
Published OnlineFirst March 5, 2015; DOI: 10.1158/0008-5472.CAN-14-2535
regulated by a circuit composed of two interconnected chimericmodules—themiR-34/SNAIL and themiR-200/ZEBmutual-inhi-bition feedback circuits (25).
We have undertaken a study of lung adenocarcinoma to deter-mine molecular and phenotypic features associated with EMTstates and their relevance to survival in early-stage disease. Exten-sive characterization of lung adenocarcinoma cell lines, includingprotein, mRNA, miRNA, DNA methylation, cell invasiveness,adhesion and motility analysis, revealed the occurrence of inter-mediate/hybrid phenotypes between epithelial and mesenchy-mal states. Gene andprotein signatures associatedwith functionalcharacteristics helped to define these hybrid states. Upregulationof cytoskeletal-related proteins was a common feature betweenmesenchymal and aggressive hybrid types. A set of transcriptsenriched for cytoskeletal and actin binding proteins was found tobe predictive of survival in independent lung adenocarcinomadatasets.
Materials and MethodsCell culture
A panel of 38 lung adenocarcinoma cell lines selected toencompass known drivers in lung adenocarcinoma were grownin RPMI-1640 with 10% FBS and 1% penicillin/streptomycinunless otherwise noted. The identity of each cell line was con-firmed byDNA fingerprinting via short tandem repeats at the timeofmRNAand total protein lysate preparation using the PowerPlex1.2 kit (Promega). Fingerprinting results were compared withreference fingerprintsmaintained by the primary source of the cellline. For SILAC labeling of cell lines, cells were grown for sevenpassages in RPMI-1640 supplemented with 13C-lysine and 10%dialyzed FBS according to the standard SILAC protocol (26).
Mass spectrometric analysisProteomic analysis was performed as previously described
(27). Detailed methods for mass spectrometric analysis can befound in the Supplementary Information section.
Protein datasets are available as Supplementary Table S1.
Determination of EMT status for cell linesA protein ratio of surface localized CDH1 (CDH1_S) and VIM
from total cell extracts was calculated from (CDH1nþ1)/(VIMnþ1) where n is the number of spectral counts. HighCDH1_S/VIM was considered to be a log2-transformed ratio >0 and low to be log2-transformed ratio < 0. Cell morphology wasassessed by plating cells at 25% to 50% confluence and acquiringphase contrast images onday 1, 2, 3, and 4 after plating. Cellswereassessed for individual cell shape (spindle for mesenchymal orcuboid for epithelial) as well cell–cell interaction. Cells wereclassified as having epithelial morphology if the individual cellswere cuboid and cells grouped to form discrete clusters withsmooth edges indicative of tight junctions. Epithelial-like mor-phology lacked complete cuboid morphology or failed to formdiscrete clusters. Mesenchymal morphology required primarilyspindle shape andno cell–cell adhesion.Cellswithmesenchymal-like morphology were primarily spindle shaped but demonstrat-ed cell–cell adhesion by forming clusters. Cells with log2-trans-formed CDH1_S/VIM ratios > 0 and an epithelial morphologywere classified as epithelial, whereas mesenchymal cells hadlog2-transformed CDH1_S/VIM ratios < 0 and a mesenchymalmorphology.
Profiling of mRNA, miRNA, and DNA methylationGene expression data were obtained using Illumina Human
WG-6 v3.0 Expression BeadChips (Illumina) and expressionvalues log2 normalized. miRNA profiling was performed usinga real-time PCR-based approach using miRCURY LNA Univer-sal RT miRNA PCR (panel IþII; Exiqon, Inc.). miRNA profilingwas not available for cell lines H1299 and H1703. IlluminaInfinium HumanMethylation27 BeadChips were used forDNA methylation analysis. DNA methylation profiling was notavailable for cell lines H1385 and H1703. mRNA, DNA meth-ylation, and miRNA datasets were deposited in the NationalCenter for Biotechnology Information's Gene ExpressionOmnibus (GEO) database (http://www.ncbi.nlm.nih.gov/geo).
Data analysisDetailed methods for data analysis can be found in the Sup-
plementary Information section.
Invasion, migration, and aggregation assaysDetailed methods for invasion, migration, and aggregation
assays can be found in the Supplementary Information section.
Western blot analysisWestern blot analysis was performed according to standard
procedures using polyvinylidene difluoride membranes and anenhanced chemiluminescence system(GEHealthcare). Followingantibodies were used for Western blot analysis: ISYNA1 (SigmaAldrich), FBXO2 (Novus), TCEAL4 (Novus), FKBP65 (BD Bio-sciences), vimentin (BD Biosciences), CDH1 (BD Biosciences),and AKAP12 (Abcam). a-tubulin (Sigma) was used as a loadingcontrol.
Immunofluorescence analysisDetailed methods for immunofluorescence analysis and IHC
analysis can be found in the Supplementary Information section.
ResultsCharacterization of cell lines based on their morphology andCDH1/VIM ratios
To define molecular features that distinguish epithelial frommesenchymal cells, a panel of 38 lung adenocarcinoma celllines representative of the genomic diversity of this disease wassubjected to proteomic, gene expression, miRNA, and DNAmethylation profiling (Supplementary Fig. S1A). Changes inCDH1 and vimentin (VIM) have been considered hallmarksof EMT. Expression of CDH1 on the cell surface and VIM inwhole-cell lysates was determined based on normalized spectralcounts from mass spectrometry data (27). We assessed ratios ofcell surface-localized CDH1 (CDH1_S) and VIM from whole-cell lysates along with cell morphology, and identified a subsetof cell lines with a distinct mesenchymal or epithelial pheno-type (28). Nine cell lines with a log2-transformed CDH1_S/VIMratio > 0 and an epithelial morphology were classified as epi-thelial, while nine cell lines with a log2-transformed CDH1_S/VIM ratio < 0 and a mesenchymal morphology were classifiedas mesenchymal (Fig. 1A and Supplementary Fig. S1B). Log2-transfomed CDH1_S/VIM protein ratio was significantly cor-related with CDH1/VIM ratios of mRNA expression (r¼ 0.8650,P < 0.0001; Spearman correlation). Common somatic genemutations that occur in lung adenocarcinoma (Kras, TP53,
Schliekelman et al.
Cancer Res; 75(9) May 1, 2015 Cancer Research1790
on April 12, 2017. © 2015 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from
Published OnlineFirst March 5, 2015; DOI: 10.1158/0008-5472.CAN-14-2535
EGFR) were not associated with a distinct EMT phenotype, withthe exception of a negative correlation between EGFR mutationand a mesenchymal type as previously reported (29). Theremaining cell lines could not be readily classified as epithelialor mesenchymal due to discordance between CDH1_S/VIMratios and morphology and were investigated further for theirhybrid properties. Immunofluorescence analysis of CDH1 andVIM revealed that both CDH1 and VIM were stained in thesame cells in hybrid cell lines (Fig. 1B). We further investigatedCDH1 and VIM protein expression in lung adenocarcinomatissues. Among 141 lung adenocarcinoma tissues in the tissuemicroarray, 29 (20.6%) tumors were both CDH1 and VIMpositive (Fig. 1C and D), indicative of a hybrid transcriptionalprogram.
Identification of distinctive gene and protein signatures formesenchymal and for epithelial cell lines
We performed a factor analysis to determine modality ofgene expression patterns among the cell lines, which revealed acontinuous rather than a modal distribution (Fig. 2A). Genescomprised in the discriminating factors included known mar-
kers of epithelial or mesenchymal cells, including CDH1, VIM,and EpCAM (data not shown). Comparison of mRNA expres-sion between the nine epithelial and nine mesenchymal celllines yielded 1,347 genes with a P value < 0.01 (t test), con-sisting of 659 with higher expression in mesenchymal and 688in epithelial cell lines (Fig. 2B and Supplementary Table S2A).Gene ontology enrichment using DAVID (28) was applied todifferentially expressed genes. The epithelial genes wereenriched for genes encoding proteins localized to the cellsurface, many of which play a role in cell adhesion, whereasmesenchymal genes were enriched for nuclear localized pro-teins and regulators of transcription (Supplementary TableS2B). Concordant findings were observed when gene ontologyfor 1,347 epithelial and mesenchymal genes was analyzed withusing Gene Set Enrichment Analysis (GSEA; ref. 30; Supple-mentary Table S2C).
Extensive proteomic analysis of cell lysates by LC/MS-MS forall 38 cell lines identified a total of 12,808 distinct proteinswith an average of 3,690 proteins identified in each cell line,pointing to substantial heterogeneity among cell lines (Sup-plementary Table S1 and Supplementary Fig. S2A). Factor
A
B
Epithelial
Epithelial-like
Mixed mesenchymal/epithelial
Mesenchymal-like
Mesenchymal
C
D
Negative Positive Total
Negative 36 35 71
Positive 41 29 70
Total 77 64 141
VIM
CDH1
H16
93
H82
0
H19
93
HC
C40
19
HC
C29
35
H14
37
H15
73
PC
-9
HC
C40
06
HC
C40
17
H13
95
H32
55
H13
85
H16
50
H96
9
DF
CI0
32
HC
C82
7
H15
68
H20
09
HC
C40
11
H13
73
H21
22
H19
44
H19
75
H64
7
HC
C22
79
H23
H24
05
H52
2
H22
28
DF
CI0
24
H17
03
H12
99
H83
8
H13
55
H17
92
H65
0
H20
30Morphology
EMT status
E-cadherin Vimentin
E-cadherin Vimentin
DAPI merge
DFCI032
Pt #1
Pt #2
H1373
PC-9
Lo
g2-
tran
sfo
rmed
CD
H1/
VIM
rat
io
–10
–5
0
5
10
Protein
mRNA
Figure 1.Classification of NSCLC cell lines. A, log2-transformed ratios of spectral counts of CDH1 on cell surface and VIM from whole-cell lysates, log2-transformed ratiosof CDH1/VIM mRNA, and morphology, with representative images. Scale, 40 mm. B, immunofluorescence analysis of E-cadherin and vimentin in hybridcell lines. C, CDH1 and VIM expression in lung adenocarcinoma tissue microarray. D, representative images of IHC analysis of E-cadherin and vimentin.
Molecular Signatures of EMT Status in Lung Adenocarcinoma
www.aacrjournals.org Cancer Res; 75(9) May 1, 2015 1791
on April 12, 2017. © 2015 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from
Published OnlineFirst March 5, 2015; DOI: 10.1158/0008-5472.CAN-14-2535
analysis performed with proteomic data similarly produced acontinuum rather than a modal distribution (Fig. 2C). Proteinscomprised in the discriminating factors included VIM but notCDH1 nor EpCAM (data not shown). Proteomic analysisresulted in 232 proteins expressed more highly in epithelialcell lines and 166 proteins expressed more highly in mesen-chymal cell lines with a P value < 0.05 (Mann–Whitney U test)and fold change > 1.5 (Fig. 2D and Supplementary Table S2D).The epithelial and mesenchymal mRNA signatures significantlyoverlapped with corresponding protein signatures, respectively(P ¼ 3.29 � 10�20, and P ¼ 6.61 � 10�10, Fisher exact test;Supplementary Fig. S2B). Significantly enriched GO categoriesin the proteomic mesenchymal signature consisted predomi-nantly of cytoskeleton and actin organization (GO:0005856~cytoskeleton, GO: 0008092~cytoskeletal proteinbinding, GO: 0003779~actin binding; Supplementary TableS2E). The enrichment in cytoskeletal and actin-related proteinswas also prominent by Ingenuity Pathway Analysis (http://www.ingenuity.com/; Supplementary Fig. S2C and S2D). The
GO terms enriched in the protein signature for epithelial cellsincluded those associated with translation and metabolism(Supplementary Table S2E). To confirm functional relevanceof cytoskeletal proteins in mesenchymal signature, we per-formed knockdown experiments of AKAP12, which is associ-ated with actin-cytoskeleton reorganization (31) and identifiedin both mRNA and protein mesenchymal signatures (Supple-mentary Tables S2A and S2C). Although no obvious changewas observed in EMT status, cell invasion was inhibited inH1299 cells by treatment with AKAP12 shRNA (SupplementaryTable S2E), indicating the functional relevance of cytoskeletalproteins in mesenchymal signature.
To explore potential regulatory factors for gene and proteinexpression, miRNA and DNA methylation profiling was per-formed. Unsupervised hierarchical clustering of microRNA dataresulted in two clusters that separated the epithelial and mesen-chymal cell lines and significantly differed in CDH1_S/VIMexpression (P ¼ 5.6 � 10�6, t test; Fig. 2E). Separation of thecell lines into two clusters was principally due to expression of
miRNA
A
E
G
C
F
mRNA
Methylation
miRNA
Protein
8 31
178 210
24
43659 688
48 75
B
D I1.0
0.8
0.6
0.4
0.2
0
-0.2
-0.2 0 0.2 0.4 0.6 0.8 1.0
5
4
3
2
1
0-5 0 5
EpithelialMesenchymal
4
2
0
-2
-4
-6
Epithelial Mesenchymal
H
P v
alue
(-lo
g 10)
Fold change (-log2)
Median β-value difference
miRNA w/o miR-200
-4 -2 0 2 4
5
4
3
2
1
0
1.0
0.8
0.6
0.4
0.2
0
-0.2
-0.2 0 0.2 0.4 0.6 0.8 1.0
Factor 1
Fac
tor
2
Factor 1
Fac
tor
2
Mesenchyal
-1 -0.5 0 0.5 1
Mesenchymal Epithelial
P v
alue
(-lo
g 10)
P v
alue
(-lo
g 10)
Fold change (-log2)-8 -4 0 4 88 -4 0 4
6
5
4
3
2
1
0
Epithelial
Epi
thel
ial
Mes
ench
yal
gene
exp
ress
ion
(-lo
g 10
P )
Methylation (-log 10 P )
Figure 2.Molecular characterization of epithelial and mesenchymal cell lines. A, factor analysis of gene expression showing distribution of cell lines. B, volcano plot ofthe differential gene expression analysis. Gray boxes, significance cutoffs. C, factor analysis of protein expression showing distribution of cell lines. D, volcano plotof the differential protein expression analysis. Gray boxes, significance cutoffs. E, unsupervised hierarchical clustering of miRNA expression from 36 lungadenocarcinoma cell lines separates the epithelial and mesenchymal cell lines into distinct clusters. F, unsupervised hierarchical clustering of miRNA expressionwithout miR-200 family members. G, volcano plot of the differential DNA methylation analysis. Gray boxes, significance cutoffs. H, starburst plot integratingdifferential DNA methylation and gene expression analyses. I, overlapping regulation of gene and protein expression. For sub figures A, C, E, and F, mesenchymalcell lines and hybrid cell lines are indicated.
Schliekelman et al.
Cancer Res; 75(9) May 1, 2015 Cancer Research1792
on April 12, 2017. © 2015 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from
Published OnlineFirst March 5, 2015; DOI: 10.1158/0008-5472.CAN-14-2535
the miR-200 family, as removal of miR-200 family membersfrom the dataset resulted in lack of clustering into two clusters(Fig. 2F). Restricting comparison of the miRNA data to theepithelial and mesenchymal cell lines resulted in 31 differen-tially expressed miRNAs at P < 0.01 (t test), with 10 expressed atlower levels in mesenchymal cell lines and 21 expressed atlower levels in epithelial cell lines (Supplementary Table S2F).Of the 10 miRNAs with lower expression in the mesenchymalcell lines, six had predicted binding sites in the mesenchymalgene signature (miR-200a, 200b, 200c, 429, 135b, and 148a)using the prediction algorithms miRanda and Targetscan (datanot shown). Of the 21 miRNAs with lower expression inepithelial cell lines, four had predicted binding sites in theepithelial gene signature (miR-30a, 330-3p, 425, 455-3p). ThemiRNAs identified by both differential expression and algo-rithmic analysis of signatures are predicted to regulate 178 and210 genes in the epithelial and mesenchymal signatures,respectively (Supplementary Table S2F).
Comparison of the DNA methylation status of the mesen-chymal and epithelial cell lines using Infinium HumanMethy-lation27 BeadChips identified 75 hypo-methylated genes in theepithelial cell lines and 48 hypo-methylated genes in themesenchymal cell lines (Fig. 2G and Supplementary TableS2G). Thirty one of the 75 hypo-methylated genes in theepithelial signature were also represented in the epithelialmRNA signature (P ¼ 2.41 � 10�26, Fisher exact test). Con-versely, eight of the 48 hypo-methylated genes in the mesen-chymal cell lines were represented in the mesenchymal mRNAsignature (P ¼ 1.281 � 10�4, Fisher exact test; Fig. 2H).Assessment of the relative contribution of DNA methylation,miRNA, and gene expression to protein levels suggested greaterconcordance between gene and protein expression for epithe-lial than mesenchymal gene sets (Fig. 2I). Comparison oftranscript and protein variance between mesenchymal andepithelial cell lines further revealed increased variance of pro-tein expression in mesenchymal versus epithelial cell lines,compared with gene expression (Supplementary Table S2G).
Heterogeneity among cell lines with a hybrid phenotypeWe next explored phenotypic differences and similarities
between epithelial, hybrid, and mesenchymal cell lines. Weobserved a significant difference in invasiveness between theepithelial and mesenchymal cell lines based on a Matrigel inva-sion assay (P ¼ 0.0001, Mann–Whitney U test; Fig. 3A). Hybridcell lines were heterogeneous in their invasive properties. Expres-sion of the mesenchymal proteins (r ¼ 0.736, Spearman corre-lation) and genes (r ¼ 0.715, Spearman correlation) had signif-icantly higher correlation with invasiveness than CDH1_S/VIMexpression alone (r ¼ �0.483, Spearman correlation; P ¼ 0.041,Fisher r-to-z transformation).We observed a significant differencein cell migration between the epithelial and mesenchymal celllines based on scratch wound assays (P¼ 0.0012,Mann–WhitneyU test; Fig. 3B). Expression of mesenchymal signature proteinsand genes but not CDH1_S/VIM expression significantly corre-lated with migration (r¼ 0.546, Spearman correlation; P¼ 7.0�10�4,Mann–WhitneyU test and r¼ 0.599, Spearman correlation;P ¼ 2.0 � 10�4, Mann–Whitney U test, respectively). Cell–celladhesion is a hallmark of epithelial cells and loss of cell–celladhesion is considered to be a critical step in metastasis (4).Analysis of cell aggregation in a liquid culture allows assessmentof the strength of cell–cell cadherin/catenin complex binding, as
cells do not have a solid surface on which to bind. We observedsignificantly more cell aggregation in epithelial cell lines com-paredwithmesenchymal cell lines (P¼0.0251,Mann–WhitneyUtest), with the hybrid cell lines distributed across the spectrum(Fig. 3C). Cell line growth rates were also assessed. We did notobserve a significant difference in cell growth rates betweenepithelial, hybrid, and mesenchymal cell lines as the variance ingrowth rates within each class was greater than the differencebetween them (data not shown).
Hierarchical clustering was performed based on the gene andprotein signatures to determine their ability to discriminatebetween epithelial, mesenchymal, and hybrid cell lines. Cluster-ing based on themRNA signature separated themesenchymal celllines into one group and eight of the nine epithelial cell lines intothe other (Fig. 3D). The two clusterswere significantly different forCDH1_S/VIM expression (P ¼ 1.49 � 10�5, Mann–Whitney Utest), invasion (P¼ 0.002, Mann–WhitneyU test), andmigration(P¼0.009,Mann–WhitneyU test). Clustering of thedifferentiallyexpressed proteins separated the epithelial and mesenchymal celllines into distinct groups that were significantly different forCDH1_S/VIM expression (P ¼ 0.003, Mann–Whitney U test),invasion (P¼ 3.0� 10�4, Mann–WhitneyU test), and migration(P ¼ 0.004, Mann–Whitney U test) and aggregation (P ¼ 0.006,Mann–Whitney U test; Fig. 3E). The large number of upregulatedproteins in mesenchymal cell lines includes many that havenot been previously associated with EMT and thus would repre-sent novel biomarkers. We selected several novel proteins(TCEAL4, FBXO2, FKBP65, and ISYNA1) and confirmed theirincreased expression in mesenchymal cell lines by Western blot-ting (Fig. 3F).
Hybrid EMT states are distinguishable by molecular featureslinked to their phenotypes
We observed phenotypic heterogeneity among the hybridcell lines that was unrelated to their CDH1 and VIM expression.Cell lines in the hybrid group exhibited phenotypic traitsof high invasion and migration that are a characteristic ofa mesenchymal phenotype, together with high aggregation, afeature of epithelial cells (Fig. 4A). A subset of hybrid cell lines(DFCI032, H1650, H1693, HCC827, and PC-9) exhibited highCDH1_S/VIM ratios but were also invasive and migratory(aggressive hybrid; Fig. 4B). Interestingly, although the mesen-chymal cell lines migrated primarily as single cell, four of thefive aggressive hybrid cell lines migrated by collective groupmigration (Fig. 4C).
Recent work by Lu and colleagues proposed Zeb/Snail/miR-200/miR-34 axis as regulators of hybrid phenotypes (25).We assessed transcript and protein expression of a large numberof genes previously identified as regulators or markers of epi-thelial or mesenchymal cell types (Fig. 5A for significantlyexpressed markers and Supplementary Table S3 for all markers).We identified upregulation of ZEB1 (P ¼ 0.042, Mann–WhitneyU test) and SNAI2 (P ¼ 0.041, Mann–Whitney U test) mRNAexpression levels in the aggressive hybrid cell lines comparedwith epithelial cell lines. We observed increased expression ofmiR-34a in the aggressive hybrid cell lines compared with theepithelial cell lines, whereas expression of miR-200 familymembers was not significantly different (Fig. 5B), concordantwith the results that four of aggressive hybrid cell lines wereclustered together with epithelial cell lines (Fig. 2E). A compar-ative analysis between the epithelial and aggressive hybrid cell
Molecular Signatures of EMT Status in Lung Adenocarcinoma
www.aacrjournals.org Cancer Res; 75(9) May 1, 2015 1793
on April 12, 2017. © 2015 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from
Published OnlineFirst March 5, 2015; DOI: 10.1158/0008-5472.CAN-14-2535
lines yielded 197 genes with upregulated mRNA levels in theaggressive hybrid cell lines out of 20,598 total genes (Fig. 5C). Aset of 135 proteins were upregulated in the aggressive hybrid celllines compared with epithelial cell lines (Supplementary TableS4A). Interestingly, the 135 protein signature was enriched forGO terms associated with cytoskeleton, actin binding, andorganization, and significantly overlapped with the mesenchy-mal protein signature (P ¼ 2.062 � 10�23 Fisher exact test; Fig.5D and E and Supplementary Table S4B). Expression of cyto-skeletal and actin-binding proteins in the aggressive hybrid celllines was significantly higher than in the epithelial (P ¼ 0.005,Mann–Whitney U test) or other hybrid cell lines (P ¼ 0.009,Mann–Whitney U test; Fig. 5F) and was the primary discrimi-nator of aggressive hybrid cell lines. Protein expression levels ofTCEAL4 and ISYNA1 that are part of the mesenchymal proteinsignature were significantly elevated in the hybrid aggressive celllines compared with epithelial type as determined by massspectrometry, further supporting overlapping molecular charac-teristics of aggressive hybrid type and mesenchymal cell lines.
TGFb-induced upregulation of cytoskeletal proteins precedesloss of E-cadherin.
We next tested induction of EMT with TGFb treatment toconfirm that proteomic differences observed between the mes-enchymal and epithelial cell lines were due to an EMT event.Following exposure to TGFb, epithelial H1437 cells elongatedand lost cell adhesion (Fig. 6A). We further observed increasedcell invasion and migration, but no changes in cell aggregation(Fig. 6B and C). We next performed a proteomic analysis ofH1437 cells after 8 days of treatment with TGFb in comparisonwith untreated cells, selecting differentially expressed proteinswith > 1.5 fold change and G score greater than 3.85 (equivalentto a P value < 0.05; Fig. 6D and Supplementary Table S5A).Changes in protein expression after TGFb treatment correlatedsignificantly with the differentially expressed epithelial andmesenchymal proteins and the aggressive hybrid protein signa-ture (P ¼ 1.14 � 10�9 and P ¼ 7.53 � 10�10, respectively,Spearman correlation; Fig. 6E). Gene Ontology analysis identi-fied cytoskeletal and actin-binding proteins, including AKAP12,
Diffuse
Aggregates
Compact
0
200
400
600
800
1,000
A
B
C
EMT status
D
E
EMT status
0
20
40
60
80
100
H13
95
H14
37
H15
73
H32
55
H82
0 H
969
HC
C29
35
HC
C40
17
HC
C40
19
DF
CI0
32
H16
50
H16
93
HC
C82
7 P
C- 9
H
1373
H
1385
H
1568
H
1944
H
1975
H
1993
H
2009
H
2122
H
CC
4006
H
CC
4011
H
1355
H
1792
H
2228
H
522
HC
C22
79
DF
CI0
24
H12
99
H17
03
H20
30
H23
H
2405
H
647
H65
0 H
838
Epithelial Mesenchymal
ISYNA1
FBXO2
TCEAL4
FKBP10
α-Tubulin
H13
95
H14
37
H15
73
H32
55
H82
0 H
969
HC
C29
35
HC
C40
17
HC
C40
19
DF
CI0
32
H16
50
H16
93
HC
C82
7 P
C- 9
H
1373
H
1385
H
1568
H
1944
H
1975
H
1993
H
2009
H
2122
H
CC
4006
H
CC
4011
H
1355
H
1792
H
2228
H
522
HC
C22
79
DF
CI0
24
H12
99
H17
03
H20
30
H23
H
2405
H
647
H13
95H
1437
H15
73H
3255
H82
0H
969
HC
C29
35H
CC
4017
HC
C40
19
DF
C10
24
H12
99
H17
03
H20
30
H23
H24
05
H64
7
H65
0
H83
8
H65
0 H
838
H13
95
H14
37
H15
73
H32
55
H82
0 H
969
HC
C29
35
HC
C40
17
HC
C40
19
DF
CI0
32
H16
50
H16
93
HC
C82
7 P
C-9
H
1373
H
1385
H
1568
H
1944
H
1975
H
1993
H
2009
H
2122
H
CC
4006
H
CC
4011
H
1355
H
1792
H
2228
H
522
HC
C22
79
DF
CI0
24
H12
99
H17
03
H20
30
H23
H
2405
H
647
H65
0 H
838
Inva
sive
cel
ls/fi
eld
% M
igra
tion
F
Protein
mRNA
Figure 3.Phenotypic characterization of cell lines. A, cell invasion through Matrigel. In at least four fields from six replicate wells, the number of cells was counted foreach cell line. B, cell migration measured by a scratch wound assay, with six scratches measured per cell line. C, cell aggregation after 24 hours in liquid cultureover agarose as a measure of cell–cell adhesion. Six replicates with assay performed in three individual wells on different days. D, hierarchical clustering ofdifferentially expressed genes in the epithelial and mesenchymal cell lines. E, hierarchical clustering of differentially expressed proteins in the epithelial andmesenchymal cell lines. For all sub figures, mesenchymal cell lines, epithelial, and hybrid cell lines are indicated.. F, confirmation of protein expression ofnovel EMT-related proteins. Western blotting of several EMT-related proteins showed increased expression of novel markers in mesenchymal cell linescompared with epithelial cell lines.
Schliekelman et al.
Cancer Res; 75(9) May 1, 2015 Cancer Research1794
on April 12, 2017. © 2015 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from
Published OnlineFirst March 5, 2015; DOI: 10.1158/0008-5472.CAN-14-2535
as the most highly enriched group among the upregulatedproteins (Supplementary Table S5B). Downregulated proteinset with TGFb treatment overlapped with the epithelial proteinsignature (P ¼ 9.88 � 10�20 Fisher exact test). Downregulatedproteins were enriched for oxidation reduction and multiplecategories related to metabolism and glycolysis. The morpho-logic and cytoskeletal protein expression alterations precededchanges in EMTmarkers, as we did not observe changes in CDH1and VIM proteins after 8 days of TGFb treatment (Fig. 6F)as confirmed by Western blotting. Loss of CDH1 expressionand increase in VIM expression were observed at 14 days. Thesimilarity between protein changes induced by TGFb treatmentof epithelial cells and the aggressive hybrid signature (invasive,migratory, high expression of cytoskeletal proteins) supportsreorganization of cytoskeletal proteins preceding loss of CDH1as an intermediate stage in EMT, with aggressive hybrid cell linesexpressing epithelial markers (Fig. 6G).
Relevance of a mesenchymal signature enriched in genesencoding for cytoskeletal proteins to survival in early stage lungadenocarcinomas
We next determined whether the mesenchymal gene signaturewe identified had predictive value in early-stage lung adenocar-cinoma by interrogating three independent gene expressiondatasets of lung adenocarcinoma annotated for outcome- Direc-tor's Challenge, Bhattacharjee and colleagues and Tomida andcolleagues datasets (32–34). Stage 1 and 2 tumors were rankedby their relative expression of genes in the signature and hazardratios were calculated by Cox regression. The mesenchymalmRNA signature significantly predicted survival in the Director'sChallenge dataset (P ¼ 7.62 � 10�3 Cox regression; Fig. 7A).Given the lack of databases for lung tumor protein expressionannotated for survival, we next tested whether a set of mRNAsspecifically encoding the mesenchymal proteomic signature hadprognostic value. We first correlated protein abundance with
C
EMT StatusMorphologyCDH1_S/VIMInvasionMigrationAggregation
H13
95H
1437
H15
73H
3255
H82
0H
969
HC
C29
35H
CC
4017
HC
C40
19
DF
CI0
32H
1650
H16
93H
CC
827
PC
- 9
H13
85H
1568
H19
93H
2009
HC
C40
06H
1373
H19
44H
1975
H21
22H
CC
4011
H13
55H
1792
H22
28H
522
HC
C22
79D
FC
I024
H12
99H
1703
H20
30H
23H
2405
H64
7H
650
H83
8
A
B
Aggressive hybrid
0 10 20 30 40 50 60 70 80 90 100
800
600
400
200
0
% migration
Inva
sive
cel
ls p
er fi
eld
Adhesion
Low HighIntermediate
Hour 0
DF
CI0
24
HC
C82
7P
C-9
Rep
rese
ntat
ive
imag
es o
f sin
gle
cell
mig
ratio
n
H16
50H
1693
Hour 12 Hour 0 Hour 12
Figure 4.Identification of aggressive hybrid celllines. A, plot of cell invasion againstmigration with cell–cell adhesionindicated by color. B, cell lines areheterogeneous for phenotypiccharacteristics with hybrid epithelial-to-mesenchymal cells lineshighlighted. Log2-transformed.CDH1_S/VIMprotein ratios:& >0, and& <0. Invasion:& > 150 cells per field;& < 150 cells per field. Aggregation,&, diffuse; &, aggregates; &,compact. Migration: & � <33% areacovered; & >33% and <63% areacovered; & > 63% area covered. Forcell morphology: EMT status: &,epithelial; &, hybrid; &,mesenchymal. For cell morphology,mesenchymal, mesenchymal-like,mixed mesenchymal/epithelialmorphology, epithelial-like, andepithelial are indicated. For EMTstatus, mesenchymal cell lines,epithelial, and hybrid cell lines areindicated. C, migration of aggressivehybrid cell lines. Aggressive hybrid celllines imaged at 0 and 12 hours during ascratch wound assay reveal collectivecell migration in four of five cell lines.
Molecular Signatures of EMT Status in Lung Adenocarcinoma
www.aacrjournals.org Cancer Res; 75(9) May 1, 2015 1795
on April 12, 2017. © 2015 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from
Published OnlineFirst March 5, 2015; DOI: 10.1158/0008-5472.CAN-14-2535
mRNA expression by Spearman correlation and demonstratedsignificant correlation between transcript and protein expressionwith a mean of correlation coefficients of 0.216 (P ¼ 1.0 �10�16 based on permutation tests; Fig. 7B). Transcript: proteinexpression increased with measures of protein abundance (Fig.7C). The set of mRNAs encoding the mesenchymal proteinsignature significantly predicted reduced survival in all threetumor datasets tested, whereas the epithelial signature wassignificantly associated with increased survival in one dataset(Fig. 7A). The Director's Challenge Consortium for the Molec-ular Classification of Lung Adenocarcinoma tested 14 differentmethods for predicting survival in early-stage non–small celllung cancer with the best predictor found to be Model A,consisting of 13,830 genes produced from clustering (32). As
a comparison, we tested Model A in the two other independentdatasets and found that the mesenchymal protein signature hadsimilar statistical significance as Model A (32). We next per-formed a Kaplan–Meier analysis and found that the mesenchy-mal signature significantly predicted survival in all three datasetstested (the Director's Challenge, Bhattacharjee and colleaguesand Tomida and colleagues datasets; Fig. 7D). Given thatupregulation of cytoskeletal proteins was associated with bothmesenchymal and aggressive hybrid phenotypes, we testedwhether mRNAs representing the cytoskeletal proteins from themesenchymal protein signature were predictive of survival. Wefound that this restricted signature also significantly predictedsurvival in the Director's Challenge and Bhattacharjee andcolleagues tumor sets (Fig. 7E).
C
Cytoskeletal proteins
P = 0.005
Mesenchymal Aggressive Hybrid
9243123
A
B
F
6
5
4
3
2
1
0-5 0 5
5
4
3
2
1
0-1 0 1
Aggressive hybrid EpithelialmRNA Protein
CD
H1
CT
NN
A1
CT
NN
B1
CT
NN
D1
TJP
2
EP
CA
M
TA
CS
TD
2
VIM
AC
TN
1
CD
44
CD
H1
CD
H3
CT
NN
A1
CT
NN
D1
ST
14
TJP
2
TJP
3
TA
CS
TD
1
TA
CS
TD
2
VIM
ZE
B2
ZE
B1
TW
IST
2
TW
IST
1
SN
AI2
TH
Y1
* * * * * * * * * * * * * * * * * * * * * * * * *x x x x x x x x x x x x x x x x x x x x x x x xx
* *
D
E
P v
alue
(-lo
g 10)
P v
alue
(-lo
g 10)
Fold change (-log2) Fold change (-log2)
Protein mRNA
Epi vs. Mes
Epi vs. AH
Epithelial
Aggressive hybrid
Mesenchymal
Mesenchymal
Aggressive hybrid
Cytoskeleton
Actin binding
Actin cytoskeleton organization
Cytoskeleton organization
Cytoskeleton protein binding
Actin cytoskeleton
30
Ave
rage
of n
orm
aliz
edM
S s
pect
ra c
ount
s
20
10
0
Aggressive hybrid Epithelial
Low High
P = 0.009
Figure 5.Molecular signatures for aggressive hybrid cell lines. A, expression of epithelial and mesenchymal markers at mRNA and protein levels in epithelial, aggressivehybrid, and mesenchymal cell lines. � , P value by t test under 0.05; X, P value� 0.05. B, expression of epithelial- and mesenchymal-related miRNAs in epithelial,aggressive hybrid, and mesenchymal cell lines. � , P value by t test under 0.05; X, P value � 0.05. C, volcano plot of the differential genes (left) and proteins(right) in a comparison of epithelial and aggressive hybrid cell lines. Gray boxes, significance cutoffs. D, overlap of mesenchymal and aggressive hybrid proteinsignatures. E, overlap of mesenchymal and aggressive hybrid protein significant gene ontology categories. F, cytoskeletal proteins are upregulated in themesenchymal and aggressive hybrid cell lines compared against epithelial cell lines. Epi, epithelial; AH, aggressive hybrid; Mes, mesenchymal.
Schliekelman et al.
Cancer Res; 75(9) May 1, 2015 Cancer Research1796
on April 12, 2017. © 2015 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from
Published OnlineFirst March 5, 2015; DOI: 10.1158/0008-5472.CAN-14-2535
Discussion
We have undertaken proteomic, gene expression, miRNA, andDNA methylation analysis of lung adenocarcinoma cell linesrepresentative of genomic heterogeneity in lung adenocarcinoma,together with their functional characterization. Subsets thatemerged from the study encompassed epithelial, mesenchymal,and an aggressive hybrid group with features of both epithelialand mesenchymal cell lines characterized by upregulation ofcytoskeleton and actin-binding proteins. Findings from TGFbtreatment of epithelial cells support the occurrence of an inter-mediate state during EMT with hybrid features. In considerationof the role that EMT plays in tumor progression, we sought toelucidate gene sets that may be predictive outcome in early-stagelung cancer based on biologic functions and identified a signatureenriched in cytoskeletal protein encoding genes predictive ofsurvival.
Molecular profiling has revealed substantial tumor heteroge-neity in many human cancers (35–38). The extensive molecularand phenotypic characterization of lung adenocarcinoma celllines likewise has revealed substantial heterogeneity amongst thecell lines. Most cell lines could not be simply categorized as eithermesenchymal or epithelial. We identified cell lines with features
of both mesenchymal and epithelial cell types, substantiating theoccurrence of a hybrid state in tumor cell populations (25).Costained E-cadherin and vimentin in hybrid cell lines suggestedthe occurrence of hybrid transcriptional program. Partial EMT hasalso been described during development, wound healing inaddition to tumorigenesis (39–42), whereas the existence ofsubpopulation of E-cadherin-positive and vimentin-positive cellshas been recently indicated in head and neck cancer (43, 44),supporting the need to further characterize this group at thegenomic and proteomic levels as we have undertaken in thisstudy. In addition, our IHC studies of CDH1 and VIM proteinwere performed using nonserial section of lung adenocarcinomatissues.On the basis of this potential limitation, further validationstudies for the occurrence of hybrid type in tumors are alsowarranted. A subset of hybrid cell lineswas highly invasive despitegene and protein expression of epithelial markers. Aggressivehybrid cell lines expressed a similar pattern of upregulated cyto-skeletal and actin binding as mesenchymal cell lines. The occur-rence of hybrid EMT states is supported by our analysis of TGFbinduction of EMT in the H1437 cell line in which we observedphenotypic alterations andupregulation of cytoskeletal and actin-binding proteins before changes in cadherin or vimentin expres-sion. Lu and colleagues proposed a regulatory switch centered
Day 0 Day 8 Day 14
Invasion Migration
TGFβ treatment
0
10
20
30
40
50
60
70
80
H1437
Cel
ls p
er f
ield
UntreatedTGFβ
0
20
40
60
80
100
H1437
% M
igra
tio
n
UntreatedTGFβ
CDH1
VIM
TGFβ, Day 14
TGFβ, Day 8Control
β-Actin
Differential protein expression after TGFβ treatment
Untreated TGFβ
Aggressivehybrid
111 257
H1437_TGFβUp
24
A
B
C
D
Untreated TGFβ
TGF
Log2-transformedCDH_S/VIM ratioMorphology
Invasion
Migration
Cell aggregation
Cytoskeletal/Actin-binding proteins
CDH1 23 21
EPCAM 48 61
TJP1 13 19
TJP2 10 9
VIM 0 0
Peptide counts by LC-MS/MS
Log2 fold change
G s
core
E
F
–10 –5 0 5 10
400
300
200
100
0
GEpithelial
Aggressivehybrid
Epithelial
>0
Low
Low
Low
Aggregate
Epithelial Mesenchymal
Mesenchymal
>0 <0
High
High
High
Aggregate
High
High
High
Diffuse
Figure 6.Proteomic analysis of TGFb induced EMT in an epithelial cell line. A, H1437 cells change from epithelial to mesenchymal morphology with TGFb treatment. Scale,40 mm. B, invasion and migration of H1437 increase with TGFb treatment. C, aggregation of H1437 cells is unaffected by TGFb treatment. Scale, 200 mm. D,volcano plot of the differential protein expression analysis. Gray boxes, significance cutoffs. E, proteins upregulated in H1437_TGFb cells overlap significantlywith the aggressive hybrid protein signatures. F, expression of epithelial or mesenchymal markers in TGFb treated H1437 by LC-MS/MS or Western blot analysis.G, alteration of molecular and functional characteristics during EMT via aggressive hybrid type in H1437 with TGFb treatment.
Molecular Signatures of EMT Status in Lung Adenocarcinoma
www.aacrjournals.org Cancer Res; 75(9) May 1, 2015 1797
on April 12, 2017. © 2015 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from
Published OnlineFirst March 5, 2015; DOI: 10.1158/0008-5472.CAN-14-2535
around miR-200/Zeb and Snail/miR-34 that regulates hybridstates of EMT (25). Our findings support the importance of ZebandmiR-34a in the regulation of hybrid phenotypes.Weobservedlimited concordance betweenmRNAandproteins for somegenes,particularly for cells with a mesenchymal phenotype, suggestingan important role for posttranscriptional regulation affectingEMT.
Cytoskeletal rearrangements emerged as the dominant featureof mesenchymal and invasive cells based on both mRNA andprotein analysis. This finding is further supported by upregulationof cytoskeletal proteins following TGFb induction of EMT in ourand other studies (45–47) and provided a rationale to investigatethe relationship between cytoskeletal gene rich signatures andsurvival. A set of 41 genes derived from the mesenchymal proteinsignature representing cytoskeletal and actin-binding proteinspredicted survival in all three tumor datasets we tested. Moreover,while clearly a key feature of EMT, our study provides supportingevidence that cytoskeletal reorganization and invasiveness occurfrequently in the absence of CDH1 loss.
We further assessed the relevance of the gene and proteinsignatures associated with EMT to survival in early-stage lung
adenocarcinoma in three independent sets. Reproducibility ofstatistical association of gene expression signatures with survivalacross independent datasets has been challenging (48). Associa-tion of the full set of mRNAs in the mesenchymal signature withsurvival was significant in the Director's Challenge dataset.Remarkably, association of the more limited set of transcriptscorresponding specifically to the mesenchymal proteomic signa-ture was significant in all three independent datasets testedwithout initial training. Thus, our findings emphasize the func-tional relevance of proteomics to integrated cancer molecularprofiling, pointing to an association between cytoskeletal andactin-binding proteins, a mesenchymal or aggressive hybrid EMTphenotype and invasive properties of lung adenocarcinomas.Byers and colleagues (49) identified an EMT gene signatureconsisting of 76 genes predictive of resistance to EGFR andPI3K/AKT inhibitors which partially overlapped with our epithe-lial andmesenchymal gene/protein signatures.We note that Byersand colleagues established their signature by selecting genes withsignificant correlation (both positive and negative) with geneexpression of CDH1, VIM, CDH2, and FN1. In our study, we firstdefined epithelial and mesenchymal cell line properties based on
BSignature
Director’schallenge Bhattacharjee Tomida
mRNAEpithelial 0.257 0.153 0.187
Mesenchymal 7.62×10 0.250 0.074
ProteinEpithelial 0.734 0.162 6.70×10
Mesenchymal 1.84×10 5.65×10 5.00×10
CA
D 100
90
80
70
60
50
40
Time to death (months)
Per
cent
sur
viva
l
100
90
80
70
60
50
40
Time to death (months)
Per
cent
sur
viva
l
100
90
80
70
60
50
40
Time to death (months)
Per
cent
sur
viva
l
Bhattacharjee TomidaP = 5.58×10 P = 0.039 P = 2.20×10
0.01
0.04
0.03
0.02
0.05
0.08
0.07
0.06
0.09
0
Protein expression (log spectral counts)
Cor
rela
tion
coef
ficie
nts
Correlation coefficients
Per
cent
age
Director’s challenge
100
90
80
70
60
50
40
Time to death (months)
Per
cent
sur
viva
l
BhattacharjeeP = 0.008
0 10 20 30 40 50 60
0 10 20 30 40 50 600 10 20 30 40 50 600 10 20 30 40 50 60
100
90
80
70
60
50
40
Time to death (months)
Per
cent
sur
viva
l
Director’s challengeP = 0.024
0 10 20 30 40 50 60
100
90
80
70
60
50
40
Time to death (months)
Per
cent
sur
viva
l
TomidaP = 0.136
0 10 20 30 40 50 60
E
–2 0 2 4 6 8–0.5 0 0.5 1
1
0.5
0
–0.5
Figure 7.Mesenchymal and cytoskeletal protein signatures predict survival in NSCLC. A, Cox regression of mRNA and protein signatures in three different gene expressiondatasets. B, correlation of mRNA and protein expression in 38 cell lines. C, correlation of mRNA and protein expression increases with protein abundance.D, Kaplan–Meier curves of mesenchymal protein signature in gene expression datasets. P values are derived from the log-rank test. Black line, the top one-thirdof tumors ranked by expression of the mesenchymal signature; gray line, the lower one-third of tumors by expression of the mesenchymal signature. E,Kaplan–Meier curves of cytoskeletal protein signature in gene expression datasets. P values are derived from the log-rank test. Black line, the top one-thirdof tumors ranked by expression of the cytoskeletal signature; gray line, the lower one-third of tumors by expression of the cytoskeletal signature.
Schliekelman et al.
Cancer Res; 75(9) May 1, 2015 Cancer Research1798
on April 12, 2017. © 2015 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from
Published OnlineFirst March 5, 2015; DOI: 10.1158/0008-5472.CAN-14-2535
cell morphology and expression at the protein level of CDH1 andVIM. Our results indicate that aggressiveness among hybridtype of cell lines is not associated with expression levels ofCDH1 and VIM.
In conclusion, an integrated systems approach that encom-passed functional and molecular characterization, of lung ade-nocarcinoma cell lines, uncovered substantial heterogeneity withrespect to epithelial and mesenchymal features among cell lines.Signatures were identified that distinguish epithelial and mesen-chymal cells as well as signatures that were shared with cells withan intermediate/hybrid phenotype. Our findings point to anassociation between cytoskeletal and actin-binding proteins, amesenchymal or hybrid EMT phenotype and invasive propertiesof lung adenocarcinomas that impact survival.
Disclosure of Potential Conflicts of InterestA.F. Gazdar has received speakers bureau honoraria from Genentech. No
potential conflicts of interest were disclosed by the other authors .
Authors' ContributionsConception and design: M.J. Schliekelman, S.M. HanashDevelopment of methodology: M.J. Schliekelman, J. Zhu, M. Celiktas,A.F. Gazdar, I.I. Wistuba, S.M. HanashAcquisition of data (provided animals, acquired and managed patients,provided facilities, etc.): M.J. Schliekelman, A. Taguchi, J. Rodriguez,M. Celiktas, A. Chin, H.Wang, S.A. Selamat, E.M. Kroh, C. Behrens, A.F. Gazdar,I.A. Laird-Offringa, I.I. Wistuba, S.M. Hanash
Analysis and interpretation of data (e.g., statistical analysis, biostatistics,computational analysis): M.J. Schliekelman, A. Taguchi, J. Zhu, X. Dai,J. Rodriguez, Q. Zhang, L. McFerrin, S.A. Selamat, C. Yang, E.M. Kroh,K.S. Garg, I.A. Laird-Offringa, M. Tewari, I.I. Wistuba, J.P. Thiery, S.M. HanashWriting, review, and/or revision of the manuscript: M.J. Schliekelman,A. Taguchi, J. Zhu, K.S. Garg, A.F. Gazdar, I.A. Laird-Offringa, M. Tewari,J.P. Thiery, S.M. HanashAdministrative, technical, or material support (i.e., reporting or organizingdata, constructing databases): M.J. Schliekelman, A. Taguchi, C.-H. Wong,S.M. HanashStudy supervision: M.J. Schliekelman, S.M. Hanash
AcknowledgmentsThe authors thank members of the Hanash lab for their invaluable sugges-
tions and Paul Schliekelman for providing statistical support and advice on thearticle.
Grant SupportThis work was supported by the Department of Defense (DOD) Congres-
sionally Mandated Lung Cancer Research Program, the National Cancer Insti-tute Early Detection Program, the Canary Foundation and the LungevityFoundation. M. Tewari was supported by the Canary Foundation. L. McFerrinwas supported by NIH grant R21/R33 CA-88245 and the Listwin FamilyFoundation.
The costs of publication of this articlewere defrayed inpart by the payment ofpage charges. This article must therefore be hereby marked advertisement inaccordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Received August 28, 2014; revised December 10, 2014; accepted January 28,2015; published OnlineFirst March 5, 2015.
References1. Boyer B, Thiery JP. Epithelium-mesenchyme interconversion as example of
epithelial plasticity. APMIS 1993;101:257–68.2. Nieto MA, Sargent MG, Wilkinson DG, Cooke J. Control of cell behavior
during vertebrate development by Slug, a zinc finger gene. Science 1994;264:835–9.
3. Ciruna B, Rossant J. FGF signaling regulates mesoderm cell fate specifica-tion and morphogenetic movement at the primitive streak. Dev Cell2001;1:37–49.
4. Thiery JP, Acloque H, Huang RY, Nieto MA. Epithelial-mesenchymaltransitions in development and disease. Cell 2009;139:871–90.
5. Kim SH, Kim JM, Shin MH, Kim CW, Huang SM, Kang DW, et al.Correlation of epithelial-mesenchymal transition markers with clinico-pathologic parameters in adenocarcinomas and squamous cell carcinomaof the lung. Histol Histopathol 2012;27:581–91.
6. 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.
7. Choi YS, Shim YM, Kim SH, Son DS, Lee HS, Kim GY, et al. Prognosticsignificance of E-cadherin and beta-catenin in resected stage I non-smallcell lung cancer. Eur J Cardiothorac Surg 2003;24:441–9.
8. Thiery JP. Epithelial-mesenchymal transitions in tumour progression. NatRev Cancer 2002;2:442–54.
9. Spaderna S, Schmalhofer O, Wahlbuhl M, Dimmler A, Bauer K, Sultan A,et al. The transcriptional repressor ZEB1 promotes metastasis and loss ofcell polarity in cancer. Cancer Res 2008;68:537–44.
10. Nawshad A, Lagamba D, Polad A, Hay ED. Transforming growth factor-beta signaling during epithelial-mesenchymal transformation: implica-tions for embryogenesis and tumor metastasis. Cells Tissues Organs2005;179:11–23.
11. Nelson WJ, Nusse R. Convergence of Wnt, beta-catenin, and cadherinpathways. Science 2004;303:1483–7.
12. Guarino M, Rubino B, Ballabio G. The role of epithelial-mesenchymaltransition in cancer pathology. Pathology 2007;39:305–18.
13. Kalluri R, Weinberg RA. The basics of epithelial-mesenchymal transition. JClin Invest 2009;119:1420–8.
14. Park SM,Gaur AB, Lengyel E, PeterME. ThemiR-200 family determines theepithelial phenotype of cancer cells by targeting the E-cadherin repressorsZEB1 and ZEB2. Genes Dev 2008;22:894–907.
15. CottonhamCL, Kaneko S, Xu L.miR-21 andmiR-31 converge on TIAM1 toregulate migration and invasion of colon carcinoma cells. J Biol Chem2010;285:35293–302.
16. Kim NH, Kim HS, Li XY, Lee I, Choi HS, Kang SE, et al. A p53/miRNA-34axis regulates Snail1-dependent cancer cell epithelial-mesenchymal tran-sition. J Cell Biol 2011;195:417–33.
17. Walter K, Holcomb T, Januario T, Du P, Evangelista M, Kartha N,et al. DNA methylation profiling defines clinically relevant biologi-cal subsets of non-small cell lung cancer. Clin Cancer Res 2012;18:2360–73.
18. Davalos V, Moutinho C, Villanueva A, Boque R, Silva P, Carneiro F, et al.Dynamic epigenetic regulation of the microRNA-200 family mediatesepithelial and mesenchymal transitions in human tumorigenesis. Onco-gene 2012;31:2062–74.
19. Neves R, Scheel C, Weinhold S, Honisch E, Iwaniuk KM, Trompeter HI,et al. Role of DNA methylation in miR-200c/141 cluster silencing ininvasive breast cancer cells. BMC Res Notes 2010;3:219.
20. McDonald OG, Wu H, Timp W, Doi A, Feinberg AP. Genome-scaleepigenetic reprogramming during epithelial-to-mesenchymal transition.Nat Struct Mol Biol 2011;18:867–74.
21. Peinado H, Olmeda D, Cano A. Snail, Zeb and bHLH factors in tumourprogression: an alliance against the epithelial phenotype? Nat Rev Cancer2007;7:415–28.
22. 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.
23. Aigner K, Dampier B, Descovich L, Mikula M, Sultan A, Schreiber M, et al.The transcription factor ZEB1 (deltaEF1) promotes tumour cell dediffer-entiation by repressing master regulators of epithelial polarity. Oncogene2007;26:6979–88.
24. Tsai JH, Donaher JL, Murphy DA, Chau S, Yang J. Spatiotemporal regu-lation of epithelial-mesenchymal transition is essential for squamous cellcarcinoma metastasis. Cancer Cell 2012;22:725–36.
www.aacrjournals.org Cancer Res; 75(9) May 1, 2015 1799
Molecular Signatures of EMT Status in Lung Adenocarcinoma
on April 12, 2017. © 2015 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from
Published OnlineFirst March 5, 2015; DOI: 10.1158/0008-5472.CAN-14-2535
25. Lu M, Jolly MK, Levine H, Onuchic JN, Ben-Jacob E. MicroRNA-basedregulationof epithelial-hybrid-mesenchymal fate determination. ProcNatlAcad Sci U S A 2013;110:18144–9.
26. Ong SE, Mann M. A practical recipe for stable isotope labeling by aminoacids in cell culture (SILAC). Nat Protoc 2006;1:2650–60.
27. Taguchi A, Politi K, Pitteri SJ, LockwoodWW, Faca VM, Kelly-Spratt K, et al.Lung cancer signatures in plasma based on proteome profiling of mousetumor models. Cancer Cell 2011;20:289–99.
28. Gemmill RM, Roche J, Potiron VA, Nasarre P, Mitas M, Coldren CD, et al.ZEB1-responsive genes in non-small cell lung cancer. Cancer Lett 2011;300:66–78.
29. Deng QF, Zhou CC, Su CX. Clinicopathological features and epidermalgrowth factor receptor mutations associated with epithelial-mesenchymaltransition in non-small cell lung cancer. Respirology 2009;14:371–6.
30. SubramanianA, TamayoP,Mootha VK,Mukherjee S, Ebert BL,GilletteMA,et al. Gene set enrichment analysis: a knowledge-based approach forinterpreting genome-wide expression profiles. Proc Natl Acad Sci U S A2005;102:15545–50.
31. Akakura S, Gelman IH. Pivotal role of AKAP12 in the regulation of cellularadhesiondynamics: control of cytoskeletal architecture, cellmigration, andmitogenic signaling. J Signal Transduct 2012;2012:529179.
32. Director's ChallengeConsortium for theMolecular Classificationof LungAShedden K, Taylor JM, Enkemann SA, Tsao MS, Yeatman TJ, et al. Geneexpression-based survival prediction in lung adenocarcinoma: amulti-site,blinded validation study. Nat Med 2008;14:822–7.
33. Bhattacharjee A, Richards WG, Staunton J, Li C, Monti S, Vasa P, et al.Classification of human lung carcinomas by mRNA expression profilingreveals distinct adenocarcinoma subclasses. Proc Natl Acad Sci U S A2001;98:13790–5.
34. Tomida S, Takeuchi T, Shimada Y, Arima C, Matsuo K, Mitsudomi T, et al.Relapse-related molecular signature in lung adenocarcinomas identifiespatients with dismal prognosis. J Clin Oncol 2009;27:2793–9.
35. Zhang J, Grubor V, Love CL, Banerjee A, Richards KL,Mieczkowski PA, et al.Genetic heterogeneity of diffuse large B-cell lymphoma. Proc Natl Acad SciU S A 2013;110:1398–403.
36. ShipitsinM,Campbell LL, Argani P,Weremowicz S, Bloushtain-QimronN,Yao J, et al. Molecular definition of breast tumor heterogeneity. Cancer Cell2007;11:259–73.
37. Budinska E, Popovici V, Tejpar S, D'Ario G, Lapique N, Sikora KO, et al.Gene expression patterns unveil a new level of molecular heterogeneity incolorectal cancer. J Pathol 2013;231:63–76.
38. Fidler IJ. Biological heterogeneity of cancer: implication to therapy. HumVaccin Immunother 2012;8:1141–2.
39. Johnen N, Francart ME, Thelen N, Cloes M, Thiry M. Evidence for a partialepithelial-mesenchymal transition in postnatal stages of rat auditoryorgan morphogenesis. Histochem Cell Biol 2012;138:477–88.
40. Debnath J, Brugge JS. Modelling glandular epithelial cancers in three-dimensional cultures. Nat Rev Cancer 2005;5:675–88.
41. Nanba D, Nakanishi Y, Hieda Y. Changes in adhesive properties ofepithelial cells during early morphogenesis of the mammary gland. DevGrowth Differ 2001;43:535–44.
42. Huber MA, Kraut N, Beug H. Molecular requirements for epithelial-mes-enchymal transition during tumor progression. Curr Opin Cell Biol2005;17:548–58.
43. La Fleur L, Johansson AC, Roberg K. A CD44high/EGFRlow subpopulationwithin head and neck cancer cell lines shows an epithelial-mesenchymaltransition phenotype and resistance to treatment. PLoS ONE 2012;7:e44071.
44. Rasanen K, Sriswasdi S, Valiga A, Tang HY, Zhang G, Perego M, et al.Comparative secretome analysis of epithelial and mesenchymal subpopula-tions of head and neck squamous cell carcinoma identifies S100A4 as apotential therapeutic target. Mol Cell Proteomics 2013;12:3778–92.
45. Keshamouni VG, Schiemann WP. Epithelial-mesenchymal transitionin tumor metastasis: a method to the madness. Future Oncol 2009;5:1109–11.
46. Yilmaz M, Christofori G. EMT, the cytoskeleton, and cancer cell invasion.Cancer Metastasis Rev 2009;28:15–33.
47. Levayer R, Lecuit T. Breaking down EMT. Nat Cell Biol 2008;10:757–9.48. Subramanian J, Simon R. Gene expression-based prognostic signatures
in lung cancer: ready for clinical use? J Natl Cancer Inst 2010;102:464–74.
49. Byers LA, Diao L, Wang J, Saintigny P, Girard L, Peyton M, et al. Anepithelial-mesenchymal transition gene signature predicts resistance toEGFR and PI3K inhibitors and identifies Axl as a therapeutic target forovercoming EGFR inhibitor resistance. Clin Cancer Res 2013;19:279–90.
Cancer Res; 75(9) May 1, 2015 Cancer Research1800
Schliekelman et al.
on April 12, 2017. © 2015 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from
Published OnlineFirst March 5, 2015; DOI: 10.1158/0008-5472.CAN-14-2535
2015;75:1789-1800. Published OnlineFirst March 5, 2015.Cancer Res Mark J. Schliekelman, Ayumu Taguchi, Jun Zhu, et al. in Lung Adenocarcinoma and Their Relevance to SurvivalMolecular Portraits of Epithelial, Mesenchymal, and Hybrid States
Updated version
10.1158/0008-5472.CAN-14-2535doi:
Access the most recent version of this article at:
Material
Supplementary
http://cancerres.aacrjournals.org/content/suppl/2015/03/07/0008-5472.CAN-14-2535.DC1
Access the most recent supplemental material at:
Cited articles
http://cancerres.aacrjournals.org/content/75/9/1789.full.html#ref-list-1
This article cites 49 articles, 16 of which you can access for free at:
Citing articles
/content/75/9/1789.full.html#related-urls
This article has been cited by 11 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
To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at
Permissions
To request permission to re-use all or part of this article, contact the AACR Publications Department at
on April 12, 2017. © 2015 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from
Published OnlineFirst March 5, 2015; DOI: 10.1158/0008-5472.CAN-14-2535