E X P E R I M E N T A L C E L L R E S E A R C H 3 1 4 ( 2 0 0 8 ) 6 9 1 – 6 9 8

ava i l ab l e a t www.sc i enced i rec t . com

www.e l sev i e r. com/ loca te /yexc r

Research Article

Human mesenchymal stem cell transformation is associatedwith a mesenchymal–epithelial transition

Daniel Rubioa, Silvia Garciab, Teresa De la Cuevac, Ma F. Pazb,Alison C. Lloydd, Antonio Bernadb,⁎, Javier Garcia-Castroe

aCentro de Biología Molecular “Severo Ochoa”, Nicolás Cabrera 1, UAM Campus de Cantoblanco, E-28049 Madrid, SpainbDepartment of Regenerative Cardiology, CNIC, Melchor Fernández Almagro 3, E-28029 Madrid, SpaincAnimalary Facility, Centro Nacional de Biotecnología/CSIC, Darwin 3, UAM Campus de Cantoblanco, E-28049 Madrid, SpaindUniversity College London, Gower Street, WC1E 6BT London, UKeSpanish Stem Cell Bank (Andalusian branch), Centro de Investigación Biomédica, Avda. del Conocimiento s/n, E-18100 Granada, Spain

A R T I C L E I N F O R M A T I O N

⁎ Corresponding author. Fax: +34 91 453 240.E-mail address: abernad@cnic.es (A. BernAbbreviations: MSC, Mesenchymal stem c

Epithelial–mesenchymal transition

0014-4827/$ – see front matter © 2007 Elsevidoi:10.1016/j.yexcr.2007.11.017

A B S T R A C T

Article Chronology:Received 8 May 2007Revised version received7 November 2007Accepted 8 November 2007Available online 16 January 2008

Carcinomasarewidely thought toderive fromepithelial cellswithmalignantprogressionoftenassociated with an epithelial–mesenchymal transition (EMT). We have characterized tumorsgenerated by spontaneously transformed human mesenchymal cells (TMC) previouslyobtained in our laboratory. Immunohistopathological analyses identified these tumors aspoorly differentiated carcinomas, suggesting that a mesenchymal–epithelial transition (MET)was involved in the generation of TMC. This was corroborated by microarray and proteinexpression analysis that showed that almost all mesenchymal-related genes were severelyrepressed in these TMC. Interestingly, TMC also expressed embryonic antigens and were ableto integrate into developing blastocysts with no signs of tumor formation, suggesting adedifferentiation process was associated with the mesenchymal stem cell (MSC)transformation. These findings support the hypothesis that some carcinomas are derivedfrom mesenchymal rather than from epithelial precursors.

© 2007 Elsevier Inc. All rights reserved.

Keywords:TransformationTumorMesenchymal stem cell

Introduction

A hierarchical model is associated with stem cell differentia-tion. In this model, totipotent stem cells differentiate tomultipotent stem cells, which in turn can give rise topluripotent stem cells, which generate committed precursorsthat differentiate terminally to mature cells [1]. Some groupshave nonetheless proposed modifications to this strict hier-archy, with processes such as cell dedifferentiation, by whicha cell could generate a precursor with greater potential than

ad).ells; TMC, Transformed m

er Inc. All rights reserved

the cell from which it derives [2], or cell transdifferentiation,that allows cell lineage switch [3].

The epithelial–mesenchymal transition (EMT) is a physio-logical mechanism present during development, and is alsoencountered in several pathological situations such as renalinterstitial fibrosis, endometrial adhesion, and cancer metas-tasis [4]. Some primary carcinoma cells undergo EMT; thesemesenchymal-like cells then secrete proteases that promotedegradation of extracellular matrix and acquire movementcapacity, allowing escape from the primary tumor, migration

esenchymal cells; MET, Mesenchymal–epithelial transition; EMT,

.

692 E X P E R I M E N T A L C E L L R E S E A R C H 3 1 4 ( 2 0 0 8 ) 6 9 1 – 6 9 8

to the bloodstream or lymph, and extravasation to generatemetastases [5–7]. These metastases, nonetheless, have anepithelial phenotype, suggesting reversibility of EMT andindicative of a mesenchymal–epithelial transition (MET). METalso takesplaceduringnormal development, inprocesses suchas somitogenesis, kidney development, and coelomic cavityformation [8–10], but to date has not been associated withtumors derived from mesenchymal cells.

Cancer stem cells (CSC) have been reported for some tumortypes, including breast and lung cancer, leukemia and glioblas-toma [11]. Adult stem cells and CSC share several featuresincluding self-renewal ability, asymmetric division and differ-entiation potential [12]. Altered stem cells are thought to giverise to CSC in certain cases [13], and we recently described an invitromodel of spontaneous CSC generation frommesenchymalstem cells [14]. A recent report has suggested that both EMT andMETmay be involved in themalignant progression of epithelialtumors. In this model, CSC in the primary epithelial tumor,undergo EMT to give rise to migrating CSC, which reconvert toa stationary epithelial-like CSC after an MET at the site ofmetastasis [15].

Here we show that spontaneously transformed MSC wheninjected into mice gave rise to primitive carcinomas with anepithelioid phenotype. These tumor cells down-regulatedalmost all mesenchymal-related transcripts and repressedvimentin, a mesenchymal protein, whereas they expressedepithelial antigens such as cytokeratins. TMC cells were ableto integrate into developing blastocysts and contributed toplacental tissue. These results show that spontaneous MSCtransformation involves a mesenchymal–epithelial transitionassociated with a dedifferentiation process.

Materials and methods

Isolation of human mesenchymal stem cell and cell culture.Human mesenchymal stem cell isolation and culture was de-scribed previously [14]. Briefly, samples of adipose tissue wereminced and digested with 1 mg collagenase P (Roche)/0.5 gsample/ml DMEM (37 °C, 1 h). Samples were clarified bysedimentation, the resulting cell suspension filtered througha 40-mm2 nylon filter (Becton Dickinson), plated onto tissueculture plastic (103 cells/cm2), allowed to adhere during 48 hand washed twice with phosphate buffer. Adipose tissue-derived MSC from C57BL/6Tg15(act-EGFP)Osb1 strain micewere isolated using the same method. Adherent cells werecultured at 37 °C and 5% CO2 in MSCmedium (DMEM plus 10%FCS, 2 mM glutamine, 50 μg/ml gentamycin). Transformedmesenchymal cells (TMC) were cultured in the same condi-tions. Humansampleswere obtainedaccording to Spanish andEuropean guidelines.

In vivo tumorigenesis

Eight-week-old male BALB/cJHanHsd-Prkdcscid (Harlan) micewere infused subcutaneously with 107 cells of EGFP-TMC. Micewere killed when tumors reached approximately 2 cm3.Tumors were removed and all organs were analyzed underUV light to detect EGFP-positive cells or fixed for immunohis-tochemistry analysis. Mice were maintained under high stan-

dards conditions in accordance with FELASA (Federation ofEuropean Laboratory Animal Science Associations) proce-dures. All the experiments were conducted according to theSpanish and European regulation about the use and treatmentof experimental animals.

Transmission electron microscopy

Semi- and ultra-thin sections were analyzed at the ElectronMicroscopy Core Facility (Universidad Complutense, Madrid,Spain). Images were captured using a JEOL microscope (JEM-2000 FX).

Microarray labeling

Total RNA was isolated from four biological replicates of pre-and post-senescence MSC and from TMC using the TriR-eagent Solution (Sigma). RNA was purified with MegaClear(Ambion), and integrity confirmed on an Agilent 2100Bioanalyzer (Agilent Technologies). Total RNA (1.5 μg/sam-ple) was amplified using the Amino Allyl MessageAmp aRNAkit (Ambion) to obtain 15–60 μg of amino-allyl amplified RNA(aRNA); meanaRNAsizewas 1500nucleotides. For each sample,aRNA (2.5 μg) was labeled with one aliquot of Cy3 or Cy5 MonoNHS Ester (CyDye Post-labeling Reactive Dye Pack, Amersham)and purified using the Amino Allyl MessageAmp aRNA kit. Cy3and Cy5 incorporation were measured using 1 μl probe in aNanodrop spectrophotometer (Nanodrop Technologies). Foreach hybridization, 80–100 pmol each of Cy3 and Cy5 probeswere mixed, dried by speed-vacuum, and resuspended in 9 μlRNase-free water. Labeled aRNA was fragmented by adding 1 μlof 10× fragmentation buffer (Ambion) and incubation (70 °C,15 min). The reaction was terminated with 1 μl stop solution(Ambion).

Slide treatment and hybridization

Slides containing 22,102 annotated genes corresponding tothe human 70-mer oligonucleotide library (V2.2) (Qiagen-Operon) were obtained from the Genomics and MicroarrayLaboratory (Cincinnati University). Information on printingand the oligo set are found at http://microarray.uc.edu.Slides were prehybridized (42 °C, 45–60 min) in 6× salinesodium citrate buffer (SSC), 0.5% sodium dodecyl sulfate(SDS) and 1% bovine serum albumin (BSA), then rinsed 10times with distilled water. Fragmented Cy3 and Cy5 aRNAprobes were mixed (80–100 pmol each) with 10 μg PolyA(Sigma) and 5 μg human Cot-DNA (Invitrogen), and dried ina speed-vacuum. Each probe mixture was then resuspendedto a final volume of 60 μl in hybridization buffer (50%formamide, 6× SSC, 0.5% SDS, 5× Denhardt's solution).Probes were denatured (95 °C, 5 min) and applied to slides,then incubated (48 °C, 16 h) in hybridization chambers(Array-It; Telechem International) in a water bath. Afterincubation, slides were washed twice with 0.5× SSC, 0.1%SDS (5 min each), three times with 0.5× SSC (5 min), andonce in 0.05× SSC (5 min), then dried by centrifugation(563×g, 1 min). Images from Cy3 and Cy5 channels wereequilibrated and captured with an Axon 4000B scanner, andspots quantified using GenePix 5.1 software.

Fig. 1 – Metastasis generation by EGFP-TMC. (a) Number andpercentage of BALB/c/SCID mice with primary tumors ormetastases affecting lung tissue after subcutaneous (s.c.)EGFP-TMC injection. (b,c) Fluorescent images (b) andhematoxylin/eosin staining (c, 40×) of lung infiltrated withTMC-derived metastases.

Fig. 2 – Transmission electron microscopy image of MSCand TMC. MSC (a) and TMC (b) transmission electronmicrophotographies at 2000× (left) and 10,000× (right).

693E X P E R I M E N T A L C E L L R E S E A R C H 3 1 4 ( 2 0 0 8 ) 6 9 1 – 6 9 8

For each experiment, we studied dye-swapped replicatesfor each of four passages (8 hybridizations). Data for replicateswere analyzed using Almazen software, which applied lowessnormalization to each replicate, and merged the log ratioswith the corresponding standard deviation and z-score.Adjusted p values were obtained using Limma (BioConductor;http://www.bioconductor.org) [16]. Differentially expressedgenes were selected by filtering signal intensity (N64), z score(N3.5 or b−3.5) and p value (b0.01).

Quantitative real-time PCR (qRT-PCR)

cDNA was generated from 100 ng of total RNA in a 10-μl finalreaction volume, using the High Capacity cDNA Archive Kit(Applied Biosystems). Real-time PCR reactions were per-formed in triplicate, using 3 μl/well for each dilution (1/50and 1/500) of each cDNA, 0.4 μMof 28s primers (28s forward, 5′-TGCCATG GTAATCCTGCTCA-3′; 28s reverse, 5′-CCTCAGC-CAAGCACATACACC-3′) or 1× TaqMan Assay-On-DemandHs00195591_m1 (Snai1), or Hs00161904_m1 (Snai2), with 1×SYBR Green PCR Master Mix or 1× TaqMan Universal PCRMaster Mix (Applied Biosystems), in a volume of 8 μl in 384-well optical plates or using Universal ProbeLibrary (Vimentin)(Roche). PCR reactions were run in an ABI PRISM 7900HT(Applied Biosystems) and SDS v2.2 software was used toanalyze results with the Comparative Ct Method (ΔΔCt).

In vitro transforming growth factor beta quantification

MSC and TMC cells were plated at 12.5×103 cells/cm2 and after72 h, supernatants were collected and TGF-β quantified usingan anti-human TGF-β ELISA (R&D Systems).

Western blot

Cell extracts were fractionated in 6–15% SDS–PAGE, andthen transferred to PVDF membranes. We used the followingantibodies: α-vimentin (M725, 1/200; Dako) and α-tubulin (9026;1/5000; Sigma). Incubation was 1 h at room temperature (RT)unless otherwise specified, followed by peroxidase-labeled goatanti-mouseor -rabbit, or rabbit anti-goat antibody (1/2000,Dako;1 h, RT). Blots were developed using ECL (Amersham).

Immunocytochemistry

Cells were formalin-fixed (5min, RT), washed extensivelywithPBS, and incubated with anti-vimentin (1/100; Dako) or anti-pan-cytokeratin (undiluted; Becton Dickinson). Samples werewashed with PBS, incubated with Cy3-goat anti-mouse anti-body (1/400, Jackson, 1 h, RT), slides mounted in Vectashieldwith DAPI (Vector Laboratories), and images were captured ona fluorescence microscope (Leica).

Immunohistochemistry

Tumors were formalin-fixed and included sections of paraf-fin-embedded were incubated with primary antibodies forvimentin (1/100; Dako) and pan-cytokeratin (1/100; Dako).Slides were washed and then incubated with a biotinylatedsecondary antibody (30 min at RT) and a streptavidinperoxidase complex (30 min at RT) (Vector Laboratories). Theimmunostaining was visualized using diaminobenzidine andcounterstained with hematoxylin.

Fig. 3 – Desmosome and filopodium formation in TMC.(a) Detail of TMC filopodium (left, 25,000×; right, 20,000×).(b) Detail of desmosomal junctions between two TMC (left,60,000×; right, 100,000×). Arrows indicate desmosomaljunctions or filopodia.

Table 1

Post-MSC vs. pre-senescent MSC

TMC vs. pre-senescent MSC

694 E X P E R I M E N T A L C E L L R E S E A R C H 3 1 4 ( 2 0 0 8 ) 6 9 1 – 6 9 8

Flow cytometry analysis

Cells were analyzed in an EPICS XL-MCL cytometer (CoulterElectronics). After washing in PBA [phosphate-buffered salinecontaining 0.1% (w/v) bovine serum albumin and 0.02% (w/v)sodium azide], 105 cells were routinely an'alyzed. Antibodieswere precalibrated to determine optimal concentration. Theantibodiesusedwere: anti-SSEA1 (MAB4301), -SSEA3 (MAB4303),-SSEA4 (MAB4304), -podocalyxin (MAB1658, R&D), -Tra 1–60(MAB4360), and -Tra 1–81 (MAB4381), all fromChemicon. Offlineanalysis was done using the WinMDI free software package.

Blastocyst microinjection

C57BL/6J blastocysts were microinjected with 8 EGFP-expres-sing TMC, generated from C57BL/6Tg15(act-EGFP)Osb1 strainmice, and transferred to pseudopregnant CD1 females.

Gene x-foldchange

pvalue

zscore

x-foldchange

pvalue

zscore

E-cadherin 1.09 0.53 0.79 1 1 0snail –1.06 0.22 –0.33 –3.37 b0.01 –2.01slug N/D N/D N/D N/D N/D N/Dvimentin 1.13 0.64 0.68 1.3 0.22 0.42ACTA2 –1.02 0.69 –0.12 –5.84 b0.01 –2.81CDH6 –3.68 b0.01 –8.61 –5.81 b0.01 –3.4MMP2 1.1 0.21 0.53 –7.01 b0.01 –3.11COLIA2 1.49 0.04 2.24 –37.15 b0.01 –5.74COL5A2 1.3 0.17 1.47 –17.53 b0.01 –4.63syndecan2 –1.2 0.41 –1.04 –5.25 b0.01 –2.73fibulin2 1.94 0.05 5.21 –1.95 b0.01 –2.52CTGF –1.05 0.63 –0.28 –4.7 b0.01 –2.35FGF2 1.1 0.51 0.6 –4.09 b0.01 –2.49

Results

Identification of primary tumors and metastases generated bytransformed mesenchymal cells

We previously reported spontaneous transformation of humanMSC and the sequence of events required for this process [14].MSC cultured in vitro bypassed senescence at high-frequency,post-senescent MSC then entered a crisis phase from whichtumorigenic derivatives, termed transformed mesenchymalcells (TMC), arose. In order to study the tumorigenic potentialof these cells, we injected TMC subcutaneously into immuno-deficient mice. These cells generated primary tumors and lung

metastases, andTMC extracted from these primary tumors alsogave rise to tumors after injection into mice (Fig. 1). Interest-ingly, histopathological analysis of the primary tumors gener-ated in vivo identified them as poorly differentiated carcinomas,rather than sarcomas.

A mesenchymal–epithelial transition is involved in TMCgeneration

Histopathological result suggested the involvement of amesenchymal–epithelial transition in TMC generation. Wepreviously showed that the transformed cells had lost theirfibroblast-like morphology and were smaller and morecompact in vitro [14]. TMC also attached less efficiently toculture dishes and detached more rapidly after trypsinization(approximately 5 min for pre- and post-senescence MSC andless than 1 min for TMC; not shown).

We obtained a similar result in morphologic analysis usingtransmission electronmicroscopy; TMCwere smaller andmorecompact than the MSC from which they derived, had a highernucleus/cytoplasm ratio, and had lost the typical fibroblast-likemorphology of mesenchymal cells (Fig. 2). TMC, but not MSC,showed filopodium (Fig. 3a) and desmosome formation (Fig. 3b),both characteristic of epithelial cells [17]; these data supportedassociation of MET with TMC generation.

Usingmicroarray techniques, we comparedmRNA levels oflineage-specific genes, genes associated with EMT or MET, inpre- andpost-senescenceMSCaswell as inTMC.We found fewdifferences between pre- and post-senescence MSC, withslight up-regulation of collagen type Iα2 (COL1A2) and clearcadherin 6 down-regulation (pb0.01; Table 1). In contrast, TMCshowedmodulation of almost all genes analyzed compared topre-senescence MSC. Many mesenchymal lineage genes weremarkedly down-regulated in TMC, including snai1, ACTA2,cadherin 6, matrix metalloprotease 2 (MMP2), collagens Iα2(COL1A2) and Vα2 (COL5A2), syndecan2, fibulin2, CTGF andFGF2, whereas vimentin and E-cadherin mRNA levels wereunaffected (Table 1).

Fig. 4 – qRT-PCR ofmesenchymal regulators. mRNA levelsmeasured by qRT-PCR of snail and slug in pre- and post-senescenceMSC and in TMC.

695E X P E R I M E N T A L C E L L R E S E A R C H 3 1 4 ( 2 0 0 8 ) 6 9 1 – 6 9 8

qRT-PCR comparison of mRNA levels for the mesenchymallineage regulator genes snail and slug confirmed that therewereno appreciable differences between pre- and post-senescenceMSC, whereas expression of both genes was repressed in TMC(Fig. 4).

To confirm that this MSC transformation reflects altera-tions in protein levels, and is not restricted to mRNAmodulation, we analyzed protein expression of vimentin, amesenchymal marker, as well as cytokeratins—which arecharacteristically expressed by epithelial cells. In contrast tothe mRNA data, immunocytochemistry showed that MSC

Fig. 5 – MET/EMT-related protein expression. (a) Immunocytochvimentin and pan-cytokeratin expression. (b) Western blot analyand in TMC; α-tubulin was used as loading control. (c) Immunohwith antibody alone (left), staining for vimentin (middle) and for

expressed vimentin, but not cytokeratins in comparison toTMC, which did not express vimentin but were positive forcytokeratins (Fig. 5a). These results were confirmed byWestern blot analysis, which showed again that whereas theMSC expressed vimentin, the TMC were vimentin-negative(Fig. 5b). Moreover, immunohistochemistry analysis of thetumors showed that TMC also expressed cytokeratins but notvimentin in vivo (Fig. 5c).

The transforming growth factor (TGF)-β is an importantfactor in the differentiation of several tissue types and TGF-βfamily members have essential roles in the plasticity process

emical analysis of pre-senescence MSC and TMC showingsis of vimentin expression in pre- and post-senescence MSCistochemistry of TMC-derived tumors. Secondary stainingpan-citokeratin (right).

Fig. 6 – TMC dedifferentiation. (a) Embryonic antigen expression in MSC (upper panel), and TMC (lower panel) using flowcytometry. (b) Two littermates of EGFP-TMC-microinjected blastocysts. (c) Fluorescence microphotography of placenta fromEGFP-TMC-microinjected blastocysts. Middle and lower panels show details from the upper panel.

696 E X P E R I M E N T A L C E L L R E S E A R C H 3 1 4 ( 2 0 0 8 ) 6 9 1 – 6 9 8

[18]. We analyzed the TGF-β concentration in supernatants ofMSC and TMC cultures, finding that TGF-β levels are lower inMSC than in TMC (Supplementary Fig. 1). TGF-β also induceEMT in carcinomas, however, TMC remains as epithelioidcells. Perhaps the pathway of TGF-β signalling would not beactive because of TGF-β-inhibitors overexpression showed inTMC, as Smad7, telomerase and Myc.

Cell dedifferentiation associated with TMC generation

Weobserved that, in comparison to theMSC, theTMCexpressedmore embryonic antigens. We analyzed specific humanembryonic stem cell (ES) markers SSEA3, SSEA4, podocalaxin,Tra 1–60 andTra 1–81; and themurine ES-specific SSEA1markeras a control. HumanMSC expressed low podocalaxin levels and

697E X P E R I M E N T A L C E L L R E S E A R C H 3 1 4 ( 2 0 0 8 ) 6 9 1 – 6 9 8

approximately 60% of MSC were SSEA4-positive. In contrast,human TMC maintained SSEA4 expression, but overexpressedpodocalaxin, Tra 1–60 and Tra 1–81, suggesting that TMC havedeveloped more “primitive” cell-characteristics than the MSCfrom which they derive (Fig. 6a).

To test their in vivo stemness potential, we used murineTMC [14] in blastocysts-contribution experiments. The mur-ine cells, like the human cells, also undergo a mesenchymal–epithelial transition during their transformation process.Murine TMC in aggregation experiments were unstable anddid not integrate into C57BL/6 developing blastocysts (notshown), so we microinjected TMC into developing blasto-cysts. We analyzed day 21 foetuses (n=3), one of whichappeared to be abnormal. The mouse was correctly formed,but was smaller than its littermates, with an intense redcolour and an enlarged abdomen (Fig. 6b). Under fluorescentlight, only the placenta of the abnormal foetus showed smallgroups of EGFP-positive cells (Fig. 6c). We did not detecttumors in this mouse by histopathological analysis. Thesedata suggest a cell dedifferentiation process associated withTMC generation.

Discussion

Wepreviously reported that humanandmurinemesenchymalstem cells spontaneously transform during long-term in vitroculture [14]. Other authors confirmed these findings, but theydescribed the generation of sarcomas after transplantation inmice [19–21]. Here, we report that our spontaneously trans-formed mesenchymal stem cells generate poorly differen-tiated carcinomas when injected into immunodeficient mice,clearly demonstrating a novelmesenchymal–epithelial transi-tion associated with tumorigenesis.

Epithelial–mesenchymal and mesenchymal–epithelialtransitions have been associated with several processesduring organism development [8,10]. In adult organisms, ithas been proposed that restrictive mechanisms repress thesecellular transitions [22]. However, during tumor development,these mechanisms appear to fail, allowing the EMT describedin metastasis generation [6]. Other potential in vivo situationswhere these mechanisms can be altered is inflammationbecause MSC are mobilized to these sites of injury and con-sequently subjected to the inflammatory response [23,24].Although there are no data about a potential MET beyond thissituation; nevertheless, other groups have published thatbone marrow-derived cells could differentiate into mature-appearing epithelial cells in response to tissue damage [25].Perhaps these kinds of stressful situations makes MSC moresusceptible to transformation which, if associated with anMET, would generate carcinomas. This hypothesis wouldsuggest that, in contrast to current thinking, not all carcino-mas are derived from epithelial cells. A report showing thatsome epithelial tumors derive from bone marrow cells in amice model of chronic inflammation supports this hypothesis[26].

Mesenchymal–epithelial transitions during in vitro cellculture is not a common event. For the study of MET, over-expression of certain genes is generally used. An example ofthat approach has been carried out using versican or BMP-7, to

force a MET in fibroblastic cell lines [27,28]. In other cases,chemical treatment of bone marrow-derived cells could giverise to different tumor types in vitro and in vivo (sarcomas,epithelial tumors, neural tumors, and tumors of poor differ-entiation) [29]. Until now, only one work showing a sponta-neousMET in vitro using a chondrosarcoma tumor cell line hasbeen reported [30]. However, the author found that onlywhen cells were left unpassed for about a month they couldgradually transform to a different phenotype. Now, we havedescribed a novel spontaneous model where long-term cul-tures of normal humanMSC suffer anMET associatedwith celltransformation.

The first approaches used to reprogram adult cells in the1990s used nuclear transfer of differentiated cell nuclei toenucleated oocytes [31]. This was followed by reports ofcancer cell reprogramming by nuclear transfer of tumornuclei to enucleated oocytes [32], or by fusion of embryonicstem cells with differentiated cells [33,34]. Herewe found thatTMC, without nuclear transfer or cell fusion approaches,overexpressed embryonic antigens and could contributepartially to blastocyst development. It therefore appearsthat in our MSC spontaneous transformation model, thetransition appears to be associated with a “regression” ordedifferentiation, with TMC representing a more “primitive”cell type. TMC contributed to placental tissue, as do otherprimitive cells such as teratocarcinomas [35]. Although theTMC did not generate tumors when microinjected intoblastocysts, their contribution apparently caused an alteredphenotype. The fetus was smaller than its littermates,reddish, and showed an enlarged abdomen. These altera-tions could be caused by VEGF secretion, in agreement withother groups who showed that a modest increase in VEGFlevels during fetal development or in newborn mice can alterorganogenesis, producing death [36, 37]. Supporting thishypothesis, we previously demonstrated, in vitro and in vivo,that TMC secrete high levels of VEGF [14].

To our knowledge, the MET linked to MSC transformationhas not been previously associated with tumor development.Our results support the hypothesis that certain carcinomasmay derive from mesenchymal cells rather than from epithe-lial precursors under stressful situations.

Acknowledgments

We thank I. Colmenero for assistance with histopathology,L. López for microarray analysis, L. Almonacid for qRT–PCRanalysis and G. Ligero for immunohystochemistry. Wewant also to thank to M.A. Nieto, M. Serrano, I. Sánchez-García and M. Ramírez for critical reading of the manuscriptand C. Mark for editorial support. DR and SG receivedSpanish Ministry of Education and Science pre-doctoralfellowships; JGC received post-doctoral fellowships fromthe MCYT and FIS. This work was partially supported bySpanish Ministry of Science and Technology (CICYT) grantsSAF2001-2262, SAF2005-0864 and GEN2001-4856-C13-02 toAB, and FIS CP03/0031 and PI052217 to JGC. The Departmentof Immunology and Oncology was founded and is sup-ported by the Spanish National Research Council (CSIC) andby Pfizer.

698 E X P E R I M E N T A L C E L L R E S E A R C H 3 1 4 ( 2 0 0 8 ) 6 9 1 – 6 9 8

Appendix A. Supplementary data

Supplementary data associated with this article can be found,in the online version, at doi:10.1016/j.yexcr.2007.11.017.

R E F E R E N C E S

[1] C.E. Eckfeldt, E.M. Mendenhall, C.M. Verfaillie, The molecularrepertoire of the “almighty” stem cell, Nat. Rev., Mol. Cell Biol.6 (2005) 726–737.

[2] G. Grafi, Y. Avivi, Stem cells: a lesson from dedifferentiation,Trends Biotechnol. 22 (2004) 388–389.

[3] L. Song, R.S. Tuan, Transdifferentiation potential of humanmesenchymal stem cells derived from bone marrow, FASEB J.18 (2004) 980–982.

[4] J.P. Thiery, Epithelial–mesenchymal transitions indevelopment and pathologies, Curr. Opin. Cell Biol. 15 (2003)740–746.

[5] J. Yang, S.A. Mani, J.L. Donaher, S. Ramaswamy, R.A. Itzykson,C. Come, P. Savagner, I. Gitelman, A. Richardson, R.A.Weinberg, Twist, a master regulator of morphogenesis,plays an essential role in tumor metastasis, Cell 117 (2004)927–939.

[6] Y. Kang, J. Massague, Epithelial–mesenchymal transitions:twist in development and metastasis, Cell 118 (2004) 77–279.

[7] S.A. Stacker, M.G. Achen, L. Jussila, M.E. Baldwin, K. Alitalo,Lymphangiogenesis and cancer metastasis, Nat. Rev., Cancer2 (2002) 573–583.

[8] E.D. Hay, The mesenchymal cell, its role in the embryo, andthe remarkable signaling mechanisms that create it, Dev.Dyn. 233 (2005) 706–720.

[9] Y. Nakaya, S. Kuroda, Y.T. Katagiri, K. Kaibuchi, Y. Takahashi,Mesenchymal–epithelial transition during somiticsegmentation is regulated by differential roles of Cdc42 andRac1, Dev. Cell 7 (2004) 425–438.

[10] S. Vainio, Y. Lin, Coordinating early kidney development:lessons from gene targeting, Nat. Rev., Genet. 3 (2002)533–543.

[11] T. Reya, S.J. Morrison, M.F. Clarke, I.L. Weissman, Stem cells,cancer, and cancer stem cells, Nature 414 (2001) 105–111.

[12] R. Pardal, M.F. Clarke, S.J. Morison, Applying the principles ofstem-cell biology to cancer, Nat. Rev., Cancer 3 (2003) 895–902.

[13] J. Marx, Cancer research. Mutant stem cells may seed cancer,Science 301 (2003) 1308–1310.

[14] D. Rubio, J. Garcia-Castro, M.C. Martin, R. de la Fuente, J.C.Cigudosa, A.C. Lloyd, A. Bernad, Spontaneous human adultstem cell transformation, Cancer Res. 65 (2005) 3035–3039.

[15] T. Brabletz, A. Jung, S. Spaderna, F. Hlubek, T. Kirchner,Opinion: migrating cancer stem cells—an integrated conceptof malignant tumour progression, Nat. Rev., Cancer 5 (2005)744–749.

[16] J.M. Wettenhall, G.K. Smyth, limmaGUI: a graphical userinterface for linearmodeling ofmicroarraydata, Bioinformatics20 (2004) 3705–3706.

[17] J. Overton, R. Meyer, E.A. Chernoff, Epithelial cell–cell and cellsubstrate contacts, Scan. Electron. Microsc. (1981) 297–305.

[18] R. Derynck, R.J. Akhurst, Differentiation plasticity regulatedby TGF-beta family proteins in development and disease, Nat.Cell Biol. 9 (2007) 1000–1004.

[19] M. Miura, Y. Miura, H.M. Padilla-Nash, A.A. Molinolo, B. Fu, V.Patel, B.M. Seo, W. Sonoyama, J.J. Zheng, C.C. Baker, W. Chen,T. Ried, S. Shi, Accumulated chromosomal instability in

murine bone marrow mesenchymal stem cells leads tomalignant transformation, Stem Cells 24 (2005) 1095–1103.

[20] J. Tolar, A.J. Nauta, M.J. Osborn, A. Panoskaltsis Mortari, R.T.McElmurry, S. Bell, L. Xia, N. Zhou, M. Riddle, T.M.Schroeder, J.J. Westendorf, R.S. McIvor, P.C. Hogendoom, K.Szuhai, L. Oseth, B. Hirsch, S.R. Yant, M.A. Kay, A. Peister,D.J. Prockop, W.E. Fibbe, B.R. Blazar, Sarcoma derived fromcultured mesenchymal stem cells, Stem Cells 25 (2007)371–379.

[21] Y. Wang, D.L. Huso, J. Harrington, J. Kellner, D.K. Jeong, J.Turney, I.K. McNiece, Outgrowth of a transformed cellpopulation derived from normal human BM mesenchymalstem cell culture, Cytotherapy 7 (2005) 509–519.

[22] G. Prindull, D. Zipori, Environmental guidance of normal andtumor cell plasticity: epithelial mesenchymal transitions as aparadigm, Blood 103 (2004) 2892–2899.

[23] D.J. Prockop, C.A. Gregory, J.L. Spees, One strategy for cell andgene therapy: harnessing the power of adult stem cells torepair tissues, Proc. Natl. Acad. Sci. U. S. A. 100 (Suppl 1) (2003)11917–11923.

[24] J.L. Spees, S.D. Olson, J. Ylostalo, P.J. Lynch, J. Smith, A. Perry,A. Peister, M.Y. Wang, D.J. Prockop, Differentiation, cellfusion, and nuclear fusion during ex vivo repair of epitheliumby human adult stem cells from bone marrow stroma, Proc.Natl. Acad. Sci. U. S. A. 100 (2003) 2397–2402.

[25] D.S. Krause, Engraftment of bone marrow-derived epithelialcells, Ann. N.Y. Acad. Sci. 1044 (2005) 117–124.

[26] J. Houghton, C. Stoicov, S. Nomura, A.B. Rogers, J. Carlson, H.Li, X. Cai, J.G. Fox, J.R. Goldenring, T.C. Wang, Gastric canceroriginating from bone marrow-derived cells, Science 306(2004) 1568–1571.

[27] S. Hosono, X. Luo, D.P. Hyink, L.M. Schnapp, P.D. Wilson, C.R.Burrow, J.C. Reddy, G.F. Atweh, J.D. Licht, WT1 expressioninduces features of renal epithelial differentiation inmesenchymal fibroblasts, Oncogene 18 (1999) 417–427.

[28] W.Sheng,G.Wang,D.P. LaPierre, J.Wen,Z.Deng,C.K.Wong,D.Y.Lee, B.B. Yang, Versican mediates mesenchymal–epithelialtransition, Mol. Biol. Cell 17 (2006) 2009–2020.

[29] C. Liu, Z. Chen, Z. Chen, T. Zhang, Y. Lu, Multiple tumor typesmay originate from bone marrow-derived cells, Neoplasia 8(2006) 716–724.

[30] P. Ouyang, An in vitromodel to studymesenchymal–epithelialtransformation, Biochem. Biophys. Res. Commun. 246 (1998)771–776.

[31] K.H. Campbell, J. McWhir, W.A. Ritchie, I. Wilmut, Sheepcloned by nuclear transfer from a cultured cell line, Nature380 (1996) 64–66.

[32] R.H. Blelloch, K. Hochedlinger, Y. Yamada, C. Brennan, M.Kim, B. Mintz, L. Chin, R. Jaenisch, Nuclear cloning ofembryonal carcinoma cells, Proc. Natl. Acad. Sci. U. S. A. 101(2004) 13985–13990.

[33] M. Tada, Y. Takahama, K. Abe, N. Nakatsuji, T. Tada, Nuclearreprogramming of somatic cells by in vitro hybridization withES cells, Curr. Biol. 11 (2001) 1553–1558.

[34] Q.L. Ying, J. Nichols, E.P. Evans, A.G. Smith, Changing potencyby spontaneous fusion, Nature 416 (2002) 545–548.

[35] B. Mintz, K. Illmensee, Normal genetically mosaic miceproduced from malignant teratocarcinoma cells, Proc. Natl.Acad. Sci. U. S. A. 72 (1975) 3585–3589.

[36] T.D. Le Cras, R.E. Spitzmiller, K.H. Albertine, J.M. Greenberg, J.A.Whitsett, A.L. Akeson, VEGF causes pulmonary hemorrhage,hemosiderosis, and air space enlargement in neonatal mice,Am. J. Physiol., Lung Cell. Mol. Physiol. 287 (2004) L134–L142.

[37] L. Miquerol, B.L. Langille, A. Nagy, Embryonic development isdisrupted bymodest increases in vascular endothelial growthfactor gene expression, Development 127 (2000) 3941–3946.