1、Articlehttps:/doi.org/10.1038/s41586-018-0040-3Identification of the tumour transition states occurring during EMTievgenia Pastushenko1,Audrey Brisebarre1,Alejandro Sifrim2,3,Marco Fioramonti1,tatiana revenco1,Soufiane Boumahdi1,Alexandra Van Keymeulen1,Daniel Brown2,4,Virginie Moers1,Sophie lemaire
2、1,Sarah De clercq5,esmeralda Minguijn5,cdric Balsat6,Youri Sokolow7,christine Dubois1,Florian De cock1,Samuel Scozzaro1,Federico Sopena8,Angel lanas9,Nicky DHaene5,isabelle Salmon5,6,Jean-christophe Marine4,10,thierry Voet2,3,Panagiota A.Sotiropoulou1,12&cdric Blanpain1,11,12*In cancer,the epithelia
3、l-to-mesenchymal transition(EMT)is associated with tumour stemness,metastasis and resistance to therapy.It has recently been proposed that,rather than being a binary process,EMT occurs through distinct intermediate states.However,there is no direct in vivo evidence for this idea.Here we screen a lar
4、ge panel of cell surface markers in skin and mammary primary tumours,and identify the existence of multiple tumour subpopulations associated with different EMT stages:from epithelial to completely mesenchymal states,passing through intermediate hybrid states.Although all EMT subpopulations presented
5、 similar tumour-propagating cell capacity,they displayed differences in cellular plasticity,invasiveness and metastatic potential.Their transcriptional and epigenetic landscapes identify the underlying gene regulatory networks,transcription factors and signalling pathways that control these differen
6、t EMT transition states.Finally,these tumour subpopulations are localized in different niches that differentially regulate EMT transition states.EMT is a cellular process in which cells lose their epithelial characteris-tics and acquire mesenchymal features,which enable them to migrate more efficien
7、tly and invade the underlying mesenchyme.EMT is essential for gastrulation,somitogenesis and neural crest delamination during embryonic development and has been associated with various diseases.In cancer,EMT is associated with tumorigenesis,invasion,metastasis,tumour stemness and resistance to thera
8、py1,2.Although EMT has traditionally been viewed as a binary switch,some in vitro data(mainly co-expression of epithelial and mesenchymal markers within the same cells)have indicated that EMT may proceed in a step-wise manner through the generation of subpopulations that represent different intermed
9、iate states between the epithelial and mesenchymal states37.However,it remains unclear whether EMT proceeds through these intermediate states in vivo,and if so how many intermediate steps exist,how plastic and reversible these intermediate states are,which mechanisms regulate the transition from one
10、 state to another and what the implications of these different EMT states are for tumour progres-sion,stemness and metastasis1,2.Different tumour EMT transition statesTo determine whether EMT in vivo occurs through a succession of different intermediate states,we used a genetic mouse model of skin s
11、quamous cell carcinoma(SCC)mediated by the conditional expres-sion of KRasG12D and p53 deletion(p53KO)in hair follicles.This model generates skin tumours that undergo spontaneous EMT,containing epithelial YFP+Epcam+and mesenchymal-like YFP+Epcam tumour cells(TCs)8.Using flow cytometry(fluorescence-a
12、ctivated cell sort-ing,FACS),we screened cells from these tumours for a large panel of cell surface markers and assessed whether these markers were heterogeneously expressed in YFP+Epcam+or YFP+Epcam popu-lations(Fig.1a).The YFP+Epcam+TCs were relatively homogenous,with only four markers being heter
13、ogeneously expressed(Fig.1b).By contrast,half of the markers were heterogeneously expressed in YFP+Epcam TCs(Fig.1b,Extended Data Fig.1a,b),suggesting that EMT is associated with important cellular heterogeneity.The mark-ers that were most frequently heterogeneously expressed during EMT included CD6
14、1(also known as Itgb3),CD51(also known as Itgav)and CD106(also known as Vcam1;Extended Data Fig.1c),which mark subpopulations of TCs associated with tumour stemness,EMT or metastasis initiation in other tumour models7,912.Other markers were not as frequently heterogeneously expressed when analysed i
15、n a larger cohort of tumours(Extended Data Fig.1c).Combinatorial multicolour FACS analysis revealed that CD106,CD61,and CD51 discriminated six distinct populations within YFP+Epcam TCs in most(75%)mixed tumours(Fig.1c,e,h,Extended Data Fig.2).About 10%of mixed tumours with a high proportion of Epcam
16、+cells presented only triple-negative(Epcam CD51CD61CD106)and Epcam CD51+CD61+populations,whereas highly mesenchymal tumours with only Epcam TCs contained CD51+,CD51+CD61+,CD106+CD51+and CD106+CD51+CD61+triple-positive tumour subpopulations with almost no triple-negative and CD106+subpopulations(Fig
17、.1cg).To define whether these different tumour populations correspond to distinct EMT transition states,we isolated the subpopulations by FACS and performed immunostaining on cytospin with epithelial(keratin 14,K14)and mesenchymal(vimentin)markers.Notably,loss of Epcam expression coincided with a ga
18、in in vimentin expression in all TCs,con-sistent with the first switch to the mesenchymal state(Fig.1i,j).However,1Laboratory of Stem Cells and Cancer,Universit Libre de Buxelles,Brussels,Belgium.2Department of Human Genetics,University of Leuven,Leuven,Belgium.3Sanger Institute-EBI Single-Cell Geno
19、mics Centre,Wellcome Trust Sanger Institute,Hinxton,UK.4Laboratory for Molecular Cancer Biology,VIB Center for Cancer Biology,Leuven,Belgium.5Pathology Department,Erasme University Hospital,Universit Libre de Bruxelles,Brussels,Belgium.6DIAPath,The Center for Microscopy and Molecular Imaging,Univers
20、it Libre de Bruxelles,Gosselies,Belgium.7Thoracic Surgery,Erasme University Hospital,Universit Libre de Bruxelles,Brussels,Belgium.8Gastroenterology Department,Hospital Clnico Universitario“Lozano Blesa”,IIS Aragon,Zaragoza,Spain.9CIBERehd,IIS Aragn,University of Zaragoza,Zaragoza,Spain.10Laboratory
21、 for Molecular Cancer Biology,Department of Oncology,KU Leuven,Leuven,Belgium.11WELBIO,Universit Libre de Bruxelles,Bruxelles,Belgium.12These authors jointly supervised this work:Panagiota A.Sotiropoulou,Cdric Blanpain.*e-mail:cedric.blanpainulb.ac.be2 6 A P r i l 2 0 1 8|V O l 5 5 6|N A t U r e|4 6
22、 3 2018 Macmillan Publishers Limited,part of Springer Nature.All rights reserved.ArticlereSeArcHsome Epcam tumour subpopulations(triple-negative,CD106+and CD51+)continued to express K14 and vimentin,whereas other Epcam subpopulations(CD51+CD61+and triple-positive)were essentially K14vimentin+,with r
23、are TCs expressing low levels of K14(Fig.1i,j).These data indicate that the subpopulations identified during sponta-neous EMT of primary skin tumours correspond to different tumour populations with different degrees of EMT,with some subpopulations corresponding to the hybrid tumour phenotypes with e
24、pithelial and mes-enchymal features that have been described in cancer cell lines in vitro37.To further assess cellular heterogeneity during spontaneous EMT,we performed single cell RNA sequencing(scRNA-seq)of FACS-isolated YFP+Epcam+and YFP+Epcam TCs.Dimensionality reduction using principal compone
25、nt analysis(PCA)revealed that the first principal component(PC1),which explained 21%of the variability,could be attributed to EMT state(Fig.1k).Our scRNA-seq data confirmed that Epcam subpopulations showed greater transcriptional heterogeneity than Epcam+subpopulations(Fig.1k).The expression of epit
26、helial and mesenchymal markers at the single-cell level confirmed the pro-gressive acquisition of EMT features with epithelial,mesenchymal and hybrid states(Extended Data Fig.3)and the distribution of CD51,CD61 and CD106 markers correlated with the degree of EMT along PC1,confirming the EMT gradient
27、 found across the different tumour subpopulations(Fig.1ln).To assess whether these different subpopulations of EMT TCs reflect a more general mechanism occurring during EMT,we assessed the expression of these cell surface markers in metaplastic-like mammary tumours arising from oncogenic Pik3ca expr
28、ession and p53 deletion and in MMTV-PyMT mammary luminal tumours,which have been reported to present EMT features1316.Notably,a subset of mammary metaplastic-like and MMTV-PyMT luminal tumours also contained Epcam+and Epcam TCs that could be subdivided into the same six subpopulations as found in KR
29、asG12D/p53KO skin tumours(Extended Data Figs.4ad,5ac).Immunostaining on cytospin and real-time PCR with reverse transcriptase(RTPCR)showed that the subpopulations isolated from the mammary tumours presented different degrees of EMT,similar to those identified in skin SCCs(Extended Data Figs.4e,f,h,5
30、d,e,g),demonstrating that the different EMT transition states identified here represent a conserved mechanism during EMT.To investigate whether these EMT transition states exist in human cancers,we assessed the expression of epithelial and mesenchymal markers in tumours derived from xenotransplantat
31、ion(PDX)of poorly differentiated human breast cancers and SCCs.After several passages in immunodeficient mice,human stroma is entirely replaced by mouse cells17,making it possible to differentiate human TCs that underwent EMT and lost the expression of epithelial markers from mouse stroma using an a
32、ntibody against human antigen(Ku-80).We detected areas expressing only epithelial markers,areas co-expressing epithelial and mesenchymal markers and areas expressing exclusively mesenchymal markers in poorly differentiated breast cancer,lung and oesophageal SCCs(Extended Data Fig.6).These data demon
33、strate that EMT in human cancers is associated with different transition states including hybrid states,as was suggested by scRNA-seq of human SCCs18.Stemness and plasticity of EMT statesEMT has been associated with cancer stemness,characterized by an increase in tumour-propagating cell(TPC)frequenc
34、y1,8,19,20.As pre-viously described8,Epcam TCs contained five times as many TPCs as did Epcam+TCs(Fig.2a).Notably,all EMT subpopulations presented similar TPC frequencies(Fig.2a).These data show that the earliest EMT state already exhibits increased TPC frequency,and tumour stemness does not increas
35、e further in later transition states.Whereas Epcam+TCs showed higher proliferation than Epcam TCs,there was no difference in proliferation rate between the different EMT subpop-ulations(Fig.2b).Thus,TPC frequency is inversely correlated to in vivo proliferation.Cancer cells have been shown to be pla
36、stic in transplantation assays,with different tumour subpopulations able to transit back and forth between different states and to recapitulate primary tumour heteroge-neity21.To determine whether the different subpopulations identified during EMT are similarly plastic,we analysed the tumour phenoty
37、pes of secondary tumours.As previously described8,secondary tumours arising from the transplantation of Epcam TCs comprise only Epcam TCs(Fig.2c).Although all EMT subpopulations presented a certain degree of plasticity,at early time points following transplantation (34 weeks),the triple-negative sub
38、population was relatively primed towards the epithelial phenotype and preferentially gave rise to jK14Vim+K14+Vim+K14+VimK14VimPer cent of YFP+TCsi020406080100aSingle-cell suspensionScreening 176 cell surface markersAnalysis of marker expressionFACS analysis YFP+Epcam+and YFP+Epcam TCsbNo.of surface
39、 markers expressedK14 VimVimK14Epcam+TNTPNegativeHomogeneousHeterogeneousPattern of expressionEpcamP=0.0001EpithelialMesen-chymaldGated on YFP+EpcamTCsPer cent of Epcam+TCscPattern 43%Pattern 312%efgh1 2 3 4020406080100Per cent of Epcam TCs0102030405060708090Per cent of Epcam TCs0102030405060708090E
40、pcam+CD106+CD51+CD106+CD51+CD51+CD61+TPTNPattern 175%Per cent of Epcam TCs0102030405060708090CD106+CD51+CD106+CD51+CD51+CD61+TPTNCD106+CD51+CD106+CD51+CD51+CD61+TPTNCD106+CD51+CD106+CD51+CD51+CD61+TPTNCD106+CD51+CD106+CD51+CD51+CD61+TPTNPer cent of Epcam TCsPattern 210%0102030405060708090Epcam+Epcam
41、HomogeneousHeterogeneous10 0 10 2010010Cdh11001010 0 10 20100 10 201001010 0 10 201010010 0 10 201010010 0 10 2010100PC2PC1PC1PC1PC1PC1PC10 1230 1 2 30 1 2 30 1 2 30 1 2 30123EpcamKrt14VimCD106CD61k20100100.251.000.750.500Cumulative densityPC110 0 10 2010010PC1PC2lnmCD51PECD106BV786TNCD106+TPCD106+C
42、D51+CD51+CD61+CD51+Pattern subpopulationdistributionCD61BV421CD61BV421CD61BV421CD61BV4210510152025301017426n=3n=6n=6n=6n=7n=8n=7Epcam+TNCD51+CD106+CD51+CD51+CD61+TPEpcam+TNCD51+CD106+CD51+CD51+CD61+TPEpcam+TNCD51+CD106+CD51+CD51+CD61+TPCD61+CD106+110621632650122Fig.1|Identification of the tumour tra
43、nsition states occurring during EMT in vivo.a,Strategy for marker screening in hair follicle-derived skin SCC model presenting spontaneous EMT.b,Percentage of homogeneously and heterogeneously expressed markers in YFP+Epcam+and YFP+Epcam TCs.Two-tailed Fishers exact test.c,Percentage of Epcam+cells
44、in the four groups of tumours that differ by the frequencies of the different subpopulations.dg,Distributions of subpopulations in tumours with pattern 1(d;mixed tumours containing Epcam+and Epcam TCs),pattern 2(e;differentiated tumours with a high percentage of Epcam+TCs),pattern 3(f)and pattern 4(
45、g;mesenchymal Epcam TCs)(n=12 mice,mean s.e.m.).TN,triple-negative;TP,triple-positive.h,FACS profile showing the six CD106/CD51/CD61 subpopulations in the most frequent pattern(pattern 1).i,Co-immunostaining of keratin 14(K14)and vimentin(Vim)in cystospin of FACS-isolated tumour cells.Scale bars,20
46、m;n=6.j,Cytospin counts of TCs based on K14 and vimentin expression(average of 90 cells per condition and tumour).k,PCA of scRNA-seq of Epcam+and Epcam TCs.Dots represent single cells;colour scale represents the normalized expression of each gene(n=66 Epcam+and n=277 Epcam cells from one tumour).l,P
47、roportion of single cells expressing Epcam,CD51,CD61 and CD106 markers.m,PCA plot coloured by expression of the markers used to define EMT subpopulations by FACS.Grey circle,epithelial tumour cells;blue,hybrid cells;yellow,early EMT cells;red,late EMT cells.n,Cumulative density of each marker combin
48、ation across PC1,which correspond to the degree of EMT(j).4 6 4|N A t U r e|V O l 5 5 6|2 6 A P r i l 2 0 1 8 2018 Macmillan Publishers Limited,part of Springer Nature.All rights reserved.Article reSeArcHtriple-negative TCs.By contrast,the most mesenchymal triple-positive subpopulation preferentiall
49、y gave rise to triple-positive TCs(Fig.2ce).The other EMT subpopulations were plastic,giving rise to each other in similar proportions(Fig.2c).Notably,at later time points(78 weeks),the triple-negative subpopulation gave rise to different EMT subpopulations,whereas the triple-positive subpopulation
50、gave rise preferentially to triple-positive cells(Fig.2f).The different tumour subpopulations had different invasive capacities,which increased with the degree of EMT(Fig.2f).Together,these data show that the different EMT tumour subpopulations are functionally distinct and have differ-ent clonogeni
