Efficient acquisition of dual metastasis organotropismto bone and lung through stable spontaneousfusion between MDA-MB-231 variantsXin Lua and Yibin Kanga,b,1

aDepartment of Molecular Biology, Princeton University, Princeton, NJ 08544; and bBreast Cancer Program, The Cancer Institute of New Jersey,New Brunswick, NJ 08903

Edited by Joan Massagué, Memorial Sloan-Kettering Cancer Center, New York, New York, and approved April 16, 2009 (received for review January 6, 2009)

Cell fusion is involved in many critical developmental processes,including zygote formation and organogenesis of placenta, bone,and skeletal muscle. In adult tissues, cell fusion has been shown toplay an active role in tissue regeneration and repair, and itsfrequency of occurrence is significantly increased during chronicinflammation. Fusion between tumor cells and normal cells, oramong tumor cells themselves, has also been speculated to con-tribute to tumor initiation, as well as phenotypic evolution duringcancer progression and metastasis. Here, we show that dualmetastasis organotropisms can be acquired in the same cellthrough in vitro or in vivo spontaneous fusion between bone- andlung-tropic sublines of the MDA-MB-231 human breast cancer cellline. The synkaryonic hybrids assimilate organ-specific metastasisgene signatures from both parental cells and are genetically andphenotypically stable. Our study suggests cell fusion as an efficientmeans of phenotypic evolution during tumor progression andadditionally demonstrates the compatibility of different metasta-sis organotropisms.

breast cancer � cell fusion � nuclear reprogramming � genomic instability �centrosome

The development of cancer is believed to be driven by theprogressive accumulation of numerous genetic and epige-netic alterations that gradually allow early hyperplasia to becomehighly malignant tumors (1, 2). Likewise, metastasis organotro-pism—the capability of tumor cells to colonize specific targetorgans—is thought to emerge via acquisition of distinct sets oforgan-specific metastasis genes in metastatic variants that aremost adapted to different target organ microenvironmentsthrough Darwinian selection (3). Indeed, genomic profiling ofmetastatic variants selected in vivo in mouse models of breastcancer has unveiled 2 separate sets of genes that promotemetastasis to bone and lung, respectively (4, 5), although it isunclear whether such distinct organ-specific metastasis genesignatures can coexist in the same cell to give rise to tumor cellscapable of colonizing both organs. An alternative theory ofmetastasis progression has also been proposed that argues forrapid acquisition of metastatic phenotypes through fusion be-tween tumor cells or between tumor cells and certain normalcells, such as macrophages (6–9), rather than requiring theprogressive accumulation of independent genetic or epigeneticalterations in a single cell lineage. Given that a 1-cm3 tumor of�109 cells is estimated to harbor as many as 105 proliferatinghybrid cells produced by spontaneous cell fusion (6, 10), thecontribution of cell fusion to the phenotypic evolution of tumorscannot be overlooked.

In this study, we used a well-characterized model system oforgan-specific breast cancer metastasis to show that spontaneouscell fusion between bone-tropic and lung-tropic cancer cells,both in vitro and in vivo, generates stable hybrids with dualmetastasis tropism to both organs. In addition to directly dem-onstrating the role of cell fusion in the rapid acquisition ofcomplex metastasis properties, our study also discovered a

surprisingly high level of chromosomal and phenotypic stabilityin hybrids during long-term passage in vitro and in vivo, despitethe existence of amplified numbers of centrioles in these cells asa consequence of cell fusion. These results suggest a potentiallyimportant role of cell fusion in the progression and phenotypicdiversity of cancer.

ResultsSpontaneous Cell Fusion Generates Synkaryonic Hybrids That InheritChromosomal Abnormalities of Parental Cells. To examine thegenetic and phenotypic consequences of cell fusion betweentumor cells with distinct metastasis characteristics, we used 2previously reported sublines of the human breast cancer cell lineMDA-MB-231: the bone-metastatic SCP2 and the lung-metastatic LM2 (4, 5). These 2 cell lines were renamed as Bmand Lm, respectively, to facilitate the presentation of results (Fig.1A). To isolate spontaneously fused cells from the coculture ofthese 2 cell lines, we labeled Bm with a GFP-Fluc (fireflyluciferase) fusion protein expression construct with a puromycinresistance marker and Lm with RFP-Rluc (Renilla luciferase)and hygromycin as markers. These 2 cell lines were coculturedfor 1 day without any fusogenic reagents and hybrid cells wereselected by either dual drug selection or dual color fluorescence-activated cell sorting (FACS) (Fig. 1 A). After 4 rounds of cellsorting, we obtained a relatively pure GFP�/RFP� population(80.3%, named as BLmFACS) whereas only 1 round of puromycin/hygromycin dual drug selection gave rise to a population with95.4% GFP�/RFP� cells (named as BLmDrug) (Fig. 1 A). We alsoisolated BBm and LLm hybrids (resulting from self-fusion of Bmand Lm cells) by using the same approach. The fused cells aresynkaryons with enlarged nuclear and cell sizes (Fig. S1 A–C),but with growth doubling times similar to the parental cell lines(Bm 34.6h, Lm 36.0h, BLmFACS 36.2h, and BLmDrug 34.2h). Flowcytometric DNA content analysis and karyotyping showednearly doubled DNA content and chromosome numbers inhybrids compared with the parental cell lines (Fig. 1B). Spectralkaryotyping (SKY) further revealed that BLm hybrids adoptedall major chromosomal abnormalities (translocations and dele-tions) of both parental cell lines (Figs. 1 C and D and S1D).

To test whether spontaneous cell fusion also occurs in vivo, weinjected an equal mixture of Bm and Lm cells s.c. into nude mice.When the tumor diameter reached 10 mm, a single-cell suspen-sion was made from the tumor by mincing and digestion with

Author contributions: X.L. and Y.K. designed research; X.L. performed research; X.L. andY.K. analyzed data; and X.L. and Y.K. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Data deposition: The data reported in this paper have been deposited in the GeneExpression Omnibus (GEO) database, www.ncbi.nlm.nih.gov/geo (accession no. GSE14244).

1To whom correspondence should be addressed. E-mail: ykang@princeton.edu.

This article contains supporting information online at www.pnas.org/cgi/content/full/0900108106/DCSupplemental.

www.pnas.org�cgi�doi�10.1073�pnas.0900108106 PNAS � June 9, 2009 � vol. 106 � no. 23 � 9385–9390

MED

ICA

LSC

IEN

CES

Dow

nloa

ded

by g

uest

on

July

2, 2

021

http://www.pnas.org/cgi/data/0900108106/DCSupplemental/Supplemental_PDF#nameddest=SF1http://www.pnas.org/cgi/data/0900108106/DCSupplemental/Supplemental_PDF#nameddest=SF1http://www.pnas.org/cgi/content/full/0900108106/DCSupplementalhttp://www.pnas.org/cgi/content/full/0900108106/DCSupplemental

collagenase A. The cells were plated directly into dual drugselective media at a very low density (104 cells/10-cm dish) toselect for surviving colonies for 14 days. By using this approach,the frequency of in vivo spontaneous fusion resulting in viableprogenies was estimated to be 2.0 � 0.8 � 10�5 in s.c. tumors,very similar to the in vitro fusion frequency of 2.6 � 0.7 � 10�5(see SI Materials and Methods, data reported as the mean � SD),and is consistent with previously reported frequencies in othertumor models (11).

Fusion Between Bone- and Lung-Tropic Metastatic Cells ProducedDual-Tropic Hybrids. Parental Bm, Lm, and hybrids BBm, LLm,BLmDrug, and BLmFACS were injected into nude mice througheither left cardiac ventricles or tail veins to test their bone or lungmetastasis abilities, respectively. Metastatic progression wasmeasured by weekly bioluminescence imaging (BLI) and ana-lyzed with Kaplan-Meier curves (Fig. 2). Bm and BBm formedaggressive bone metastases but no lung metastasis. Conversely,Lm and LLm displayed high metastasis potential to lung and amuch weaker ability to form bone metastasis. In contrast,BLmDrug and BLmFACS showed strong metastasis ability to bothbone and lung, indicating that BLm cells have inherited metas-tasis phenotypes from both parental cell lines. Formation ofmassive metastases in both bone and lung by the hybrid cellswere confirmed by X-ray and histological analyses of the lesions(Fig. 2 C and F). The acquisition of dual metastasis organotro-pism was also confirmed when bone and lung metastasis assayswere performed with 2 hybrid clones (BLmDrug-i.v.#1 andBLmDrug-i.v.#2) obtained from in vivo fusion by Bm and Lmcells in s.c. tumors as described above (Fig. S2).

Interpretation of the finding that hybrid cells can form ag-gressive metastases in both organs may be complicated by thefact that the hybrids may contain a very small fraction of unfusedparental cells, which may separately form metastases in bone andlung. This possibility was ruled out by 2 pieces of evidence. First,

firefly luciferase and Renilla luciferase BLI performed on con-secutive days showed exactly matched metastasis locations in themice injected with BLmDrug (Fig. S3). This result suggested thatmost, if not all, of metastases were derived from hybrids withdual markers instead of rare singly labeled parental cells thatsurvived the drug selection. A more rigorous test was performedby analyzing single cell isolates from BLm. By using single-cellsorting, we isolated 21 single cell clones from BLmDrug. Werandomly picked 5 clones (#6, #7, #8, #12, and #18) to testtheir tissue-specific metastasis abilities (Fig. S4). All of theclones were bone metastatic, although clones #6 and #7 werenot as strong as the pooled BLmDrug population. All of the clonesexcept clone #7 were also lung metastatic, with clone #18showing an even more aggressive phenotype than the pooledpopulation. Taken together, these results indicated that bothbone and lung metastasis phenotypes can be efficiently acquiredin a single tumor cell through cell fusion.

Bone and Lung Metastasis Gene Signatures Are Coexpressed byHybrids. As cell fusion did not confer growth advantages, wereasoned that the acquired dual metastasis tropism may bebecause of the coexpression of tissue-specific metastasis genesthat have been previously defined for both parental cell lines (4,5). We performed microarray analyses on various cell linesderived from our current study and used the previously identifiedbone and lung metastasis gene signatures to perform unsuper-vised clustering of these cell lines as well as several previouslycharacterized MDA-MB-231 sublines with different metastasisorganotropisms (Fig. 3 A and B). Both signatures were able tosegregate various cell lines into a lowly metastatic group and ahighly metastatic group. BLm and its single cell-derived cloneswere consistently clustered into the strongly metastatic groupsbased on either signature, although not all of the genes in eachsignature retained their expression level in the BLm cells. Asexpected, the self-fusion hybrids, BBm and LLm, always clus-

Fig. 1. Spontaneous fusion hybrids are hyperploid and inherit chromosomal abnormalities from both parental cell lines. (A) Experimental design for using FACSor drug selection to isolate spontaneous fusion hybrids (BLmFACS or BLmDrug) from the coculture of Lm and Bm cells that were differentially labeled with variousmarkers. The fluorescence dot plots for BLmDrug and BLmFACS represent the results after 1 round of drug selection and 4 rounds of FACS, respectively. (B) DNAcontent and chromosome numbers of the parental and hybrids evaluated by flow cytometric (propidium iodide staining) and karyotyping (Giemsa staining)methods (n � haploid chromosome number of MDA-MB-231). (C) Representative SKY image showing the chromosomal composition of BLmDrug. (D) Summaryof chromosomal translocations and deletions in Bm, Lm, and BLmDrug.

9386 � www.pnas.org�cgi�doi�10.1073�pnas.0900108106 Lu and Kang

Dow

nloa

ded

by g

uest

on

July

2, 2

021

http://www.pnas.org/cgi/data/0900108106/DCSupplemental/Supplemental_PDF#nameddest=STXThttp://www.pnas.org/cgi/data/0900108106/DCSupplemental/Supplemental_PDF#nameddest=SF2http://www.pnas.org/cgi/data/0900108106/DCSupplemental/Supplemental_PDF#nameddest=SF3http://www.pnas.org/cgi/data/0900108106/DCSupplemental/Supplemental_PDF#nameddest=SF4

tered together with their parental cell lines of origin, Bm and Lm.When the expression profiles of MDA-MB-231 variants andhybrids were scored for their similarity to the lung and bone

metastasis signatures, the hybrids were positive for both signa-tures, albeit with scores slightly lower than those from thesublines with only one metastasis organ tropism (Fig. 3C).

Fig. 2. BLm hybrid cells are highly metastatic to both bone and lung. (A and D) Kaplan-Meier curves show the metastasis propensities of the parental and hybridsto bone (A) or the lung (D). *, P � 0.05 with log-rank test when compared with Lm (A) and Bm (D) (n � 8 in A or n � 5 in D). (B and E) BLI images of representativemice from each experimental group showing development of metastases in the hindlimbs (B) or the lungs (E). (C) X-ray radiography of the hindlimbs and H&Estaining of the femurs from mice injected with parental and BLm cells. Osteolytic bone lesions are marked with arrows, and metastases are marked with dottedlines. (F) Photographs and H&E staining of the lung from mice injected with parental and BLm cells. Metastasis nodules are indicated with dotted lines in theH&E images. (Scale bar, 500 �m in C and F.)

Fig. 3. Coexpression of both bone and lung metastasis signatures in the BLm hybrid cells. (A and B) Hierarchical clustering of the BLm cells and MDA-MB-231sublines is shown with different metastatic abilities by using the bone-metastasis gene signature (A) or the lung-metastasis gene signature (B). Cell lines underthe red line are strongly metastatic, whereas those under the green line are weakly metastatic (4, 5). ‘‘Parental’’ refers to the original MDA-MB-231 cell lineacquired from ATCC and used to generate all other sublines. Cell lines started with ‘‘SCP’’ were single-cell clones of MDA-MB-231. MDA-MB-231 derivatives 1833,3481, 4173, 4175, 4180, and 4142 were obtained by in vivo selection (4, 5). In both heatmaps, the BLm lines clustered together with the strongly metastaticsublines. (C) The metastasis signature correlation score analysis shows the grouping of the weakly metastatic cell lines (negative for both signatures), cell lineswith a single metastasis tropism (positive for either one, but not both, of the signatures), and BLm cells (positive for both signatures). Cell lines used in this analysisare the same as in A and B. (D) Northern blot showing BLm cells coexpress representative lung- and bone-metastatic signature genes. MMP1 is a shared genefor both signatures.

Lu and Kang PNAS � June 9, 2009 � vol. 106 � no. 23 � 9387

MED

ICA

LSC

IEN

CES

Dow

nloa

ded

by g

uest

on

July

2, 2

021

Northern blot analysis confirmed the coexpression of typicalbone metastasis genes and lung metastasis genes in the BLm cellsand BLmDrug clones (Figs. 3D and S5). These results indicate thatthe dual metastasis organ tropism displayed by BLm reflects thecoexpression of distinct sets of tissue-specific metastasis genesand that fusion as a cellular process itself did not significantlyalter gene expression or metastatic behavior.

Chromosomal and Phenotypic Stability of Hybrids. Cell fusion mayresult in dramatic chromosomal instability and hybrid cells maygradually lose extra chromosomes through reductive mitosis(12). We tested the long-term genomic and phenotypic stabilityof BLm by culturing the cells continuously for nearly a year (upto the time of manuscript preparation). Remarkably, BLmFACS,BLmDrug and its clones maintained the nearly combined genomesize (Fig. 4A) and the expression of the characteristic metastasisgenes (Fig. 4B) throughout the duration of the experiment. Inaddition, the tissue-specific metastasis ability to both bone andlung was maintained in BLmDrug cells that were cultured for 1,3, and 6 months (Fig. 4 C and D). To test whether hybrids alsomaintained the chromosomal and phenotypic stability in vivoduring metastasis assays in mice, we isolated tumor cells (1241Rand 1241L) from bone metastases formed by BLmDrug andconfirmed their sustained strong metastatic ability to bothorgans (Fig. 4 C and D). DNA content analysis of 1241R, 1241Land their secondary in vivo derivatives (1241R-d and 1241L-d)consistently revealed a combined genome size (Fig. 4E).

DiscussionAlthough cell fusion is generally considered to be a highlyspecialized event that only occurs in limited scenarios duringdevelopment (13, 14), a broader role for cell fusion in repair andregeneration of adult tissues has been increasingly recognizedsince the discovery of cell fusion as one of the underlyingmechanisms of the so-called ‘‘transdifferentiation’’ phenome-non—the ability of one lineage of differentiated cells, such asbone marrow cells, to give rise to a different lineage of cells, suchas liver or neural cells (15–21). In recent years, cell fusion hasalso become an important research platform to study epigeneticreprogramming, cell fate manipulation and pluripotency (14, 22,23). In cancer research, cell fusion has been speculated to play

important roles in several aspects of tumor progression, includ-ing generation of cancer stem cells (8), acquisition of metastasisability (7), and multidrug resistance (11). Cancer cells can haveconsiderably high rate (up to 1%) of fusion in vivo in experi-mental tumor models (10). Recent findings that chronic inflam-mation dramatically increases the frequency of cell fusion be-tween hematopoietic cells with various cell lineages (24, 25) mayhelp explain the high fusogenicity of tumor cells, as inflamma-tion is often associated with the tumor microenvironment (26).Fusion between cancer cells may be further facilitated by fuso-genic viruses (27, 28) and entosis (29). Despite the technicaldifficulty in directly detecting cell fusion in cancer patients,several clinical observations support the existence and potentialimportance of cell fusion in cancer development. First, in kidneycancer patients who were also recipients of previous bonemarrow transplantation, the renal cell carcinoma cells werefound to contain genetic material from both the recipients andthe donor (30, 31). Secondly, premature chromosome conden-sation, a consequence of fusion of cells at different stages of cellcycle, has been observed in a large number of tumor types(32–35). Finally, binucleated and multinucleated cells are fre-quently observed in many types of tumors (9), and increasedDNA content and aneuploidy in tumor cells, which may resultfrom cell fusion, is associated with poor prognosis and metastasis(36–38). Indeed, we estimated that cell fusion accounts for65–88% of the hyperploid cells in our current experimentalsetting (see calculation in SI Materials and Methods and Fig. S6and Table S1).

Compared with mutation-based phenotypic evolution duringcancer progression, the main advantage of cell fusion is itsefficiency in generating drastic genomic changes and phenotypicheterogeneity (11). Although fusion of 2 genetically identicaltumor cells may not benefit tumor cell progression, fusionbetween tumor cells with distinct phenotypes can generatehybrids with diversified phenotypes to allow better adaptation tovarious secondary environments during metastasis or to survivedifferent therapeutic challenges. In this study, we derived spon-taneous fusion hybrids from the mixed population of 2 breastcancer sublines with distinct organ-specific metastasis propen-sities (Bm and Lm) and found that the hybrids (BLm) efficientlyacquired the ability to metastasize to both organs (Fig. 2).

Fig. 4. Genomic and phenotypic stability of hybrid BLm cells during long-term in vitro culture and in vivo passage. (A) DNA content histograms of the pooledBLmDrug and its single cell derived clones measured after 1, 3, and 6 months of culture (n � haploid number of chromosomes in MDA-MB-231). (B) Expressionof selected genes from the lung and bone metastasis signatures in the pooled BLmDrug at 3 time points during the long-term in vitro culture. (C) Kaplan-Meiercurves showing bone metastasis (Left) and lung metastasis (Right) developed after injection of BLmDrug cultured in vitro for 1, 3, or 6 months, or isolated frombone metastases formed by BLmDrug (1241L, 1241R). *, P � 0.05 with log-rank test when compared with Lm or Bm (n � 8 for bone metastasis assays and n � 5for lung metastasis assays). (D) Representative mice from experimental groups in (C) with hindlimb bone metastases on day 30 after injection and lung metastaseson day 45 after injection. (E) Stable DNA content of BLmDrug after 1 round (1241L and 1241R) or 2 rounds (1241R-d, 1241L-d) of in vivo passage as bone metastasis(n � haploid chromosome number of MDA-MB-231).

9388 � www.pnas.org�cgi�doi�10.1073�pnas.0900108106 Lu and Kang

Dow

nloa

ded

by g

uest

on

July

2, 2

021

http://www.pnas.org/cgi/data/0900108106/DCSupplemental/Supplemental_PDF#nameddest=SF5http://www.pnas.org/cgi/data/0900108106/DCSupplemental/Supplemental_PDF#nameddest=STXThttp://www.pnas.org/cgi/data/0900108106/DCSupplemental/Supplemental_PDF#nameddest=SF6http://www.pnas.org/cgi/data/0900108106/DCSupplemental/Supplemental_PDF#nameddest=ST1

Similar results were observed when hybrid cells were obtainedfrom the fusion of Lm and SCP28, another bone-tropic meta-static subline of MDA-MB-231 (Fig. S7).

The gain of dual tropism in the BLm cells is not simply due tolarger cell size and increased mechanical arrests in capillary bedsof distant organs, because hybrids of identical parental cells(LLm and BBm) did not show significant change in metastasispropensity. Instead, dual metastasis organotropism of hybrids islikely because of the expression of both sets of genes importantfor metastasis to each organ. It should be noted that the modestbut statistically insignificant (P � 0.0849, log-rank test) increaseof bone metastatic potential by LLm as compared with Lm (Fig.3D) may be because of elevated expression of 2 bone metastasisgenes, MMP1 and IL11. The ability of hybrid cancer cells toretain the genomic, transcriptomic, and phenotypic characteris-tics of both parents without a significant bias (Figs. 3 and S7C)is distinctively different from the dramatic nuclear reprogram-ming of differentiated somatic cells to the embryonic state aftertheir fusion with embryonic stem (ES) cells (22, 23). Thedifference in transcriptional reprogramming in hybrids may be inpart because of the difference in epigenetic regulation betweenthe somatic/ES cell hybrids and the BLm hybrids. Although EScells may reprogram the epigenetic regulation landscape (e.g.,methylation pattern of the OCT4 promoter) to a predominant ESstate after fusion with fibroblasts (22), coexistence of epigeneticregulatory mechanisms from both fusion partners in the BLmhybrids may explain the moderately elevated expression of bothsets of organ-specific metastasis genes in the hybrids (Fig. 3 Aand D). Although the level of expression of these genes in thehybrid are modest compared with the generally higher level oforganotropic gene expression in Bm or Lm cells, it is sufficientto promote metastasis of hybrids to both bone and lung.

The fact that cell fusion can generate tumor cells with dualmetastasis tropism to bone and lung indicates that differentmetastasis organotropism and the related metastasis signaturescan coexist in the same tumor cell to facilitate simultaneousmetastasis to multiple organs (4, 5, 39). Hybrid cells withmultiple metastasis organotropism may confer several advan-tages for disseminated tumor cells to proliferate and survivetherapeutic treatments. First, tumor cells with multiple metas-tasis organotropism will be able to colonize several differentorgans when they are disseminated systemically through bloodcirculation, whereas tumor cells with only a single metastasisorganotropism can only colonize one organ and will be lost whenthey are distributed to an inhospitable microenvironment.Therefore, fusion between tumor cells with different metastaticability can increase the efficiency of metastatic dissemination.Secondly, because metastases in different organs often havedifferent responses to therapeutic treatments (40), tumor cellswith multiple metastasis organotropisms can survive in a differ-ent organ and then quickly re-colonize the organ that haspositively responded to therapies. Finally, cell fusion could alsolead to combinatorial advantage in other phenotypes, like che-moresistance, and may therefore, confer additional survivaladvantage for hybrid cells.

We also found that hybrid cells resulted from spontaneousfusion are remarkably robust in maintaining their genetic andphenotypic stability, at least for several months in our currentexperiment (Figs. 4 and S7 G and H). The chromosome stabilityafter cell fusion was also observed in various other experimentalsystems, including fusion beween HeLa and fibroblasts (41),between mouse mammary tumor cells (11, 42), or between EScells and fibroblasts (22). Cell fusion automatically leads toincreased number of centrosomes in the hybrids. Supernumerarycentrosomes can cause multipolar spindles in mitosis which willresult in either aborted mitosis or daughter cells with imbalancedchromosome numbers (12, 43). However, the near completeretention of all chromosomes from parent cell lines in our

hybrids and their remarkable chromosomal stability duringlong-term passage suggested that they may have intrinsic mech-anism(s) to undergo bipolar segregation of chromosomes despitethe existence of additional centrosomes resulting from cellfusion. Indeed, when we quantified the number of centrosomesin postprophase mitosis in Bm, Lm, and various BLm cell linesby staining the cells with the centrosome marker pericentrin, allcells tested displayed a similar distribution of centrosome num-bers: �90% of cells in meta/ana/telophase harbor 2 centrosomesas microtubule-organizing centers; �10% of the cells have morethan 2 centrosomes, which may be accounted for by the basallevel of centrosome amplification present in MDA-MB-231 cells(Fig. S8). This result suggests that BLm hybrids are able tomaintain bipolar cell division despite initially gaining additionalcentrosomes after cell fusion. When we stained cells for both thecentrosome marker (pericentrin) and a centriole marker (cen-trin) and examined the centriole distribution in the cells with 2centrosomes during mitosis (Fig. S9), we observed 3 majorconfigurations: (i) 2 standard centrosomes each containing 2centrioles; (ii) 1 or both centrosomes containing �2 centrioles(coalesced centrioles); and (iii) centrioles dispersed in the cy-toplasm without pericentrolar materials (orphan centrioles) inaddition to the centrioles associated with the 2 centrosomes.Therefore, we envision 3 corresponding scenarios that may haveallowed hybrids to divide successfully despite supernumerarycentrioles introduced by cell fusion: (i) hybrids retain only 2structurally and functionally intact centrosomes by discardingextra centrioles; (ii) multiple centrioles coalesce into 2 spindle,similar to what was observed in neuroblastoma (43, 44); and (iii)multipolar mitosis is suppressed by keeping extra centrioles in apericentriolar material-free, inactive status (orphan centrioles).Interestingly, when these 3 configurations were quantified, asignificantly higher percentage (50.0%) of cells with orphancentrioles was seen in BLmDrug compared with Bm (22.1%) andLm (19.4%), perhaps representing a vestige of the fusion eventsresponsible for the formation of BLmDrug. Orphan centriolesmight therefore be an important approach for hybrids to main-tain relative chromosomal stability following cell fusion.

Overall, our study demonstrates a role for cell fusion in therapid acquisition of complex malignant traits such as metastasisorganotropism in tumor cells. Genomic and phenotypic charac-teristics of both parental cells were assimilated in the resultanthybrids without an overall nuclear reprogramming toward apredominant fusion partner and were stably maintained inhybrids after long-term passage in vitro and in vivo. Theseobservations, together with recent findings of the involvement ofcell fusion in many physiological and pathological conditions,suggest a potentially important role of cell fusion in cancerprogression that warrants further investigation.

Materials and MethodsCell Culture. SCP2 (Bm) and LM2 (Lm) sublines were derived from the parentalcell line MDA-MB-231 (ATCC) (4, 5) and were generously provided by JoanMassaugé (Sloan-Kettering Institute, New York). These sublines and theirfusion variants were maintained in DMEM with10% FBS and antibioticssupplemented with appropriate selective drugs. H29 cells, a packaging cellline used for retrovirus production, were maintained in DMEM supplementedwith 10% FBS, 1% glutamine, and antibiotics.

Generation of Hybrids by Spontaneous Cell Fusion. Bm and Lm cells werelabeled with different fluorescence, bioluminescence, and drug selectionmakers before fusion. The retroviral construct pMSCV-hyg-hrl-mrfp-ttk wasgenerated by subcloning the coding sequence of the Rluc-RFP-TK triple fusionreporter from pcDNA3.1-CMV-hrl-mrfp-ttk (45) to the BglII and HpaI sites ofpMSCV-hyg (Clontech). Retroviruses were generated from the H29 packingcell lines and used to transduce Lm cells. Bm cells were first labeled with theTK-GFP-Fluc triple reporter encoded in retroviruses produced from SFG-NESTGL(46) (tk-gfp-fluc) and then stably transfected with pPUR (Clontech) to becomeresistant to puromycin. To obtain fusion hybrids, 106 RFP-Rluc/hyg labeled Lm

Lu and Kang PNAS � June 9, 2009 � vol. 106 � no. 23 � 9389

MED

ICA

LSC

IEN

CES

Dow

nloa

ded

by g

uest

on

July

2, 2

021

http://www.pnas.org/cgi/data/0900108106/DCSupplemental/Supplemental_PDF#nameddest=SF7http://www.pnas.org/cgi/data/0900108106/DCSupplemental/Supplemental_PDF#nameddest=SF7http://www.pnas.org/cgi/data/0900108106/DCSupplemental/Supplemental_PDF#nameddest=SF7http://www.pnas.org/cgi/data/0900108106/DCSupplemental/Supplemental_PDF#nameddest=SF8http://www.pnas.org/cgi/data/0900108106/DCSupplemental/Supplemental_PDF#nameddest=SF9

cells were mixed with 106 GFP-Fluc/pur labeled Bm cells in a 10-cm dish. After24 h, the coculture was either selected with 1.5 mg/mL hygromycin B and 0.3�g/mL puromycin to obtain dual resistant cells, or subjected to FACS to purifyRFP/GFP double-positive cells. For analysis of DNA content, 106 cells were fixedwith 70% ethanol, treated with 0.12 mg/mL RNase A (Sigma), stained with 8�g/mL propidium iodide (Sigma), and analyzed by flow cytometry.

Karyotype Analysis. To obtain metaphase spreads to quantify chromosomenumbers, the cell culture was treated with 0.1 �g/mL Colcemid (GIBCO) for 1 h.The cells were harvested, treated with 0.075 M KCl for 10 min, fixed with amethanol/acetic acid mixture for 30 min, and dropped on precooled slides. Airdried slides were stained with Giemsa solution (GIBCO). SKY was performed bythe SKY/FISH facility in the Roswell Park Cancer Institute as described (47).

Tumor Xenografts and Analysis. All procedures involving mice, such as housingand care, and all experimental protocols were approved by InstitutionalAnimal Care and Use Committee (IACUC) of Princeton University. For intra-cardiac injections, 105 cells in PBS were injected into the left cardiac ventricleof 4-week-old, female nude mice (NCI) as described (4). For i.v. injection, 2 �105 cells in PBS were injected into the tail vein of nude mice as described (5).Development of metastases in bone and lung was monitored by BLI with theIVIS Imaging System (Xenogen) as described (4, 5). BLI analysis was performedwith Living Image software (Xenogen) by measuring photon flux of the regionof interest. Metastasis status was recorded as positive when BLI signal washigher than that of the basal BLI signal of total injected cells on day 0, and usedfor constructing Kaplan-Meier curves. X-ray radiography analysis of bonelesions was performed by using procedures as described (4). For the orthotopicxenograft model, mammary fat pad injections and tumor size measurementswere performed following the procedure described previously (5).

Accession Number. Microarray data reported herein have been deposited atthe National Center for Biotechnology Information Gene Expression Omnibus(http://www.ncbi.nlm.nih.gov/geo/) with the accession no. GSE14244.

Statistical Analysis. Results were reported as mean � SD. Kaplan-Meier curveswere created by using Stata 7.0 software (Stata Corporation). Log-rank testwas used to calculate the statistic significance of difference between metas-tasis curves. Other comparisons were performed by using unpaired 2-sidedStudent’s t test without equal variance assumption or nonparametric Mann-Whitney test.

Additional experimental procedures and discussion, including measure-ment of spontaneous fusion frequency, histological and immunofluorescenceanalyses, microarray analysis, Northern blot analysis, and methods to estimatethe contribution of cell fusion to hyperploidy are listed in SI Materials andMethods.

ACKNOWLEDGMENTS. We thank C. DeCoste for assistance on FACS; RoswellPark Cancer Institute for SKY analysis; G. Hu and M. Yuan for technicalassistance in statistical analysis and animal experiments; G. Hu, Y. Wei, M.A.Blanco, and members of our laboratories for helpful discussions; T.A. Guiseand K.S. Mohammad for training and technical advice in bone histology; J.Massagué (Sloan-Kettering Institute, New York) for Lm and Bm cell lines; R.Blasberg (Sloan-Kettering Institute, New York) for SFG-NESTGL plasmid; S.S.Gambhir (University of California, Los Angeles) for triple-reporter plasmids;and J. L. Salisbury (Mayo Clinic, Rochester, MN) for centrin antibody. Y.K. is aChampalimaud Investigator and was funded by grants from the Departmentof Defense, The American Cancer Society, The National Institutes of Health,and the New Jersey Commission on Cancer Research. X.L. is a recipient of aHarold W. Dodds Fellowship from Princeton University.

1. Chin L, Gray JW (2008) Translating insights from the cancer genome into clinicalpractice. Nature 452:553–563.

2. Hanahan D, Weinberg RA (2000) The hallmarks of cancer. Cell 100:57–70.3. Fidler IJ (2003) The pathogenesis of cancer metastasis: The ‘seed and soil’ hypothesis

revisited. Nat Rev Cancer 3:453–458.4. Kang Y, et al. (2003) A multigenic program mediating breast cancer metastasis to bone.

Cancer Cell 3:537–549.5. Minn AJ, et al. (2005) Genes that mediate breast cancer metastasis to lung. Nature

436:518–524.6. Duelli D, Lazebnik Y (2003) Cell fusion: A hidden enemy? Cancer Cell 3:445–448.7. Pawelek JM, Chakraborty AK (2008) Fusion of tumour cells with bone marrow-derived

cells: A unifying explanation for metastasis. Nat Rev Cancer 8:377–386.8. Bjerkvig R, Tysnes BB, Aboody KS, Najbauer J, Terzis AJA (2005) The origin of the cancer

stem cell: Current controversies and new insights. Nat Rev Cancer 5:899–904.9. Larizza L, Schirrmacher V (1984) Somatic cell fusion as a source of genetic rearrange-

ment leading to metastatic variants. Cancer Metastasis Rev 3:193–222.10. Fortuna MB, Dewey MJ, Furmanski P (1989) Cell fusion in tumor development and

progression: Occurrence of cell fusion in primary methylcholanthrene-induced tumor-igenesis. Int J Cancer 44:731–737.

11. Miller FR, McInerney D, Rogers C, Miller BE (1988) Spontaneous fusion betweenmetastatic mammary tumor subpopulations. J Cell Biochem 36:129–136.

12. Nigg EA (2002) Centrosome aberrations: Cause or consequence of cancer progression?Nat Rev Cancer 2:815–825.

13. Ogle BM, Cascalho M, Platt JL (2005) Biological implications of cell fusion. Nat Rev MolCell Biol 6:567–575.

14. Singec I, Snyder EY (2008) Inflammation as a matchmaker: Revisiting cell fusion. NatCell Biol 10:503–505.

15. Bjornson CR, Rietze RL, Reynolds BA, Magli MC, Vescovi AL (1999) Turning brain intoblood: A hematopoietic fate adopted by adult neural stem cells in vivo. Science283:534–537.

16. Mezey E, Chandross KJ, Harta G, Maki RA, McKercher SR (2000) Turning blood intobrain: Cells bearing neuronal antigens generated in vivo from bone marrow. Science290:1779–1782.

17. Brazelton TR, Rossi FM, Keshet GI, Blau HM (2000) From marrow to brain: Expressionof neuronal phenotypes in adult mice. Science 290:1775–1779.

18. Ying QL, Nichols J, Evans EP, Smith AG (2002) Changing potency by spontaneous fusion.Nature 416:545–548.

19. Terada N, et al. (2002) Bone marrow cells adopt the phenotype of other cells byspontaneous cell fusion. Nature 416:542–545.

20. Wang X, et al. (2003) Cell fusion is the principal source of bone-marrow-derivedhepatocytes. Nature 422:897–901.

21. Alvarez-Dolado M, et al. (2003) Fusion of bone-marrow-derived cells with Purkinjeneurons, cardiomyocytes, and hepatocytes. Nature 425:968–973.

22. Cowan CA, Atienza J, Melton DA, Eggan K (2005) Nuclear reprogramming of somaticcells after fusion with human embryonic stem cells. Science 309:1369–1373.

23. Silva J, Chambers I, Pollard S, Smith A (2006) Nanog promotes transfer of pluripotencyafter cell fusion. Nature 441:997–1001.

24. Johansson CB, et al. (2008) Extensive fusion of haematopoietic cells with Purkinjeneurons in response to chronic inflammation. Nat Cell Biol 10:575–583.

25. Nygren JM, et al. (2008) Myeloid and lymphoid contribution to non-haematopoieticlineages through irradiation-induced heterotypic cell fusion. Nat Cell Biol 10:584–592.

26. Coussens LM, Werb Z (2002) Inflammation and cancer. Nature 420:860–867.27. Duelli DM, et al. (2007) A virus causes cancer by inducing massive chromosomal

instability through cell fusion. Curr Biol 17:431–437.28. Duelli D, Lazebnik Y (2007) Cell-to-cell fusion as a link between viruses and cancer. Nat

Rev Cancer 7(12):968–976.29. Overholtzer M, et al. (2007) A nonapoptotic cell death process, entosis, that occurs by

cell-in-cell invasion. Cell 131(5):966–979.30. Chakraborty A, et al. (2004) Donor DNA in a renal cell carcinoma metastasis from a bone

marrow transplant recipient. Bone Marrow Transplant 34:183–186.31. Yilmaz Y, Lazova R, Qumsiyeh M, Cooper D, Pawelek J (2005) Donor Y chromosome in

renal carcinoma cells of a female BMT recipient: Visualization of putative BMT-tumorhybrids by FISH. Bone Marrow Transplant 35:1021–1024.

32. Johnson RT, Rao PN (1970) Mammalian cell fusion: Induction of premature chromo-some condensation in interphase nuclei. Nature 226:717–722.

33. Atkin NB (1979) Premature chromosome condensation in carcinoma of the bladder:Presumptive evidence for fusion of normal and malignant cells. Cytogenet Cell Genet23:217–219.

34. Kovacs G (1985) Premature chromosome condensation: Evidence for in vivo cell fusionin human malignant tumours. Int J Cancer 36:637–641.

35. Williams DM, Scott CD, Beck TM (1976) Premature chromosome condensation inhuman leukemia. Blood 47:687–693.

36. Kang HS, Youn YK, Oh SK, Choe KJ, Noh DY (2000) Flow cytometric analysis of primarytumors and their corresponding metastatic nodes in breast cancer. Breast Cancer ResTreat 63:81–87.

37. Isharwal S, et al. (2008) Prognostic value of Her-2/neu and DNA index for progression,metastasis and prostate cancer-specific death in men with long-term follow-up afterradical prostatectomy. Int J Cancer 123:2636–2643.

38. Horn LC, Raptis G, Nenning H (2002) DNA cytometric analysis of surgically treatedsquamous cell cancer of the uterine cervix, stage pT1b1-pT2b. Anal Quant Cytol Histol24:23–29.

39. Gupta GP, Massagué J (2006) Cancer metastasis: Building a framework. Cell 127:679–695.

40. Wei Y, Au JL-S (2005) Role of tumor microenvironment in mediating chemoresistance.Cancer growth and progression. In Integration/Interaction of Oncologic Growth, edMeadows GG (Springer, New York), Vol 15, pp 285–321.

41. Stanbridge EJ, et al. (1982) Human cell hybrids: Analysis of transformation and tumor-igenicity. Science 215:252–259.

42. Miller FR, Mohamed AN, McEachern D (1989) Production of a more aggressive tumorcell variant by spontaneous fusion of two mouse tumor subpopulations. Cancer Res49:4316–4321.

43. Brinkley BR (2001) Managing the centrosome numbers game: From chaos to stabilityin cancer cell division. Trends Cell Biol 11:18–21.

44. Ring D, Hubble R, Kirschner M (1982) Mitosis in a cell with multiple centrioles. J Cell Biol94:549–556.

45. Ray P, De A, Min J-J, Tsien RY, Gambhir SS (2004) Imaging tri-fusion multimodalityreporter gene expression in living subjects. Cancer Res 64:1323–1330.

46. Ponomarev V, et al. (2004) A novel triple-modality reporter gene for whole-bodyfluorescent, bioluminescent, and nuclear noninvasive imaging. Eur J Nucl Med MolImaging 31:740–751.

47. Bartos JD, et al. (2004) Genomic heterogeneity and instability in colorectal cancer:Spectral karyotyping, glutathione transferase-Ml and ras. Mutat Res 568:283–292.

9390 � www.pnas.org�cgi�doi�10.1073�pnas.0900108106 Lu and Kang

Dow

nloa

ded

by g

uest

on

July

2, 2

021

http://www.pnas.org/cgi/data/0900108106/DCSupplemental/Supplemental_PDF#nameddest=STXThttp://www.pnas.org/cgi/data/0900108106/DCSupplemental/Supplemental_PDF#nameddest=STXT